Periodic Reporting for period 1 - MultiMiner (MULTI-SOURCE AND MULTI-SCALE EARTH OBSERVATION AND NOVEL MACHINE LEARNING METHODS FOR MINERAL EXPLORATION AND MINE SITE MONITORING)
Reporting period: 2023-01-01 to 2024-06-30
European Union has set a goal to become a carbon-neutral economy by 2050. To achieve the goal industry transformation is required in the raw materials sector to supply critical (CRM) and strategic raw materials for the European market. Thus, the EU Action Plan for CRM 2020 and the CRM Act of 2023 were formulated to list concrete actions aimed to mitigate supply risks, ensure environmental protection, and the EU’s strategic autonomy in the raw materials market. In addition, the European Green Deal and the Circular Economy Action Plan and a number of international, EU and state level regulations provide frameworks to support integration of sustainability and circularity principles across the raw materials sector. The resilient supply of raw materials also requires exploration of previously undiscovered mineralization within EU but mineral exploration should be conducted preserving the environment. Thus, traditional CRM exploration methods with low environmental impacts should be complemented with new advanced methods. Additionally, the extraction of raw materials should be conducted sustainably to align with the above mentioned regulatory and policy frameworks.
Remote sensing provides a set of low-impact tools to support the entire mining life cycle including mineral exploration, operational, closure and post-closure stages. The use of remote sensing in its full potential has been hampered by lack of understanding of the method applicability in mining industry, need for easy-to-use data processing tools and availability of suitable in situ data. The goal of the project Multi-source and Multi-scale Earth observation and Novel Machine Learning Methods for Mineral Exploration and Mine Site Monitoring (MultiMiner) is to develop novel data processing algorithms for efficient utilization of Earth Observation (EO) technologies for mineral exploration and mine site monitoring. MultiMiner unlocks the potential of EO data, including Copernicus, commercial satellites, airborne, drone and in situ data, to support the entire mining life cycle. The general objectives of MultiMiner are to 1) develop scalable and automated approaches for mineral exploration based on multi-source EO data and sparse in situ data, focused on mineral deposits hosting CRMs across EU, and to 2) leverage novel EO data analysis methods to make the most of scarcely available in situ data for timely mine site monitoring, reducing both disruptions to mining activities and environmental impacts.
The goals are achieved by creating generic but innovative unsupervised and weakly supervised Machine Learning (ML) approaches and tools for multi-scale image processing which augment and potentially substitute in situ data collected on the ground. For exploration purposes, we will develop of a novel mineral mapping algorithm performing automatic spectral feature extraction from a comprehensive mineral spectral library. In addition, a mineral prospectivity wizard facilitating multi-scale mineral mapping and automatic interpretation for non-experts will be constructed.
For environmental monitoring of mining activities, we will develop EO solutions for 1) 3D/4D monitoring of composition and volume of tailings storage facilities, 2) automatic interpretation of European Ground Motion Service products, producing maps for dam stability and open pit stability, 3) algorithms and maps of ground moisture, supporting dam stability as well, 4) monitoring algorithms for water quality and acid mine drainage, 5) algorithms and maps of revegetation status, including vegetation structure and biodiversity and 6) multi-source models for atmospheric and surface dust monitoring. Overarching most of these thematic applications, an ambitious Generic Mine Site Monitoring algorithm (GMSM) will be developed, that can perform these monitoring tasks with advanced deep learning methods making the most of the vast amounts of available EO data and sparser in situ data in respective applications.
The project focuses on new EO based exploration technologies for CRM to increase the probability of finding new sources within EU thereby strengthening the EU autonomy in the area of raw materials. MultiMiner EO-based exploration solutions are fast, economically efficient and have an extremely low environmental impact. They improve safety and are environmentally safe, making them potentially more socially accepted. The project’s solutions for mine site monitoring increase the transparency of mining operations as environmental impacts can be detected as early as possible and digital information of the currently unexploitable raw materials can be stored for future generations. MultiMiner aims to create new, innovative tools utilizing EO data that can be used to discover additional primary resources within Europe, thus ensuring a legacy of EU raw material self-sufficiency.
Remote sensing provides a set of low-impact tools to support the entire mining life cycle including mineral exploration, operational, closure and post-closure stages. The use of remote sensing in its full potential has been hampered by lack of understanding of the method applicability in mining industry, need for easy-to-use data processing tools and availability of suitable in situ data. The goal of the project Multi-source and Multi-scale Earth observation and Novel Machine Learning Methods for Mineral Exploration and Mine Site Monitoring (MultiMiner) is to develop novel data processing algorithms for efficient utilization of Earth Observation (EO) technologies for mineral exploration and mine site monitoring. MultiMiner unlocks the potential of EO data, including Copernicus, commercial satellites, airborne, drone and in situ data, to support the entire mining life cycle. The general objectives of MultiMiner are to 1) develop scalable and automated approaches for mineral exploration based on multi-source EO data and sparse in situ data, focused on mineral deposits hosting CRMs across EU, and to 2) leverage novel EO data analysis methods to make the most of scarcely available in situ data for timely mine site monitoring, reducing both disruptions to mining activities and environmental impacts.
