Final Report Summary - EUPORIAS (EUropean Provision Of Regional Impact Assessment on a Seasonal-to-decadal timescale)
Executive Summary:
Advances in scientific understanding and the ability to forecast climate variability and climate change mean that skilful predictions are routinely made on seasonal to multiannual timescales. Such forecasts can be of great value to a wide range of decision-making, where outcomes are influenced by climatic variations. EUPORIAS (EUropean Provision Of Regional Impact Assessment on a Seasonal-to-decadal timescale) set out to develop tools and services to exploit these emerging prediction capabilities, to extract useful and useable information tailored to specific sectoral needs, and to engage with potential users to help decision-makers make better informed decisions.
The consortium of 24 organisations represented world-class European climate research and climate service centres; expertise in impact assessments and seasonal predictions; two United Nations agencies; specialists in new media; and commercial companies in climate-vulnerable sectors such as energy, water, and tourism. The consortium also developed strong links with end users external to the project consortium, many of whom were subsequently directly involved in the project’s activities.
The vision of EUPORIAS was to develop climate services, operating on seasonal-to-multiannual timescales, and demonstrate their value in informing decision-making, thereby helping stimulate a market for climate services and improve the resilience of society to climate variability and change.
The project successfully delivered its top-level objectives, and some key highlights are as follows:
1. Delivered six prototype climate services, developed through close interaction with end users, for key sectors (wind energy, transport, agriculture, food security and water management), pulling the project results through to demonstrated societal benefit.
2. Produced an extensive survey and analysis of the landscape of users of climate predictions in Europe. Previous analyses focussed on longer timescale climate change projections, and little was known about who was using such shorter-term information in Europe and how they were using it.
3. Working closely with the Framework Programme 7 ‘Seasonal to decadal climate Prediction for the improvement of European Climate Services’ (SPECS) project, developed and delivered standard tools for calibrating, downscaling, bias correcting and visualising the skill of climate predictions. A series of R software packages was developed and made freely available, helping make complex operations easier.
4. Researched methods to transform seasonal forecast data into user relevant information for informing decisions. The methods included transforming raw variables into user relevant indices through statistical post-processing, and using impact models and regional climate models.
5. Forged engagement between various actors, particularly providers and users of climate services (a key concept of the Global Framework for Climate Services), to improve communication, mutual understanding, and co-development of prototypes, often using social science techniques.
EUPORIAS leaves several legacies including published research findings, the prototypes (some being developed further), and sustained engagement between climate service developers and users. Some of these legacies will benefit other projects under Horizon 2020 and Copernicus, some will be maintained through the Climateurope Coordination and Support Action’s network activities, while others will lead to new or stronger engagement between users and providers of climate services.
Project Context and Objectives:
The aim of the EUPORIAS project has been to develop new scientific capability, and prototype climate services through close engagement between developers and users, to help decision-makers in climate-sensitive sectors on seasonal to decadal timescales.
Societies have always faced challenges and opportunities arising from variations in climate, and have often flourished or collapsed depending on their ability to adapt to such changes. In today’s society we are seeing a growing need to improve our resilience to climate-related hazards and better manage the risks and opportunities from climate variability and climate change. This situation was recognised by governments, scientists and decision-makers at the World Climate Conference-3 in 2009, subsequently leading to the creation of the Global Framework for Climate Services (GFCS) under the leadership of several United Nations agencies.
Advances in scientific understanding and the ability to forecast climate variability and climate change mean that skilful predictions are now being routinely made on seasonal to decadal timescales. Such forecasts have the potential to be of great value to a wide range of decision-making, where outcomes are strongly influenced by variations in the climate. A clear opportunity therefore exists to develop new and improved methodologies that exploit the emerging prediction capabilities in climate science and, more importantly, to engage with potential users of such predictions in developing tools to extract useful and useable information tailored to specific sectoral needs.
To this end, the European Commission commissioned the four-year-long EUPORIAS project to develop prototype end-to-end climate impact prediction services operating on a seasonal to decadal timescale and to assess their value in informed decision-making.
The project began in November 2012, coordinated by the Met Office, with a consortium of 24 organisations representing world-class European climate research and climate service centres; expertise in impact assessments and seasonal predictions; two United Nations agencies; specialists in new media (such as Twitter, Facebook, smartphones, and YouTube); and commercial companies in climate-vulnerable sectors such as energy, water, and tourism. Additionally, the consortium developed strong links with end user organisations external to the project consortium, many of which were subsequently directly involved in the project’s activities.
The concept (or vision) behind EUPORIAS was that developing end-to-end climate impact prediction services, operating on seasonal-to-decadal timescales, and demonstrating their value in informing decision-making, would help stimulate a market for climate services and improve the resilience of society to climate variability and change. More specifically, through the development of prototype climate services, each with a very specific end user, EUPORIAS has developed a suite of climate services for key sectors (wind energy, transport, agriculture, food security and water management), pulling the scientific methodologies, tools, and techniques through to demonstrated societal benefit.
Over the past four years, EUPORIAS has largely delivered this vision. EUPORIAS’ research findings, development of the prototypes, outputs (such as project results and papers, workshops, meetings), and close engagement with a range of stakeholders and end users has certainly had an impact on national, European and global discussions and development of climate services. EUPORIAS has helped promote the emergence of a market for those services.
EUPORIAS had six top-level objectives, described below along with a brief summary of progress:
1. Develop and deliver reliable and trusted impact prediction systems for a number of carefully selected case studies. These will provide working examples of end-to-end climate-to-impacts-to-decision-making services operating on seasonal-to-decadal (S2D) timescales.
EUPORIAS delivered six prototypes of climate services operating on a seasonal time scale. The services were defined and developed as much as possible through close interaction with the target users. The prototypes addressed sectors as different as food security in Africa and transport in the British Isles and also looked at agriculture, renewable energy and water resources. They provide working example of how to transform raw climate prediction data, mostly coming from the outputs of European seasonal predictions models, into decision relevant advice for specific economic sectors. Such a transformation went beyond the post-processing of the data (such as bias correction, downscaling and calibration) and included workshop design, training material and mobile phone applications.
2. Assess and document key knowledge gaps and vulnerabilities of important sectors (e.g. water, energy, health, transport, agriculture, tourism) along with the needs of specific users within these sectors, through close collaboration with project stakeholders.
EUPORIAS conducted one of the first and most extensive analyses of the landscape of users of climate predictions in Europe. Although other analyses had been conducted before, they tended to focus on the longer climate change projection time-scale, and prior to the project relatively little was known about who was using shorter-term information in Europe and how they were using it. An online questionnaire and 80 in-depth interviews provided an initial assessment of the main perceived gaps in the provision of climate information from key sectors such as water and energy. The analysis also revealed the striking disparity that exists among sectors in the ability to deal with and digest climate data. EUPORIAS also developed a vulnerability framework specifically designed for climate prediction time-scales.
3. Develop a set of standard tools tailored to the needs of stakeholders for calibrating, downscaling, and modelling sector-specific impacts on S2D timescales.
EUPORIAS worked closely with its sister project SPECS to develop and deliver a set of standard tools for calibrating, downscaling, bias-correcting and visualising the skill of the climate predictions. One of the challenges the project participants faced in developing user-relevant products was the access to high quality information in a standard format. The European Climate Observations, Modelling and Services (ECOMS) User Data Gateway developed and hosted by the University of Cantabria played a major role in alleviating the problem as it gave all project partners access to observations, and predictions through a standardised interface. A series of software packages developed in R was made publicly and freely available and made complex operations possible in a few simple steps.
EUPORIAS built upon some of the technical developments undertaken in the SPECS project. It looked at ways to represent verification attributes, such as ranked probability skill scores or reliability, in a graphical format. From simple site-specific validation plots to visually compelling visualisation such as Project Ukko (http://project-ukko.net/) EUPORIAS transformed complex data streams and mathematical concepts into information that users can act upon.
4. Develop techniques to map the meteorological variables from the prediction systems provided by the World Meteorological Organization (WMO) Global Producing Centres (two of which [Met Office and Météo-France] are partners in the project) into variables which, are directly relevant to the needs of specific stakeholders.
Central to the success of the project was the ability to transform the data coming from the seasonal prediction systems into user relevant information that can be used to inform decisions. To achieve this, EUPORIAS worked along two distinct and complementary lines. One was to transform the raw variables emerging from the forecasting systems into user relevant indices through statistical post-processing. The other was to employ impact models as well as regional climate models to increase the user relevance of the raw variables.
The aim was that whilst increasing user relevance, the derived variables could in some cases increase the quality (as defined by the verification attributes) of the variables being forecast. The results of the project indicate that in some cases, for example for some hydrological variables, this is indeed the case. However, in the majority of the circumstances the post-processing doesn’t significantly alter the verification characteristics of the underpinning variables. The exceptions appear to be related to either cases in which the extra predictability comes from storage terms (e.g. lake and snow accumulation) that are not explicitly included in the climate prediction forcing, or to indices that are calculated through a non-linear combination of a number of raw model outputs (e.g. Fire Weather Indices).
5. Develop a knowledge-sharing protocol necessary to promote the use of these technologies. This will include making uncertain information fit into the decision support systems used by stakeholders to take decisions on the S2D horizon. This objective will place Europe at the forefront of the implementation of the GFCS, through the GFCS's ambitions to develop climate services research, a climate services information system and a user interface platform.
One of the challenges faced in the development of climate services is the interface between climate information and decision-making. EUPORIAS looked at this challenge in a variety of ways. Effort went into identifying the common cognitive biases that can affect the perception of uncertain information, including social science techniques. This led to the development of a series of communication guidelines such as incremental disclosure of information. Effort was also put into developing novel approaches for user engagement. In some cases the project drew on ideas and practices from other fields and disciplines, such as the placebo effect. In other cases the project built upon concepts that had been suggested in the past, such as climate roulette, but pushed the concepts further to make quantitative economic evaluations.
What became clear early on in the project was that a lot of the value of the services being developed was not so much in the services themselves but rather in the continuous interaction with the users and the user support this implied. This allowed EUPORIAS to give substance to some of the concepts proposed by the GFCS, in particular the GFCS Climate User Interface Platform (CUIP). Contrary to the original expectations of most of the project participants, the CUIP EUPORIAS developed is not an online portal where users can get access to a climate data stream relevant to their business or decision, but a useful and active collection of user-provider engagement events. EUPORIAS re-defined the user interface as more than a collection of general decision support systems. It instead designed a system able to support a continuous dialogue between users and providers of climate information. To use the words of one of our external international advisors, Professor Roger Stone, it created a ‘discussion support system’.
6. Assess and document the current marketability of climate services in Europe and demonstrate how climate services on S2D time horizons can be made useful to end users.
To develop a market for climate services in Europe it is important to consider the size of the market and identify the major barriers preventing its growth. Whilst an in-depth analysis of the climate service market was out of the scope of the project (but is the scope of other European Commission-funded projects) EUPORIAS worked on a generic methodology that could be used to assess the marketability of services. This was done by analysing the water sector in some detail, and the methodology could be easily generalised and applied to other sectors. The results the project generated could be used as a basis for the new Horizon 2020 projects that will explicitly look into the issue of the climate service market.
Project Results:
Key areas of progress relating to the project objectives have been summarised in section 2. Further information is now given on the main areas of progress and noteworthy results from each work package.
3.1 WP11 – Assess sector-specific vulnerability
The objectives of WP11 were:
• To raise awareness of seasonal and decadal predictions and their limitations through a close interaction with a number of stakeholders;
• To identify critical sector-specific vulnerabilities operating on seasonal to decadal time scales;
• To contribute to a prototype component of the CUIP within the GFCS in terms of targeted vulnerability information and tailored data-flow in relevant European economic sectors for seasonal to decadal time scales.
The main focus of WP11 has been the identification and analysis of sector-specific vulnerabilities based on an effective dialogue with stakeholders. This dialogue started early in the process and was sustained throughout the project. A theoretical framework for consistent identification of the relevant vulnerabilities for the project’s prototypes was developed. The main outcomes of this analysis have been fed into the EUPORIAS contribution to the CUIP through a collection of different microsites and a series of sectoral workshops. The main findings of WP11 were as follows:
a) Outcomes from European Stakeholder Climate Services Conference:
• In January 2013 the first EUPORIAS Stakeholder Meeting took place in Rome, attended by 43 people from across Europe representing ten different sectors (e.g. water, health, transport, energy). The main objective of the workshop, in line with all EUPORIAS activities, was to encourage communication between the producers and users of S2D information.
• The interaction and dialogue with the stakeholders made it possible to gain an understanding of sector-specific information regarding: how stakeholders currently use S2D climate predictions; critical/relevant choices in their business that could be affected by climate; how climate influences their business choices; and how climate information enters into the decision-making procedures.
• This information was gathered through questionnaires issued prior to the meeting, the Stakeholder Meeting itself, and subsequent discussions and workshops resulting from the growing relationships between stakeholders and EUPORIAS project partners.
• Conclusions emerging from the workshop included: temperature and precipitation are the most relevant climate parameters requested by the EUPORIAS stakeholders. This is especially so in the water, energy, health and agriculture sectors. The most valuable parameters for the surface-transport sector are ground temperature (influenced by air temperature, wind, and soil moisture) and the number of marginal nights (zero-crossing). Important parameters for the insurance sector include: the number of land-falling tropical storms, extreme precipitation, river runoff over specific thresholds, insurance-specific drought indices, weather profile over the year (including lack of snow and late frosts), general “crop failure indices” with a focus on drought lengths or dry spells.
• The business-critical decisions that can be informed by weather tend to cluster in spring (for the summer outlook) and autumn (for the winter outlook). The exceptions are the agricultural sector, which would benefit from seasonal predictions throughout the year, and the insurance sector, for which the beginning of January and the beginning of April are crucial dates.
• While downscaling is seen as important for most stakeholders, they appeared to be keen to prioritise resource-investment in improving the predictability of the large-scale drivers rather than increasing the granularity of the data.
• While climate predictions (seasonal and decadal) are an interesting and potentially useful area for the stakeholders, and while many sectors use them, there is still a huge need for education and training. Direct access to expertise, for instance via sector-specific workshops or seminars, is seen as a vital way of providing this.
• Despite the fact that a significant fraction of the audience was aware of climate predictions, and whilst some of the participants were using these predictions, there was a clear language barrier on a series of crucial definitions. The primary example of this was around the communication of risk and uncertainty. A number of stakeholders stated that they would not have used the predictive information unless its level of confidence (no definition provided) exceeded 95%.
• The Stakeholder Meeting was largely a success and it certainly helped to inform a number of other activities within the project, but in hindsight it might have been better to have a larger number of more sector-specific meetings. Subsequently, EUPORIAS tried to address this challenge by, whenever feasible, supporting sector-specific workshops, which could have helped the interaction between users and providers of climate information.
b) Developing a method to assess sector-specific vulnerabilities:
• A common methodology to assess sector-specific vulnerabilities has been developed and the resulting Vulnerability Assessment Framework (VAF) was applied to each sector of interest. This characterised sector-specific vulnerabilities in such a way as to provide valuable information for decision-makers with respect to the use of climate predictions.
• Challenges for the vulnerability assessment of economic sectors to S2D climate events originate from the inherent role of climate for different economic sectors. Climate may be viewed by these sectors as a hazard-, resource-, production- or regulation-factor.
• One of the major findings of this study is the realisation that the Decision-Making Process (DMP) is a very sensitive factor, which significantly influences: the determination of threshold, which define critical situations; the characteristics of buffer functions and thus the temporal scale of critical climate conditions; and the specification of climate information needs and thus their potential usability (value) for decision-making. Decision-making processes should be considered in detail when assessing sector-specific vulnerabilities in the context of climate information.
• Another major finding is the classification system of ‘climate-impact types’ that classifies sector-specific problems in a systemic way. This system proves to be a promising concept because: it reflects and differentiates the cause for the climate relevance of a specific problem (compositions of buffer factors); it integrates DMPs, which proved to be a significant factor; it indicates a potential usability of S2D climate service products and thus integrates coping options; and its systemic approach goes beyond the established ‘snap-shot’ of vulnerability assessments.
• The EUPORIAS VAF has been applied to all the prototypes and case-studies developed in the context of the EUPORIAS project prototype and inserted into the EUPORIAS microsites.
c) Contribution to GFCS Climate User Interface Platform
• One of the most important (but least developed) components of the GFCS is the Climate User Interface Platform (CUIP). The understanding of what a CUIP should be has changed significantly over the years and it has become evident that an online tool that is not supported by close interaction with the users tends to be of limited use. For this reason, EUPORIAS developed a CUIP based on a revised scope and consistent of two distinct parts.
• On the one hand there are the microsites, a collection of user specific web pages describing the climate service prototypes, and four of the case studies, including information on their associated sector-specific vulnerabilities.
• However, the CUIP should not be just a simple ‘portal’, but rather a place where the different communities can share lessons learnt materials, and know-how, together with the main outcomes of the EUPORIAS project. Face to face meetings and ongoing dialogue are often required. It has been recognised that person-to-person meetings help build trust, legitimacy and saliency into the service being developed. A series of workshops was developed with the users, who have played a pivotal role in scoping the prototypes, their functionalities and interfaces. These events have helped the developers to go further than they initially planned, and also helped the users understand why some of their needs cannot be fulfilled.
3.2 WP12 – Assessment of users’ needs
The objective of WP 12 was to assess the needs of users from across European society for seasonal-to-decadal climate predictions.
The WP assessed the needs and requirements of users regarding seasonal-to-decadal climate predictions (S2DCP) across Europe. The tasks resulted in the following successful outcomes: a systematic literature review of the use of S2DCP in Europe was produced; a workshop with National Meteorological and Hydrological Services (NMHSs) and other stakeholders was conducted; around 80 interviews with key European organisations working across the sectors identified in EUPORIAS were conducted; a multi-lingual online survey of users’ needs was undertaken with a total of 489 responses received; a workshop with S2DCP developers was conducted.
The main findings of the WP were as follows:
• The majority of the organisations involved in this study were aware, to differing degrees, of seasonal climate forecasts (SCF), whilst decadal predictions were less well-known.
• The use of SCF across Europe to inform decision-making is still an emerging venture, mainly due to the lack of skill and reliability (‘reliability’ is used here as a synonym of trustworthiness and can be mapped onto a number of other technical concepts such as skill, reliability, and sharpness) over Europe. Decadal predictions are regarded as unchartered territory and thus not used.
• Some sectors were identified as using SCF, such as the energy, water, transport, and insurance sectors. However, the use of SCF is limited and only used in a qualitative manner i.e. the forecasts are not used operationally or in a systematic way (Bruno Soares and Dessai, 2016).
• European Centre for Medium-Range Weather Forecasts (ECMWF) and NMHSs are regarded as the main providers of climate information in Europe including SCF.
• Most organisations involved are sensitive to weather and climate conditions, although some are more concerned with the impacts of extreme weather (e.g. drought, floods), whilst others are more interested in weather variables such as temperature and precipitation (see D12.3).
• Many organisations already use weather/climate information (mainly historical data/past observations and weather forecasts). Such information tends to be used to either develop/feed operational models, forecast seasonal variability based on past data, and/or understand future weather conditions, all of which are used to inform decision-making processes within organisations. Many perform some kind of post-processing in-house.
• The main perceived barriers to the use of SCF are mainly related to low reliability, but also related to other factors such as usability of the information provided and accessibility to such information by the end-users (Bruno Soares and Dessai, 2016).
• The main enabler to the use of SCF in those organisations was related to the interactions with the producers of SCF (e.g. National Met. Services). Other non-technical aspects such as the relevance of SCF to the organisation’s activities and operations and the existing resources and capacity to assimilate and act upon that information also acted as enabling conditions to the use of SCF (Bruno Soares and Dessai, 2016).
• Difficulties in integrating SCF into existing operational models and structures, existing traditions of performing historical analyses, and not needing this type of information were also pointed out as reasons for not using SCF.
• Understanding decision-making processes within organisations and how SCF is or can potentially be used can be difficult at times, as responses provided by respondents are influenced not only by the size (i.e. multiplicity of activities) and nature (i.e. end-user versus intermediate organisations) of the organisation itself, but also by the role of the interviewee in it (e.g. modeller, head of department).
• There is a general understanding that information on the uncertainty of the information provided is a fundamental component of S2DCP. Many organisations would prefer information on the uncertainty of forecasts to be provided using a deterministic approach. The preferred method for representing uncertainty is numerical (e.g. one figure, percentages) as this would facilitate the quantification and integration of uncertainty into models.
• The final workshop with the S2DCP developers also provided a range of interesting findings although these were specific to the case studies explored during the workshop. One of the main findings was that for all of the cases of users’ needs discussed at the workshop there were no products available that could fully satisfy their needs (e.g. spatial or temporal resolution required, limited skill). As a result, a better understanding of the types and chains of decision-making by the users and how this data is/would be utilised to inform such processes would also help to refine the quest for (and potentially the provision of) data that is able to satisfy those particular needs.
3.3 WP21 – Calibration and downscaling
The objectives of WP21 were to develop and apply a set of bias-correction and downscaling methods for use with seasonal to decadal forecasts; to downscale standard climate variables and advanced climate indices for use in the EUPORIAS case studies; to assess the uncertainty associated with the downscaling methods in collaboration with WP33; and to make available, through the EUPORIAS web portal, downscaled data, along with a number of calibrated methodologies.
Seasonal forecasts are produced from global prediction systems at typical resolution of 100-200km while climate service users often require information at significantly finer spatial resolution. Furthermore, forecasts contain sizeable errors when compared to observations and such biases need to be corrected for successful use of forecast data in impact studies. Several statistical downscaling and bias correction methods have been applied to global seasonal forecasts over Europe to provide inputs for impact models used in several case studies and climate service prototypes. Both approaches show similar performance in removing biases and preserving predictability skill.
In Eastern Africa a number of regional climate models and statistical downscaling methods were used to downscale a global forecast to focus on Ethiopia in June-September. Verification results revealed that both dynamical and statistical downscaling are able, in general, to capture and reproduce the predictive skill evident in the global seasonal hindcast it is derived from. However, on average, the dynamically and statistically downscaled hindcasts show no added value in Eastern Africa if we define the added value as a higher predictive skill in the downscaled hindcast. Similar results were found for the World Food Programme (WFP) Livelihoods, Early Assessment and Protection (LEAP) platform (drought early-warning system) for Ethiopia: the performance of LEAP in predicting humanitarian needs at the national level is not improved by using downscaled seasonal forecasts. These results indicate that, in the EUPORIAS case studies over Europe and Eastern Africa, downscaling only reproduces the forecast skill of the global forecasts it’s derived from but does not improve it.
The main findings of the WP were as follows:
Task 21.1: Statistical downscaling and bias correction of General Circulation Model (GCM) forecasts over Europe
• The ECMWF system 4 (S4) forecast was bias corrected using different approaches with focus on i) pan-European hydrological modelling for flood risk estimations, ii) supporting logistical decisions for inland waterway transport on the river Rhine and the electricity demand case study over Southern Italy, and the Land Management Tool prototype over Southwest UK.
• All bias correction methods applied are efficient in removing seasonal mean S4 biases, even if some biases in precipitation are still evident during test periods not used in their calibration. In general, bias correction methods preserve the predictive skill of the original S4 hindcast, but they do not improve it.
• Perfect prognosis statistical downscaling has also been applied over different regions in Europe to support several case studies and the climate service prototypes.
• The predictive skill of the perfect prognosis downscaled forecasts over different European regions, in general, is not improved. Only the analog method has shown a slight increase in reliability, as found in different analyses over France and Italy. However, this may be related to a limited forecast resolution. Some improvement of the forecast skill for CII was found, but it is not significant in all cases.
Task 21.2: Combined statistical and dynamical downscaling over eastern Africa
• The utility of dynamical and statistical downscaling to provide seasonal forecast data for impact models over eastern Africa was assessed, in particular for the Drought Early-Warning System LEAP. After consultations with the WFP, it was decided to focus on the Kiremt rainy season (June-September) in Ethiopia using seasonal hindcast initialised in May, which can be used as input to the LEAP system.
• The downscaled ensemble created consists of five regional climate models – (RCMs) (DWD-CCLM4-8-21, SMHI-RCA4, ENEA-RegCM-4-3, UCAN-WRF341G and UL-IDL-WRF360D) at about 25 km resolution and two empirical-statistical downscaling (ESD) methods (the analogue technique - AN1 and a generalised linear model - GLM) at about 50km resolution (defined by observations). Analysis of eight observational datasets showed that there are large uncertainties in interannual variability of precipitation among these datasets (which may have a considerable impact for the case of ESD methods) in eastern Africa.
• Using a number of deterministic and probabilistic verification metrics, it was found that there are two distinct regions where some predictive skill is evident in the driving EC-EARTH seasonal hindcast: northern Ethiopia - North-East Sudan and southern Sudan - northern Uganda while there is no skill elsewhere (Fig. 1).
• Focusing on Ethiopia, verification results revealed that the RCM and ESD methods are able to capture and reproduce the signal evident in EC-EARTH over northern Ethiopia in June-September showing about the same performance as their driving GCM (Fig. 1). However, on average, the RCM and ESD hindcasts show no higher skill in predicting future seasonal anomalies.
