multivariate linear regression

ET0 plays an essential role in water resources management. In this study, nine temperature-based, ten radiation-based and six mass transfer-based methods at three-time scales (monthly, seasonal, and annual) were evaluated versus the FAO56-PM model, and the best and worst methods were selected from each group for the cold moderate and semi-humid climate of Iran. At annual and seasonal time scales (spring, summer and autumn seasons) Hargreaves-M4 and Trajkovic methods (temperature-based methods), Priestley-Taylor and IRMAK1 equations (radiation-based methods) in winter season and for monthly time scale: Meyer and Rohwer, Hargreaves-M4, Trajkovic, IRMAK1, and Priestley-Taylor methods were shown to have the best performance in comparison to FAO56 PM. These equations are the best ET0 estimators in cold moderate, and semi-humid climate. In this study, two equations with the best performance, namely Hargreaves-M4 and Trajkovic were optimized as linear and power functions. The results showed that the power equation had better performance than the linear equation. Finally, two equations were presented based on the linear multivariate regression to simplify ET0 estimation. The results revealed that those kinds of equations, which used Tmax, Tmin as input parameters have high accuracy in comparison with Hargreaves-M4 and Trajkovic equations.

Salinity (EC) is an important factor, especially in the river engineering and water quality studies. In recent decades, the artificial intelligent systems such as artificial neural networks have many applications in various researches, including water resources management. In this study, water electrical conductivity was estimated using artificial neural networks for Simineh-Rood, West Azerbaijan province, Iran. After that, the result were compared with statistical results using conventional methods such as multivariate linear regression. Hydrogen carbonate, chloride sulfate, and the flow rate in the monthly time scale of the period (1992- 2009) were used as the input for linear regression and EC was considered as the output. Criteria of correlation coefficient, root mean square error and coefficient of Nash Sutcliff were used as evaluating criteria for model performance. The results showed that the artificial neural network model has the highest correlation coefficient (0.967), the lowest root mean square error (0.047ds/m) and the Nash Sutcliff criteria (0.706). Overall, the results showed that the artificial neural network model can be used for estimating minimum and maximum rates of water salinity.

Simulation and evaluation of river sediment is important in water resources management. The common ways of measuring the sediment concentration are generally time consuming and costly. Moreover, they sometimes do not possess sufficient accuracy. In this study, the sediment of the Garm Roud river was estimated using artificial neural network and the results were compared via common statistical procedures e.g., multivariate linear regression. River discharge parameter on a monthly time scale in the period (1990 to 2001) and the sedimentation rate were taken as input and output, respectively. Correlation coefficient, root mean square error and Nash-Sutcliffe coefficient were utilized to evaluate and compare the performance of models. The results showed that both models used to estimate discharge had acceptable accuracy, but the ANN model with the highest correlation (0.894), minimum root mean square error (0.062 ton/day) and Nash Sutcliffe coefficient of 0.756 was prioritized in the verification phase. The results showed that the artificial neural network had a high ability to estimate the minimum and maximum values of sedimentation rate.