Evaluation of crop growth simulation by use of satellite rainfall estimates in Lake Victoria basin

Crop growth simulations often rely on rainfall data from in-situ rain gauges, but limited access to quality gauged data in many regions necessitates the use of satellite rainfall estimates (SREs). However, SREs are affected by estimation errors that can propagate into crop growth simulations. Improving the reliability of SREs data sources by reducing the rainfall estimation errors can bring crop simulation results closer to simulations that rely on in-situ rainfall. This study assesses the effectiveness of a weighted ensemble ( WEsc ) of bias and random error corrected SREs in simulating maize biomass using the AquaCrop-OSPy model. Biomass estimates based on WEsc were compared with those simulated using: (i) four uncorrected SREs sources (CHIRPS, CMORPH, MSWEP, and RFE2), (ii) an arithmetic mean-based ensemble of uncorrected SREs data ( AEs ), and (iii) gauged rainfall as reference across six cropping seasons (2012–2017) in the Lake Victoria basin, Kenya. WEsc consistently outperformed individual SREs and AEs , closely matching biomass estimates obtained using gauged rainfall. WEsc also better represented crop failure events and showed potential for biomass estimation in ungauged areas. These results advocate for using ensemble estimates of bias and random error corrected SREs sources in agro-hydrological applications, particularly where rain gauge data is limited. • Weighted ensemble SRE ( WEsc ) outperforms uncorrected SREs in biomass simulation. • Dynamic bias correction windows reduce uncertainty in crop growth estimates. • WEsc achieves 89 % crop growth failure detection with reduced false alarms. • WEsc attenuates rainfall error propagation to biomass compared to individual SREs. • WEsc provides reliable rainfall input for crop growth simulations in ungauged areas.