Estimation of Soil Temperature Based on Meteorological Parameters by the HYBRID INVASIVE Weed Optimization Algorithm Model

Abstract

Soil temperature (ST) is considered as one of the crucial characteristics of soil affecting physical and chemical processes of soil, agricultural products, the optimal time for planting seeds, etc and land surface ecological system. Hence, estimating this parameter can play an important role in agricultural and hydrology engineering. In this study, soil temperature was estimated using multilayer perceptron (MLP) and hybrid models of multilayer perceptron with invasive weed optimization algorithm (MLP-IWO). For training and testing of intelligence models meteorological data such as air temperature, relative moisture, wind speed, sunshine hours with a ten-year statistical period during 2007-2017 related to the meteorological station of Van in Turkey were used. The results showed for nominated scenario that the MLP-IWO performed best with lowest root mean square error (RMSE) at 2.057 °C when compared with original MLP model with RMSE at 2.426 °C, the MLP-IWO model ranked first. With greatly reduced modeling error rate, the MLP-IWO is suggested as superior models in soil temperature modeling and may be extended to other aspect of soil science. Overall, new hybrid MLP-IWO methods showed the potential to estimate soil temperature from freely available meteorological data, which may help reduce costs and labour in the field and this model has provided as a high accurate model for estimation soil temperature.