Early season prediction of within-field crop yield variability by assimilating CubeSat data into a crop model

Accurate early season predictions of crop yield at the within-field scale can be used to address a range production, management, and precision agricultural challenges. While remote sensing insights has been research goal for many years, it is only recently that observations with required spatio-temporal resolutions, together efficient assimilation methods integrate these into modeling frameworks, have become available advance prediction efforts. Here we explore approach combines daily high-resolution CubeSat imagery APSIM model. The employs train linear regression relates simulated leaf area index (LAI). That relationship then identify optimal date which LAI provides best yield: in this case, approximately 14 weeks prior harvest. Instead applying on satellite coincident, or closest to, date, our method implements particle filter integrates CubeSat-based provide end-of-season (3 m) maps before date. demonstrated rainfed maize field located Nebraska, USA, where suitable collections both in-situ data were assessment. procedure does not require in-field calibrate model, results showing even single step, possible estimates good accuracy up 21 days Yield spatial variability was reproduced reasonably well, strong correlation independently collected measurements (R2 = 0.73 rRMSE 12%). When averaged compared, reduced error from 1 Mg/ha (control case based calibrated model), 0.5 (using alone), 0.2 (results three date). Such capacity spatially explicit considerable potential enhance digital goals improve predictions.