GeoAI Flood Hazard & Satellite Response Prototype

Map Basemap

Research Alignment

1. Hazard dataset development This demo converts historical flood observations and terrain/rainfall predictors into a wall-to-wall flood-hazard probability layer.

2. Satellite response expansion The same framework can be expanded with Sentinel-1 flood extent, Sentinel-2 damage cues, VIIRS/Black Marble night-time lights, roads, shelters, and rescue-priority indicators.

Note: Current map is an exposure-weighted hazard index, not a full vulnerability-based risk map.

Hazard Dataset Layers

Observed Flood Evidence

GFD historical inundation, 2000–2018

97,749 px

Severe Hazard Alert

P₉₅ · p̂ ≥ 0.9193 · lift 8.30×

99,565 px

Hazard Watch Zone

P₉₀ · p̂ ≥ 0.8877 · planning buffer

99,854 px

Exposure Priority Zone

Top 5% hazard × population index

54,575 px

Opacity 72%

How To Read The Evidence

Red: severe physical hazard

Locations with very high modelled flood probability. Use for embankment, river-corridor, and warning checks.

Orange: exposure priority

Places where flood hazard overlaps population. Use first for evacuation, shelter, and relief logistics screening.

Blue: observed flood evidence

Historical inundation pixels from GFD. Use to cross-check whether hotspot zones match past floods.

Priority rule: inspect orange zones that overlap or touch red hazard corridors first.

Click any point to inspect hazard and exposure evidence

LOCATION EVIDENCE INSPECTOR

Click any point on the map to inspect its mapped hazard, exposure-priority, and historical flood evidence layers.

Expansion Path

Start with hazard data

Use flood probability and historical flood evidence as the core disaster hazard dataset.

Add satellite response evidence

Add Sentinel-1/2, VIIRS night-time lights, roads, shelters, and facility layers to screen response needs.

Validate response indicators

Compare mapped signals with rescue locations, relief camps, field reports, or government records.

Hazard Drivers

Mean |SHAP| — contribution to flood probability

Hazard Dataset Metrics

1.08MBasin Area
(km²)

~500MPeople Screened
in basin

24.8%Pop. in
Priority Zone

399.6Priority density
(ppl/km²)

0.6460Average
Precision

61,703Priority Zone
Area (km²)

Evidence Legend

Observed flood evidence, 2000–2018

Severe hazard alert, P₉₅

Hazard watch zone, P₉₀

Exposure priority zone, top 5%

Research Message:
This prototype shows how hazard datasets can be developed from flood evidence and predictors. The next improvement is to add satellite-derived response indicators such as flooded roads, isolated settlements, shelters, NTL outage/recovery, and rescue-priority zones.

Singh · Avtar · Pramanik · Eregno · Thakur · Calay
Hokkaido Univ. · Poornima Univ. · AIT Thailand · UiT Norway
JST SICORP JPMJSC23A1 · INTPART BRIDGE No. 322325

GeoAI Flood Hazard and Response Prototype

Ganges Basin GeoAI Flood Hazard & Response-Evidence Prototype _{Hazard probability · Historical flood evidence · Exposure-priority screening · Future satellite response indicators}

GeoAI Flood Hazard and Response Prototype

Ganges Basin GeoAI Flood Hazard & Response-Evidence Prototype Hazard probability · Historical flood evidence · Exposure-priority screening · Future satellite response indicators

Ganges Basin GeoAI Flood Hazard & Response-Evidence Prototype _{Hazard probability · Historical flood evidence · Exposure-priority screening · Future satellite response indicators}