• Two RCM hindcasts (DWD-CCLM4-8-21 and SMHI-RCA4) were used to assess the utility of dynamically downscaled seasonal hindcast data as an input to the LEAP platform. Consistent with the results found for the RCM hindcast, the performance of the LEAP system in predicting humanitarian needs at the national level in Ethiopia is not improved by using downscaled seasonal forecasts.
• The effect of downscaling on the quality of seasonal forecasts of climate information indices in eastern Africa was considered, using a range of indices relating to the water availability (potential evapotranspiration and water balance) and rainfall regimes (wet day frequency and simple daily intensity index). This analysis shows that the downscaling does not improve the forecast skill of global forecasts of climate indices in eastern Africa. Consequently, it is recommended to derive forecasts of indices from the bias corrected global forecasting systems rather than from dynamically downscaled forecasts.
• It should be noted that these conclusions are only for Ethiopia in the June-September season and cannot be generalised for other regions and seasons. For example, some skill improvement in a statistically downscaled forecast was found in the Philippines (see WP32). Additionally, large observational uncertainties can potentially prevent accurate verification of the downscaled high-resolution hindcast in eastern Africa.
3.4 WP22 – Impact-relevant climate information indices (CII)
The aim of this work package was to generate a collection of impact relevant climate information indices (CIIs) and to provide the best estimate of such CIIs for the current and near-term climate.
Based on user needs identified in WP11 and WP12, CIIs relevant for various economic sectors and a large variety of stakeholders were selected (D22.1 D22.2). Seasonal forecast skill of these CIIs has been analysed extensively, including an assessment of downscaled CII forecasts as a contribution to WP21 (D22.3). This work package has also attempted to identify key windows of opportunity where CIIs can provide quality and value from an end-user perspective (D22.4). Forecasts of CIIs evaluated in WP22 were used to complement the prototypes and case studies developed in WP42 and WP43.
In addition, this work package has developed software tools (R packages) to facilitate download, bias correction, index computation and the validation of seasonal forecasts, enabling the further development of CII forecasts within the project and beyond.
The main findings of the WP were as follows:
• Forecast skill in Europe is limited with enhanced skill in summer compared to winter and higher skill for temperature and related indices than for precipitation. Skill varies strongly by season, region, lead time, climate index, and spatio-temporal aggregation.
• Some regions show skill in specific seasons, which may open a window of opportunity for forecasts to be useful in specific cases. Hence, this WP has developed an interactive, publicly available web-platform (https://meteoswiss-climate.shinyapps.io/skill_metrics) that allows users to explore this variability in forecast skill interactively to identify opportunities for skilful and thus useful seasonal forecasts.
• The limited predictability of climatic conditions in Europe poses challenges for the use of seasonal forecast information. Cases from various sectors such as hydrology, viticulture, and energy show that it is very difficult to derive decision-relevant information from seasonal forecasts in Europe due to the lack of skill.
• In general, forecasts of CIIs are, at most, as skilful as forecasts of the seasonal mean of the underlying meteorological variables. As this reduction in forecast quality is moderate for all except indices describing rare events, CII forecasts can often be issued without major loss in skill. The enhanced user relevance of forecasts of CIIs outweighs the slight reduction in forecast skill.
• CII forecast skill is found to be rather insensitive to the choice of bias correction method, but basic calibration is very important. In some cases, however, using more sophisticated calibration methods (e.g. quantile mapping compared to mean de-biasing) improves the skill in CII forecasts.
• We further find that indices defined with respect to percentiles of the forecast and observed distribution respectively (e.g. percentage of time with wind speed above the 90th percentile) are advantageous in that such indices are less prone to systematic model errors than indices defined with respect to absolute physical thresholds (e.g. frost days).
• While the climatic conditions can be difficult to forecast, forecasts of impacts dependent on these conditions can have considerable value (demonstrated by forecasts of humanitarian needs for Ethiopia). Such added value is partly due to the sensitivity of the impact to large-scale long-term climatic conditions (drought) that are more predictable and partly due to the co-location of forecast skill in areas sensitive to drought.
• Enhanced skill is also found in forecasts of river flow based on a hydrological model resulting from the long-term memory of initial conditions for the hydrological model including soil moisture and snow. This highlights the presence of sources of predictability not included in the operational seasonal forecasting system that may be exploited for targeted applications through coupled and initialised impact models.
• It is worth exploring the value of CII forecasts even if there is marginal skill in forecasts of the underlying variables. Implicit aggregation or weighting in the calculation of the impact variable can amplify partial skill in the underlying variables and thus exploit windows or pockets of predictability while transforming meteorological information to more user relevant quantities.
3.5 WP23 – Impact models for impact predictions and WP31 – Quantifying uncertainty in impact models
WP23 and WP31 were intricately linked together in the project, both jointly led by the Met Office and WU, and were delivered by mostly the same set of partners (with the exception of CETAqua who only participated in WP23). The work delivered under WP31is a natural extension of that began under WP23, so a joint report for both work packages has been provided to avoid duplication.
The objectives of WP23 were:
• To further develop complex impact models able to address the users’ needs and inform the case studies and the prototypes:
• To develop a prototype operational workflow to use these models in S2D forecast mode;
• To assess and improve their predictive skill by analysing hindcasts of low- and high-end impact events (hi/lo discharge, crop yields, etc);
• To develop optimal geographical forecasting units, as a function of model physics and stakeholder needs.
The objective of WP31 was then:
• To fully characterise the level of confidence we can associate with specific impact models. Multi-model, multi-driver and perturbed physics parameter ensembles will be used for this purpose.
The main findings of the WPs were as follows:
● WP23 and WP31 have completed a coordinated set of model studies across different sectors (agriculture, water, forestry and transport), over different regions and scales (Europe, East Africa, specific areas and basins), driven by baseline forcing data and different seasonal hindcast datasets. Some of the studies have investigated the role of particular parts of the uncertainty chain (e.g. role of bias correction, initial conditions, seasonal hindcast skill, model parameters) in overall impact skill.
● In general, it appears that while bias correction is important for some applications (e.g. where thresholds are important in dynamic models using crop temperature thresholds, or indices), it does not affect probabilistic skill in all cases.
● There are remaining issues to be dealt with regarding application of bias correction in impact modelling, such as appropriate choice of window sizes for Quantile-Quantile (QQ) mapping, and whether to apply bias correction to model inputs or outputs.
● There is not necessarily a monotonic decay in skill with lead time in all cases, although this is true for some impacts sectors/models.
● Skill varies with variable (e.g. greater skill for temperature than precipitation, and least skill for radiation), region, season, lead time and forcing dataset.
● Skill for impact fluctuates (e.g. terrestrial Gross Primary Production, (GPP)) may be greater than that for state variables (e.g. crop yield).
● Skill for some impacts, particularly river discharge, may be larger than that of its meteorological drivers. In such cases memory effects from proper initialisation of state variables (e.g. soil moisture, snow mass for hydrology) explain the skill.
● A significant part of WP23/31 involved further development of impact models, including those for crops, water, forestry and energy and development of workflows for impact hindcast simulations and analysis.
● For some impacts proper skill assessment is hindered by lack of high quality, spatially distributed observations, most notably so for crop yields (in eastern Africa) and quality assessment of such data may require significant efforts.
Figure 2 outlines the modelling studies performed by WP23 and WP31. A complete scenario set consists of: one (30 yrs) baseline/reference run forced by WFDEI; one hindcast with System 4 forcing, not bias corrected (30 yrs x 12 months x 15 members = 5,400 runs of seven months each) and one identical hindcast with bias corrected forcing. Scenarios are available as a set of self-descriptive NetCDF files in a format agreed between the relevant partners, to facilitate intercomparison of results and allow for the creation of an MME.
In terms of assessing the influence of lead time and other factors on seasonal impact forecast skill, some sector-specific tentative findings are that:
For crops:
• There is reasonable skill for some models (in predicting Above/Normal/Below events), but poor skill for others (Ogutu et al. 2016);
• It is difficult to evaluate skill over East Africa due to poor availability of observations and because other factors (human, socio-economic) may be more important in driving changes in crop yields than weather;
• Strong links between precipitation and yield in observations and models implies that climate impact model initialisation may not be needed, and that models can be more confident where seasonal forecast predictability for precipitation is strong (Williams and Falloon, 2015).
For water:
• There is better skill in river discharge and runoff for North Europe, better skill during winter than in summer, and better skill for high flows versus low flows; skill is dependent on the basin’s hydrologic regime;
• Climate skill is not the only, or dominant driver, of overall skill, due to the importance of other processes and factors (e.g. snow, lakes, wetlands, reservoirs, human factors). For river hydrology in Europe, overall skill appears to be dominated by skill derived from initial conditions of soil moisture and snow mass;
• Skill for river discharge extends to longer timescales (and more downstream) than that for runoff; which is useful for analysing skill with respect to pseudo observations (i.e. hydrological reanalysis data);
• Since summertime drinking water demands correlate with temperature, skilful seasonal water demand forecasts in Spain are feasible. Winter time dam inflow predictions in Spain do not provide additional information over climatology.
For forestry:
• In the Swedish case study regions, the greatest skill was found for bias-corrected temperature during January-March, potentially supporting planning of harvest operations depending on frozen soil, or to support the planning of planting activities depending on soil moisture;
• Fire Weather Index showed considerable skill in especially the Eastern Mediterranean, a skill that, however, cannot be attributed to forecast models reproducing variability in summer North Atlantic Oscillation.
For other sectors
• The future variability of wind energy resources is better addressed with wind speed and temperature transformed into a capacity factor, predictions of which can be integrated more easily in decision-making processes;
• For road transport, predicted winter air temperatures need translation into colder road skin temperatures.
A significant part of the WP23/31 involved further development of impact models, which included:
• Crop models - new GLAM version, setup for Africa; New JULES-crop model and testing (Osborne et al. 2015; Williams et al. 2017); nutrient representation in WOFOST; Regional setup for LPJmL, Wine production model (Fraga et al. 2014);
• Water models - New version of HYPE, re-calibrated (Donnely et al. 2015); development and testing of an impact model for dam management, and testing on the use of predictions for urban water demand prediction (Pouget et al. 2015);
• Forestry models – development of extreme temperature and precipitation metrics using a soil moisture and soil temperature model;
• Transport sector - development of METRo impact model to be used to predict winter road temperatures for the Iberian Peninsula;
• Energy models – development of capacity factor seasonal predictions based on ECMWF System 4 wind speed and temperature.
3.6 WP32 – Uncertainty framework
The main goal of WP32 was to look at the best way of combining the main sources of uncertainty in seasonal impact predictions (i.e. uncertainty from the climate predictions and uncertainty in the impact model formulation) in a single coherent assessment. Early in the project (see e.g. the internal report on S2D uncertainty assessment protocol for impact models, released as milestone M26), it was found that a general framework exploring all sources of uncertainty in a coherent way across application sectors was unfeasible due to their heterogeneity and different sensitivities. Therefore, the analyses and results focused on specific sectors and steps of the climate-impacts uncertainty chain.
The main findings of the WP were as follows.
Task 32.1: Assessment and combination of climate predictions:
• New tools at the interface between climate forecasts and impacts (ECOMS-UDG, visualize https://github.com/SantanderMetGroup/visualizeR) have been developed, enabling providers and end-users of seasonal forecasts to access, analyse and visualise different forecasts in a homogeneous way. This data gateway and tools replace standard file sharing, increasing flexibility and adaptation to users’ needs while reducing the data transfer burden.
• This technical development required the cooperation of different work packages (WP4, WP21, WP22, WP23, WP31, WP32 and WP33) and projects (SPECS, WP52) under the European Climate Observations, Modelling and Services (ECOMS) umbrella. Moreover, link functions make these tools take advantage of other tools developed in ECOMS (easyVerification, SpecsVerification, esd). The tools have been used by many partners, as a convenient way to guarantee the homogeneity of input and presentation of results (e.g. Ogutu et al. 2016).
• The skill of several seasonal forecasting systems (ECMWF System 4, NCEP CFSv2, MetOffice GloSea5) has been assessed, both worldwide and on specific regions (Europe, East Africa and the Philippines). The comprehensive analysis has been made possible by the availability of these forecast systems through the ECOMS-UDG.
• The analysis of ECMWF System 4 seasonal forecast (selected as the initial model to assess the skill of the global models) enables users to identify the most suitable regions for impact applications on seasonal time scales using different observational datasets as reference. Results indicate that the skill of the global seasonal forecasting systems is mainly located in the tropics (especially for precipitation) and no significant skill is found over Europe (Manzanas et al. 2014), where the small regional skill does not agree across different forecasting systems.
• Important season and lead-time-depending biases were found in prototype regions (East Africa regarding crop models, and Europe for hydrology). In this sense, bias-correction and or downscaling methods are required to make key variables, such as precipitation and temperature, suitable for impact applications. However, the performance of perfect prognosis statistical downscaling was found to be limited in those regions exhibiting significant reanalysis uncertainty, such as East Africa or the Philippines (Manzanas et al. 2015).
• Moreover, contrary to simple bias correction techniques (which are limited by the quality of the model), most sophisticated perfect prognosis techniques were found to improve the global model seasonal forecasts when/where the model is more skilful on large-scale predictor variables than on the target variable (Manzanas et al. 2017).
• Climate forecast combinations by sophisticated techniques did not significantly improve simple, equal weight combinations. In an attempt to improve the skill of individual seasonal forecasting models, a Bayesian calibration and combination method (as described by Stephenson et al. 2005) was applied over Europe and the Philippines for different seasonal forecast systems (EUROSIP). The different settings lead to a slight improvement upon the skill of the individual models. However, no clear benefit was observed when applying sophisticated model combination techniques; thus, simple model combination obtained by equal weighting of all contributing systems is recommended in general.
Task 32.2: Uncertainty framework for seasonal impact predictions:
• The uncertainty of impact model predictions was compared with that of their climate model forcing. This is strongly linked with the activities on WP31, which quantified the uncertainty in impact models.
• An uncertainty framework, which enables the user to identify the contribution of each source of uncertainty, was developed for the wine production industry. Results indicated that the uncertainty associated to the seasonal forecast (measured through the error variance, according to Beven, 2012) is reduced as the lead time decreases. These improvements were mostly felt in the lower percentiles of wine production. Although the model variance increased with the reduction of the lead time, indicating an improvement in the forecast, the model variability was always smaller than the observations and the uncertainty was very high.
• An efficient tool to estimate predictive hydrological uncertainty has been deployed, with a graphical interface for end-users. The method assumes that the error in the forecast is a function of the forecast itself and therefore relies on having both hindcasts and observations at the site of interest. A methodology using quantile regression was implemented as a forecast post-processing tool in DHI’s generic operational decision framework. This flexible and efficient uncertainty estimation technique is available to DHI’s software users.
• Four sources of uncertainty occur in deterministic flow modelling; random or systematic errors in the model inputs or boundary condition data, random or systematic errors in the recorded output data, uncertainty due to sub-optimal parameter values and errors due to incomplete or biased model structure. While Bayesian methods are widely used to quantify model parameter uncertainty, a more straightforward Monte Carlo approach can be applied (Butts et al. 2016). This approach was tested on hydrological model parameters applied to the Añarbe reservoir. Model performance was assessed using the absolute error average and variance. Assuming 10% uncertainty in the discharge then all parameter sets within an acceptance interval of 0.36 m3/s of the Pareto front define the parametric uncertainty.
• For certain applications (e.g. hydrology) seasonal impact forecast uncertainty can be smaller than the atmospheric seasonal forecast uncertainty used as input. This is due to slow-varying initial non-atmospheric variables, usually related to initial reservoirs (e.g. mountain snowpack, groundwater). Despite the low skill of seasonal climate forecasts over Europe, some examples of reduction of the uncertainty and slight skill improvement in river flow forecasts were found. These improvements are not significant, however, and a protocol has been agreed with the stakeholders to provide both seasonal climate- and climatology-driven river flow forecasts along with expert advice encouraging stakeholders to use the seasonal prediction when skill is expected.
3.7 WP33 – Communicating levels of confidence
The objective of WP33 was to test the effectiveness of different approaches of communicating the confidence and uncertainty associated with S2D predictions and its impacts and to make recommendations for good practice.
To address this challenge, the work package brought together expertise and insight from climate service providers, end-users, and social scientists. A preliminary user needs survey (Taylor et al. 2015) and review of existing approaches to communicating uncertainty, informed the development of a set of strategies for communicating information about probability and skill in seasonal predictions. Subsets of these strategies were then formally tested with decision-makers from relevant sectors in a series of online decision labs. Finally, key findings from the preceding work were brought together with key lessons from four prototypes in a final report outlining the recommendations for the communication of confidence and uncertainty in S2D predictions (and climate information more broadly) that can be drawn from the EUPORIAS project as a whole.
The main findings of the WP were as follows:
a) Identifying user needs and challenges
• To explore user needs with respect to the communication of confidence and uncertainty, a survey was conducted with current and potential users of S2D.
• Survey findings indicated that: (i) while current S2D users perceived these forecasts to be useful they were not felt to be accessible or understandable; (ii) users tended to prefer familiar formats when receiving information about forecast probabilities; (iii) preference for receiving information about forecast probabilities in different formats was linked to level of statistical expertise (with those with greater expertise having a stronger preference for receiving full confidence intervals); and (iv) a large proportion of current users were not receiving information about how well forecasts performed relative to observations (Figure 3), suggesting that information about forecast quality (indicated by measures of reliability and skill) was either not being communicated to many users, or not being communicated in a way that was well understood.
• Taken together, this highlighted a pressing need to identify more understandable formats for communicating forecasts, especially information related to forecast skill.
b) A review of existing approaches
• In reviewing existing approaches to communicating uncertainty, we examined those formats and visualisations that were currently in use with respect to S2D predictions, and reviewed relevant findings from the broader literature on risk communication and visualisation.
• Through examining existing communication strategies, it was established that while many ways of representing uncertainty in seasonal forecasts existed, relatively little end-user testing had been conducted.
• Our review of literature on the communication of risk and uncertainty identified several factors to consider in devising communication strategies, including: the need to ensure that climate visualisations do not use counterintuitive colour schemes (e.g. red to denote lower temperatures) (e.g. Kaye et al. 2012); that while aversion to uncertainty may lead some users to seek decision aids offering act/don’t act guidelines, it is important to ensure that the presence of uncertainty is not masked by this; the potential for “cognitive biases” to lead to systematic errors in interpreting information (e.g. Kahneman, 2011); and that decision-makers can vary considerably in their ability to use complex statistical and graphical information (e.g. Reyna, 2009).
• This review also identified the ‘progressive disclosure of information’ framework (Kloprogge et al. 2007) in which information is made available with different levels of technical detail, as a potential strategy for addressing the challenge of communicating forecast information to audiences where members differ in statistical/technical expertise.
c) Developing strategies for communicating levels of confidence
• Focussing on the challenge of communicating information about forecast probability and forecast quality (skill) to end users, a set of strategies that aimed to communicate this information to end-users were developed.
• Development of these strategies drew on the findings of the previous two tasks, along with input from work package partners, colleagues from WP32, and an external expert in visualisation.
d) Testing communication strategies
• To empirically test the effectiveness of a subset of the communication strategies developed in the preceding task, online decision labs were conducted with a broad group of decision-makers from climate sensitive sectors (Decision Lab 1), and a group of 98 highly engaged climate information users (Decision Lab 2).
• In the decision labs, participants were presented with forecasts using different formats, and asked questions to ascertain preference, understanding, subjective interpretation, perceived familiarity, perceived usefulness and how they might be used in decision-making.
• Key findings included: 1) preference for specific communication formats is related to familiarity but not objective understanding; 2) forecasts with no skill can have an undue effect on expectations about future conditions; 3) using qualitative categories (e.g. none, low, medium, high) to label skill scores appears to aid understanding amongst those with less experience of using this information.
• This task also enabled us to identify specific aspects of the different communication strategies that were being systematically misunderstood. For instance, white space on the bubble map (Figure 3a) was not always interpreted to mean ‘no useful information’, while it was found that participants presented with the bar graph (Figure 3d) mistook information about skill (Relative Operating Characteristic Skill Score, ROCSS) for information about likelihood.
e) Distilling key lessons and recommendations on the communication of confidence and uncertainty from the EUPORIAS project
In the final phase of WP33 we collated key lessons about the communication of uncertainty in S2D from both this work package and four of the EUPORIAS prototypes. Core lessons and recommendations include:
• No “one size fits all” solution exists for communicating confidence and uncertainty in S2D to end users, who have diverse needs and expertise;
• Where feasible, tailored communication strategies that are co-produced with users work best.
• Where tailoring is not feasible because users groups are too large and diverse, a layered approach to presenting information may address differing user needs;
• Forecast quality metrics such as skill can be difficult to understand. Guidance should be provided as to what these mean (e.g. through descriptions, ordinal categories);
• It should not be assumed that those communication strategies that are most preferred by users will be best understood. It is important to ensure that both requirements are met;
• Even when forecasts have no skill relative to climatology, they may have an undue influence on expectations about future states. We recommend that forecasts without skill not be provided by default unless specifically requested by users who understand that they do not provide ‘added value’;
• It is not always obvious that a particular aspect of a forecast visualisation will be systematically misinterpreted. New communications should always be tested with members of the intended end-user group to ensure that misunderstandings are addressed.
3.8 WP41 – Climate information and decision-making processes
The objectives of WP41 were to assess:
• The climate and non-climate-related options available to decision-makers;
• The value of Climate Information within the decision-making processes (DMP);
• The impact of S2D forecasts developed in RT2 onto DMPs.
The initial aim of the work package of assessing the value of decision-making per se could not be pursued within the EUPORIAS prototypes. This was mainly due to the novelty of SCF for those organisations involved in the prototypes. The lack of experience using SCF and the need for more time to build trust among the users in the different sectors made it difficult for them to identify a specific decision-making to be evaluated. A deeper understanding of users’ decision-making processes requires time and this time is needed to build trust (since details on decision-making constitute sensitive information in many sectors) but also to allow an understanding between scientists and users in novel topics such as SCF.
Given this situation, WP41 focused on the assessment of the value of using SCF in defined decision-making processes, a goal that was tackled from multiple perspectives and methodological approaches presented in a preliminary guidance document (D41.2; Soubeyroux et al. 2016). To fully understand the value of SCF, it was important to account for both the quantitative and the qualitative benefits. To do so, the guidance document presented a bibliographic analysis of different relevant methods for assessing the value and impact of SCF in specific decision-making contexts. A critical aspect found in WP41 was how to effectively distinguish between “value” and “impact” and how to define these terms across the different methodologies. Since addressing both value and impact of the SCF proved to be a challenging objective due to the intrinsic complexities of each prototype in measuring either economic value in the sector of analysis (e.g. benefits and costs associated to multiple water uses of a dam), by providing an indication of the effect that SCF can have on highly political decision-making processes, each methodology focused its attention mainly on one of the two aspects – the value or impact of SCF in the DMP. The detailed results and information from each methodology can be found in D41.3/D41.4 (Bruno Soares et al. 2016).
The main findings of the WP were as follows:
a) Decision maps (Land Management tool prototype)
• The assessment of the impact of the SCF in the farmers’ decisions was pursued via an interactive workshop in January 2016 followed by in-depth interviews in April 2016. Of the six farmers interviewed two farmers changed their normal course of action based on the SCF provided to them.
• The two farmers that used the SCF did it to help them plan their activities and, in one case, also to help manage their expectations with regard to their financial conditions. Although it was not possible to determine an economic value in terms of the decisions taken based on the information provided by the SCF, both farmers claimed to have saved money as changing the decisions based on the SCF prevented them from spending unnecessary money. Both farmers agreed that if the SCF were available they would keep using it as an additional piece of information to their decision-making processes.
• Another critical finding from this study was the need to build trust in the SCF. The novelty of this type of forecasts and the content of the information provided (i.e. average conditions over 3-month period) meant that most of the farmers involved in this analysis mentioned that they needed more time to gain confidence and trust in the information provided.
b) The Weather Roulette (RESILIENCE prototype)
• The Weather Roulette is a diagnostic tool that provides information about the skill of a forecast by providing an economic and financially oriented assessment in terms of effective interest rate and return on investment.
• One of the results was the demonstration that effective interest rate is a direct measure of the performance of SCF compared to climatology and that return on investment increases exponentially with the skill. This upholds the methodology as an engaging tool to explain the concept of skill and probabilistic predictions value to non-experts within the energy industry.
• More importantly, it provides a tool to analysts and technical staff within energy companies so they can use them internally to report their own benchmarks of SCF against climatology.
c) The Placebo concept (RIFF prototype)
• The Placebo protocol was followed by Etablissement Public Territorial de Bassin Seine Grands Lacs over the past 29 years and SCF shows a little added value versus RAF (historical simulation) but another important result has been that the current method used by the stakeholder is significantly worse than SCF and RAF.
d) Cost-Benefit Analysis (LEAP prototype)
• LEAP forecasts can be a critical component of raising funds in time to facilitate an early response to food security emergency. When this response is funded and triggered early, benefits arise as households avoid negative coping strategies, engage in greater investment, and avoid long term impacts to household growth, nutrition and educational outcomes.
