Wheat water footprint modeling using machine learning models in Fars province

This study was conducted with the aim of estimating and modeling the green and blue water footprint of wheat crop using machine learning models in irrigated lands during (2004-2016). Therefore, using climatic and crop data and the fuzzy cluster method, the irrigated wheat cultivation areas in Fars province were divided into four homogeneous regions. Blue, green and gray water footprints were estimated in each region based on the Hoekstra framework. Then, the water footprint in the homogeneous climate was divided into two categories: training (70%) and testing (30%) and using the neural network model and two kernel such as log logistic and hyperbolic tangent (50 input combinations), random forest model and support vector regression (Sigmoid kernel function) was predicted with climatic and plant variables and the results of the models were compared with error evaluation indices and Taylor diagram. The results showed that the best model for estimating the water footprint of wheat in Fars province is the artificial neural network model with logistic log function with a correlation coefficient of more than 0.72 and an average absolute error of less than 0.48. This model can help improve the decision-making process for water managers and planners in the agricultural sector.