Modeling of daily reference evapotranspiration using deterministic and stochastic methods of Sudan

Plant water consumption is the basic data of an irrigation project and one of the most important elements of irrigation applications. The investment costs of irrigation systems are very high, and it is very important to consider climate parameters for determining the water consumption of plants in the planning of these systems. The development of accurate models to predict reference evapotranspiration (ETo) is necessary for effective water management, especially in arid and semi-arid regions where water resources are limited. In this study, reference plant water consumption (ETo) was calculated by considering the daily climate data of 120 stations in Sudan using the ASCE Penman-Monteith equation. The study was carried out in three stages. In the first stage, estimating of ETo by time series, and in the second stage, maps of ETo for Sudan were obtained by different interpolation techniques. In the third stage, the estimation of ETo was carried out with multiple regression and artificial intelligence applications. Autoregressive Integrated Moving Average (ARIMA) models were used for analyses in 33 provinces of Sudan. In the second method, Ordinary Kriging, Simple Kriging, and Inverse Distance Weight (IDW) methods were used from interpolation methods. According to the results of geostatistical analysis, Gaussian, Exponential, and Spherical are the best semi-variogram models. In the third method, multiple linear regression, multilayer artificial neural networks (MLP), and fuzzy artificial neural network system (ANFIS) were applied. ETo estimation was successfully performed using relative humidity, temperature, wind speed, and solar radiation variables as input data in both artificial intelligence models and traditional multiple regression models, and ET0 was used as output data. Deterministic and stochastic models will serve as useful tools for Sudan's water resource management and irrigation planning.