Monthly 1-km rice yields for 2012-2023 in Madagascar, Sri Lanka and the Philippines, estimated uisng satellite data, process model, field observations and machine learning

Iizumi and co-authors developed a method for deriving a monthly 1-km gridded yield product that is theoretically applicable worldwide and workable without national and subnational yield statistics. The method combines objective sources of information—satellite remote sensing, process model simulation and field measurements—via machine learning residual model and does not rely on yield statistics. The method was tested for feasibility in Madagascar, Sri Lanka and the Philippines, where diverse rice cropping systems are in operation, with three satellites (MODIS, VIIRS and GCOM-C). The gridded products derived using the proposed method offer plausible yield estimates, but the consistency with the independent yield data varies depending on the season, satellite used and type of data being compared (crop cuts, household surveys, yield statistics and other gridded yield products). The yield products developed here are unique in that they provide information on season-specific water management conditions and yield responses to changes in nitrogen input and planting dates, offering guidance on adaptation to climate change through agronomic adjustments.