Deciphering the Interplay of Precipitation and NDVI for Future Enhanced Wheat Production Strategies

After conducting an in-depth study on the relationship between NDVI (Normalized Difference Vegetation Index) and precipitation in the Ain Aïcha region in Taounate, several modeling methods were explored to quantify this relationship. Initially, a polynomial regression analysis was performed, revealing a coefficient of determination (R2) of 0.70 and a Root Mean Square Error (RMSE) of 0.032. This indicates a moderate correlation between NDVI and precipitation in the region. Subsequently, an approach based on the Random Forest (RF) algorithm was applied, significantly increasing the performance of the model with an R2 of 0.92 and an RMSE of 0.017. These results reinforce the idea of ​​a significant relationship between NDVI and precipitation. Then, the K-Nearest Neighbors (KNN) Regressor model was used, obtaining an R2 of 0.80 and an RMSE of 0.031. Although slightly lower than the RF algorithm, this result still supports the presence of a substantial correlation between NDVI and precipitation. To determine the optimal neighbors’ number for the K-Nearest Neighbors model, the elbow method was employed, providing a more robust parameter selection. Finally, the Gradient Boosting Machine (GBM) approach was adopted, leading to a remarkable performance with an R2 of 0.99 and an RMSE of 0.001. These results convincingly confirm the strong correlation between NDVI and precipitation in the study region.