• As a result, when the benefits of early response are incorporated into the analysis, responding based on forecasts becomes the most cost effective option in this analysis, allowing for savings of US$125M over the “business as usual” option.
e) Avoided cost (Climaware case study)
• Different types of SCF were introduced in a simple water management model and compared with results using only climatology. The hindcast period considered more than 40 years – for some of them there was a clear opportunity in using SCF but the change in DMP and avoided cost (related to flood and drought) could not be evaluated very precisely.
In parallel to the development and assessment of these methodologies, WP41 assessed the potential of the Regional Climate Centres (RCC) of the WMO European Regional Association for disseminating climate information products and their impacts on key stakeholders, especially the NMHSs (D41.1; Funk, 2016). This was assessed by an online survey that examined the characteristics and extent of products developed by NMHSs with respect to their sector-specificity and time-scales.
Furthermore, the impact and relevance of RCC products was also assessed. One product was particularly assessed since it had special relevance for the EUPORIAS project: the Climate Watch Advisories (early warnings) and its use and relevance for NMHSs. The results indicate a broad use of RCC products by NMHSs. Although all RCC products are considered to have a minimum relevance for most of the NMHSs, they are mostly used for the development of general national and sub-regional climate information products.
These products have to be further specified and tailored for the development of user- or sector-specific products by the NMHSs by using further sector-specific information on a national basis, which is mostly not available at RCCs because end-users are instead served by the NMHSs. Exceptions are seasonal forecast products, which are only used by a few NMHSs but are used permanently and have high relevance for the users. The potential use for user-specific applications is generally high. Climate Watch Advisories also have a general relevance for all NMHSs and have a high potential use covering weekly to seasonal time-scales as well as impact-related information relevant for sector- and user-specific climate information. RCCs, as providers of regional climate information, may become key institutions for laborious and capacity intensive tasks like seasonal forecast products. However, the pathway of regional climate service development and dissemination conditioned by RCC structures (top-down) may always require an additional user- or sector-specific tailoring of climate information on a national level.
3.9 WP42 – Climate service prototypes
The objective of WP42 was to develop an experimental operational prototype of a climate impact prediction system for a subset of the case-studies/sectors selected.
Given the ambitions of the project in terms of co-development it was decided that the definition of the climate service prototypes needed to be based on an objective evaluation of their specific merits in addressing user requirements. This meant that, contrary to other projects of this kind, no prototype was defined at the outset but they were instead dynamically defined during the course of the project. WP42, in close coordination with the project management team, opened a discussion on the criteria that make a climate service successful. Two of them were given disproportionate weights. On the one hand it was felt that the existence of a well identified user with a clear decision that can be informed should be considered as a necessary condition, on the hand it was felt that also the strength of the relationship linking the climate driver to the impacts and the predictability of that driver should be considered as an essential element in the evaluation of the prototype.
The criteria were defined and agreed upon ahead of the first project general assembly. The criteria were then passed to a panel of experts external to the consortium alongside the proposal for evaluation. All the proposals were considered eligible, i.e. above the threshold for the two required criteria, but in order to maintain an in-depth focus only a subset of them were selected for development. These were: LMTool, led by the Met Office and focusing on agricultural applications in the UK; RESILIENCE, led by BSC and focusing on wind power production world-wide; RIFF, led by Meteo France and focusing on freshwater resources for Paris; LEAP, co-led by ENEA and WFP and focussing on food security in Ethiopia; SPRINT, led by the Met Office and focusing on climate advice for the transport sector in the UK.
The dynamic development of the prototypes had forced all people involved to identify and involve end-users from the very beginning, something that has definitely improved the interaction between providers and users of climate services. Being part of the process from day one has gained buy-in to the project from the users that they would not have done so otherwise. The fact that some of those users such as the Clinton Devon Estate who we targeted in the LMTool, attended all our meetings and used their own communication channel to promote the work we were doing as part of the project is a measure of the success of this approach.
The team behind each prototype then worked at the assessment of the potential benefits the target users could get from the use of the prototype. This was first done at an abstract level then it was repeated using historical data as an input. Towards the end of the project each prototype was supposed to provide a real-time forecast to the users. This effectively happened for all but one prototype, and in at least three cases (RESILIENCE, SPRINT, and LMtool) the work done within EUPORIAS has then continued beyond the end of the project.
As well as the climate service prototypes the WP has also delivered a number of case-studies based on the prototype proposals that had not been selected for development. Whilst these case-studies did not have the same level of support and funding from the project, the work done within EUPORIAS provided the basis for projects that in some cases have then been developed into full projects elsewhere, as is the case for the Pro-Snow proposal led by TEC.
3.10 WP43 – Stakeholder engagement
The objectives of WP43 were to engage with European citizens about the use of S2D data in everyday decisions, to demonstrate ways in which a climate service can be developed to address specific users’ needs, to facilitate clear communication and exchange of information with stakeholder groups; and to empower SMEs, allowing them to develop their own climate services.
The major outcomes of this WP are as follows:
a) Stakeholder engagement
• A large number of meetings and workshops with stakeholders were held as part of this WP. The workshops were used to assess the users' understanding of the products delivered and provide an invaluable opportunity to understand ways in which we can improve forecast content, presentation and relevance, therefore helping to improve the service provided.
• FutureEverything organised a EUPORIAS lab during their festival in Manchester in March 2016. As part of this lab, a highly visible workshop with EUPORIAS partners was held. The target audience included non-specialists in industry and the general public. The approach and activity consisted in use of data design to communicate probabilistic information, and use of participatory experiences and events to engage various publics in science and innovation.
• In order to communicate some of the results of the project to an audience of non-specialists, two short video documentaries were produced by FutureEverything. The first of these videos presents project Ukko (http://project-ukko.net) the interactive web-browser based visualisation interface designed to provide stakeholders with access to the seasonal wind forecasts given by the RESILIENCE prototype. This video (http://vimeo.com/152558782) has been embedded in the project Ukko homepage and has been viewed nearly 10,000 times since being published in March 2016. The second video (http://vimeo.com/185503156) summarises the whole EUPORIAS project. It tells the story of the project through its aims, processes and outputs. The film includes interviews and narration from project partners and stakeholders.
b) Technical facilities
• During workshops with stakeholders, the possibility of having a mobile phone application was identified as key to bring the climate services developed closer to the user. As a result, in parallel to the corresponding microsite (http://lmtool.euporias.eu) a mobile app has been created for the LMTool prototype (Figure 1). This app is freely available for Android (http://play.google.com/store/apps/details?id=es.predictia.lmtool&hl=es) and Apple iOS (http://itunes.apple.com/us/app/lmtool/id1118042668?mt=8). It allows for visualising all the contents of the microsite and provides password-protected (as required by the Met Office) access to the short-term and seasonal forecasts delivered by the prototype for South West England.
• A second mobile app related to the RESILIENCE prototype has been created. This app (which will be published both in Google Play and the App Store) allows the user to assess the economic value of the wind speed seasonal predictions provided by the prototype, as compared to a climatological forecast. This is done by playing a roulette game from an interactive screen. In particular, the user can introduce the amount of money they would like to bet and the app simulates how much they would have won/lost if using the available probabilistic predictions. As in the case of the LMTool app, the EUPORIAS Application Planning Interface (API) is used to access the predictions provided by RESILIENCE.
c) A guideline for developing climate services
• As a result of the work carried out in this WP, a broad community of climate service developers (including forecast providers, scientists and communicators) and users has been established, which constitutes one of the major outcomes of this WP.
• The scientific paper produced to meet deliverable 43.4 (“Develop a mobile application to help access information produced by climate prototypes”) describes, for the illustrative case of the LMTool prototype, the steps that need to be followed in order to take advantage of the S2D climate information in land use decision-making. The complete development of the service is presented, from the first user engagement phase to the final products that are being delivered. The paper is currently being reviewed by the different co-authors and will be submitted to Climate Services for its inclusion in the EUPORIAS special issue that this journal is preparing for 2017.
3.11 WP44 – Delivery tools
The objectives of WP44 were to identify, develop, and maintain an interface that will allow an effective delivery of the climate services developed in WP 42 to both the general public and the relevant decision-makers. In this manner, EUPORIAS goes beyond merely improving the reliability of the underlying prediction systems to enhancing the usability of these forecasts in practical applications.
WP44 started in month 21, but WP leaders joined the earlier WP42 workshops and contributed to the development of the prototypes, enabling them to follow the developments in EUPORIAS before the official WP44 start. Their flexible approach allowed them to change and adapt quickly to achieve results. The main findings of the WP were as follows:
Task 44.1: Inventory of existing climate data portals and tools
• This task provides an inventory of existing climate data portals and gathers interface requirements on how the project climate service prototypes can be integrated into these existing portals;
• There are relatively few portals (Climate-ADAPT & climate4impact) that could host almost all prototypes and only Climate-ADAPT could host them all;
• Private company portals were excluded because of their data policy limitations;
• Many of the other portals are relevant for specific sectors or regions and it would therefore be desirable to formulate an architecture that would also allow appropriate prototypes to interface to these specific portals;
• To interface with these platforms both administrative and technical requirements must be addressed. Taking the Climate-ADAPT portal as an example the administrative effort includes obtaining permission from the portal administrators, ensuring the EUPORIAS prototypes meet the stipulated criteria for case studies, undergoing a review and annotation by experts outside of EUPORIAS. The technical effort would include assessing a range of integration approaches ranging from links to a dedicated website, integration of EUPORIAS datasets into their database or the addition of a tool.
Task 44.2: Climate service architecture development and improvements
• This task provided a plan detailing how to integrate the EUPORIAS prototypes into existing protocols inventoried in D44.1. In order to ensure flexibility in the data types, external portals, and external devices and to support efficient data exchange while ensuring proper control an API is proposed for the climate services.
• The EUPORIAS API was implemented (D44.5) and used by the LMTool prototype and in a mobile App related to the RESILIENCE prototype. The API is also used to publish EUPORIAS data on climate4impact.eu (C4I) and KNMI data centre (KDC).
• Making use of the EUSurvey service (http://ec.europa.eu/eusurvey) the API also allows the prototypes/case-studies developers to access the end-users' feedback in a comprehensive way.
• By providing an API to portal developers we can provide an abstracted interface, which hides the EUPORIAS prototype and case studies implementation details. In this way, portal developers can use a stable API, while EUPORIAS maintains flexibility in adding new or changed prototypes and case studies. This is shown in figure 4 below.
Task 44.3: Service delivery and operation
• This task aimed to enable a flow of relevant data from EUPORIAS to existing climate portals. The Service architecture (API) described above is implemented and tools have been created to bring the prototypes and case-studies developed in the project closer to the end-user, in particular the EUPORIAS microsites.
• A total of 11 microsites (one for each of the five prototypes and six case-studies developed in the project) have been created and delivered under the EUPORIAS domain. All of them share a common design and appearance, inheriting the visual identity created for the EUPORIAS website, and have the same general structure.
• They are user friendly interfaces for demonstration and promotion of the services developed within the project to both general public and relevant decision-makers. All these microsites have been gathered under the online component of the CUIP developed within the project, which can be reached at http://www.euporias.eu/cuip.
Task 44.4: Web interface for end users of the EUPORIAS service prototypes
• The API has been developed to ensure both the data and results emerging from EUPORIAS can be shared and, whenever appropriate, used by others, helping therefore to promote the project beyond its natural reach.
• Additionally, by making use of the EUSurvey service (http://ec.europa.eu/eusurvey) it also allows the prototypes/case-studies developers to access the end-users' feedback in a comprehensive way. All the code developed for this API is free and can be found here: http://github.com/Predictia/euporias-api. Additionally, the following page documents in details the API and provides some practical examples of use (for instance, how to connect to the API via bash, Java, Javascript and PHP protocols): http://api.euporias.eu/docs/api-guide.html.
Task 44.5: Generation and dissemination of easily accessible services.
• The aim was to reach a large audience of non-specialists. The process entailed selecting the relevant project results and services, translating these into easily accessible formats, and the production and publication of the messages.
• During the project, it was determined there was scope for T44.5 to draw on innovative methodologies and media in art and design to realise the WP objectives, above and beyond the tools described in the EUPORIAS description of work (YouTube, podcasts). A data visualisation interface was developed for one of the prototypes, along with a range of means to engage wider audiences, including a digital culture festival, in addition to the media described in the description of work.
• A design process was undertaken, through which it was determined a climate service visualisation interface would be developed with one of the prototypes. The result was the Project Ukko (www.project-ukko.net) visualisation interface developed for the RESILIENCE prototype.
• High visibility events were utilised, centred on a special edition of the FutureEverything festival, the UK's digital culture festival, established in 1995. Other events included Information+ in Vancouver and Liverpool International Business Festival. PR, media and podcasts were deployed to generate wider reach and engagement.
3.12 WP45 – Climate services as a business opportunity
This WP aimed to assess whether a market for climate services in Europe exists and whether this can be effectively and efficiently used by SMEs and other relevant stakeholders. A general methodology was developed and applied to specific key sectors to check the economic viability of climate service prototypes produced in WP 42 of this project.
The main findings of the WP are as follows:
• WP42 developed an experimental operational prototype of a climate impact prediction system for a subset of the case studies/sectors selected. WP45 has developed a general market analysis methodology to assess business opportunities of climate services and has applied this methodology to some of the prototypes developed in WP42 (water sector, renewable energy, road maintenance, and tourism).
• The methodology developed: (a) affords a general overview of a market (both the variables that affect an industry, and analysing the most relevant aspects of current and potential demand and supply); (b) an assessment of new services developed, from the point of view of their economic feasibility; (c) gives insight into a range of current innovative business models used, and hence helps assess which could be more suitable for the prototypes.
• The climate services industry in Europe is considered to be immature for the seasonal to decadal timescale in the sectors studied (water sector, renewable energy, transport and tourism). The wind energy sector seems to be somewhat more established than the others.
• All sectors identify the same current gaps in climate services information (e.g. lack of skill of forecasts, insufficient accuracy, etc.) and possible actions are suggested to overcome these gaps.
The main conclusions regarding the methodology developed are as follows:
• For each step, different business analysis tools and approaches have been proposed, tested and adapted. Thus, the methodology is now robust enough to be used for assessing other climate services external to the project.
• Relevant information has been gathered about markets (needs, gaps, competition, etc.) for the key sectors studied. This could be an excellent basis for the realisation of future market studies of new climate services in Europe.
The main results of the application of the methodology are as follows:
• All the sectors analysed are sensitive to weather and climate conditions, to a greater or lesser degree;
• All the sectors analysed think that the Copernicus Climate Change Service is going to have a positive influence on the development of the climate services industry;
• The information gathered during demand analyses based on WP12 interviews is of great significance for the market analysis (especially the information on willingness to pay). To have a better representativeness, additional interviews should take place;
• All sectors identify the same current gaps in climate services information, which are the lack of skill of forecasts, and the insufficient accuracy;
• The feasibility study conducted for the key sectors shows significant socio-economic benefits. For example, for the UK transport sector prototype (SPRINT) there are potential cost savings of ~£304 million/year using the 1-month temperature forecasts, and potential cost savings of ~€13,500 from scheduling five days of maintenance of a wind farm of 10.5MW turbines on days that turn to have good wind conditions for producing electricity.
References
Beven, K. (2012) Rainfall-Runoff Modelling: The Primer, Second Edition. Willey Library. Print ISBN: 9780470714591, doi: 10.1002/9781119951001
Bruno Soares, M., Bosi, L., Pouget, L., Reitg, E., Jiménez, I., Lledó, L., Soubeyroux, J.M. Viel, C., Creswick, J. (2016) Report on the evaluation of the value and impacts of using seasonal climate forecasts in decision-making processes. EUPORIAS D41.3 and D41.4. Available at: http://www.euporias.eu/system/files/D41.3%20and%2041.4_Final.pdf
Bruno Soares, M., Dessai, S. (2016) Barriers and enablers to the use of seasonal climate forecasts amongst organisations in Europe. Climatic Change, 137(1-2), 89-103
Butts, M., Xu, Y., Mueller, H.G. Jensen, R. A., Larsen, J.K. Godiksen, P.N. Richaud, B., Larsen, P. T., Corcuera Barrera, M.Á. (2016) Seasonal Forecasting for Water Resource Management: Reservoir inflows and uncertainties. Presentation at the XXI International Conference Computational Methods in Water Resources, CMWR 2016, University of Toronto, Canada, 20–24 June 2016
Donnelly, C., Andersson, J.C. Arheimer, B. (2016) Using flow signatures and catchment similarities to evaluate the E-HYPE multi-basin model across Europe. Hydrological Sciences Journal, 61(2), 255-273
Fraga, H., Malheiro, A.C. Moutinho-Pereira, J., Santos, J.A. (2014) Climate factors driving wine production in the Portuguese Minho region. Agricultural and Forest Meteorology, 185, 26-36
Funk, D. (2016) Review of the climate information products of the WMO RA VI RCC-network useful to the decision-maker. EUPORIAS D41.1. Available at: http://www.euporias.eu/system/files/D41.1_Final.pdf
Kahneman, D. (2011) Thinking, fast and slow. Macmillan
Kaye, N.R. Hartley, A. Hemming, D. (2012) Mapping the climate: guidance on appropriate techniques to map climate variables and their uncertainty. Geoscientific Model Development, 5(1), 245
Kloprogge, P., van der Sluijs, J., Wardekker, A. (2007) Uncertainty communication. Copernicus Institute for Sustainable Development and Innovation, Report NWS-E-2007-199
Manzanas, R., Frías, M.D. Cofiño, A.S. Gutiérrez, J.M. (2014): Validation of 40 year multimodel seasonal precipitation forecasts: The role of ENSO on the global skill. Journal of Geophysical Research: Atmospheres, 119(4), 1708-1719
Manzanas, R., Brands, S., San-Martín, D., Lucero, A., Limbo, C., Gutiérrez, J.M. (2015) Statistical Downscaling in the Tropics can be Sensitive to Reanalysis Choice: A Case Study for Precipitation in the Philippines. J. Climate, 28, 4171-4184
Manzanas, R., Lucero, A., Weisheimer, A.,Gutiérrez, J.M. (2017) Can bias correction and statistical downscaling methods improve the skill of seasonal precipitation forecasts? Clim Dyn, doi:10.1007/s00382-017-3668-z
Ogutu, G.E.O. W.H.P. Franssen, I. Supit, P. Omondi, and R.W.A. Hutjes (2016) Skill of ECMWF System-4 Ensemble Seasonal Climate Forecasts for East Africa. Int. J. Climatol., doi:10.1002/joc.4876
Osborne, T., Gornall, J. L., Hooker, J., Williams, K., Wiltshire, A., Betts, R. A., Wheeler, T., (2015) JULES-crop: a parametrisation of crops in the Joint UK Land Environment Simulator. Geoscientific Model Development, 8(4), 1139-1155
Pouget, L., Roldán, T., Gómez, M., Cabello, À., Camino, E.R. Navascués, B., Voces, J., Comas, P., Pastor, F., Gómez, M.C.G. Gil, J.J. (2015) Use of seasonal climate predictions in the water sector—preliminary results from the EUPORIAS project. Drought: Research and Science-Policy Interfacing, 321
Reyna, V.F. Nelson, W.L. Han, P.K. Dieckmann, N.F. (2009) How numeracy influences risk comprehension and medical decision-making. Psychological bulletin, 135(6), 943
Soubeyroux, J.M. Bruno Soares, M., Pouget, L., Jiménez, I., Torralba, V., Ortega, D., Viel, C., Bosi, L., Buontempo, C. (2016) Preliminary Guidance Document on the evaluation of the value of DMP. EUPORIAS D41.2. Available at: accessed 01.12.16
Taylor, A.L. Dessai, S., de Bruin, W.B. (2015) Communicating uncertainty in seasonal and interannual climate forecasts in Europe. Phil. Trans. R. Soc. A, 373(2055), 20140454
Williams, K., Falloon, P. (2015) Sources of interannual yield variability in JULES-crop and implications for forcing with seasonal weather forecasts. Geoscientific Model Development, 8(12), 3987-3997
Williams, K., Gornall, J., Harper, A., Wiltshire, A., Hemming, D., Quaife, T., Arkebauer, T., Scoby, D. (2017) Evaluation of JULES-crop performance against site observations of irrigated maize from Mead, Nebraska. Geoscientific Model Development, 10(3), 1291-1320
Potential Impact:
We have identified four main areas of potential impact, and main dissemination activities and exploitation of results:
1. Science: particularly downscaling methods and their value in climate services; the use of climate information indices; communication; impact modelling
2. Service development: particularly the user landscape; benefits and impacts of the EUPORIAS user data gateway and the R statistical software language; development of climate service principles; lessons to be learned from the development of the prototypes
3. Policy impact: particularly the ECOMS white paper; project governance; a cost benefit analysis
4. Business impact: particularly Project Ukko; the weather roulette; a mobile app; benefits and impact of the business analysis of climate services
4.1 Science
4.1.1 Downscaling methods and their value in climate services
There is great demand for user-relevant climate information to guide planning decisions at regional and local scales. This demand includes information about past weather and climate from observations, and future predictions. EUPORIAS examined the role of different downscaling methods in improving the skill and user-relevance of information at the regional and local level.
In broad terms, downscaling involves mathematical methods that take in data from large-scale simulations of the climate system, using global climate models, and outputs finer scale data that is more applicable to regional and local scales. Although there is a growing body of evidence on the relative strengths and limitations of different downscaling approaches, most studies focus on long-term climate change time-scales. In EUPORIAS we assessed the value of downscaling seasonal predictions, using different methods, to determine if downscaling approaches add value to the regional information provided by global climate model simulations. Computer model experiments were conducted using two different types of downscaling: statistical and dynamical.
Statistical models use the knowledge about the statistical characteristics of local observations to alter the outputs of a global model. These procedures allow the model outputs to look similar to the observations. Dynamical methods, using regional climate models, apply similar mathematical and computational approaches used in global climate models but over a limited area (usually covering a number of countries) to produce finer scale information. Dynamical and statistical methods were tested for the Greater Horn of Africa region, since this region provides a complex topography, has a reasonably strong influence from the climate phenomena El Nino, and, through the WFP, has a strong use-case where the impacts of the climate are experienced by society. Analysis focused on rainfall, as a key variable of interest for decision makers, and was conducted on a series of years that included both dry and wet years.
A significant challenge in the assessment of “added value” was evaluating the real observed state of the atmosphere. Observations in this region of the world are relatively poor, placing significant limitations on the quality of the analysis and strength of the conclusions. This stresses the importance of reliable long-term observational datasets in underpinning the development of climate services. Despite the challenges provided by insufficient observations, it was possible to assess the skill of the downscaled simulations compared to the skill of the global climate models.
Results show that there is no evidence of additional skill from the regional climate models in predicting the regional rainfall on seasonal time-scales. This result holds true even when looking at sub-national regions. This implies that there is no added value of using regional climate models compared to global climate models in this region (within the context of the experiments conducted) to feed into downstream products and services, such as the World Food Programme LEAP platform (drought early-warning system). On the other hand, the results suggest that regional climate models may provide improvements in the “reliability” of seasonal forecasts – that is to say the climate system is more realistically simulated, even if this doesn’t result in improvements in skill.
The results for statistical downscaling may show more promise. For the summer season the results suggest that the use of statistical downscaling has similar performance to regional climate models at a fraction of the cost and thus provide a more cost-effective solution to downscaling seasonal forecast information to provide more detailed spatial information for use at the regional and local scale. Nevertheless, it is difficult to generalise the results obtained from these experiments and the conclusions may differ for other regions and climate variables.
4.1.2 The use of climate information indices
The lack of user-relevant variables has long been recognised as one of the barriers limiting the use of climate information in sectoral decision making. Climate data, as supplied from models or forecast systems, is often not directly related to user decision needs, and can therefore be difficult for users to apply in their decision making. Climate variables, such as temperature, humidity or wind can easily be translated into Climate Information Indices, which offer the possibility of describing climate in a more user-relevant way. As an example, agriculture or water resource users may find the index ‘consecutive dry days’ more directly useful and relevant than mean precipitation data. Similarly, indices such as heating degree days can be linked directly to user-relevant quantities such as energy demand. As an additional factor, these indices can replace more complex and computationally expensive impact models, and may be of similar value to users.
To address this challenge EUPORIAS devoted a work package to the transformation of raw climate data into sector-relevant indicators and to the evaluation of their skill. The expectation was that the indicators could allow more users to interact with the data directly. There was also the hope that in some cases the calculation of an appropriate index could positively impact on key verification attributes of the forecasts, such as reliability and skill. Considering that most user-relevant indices are based on daily data, EUPORIAS developed a procedure to bias-correct daily time-series. All indices were then assessed using observations or reanalysis as a benchmark. The results suggest that, leaving aside the evident engagement value that using a tailored index brings, the verification attributes of the forecast are not significantly affected by the calculation of such indices. For simple post-processing of single variables (e.g. accumulation over time or exceeding a threshold) EUPORIAS demonstrated that the indices cannot have more skill than the underpinning variable. On the other hand EUPORIAS showed that in the case of complex multi-variable indices there might be occasions locally where the index shows more value than the variables used for its calculation.
