Application of wavelet neural network in estimation of average air-temperature

Standard weather station evaluates air-temperature which is a major descriptor of earth environmental condition. Estimation of average daily temperature is one of the main perquisites for agricultural programming and water resource management which is possible by empirical, quasi-empirical and intelligent methods. This study evaluates the application of wavelet neural network (WNN) to estimate the average daily air-temperature in Sari weather station and also compares its efficiency with artificial neural network (ANN). We used thermograph data of Sari weather station for modeling. Relative humidity, maximum temperature, minimum temperature, wind velocity and daily evaporation were considered as network input and air-temperature was considered as network output for the years 2010 to 2020 years. Criteria including correlation coefficient, root mean square error (RMSE), NashSutcliffe (NS) coefficient were used to evaluate and comparison the models efficiency. Results showed that WNN model had better performance than ANN for modeling with the coefficient of determination 0.999, RMSE 0.001 and NS 0.998. In conclusion, results showed reliability of WNN model in estimation of air-temperature.