Mapping Paddy Fields, Rice Growth & Irrigation Performance with Sentinel-1 Radar

Java grows much of Indonesia’s rice, but clouds block optical satellites for most of the growing season. Field surveys can’t cover millions of hectares, and irrigation managers need to know whether water actually reaches every block.

• Radar data: VH time series from ~1,000 Sentinel-1 scenes covering Java. Radar sees through clouds and records rice’s distinctive signature from flooding to harvest.
• Paddy map: phenology features, a neural network with SMOTE class balancing, and a two-year consensus to filter out false detections.
• Growth stage & cropping intensity: time-series models that update each field’s stage every 12-day period and count how many rice cycles it has per year (1×/2×/3×).
• Irrigation performance: a water balance fusing Sentinel-1, Sentinel-2/Landsat, CHIRPS rainfall and evapotranspiration to compute satisfaction (SI), uniformity (CU) and reliability (RI) per tertiary block.

• A 3.01 million ha consensus paddy map, with 2.28 million ha actively cropped in the 2024/25 season.
• Java-wide rice growth-stage maps updated every 12 days, plus cropping-intensity maps.
• In the DI Klambu irrigation area (2023/24 season), 653 tertiary blocks assessed, averaging SI 0.83, CU 0.94 and RI 0.98.
• The full method is packaged as an open tutorial that runs on a laptop without a GPU.