Prediction of some chemical properties of rangeland soils under exclosure and grazed conditions using adaptive fuzzy-neural inference system in Ghoshchi rangelands of Urmia

Soil is the most important component of rangeland ecosystems and by preserving it and its characteristics, In the present study, the amount of potassium and phosphorus in the soil of Ghoshchi rangelands of Urmia located in West Azerbaijan province from 2019 to 2021 under the influence of grazing and grazing conditions was investigated. In addition, the development and evaluation of an adaptive fuzzy-neural inference model (ANFIS) was presented in order to predict the amount of potassium and phosphorus in the soil and compare its results with the regression model. The mean squared error (RMSE) and the coefficient of explanation (R2) were used to evaluate the regression and inference models. The results of analysis of variance showed that different years and conditions under confinement and grazing had a significant effect on the amount of potassium and phosphorus in the soil, but their interaction was meaningless. The highest amount of soil potassium is related to the year 2021 and the conditions under grazing. While the highest amount of soil phosphorus was related to 2020. In the phosphorus factor modeling section, the ANFIS model with higher accuracy (R2 = 59.5) and less error (RMSE = 0.087) than the regression model (R2=0.38) with more error (RMSE = 0.089) was able to determine the amount of P to predict. Regarding potassium factor, ANFIS model with higher accuracy (R2 = 0.62 and less error (RMSE = 0.017) than regression model (R2 = 0.42) with more error (RMSE = 0.097) was able to measure soil potassium.