One application of the findings of this work is in the forecast of wildfire risk. Wildfires represent a significant hazard in the Euro-Mediterranean region, in particular to land managers and populations in the rural-urban interface. Conversations and a survey circulated to a range of decision makers suggested that estimating fire risk a few months in advance could provide significant benefit, allowing appropriate pre-season land management, and allowing fire protection agencies a timely reaction and an adequate provision of resources. Within EUPORIAS we considered whether the skill in dynamical seasonal predictions of a climate information index describing fire weather risk (the Canadian Fire Weather Index, [FWI]) is sufficient to provide value to such decision-makers. It was found that there is significant skill in predicting above average summer FWI in some parts of the Euro-Mediterranean region at 1 month ahead. This skill was largely limited to an area covering Greece, Bulgaria and Turkey and displayed the potential for development of a useful seasonal forecast service in this area; however skill elsewhere was not sufficient to support service development. There is some suggestion from this work that skill levels could be improved by using a subset of the FWI components, some of which may more closely link climate to fire risk in this region. In addition future work may consider whether the skill of seasonal FWI predictions could be improved for some locations through the use of alternative statistical downscaling techniques where longer historical records are available.
A further study considered the value of sub-seasonal-to-seasonal climate forecasts of extreme temperature indices, and questioned whether reliable forecasts could allow for improved resource management for heat-health planning, protecting vulnerable populations and potentially preventing fatalities from temperature-related illness in Europe. In particular this study considered the extent to which such climate forecasts could be mainstreamed into heat-health planning, to support timely decision-making in advance of heat waves in Europe. Forecast data at lead times of 1 day to 3 months were used to drive a mortality model which produced probabilistic mortality forecasts running up to the 2003 heat wave in Europe. Through comparison with observed climate, this work found a range of levels of skill across the region and with different lead times. However, even at lead times of up to 3 months, some regions, particularly within Spain and the UK, did display significant skill in forecasting excess mortality. This work has suggested that for some regions at least, there is the potential for value in provision of mortality forecasts on sub-seasonal-seasonal timescales.
4.1.3 Communication
a) Communication with end users of climate information
For climate information to be useful in decision making, it has to be communicated in an understandable and actionable way whilst ensuring it does not lead to a false sense of certainty. EUPORIAS has focused on the communication of seasonal to decadal forecasts, although many of our findings are equally applicable to the presentation of climate information at other timescales.
To help us assess the strengths and limitations of different approaches to communicating climate information, including the representation of skill and uncertainty of the forecasts, researchers within EUPORIAS worked closely with the end users of the information, including those involved with the development of our prototype climate services. An important driver was that this research was led by experts from the social sciences who worked alongside climate scientists and other project partners to gather and analyse evidence. Some key findings were:
• No “one size fits all” solution exists for communicating seasonal to decadal climate information to end users, who have diverse needs and expertise. If possible, it is best to co-design and test the product with the end user, helping to ensure that the communication technique chosen does not mislead but aids the decision making process.
• Where tailoring is not feasible, a layered approach to presenting information alongside clear guidance to interpretation may address differing user needs.
• Even when forecasts have no skill relative to climatology, they may have influence on expectations about future states. We recommend that forecasts without skill not be provided unless specifically requested by users who understand that they do not provide ‘added value’.
Several research activities have developed a similar user-centric approach, with some directly inspired by EUPORIAS. The research techniques, sample products and tools developed in EUPORIAS will help the producers of climate information (co-)design their products to be more useful, actionable and robust.
b) Communication with the climate services and climate research communities
As well as journal publications, conferences and other traditional forms of communication, EUPORIAS engaged with other researchers though social media, including Twitter (@euporias). It created a summary infographic and supporting factsheets about the prototypes. We ran two masterclasses, particularly for early career researchers interested in the development and delivery of climate services. These were held at EURAC in Italy in May 2015 and May 2016. Alongside a varied programme of lectures, the students created example climate services for real-life user requirements. Such a hands-on a formula worked well and both masterclasses were very well received.
c) BellHouse – a novel communication activity within EUPORIAS
In the Spring of 2016, the UK-based arts organisation Kaleider and EUPORIAS made an International Commission Call inviting artists to submit ideas for playable artworks to be debuted at the EUPORIAS General Assembly at the Met Office in October 2016. We received over 60 applications worldwide and were overwhelmed with the quality of ideas. We commissioned Roop Johnstone from RAMP Ceramics to create his exquisite playable artwork – BellHouse.
BellHouse is a playful, interactive sound sculpture that translated the non-verbal communication of the delegates presenting at the General Assembly into the chimes of 35 bells in an opened sided house. A motion capture system devised by the Met Office Informatics Lab activated striking mechanisms associated with each ceramic bell generating a continuous chiming whilst each speaker at the assembly presented their research.
BellHouse was a novel collaboration. It put climate scientists in the shoes of others. Speakers at the meeting could listen to a playback of their talk, as played by the BellHouse. Although the sound they heard was a true reflection of their movements as they spoke, it was almost unrecognisable as their speech – perhaps reflecting how misunderstood even the most simple of messages can be. BellHouse explored how the arts and research can work together to create something beautiful, memorable and engaging and is one of the more unusual legacies of EUPORIAS.
4.1.4 Impact modelling
EUPORIAS completed a coordinated set of impact model studies across different sectors (agriculture, water, forestry and transport), over different regions and scales (Europe, East Africa, specific areas and basins), driven by baseline forcing data and different seasonal hindcast datasets (Figure 2). Some of the studies have also investigated the role of uncertainty in overall impact skill, such as arises from bias correction, initial conditions, seasonal hindcast skill, and model parameters.
Impact models for crops, water, forestry and energy were further developed, and impact hindcast simulations and analysis were defined, contributing to improved European capability for assessing long-range weather and climate impacts across these sectors. The findings help inform recommendations for impact assessments using seasonal forecasts across Europe and East Africa including the need to:
• Consider appropriate methods of bias correction of forecast data for applications where thresholds are important
• Consider how impact skill varies with variable, region, season, lead time and forcing dataset
• Note that skill for impact fluxes may be greater than that for state variables and that skill for some impacts such as hydrology, may be larger than that of its meteorological drivers
• Consider requirements for impact model initialisation. For example, skill for hydrology is often related to memory effects from initialisation of soil moisture and snow mass; in contrast, in East Africa where soils are often dry prior to planting, crop model initialisation may not be necessary.
In particular, some sectors, regions and seasons show greater potential for confident application of seasonal impact forecasts, including:
• For hydrology in North Europe, with better skill during winter than in summer, and better skill for high flows versus low flows
• For Swedish forestry, temperature forecasts have the potential to support planning of harvest operations, and soil moisture forecasts could support the planning of planting activities
• Fire Weather Index showed considerable skill in the eastern Mediterranean
• A capacity factor based on wind speed and temperature has the potential for assessing the future variability of wind energy resources
• Translation of predicted winter air temperatures, translated to road skin temperatures, show potential for road transport applications
For such applications, appropriate use of seasonal impact forecasts could aid preparedness and planning, and may avoid negative impacts such as: water shortages, flooding, crop failure and food shortages, poorly planned forestry operations, fire damage, energy and transport planning. For some impacts proper skill assessment is hindered by lack of high quality, spatially distributed observations, most notably so for crop yields (in Eastern Africa) so obtaining suitable datasets would strongly increase the potential for seasonal timescale food security assessments in these regions.
4.2 Service Development
4.2.1 Our user landscape
Before EUPORIAS, few research projects truly put the users of climate information at the heart of the research. Today it is not uncommon to see users helping shape research proposals and working closely alongside the providers of climate information within projects. EUPORIAS helped lead the way, working closely with our users throughout the four year project. In recognition of this, Steve Zebiak, former Director of the International Research Institute for Climate and Society and Chair of the international Climate Service Partnership, said ‘EUPORIAS is the most comprehensive research program in climate services today, anywhere.’ at one of the EUPORIAS General Assemblies.
Through exploring diverse sectors and gathering a robust evidence base to help inform the development of the prototypes, EUPORIAS engaged with the users, and providers, of climate information across Europe. It did this through online surveys, an extensive programme of interviews alongside a review of the existing literature. The results suggest that the user landscape, and potential user landscape, within EUPORIAS was undoubtedly complex.
The knowledge gathered was analysed, communicated through the project deliverables and journal articles and presented at conferences. As the climate services landscape develops, it is vital that we share such knowledge, while acknowledging commercial considerations, so that the climate services community can benefit from such in-depth research. User fatigue is of concern but by working across providers and sharing our understanding of the user landscape, we can seek to maximise the relevance of climate information across sectors and improve its use in decision-making.
4.2.2 Benefits and impacts of the User Data Gateway (UDG) and R statistical software language
The ability of research partners to access and use the data produced by EUPORIAS was key to the success of the project. The ECOMS User Data Gateway (ECOMS-UDG), combined with the use of R (an open source programming language and software environment for statistical computing and graphics) to explore the data, ensured that all data used within EUPORIAS was available in a common accessible location.
Tested and quality controlled hindcast seasonal datasets made available to EUPORIAS, SPECS and North Atlantic Climate (NACLIM) partners through the ECOMS-UDG include ECMWF System 4, Met Office GloSea5, Météo-France System 4, and National Centers for Environmental Prediction Climate Forecast System v2 (NCEP CFSv2), along with access to SMHI-EC-Earth hindcast data, ECMWF ERA-Interim, and E-OBS. The ECOMS-UDG is hosted by the IT infrastructure of the University of Cantabria. A set of tools provided within the ECOMS-UDG, written in R, allows data to be manipulated. Predictia has provided user support for the R data access tools.
Benefits of the ECOMS-UDG include: project data being available from a common location; data are available in a predefined format; development and access within the ECOMS-UDG to a standard set of data processing tools, written in R, tailored to the needs of stakeholders; use of a standard input/output protocol to prevent duplication of pre-processing steps and facilitating shared analysis and multi model ensemble creation.
The ECOMS-UDG provides over 64TB of data to forty users from EUPORIAS and fifteen users from the SPECS projects, allowing both cutting-edge research and prototype climate service development. To gain access to ECOMS-UDG the user has to agree to the licensing and distribution terms and conditions required by each institution providing data. This highlighted the restrictions and limited availability of data from Global Producing Centres. For a number of impact applications, both daily data and continental geographical scale is required. However, restrictions are often placed on the temporal and spatial extent of released data. This led to some agreements being signed to make more data easily available to partners, and the ambition is that the Copernicus Climate Change Service will provide further free access to useful data for impact applications.
4.2.3 Development of climate service principles
The production of the prototypes highlighted several challenges and best practice ways of working. These were considered and collected in November 2014 at a workshop of thirty international experts in climate services. Through an interactive and dynamic workshop the attendees identified seven principles that should be considered when developing successful climate services. These principles helped to influence the prototype development within EUPORIAS and will be of use to the wider climate services community. The seven principles are illustrated in Figure 7, and covered the following broad topics.
Who are the users and possible users of the climate service? What is the user-chain? What are the motivations of each participant to take part in the project? Are all the relevant people involved in the discussion? Do the providers have all the skills needed to deliver the service? What expertise will the users bring to the service development? It is essential that the scope is clearly defined at the beginning of the project AND to ensure there is a common understanding about how it is evolving. Be honest about what is and is not achievable within the project. Be open about new ideas that can alter the course of the project. Communicate all the possible issues (scientific, technical, legal, political or commercial) which could limit the service at the start and throughout development. Take the journey together. The service should provide value to users but it is also important to identify value (not necessarily monetary) to the provider. Make clear what each actor involved is expecting to get out of the service. Be flexible, expect changes in the scope. Maintaining a highly interactive and flexible work-programme you will be able to account for some of those changes, while considering the boundaries of flexibility. Scope-deliver-evaluate: iterate. If possible divide the service in small components that can be delivered separately. Scope each of these, deliver and evaluate them with the users and then, if necessary, re-scope. Some project management practices (e.g. agile) are intrinsically designed for these sort of applications.
4.2.4 Lessons to be learned from the development of prototypes
EUPORIAS developed five climate service prototypes (with a sixth funded outside of the project and developed by SMHI). These were LEAP (for food security), RIFF (water management), SPRINT (transport), LMTool (land management) and RESILIENCE (wind energy). EUPORIAS sought to involve users directly in the governance of the project and include them in the problem definition. The prototypes were selected through a competitive process focussing on ensuring a tangible user drive and being likely to provide useful results. Many challenges arose during the production of the prototypes and valuable lessons, applicable to many other climate service development activities, are summarised here:
• The interaction with users during the development of a climate service cannot be sporadic and cannot simply occur at the beginning (e.g. service definition) and at the end (e.g. service evaluation) of the service development.
• It is essential to allocate sufficient time to the dialogue with the user and to any required changes in the definition and scope of the service being developed.
• In addition to the benefits that users could gain from a climate service tailored to their needs, access to climate expertise during the development of the service is an important added value to the users.
• Top-down management practices are not necessarily the most suitable for developing climate services. Adopting a flexible management approach (e.g. Agile) can be an advantage in an environment where changes in scope in response to user feedback are to be expected.
• User representation (or lack thereof) in the governance structures of climate service projects and the way in which these projects are linked to downstream business opportunities have a direct impact on their ultimate usefulness to society.
4.3 Policy impact
4.3.1 White paper
EUPORIAS, through the ECOMS coordinating mechanism, has been central to identifying and providing priorities, and research and investment needs in the field of climate observations, climate modelling and climate services. The ECOMS group (coordinators of EU FP7 climate modelling and climate service projects, and representatives from European climate research and climate service centres) met in February 2013 and the conclusions of that meeting are summarised in a public white paper (Deliverable 2.1) which provides recommendations from ECOMS on the priorities for Horizon 2020 (Societal Challenge 5) in the field of climate modelling and climate service development, underpinned by observations. The EC has managed to take on board all of the recommendations from EUPORIAS and ECOMS in launching subsequent calls under Horizon 2020. ECOMS has been superseded and replaced now by the Climateurope (www.climateurope.eu) Coordination and Support Action under Horizon 2020, ensuring the legacy of ECOMS will be maintained and built upon.
The EUPORIAS coordinator has been involved in his capacity as ECOMS chair, and through his role under the UN’s GFCS, at the invitation of the EC’s DG (Environment), to act as an expert in a series of climate service-related workshops. These include the Belmont Forum (Goa India, November 2013), H2020 Challenge 5 stakeholder meeting (Brussels, April 2013) and ‘Towards a European market of climate services’ (Brussels March 2014).
4.3.2 Project governance
EUPORIAS has been built on the concept of co-design of climate services between users and providers. At times this meant stretching the limit of what was possible in the context of an EU project and we are very grateful to the EC’s project officers for their support in what we were trying to achieve. The dynamic allocation of the prototypes gave the end users an opportunity to influence and define the scope and structure of what we were trying to do. However, the project was perhaps less user-driven than desired (although not necessarily less than originally envisaged) and this provided an important learning opportunity for climate services in general. Despite the presence of a stakeholder board, the governance of the project was still skewed in favour of climate providers and users did not have sufficient leverage to alter the course of the project towards something they could immediately benefit from, perhaps an inevitable outcome given the contractual need for the project to reach its agreed milestones and deliver its agreed deliverables. This is understandable in the context of a research and demonstration project, but a different style of project governance should be considered when developing operational services to meet user requirements.
The results of EUPORIAS suggest that co-design works well when all parties involved have a clear interest in the developing of the service but this interest should go beyond the financial benefits that partners may gain from being part of the project in the first place. A good measure of success would be the emergence of joint ventures whereby prospective users are prepared to invest and risk their own money on the success of an emerging service.
4.3.3 WFP LEAP Cost Benefit Analysis
The LEAP Cost Benefit Analysis (CBA) provided a quantifiable estimate of the benefits of introducing seasonal forecast capability within Ethiopia’s national early warning system for food security. The analysis was carried out by comparing three scenarios, using data from 2003-2010:
• Late humanitarian response via needs assessments, using historic data on the numbers in need;
• LEAP Current: Early humanitarian response based on current LEAP predictions at the end of the crop season (August/September);
• LEAP Forecast: Early humanitarian response based on LEAP Forecast, with numbers in need forecast four months before the first failed rains (February/March).
Based on the cost of food/cash alone, the LEAP forecasts result in a higher total cost over the eight years of analysis as compared with a late response based on needs assessments. This is because costs are allocated to the full number of beneficiaries when LEAP over-predicts the number in need (as compared with needs assessments) and is topped up to needs assessment figures whenever it under-predicts. Therefore, responding based on LEAP forecasts will always appear more expensive based on a cost analysis alone. However, LEAP forecasts can be a critical component of raising funds in time to facilitate an early response. When this response is funded and triggered early, benefits arise as households avoid negative coping strategies, engage in greater investment, and avoid long term impacts to household growth, nutrition and educational outcomes.
Early action can cover a wide range of activities, from agriculture conservation practices to prevent loss of crops, to livestock initiatives that prevent animal death, as well as temporary water measures and preventative nutrition interventions that are initiated at the first signs of an impending drought. LEAP is not currently structured to trigger early action measures as a result of forecasts; and yet this is part of the design of LEAP and likely to be the area of greatest benefit – benefits that will more than outweigh the additional costs of LEAP.
The cost of response over the eight years is summarised in Table 1. The investment in early action through LEAP Forecast would save $125m over the cost of responding at the time of needs assessments. As a result, the CBA outcomes so far point out the considerable quantifiable benefits of using LEAP seasonal forecasts, as well as additional benefits that we are currently only able to estimate, which would have strong positive repercussions on how humanitarian crises are managed, and represent a considerable step towards increased disaster risk reduction and prevention of the worst impacts of hunger and malnutrition on vulnerable communities in Ethiopia.
These results should persuade LEAP stakeholders, including the Ethiopian Government and international donors, to introduce seasonal forecasting to improve current early response initiative. Unfortunately, while it is part of LEAP’s vision, LEAP is not currently established to trigger contingency planning and early action based on forecasts. And yet this is a significant component of the benefits that can be realised. Therefore, it is recommended that key stakeholders for LEAP from the Ethiopian Government and other humanitarian and development organisations discuss and lay out a plan on how LEAP can most effectively be used to trigger early action, appropriate measures, and estimated costs and benefits, to incorporate into a revised version of this analysis, and eventually make the seasonal forecasting capabilities of LEAP operational.
4.4 Business Impact
4.4.1 Project Ukko
Project Ukko presents an entirely new way to look at seasonal wind predictions. Recognising opportunities in the emerging climate services market, and the potential utility of seasonal wind speed forecasts for the energy sector, EUPORIAS and the RESILIENCE prototype developed an online visualisation platform http://project-ukko.net/ to inform assessments of wind energy production for the season ahead.
The project involved an interdisciplinary team, composed of design researchers, climate scientists and a data designer, working to integrate design and visualisation approaches to communicate complex and uncertain data in an engaging and accessible way. The interactive online user interface provides forecast wind information overlaid on a world map, using symbols and different visualisation techniques to summarise forecast parameters. These include the predicted wind speeds (given by line thickness), prediction skill (given by line opacity), and the most probable trend taken from an ensemble of forecasts (given by line colour and rotation). The interactive map includes the locations of major wind farms, drawing attention to those areas of interest where the forecast shows high probabilities of significant changes in wind speed. Selecting a location enables the user to obtain detailed information about the past observations and the future predictions.
One of the greatest challenges was dealing with the large volume of data required to support the interface. The seasonal forecast data is provided by a large number of model simulations taken from ECMWF’s System 4 seasonal prediction system. To summarise and communicate the seasonal forecast information, integrated approaches were required to ensure probabilistic information is communicated in a scientifically credible way, while ensuring accessibility to non-scientists.
In the development of Project Ukko, end-users were engaged throughout using various approaches to help inform and direct the climate service design and visualisation process, strengthening the relevance and usability of the service. Ultimately, the visualisation approach helps in closing the gap between the users and the scientists.
Lessons to be learned include adopting a common language and clear context for collaboration across disciplines, having a sustained dialogue with end-users and other actors using conventional and novel user-engagement methods, compromising between scientific soundness, functionality and aesthetics, and having a well-structured dissemination and engagement strategy, executed through different communication and discourse channels.
4.4.2 Weather Roulette
A “weather roulette” tool has been developed in EUPORIAS to help decision makers understand the potential economic value of weather and climate forecast information. The tool provides information about the skill of forecasts and their value as a basis for guiding risk management decisions. It has been demonstrated for use in the wind energy sector using seasonal wind speed forecast information from the RESILIENCE prototype forecasting system.
In the case explored, the weather roulette tool works by using slots to represent three tercile categories - normal, below normal and above normal seasonal wind speeds. Using past observations, threshold wind speeds for each of these categories are selected. A climatology “forecast” then assumes equal probabilities for each category – i.e. a 1/3 chance of occurring – and this is compared to adjusted probabilities taken from the RESILIENCE forecast system (based on calibrated forecasts from ECMWF’s System 4 seasonal prediction system).
The tool was tested for 37 locations around the world where wind farms have been installed. Using different statistical skills scores, the tool demonstrates where using seasonal forecast information can increase the value of investments for individual wind farms and for portfolios of wind farms. Results were mixed, showing value in some locations and portfolios but not for others.
To further demonstrate and test the weather roulette tool with stakeholders, a Smartphone app has been designed. This enables people to experience making investment decisions using the RESILIENCE forecast system and climatology for different locations across the world. The user reinvests an initial capital of 10€ over a single year, or a succession of multiple years, to see if investments are more or less profitable when made using seasonal forecasts or climatology.
The weather roulette tool has proved to be an engaging way of explaining the concept of skill and the potential value of probabilistic predictions. Interaction with stakeholders to assess the tool has explored how such a methodology could impact on the understanding and uptake of seasonal forecast information. Whilst there is a general agreement that the lack of predictability precludes the use of seasonal wind predictions at present, users expressed their willingness to see more results from the weather roulette approach on the performance of real-world wind farm portfolios.
4.4.3 Clinton Devon Estates mobile app
The Met Office, University of Leeds, Predictia and KNMI developed the Land Management Tool (LMTool) in close collaboration with Clinton Devon Estates and the National Farmers Union in the UK. LMTool is a prototype climate service providing seasonal climate forecasts to support land management-related decision making for Southwest UK. This service focuses on the winter months since recent advances in the prediction of the North Atlantic Oscillation (NAO) allows for better seasonal forecasts of the Northern Europe winter climate.
The LMTool was iteratively developed between January 2014 and May 2016, building strongly on a range of stakeholder engagement activities (workshops, interviews, surveys and feedback gathering) carried out with farmers and land managers. During the first winter (2014/2015), the project worked closely with a small, representative subset of farmers to blueprint the prototype service, providing three month outlooks of temperature and precipitation in hardcopy and email. Insights gained during the first winter were then taken forward, alongside engaging a wider farmer group, in developing forecast products for the following winter (2015/2016): three month outlooks of temperature and rainfall for the whole UK, and also 14-day forecasts of rain, temperature and winds for a set of weather stations across South West UK. These products were delivered via an interactive password-protected website (which forms part of a more general micro site (http://lmtool.euporias.eu/) and a mobile app. These platforms have been found to be very useful to carry the prototype to the public.
4.4.4 Benefits and impact of the business analysis of climate services
A five step climate service market analysis methodology developed analyses of industry, demand, supply, feasibility and business models. This approach considers both an overview of the market, and assessment of newly developed services (i.e. the prototypes) in terms of economic feasibility. The analysis showed that all of the sectors analysed are sensitive to weather and climate conditions. The climate services industry in Europe is considered an immature emerging market, especially for the seasonal to decadal timescale, and for the sectors assessed. There is potential demand for these services which is currently unmet. All sectors identify the same gaps in current climate services provision, e.g. lack of forecast skill and insufficient accuracy. Possible actions are suggested to overcome these gaps. All of the sectors analysed think that the Copernicus Climate Change Service will have a positive influence on the development of the climate services industry. The feasibility study conducted for the key sectors shows significant socio-economic benefits, for example, potential cost savings of ~£304 million/year using the 1-month temperature forecasts for the transport sector prototype, and potential cost savings of ~€13,500 from the maintenance of a wind farm of 10.5 MW turbines scheduled for a period of five days that has good wind conditions for producing electricity.
The benefits of the climate services business analysis work include the provision of a well defined holistic methodology for assessing the EUPORIAS prototypes, which can be used for assessment of any climate service. Also the analysis provided a comprehensive market assessment of climate services in key sectors that can be referred to when developing legislations and policy, and justifying the economic viability of future climate service developments, as well as the potential market gaps and risks in service development.
