Kenya Space Agency · Earth observation
Disaster Monitor
Turn satellite and climate data into rangeland decisions.
Run analysis over Kenya's rangelands in four steps, one screen each: choose the scope, pick the model, select the areas, then review and run a request you can trace back to its inputs.
Choose a topic
Each topic answers one question and returns one kind of result.
Flood risk
Which areas are likely to flood, and how badly?
A per-pixel flood susceptibility score with the contributing drivers ranked.
Inputs
RainfallTerrain and slopeDrainage densitySoil moisture
Start analysis
Drought monitoring
Where is drought setting in, and how far has it progressed?
Drought severity classes over time, with the onset date per selected area.
Inputs
Rainfall anomalyLand surface temperatureNDVI anomalyEvapotranspiration
Start analysis
Rangeland dynamics
How is grazing land changing across seasons and years?
Vegetation condition and cover-change trajectories, with degradation flagged.
Inputs
NDVI and EVI seriesLand coverGrazing pressureRainfall seasonality
Start analysis
Food security assessment
Which populations are at risk of food insecurity, and when?
An IPC-style phase estimate per area with the strongest risk drivers named.
Inputs
Crop and pasture conditionMarket pricesLivestock body conditionRainfall forecast
Start analysis
Four steps, one screen each
Each decision gets the screen to itself, and the rail across the top keeps the other three one click away. Nothing is submitted until every one of them is made.
Scope
What is the scope?
Model
Which model should run?
Areas
Which areas?
Review
Review and run
The models
The same three are available for every topic. Pick on the trade-off, not the name.
- Random Forest
- Steady on small or noisy samples and easy to interpret. The safe default.
- XGBoost
- Usually the sharpest accuracy on dense feature stacks, at the cost of tuning.
- Combined model
- Ensembles both and reports their agreement, so disputed areas are visible.