Modeling the impact of climate change on wheat yield in Morocco based on stacked ensemble learning

Climate change increases the frequency and intensity of extreme events such as droughts, heat waves, and floods, posing a significant challenge to Morocco’s agriculture and food security. Understanding the future impact of climate on crop yield is crucial for long-term agricultural planning. However, this area has been underexplored due to various challenges, including data constraints. This study aimed to project wheat yield in Morocco at a provincial scale from 2021 to 2040 by using multiple climate model datasets, and advanced Machine Learning (ML) algorithms. An ensemble of five global climate models (MIROC6, CanESM5, IPSL-CM6A-LR, INM-CM5-0, NESM3) was employed to project changes in temperature (Tmax, Tmin) and precipitation (Pr). The climate projections were bias corrected using quantile-quantile approach. Four advanced ML algorithms: Random Forest, XGBoost, LightGBM, and Gradient Boosting Regressor, were utilized to develop a stacked ensemble learning model for wheat yield prediction at provincial scale in Morocco. The stacked ensemble learning model was calibrated and validated using historical wheat yield data. Results show that the stacked ensemble learning approach significantly reduced prediction errors compared to individual models, achieving high coefficient of determination of 0.82 and low root mean square error (RMSE) of 300.51 kg/ha. Wheat yields are projected to decline by an average of 10% by 2040 under the modest shared socioeconomic pathways (SSP2-4.5) scenario while under high emission scenario (SSP5-8.5), yield could decrease by up to 60% across some provinces such as Essaouira, Youssoufia, Ouezzane, Rehamna, and Sidi Kacem. Temperature (Tmax and Tmin) and precipitation (Pr) were identified as the critical climate variables influencing wheat yield, with Tmax being the most impactful. Regional projections revealed that provinces inland and in southern Morocco may experience a significant yield reduction of up to 60%. This study highlights the need for implementing effective climate change mitigation measures to avert food insecurity in Morocco and other northern African countries. The primary findings indicate that climate variables, particularly Tmax, play a crucial role in wheat yield projections, emphasizing the importance of detailed climate data and advanced modeling techniques in agricultural planning.