List of Websites:
www.euporias.eu
EUPORIAS Coordinator, Dr Chris Hewitt (chris.hewitt@metoffice.gov.uk)
EUPORIAS Science Coordinator, Dr Carlo Buontempo (carlo.buontempo@metoffice.gov.uk carlo.buontempo@ecmwf.int)
EUPORIAS Project Manager, Paula Newton (paula.newton@metoffice.gov.uk)
Advances in scientific understanding and the ability to forecast climate variability and climate change mean that skilful predictions are routinely made on seasonal to multiannual timescales. Such forecasts can be of great value to a wide range of decision-making, where outcomes are influenced by climatic variations. EUPORIAS (EUropean Provision Of Regional Impact Assessment on a Seasonal-to-decadal timescale) set out to develop tools and services to exploit these emerging prediction capabilities, to extract useful and useable information tailored to specific sectoral needs, and to engage with potential users to help decision-makers make better informed decisions.
The consortium of 24 organisations represented world-class European climate research and climate service centres; expertise in impact assessments and seasonal predictions; two United Nations agencies; specialists in new media; and commercial companies in climate-vulnerable sectors such as energy, water, and tourism. The consortium also developed strong links with end users external to the project consortium, many of whom were subsequently directly involved in the project’s activities.
The vision of EUPORIAS was to develop climate services, operating on seasonal-to-multiannual timescales, and demonstrate their value in informing decision-making, thereby helping stimulate a market for climate services and improve the resilience of society to climate variability and change.
The project successfully delivered its top-level objectives, and some key highlights are as follows:
1. Delivered six prototype climate services, developed through close interaction with end users, for key sectors (wind energy, transport, agriculture, food security and water management), pulling the project results through to demonstrated societal benefit.
2. Produced an extensive survey and analysis of the landscape of users of climate predictions in Europe. Previous analyses focussed on longer timescale climate change projections, and little was known about who was using such shorter-term information in Europe and how they were using it.
3. Working closely with the Framework Programme 7 ‘Seasonal to decadal climate Prediction for the improvement of European Climate Services’ (SPECS) project, developed and delivered standard tools for calibrating, downscaling, bias correcting and visualising the skill of climate predictions. A series of R software packages was developed and made freely available, helping make complex operations easier.
4. Researched methods to transform seasonal forecast data into user relevant information for informing decisions. The methods included transforming raw variables into user relevant indices through statistical post-processing, and using impact models and regional climate models.
5. Forged engagement between various actors, particularly providers and users of climate services (a key concept of the Global Framework for Climate Services), to improve communication, mutual understanding, and co-development of prototypes, often using social science techniques.
EUPORIAS leaves several legacies including published research findings, the prototypes (some being developed further), and sustained engagement between climate service developers and users. Some of these legacies will benefit other projects under Horizon 2020 and Copernicus, some will be maintained through the Climateurope Coordination and Support Action’s network activities, while others will lead to new or stronger engagement between users and providers of climate services.
Project Context and Objectives:
The aim of the EUPORIAS project has been to develop new scientific capability, and prototype climate services through close engagement between developers and users, to help decision-makers in climate-sensitive sectors on seasonal to decadal timescales.
Societies have always faced challenges and opportunities arising from variations in climate, and have often flourished or collapsed depending on their ability to adapt to such changes. In today’s society we are seeing a growing need to improve our resilience to climate-related hazards and better manage the risks and opportunities from climate variability and climate change. This situation was recognised by governments, scientists and decision-makers at the World Climate Conference-3 in 2009, subsequently leading to the creation of the Global Framework for Climate Services (GFCS) under the leadership of several United Nations agencies.
Advances in scientific understanding and the ability to forecast climate variability and climate change mean that skilful predictions are now being routinely made on seasonal to decadal timescales. Such forecasts have the potential to be of great value to a wide range of decision-making, where outcomes are strongly influenced by variations in the climate. A clear opportunity therefore exists to develop new and improved methodologies that exploit the emerging prediction capabilities in climate science and, more importantly, to engage with potential users of such predictions in developing tools to extract useful and useable information tailored to specific sectoral needs.
To this end, the European Commission commissioned the four-year-long EUPORIAS project to develop prototype end-to-end climate impact prediction services operating on a seasonal to decadal timescale and to assess their value in informed decision-making.
The project began in November 2012, coordinated by the Met Office, with a consortium of 24 organisations representing world-class European climate research and climate service centres; expertise in impact assessments and seasonal predictions; two United Nations agencies; specialists in new media (such as Twitter, Facebook, smartphones, and YouTube); and commercial companies in climate-vulnerable sectors such as energy, water, and tourism. Additionally, the consortium developed strong links with end user organisations external to the project consortium, many of which were subsequently directly involved in the project’s activities.
The concept (or vision) behind EUPORIAS was that developing end-to-end climate impact prediction services, operating on seasonal-to-decadal timescales, and demonstrating their value in informing decision-making, would help stimulate a market for climate services and improve the resilience of society to climate variability and change. More specifically, through the development of prototype climate services, each with a very specific end user, EUPORIAS has developed a suite of climate services for key sectors (wind energy, transport, agriculture, food security and water management), pulling the scientific methodologies, tools, and techniques through to demonstrated societal benefit.
Over the past four years, EUPORIAS has largely delivered this vision. EUPORIAS’ research findings, development of the prototypes, outputs (such as project results and papers, workshops, meetings), and close engagement with a range of stakeholders and end users has certainly had an impact on national, European and global discussions and development of climate services. EUPORIAS has helped promote the emergence of a market for those services.
EUPORIAS had six top-level objectives, described below along with a brief summary of progress:
1. Develop and deliver reliable and trusted impact prediction systems for a number of carefully selected case studies. These will provide working examples of end-to-end climate-to-impacts-to-decision-making services operating on seasonal-to-decadal (S2D) timescales.
EUPORIAS delivered six prototypes of climate services operating on a seasonal time scale. The services were defined and developed as much as possible through close interaction with the target users. The prototypes addressed sectors as different as food security in Africa and transport in the British Isles and also looked at agriculture, renewable energy and water resources. They provide working example of how to transform raw climate prediction data, mostly coming from the outputs of European seasonal predictions models, into decision relevant advice for specific economic sectors. Such a transformation went beyond the post-processing of the data (such as bias correction, downscaling and calibration) and included workshop design, training material and mobile phone applications.
2. Assess and document key knowledge gaps and vulnerabilities of important sectors (e.g. water, energy, health, transport, agriculture, tourism) along with the needs of specific users within these sectors, through close collaboration with project stakeholders.
EUPORIAS conducted one of the first and most extensive analyses of the landscape of users of climate predictions in Europe. Although other analyses had been conducted before, they tended to focus on the longer climate change projection time-scale, and prior to the project relatively little was known about who was using shorter-term information in Europe and how they were using it. An online questionnaire and 80 in-depth interviews provided an initial assessment of the main perceived gaps in the provision of climate information from key sectors such as water and energy. The analysis also revealed the striking disparity that exists among sectors in the ability to deal with and digest climate data. EUPORIAS also developed a vulnerability framework specifically designed for climate prediction time-scales.
3. Develop a set of standard tools tailored to the needs of stakeholders for calibrating, downscaling, and modelling sector-specific impacts on S2D timescales.
EUPORIAS worked closely with its sister project SPECS to develop and deliver a set of standard tools for calibrating, downscaling, bias-correcting and visualising the skill of the climate predictions. One of the challenges the project participants faced in developing user-relevant products was the access to high quality information in a standard format. The European Climate Observations, Modelling and Services (ECOMS) User Data Gateway developed and hosted by the University of Cantabria played a major role in alleviating the problem as it gave all project partners access to observations, and predictions through a standardised interface. A series of software packages developed in R was made publicly and freely available and made complex operations possible in a few simple steps.
EUPORIAS built upon some of the technical developments undertaken in the SPECS project. It looked at ways to represent verification attributes, such as ranked probability skill scores or reliability, in a graphical format. From simple site-specific validation plots to visually compelling visualisation such as Project Ukko (http://project-ukko.net/) EUPORIAS transformed complex data streams and mathematical concepts into information that users can act upon.
4. Develop techniques to map the meteorological variables from the prediction systems provided by the World Meteorological Organization (WMO) Global Producing Centres (two of which [Met Office and Météo-France] are partners in the project) into variables which, are directly relevant to the needs of specific stakeholders.
Central to the success of the project was the ability to transform the data coming from the seasonal prediction systems into user relevant information that can be used to inform decisions. To achieve this, EUPORIAS worked along two distinct and complementary lines. One was to transform the raw variables emerging from the forecasting systems into user relevant indices through statistical post-processing. The other was to employ impact models as well as regional climate models to increase the user relevance of the raw variables.
The aim was that whilst increasing user relevance, the derived variables could in some cases increase the quality (as defined by the verification attributes) of the variables being forecast. The results of the project indicate that in some cases, for example for some hydrological variables, this is indeed the case. However, in the majority of the circumstances the post-processing doesn’t significantly alter the verification characteristics of the underpinning variables. The exceptions appear to be related to either cases in which the extra predictability comes from storage terms (e.g. lake and snow accumulation) that are not explicitly included in the climate prediction forcing, or to indices that are calculated through a non-linear combination of a number of raw model outputs (e.g. Fire Weather Indices).
5. Develop a knowledge-sharing protocol necessary to promote the use of these technologies. This will include making uncertain information fit into the decision support systems used by stakeholders to take decisions on the S2D horizon. This objective will place Europe at the forefront of the implementation of the GFCS, through the GFCS's ambitions to develop climate services research, a climate services information system and a user interface platform.
One of the challenges faced in the development of climate services is the interface between climate information and decision-making. EUPORIAS looked at this challenge in a variety of ways. Effort went into identifying the common cognitive biases that can affect the perception of uncertain information, including social science techniques. This led to the development of a series of communication guidelines such as incremental disclosure of information. Effort was also put into developing novel approaches for user engagement. In some cases the project drew on ideas and practices from other fields and disciplines, such as the placebo effect. In other cases the project built upon concepts that had been suggested in the past, such as climate roulette, but pushed the concepts further to make quantitative economic evaluations.
What became clear early on in the project was that a lot of the value of the services being developed was not so much in the services themselves but rather in the continuous interaction with the users and the user support this implied. This allowed EUPORIAS to give substance to some of the concepts proposed by the GFCS, in particular the GFCS Climate User Interface Platform (CUIP). Contrary to the original expectations of most of the project participants, the CUIP EUPORIAS developed is not an online portal where users can get access to a climate data stream relevant to their business or decision, but a useful and active collection of user-provider engagement events. EUPORIAS re-defined the user interface as more than a collection of general decision support systems. It instead designed a system able to support a continuous dialogue between users and providers of climate information. To use the words of one of our external international advisors, Professor Roger Stone, it created a ‘discussion support system’.
6. Assess and document the current marketability of climate services in Europe and demonstrate how climate services on S2D time horizons can be made useful to end users.
To develop a market for climate services in Europe it is important to consider the size of the market and identify the major barriers preventing its growth. Whilst an in-depth analysis of the climate service market was out of the scope of the project (but is the scope of other European Commission-funded projects) EUPORIAS worked on a generic methodology that could be used to assess the marketability of services. This was done by analysing the water sector in some detail, and the methodology could be easily generalised and applied to other sectors. The results the project generated could be used as a basis for the new Horizon 2020 projects that will explicitly look into the issue of the climate service market.
Project Results:
Key areas of progress relating to the project objectives have been summarised in section 2. Further information is now given on the main areas of progress and noteworthy results from each work package.
3.1 WP11 – Assess sector-specific vulnerability
The objectives of WP11 were:
• To raise awareness of seasonal and decadal predictions and their limitations through a close interaction with a number of stakeholders;
• To identify critical sector-specific vulnerabilities operating on seasonal to decadal time scales;
• To contribute to a prototype component of the CUIP within the GFCS in terms of targeted vulnerability information and tailored data-flow in relevant European economic sectors for seasonal to decadal time scales.
The main focus of WP11 has been the identification and analysis of sector-specific vulnerabilities based on an effective dialogue with stakeholders. This dialogue started early in the process and was sustained throughout the project. A theoretical framework for consistent identification of the relevant vulnerabilities for the project’s prototypes was developed. The main outcomes of this analysis have been fed into the EUPORIAS contribution to the CUIP through a collection of different microsites and a series of sectoral workshops. The main findings of WP11 were as follows:
a) Outcomes from European Stakeholder Climate Services Conference:
• In January 2013 the first EUPORIAS Stakeholder Meeting took place in Rome, attended by 43 people from across Europe representing ten different sectors (e.g. water, health, transport, energy). The main objective of the workshop, in line with all EUPORIAS activities, was to encourage communication between the producers and users of S2D information.
• The interaction and dialogue with the stakeholders made it possible to gain an understanding of sector-specific information regarding: how stakeholders currently use S2D climate predictions; critical/relevant choices in their business that could be affected by climate; how climate influences their business choices; and how climate information enters into the decision-making procedures.
• This information was gathered through questionnaires issued prior to the meeting, the Stakeholder Meeting itself, and subsequent discussions and workshops resulting from the growing relationships between stakeholders and EUPORIAS project partners.
• Conclusions emerging from the workshop included: temperature and precipitation are the most relevant climate parameters requested by the EUPORIAS stakeholders. This is especially so in the water, energy, health and agriculture sectors. The most valuable parameters for the surface-transport sector are ground temperature (influenced by air temperature, wind, and soil moisture) and the number of marginal nights (zero-crossing). Important parameters for the insurance sector include: the number of land-falling tropical storms, extreme precipitation, river runoff over specific thresholds, insurance-specific drought indices, weather profile over the year (including lack of snow and late frosts), general “crop failure indices” with a focus on drought lengths or dry spells.
• The business-critical decisions that can be informed by weather tend to cluster in spring (for the summer outlook) and autumn (for the winter outlook). The exceptions are the agricultural sector, which would benefit from seasonal predictions throughout the year, and the insurance sector, for which the beginning of January and the beginning of April are crucial dates.
• While downscaling is seen as important for most stakeholders, they appeared to be keen to prioritise resource-investment in improving the predictability of the large-scale drivers rather than increasing the granularity of the data.
• While climate predictions (seasonal and decadal) are an interesting and potentially useful area for the stakeholders, and while many sectors use them, there is still a huge need for education and training. Direct access to expertise, for instance via sector-specific workshops or seminars, is seen as a vital way of providing this.
• Despite the fact that a significant fraction of the audience was aware of climate predictions, and whilst some of the participants were using these predictions, there was a clear language barrier on a series of crucial definitions. The primary example of this was around the communication of risk and uncertainty. A number of stakeholders stated that they would not have used the predictive information unless its level of confidence (no definition provided) exceeded 95%.
• The Stakeholder Meeting was largely a success and it certainly helped to inform a number of other activities within the project, but in hindsight it might have been better to have a larger number of more sector-specific meetings. Subsequently, EUPORIAS tried to address this challenge by, whenever feasible, supporting sector-specific workshops, which could have helped the interaction between users and providers of climate information.
b) Developing a method to assess sector-specific vulnerabilities:
• A common methodology to assess sector-specific vulnerabilities has been developed and the resulting Vulnerability Assessment Framework (VAF) was applied to each sector of interest. This characterised sector-specific vulnerabilities in such a way as to provide valuable information for decision-makers with respect to the use of climate predictions.
• Challenges for the vulnerability assessment of economic sectors to S2D climate events originate from the inherent role of climate for different economic sectors. Climate may be viewed by these sectors as a hazard-, resource-, production- or regulation-factor.
• One of the major findings of this study is the realisation that the Decision-Making Process (DMP) is a very sensitive factor, which significantly influences: the determination of threshold, which define critical situations; the characteristics of buffer functions and thus the temporal scale of critical climate conditions; and the specification of climate information needs and thus their potential usability (value) for decision-making. Decision-making processes should be considered in detail when assessing sector-specific vulnerabilities in the context of climate information.
• Another major finding is the classification system of ‘climate-impact types’ that classifies sector-specific problems in a systemic way. This system proves to be a promising concept because: it reflects and differentiates the cause for the climate relevance of a specific problem (compositions of buffer factors); it integrates DMPs, which proved to be a significant factor; it indicates a potential usability of S2D climate service products and thus integrates coping options; and its systemic approach goes beyond the established ‘snap-shot’ of vulnerability assessments.
• The EUPORIAS VAF has been applied to all the prototypes and case-studies developed in the context of the EUPORIAS project prototype and inserted into the EUPORIAS microsites.
c) Contribution to GFCS Climate User Interface Platform
• One of the most important (but least developed) components of the GFCS is the Climate User Interface Platform (CUIP). The understanding of what a CUIP should be has changed significantly over the years and it has become evident that an online tool that is not supported by close interaction with the users tends to be of limited use. For this reason, EUPORIAS developed a CUIP based on a revised scope and consistent of two distinct parts.
• On the one hand there are the microsites, a collection of user specific web pages describing the climate service prototypes, and four of the case studies, including information on their associated sector-specific vulnerabilities.
• However, the CUIP should not be just a simple ‘portal’, but rather a place where the different communities can share lessons learnt materials, and know-how, together with the main outcomes of the EUPORIAS project. Face to face meetings and ongoing dialogue are often required. It has been recognised that person-to-person meetings help build trust, legitimacy and saliency into the service being developed. A series of workshops was developed with the users, who have played a pivotal role in scoping the prototypes, their functionalities and interfaces. These events have helped the developers to go further than they initially planned, and also helped the users understand why some of their needs cannot be fulfilled.
3.2 WP12 – Assessment of users’ needs
The objective of WP 12 was to assess the needs of users from across European society for seasonal-to-decadal climate predictions.
The WP assessed the needs and requirements of users regarding seasonal-to-decadal climate predictions (S2DCP) across Europe. The tasks resulted in the following successful outcomes: a systematic literature review of the use of S2DCP in Europe was produced; a workshop with National Meteorological and Hydrological Services (NMHSs) and other stakeholders was conducted; around 80 interviews with key European organisations working across the sectors identified in EUPORIAS were conducted; a multi-lingual online survey of users’ needs was undertaken with a total of 489 responses received; a workshop with S2DCP developers was conducted.
The main findings of the WP were as follows:
• The majority of the organisations involved in this study were aware, to differing degrees, of seasonal climate forecasts (SCF), whilst decadal predictions were less well-known.
• The use of SCF across Europe to inform decision-making is still an emerging venture, mainly due to the lack of skill and reliability (‘reliability’ is used here as a synonym of trustworthiness and can be mapped onto a number of other technical concepts such as skill, reliability, and sharpness) over Europe. Decadal predictions are regarded as unchartered territory and thus not used.
• Some sectors were identified as using SCF, such as the energy, water, transport, and insurance sectors. However, the use of SCF is limited and only used in a qualitative manner i.e. the forecasts are not used operationally or in a systematic way (Bruno Soares and Dessai, 2016).
• European Centre for Medium-Range Weather Forecasts (ECMWF) and NMHSs are regarded as the main providers of climate information in Europe including SCF.
• Most organisations involved are sensitive to weather and climate conditions, although some are more concerned with the impacts of extreme weather (e.g. drought, floods), whilst others are more interested in weather variables such as temperature and precipitation (see D12.3).
• Many organisations already use weather/climate information (mainly historical data/past observations and weather forecasts). Such information tends to be used to either develop/feed operational models, forecast seasonal variability based on past data, and/or understand future weather conditions, all of which are used to inform decision-making processes within organisations. Many perform some kind of post-processing in-house.
• The main perceived barriers to the use of SCF are mainly related to low reliability, but also related to other factors such as usability of the information provided and accessibility to such information by the end-users (Bruno Soares and Dessai, 2016).
• The main enabler to the use of SCF in those organisations was related to the interactions with the producers of SCF (e.g. National Met. Services). Other non-technical aspects such as the relevance of SCF to the organisation’s activities and operations and the existing resources and capacity to assimilate and act upon that information also acted as enabling conditions to the use of SCF (Bruno Soares and Dessai, 2016).
• Difficulties in integrating SCF into existing operational models and structures, existing traditions of performing historical analyses, and not needing this type of information were also pointed out as reasons for not using SCF.
• Understanding decision-making processes within organisations and how SCF is or can potentially be used can be difficult at times, as responses provided by respondents are influenced not only by the size (i.e. multiplicity of activities) and nature (i.e. end-user versus intermediate organisations) of the organisation itself, but also by the role of the interviewee in it (e.g. modeller, head of department).
• There is a general understanding that information on the uncertainty of the information provided is a fundamental component of S2DCP. Many organisations would prefer information on the uncertainty of forecasts to be provided using a deterministic approach. The preferred method for representing uncertainty is numerical (e.g. one figure, percentages) as this would facilitate the quantification and integration of uncertainty into models.
• The final workshop with the S2DCP developers also provided a range of interesting findings although these were specific to the case studies explored during the workshop. One of the main findings was that for all of the cases of users’ needs discussed at the workshop there were no products available that could fully satisfy their needs (e.g. spatial or temporal resolution required, limited skill). As a result, a better understanding of the types and chains of decision-making by the users and how this data is/would be utilised to inform such processes would also help to refine the quest for (and potentially the provision of) data that is able to satisfy those particular needs.
3.3 WP21 – Calibration and downscaling
The objectives of WP21 were to develop and apply a set of bias-correction and downscaling methods for use with seasonal to decadal forecasts; to downscale standard climate variables and advanced climate indices for use in the EUPORIAS case studies; to assess the uncertainty associated with the downscaling methods in collaboration with WP33; and to make available, through the EUPORIAS web portal, downscaled data, along with a number of calibrated methodologies.
Seasonal forecasts are produced from global prediction systems at typical resolution of 100-200km while climate service users often require information at significantly finer spatial resolution. Furthermore, forecasts contain sizeable errors when compared to observations and such biases need to be corrected for successful use of forecast data in impact studies. Several statistical downscaling and bias correction methods have been applied to global seasonal forecasts over Europe to provide inputs for impact models used in several case studies and climate service prototypes. Both approaches show similar performance in removing biases and preserving predictability skill.
In Eastern Africa a number of regional climate models and statistical downscaling methods were used to downscale a global forecast to focus on Ethiopia in June-September. Verification results revealed that both dynamical and statistical downscaling are able, in general, to capture and reproduce the predictive skill evident in the global seasonal hindcast it is derived from. However, on average, the dynamically and statistically downscaled hindcasts show no added value in Eastern Africa if we define the added value as a higher predictive skill in the downscaled hindcast. Similar results were found for the World Food Programme (WFP) Livelihoods, Early Assessment and Protection (LEAP) platform (drought early-warning system) for Ethiopia: the performance of LEAP in predicting humanitarian needs at the national level is not improved by using downscaled seasonal forecasts. These results indicate that, in the EUPORIAS case studies over Europe and Eastern Africa, downscaling only reproduces the forecast skill of the global forecasts it’s derived from but does not improve it.
The main findings of the WP were as follows:
Task 21.1: Statistical downscaling and bias correction of General Circulation Model (GCM) forecasts over Europe
• The ECMWF system 4 (S4) forecast was bias corrected using different approaches with focus on i) pan-European hydrological modelling for flood risk estimations, ii) supporting logistical decisions for inland waterway transport on the river Rhine and the electricity demand case study over Southern Italy, and the Land Management Tool prototype over Southwest UK.
• All bias correction methods applied are efficient in removing seasonal mean S4 biases, even if some biases in precipitation are still evident during test periods not used in their calibration. In general, bias correction methods preserve the predictive skill of the original S4 hindcast, but they do not improve it.
• Perfect prognosis statistical downscaling has also been applied over different regions in Europe to support several case studies and the climate service prototypes.
• The predictive skill of the perfect prognosis downscaled forecasts over different European regions, in general, is not improved. Only the analog method has shown a slight increase in reliability, as found in different analyses over France and Italy. However, this may be related to a limited forecast resolution. Some improvement of the forecast skill for CII was found, but it is not significant in all cases.
Task 21.2: Combined statistical and dynamical downscaling over eastern Africa
• The utility of dynamical and statistical downscaling to provide seasonal forecast data for impact models over eastern Africa was assessed, in particular for the Drought Early-Warning System LEAP. After consultations with the WFP, it was decided to focus on the Kiremt rainy season (June-September) in Ethiopia using seasonal hindcast initialised in May, which can be used as input to the LEAP system.
• The downscaled ensemble created consists of five regional climate models – (RCMs) (DWD-CCLM4-8-21, SMHI-RCA4, ENEA-RegCM-4-3, UCAN-WRF341G and UL-IDL-WRF360D) at about 25 km resolution and two empirical-statistical downscaling (ESD) methods (the analogue technique - AN1 and a generalised linear model - GLM) at about 50km resolution (defined by observations). Analysis of eight observational datasets showed that there are large uncertainties in interannual variability of precipitation among these datasets (which may have a considerable impact for the case of ESD methods) in eastern Africa.
• Using a number of deterministic and probabilistic verification metrics, it was found that there are two distinct regions where some predictive skill is evident in the driving EC-EARTH seasonal hindcast: northern Ethiopia - North-East Sudan and southern Sudan - northern Uganda while there is no skill elsewhere (Fig. 1).
• Focusing on Ethiopia, verification results revealed that the RCM and ESD methods are able to capture and reproduce the signal evident in EC-EARTH over northern Ethiopia in June-September showing about the same performance as their driving GCM (Fig. 1). However, on average, the RCM and ESD hindcasts show no higher skill in predicting future seasonal anomalies.
