Periodic Reporting for period 2 - TEMA (Trusted Extremely Precise Mapping and Prediction for Emergency Management)
Okres sprawozdawczy: 2024-06-01 do 2025-11-30
TEMA project aims to transform Natural Disaster Management (NDM), by developing and integrating rapid, accurate methods for large extracting data-driven NDM insights. By optimizing decision-making through high-speed analytics, the project delivers significant benefits to society, the environment, and the global economy.
TEMA R&D activities are organized around three major TEMA objective sets: a) OA (WP3) Improve and accelerate extreme data analytics, b) OB (WP4, WP5) Improve and accelerate emergency phenomenon modelling, evolution predictions, simulation and interactive visualization and c) OC (WP2-6) Improve NDM using new digital technologies and extreme data analytics. Work packages WP3-WP5 being the three core TEMA research work-packages, are driven by: a) end-user specifications and TEMA platform design (WP2), b) technology/regulations/ethics monitoring (WP2) and risk assessment/mitigation (WP1). Work package WP6 a) unifies and integrates TEMA technologies developed in WP3-WP5, b) tailors the research outputs towards the end-user requirements (WP2), and c) validates the TEMA platform in 8 TEMA pilot trials. Work package WP7 implements an effective communication of the project objectives, progress and results, ensures the engagement of external stakeholders, builds a TEMA community, enhances collaboration and clustering, and contributes to the exploitation of TEMA research results. Moreover, WP7 also provides standardization recommendations concerning massive extreme data analysis and privacy-by-design principles.
TEMA R&D activities are organized around three major TEMA objective sets: a) OA (WP3) Improve and accelerate extreme data analytics, b) OB (WP4, WP5) Improve and accelerate emergency phenomenon modelling, evolution predictions, simulation and interactive visualization and c) OC (WP2-6) Improve NDM using new digital technologies and extreme data analytics. Work packages WP3-WP5 being the three core TEMA research work-packages, are driven by: a) end-user specifications and TEMA platform design (WP2), b) technology/regulations/ethics monitoring (WP2) and risk assessment/mitigation (WP1). Work package WP6 a) unifies and integrates TEMA technologies developed in WP3-WP5, b) tailors the research outputs towards the end-user requirements (WP2), and c) validates the TEMA platform in 8 TEMA pilot trials. Work package WP7 implements an effective communication of the project objectives, progress and results, ensures the engagement of external stakeholders, builds a TEMA community, enhances collaboration and clustering, and contributes to the exploitation of TEMA research results. Moreover, WP7 also provides standardization recommendations concerning massive extreme data analysis and privacy-by-design principles.
During the M19–M36 reporting period, the TEMA project successfully transitioned from laboratory research to operational field validation. The core R&D work packages (WP3, WP4, and WP5) met all technical objectives and KPIs, with results extensively documented in 104 scientific publications and public deliverables.
Technical Achievements (WP3, WP4, WP5) include:
Extreme Data Analytics (WP3): Developed methods providing high accuracy, trustworthiness, and real-time processing, specifically optimized for low-latency and computational frugality.
Phenomenon Modeling & Response (WP4): Significantly improved simulations for wildfire fronts, smoke dispersion, and flood progression while reducing computational costs. It enhanced situational awareness by fusing 10 distinct data sources and automated response strategies for drone and satellite data acquisition.
Digital Twin & Visualization (WP5): Evolved the TEMA Digital Twin into a high-fidelity, interactive platform featuring 3D semantic mapping and eXtended Reality (XR) interfaces. The SmartDesk was finalized as the primary dashboard for high-pressure decision-making.
A critical milestone was the integration of 22 individual technologies into a unified Cloud/Edge/IoT ecosystem. This was managed via Agile methodologies and intensive hackathons focused on containerization and API compliance.
TEMA implemented the Trial Guided Methodology, a standardized procedure for trial execution and evaluation. A major output is the TEMA Trial Action Plan, proposed as a future standard for European NDM pilot trials. Supported by WP2 (ethics/legal) and WP7 (communication/stakeholder engagement), the project successfully executed four large-scale pilot trials in 2026 across various environments and disaster types. The final year will focus on system robustness and the integration of feedback from the remaining 2026 trials.
Technical Achievements (WP3, WP4, WP5) include:
Extreme Data Analytics (WP3): Developed methods providing high accuracy, trustworthiness, and real-time processing, specifically optimized for low-latency and computational frugality.
Phenomenon Modeling & Response (WP4): Significantly improved simulations for wildfire fronts, smoke dispersion, and flood progression while reducing computational costs. It enhanced situational awareness by fusing 10 distinct data sources and automated response strategies for drone and satellite data acquisition.
Digital Twin & Visualization (WP5): Evolved the TEMA Digital Twin into a high-fidelity, interactive platform featuring 3D semantic mapping and eXtended Reality (XR) interfaces. The SmartDesk was finalized as the primary dashboard for high-pressure decision-making.