The goals are achieved by creating generic but innovative unsupervised and weakly supervised Machine Learning (ML) approaches and tools for multi-scale image processing which augment and potentially substitute in situ data collected on the ground. For exploration purposes, we will develop of a novel mineral mapping algorithm performing automatic spectral feature extraction from a comprehensive mineral spectral library. In addition, a mineral prospectivity wizard facilitating multi-scale mineral mapping and automatic interpretation for non-experts will be constructed.
For environmental monitoring of mining activities, we will develop EO solutions for 1) 3D/4D monitoring of composition and volume of tailings storage facilities, 2) automatic interpretation of European Ground Motion Service products, producing maps for dam stability and open pit stability, 3) algorithms and maps of ground moisture, supporting dam stability as well, 4) monitoring algorithms for water quality and acid mine drainage, 5) algorithms and maps of revegetation status, including vegetation structure and biodiversity and 6) multi-source models for atmospheric and surface dust monitoring. Overarching most of these thematic applications, an ambitious Generic Mine Site Monitoring algorithm (GMSM) will be developed, that can perform these monitoring tasks with advanced deep learning methods making the most of the vast amounts of available EO data and sparser in situ data in respective applications.
The project focuses on new EO based exploration technologies for CRM to increase the probability of finding new sources within EU thereby strengthening the EU autonomy in the area of raw materials. MultiMiner EO-based exploration solutions are fast, economically efficient and have an extremely low environmental impact. They improve safety and are environmentally safe, making them potentially more socially accepted. The project’s solutions for mine site monitoring increase the transparency of mining operations as environmental impacts can be detected as early as possible and digital information of the currently unexploitable raw materials can be stored for future generations. MultiMiner aims to create new, innovative tools utilizing EO data that can be used to discover additional primary resources within Europe, thus ensuring a legacy of EU raw material self-sufficiency.
The project has five work packages (WP) which all have progressed according to the project plan during the first 18 months. Delays in WP execution, especially in field work done in WP4 and related availability issues of validation data, were caused by the departure of one key industrial partner from the consortium and mitigating actions caused by it.
The main achievements of WP1, Project management, were the organization of four project meetings: the kick-off meeting in Finland 1/2023 and the following progress meetings: Austria 5/2023, online 11/2023 and Greece 5/2024. Coordination of project activities and reporting have also been proceeded as planned. MultiMiner has produced three internal progress reports and seven deliverables during the first 18 months of operation.
In WP3, Timely Mine Site Monitoring Methods, a first prototype of GMSM algorithm was developed, as a minimal version addressing two common EO data combinations. In all thematic applications, most available free data has been downloaded, pre-processed and partially analysed for most sites, and all thematic applications have presented preliminary results of their progress. Thematic applications in WP3 include applications related to environment monitoring and to mine safety.
In WP4, Site Demonstrations, the field work collection guidelines for in situ validation data were formulated and the database structure of the key in situ metadata were planned and executed as the MultiMiner in situ metadata database. First metadata are entered into the database. Additionally, a literature survey of the commonly accepted accuracy assessment methods for EO data interpretation was conducted, and finally adopted as an application specific plan how to use them specifically in cases when unsupervised and weakly supervised ML methods are applied and where training data are limited. Field and laboratory work were successfully conducted in all five test sites (Hochfilzen (Austria), Ihalainen (Finland) and Kallyntiri, Chalkidiki and Kirki (Greece)) during summer and fall 2023. The remaining field campaigns will be completed by spring 2025.
The main achievements of WP1, Project management, were the organization of four project meetings: the kick-off meeting in Finland 1/2023 and the following progress meetings: Austria 5/2023, online 11/2023 and Greece 5/2024. Coordination of project activities and reporting have also been proceeded as planned. MultiMiner has produced three internal progress reports and seven deliverables during the first 18 months of operation.
In WP3, Timely Mine Site Monitoring Methods, a first prototype of GMSM algorithm was developed, as a minimal version addressing two common EO data combinations. In all thematic applications, most available free data has been downloaded, pre-processed and partially analysed for most sites, and all thematic applications have presented preliminary results of their progress. Thematic applications in WP3 include applications related to environment monitoring and to mine safety.
In WP4, Site Demonstrations, the field work collection guidelines for in situ validation data were formulated and the database structure of the key in situ metadata were planned and executed as the MultiMiner in situ metadata database. First metadata are entered into the database. Additionally, a literature survey of the commonly accepted accuracy assessment methods for EO data interpretation was conducted, and finally adopted as an application specific plan how to use them specifically in cases when unsupervised and weakly supervised ML methods are applied and where training data are limited. Field and laboratory work were successfully conducted in all five test sites (Hochfilzen (Austria), Ihalainen (Finland) and Kallyntiri, Chalkidiki and Kirki (Greece)) during summer and fall 2023. The remaining field campaigns will be completed by spring 2025.
The work is proceeding in schedule to reach the exploitable results.