• Two RCM hindcasts (DWD-CCLM4-8-21 and SMHI-RCA4) were used to assess the utility of dynamically downscaled seasonal hindcast data as an input to the LEAP platform. Consistent with the results found for the RCM hindcast, the performance of the LEAP system in predicting humanitarian needs at the national level in Ethiopia is not improved by using downscaled seasonal forecasts.
• The effect of downscaling on the quality of seasonal forecasts of climate information indices in eastern Africa was considered, using a range of indices relating to the water availability (potential evapotranspiration and water balance) and rainfall regimes (wet day frequency and simple daily intensity index). This analysis shows that the downscaling does not improve the forecast skill of global forecasts of climate indices in eastern Africa. Consequently, it is recommended to derive forecasts of indices from the bias corrected global forecasting systems rather than from dynamically downscaled forecasts.
• It should be noted that these conclusions are only for Ethiopia in the June-September season and cannot be generalised for other regions and seasons. For example, some skill improvement in a statistically downscaled forecast was found in the Philippines (see WP32). Additionally, large observational uncertainties can potentially prevent accurate verification of the downscaled high-resolution hindcast in eastern Africa.
3.4 WP22 – Impact-relevant climate information indices (CII)
The aim of this work package was to generate a collection of impact relevant climate information indices (CIIs) and to provide the best estimate of such CIIs for the current and near-term climate.
Based on user needs identified in WP11 and WP12, CIIs relevant for various economic sectors and a large variety of stakeholders were selected (D22.1 D22.2). Seasonal forecast skill of these CIIs has been analysed extensively, including an assessment of downscaled CII forecasts as a contribution to WP21 (D22.3). This work package has also attempted to identify key windows of opportunity where CIIs can provide quality and value from an end-user perspective (D22.4). Forecasts of CIIs evaluated in WP22 were used to complement the prototypes and case studies developed in WP42 and WP43.
In addition, this work package has developed software tools (R packages) to facilitate download, bias correction, index computation and the validation of seasonal forecasts, enabling the further development of CII forecasts within the project and beyond.
The main findings of the WP were as follows:
• Forecast skill in Europe is limited with enhanced skill in summer compared to winter and higher skill for temperature and related indices than for precipitation. Skill varies strongly by season, region, lead time, climate index, and spatio-temporal aggregation.
• Some regions show skill in specific seasons, which may open a window of opportunity for forecasts to be useful in specific cases. Hence, this WP has developed an interactive, publicly available web-platform (https://meteoswiss-climate.shinyapps.io/skill_metrics) that allows users to explore this variability in forecast skill interactively to identify opportunities for skilful and thus useful seasonal forecasts.
• The limited predictability of climatic conditions in Europe poses challenges for the use of seasonal forecast information. Cases from various sectors such as hydrology, viticulture, and energy show that it is very difficult to derive decision-relevant information from seasonal forecasts in Europe due to the lack of skill.
• In general, forecasts of CIIs are, at most, as skilful as forecasts of the seasonal mean of the underlying meteorological variables. As this reduction in forecast quality is moderate for all except indices describing rare events, CII forecasts can often be issued without major loss in skill. The enhanced user relevance of forecasts of CIIs outweighs the slight reduction in forecast skill.
• CII forecast skill is found to be rather insensitive to the choice of bias correction method, but basic calibration is very important. In some cases, however, using more sophisticated calibration methods (e.g. quantile mapping compared to mean de-biasing) improves the skill in CII forecasts.
• We further find that indices defined with respect to percentiles of the forecast and observed distribution respectively (e.g. percentage of time with wind speed above the 90th percentile) are advantageous in that such indices are less prone to systematic model errors than indices defined with respect to absolute physical thresholds (e.g. frost days).
• While the climatic conditions can be difficult to forecast, forecasts of impacts dependent on these conditions can have considerable value (demonstrated by forecasts of humanitarian needs for Ethiopia). Such added value is partly due to the sensitivity of the impact to large-scale long-term climatic conditions (drought) that are more predictable and partly due to the co-location of forecast skill in areas sensitive to drought.
• Enhanced skill is also found in forecasts of river flow based on a hydrological model resulting from the long-term memory of initial conditions for the hydrological model including soil moisture and snow. This highlights the presence of sources of predictability not included in the operational seasonal forecasting system that may be exploited for targeted applications through coupled and initialised impact models.
• It is worth exploring the value of CII forecasts even if there is marginal skill in forecasts of the underlying variables. Implicit aggregation or weighting in the calculation of the impact variable can amplify partial skill in the underlying variables and thus exploit windows or pockets of predictability while transforming meteorological information to more user relevant quantities.
3.5 WP23 – Impact models for impact predictions and WP31 – Quantifying uncertainty in impact models
WP23 and WP31 were intricately linked together in the project, both jointly led by the Met Office and WU, and were delivered by mostly the same set of partners (with the exception of CETAqua who only participated in WP23). The work delivered under WP31is a natural extension of that began under WP23, so a joint report for both work packages has been provided to avoid duplication.
The objectives of WP23 were:
• To further develop complex impact models able to address the users’ needs and inform the case studies and the prototypes:
• To develop a prototype operational workflow to use these models in S2D forecast mode;
• To assess and improve their predictive skill by analysing hindcasts of low- and high-end impact events (hi/lo discharge, crop yields, etc);
• To develop optimal geographical forecasting units, as a function of model physics and stakeholder needs.
The objective of WP31 was then:
• To fully characterise the level of confidence we can associate with specific impact models. Multi-model, multi-driver and perturbed physics parameter ensembles will be used for this purpose.
The main findings of the WPs were as follows:
● WP23 and WP31 have completed a coordinated set of model studies across different sectors (agriculture, water, forestry and transport), over different regions and scales (Europe, East Africa, specific areas and basins), driven by baseline forcing data and different seasonal hindcast datasets. Some of the studies have investigated the role of particular parts of the uncertainty chain (e.g. role of bias correction, initial conditions, seasonal hindcast skill, model parameters) in overall impact skill.
● In general, it appears that while bias correction is important for some applications (e.g. where thresholds are important in dynamic models using crop temperature thresholds, or indices), it does not affect probabilistic skill in all cases.
● There are remaining issues to be dealt with regarding application of bias correction in impact modelling, such as appropriate choice of window sizes for Quantile-Quantile (QQ) mapping, and whether to apply bias correction to model inputs or outputs.
● There is not necessarily a monotonic decay in skill with lead time in all cases, although this is true for some impacts sectors/models.
● Skill varies with variable (e.g. greater skill for temperature than precipitation, and least skill for radiation), region, season, lead time and forcing dataset.
● Skill for impact fluctuates (e.g. terrestrial Gross Primary Production, (GPP)) may be greater than that for state variables (e.g. crop yield).
● Skill for some impacts, particularly river discharge, may be larger than that of its meteorological drivers. In such cases memory effects from proper initialisation of state variables (e.g. soil moisture, snow mass for hydrology) explain the skill.
● A significant part of WP23/31 involved further development of impact models, including those for crops, water, forestry and energy and development of workflows for impact hindcast simulations and analysis.
● For some impacts proper skill assessment is hindered by lack of high quality, spatially distributed observations, most notably so for crop yields (in eastern Africa) and quality assessment of such data may require significant efforts.
Figure 2 outlines the modelling studies performed by WP23 and WP31. A complete scenario set consists of: one (30 yrs) baseline/reference run forced by WFDEI; one hindcast with System 4 forcing, not bias corrected (30 yrs x 12 months x 15 members = 5,400 runs of seven months each) and one identical hindcast with bias corrected forcing. Scenarios are available as a set of self-descriptive NetCDF files in a format agreed between the relevant partners, to facilitate intercomparison of results and allow for the creation of an MME.
In terms of assessing the influence of lead time and other factors on seasonal impact forecast skill, some sector-specific tentative findings are that:
For crops:
• There is reasonable skill for some models (in predicting Above/Normal/Below events), but poor skill for others (Ogutu et al. 2016);
• It is difficult to evaluate skill over East Africa due to poor availability of observations and because other factors (human, socio-economic) may be more important in driving changes in crop yields than weather;
• Strong links between precipitation and yield in observations and models implies that climate impact model initialisation may not be needed, and that models can be more confident where seasonal forecast predictability for precipitation is strong (Williams and Falloon, 2015).
For water:
• There is better skill in river discharge and runoff for North Europe, better skill during winter than in summer, and better skill for high flows versus low flows; skill is dependent on the basin’s hydrologic regime;
• Climate skill is not the only, or dominant driver, of overall skill, due to the importance of other processes and factors (e.g. snow, lakes, wetlands, reservoirs, human factors). For river hydrology in Europe, overall skill appears to be dominated by skill derived from initial conditions of soil moisture and snow mass;
• Skill for river discharge extends to longer timescales (and more downstream) than that for runoff; which is useful for analysing skill with respect to pseudo observations (i.e. hydrological reanalysis data);
• Since summertime drinking water demands correlate with temperature, skilful seasonal water demand forecasts in Spain are feasible. Winter time dam inflow predictions in Spain do not provide additional information over climatology.
For forestry:
• In the Swedish case study regions, the greatest skill was found for bias-corrected temperature during January-March, potentially supporting planning of harvest operations depending on frozen soil, or to support the planning of planting activities depending on soil moisture;
• Fire Weather Index showed considerable skill in especially the Eastern Mediterranean, a skill that, however, cannot be attributed to forecast models reproducing variability in summer North Atlantic Oscillation.
For other sectors
• The future variability of wind energy resources is better addressed with wind speed and temperature transformed into a capacity factor, predictions of which can be integrated more easily in decision-making processes;
• For road transport, predicted winter air temperatures need translation into colder road skin temperatures.
A significant part of the WP23/31 involved further development of impact models, which included:
• Crop models - new GLAM version, setup for Africa; New JULES-crop model and testing (Osborne et al. 2015; Williams et al. 2017); nutrient representation in WOFOST; Regional setup for LPJmL, Wine production model (Fraga et al. 2014);
• Water models - New version of HYPE, re-calibrated (Donnely et al. 2015); development and testing of an impact model for dam management, and testing on the use of predictions for urban water demand prediction (Pouget et al. 2015);
• Forestry models – development of extreme temperature and precipitation metrics using a soil moisture and soil temperature model;
• Transport sector - development of METRo impact model to be used to predict winter road temperatures for the Iberian Peninsula;
• Energy models – development of capacity factor seasonal predictions based on ECMWF System 4 wind speed and temperature.
3.6 WP32 – Uncertainty framework
The main goal of WP32 was to look at the best way of combining the main sources of uncertainty in seasonal impact predictions (i.e. uncertainty from the climate predictions and uncertainty in the impact model formulation) in a single coherent assessment. Early in the project (see e.g. the internal report on S2D uncertainty assessment protocol for impact models, released as milestone M26), it was found that a general framework exploring all sources of uncertainty in a coherent way across application sectors was unfeasible due to their heterogeneity and different sensitivities. Therefore, the analyses and results focused on specific sectors and steps of the climate-impacts uncertainty chain.
The main findings of the WP were as follows.
Task 32.1: Assessment and combination of climate predictions:
• New tools at the interface between climate forecasts and impacts (ECOMS-UDG, visualize https://github.com/SantanderMetGroup/visualizeR) have been developed, enabling providers and end-users of seasonal forecasts to access, analyse and visualise different forecasts in a homogeneous way. This data gateway and tools replace standard file sharing, increasing flexibility and adaptation to users’ needs while reducing the data transfer burden.
• This technical development required the cooperation of different work packages (WP4, WP21, WP22, WP23, WP31, WP32 and WP33) and projects (SPECS, WP52) under the European Climate Observations, Modelling and Services (ECOMS) umbrella. Moreover, link functions make these tools take advantage of other tools developed in ECOMS (easyVerification, SpecsVerification, esd). The tools have been used by many partners, as a convenient way to guarantee the homogeneity of input and presentation of results (e.g. Ogutu et al. 2016).
• The skill of several seasonal forecasting systems (ECMWF System 4, NCEP CFSv2, MetOffice GloSea5) has been assessed, both worldwide and on specific regions (Europe, East Africa and the Philippines). The comprehensive analysis has been made possible by the availability of these forecast systems through the ECOMS-UDG.
• The analysis of ECMWF System 4 seasonal forecast (selected as the initial model to assess the skill of the global models) enables users to identify the most suitable regions for impact applications on seasonal time scales using different observational datasets as reference. Results indicate that the skill of the global seasonal forecasting systems is mainly located in the tropics (especially for precipitation) and no significant skill is found over Europe (Manzanas et al. 2014), where the small regional skill does not agree across different forecasting systems.
• Important season and lead-time-depending biases were found in prototype regions (East Africa regarding crop models, and Europe for hydrology). In this sense, bias-correction and or downscaling methods are required to make key variables, such as precipitation and temperature, suitable for impact applications. However, the performance of perfect prognosis statistical downscaling was found to be limited in those regions exhibiting significant reanalysis uncertainty, such as East Africa or the Philippines (Manzanas et al. 2015).
• Moreover, contrary to simple bias correction techniques (which are limited by the quality of the model), most sophisticated perfect prognosis techniques were found to improve the global model seasonal forecasts when/where the model is more skilful on large-scale predictor variables than on the target variable (Manzanas et al. 2017).
• Climate forecast combinations by sophisticated techniques did not significantly improve simple, equal weight combinations. In an attempt to improve the skill of individual seasonal forecasting models, a Bayesian calibration and combination method (as described by Stephenson et al. 2005) was applied over Europe and the Philippines for different seasonal forecast systems (EUROSIP). The different settings lead to a slight improvement upon the skill of the individual models. However, no clear benefit was observed when applying sophisticated model combination techniques; thus, simple model combination obtained by equal weighting of all contributing systems is recommended in general.
Task 32.2: Uncertainty framework for seasonal impact predictions:
• The uncertainty of impact model predictions was compared with that of their climate model forcing. This is strongly linked with the activities on WP31, which quantified the uncertainty in impact models.
• An uncertainty framework, which enables the user to identify the contribution of each source of uncertainty, was developed for the wine production industry. Results indicated that the uncertainty associated to the seasonal forecast (measured through the error variance, according to Beven, 2012) is reduced as the lead time decreases. These improvements were mostly felt in the lower percentiles of wine production. Although the model variance increased with the reduction of the lead time, indicating an improvement in the forecast, the model variability was always smaller than the observations and the uncertainty was very high.
• An efficient tool to estimate predictive hydrological uncertainty has been deployed, with a graphical interface for end-users. The method assumes that the error in the forecast is a function of the forecast itself and therefore relies on having both hindcasts and observations at the site of interest. A methodology using quantile regression was implemented as a forecast post-processing tool in DHI’s generic operational decision framework. This flexible and efficient uncertainty estimation technique is available to DHI’s software users.
• Four sources of uncertainty occur in deterministic flow modelling; random or systematic errors in the model inputs or boundary condition data, random or systematic errors in the recorded output data, uncertainty due to sub-optimal parameter values and errors due to incomplete or biased model structure. While Bayesian methods are widely used to quantify model parameter uncertainty, a more straightforward Monte Carlo approach can be applied (Butts et al. 2016). This approach was tested on hydrological model parameters applied to the Añarbe reservoir. Model performance was assessed using the absolute error average and variance. Assuming 10% uncertainty in the discharge then all parameter sets within an acceptance interval of 0.36 m3/s of the Pareto front define the parametric uncertainty.
• For certain applications (e.g. hydrology) seasonal impact forecast uncertainty can be smaller than the atmospheric seasonal forecast uncertainty used as input. This is due to slow-varying initial non-atmospheric variables, usually related to initial reservoirs (e.g. mountain snowpack, groundwater). Despite the low skill of seasonal climate forecasts over Europe, some examples of reduction of the uncertainty and slight skill improvement in river flow forecasts were found. These improvements are not significant, however, and a protocol has been agreed with the stakeholders to provide both seasonal climate- and climatology-driven river flow forecasts along with expert advice encouraging stakeholders to use the seasonal prediction when skill is expected.
3.7 WP33 – Communicating levels of confidence
The objective of WP33 was to test the effectiveness of different approaches of communicating the confidence and uncertainty associated with S2D predictions and its impacts and to make recommendations for good practice.
To address this challenge, the work package brought together expertise and insight from climate service providers, end-users, and social scientists. A preliminary user needs survey (Taylor et al. 2015) and review of existing approaches to communicating uncertainty, informed the development of a set of strategies for communicating information about probability and skill in seasonal predictions. Subsets of these strategies were then formally tested with decision-makers from relevant sectors in a series of online decision labs. Finally, key findings from the preceding work were brought together with key lessons from four prototypes in a final report outlining the recommendations for the communication of confidence and uncertainty in S2D predictions (and climate information more broadly) that can be drawn from the EUPORIAS project as a whole.
The main findings of the WP were as follows:
a) Identifying user needs and challenges
• To explore user needs with respect to the communication of confidence and uncertainty, a survey was conducted with current and potential users of S2D.
• Survey findings indicated that: (i) while current S2D users perceived these forecasts to be useful they were not felt to be accessible or understandable; (ii) users tended to prefer familiar formats when receiving information about forecast probabilities; (iii) preference for receiving information about forecast probabilities in different formats was linked to level of statistical expertise (with those with greater expertise having a stronger preference for receiving full confidence intervals); and (iv) a large proportion of current users were not receiving information about how well forecasts performed relative to observations (Figure 3), suggesting that information about forecast quality (indicated by measures of reliability and skill) was either not being communicated to many users, or not being communicated in a way that was well understood.
• Taken together, this highlighted a pressing need to identify more understandable formats for communicating forecasts, especially information related to forecast skill.
b) A review of existing approaches
• In reviewing existing approaches to communicating uncertainty, we examined those formats and visualisations that were currently in use with respect to S2D predictions, and reviewed relevant findings from the broader literature on risk communication and visualisation.
• Through examining existing communication strategies, it was established that while many ways of representing uncertainty in seasonal forecasts existed, relatively little end-user testing had been conducted.
• Our review of literature on the communication of risk and uncertainty identified several factors to consider in devising communication strategies, including: the need to ensure that climate visualisations do not use counterintuitive colour schemes (e.g. red to denote lower temperatures) (e.g. Kaye et al. 2012); that while aversion to uncertainty may lead some users to seek decision aids offering act/don’t act guidelines, it is important to ensure that the presence of uncertainty is not masked by this; the potential for “cognitive biases” to lead to systematic errors in interpreting information (e.g. Kahneman, 2011); and that decision-makers can vary considerably in their ability to use complex statistical and graphical information (e.g. Reyna, 2009).
• This review also identified the ‘progressive disclosure of information’ framework (Kloprogge et al. 2007) in which information is made available with different levels of technical detail, as a potential strategy for addressing the challenge of communicating forecast information to audiences where members differ in statistical/technical expertise.
c) Developing strategies for communicating levels of confidence
• Focussing on the challenge of communicating information about forecast probability and forecast quality (skill) to end users, a set of strategies that aimed to communicate this information to end-users were developed.
• Development of these strategies drew on the findings of the previous two tasks, along with input from work package partners, colleagues from WP32, and an external expert in visualisation.
d) Testing communication strategies
• To empirically test the effectiveness of a subset of the communication strategies developed in the preceding task, online decision labs were conducted with a broad group of decision-makers from climate sensitive sectors (Decision Lab 1), and a group of 98 highly engaged climate information users (Decision Lab 2).
• In the decision labs, participants were presented with forecasts using different formats, and asked questions to ascertain preference, understanding, subjective interpretation, perceived familiarity, perceived usefulness and how they might be used in decision-making.
• Key findings included: 1) preference for specific communication formats is related to familiarity but not objective understanding; 2) forecasts with no skill can have an undue effect on expectations about future conditions; 3) using qualitative categories (e.g. none, low, medium, high) to label skill scores appears to aid understanding amongst those with less experience of using this information.
• This task also enabled us to identify specific aspects of the different communication strategies that were being systematically misunderstood. For instance, white space on the bubble map (Figure 3a) was not always interpreted to mean ‘no useful information’, while it was found that participants presented with the bar graph (Figure 3d) mistook information about skill (Relative Operating Characteristic Skill Score, ROCSS) for information about likelihood.
e) Distilling key lessons and recommendations on the communication of confidence and uncertainty from the EUPORIAS project
In the final phase of WP33 we collated key lessons about the communication of uncertainty in S2D from both this work package and four of the EUPORIAS prototypes. Core lessons and recommendations include:
• No “one size fits all” solution exists for communicating confidence and uncertainty in S2D to end users, who have diverse needs and expertise;
• Where feasible, tailored communication strategies that are co-produced with users work best.
• Where tailoring is not feasible because users groups are too large and diverse, a layered approach to presenting information may address differing user needs;
• Forecast quality metrics such as skill can be difficult to understand. Guidance should be provided as to what these mean (e.g. through descriptions, ordinal categories);
• It should not be assumed that those communication strategies that are most preferred by users will be best understood. It is important to ensure that both requirements are met;
• Even when forecasts have no skill relative to climatology, they may have an undue influence on expectations about future states. We recommend that forecasts without skill not be provided by default unless specifically requested by users who understand that they do not provide ‘added value’;
• It is not always obvious that a particular aspect of a forecast visualisation will be systematically misinterpreted. New communications should always be tested with members of the intended end-user group to ensure that misunderstandings are addressed.
3.8 WP41 – Climate information and decision-making processes
The objectives of WP41 were to assess:
• The climate and non-climate-related options available to decision-makers;
• The value of Climate Information within the decision-making processes (DMP);
• The impact of S2D forecasts developed in RT2 onto DMPs.
The initial aim of the work package of assessing the value of decision-making per se could not be pursued within the EUPORIAS prototypes. This was mainly due to the novelty of SCF for those organisations involved in the prototypes. The lack of experience using SCF and the need for more time to build trust among the users in the different sectors made it difficult for them to identify a specific decision-making to be evaluated. A deeper understanding of users’ decision-making processes requires time and this time is needed to build trust (since details on decision-making constitute sensitive information in many sectors) but also to allow an understanding between scientists and users in novel topics such as SCF.
Given this situation, WP41 focused on the assessment of the value of using SCF in defined decision-making processes, a goal that was tackled from multiple perspectives and methodological approaches presented in a preliminary guidance document (D41.2; Soubeyroux et al. 2016). To fully understand the value of SCF, it was important to account for both the quantitative and the qualitative benefits. To do so, the guidance document presented a bibliographic analysis of different relevant methods for assessing the value and impact of SCF in specific decision-making contexts. A critical aspect found in WP41 was how to effectively distinguish between “value” and “impact” and how to define these terms across the different methodologies. Since addressing both value and impact of the SCF proved to be a challenging objective due to the intrinsic complexities of each prototype in measuring either economic value in the sector of analysis (e.g. benefits and costs associated to multiple water uses of a dam), by providing an indication of the effect that SCF can have on highly political decision-making processes, each methodology focused its attention mainly on one of the two aspects – the value or impact of SCF in the DMP. The detailed results and information from each methodology can be found in D41.3/D41.4 (Bruno Soares et al. 2016).
The main findings of the WP were as follows:
a) Decision maps (Land Management tool prototype)
• The assessment of the impact of the SCF in the farmers’ decisions was pursued via an interactive workshop in January 2016 followed by in-depth interviews in April 2016. Of the six farmers interviewed two farmers changed their normal course of action based on the SCF provided to them.
• The two farmers that used the SCF did it to help them plan their activities and, in one case, also to help manage their expectations with regard to their financial conditions. Although it was not possible to determine an economic value in terms of the decisions taken based on the information provided by the SCF, both farmers claimed to have saved money as changing the decisions based on the SCF prevented them from spending unnecessary money. Both farmers agreed that if the SCF were available they would keep using it as an additional piece of information to their decision-making processes.
• Another critical finding from this study was the need to build trust in the SCF. The novelty of this type of forecasts and the content of the information provided (i.e. average conditions over 3-month period) meant that most of the farmers involved in this analysis mentioned that they needed more time to gain confidence and trust in the information provided.
b) The Weather Roulette (RESILIENCE prototype)
• The Weather Roulette is a diagnostic tool that provides information about the skill of a forecast by providing an economic and financially oriented assessment in terms of effective interest rate and return on investment.
• One of the results was the demonstration that effective interest rate is a direct measure of the performance of SCF compared to climatology and that return on investment increases exponentially with the skill. This upholds the methodology as an engaging tool to explain the concept of skill and probabilistic predictions value to non-experts within the energy industry.
• More importantly, it provides a tool to analysts and technical staff within energy companies so they can use them internally to report their own benchmarks of SCF against climatology.
c) The Placebo concept (RIFF prototype)
• The Placebo protocol was followed by Etablissement Public Territorial de Bassin Seine Grands Lacs over the past 29 years and SCF shows a little added value versus RAF (historical simulation) but another important result has been that the current method used by the stakeholder is significantly worse than SCF and RAF.
d) Cost-Benefit Analysis (LEAP prototype)
• LEAP forecasts can be a critical component of raising funds in time to facilitate an early response to food security emergency. When this response is funded and triggered early, benefits arise as households avoid negative coping strategies, engage in greater investment, and avoid long term impacts to household growth, nutrition and educational outcomes.
