Environment, 1992-1994

A project has been set up to provide timely information on vegetation cover status to decision makers at both qualitative and quantitative levels. The information allows agricultural surveillance at the level of the European Union and contains indications on eventual agricultural alarm situations which might occur due to extraordinary agrometerological conditions. The primary sources of information are data acquired by the advanced very high resolution radiometer on board a series of meteorological satellites. A regional up to continental synoptic view is provided. The data acquisition rate is high and the data are cheap in comparison to high resolution remote sensing products. The information extracted covers 3 types of thematic fields of application: qualitative status of the vegetation cover, where emphasis is placed on the detection of abnormal vegetation cover, mapping of land cover units; plant physiological characteristics data which can be used to attempt derivation and monitoring of quantitative yield estimates. A prototype data preprocessing software system has been finalized and an archive of three years of daily mosaic images covering the territory of the European Union generated. Automatic procedures have been implemented to extract from this image archive time series of geophysical parameters which can be elaborated further in a geographical information system environment. Problems have been addressed by the processing system where state of the art geometric, calibration and atmospheric correction algorithms have been implemented. The product generated by the processing system consists of a daily mosaic of radiometrically and geometrically corrected 5 band images. The availability of a historical archive of data allows comparison of the prevailing vegetation state with any previous monitoring period. The cartographic products of these geophysical indices can be linked with other statistical or geographical information. This type of information is then used by the agronomist to compile the timely vegetation status bulletins.

Microwave imagery has been investigated as a replacement for the present optical data collection for the rapid estimating of crop acreage by remote sensing techniques. As the earth remote sensing satellite-synthetic aperture radar (ERS-SAR) imagery is significantly cheaper than the normally used systeme probatoire d'observation de la terre (SPOT) or thematic mapper (TM) data, a successful substitution of the optical by radar data, will improve the cost effectiveness of the rapid estimates project. The all weather capabilities of the SAR make image acquisition during autumn and winter time possible. The crop specific field cultivation practices and the resulting effects on the radar surface roughness, make it highly probable that these data can be used to perform an early estimation of the crop acreage. Work began with establishment of a Seville test site data base covering an area of 1600 square kilometres. The results of the 1992 ground survey based on 17 segments were georeferences and included in this database. After an initial evaluation of fast delivery (FD) data, work concentrated on the preprocessing for ERS-1 precision images (PRI) data. Software for image resampling was developed. A second research axis was oriented towards the definition of occurrence masks. Occurrence masks define areas, where certain types or associations of crops have a high likelihood of occurrence. First results indicated that the performance of standard classifiers can be significantly increased by the use of a priori occurrence masks. A third research area was the use of high (geometric) resolution Russian satellite imagery. KFA-1000 data with 5 m resolution began. The idea was to use these data for a high precision base segmentation in order to derive field boundaries. As field boundaries will not change dramatically from one season to the other, the field boundaries will be a valid reference system for further data analysis. The segmentation of the KFA-1000 however, did not show satisfying results due to image noise present in the data.

Exploratory research has been carried out to undertake an inventory of radiation balance component data and instrumentation available in Europe, examine the possibility of interpolating global radiation measurements, and formulate practical techniques to estimate solar radiation from existing observations of readily available parameters. Although radiation balance in not usually available over the global telecommunication system (GTS), sunshine hours are often reported. Radiation balance can usually be calculated from sunshine hours using the Angstrom formula. This formula, however, contains 2 empirical constants which are specific to individual locations. The first phase of this exploratory research program included a review of the radiation balance component data and instrumentation available in Europe and resulted in a comprehensive list of stations which collect data on either radiation or sunshine hours. On the basis of the expanded dataset, the 2 empirical constants were estimated on a monthly basis for several hundred European stations. Using these estimations, radiation could be derived at any of the studied stations on a real time basis from sunshine hour values reported over the GTS. To estimate radiation values at other points, it was necessary to examine the seasonal and spatial variation of the 2 constants. It was found that the seasonal variation of the 2 constants fitted various sets of sine and cosine curves. The spatial distribution of the factors was best represented by isolines across Europe. On the basis of these isolines and the trigonometric curves, the values for the constants for any point in Europe could be estimated. So, wherever sunshine hours were reported, radiation could be estimated. Calculation of radiation in this manner was found to be more accurate than direct interpolation of radiation values. In some areas even sunshine hours are not reported, and the research program also developed equations for the estimation of radiation values from cloud cover and daily minimum and maximum temperatures and from daily minimum and maximum temperatures only. The former gives somewhat more reliable results than the latter.

A project has been initiated to establish a regional network and to develop a methodology for the real time monitoring of the hydrological resources in the mediterranean area of the European Community. The methodology which has been adopted is based on the combined use of agrometeorological models and remote sensing data for the monitoring of the vegetation development and the water balance in the soil: the project runs agrometeorological models based on the input of data from a large meteorological data base, a soils data base and a digital terrain model of the region under study. The meteorological data base is being updated on a daily basis. In addition, daily data are processed into a mosaic covering the entire region of study. The data are calibrated, atmospherically and geometrically corrected according to the most up to date algorithms. Finally several derived parameters, like the surface temperature and vegetation indices are used to monitor the environmental conditions. Establishment of the network and the collection of additional data has been completed. In addition, the agrometeorological and satellite processing chains have established in order to provide first products which allow for a qualitative monitoring of the hydrological situation. Since April 1993 the project has edited a monthly bulletin giving information on the hydrological status throughout the region as well as on the development of the agricultural season. The bulletin contains maps of several meteorological parameters, such as temperature, rainfall, evapotranspiration and climatological water balance as well as maps on the current status of the main agricultural crops. The comparison of the model outputs with the actual situation, as reported by the regional teams involved in the project, has shown that the agrometeorological as well as the remote sensing tools are able to provide qualitative monitoring of the seasonal development and to pick up adverse developments at an early stage.

