Modeling of soil water content for vegetated surface by artificial neural network and adaptive neuro-fuzzy inference system

Soil water content is one of the critical and dynamic factors for controlling many processes in plant growth and understanding agricultural drought status. It also influences the management of water. Unfortunately, it hasn’t been a routinely measured variable in the world, yet. Therefore, this variable is subject to be estimated using related approaches. In this study, an artificial neural network (ANN), a suitable adaptive neuro-fuzzy inference system (ANFIS) and a multiple linear regression (MLR) model were applied and compared for modeling the variation in the measured soil water content for a vegetated surface by meteorological and plant factors such as air temperature, relative humidity, vapor pressure deficit, precipitation and leaf area index. Measurements were carried out over an irrigated field. The results indicated that the best determination coefficient (r²=0.98) between the measured soil water content and all considered variables was estimated by the ANFIS, whereas weaker relationships were calculated between the same variables by MLR as r²=0.38 and ANN as r²=0.56. Comparisons showed that ANFIS approach had a better modeling potential of the soil water content compared to the MLR and ANN model in the trial period, though weaker relationships in the testing period were found by all approaches.