• As a result, when the benefits of early response are incorporated into the analysis, responding based on forecasts becomes the most cost effective option in this analysis, allowing for savings of US$125M over the “business as usual” option.
e) Avoided cost (Climaware case study)
• Different types of SCF were introduced in a simple water management model and compared with results using only climatology. The hindcast period considered more than 40 years – for some of them there was a clear opportunity in using SCF but the change in DMP and avoided cost (related to flood and drought) could not be evaluated very precisely.
In parallel to the development and assessment of these methodologies, WP41 assessed the potential of the Regional Climate Centres (RCC) of the WMO European Regional Association for disseminating climate information products and their impacts on key stakeholders, especially the NMHSs (D41.1; Funk, 2016). This was assessed by an online survey that examined the characteristics and extent of products developed by NMHSs with respect to their sector-specificity and time-scales.
Furthermore, the impact and relevance of RCC products was also assessed. One product was particularly assessed since it had special relevance for the EUPORIAS project: the Climate Watch Advisories (early warnings) and its use and relevance for NMHSs. The results indicate a broad use of RCC products by NMHSs. Although all RCC products are considered to have a minimum relevance for most of the NMHSs, they are mostly used for the development of general national and sub-regional climate information products.
These products have to be further specified and tailored for the development of user- or sector-specific products by the NMHSs by using further sector-specific information on a national basis, which is mostly not available at RCCs because end-users are instead served by the NMHSs. Exceptions are seasonal forecast products, which are only used by a few NMHSs but are used permanently and have high relevance for the users. The potential use for user-specific applications is generally high. Climate Watch Advisories also have a general relevance for all NMHSs and have a high potential use covering weekly to seasonal time-scales as well as impact-related information relevant for sector- and user-specific climate information. RCCs, as providers of regional climate information, may become key institutions for laborious and capacity intensive tasks like seasonal forecast products. However, the pathway of regional climate service development and dissemination conditioned by RCC structures (top-down) may always require an additional user- or sector-specific tailoring of climate information on a national level.
3.9 WP42 – Climate service prototypes
The objective of WP42 was to develop an experimental operational prototype of a climate impact prediction system for a subset of the case-studies/sectors selected.
Given the ambitions of the project in terms of co-development it was decided that the definition of the climate service prototypes needed to be based on an objective evaluation of their specific merits in addressing user requirements. This meant that, contrary to other projects of this kind, no prototype was defined at the outset but they were instead dynamically defined during the course of the project. WP42, in close coordination with the project management team, opened a discussion on the criteria that make a climate service successful. Two of them were given disproportionate weights. On the one hand it was felt that the existence of a well identified user with a clear decision that can be informed should be considered as a necessary condition, on the hand it was felt that also the strength of the relationship linking the climate driver to the impacts and the predictability of that driver should be considered as an essential element in the evaluation of the prototype.
The criteria were defined and agreed upon ahead of the first project general assembly. The criteria were then passed to a panel of experts external to the consortium alongside the proposal for evaluation. All the proposals were considered eligible, i.e. above the threshold for the two required criteria, but in order to maintain an in-depth focus only a subset of them were selected for development. These were: LMTool, led by the Met Office and focusing on agricultural applications in the UK; RESILIENCE, led by BSC and focusing on wind power production world-wide; RIFF, led by Meteo France and focusing on freshwater resources for Paris; LEAP, co-led by ENEA and WFP and focussing on food security in Ethiopia; SPRINT, led by the Met Office and focusing on climate advice for the transport sector in the UK.
The dynamic development of the prototypes had forced all people involved to identify and involve end-users from the very beginning, something that has definitely improved the interaction between providers and users of climate services. Being part of the process from day one has gained buy-in to the project from the users that they would not have done so otherwise. The fact that some of those users such as the Clinton Devon Estate who we targeted in the LMTool, attended all our meetings and used their own communication channel to promote the work we were doing as part of the project is a measure of the success of this approach.
The team behind each prototype then worked at the assessment of the potential benefits the target users could get from the use of the prototype. This was first done at an abstract level then it was repeated using historical data as an input. Towards the end of the project each prototype was supposed to provide a real-time forecast to the users. This effectively happened for all but one prototype, and in at least three cases (RESILIENCE, SPRINT, and LMtool) the work done within EUPORIAS has then continued beyond the end of the project.
As well as the climate service prototypes the WP has also delivered a number of case-studies based on the prototype proposals that had not been selected for development. Whilst these case-studies did not have the same level of support and funding from the project, the work done within EUPORIAS provided the basis for projects that in some cases have then been developed into full projects elsewhere, as is the case for the Pro-Snow proposal led by TEC.
3.10 WP43 – Stakeholder engagement
The objectives of WP43 were to engage with European citizens about the use of S2D data in everyday decisions, to demonstrate ways in which a climate service can be developed to address specific users’ needs, to facilitate clear communication and exchange of information with stakeholder groups; and to empower SMEs, allowing them to develop their own climate services.
The major outcomes of this WP are as follows:
a) Stakeholder engagement
• A large number of meetings and workshops with stakeholders were held as part of this WP. The workshops were used to assess the users' understanding of the products delivered and provide an invaluable opportunity to understand ways in which we can improve forecast content, presentation and relevance, therefore helping to improve the service provided.
• FutureEverything organised a EUPORIAS lab during their festival in Manchester in March 2016. As part of this lab, a highly visible workshop with EUPORIAS partners was held. The target audience included non-specialists in industry and the general public. The approach and activity consisted in use of data design to communicate probabilistic information, and use of participatory experiences and events to engage various publics in science and innovation.
• In order to communicate some of the results of the project to an audience of non-specialists, two short video documentaries were produced by FutureEverything. The first of these videos presents project Ukko (http://project-ukko.net) the interactive web-browser based visualisation interface designed to provide stakeholders with access to the seasonal wind forecasts given by the RESILIENCE prototype. This video (http://vimeo.com/152558782) has been embedded in the project Ukko homepage and has been viewed nearly 10,000 times since being published in March 2016. The second video (http://vimeo.com/185503156) summarises the whole EUPORIAS project. It tells the story of the project through its aims, processes and outputs. The film includes interviews and narration from project partners and stakeholders.
b) Technical facilities
• During workshops with stakeholders, the possibility of having a mobile phone application was identified as key to bring the climate services developed closer to the user. As a result, in parallel to the corresponding microsite (http://lmtool.euporias.eu) a mobile app has been created for the LMTool prototype (Figure 1). This app is freely available for Android (http://play.google.com/store/apps/details?id=es.predictia.lmtool&hl=es) and Apple iOS (http://itunes.apple.com/us/app/lmtool/id1118042668?mt=8). It allows for visualising all the contents of the microsite and provides password-protected (as required by the Met Office) access to the short-term and seasonal forecasts delivered by the prototype for South West England.
• A second mobile app related to the RESILIENCE prototype has been created. This app (which will be published both in Google Play and the App Store) allows the user to assess the economic value of the wind speed seasonal predictions provided by the prototype, as compared to a climatological forecast. This is done by playing a roulette game from an interactive screen. In particular, the user can introduce the amount of money they would like to bet and the app simulates how much they would have won/lost if using the available probabilistic predictions. As in the case of the LMTool app, the EUPORIAS Application Planning Interface (API) is used to access the predictions provided by RESILIENCE.
c) A guideline for developing climate services
• As a result of the work carried out in this WP, a broad community of climate service developers (including forecast providers, scientists and communicators) and users has been established, which constitutes one of the major outcomes of this WP.
• The scientific paper produced to meet deliverable 43.4 (“Develop a mobile application to help access information produced by climate prototypes”) describes, for the illustrative case of the LMTool prototype, the steps that need to be followed in order to take advantage of the S2D climate information in land use decision-making. The complete development of the service is presented, from the first user engagement phase to the final products that are being delivered. The paper is currently being reviewed by the different co-authors and will be submitted to Climate Services for its inclusion in the EUPORIAS special issue that this journal is preparing for 2017.
3.11 WP44 – Delivery tools
The objectives of WP44 were to identify, develop, and maintain an interface that will allow an effective delivery of the climate services developed in WP 42 to both the general public and the relevant decision-makers. In this manner, EUPORIAS goes beyond merely improving the reliability of the underlying prediction systems to enhancing the usability of these forecasts in practical applications.
WP44 started in month 21, but WP leaders joined the earlier WP42 workshops and contributed to the development of the prototypes, enabling them to follow the developments in EUPORIAS before the official WP44 start. Their flexible approach allowed them to change and adapt quickly to achieve results. The main findings of the WP were as follows:
Task 44.1: Inventory of existing climate data portals and tools
• This task provides an inventory of existing climate data portals and gathers interface requirements on how the project climate service prototypes can be integrated into these existing portals;
• There are relatively few portals (Climate-ADAPT & climate4impact) that could host almost all prototypes and only Climate-ADAPT could host them all;
• Private company portals were excluded because of their data policy limitations;
• Many of the other portals are relevant for specific sectors or regions and it would therefore be desirable to formulate an architecture that would also allow appropriate prototypes to interface to these specific portals;
• To interface with these platforms both administrative and technical requirements must be addressed. Taking the Climate-ADAPT portal as an example the administrative effort includes obtaining permission from the portal administrators, ensuring the EUPORIAS prototypes meet the stipulated criteria for case studies, undergoing a review and annotation by experts outside of EUPORIAS. The technical effort would include assessing a range of integration approaches ranging from links to a dedicated website, integration of EUPORIAS datasets into their database or the addition of a tool.
Task 44.2: Climate service architecture development and improvements
• This task provided a plan detailing how to integrate the EUPORIAS prototypes into existing protocols inventoried in D44.1. In order to ensure flexibility in the data types, external portals, and external devices and to support efficient data exchange while ensuring proper control an API is proposed for the climate services.
• The EUPORIAS API was implemented (D44.5) and used by the LMTool prototype and in a mobile App related to the RESILIENCE prototype. The API is also used to publish EUPORIAS data on climate4impact.eu (C4I) and KNMI data centre (KDC).
• Making use of the EUSurvey service (http://ec.europa.eu/eusurvey) the API also allows the prototypes/case-studies developers to access the end-users' feedback in a comprehensive way.
• By providing an API to portal developers we can provide an abstracted interface, which hides the EUPORIAS prototype and case studies implementation details. In this way, portal developers can use a stable API, while EUPORIAS maintains flexibility in adding new or changed prototypes and case studies. This is shown in figure 4 below.
Task 44.3: Service delivery and operation
• This task aimed to enable a flow of relevant data from EUPORIAS to existing climate portals. The Service architecture (API) described above is implemented and tools have been created to bring the prototypes and case-studies developed in the project closer to the end-user, in particular the EUPORIAS microsites.
• A total of 11 microsites (one for each of the five prototypes and six case-studies developed in the project) have been created and delivered under the EUPORIAS domain. All of them share a common design and appearance, inheriting the visual identity created for the EUPORIAS website, and have the same general structure.
• They are user friendly interfaces for demonstration and promotion of the services developed within the project to both general public and relevant decision-makers. All these microsites have been gathered under the online component of the CUIP developed within the project, which can be reached at http://www.euporias.eu/cuip.
Task 44.4: Web interface for end users of the EUPORIAS service prototypes
• The API has been developed to ensure both the data and results emerging from EUPORIAS can be shared and, whenever appropriate, used by others, helping therefore to promote the project beyond its natural reach.
• Additionally, by making use of the EUSurvey service (http://ec.europa.eu/eusurvey) it also allows the prototypes/case-studies developers to access the end-users' feedback in a comprehensive way. All the code developed for this API is free and can be found here: http://github.com/Predictia/euporias-api. Additionally, the following page documents in details the API and provides some practical examples of use (for instance, how to connect to the API via bash, Java, Javascript and PHP protocols): http://api.euporias.eu/docs/api-guide.html.
Task 44.5: Generation and dissemination of easily accessible services.
• The aim was to reach a large audience of non-specialists. The process entailed selecting the relevant project results and services, translating these into easily accessible formats, and the production and publication of the messages.
• During the project, it was determined there was scope for T44.5 to draw on innovative methodologies and media in art and design to realise the WP objectives, above and beyond the tools described in the EUPORIAS description of work (YouTube, podcasts). A data visualisation interface was developed for one of the prototypes, along with a range of means to engage wider audiences, including a digital culture festival, in addition to the media described in the description of work.
• A design process was undertaken, through which it was determined a climate service visualisation interface would be developed with one of the prototypes. The result was the Project Ukko (www.project-ukko.net) visualisation interface developed for the RESILIENCE prototype.
• High visibility events were utilised, centred on a special edition of the FutureEverything festival, the UK's digital culture festival, established in 1995. Other events included Information+ in Vancouver and Liverpool International Business Festival. PR, media and podcasts were deployed to generate wider reach and engagement.
3.12 WP45 – Climate services as a business opportunity
This WP aimed to assess whether a market for climate services in Europe exists and whether this can be effectively and efficiently used by SMEs and other relevant stakeholders. A general methodology was developed and applied to specific key sectors to check the economic viability of climate service prototypes produced in WP 42 of this project.
The main findings of the WP are as follows:
• WP42 developed an experimental operational prototype of a climate impact prediction system for a subset of the case studies/sectors selected. WP45 has developed a general market analysis methodology to assess business opportunities of climate services and has applied this methodology to some of the prototypes developed in WP42 (water sector, renewable energy, road maintenance, and tourism).
• The methodology developed: (a) affords a general overview of a market (both the variables that affect an industry, and analysing the most relevant aspects of current and potential demand and supply); (b) an assessment of new services developed, from the point of view of their economic feasibility; (c) gives insight into a range of current innovative business models used, and hence helps assess which could be more suitable for the prototypes.
• The climate services industry in Europe is considered to be immature for the seasonal to decadal timescale in the sectors studied (water sector, renewable energy, transport and tourism). The wind energy sector seems to be somewhat more established than the others.
• All sectors identify the same current gaps in climate services information (e.g. lack of skill of forecasts, insufficient accuracy, etc.) and possible actions are suggested to overcome these gaps.
The main conclusions regarding the methodology developed are as follows:
• For each step, different business analysis tools and approaches have been proposed, tested and adapted. Thus, the methodology is now robust enough to be used for assessing other climate services external to the project.
• Relevant information has been gathered about markets (needs, gaps, competition, etc.) for the key sectors studied. This could be an excellent basis for the realisation of future market studies of new climate services in Europe.
The main results of the application of the methodology are as follows:
• All the sectors analysed are sensitive to weather and climate conditions, to a greater or lesser degree;
• All the sectors analysed think that the Copernicus Climate Change Service is going to have a positive influence on the development of the climate services industry;
• The information gathered during demand analyses based on WP12 interviews is of great significance for the market analysis (especially the information on willingness to pay). To have a better representativeness, additional interviews should take place;
• All sectors identify the same current gaps in climate services information, which are the lack of skill of forecasts, and the insufficient accuracy;
• The feasibility study conducted for the key sectors shows significant socio-economic benefits. For example, for the UK transport sector prototype (SPRINT) there are potential cost savings of ~£304 million/year using the 1-month temperature forecasts, and potential cost savings of ~€13,500 from scheduling five days of maintenance of a wind farm of 10.5MW turbines on days that turn to have good wind conditions for producing electricity.
References
Beven, K. (2012) Rainfall-Runoff Modelling: The Primer, Second Edition. Willey Library. Print ISBN: 9780470714591, doi: 10.1002/9781119951001
Bruno Soares, M., Bosi, L., Pouget, L., Reitg, E., Jiménez, I., Lledó, L., Soubeyroux, J.M. Viel, C., Creswick, J. (2016) Report on the evaluation of the value and impacts of using seasonal climate forecasts in decision-making processes. EUPORIAS D41.3 and D41.4. Available at: http://www.euporias.eu/system/files/D41.3%20and%2041.4_Final.pdf
Bruno Soares, M., Dessai, S. (2016) Barriers and enablers to the use of seasonal climate forecasts amongst organisations in Europe. Climatic Change, 137(1-2), 89-103
Butts, M., Xu, Y., Mueller, H.G. Jensen, R. A., Larsen, J.K. Godiksen, P.N. Richaud, B., Larsen, P. T., Corcuera Barrera, M.Á. (2016) Seasonal Forecasting for Water Resource Management: Reservoir inflows and uncertainties. Presentation at the XXI International Conference Computational Methods in Water Resources, CMWR 2016, University of Toronto, Canada, 20–24 June 2016
Donnelly, C., Andersson, J.C. Arheimer, B. (2016) Using flow signatures and catchment similarities to evaluate the E-HYPE multi-basin model across Europe. Hydrological Sciences Journal, 61(2), 255-273
Fraga, H., Malheiro, A.C. Moutinho-Pereira, J., Santos, J.A. (2014) Climate factors driving wine production in the Portuguese Minho region. Agricultural and Forest Meteorology, 185, 26-36
Funk, D. (2016) Review of the climate information products of the WMO RA VI RCC-network useful to the decision-maker. EUPORIAS D41.1. Available at: http://www.euporias.eu/system/files/D41.1_Final.pdf
Kahneman, D. (2011) Thinking, fast and slow. Macmillan
Kaye, N.R. Hartley, A. Hemming, D. (2012) Mapping the climate: guidance on appropriate techniques to map climate variables and their uncertainty. Geoscientific Model Development, 5(1), 245
Kloprogge, P., van der Sluijs, J., Wardekker, A. (2007) Uncertainty communication. Copernicus Institute for Sustainable Development and Innovation, Report NWS-E-2007-199
Manzanas, R., Frías, M.D. Cofiño, A.S. Gutiérrez, J.M. (2014): Validation of 40 year multimodel seasonal precipitation forecasts: The role of ENSO on the global skill. Journal of Geophysical Research: Atmospheres, 119(4), 1708-1719
Manzanas, R., Brands, S., San-Martín, D., Lucero, A., Limbo, C., Gutiérrez, J.M. (2015) Statistical Downscaling in the Tropics can be Sensitive to Reanalysis Choice: A Case Study for Precipitation in the Philippines. J. Climate, 28, 4171-4184
Manzanas, R., Lucero, A., Weisheimer, A.,Gutiérrez, J.M. (2017) Can bias correction and statistical downscaling methods improve the skill of seasonal precipitation forecasts? Clim Dyn, doi:10.1007/s00382-017-3668-z
Ogutu, G.E.O. W.H.P. Franssen, I. Supit, P. Omondi, and R.W.A. Hutjes (2016) Skill of ECMWF System-4 Ensemble Seasonal Climate Forecasts for East Africa. Int. J. Climatol., doi:10.1002/joc.4876
Osborne, T., Gornall, J. L., Hooker, J., Williams, K., Wiltshire, A., Betts, R. A., Wheeler, T., (2015) JULES-crop: a parametrisation of crops in the Joint UK Land Environment Simulator. Geoscientific Model Development, 8(4), 1139-1155
Pouget, L., Roldán, T., Gómez, M., Cabello, À., Camino, E.R. Navascués, B., Voces, J., Comas, P., Pastor, F., Gómez, M.C.G. Gil, J.J. (2015) Use of seasonal climate predictions in the water sector—preliminary results from the EUPORIAS project. Drought: Research and Science-Policy Interfacing, 321
Reyna, V.F. Nelson, W.L. Han, P.K. Dieckmann, N.F. (2009) How numeracy influences risk comprehension and medical decision-making. Psychological bulletin, 135(6), 943
Soubeyroux, J.M. Bruno Soares, M., Pouget, L., Jiménez, I., Torralba, V., Ortega, D., Viel, C., Bosi, L., Buontempo, C. (2016) Preliminary Guidance Document on the evaluation of the value of DMP. EUPORIAS D41.2. Available at: accessed 01.12.16
Taylor, A.L. Dessai, S., de Bruin, W.B. (2015) Communicating uncertainty in seasonal and interannual climate forecasts in Europe. Phil. Trans. R. Soc. A, 373(2055), 20140454
Williams, K., Falloon, P. (2015) Sources of interannual yield variability in JULES-crop and implications for forcing with seasonal weather forecasts. Geoscientific Model Development, 8(12), 3987-3997
Williams, K., Gornall, J., Harper, A., Wiltshire, A., Hemming, D., Quaife, T., Arkebauer, T., Scoby, D. (2017) Evaluation of JULES-crop performance against site observations of irrigated maize from Mead, Nebraska. Geoscientific Model Development, 10(3), 1291-1320
Potential Impact:
We have identified four main areas of potential impact, and main dissemination activities and exploitation of results:
1. Science: particularly downscaling methods and their value in climate services; the use of climate information indices; communication; impact modelling
2. Service development: particularly the user landscape; benefits and impacts of the EUPORIAS user data gateway and the R statistical software language; development of climate service principles; lessons to be learned from the development of the prototypes
3. Policy impact: particularly the ECOMS white paper; project governance; a cost benefit analysis
4. Business impact: particularly Project Ukko; the weather roulette; a mobile app; benefits and impact of the business analysis of climate services
4.1 Science
4.1.1 Downscaling methods and their value in climate services
There is great demand for user-relevant climate information to guide planning decisions at regional and local scales. This demand includes information about past weather and climate from observations, and future predictions. EUPORIAS examined the role of different downscaling methods in improving the skill and user-relevance of information at the regional and local level.
In broad terms, downscaling involves mathematical methods that take in data from large-scale simulations of the climate system, using global climate models, and outputs finer scale data that is more applicable to regional and local scales. Although there is a growing body of evidence on the relative strengths and limitations of different downscaling approaches, most studies focus on long-term climate change time-scales. In EUPORIAS we assessed the value of downscaling seasonal predictions, using different methods, to determine if downscaling approaches add value to the regional information provided by global climate model simulations. Computer model experiments were conducted using two different types of downscaling: statistical and dynamical.
Statistical models use the knowledge about the statistical characteristics of local observations to alter the outputs of a global model. These procedures allow the model outputs to look similar to the observations. Dynamical methods, using regional climate models, apply similar mathematical and computational approaches used in global climate models but over a limited area (usually covering a number of countries) to produce finer scale information. Dynamical and statistical methods were tested for the Greater Horn of Africa region, since this region provides a complex topography, has a reasonably strong influence from the climate phenomena El Nino, and, through the WFP, has a strong use-case where the impacts of the climate are experienced by society. Analysis focused on rainfall, as a key variable of interest for decision makers, and was conducted on a series of years that included both dry and wet years.
A significant challenge in the assessment of “added value” was evaluating the real observed state of the atmosphere. Observations in this region of the world are relatively poor, placing significant limitations on the quality of the analysis and strength of the conclusions. This stresses the importance of reliable long-term observational datasets in underpinning the development of climate services. Despite the challenges provided by insufficient observations, it was possible to assess the skill of the downscaled simulations compared to the skill of the global climate models.
Results show that there is no evidence of additional skill from the regional climate models in predicting the regional rainfall on seasonal time-scales. This result holds true even when looking at sub-national regions. This implies that there is no added value of using regional climate models compared to global climate models in this region (within the context of the experiments conducted) to feed into downstream products and services, such as the World Food Programme LEAP platform (drought early-warning system). On the other hand, the results suggest that regional climate models may provide improvements in the “reliability” of seasonal forecasts – that is to say the climate system is more realistically simulated, even if this doesn’t result in improvements in skill.
The results for statistical downscaling may show more promise. For the summer season the results suggest that the use of statistical downscaling has similar performance to regional climate models at a fraction of the cost and thus provide a more cost-effective solution to downscaling seasonal forecast information to provide more detailed spatial information for use at the regional and local scale. Nevertheless, it is difficult to generalise the results obtained from these experiments and the conclusions may differ for other regions and climate variables.
4.1.2 The use of climate information indices
The lack of user-relevant variables has long been recognised as one of the barriers limiting the use of climate information in sectoral decision making. Climate data, as supplied from models or forecast systems, is often not directly related to user decision needs, and can therefore be difficult for users to apply in their decision making. Climate variables, such as temperature, humidity or wind can easily be translated into Climate Information Indices, which offer the possibility of describing climate in a more user-relevant way. As an example, agriculture or water resource users may find the index ‘consecutive dry days’ more directly useful and relevant than mean precipitation data. Similarly, indices such as heating degree days can be linked directly to user-relevant quantities such as energy demand. As an additional factor, these indices can replace more complex and computationally expensive impact models, and may be of similar value to users.
To address this challenge EUPORIAS devoted a work package to the transformation of raw climate data into sector-relevant indicators and to the evaluation of their skill. The expectation was that the indicators could allow more users to interact with the data directly. There was also the hope that in some cases the calculation of an appropriate index could positively impact on key verification attributes of the forecasts, such as reliability and skill. Considering that most user-relevant indices are based on daily data, EUPORIAS developed a procedure to bias-correct daily time-series. All indices were then assessed using observations or reanalysis as a benchmark. The results suggest that, leaving aside the evident engagement value that using a tailored index brings, the verification attributes of the forecast are not significantly affected by the calculation of such indices. For simple post-processing of single variables (e.g. accumulation over time or exceeding a threshold) EUPORIAS demonstrated that the indices cannot have more skill than the underpinning variable. On the other hand EUPORIAS showed that in the case of complex multi-variable indices there might be occasions locally where the index shows more value than the variables used for its calculation.
