Methodology & Data Sources
Every number in this system can be traced to a source, a method and an observation date.
Processing chain
1. Scheme register
Gazetted command-area boundaries and design records from basin authorities form the spatial backbone.
2. Image acquisition
Sentinel-2 L2A optical scenes (10 m, 5-day revisit) and MODIS/WaPOR products are ingested for each command area.
3. Cloud screening
Scenes above 20% cloud over the command area are excluded; remaining pixels are masked per the scene classification layer.
4. Compositing
Monthly median composites are generated to suppress residual noise and gaps.
5. Classification
Cultivated area is classified from growing-season spectral profiles and separated into irrigated and rainfed using dry-season water consumption.
6. Indicator derivation
Utilisation, cropping intensity, biomass, water productivity and stress indicators are computed from the classified layers and ET products.
7. Validation
Outputs are cross-checked against reported figures and, where available, field observations; divergences are published rather than smoothed.
8. Confidence grading
Each value is graded High, Medium or Low based on observation count, cloud cover and validation agreement.
Confidence grading
Low-confidence values are still published, clearly labelled. Nothing is hidden because it is uncertain.
Data sources
Known limitations
- ▸Small plots below the sensor resolution may be missed or mixed with surrounding land cover.
- ▸Persistent cloud during the wet season reduces optical observation frequency in southern basins.
- ▸Evapotranspiration products are coarse relative to individual command areas and are best read at scheme scale.
- ▸Irrigation is inferred from water consumption patterns, not directly observed; supplementary irrigation in rainy periods is hard to separate.
- ▸Command-area boundaries depend on the accuracy of the scheme register supplied by basin authorities.
- ▸Indicators describe biophysical condition, not institutional performance, cost recovery or governance.

