Estimating the water stress in soil using HYDRUS2D and Adaptive Neuro-Fuzzy Interference System

Most In this research, the ability of HYDRUS2D and ANFIS models for simulating temporal variations of soil water content and soil water balance components under full irrigation and water deficit irrigation with two levels of 75 and 55 percentage in a maize field were compared to determine water stress duration in the growing season. To do so, soil water content was measured using TRIME-FM TDR sensors during two growing seasons for calibrating and validating HYDRUS2D model. Also, soil water content was simulated using ANFIS with different type of membership functions and using independent variables of days after planting, GDD, irrigation depth and water stress level. Comparing root mean square error, mean bias error and model efficiency coefficient indices for simulating soil water content stress period duration, soil water content and soil water balance components demonstrated the possibility of using ANFIS instead of a complicated model such as HYDRUS2D when defining the suitable independent variables. Despite 10 days sooner application of treatments in second growing season, the same water stress duration under DI75 treatment for both growing season (i.e. since 82 DAP till harvest) shows that it is possible to apply treatments either sooner or with higher intensity when applying deficit irrigation. Based on the results, ANFIS model could be used for these purposes.