One application of the findings of this work is in the forecast of wildfire risk. Wildfires represent a significant hazard in the Euro-Mediterranean region, in particular to land managers and populations in the rural-urban interface. Conversations and a survey circulated to a range of decision makers suggested that estimating fire risk a few months in advance could provide significant benefit, allowing appropriate pre-season land management, and allowing fire protection agencies a timely reaction and an adequate provision of resources. Within EUPORIAS we considered whether the skill in dynamical seasonal predictions of a climate information index describing fire weather risk (the Canadian Fire Weather Index, [FWI]) is sufficient to provide value to such decision-makers. It was found that there is significant skill in predicting above average summer FWI in some parts of the Euro-Mediterranean region at 1 month ahead. This skill was largely limited to an area covering Greece, Bulgaria and Turkey and displayed the potential for development of a useful seasonal forecast service in this area; however skill elsewhere was not sufficient to support service development. There is some suggestion from this work that skill levels could be improved by using a subset of the FWI components, some of which may more closely link climate to fire risk in this region. In addition future work may consider whether the skill of seasonal FWI predictions could be improved for some locations through the use of alternative statistical downscaling techniques where longer historical records are available.
A further study considered the value of sub-seasonal-to-seasonal climate forecasts of extreme temperature indices, and questioned whether reliable forecasts could allow for improved resource management for heat-health planning, protecting vulnerable populations and potentially preventing fatalities from temperature-related illness in Europe. In particular this study considered the extent to which such climate forecasts could be mainstreamed into heat-health planning, to support timely decision-making in advance of heat waves in Europe. Forecast data at lead times of 1 day to 3 months were used to drive a mortality model which produced probabilistic mortality forecasts running up to the 2003 heat wave in Europe. Through comparison with observed climate, this work found a range of levels of skill across the region and with different lead times. However, even at lead times of up to 3 months, some regions, particularly within Spain and the UK, did display significant skill in forecasting excess mortality. This work has suggested that for some regions at least, there is the potential for value in provision of mortality forecasts on sub-seasonal-seasonal timescales.
4.1.3 Communication
a) Communication with end users of climate information
For climate information to be useful in decision making, it has to be communicated in an understandable and actionable way whilst ensuring it does not lead to a false sense of certainty. EUPORIAS has focused on the communication of seasonal to decadal forecasts, although many of our findings are equally applicable to the presentation of climate information at other timescales.
To help us assess the strengths and limitations of different approaches to communicating climate information, including the representation of skill and uncertainty of the forecasts, researchers within EUPORIAS worked closely with the end users of the information, including those involved with the development of our prototype climate services. An important driver was that this research was led by experts from the social sciences who worked alongside climate scientists and other project partners to gather and analyse evidence. Some key findings were:
• No “one size fits all” solution exists for communicating seasonal to decadal climate information to end users, who have diverse needs and expertise. If possible, it is best to co-design and test the product with the end user, helping to ensure that the communication technique chosen does not mislead but aids the decision making process.
• Where tailoring is not feasible, a layered approach to presenting information alongside clear guidance to interpretation may address differing user needs.
• Even when forecasts have no skill relative to climatology, they may have influence on expectations about future states. We recommend that forecasts without skill not be provided unless specifically requested by users who understand that they do not provide ‘added value’.
Several research activities have developed a similar user-centric approach, with some directly inspired by EUPORIAS. The research techniques, sample products and tools developed in EUPORIAS will help the producers of climate information (co-)design their products to be more useful, actionable and robust.
b) Communication with the climate services and climate research communities
As well as journal publications, conferences and other traditional forms of communication, EUPORIAS engaged with other researchers though social media, including Twitter (@euporias). It created a summary infographic and supporting factsheets about the prototypes. We ran two masterclasses, particularly for early career researchers interested in the development and delivery of climate services. These were held at EURAC in Italy in May 2015 and May 2016. Alongside a varied programme of lectures, the students created example climate services for real-life user requirements. Such a hands-on a formula worked well and both masterclasses were very well received.
c) BellHouse – a novel communication activity within EUPORIAS
In the Spring of 2016, the UK-based arts organisation Kaleider and EUPORIAS made an International Commission Call inviting artists to submit ideas for playable artworks to be debuted at the EUPORIAS General Assembly at the Met Office in October 2016. We received over 60 applications worldwide and were overwhelmed with the quality of ideas. We commissioned Roop Johnstone from RAMP Ceramics to create his exquisite playable artwork – BellHouse.
BellHouse is a playful, interactive sound sculpture that translated the non-verbal communication of the delegates presenting at the General Assembly into the chimes of 35 bells in an opened sided house. A motion capture system devised by the Met Office Informatics Lab activated striking mechanisms associated with each ceramic bell generating a continuous chiming whilst each speaker at the assembly presented their research.
BellHouse was a novel collaboration. It put climate scientists in the shoes of others. Speakers at the meeting could listen to a playback of their talk, as played by the BellHouse. Although the sound they heard was a true reflection of their movements as they spoke, it was almost unrecognisable as their speech – perhaps reflecting how misunderstood even the most simple of messages can be. BellHouse explored how the arts and research can work together to create something beautiful, memorable and engaging and is one of the more unusual legacies of EUPORIAS.
4.1.4 Impact modelling
EUPORIAS completed a coordinated set of impact model studies across different sectors (agriculture, water, forestry and transport), over different regions and scales (Europe, East Africa, specific areas and basins), driven by baseline forcing data and different seasonal hindcast datasets (Figure 2). Some of the studies have also investigated the role of uncertainty in overall impact skill, such as arises from bias correction, initial conditions, seasonal hindcast skill, and model parameters.
Impact models for crops, water, forestry and energy were further developed, and impact hindcast simulations and analysis were defined, contributing to improved European capability for assessing long-range weather and climate impacts across these sectors. The findings help inform recommendations for impact assessments using seasonal forecasts across Europe and East Africa including the need to:
• Consider appropriate methods of bias correction of forecast data for applications where thresholds are important
• Consider how impact skill varies with variable, region, season, lead time and forcing dataset
• Note that skill for impact fluxes may be greater than that for state variables and that skill for some impacts such as hydrology, may be larger than that of its meteorological drivers
• Consider requirements for impact model initialisation. For example, skill for hydrology is often related to memory effects from initialisation of soil moisture and snow mass; in contrast, in East Africa where soils are often dry prior to planting, crop model initialisation may not be necessary.
In particular, some sectors, regions and seasons show greater potential for confident application of seasonal impact forecasts, including:
• For hydrology in North Europe, with better skill during winter than in summer, and better skill for high flows versus low flows
• For Swedish forestry, temperature forecasts have the potential to support planning of harvest operations, and soil moisture forecasts could support the planning of planting activities
• Fire Weather Index showed considerable skill in the eastern Mediterranean
• A capacity factor based on wind speed and temperature has the potential for assessing the future variability of wind energy resources
• Translation of predicted winter air temperatures, translated to road skin temperatures, show potential for road transport applications
For such applications, appropriate use of seasonal impact forecasts could aid preparedness and planning, and may avoid negative impacts such as: water shortages, flooding, crop failure and food shortages, poorly planned forestry operations, fire damage, energy and transport planning. For some impacts proper skill assessment is hindered by lack of high quality, spatially distributed observations, most notably so for crop yields (in Eastern Africa) so obtaining suitable datasets would strongly increase the potential for seasonal timescale food security assessments in these regions.
4.2 Service Development
4.2.1 Our user landscape
Before EUPORIAS, few research projects truly put the users of climate information at the heart of the research. Today it is not uncommon to see users helping shape research proposals and working closely alongside the providers of climate information within projects. EUPORIAS helped lead the way, working closely with our users throughout the four year project. In recognition of this, Steve Zebiak, former Director of the International Research Institute for Climate and Society and Chair of the international Climate Service Partnership, said ‘EUPORIAS is the most comprehensive research program in climate services today, anywhere.’ at one of the EUPORIAS General Assemblies.
Through exploring diverse sectors and gathering a robust evidence base to help inform the development of the prototypes, EUPORIAS engaged with the users, and providers, of climate information across Europe. It did this through online surveys, an extensive programme of interviews alongside a review of the existing literature. The results suggest that the user landscape, and potential user landscape, within EUPORIAS was undoubtedly complex.
The knowledge gathered was analysed, communicated through the project deliverables and journal articles and presented at conferences. As the climate services landscape develops, it is vital that we share such knowledge, while acknowledging commercial considerations, so that the climate services community can benefit from such in-depth research. User fatigue is of concern but by working across providers and sharing our understanding of the user landscape, we can seek to maximise the relevance of climate information across sectors and improve its use in decision-making.
4.2.2 Benefits and impacts of the User Data Gateway (UDG) and R statistical software language
The ability of research partners to access and use the data produced by EUPORIAS was key to the success of the project. The ECOMS User Data Gateway (ECOMS-UDG), combined with the use of R (an open source programming language and software environment for statistical computing and graphics) to explore the data, ensured that all data used within EUPORIAS was available in a common accessible location.
Tested and quality controlled hindcast seasonal datasets made available to EUPORIAS, SPECS and North Atlantic Climate (NACLIM) partners through the ECOMS-UDG include ECMWF System 4, Met Office GloSea5, Météo-France System 4, and National Centers for Environmental Prediction Climate Forecast System v2 (NCEP CFSv2), along with access to SMHI-EC-Earth hindcast data, ECMWF ERA-Interim, and E-OBS. The ECOMS-UDG is hosted by the IT infrastructure of the University of Cantabria. A set of tools provided within the ECOMS-UDG, written in R, allows data to be manipulated. Predictia has provided user support for the R data access tools.
Benefits of the ECOMS-UDG include: project data being available from a common location; data are available in a predefined format; development and access within the ECOMS-UDG to a standard set of data processing tools, written in R, tailored to the needs of stakeholders; use of a standard input/output protocol to prevent duplication of pre-processing steps and facilitating shared analysis and multi model ensemble creation.
The ECOMS-UDG provides over 64TB of data to forty users from EUPORIAS and fifteen users from the SPECS projects, allowing both cutting-edge research and prototype climate service development. To gain access to ECOMS-UDG the user has to agree to the licensing and distribution terms and conditions required by each institution providing data. This highlighted the restrictions and limited availability of data from Global Producing Centres. For a number of impact applications, both daily data and continental geographical scale is required. However, restrictions are often placed on the temporal and spatial extent of released data. This led to some agreements being signed to make more data easily available to partners, and the ambition is that the Copernicus Climate Change Service will provide further free access to useful data for impact applications.
4.2.3 Development of climate service principles
The production of the prototypes highlighted several challenges and best practice ways of working. These were considered and collected in November 2014 at a workshop of thirty international experts in climate services. Through an interactive and dynamic workshop the attendees identified seven principles that should be considered when developing successful climate services. These principles helped to influence the prototype development within EUPORIAS and will be of use to the wider climate services community. The seven principles are illustrated in Figure 7, and covered the following broad topics.
Who are the users and possible users of the climate service? What is the user-chain? What are the motivations of each participant to take part in the project? Are all the relevant people involved in the discussion? Do the providers have all the skills needed to deliver the service? What expertise will the users bring to the service development? It is essential that the scope is clearly defined at the beginning of the project AND to ensure there is a common understanding about how it is evolving. Be honest about what is and is not achievable within the project. Be open about new ideas that can alter the course of the project. Communicate all the possible issues (scientific, technical, legal, political or commercial) which could limit the service at the start and throughout development. Take the journey together. The service should provide value to users but it is also important to identify value (not necessarily monetary) to the provider. Make clear what each actor involved is expecting to get out of the service. Be flexible, expect changes in the scope. Maintaining a highly interactive and flexible work-programme you will be able to account for some of those changes, while considering the boundaries of flexibility. Scope-deliver-evaluate: iterate. If possible divide the service in small components that can be delivered separately. Scope each of these, deliver and evaluate them with the users and then, if necessary, re-scope. Some project management practices (e.g. agile) are intrinsically designed for these sort of applications.
4.2.4 Lessons to be learned from the development of prototypes
EUPORIAS developed five climate service prototypes (with a sixth funded outside of the project and developed by SMHI). These were LEAP (for food security), RIFF (water management), SPRINT (transport), LMTool (land management) and RESILIENCE (wind energy). EUPORIAS sought to involve users directly in the governance of the project and include them in the problem definition. The prototypes were selected through a competitive process focussing on ensuring a tangible user drive and being likely to provide useful results. Many challenges arose during the production of the prototypes and valuable lessons, applicable to many other climate service development activities, are summarised here:
• The interaction with users during the development of a climate service cannot be sporadic and cannot simply occur at the beginning (e.g. service definition) and at the end (e.g. service evaluation) of the service development.
• It is essential to allocate sufficient time to the dialogue with the user and to any required changes in the definition and scope of the service being developed.
• In addition to the benefits that users could gain from a climate service tailored to their needs, access to climate expertise during the development of the service is an important added value to the users.
• Top-down management practices are not necessarily the most suitable for developing climate services. Adopting a flexible management approach (e.g. Agile) can be an advantage in an environment where changes in scope in response to user feedback are to be expected.
• User representation (or lack thereof) in the governance structures of climate service projects and the way in which these projects are linked to downstream business opportunities have a direct impact on their ultimate usefulness to society.
4.3 Policy impact
4.3.1 White paper
EUPORIAS, through the ECOMS coordinating mechanism, has been central to identifying and providing priorities, and research and investment needs in the field of climate observations, climate modelling and climate services. The ECOMS group (coordinators of EU FP7 climate modelling and climate service projects, and representatives from European climate research and climate service centres) met in February 2013 and the conclusions of that meeting are summarised in a public white paper (Deliverable 2.1) which provides recommendations from ECOMS on the priorities for Horizon 2020 (Societal Challenge 5) in the field of climate modelling and climate service development, underpinned by observations. The EC has managed to take on board all of the recommendations from EUPORIAS and ECOMS in launching subsequent calls under Horizon 2020. ECOMS has been superseded and replaced now by the Climateurope (www.climateurope.eu) Coordination and Support Action under Horizon 2020, ensuring the legacy of ECOMS will be maintained and built upon.
The EUPORIAS coordinator has been involved in his capacity as ECOMS chair, and through his role under the UN’s GFCS, at the invitation of the EC’s DG (Environment), to act as an expert in a series of climate service-related workshops. These include the Belmont Forum (Goa India, November 2013), H2020 Challenge 5 stakeholder meeting (Brussels, April 2013) and ‘Towards a European market of climate services’ (Brussels March 2014).
4.3.2 Project governance
EUPORIAS has been built on the concept of co-design of climate services between users and providers. At times this meant stretching the limit of what was possible in the context of an EU project and we are very grateful to the EC’s project officers for their support in what we were trying to achieve. The dynamic allocation of the prototypes gave the end users an opportunity to influence and define the scope and structure of what we were trying to do. However, the project was perhaps less user-driven than desired (although not necessarily less than originally envisaged) and this provided an important learning opportunity for climate services in general. Despite the presence of a stakeholder board, the governance of the project was still skewed in favour of climate providers and users did not have sufficient leverage to alter the course of the project towards something they could immediately benefit from, perhaps an inevitable outcome given the contractual need for the project to reach its agreed milestones and deliver its agreed deliverables. This is understandable in the context of a research and demonstration project, but a different style of project governance should be considered when developing operational services to meet user requirements.
The results of EUPORIAS suggest that co-design works well when all parties involved have a clear interest in the developing of the service but this interest should go beyond the financial benefits that partners may gain from being part of the project in the first place. A good measure of success would be the emergence of joint ventures whereby prospective users are prepared to invest and risk their own money on the success of an emerging service.
4.3.3 WFP LEAP Cost Benefit Analysis
The LEAP Cost Benefit Analysis (CBA) provided a quantifiable estimate of the benefits of introducing seasonal forecast capability within Ethiopia’s national early warning system for food security. The analysis was carried out by comparing three scenarios, using data from 2003-2010:
• Late humanitarian response via needs assessments, using historic data on the numbers in need;
• LEAP Current: Early humanitarian response based on current LEAP predictions at the end of the crop season (August/September);
• LEAP Forecast: Early humanitarian response based on LEAP Forecast, with numbers in need forecast four months before the first failed rains (February/March).
Based on the cost of food/cash alone, the LEAP forecasts result in a higher total cost over the eight years of analysis as compared with a late response based on needs assessments. This is because costs are allocated to the full number of beneficiaries when LEAP over-predicts the number in need (as compared with needs assessments) and is topped up to needs assessment figures whenever it under-predicts. Therefore, responding based on LEAP forecasts will always appear more expensive based on a cost analysis alone. However, LEAP forecasts can be a critical component of raising funds in time to facilitate an early response. When this response is funded and triggered early, benefits arise as households avoid negative coping strategies, engage in greater investment, and avoid long term impacts to household growth, nutrition and educational outcomes.
Early action can cover a wide range of activities, from agriculture conservation practices to prevent loss of crops, to livestock initiatives that prevent animal death, as well as temporary water measures and preventative nutrition interventions that are initiated at the first signs of an impending drought. LEAP is not currently structured to trigger early action measures as a result of forecasts; and yet this is part of the design of LEAP and likely to be the area of greatest benefit – benefits that will more than outweigh the additional costs of LEAP.
The cost of response over the eight years is summarised in Table 1. The investment in early action through LEAP Forecast would save $125m over the cost of responding at the time of needs assessments. As a result, the CBA outcomes so far point out the considerable quantifiable benefits of using LEAP seasonal forecasts, as well as additional benefits that we are currently only able to estimate, which would have strong positive repercussions on how humanitarian crises are managed, and represent a considerable step towards increased disaster risk reduction and prevention of the worst impacts of hunger and malnutrition on vulnerable communities in Ethiopia.
These results should persuade LEAP stakeholders, including the Ethiopian Government and international donors, to introduce seasonal forecasting to improve current early response initiative. Unfortunately, while it is part of LEAP’s vision, LEAP is not currently established to trigger contingency planning and early action based on forecasts. And yet this is a significant component of the benefits that can be realised. Therefore, it is recommended that key stakeholders for LEAP from the Ethiopian Government and other humanitarian and development organisations discuss and lay out a plan on how LEAP can most effectively be used to trigger early action, appropriate measures, and estimated costs and benefits, to incorporate into a revised version of this analysis, and eventually make the seasonal forecasting capabilities of LEAP operational.
4.4 Business Impact
4.4.1 Project Ukko
Project Ukko presents an entirely new way to look at seasonal wind predictions. Recognising opportunities in the emerging climate services market, and the potential utility of seasonal wind speed forecasts for the energy sector, EUPORIAS and the RESILIENCE prototype developed an online visualisation platform http://project-ukko.net/ to inform assessments of wind energy production for the season ahead.
The project involved an interdisciplinary team, composed of design researchers, climate scientists and a data designer, working to integrate design and visualisation approaches to communicate complex and uncertain data in an engaging and accessible way. The interactive online user interface provides forecast wind information overlaid on a world map, using symbols and different visualisation techniques to summarise forecast parameters. These include the predicted wind speeds (given by line thickness), prediction skill (given by line opacity), and the most probable trend taken from an ensemble of forecasts (given by line colour and rotation). The interactive map includes the locations of major wind farms, drawing attention to those areas of interest where the forecast shows high probabilities of significant changes in wind speed. Selecting a location enables the user to obtain detailed information about the past observations and the future predictions.
One of the greatest challenges was dealing with the large volume of data required to support the interface. The seasonal forecast data is provided by a large number of model simulations taken from ECMWF’s System 4 seasonal prediction system. To summarise and communicate the seasonal forecast information, integrated approaches were required to ensure probabilistic information is communicated in a scientifically credible way, while ensuring accessibility to non-scientists.
In the development of Project Ukko, end-users were engaged throughout using various approaches to help inform and direct the climate service design and visualisation process, strengthening the relevance and usability of the service. Ultimately, the visualisation approach helps in closing the gap between the users and the scientists.
Lessons to be learned include adopting a common language and clear context for collaboration across disciplines, having a sustained dialogue with end-users and other actors using conventional and novel user-engagement methods, compromising between scientific soundness, functionality and aesthetics, and having a well-structured dissemination and engagement strategy, executed through different communication and discourse channels.
4.4.2 Weather Roulette
A “weather roulette” tool has been developed in EUPORIAS to help decision makers understand the potential economic value of weather and climate forecast information. The tool provides information about the skill of forecasts and their value as a basis for guiding risk management decisions. It has been demonstrated for use in the wind energy sector using seasonal wind speed forecast information from the RESILIENCE prototype forecasting system.
In the case explored, the weather roulette tool works by using slots to represent three tercile categories - normal, below normal and above normal seasonal wind speeds. Using past observations, threshold wind speeds for each of these categories are selected. A climatology “forecast” then assumes equal probabilities for each category – i.e. a 1/3 chance of occurring – and this is compared to adjusted probabilities taken from the RESILIENCE forecast system (based on calibrated forecasts from ECMWF’s System 4 seasonal prediction system).
The tool was tested for 37 locations around the world where wind farms have been installed. Using different statistical skills scores, the tool demonstrates where using seasonal forecast information can increase the value of investments for individual wind farms and for portfolios of wind farms. Results were mixed, showing value in some locations and portfolios but not for others.
To further demonstrate and test the weather roulette tool with stakeholders, a Smartphone app has been designed. This enables people to experience making investment decisions using the RESILIENCE forecast system and climatology for different locations across the world. The user reinvests an initial capital of 10€ over a single year, or a succession of multiple years, to see if investments are more or less profitable when made using seasonal forecasts or climatology.
The weather roulette tool has proved to be an engaging way of explaining the concept of skill and the potential value of probabilistic predictions. Interaction with stakeholders to assess the tool has explored how such a methodology could impact on the understanding and uptake of seasonal forecast information. Whilst there is a general agreement that the lack of predictability precludes the use of seasonal wind predictions at present, users expressed their willingness to see more results from the weather roulette approach on the performance of real-world wind farm portfolios.
4.4.3 Clinton Devon Estates mobile app
The Met Office, University of Leeds, Predictia and KNMI developed the Land Management Tool (LMTool) in close collaboration with Clinton Devon Estates and the National Farmers Union in the UK. LMTool is a prototype climate service providing seasonal climate forecasts to support land management-related decision making for Southwest UK. This service focuses on the winter months since recent advances in the prediction of the North Atlantic Oscillation (NAO) allows for better seasonal forecasts of the Northern Europe winter climate.
The LMTool was iteratively developed between January 2014 and May 2016, building strongly on a range of stakeholder engagement activities (workshops, interviews, surveys and feedback gathering) carried out with farmers and land managers. During the first winter (2014/2015), the project worked closely with a small, representative subset of farmers to blueprint the prototype service, providing three month outlooks of temperature and precipitation in hardcopy and email. Insights gained during the first winter were then taken forward, alongside engaging a wider farmer group, in developing forecast products for the following winter (2015/2016): three month outlooks of temperature and rainfall for the whole UK, and also 14-day forecasts of rain, temperature and winds for a set of weather stations across South West UK. These products were delivered via an interactive password-protected website (which forms part of a more general micro site (http://lmtool.euporias.eu/) and a mobile app. These platforms have been found to be very useful to carry the prototype to the public.
4.4.4 Benefits and impact of the business analysis of climate services
A five step climate service market analysis methodology developed analyses of industry, demand, supply, feasibility and business models. This approach considers both an overview of the market, and assessment of newly developed services (i.e. the prototypes) in terms of economic feasibility. The analysis showed that all of the sectors analysed are sensitive to weather and climate conditions. The climate services industry in Europe is considered an immature emerging market, especially for the seasonal to decadal timescale, and for the sectors assessed. There is potential demand for these services which is currently unmet. All sectors identify the same gaps in current climate services provision, e.g. lack of forecast skill and insufficient accuracy. Possible actions are suggested to overcome these gaps. All of the sectors analysed think that the Copernicus Climate Change Service will have a positive influence on the development of the climate services industry. The feasibility study conducted for the key sectors shows significant socio-economic benefits, for example, potential cost savings of ~£304 million/year using the 1-month temperature forecasts for the transport sector prototype, and potential cost savings of ~€13,500 from the maintenance of a wind farm of 10.5 MW turbines scheduled for a period of five days that has good wind conditions for producing electricity.
The benefits of the climate services business analysis work include the provision of a well defined holistic methodology for assessing the EUPORIAS prototypes, which can be used for assessment of any climate service. Also the analysis provided a comprehensive market assessment of climate services in key sectors that can be referred to when developing legislations and policy, and justifying the economic viability of future climate service developments, as well as the potential market gaps and risks in service development.
List of Websites:
www.euporias.eu
EUPORIAS Coordinator, Dr Chris Hewitt (chris.hewitt@metoffice.gov.uk)
EUPORIAS Science Coordinator, Dr Carlo Buontempo (carlo.buontempo@metoffice.gov.uk carlo.buontempo@ecmwf.int)
EUPORIAS Project Manager, Paula Newton (paula.newton@metoffice.gov.uk)
