Evaluation of MODIS, Landsat 8 and Sentinel-2 Data for Accurate Crop Yield Predictions: A Case Study Using STARFM NDVI in Bavaria, Germany

The increasing availability and variety of global satellite products the rapid development new algorithms has provided great potential to generate a level data with different spatial, temporal, spectral resolutions. However, ability these synthetic spatiotemporal datasets accurately map monitor our planet on field or regional scale remains underexplored. This study aimed support future research efforts in estimating crop yields by identifying optimal spatial (10 m, 30 250 m) temporal (8 16 days) resolutions scale. current explored discussed suitability four (Landsat (L)-MOD13Q1 (30 8 Sentinel-2 (S)-MOD13Q1 days)) two real (MOD13Q1 (250 NDVI combined separately widely used growth models (CGMs) (World Food Studies (WOFOST), semi-empiric Light Use Efficiency approach (LUE)) for winter wheat (WW) oil seed rape (OSR) yield forecasts Bavaria (70,550 km2) year 2019. For WW OSR, products’ high resolution resulted higher accuracies using LUE WOFOST. observations (8-day) both S-MOD13Q1 L-MOD13Q1 played significant role measuring OSR. example, L- an R2 = 0.82 0.85, RMSE 5.46 5.01 dt/ha WW, 0.89 0.82, 2.23 2.11 OSR model, respectively. Similarly, 8- 16-day products, simple model (R2 0.77 relative (RRMSE) 8.17%) required fewer input parameters simulate was highly accurate, reliable, more precise than complex WOFOST 0.66 RRMSE 11.35%) parameters. Conclusively, L-MOD13Q1, combination LUE, were prominent predicting products; however, advantageous generating exploring long-term time series due Landsat since 1982, maximum m. In addition, this recommended further use its findings implementing validating regions world.