The Prediction of Wheat Yield in the North China Plain by Coupling Crop Model with Machine Learning Algorithms

The accuracy prediction for the crop yield is conducive to food security in regions and/or nations. To some extent, model yields combining mechanism with statistical regression (SRM) can improve timeliness and robustness of final prediction. In this study, accumulated biomass (AB) simulated by Agricultural Production Systems sIMulator (APSIM) multiple climate indices (e.g., suitability extreme indices) were incorporated into SRM predict wheat North China Plain (NCP). results showed that based on random forest (RF) algorithm outperformed models using other algorithms. at SM (the period from start grain filling milky stage) RF obtain a higher (r = 0.86, RMSE 683 kg ha?1 MAE 498 ha?1). With progression growth, performances improved gradually. FS flowering filling) achieve precision longer lead time, which be viewed as optimum providing decent performance about one month’s time. addition, predicted irrigated sites was than rainfed sites. APSIM-simulated AB had an importance above 30% last three events, including FIF event floral initiation flowering), stage), ranked first model. indices, rank every event, played important role winter NCP seriously affected low temperature events before flowering, high after water stress. We hope used develop adaptation strategies mitigate negative effects change productivity provide data support security.