The GEORES WebGIS is an interactive mapping platform that visualizes and analyzes multiple environmental risks affecting urban areas. It combines satellite data, AI modeling, and environmental indicators to assess and predict risks like floods, wildfires, landslides, and connectivity issues.

XAI stands for "eXplainable Artificial Intelligence." It's a set of methods that make AI models more transparent and interpretable. In the GEORES project, XAI shows how different environmental factors contribute to risk predictions through SHAP (SHapley Additive exPlanations) values. This transparency is crucial for decision-makers to understand and trust the model recommendations.

The WebGIS includes four main types of layers:

• Target Layers: Show original categorical risk classes (low/medium/high).
• Predicted Risk Layers: Display model-generated risk predictions and combinations.
• Predictor Layers: Present individual environmental and socio-economic factors used as inputs.
• XAI Layers: Visualize how much ach predictor contributes to the final risk assessment.

Use CTRL+Click on any point to query actual values in all visible layers.

The four-map layout allows you to compare different aspects of the risk assessment simultaneously:

1. Target Maps (original risk classifications)
2. Predicted Risk Maps (AI model outputs)
3. Predictor Maps (input variables)
4. XAI Maps (eXplanation of model decisions)

Synchronization ensures that all maps stay centered on the same location as you navigate, making comparison easier. You can disable synchronization if needed.

The GEORES WebGIS provides several key benefits for urban planning and decision-making:

• Comprehensive multi-risk assessment in a single platform
• Evidence-based predictions to guide resource allocation and preventive measures
• Transparent explanation of risk factors to support policy development
• Visual identification of high-risk areas for prioritized intervention
• Ability to explore how different environmental factors interact to create compound risks

This information aligns with UN Sustainable Development Goal 11 (Sustainable Cities and Communities) and EU disaster resilience frameworks.

The GEORES project incorporates data from multiple sources:

• Earth Observation data: Sentinel-1, Sentinel-2, and COSMO-SkyMed satellite imagery
• Environmental data: Digital Elevation Models, land use maps, vegetation indices, soil data
• Meteorological data: Rainfall patterns, temperature records, climate indicators
• Socio-economic data: Population density, road networks, infrastructure proximity
• Historical data: Previous disaster events, land cover changes over time

These diverse datasets are processed and integrated to create the predictor variables used in the risk assessment models.

The prediction models used in GEORES have been rigorously validated using historical data and ground truth measurements. The accuracy of predictions varies by risk type:

• Wildfire risk models typically achieve 75-85% accuracy
• Flood risk models reach 80-90% accuracy in most scenarios
• Land displacement models have 70-80% accuracy
• Connectivity assessments achieve 75-85% accuracy

It's important to note that these are decision support tools and should be used alongside expert knowledge and field observations for critical decision-making.