Agrometeorological models based as a comprehensive database containing information about weather, soils, topography and crop parameters have been developed to improve assessments of crop state and yield forecasts at the level of the European Community member states. A prototype model is now installed and being used to produce preoperational outputs every month. The geographical data which are the input to the agrometeorological models can be divided into 5 main categories: weather, soils, crop parameters, topography and historical crop statistics. In some cases the agrometeorological models can be improved by the inclusion of data which are not readily available. Parameters such as potential evapotranspiration, solar radiation and soil water balance are not normally available in meteorological bulletins. Work has therefore been undertaken to develop algorithms to derive these data from information which is available. The agrometeorological models are designed to produce results on a regular spatial grid, initially containing cells of 50 kms by 50 kms. The input data, however, are not received on the basis of this grid. The meteorolgical data, for instance, are collected from stations which are irregularly distributed across Europe. Work has been undertaken to develop methodologies to interpolate weather measurements from the stations to the grid squares. The present algorithm which interpolates basic data such as temperature and wind speeds is being validated and refined. The algorithms developed for crop modelling have been implemented in a prototype system. This system guides the user through the processing chain and presents the results in the form of maps and tabular reports. The database developed for the main agrometeorological model has also been used as the basis of an exercise to develop an algorithm for estimation of long term mean grape sugar content.

A project has been established to define and demonstrate how remote sensing could be used operationally to supplement, interpret and standardize agricultural statistical data provided by conventional techniques. Using the methodology of computer assisted photointerpretation, the work was directed towards analysis at a field level, so meeting the requirements of individual declarations verifications. Control zones were defined either as squares of 50 by 50 km or circles with a radius of 25 km so as to be compatible with a single systeme probatoire d'observation de la terre (SPOT) image. Usually, 3 dates for the acquisition of multispectral images are defined during the crop year (September/October, April/May, June/July). A panchromatic image is also acquired in April. Field boundaries from the farmers' crop plans are digitized either from cadastral or topographic maps, from aerial photos or directly on screen from satellite images. Computer assisted photointerpretation has always been used for classification of the land use within each digitized parcel. Automatic classification has now been introduced. Each dossier is classified as conforming to declaration, nonconforming or doubtful. Nonconforming and doubtful dossiers are selected and an image hard copy with the boundaries of problematic plots provided to field controllers. Remote sensing techniques optimize efficiency by guiding field controllers to nonconforming sites, thus reducing the time (and cost) of on the spot checks as only certain fields have to be measured.

An operational method has been developed for regional inventories of crops or other land use, based on area frame sampling and high resolution satellite images. The main software used to carry out the estimation method is operational, but not user friendly. A functional analysis has started for an in depth renewal of the software. A working scheme to produce frame stratifications has been developed. Support is being provided to digitizing segment drawings from field survey through television tracking system. A study has been undertaken to establish a working scheme for image classification and correction of estimates. The main conclusions of the comparisons of results are that a sample of n TER-UTI segments of 36 points on a squared grid covering 1800 m by 1800 m, gives better precision than a sample of n complete segments of 700 m by 700 m, involving field drawing and digitizing. However, point sampling is less suitable than segment sampling for matching with remote sensing. The relative efficiencies of regression correction of area estimates by remote sensing on point segments is much lower than that on complete segments. In most regions, samples have been drawn with a systematic aligned method. A method is being studied to improve the estimation by considering clusters along the sampling blocks and permutations within the blocks. A subsample of the segments of the ground survey is being revisited several times during the year to obtain a fast assessment of the crop rotation, state of crops and area evolution.

A project has been set up to: provide decision makers with rapid estimates of changes in acreage for various economically important crops using high resolution satellite data; design, develop and implement a transportable operational information system in a computerized environment for future use by other services of the Commission. High resolution satellite imagery has been used to provide rapid estimates of annual changes in areas of various important crops in Europe, and estimates of the potential yield of these crops. The system was designed to give results only at the European scale, and not to provide accurate statistics at a more local level. The project now monitors 53 sites scattered over Europe each 40 by 40 km (6% of areas used for farming in Europe). The agricultural information must be delivered rapidly in order to be of use, a target of 10 days has been set. This requires a novel and industrial approach to the image analysis and the project has therefore developed automatic geometric and radiometric correction of the incoming data, computer aided image interpretation without access to up to date ground data, a knowledge oriented data base, automatic classification using a merge and split algorithm, multitemporal cross classification, and validation of the interpretation once a year, using ground data collected specifically for this purpose. Each site includes anything up to 10 000 fields. It would be an enormous task to examine each field individually. Instead, within each site the interpreters analyse in detail about 16 areas, known as segments, of just under 50 ha each. These same segments are used by ground teams to collect information on the crops growing in each field in every segment. The data collected by the field teams are used to check the work of the interpreters at the end of the year. Overall results seem to confirm the operational character of this technique to acquire agricultural statistics.

Exploitable results

Share this page

Download