Application of data driven models in estimating daily reference evapotranspiration in a coastal region

An accurate calculation of the amount of water requirements for plants can create a more effective irrigation program. In this study, the daily reference evapotranspiration (ETo) was calculated by FAO-Penman-Monteith method and also estimated by three data-driven based models; M5Rule, support vector regression, K-nearest neighbours and a long-short term memory (LSTM) model based on deep learning. Eight meteorological variables (maximum and minimum daily temperature, maximum and minimum relative humidity, wind speed, sunshine duration, dew point temperature and monthly time index) and 15 different input scenarios were considered for modelling in a coastal agricultural land, namely, Tekirdag, Turkey. The results showed that all the models used presented highly accurate estimations. However, the deep learning based LSTM model yielded the most accurate result with 0.99 as the correlation coefficient and 0.25 as the RMSE. The results concluded that, by using only the maximum temperature or minimum temperature, the amount of ETo can be estimated with a high degree of accuracy without the need for other meteorological variables and physically based equations.