You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Multi-hazard risk intelligence platform with real-time weather monitoring, ML-powered risk scoring, and personalized emergency guidance for every province in Spain.
7 ML risk models — Flood, wildfire, drought, heatwave, seismic, cold wave, and windstorm risk scored 0–100 for all 52 Spanish provinces
Real-time data fusion — 16 live data sources including AEMET weather alerts, Open-Meteo forecasts, IGN and USGS seismic feeds, NASA FIRMS fire hotspots, Copernicus CAMS/EFAS/EMS, OpenAQ air quality, SAIH river gauges (9 basins), MITECO reservoirs, REE energy grid, and ECMWF seasonal outlooks
Interactive risk map — Province-level risk visualization with MapLibre GL, alert overlays, and seismic activity markers
AI emergency advisor — Context-aware safety guidance powered by OpenAI, tailored to current conditions and location
Community hazard reports — Citizens can submit and view local hazard observations with photo evidence
Model explainability — Per-feature importance breakdown showing why each risk score was computed
Temporal Fusion Transformer forecasting — Deep learning sequence models for multi-day risk prediction
Multi-channel alerts — Web Push notifications, email (Resend), SMS (Twilio), and Telegram bot integration
Drought dashboard — Dedicated monitoring with SPEI index, reservoir levels, and LSTM-based predictions
Property risk assessment — Location-specific risk analysis for real estate and insurance applications
Admin backoffice — Alert management, data records, and system monitoring dashboard
Bilingual — Full English and Spanish support via next-intl
Authentication — NextAuth v5 with JWT, Google OAuth, and GitHub OAuth
Architecture
graph LR
subgraph Weather
A1[AEMET]
A2[Open-Meteo]
A3[ECMWF Seasonal]
end
subgraph Hazards
B1[IGN Seismic]
B2[USGS Earthquake]
B3[NASA FIRMS]
B4[Copernicus EMS]
end
subgraph Environment
C1[Copernicus CAMS]
C2[Copernicus EFAS]
C3[Copernicus Land / NDVI]
C4[OpenAQ]
C5[NASA POWER]
end
subgraph Infrastructure
D1[REE Energy Grid]
D2[MITECO Reservoirs]
D3[SAIH River Basins]
D4[INE Demographics]
end
Weather --> BE[FastAPI Backend]
Hazards --> BE
Environment --> BE
Infrastructure --> BE
BE --> E[Feature Engineering]
E --> F1[XGBoost — Flood]
E --> F2[RF + LightGBM — Wildfire]
E --> F3[SPEI + LSTM — Drought]
E --> F4[XGBoost — Heatwave]
E --> F5[Rule-based — Seismic]
E --> F6[Rule-based — Cold Wave]
E --> F7[Rule-based — Windstorm]
F1 --> G[Composite Risk Engine]
F2 --> G
F3 --> G
F4 --> G
F5 --> G
F6 --> G
F7 --> G
G --> H[Risk Scores 0-100]
H --> I[Next.js Frontend]
BE --> I
Loading
Model Performance
Hazard
Method
Features
Accuracy
F1
AUC-ROC
Flood
XGBoost
23
89%
0.84
0.93
Wildfire
RF + LightGBM
20
91%
0.87
0.95
Drought
SPEI + LSTM
6 (90-day seq)
86%
0.81
0.90
Heatwave
XGBoost + WBGT
18
88%
0.83
0.92
Seismic
Rule-based
8
92%
0.78
—
Cold Wave
Rule-based
14
90%
0.76
—
Windstorm
Rule-based
14
91%
0.79
—
ML Pipeline
Data Ingestion — Weather from AEMET and Open-Meteo, seismic data from IGN and USGS, fire hotspots from NASA FIRMS, air quality from OpenAQ and Copernicus CAMS, flood indicators from Copernicus EFAS and SAIH river gauges, vegetation health from Copernicus Land NDVI, reservoir levels from MITECO, seasonal outlooks from ECMWF, and energy grid state from REE
Feature Engineering — 26+ temporal features from hourly history (precipitation accumulation, consecutive hot/cold/dry days, pressure dynamics, soil moisture trends)
Model Inference — 7 hazard-specific models run independently, each producing a 0-100 risk score
Multi-hazard risk intelligence platform with real-time weather monitoring, ML-powered risk scoring, and personalized emergency guidance for every province in Spain.
