About Workshop
Aim
What Participants Will Learn
-
Introduce core concepts of AI-enabled remote sensing for real-time monitoring.
-
Demonstrate ingestion & preprocessing of satellite, drone, and sensor data.
-
Apply change detection and early-warning models for deforestation and wildfires.
-
Fuse multi-resolution data (UAV + EO + IoT) for robust signal detection.
-
Estimate air-quality proxies (e.g., AOD) and vegetation carbon stocks.
-
Evaluate model performance (IoU/F1/PR) and calibrate alert thresholds.
-
Explore the Silvanet & Silvaguard case to derive deployable design patterns.
-
Address data gaps, accuracy limits, and scaling in low-resource contexts.
-
Prototype a minimal alerting workflow and web map/dashboard for decision support.
Structure
📅 Day 1 – Platforms, Sensors & Mission Design
- Scope & requirements: deforestation, wildfire, air quality (latency, resolution, revisit)
- Platforms/payloads: UAS (RGB/TIR), public satellites (Sentinel-1/2, Landsat, MODIS/VIIRS), ground AQ (PM₂.₅/NO₂/O₃)
- Data plumbing: orthorectification, tiling, STAC, spatiotemporal indexing, cloud/gap handling
- Case study overview: Silvanet & Silvaguard (Germany) — early warning, integration, limitations
- Governance/ethics: privacy, community data rights, responsible alerting
📅 Day 2 – Analytics & Fusion
- Wildfire: TIR/VIIRS anomaly flags, smoke segmentation, change detection, alert thresholds
- Deforestation: time-series change (BFAST/Delta), semantic segmentation, accuracy assessment
- Air quality: EO AOD ↔ ground fusion, bias correction, nowcasting under missing data
- Carbon/emissions: multispectral+SAR biomass, FRP→emissions, uncertainty bands
- Deployment at scale: robustness, drift monitoring, human-on-the-loop, low-bandwidth constraints
📅 Day 3 – Hands-On: End-to-End AOI Build (single lab)
- Create STAC-indexed AOI data lake (Sentinel-2 + VIIRS + one drone scene)
- Run pipeline: cloud mask → wildfire/smoke flags → forest-loss polygons (with confidence)
- Fuse EO + ground AQ to produce daily PM map (bias-corrected)
- Estimate stand-level carbon with basic uncertainty
- Publish a lightweight dashboard (alerts, loss, AQ index, carbon snapshot)
Important Dates
Registration Ends
Workshop Dates
What You Will Gain
Outcomes
-
Build a real-time EO pipeline (satellite + drones + sensors).
-
Train models for deforestation change & early wildfire signals.
-
Fuse multi-resolution data; detect smoke/thermal anomalies.
-
Estimate air-quality (AOD) and vegetation carbon stocks.
-
Evaluate (IoU/F1), tune thresholds to cut false alarms.
-
Deliver: alerting workflow, mini dashboard, and a one-page rollout for low-resource settings.
Who Should Attend
-
Environmental & forestry professionals, wildfire/disaster management teams
-
Remote sensing & GIS analysts (QGIS/ArcGIS/Google Earth Engine users)
-
Data scientists & ML engineers working with geospatial data
-
Air-quality, climate, and sustainability researchers/practitioners
-
UAV/drone operators and Earth-observation startups
-
Government agencies, utilities, and NGOs involved in monitoring & response
-
Graduate students and faculty in Environmental Science, Geography, CS/AI