A critical milestone was the integration of 22 individual technologies into a unified Cloud/Edge/IoT ecosystem. This was managed via Agile methodologies and intensive hackathons focused on containerization and API compliance.
TEMA implemented the Trial Guided Methodology, a standardized procedure for trial execution and evaluation. A major output is the TEMA Trial Action Plan, proposed as a future standard for European NDM pilot trials. Supported by WP2 (ethics/legal) and WP7 (communication/stakeholder engagement), the project successfully executed four large-scale pilot trials in 2026 across various environments and disaster types. The final year will focus on system robustness and the integration of feedback from the remaining 2026 trials.
The project achieved a major increase in European analytics capacity by delivering explainable, high-speed extreme data analytics and precise natural disaster phenomenon prediction, aiming to reduce emergency response times from hours to minutes.
TEMA increased the trustworthiness of extreme data analysis algorithms through explainable artificial intelligence (XAI) frameworks that generate interpretable local explanations at a factor of 2.5 factor compared to standard neural network inference. In terms of accuracy, the project attained performance gains of over 15% in wildfire segmentation in supervised and unsupervised learning settings, compared to previous state-of-the-art baselines. For flood region segmentation, a novel self-knowledge distillation framework improved accuracy by 4.5% over previous benchmarks. Computational throughput in satellite-data analysis was scaled to process a full Sentinel-2 scene in under 10 seconds, which is substantially faster than standard manual or semi-automated operational methods. Additionally, TEMA demonstrated that federated execution at the edge achieves a 15% reduction in computational latency and at least a 10% reduction in data migration compared to cloud-centralized approaches.
TEMA developed methodologies that significantly increase the accuracy of evolving phenomenon simulations. In operational pilot trials, real-time fire behavior simulations calibrated with field observations matched actual burned perimeters almost perfectly, whereas standard models typically reached only half the actual perimeter. Smoke dispersion models now provide updates in less than one second, and the digital twin environment achieved a 30% reduction in root mean square error compared to existing state-of-the-art solutions while optimizing processing speed from 30 minutes to just 5 minutes. These advancements are integrated into the SmartDesk dashboard and XR Viewer, which handle critical emergency data with sub-second update times and provide more than five interactive tools for content customization, enhancing situational comprehension and reducing cognitive load for operators.
TEMA focused on reducing latency in natural disaster management by automating steps in the satellite-based emergency mapping workflow, reducing the time for area of interest detection to minutes. The use of direct downlink at the DLR receiving station potentially reduces the time between sensing and satellite data availability by a factor of 5 compared to existing ecosystems. Furthermore, TEMA increased situational awareness by jointly fusing at least 10 heterogeneous data modalities, including satellite, drone, sensor networks, and social media, into semantically rich hazard maps and 3D digital twin visualizations.
TEMA increased the trustworthiness of extreme data analysis algorithms through explainable artificial intelligence (XAI) frameworks that generate interpretable local explanations at a factor of 2.5 factor compared to standard neural network inference. In terms of accuracy, the project attained performance gains of over 15% in wildfire segmentation in supervised and unsupervised learning settings, compared to previous state-of-the-art baselines. For flood region segmentation, a novel self-knowledge distillation framework improved accuracy by 4.5% over previous benchmarks. Computational throughput in satellite-data analysis was scaled to process a full Sentinel-2 scene in under 10 seconds, which is substantially faster than standard manual or semi-automated operational methods. Additionally, TEMA demonstrated that federated execution at the edge achieves a 15% reduction in computational latency and at least a 10% reduction in data migration compared to cloud-centralized approaches.
TEMA developed methodologies that significantly increase the accuracy of evolving phenomenon simulations. In operational pilot trials, real-time fire behavior simulations calibrated with field observations matched actual burned perimeters almost perfectly, whereas standard models typically reached only half the actual perimeter. Smoke dispersion models now provide updates in less than one second, and the digital twin environment achieved a 30% reduction in root mean square error compared to existing state-of-the-art solutions while optimizing processing speed from 30 minutes to just 5 minutes. These advancements are integrated into the SmartDesk dashboard and XR Viewer, which handle critical emergency data with sub-second update times and provide more than five interactive tools for content customization, enhancing situational comprehension and reducing cognitive load for operators.
TEMA focused on reducing latency in natural disaster management by automating steps in the satellite-based emergency mapping workflow, reducing the time for area of interest detection to minutes. The use of direct downlink at the DLR receiving station potentially reduces the time between sensing and satellite data availability by a factor of 5 compared to existing ecosystems. Furthermore, TEMA increased situational awareness by jointly fusing at least 10 heterogeneous data modalities, including satellite, drone, sensor networks, and social media, into semantically rich hazard maps and 3D digital twin visualizations.