Modeling of river water temperature using Gene Expression Programming (Case Study: MohammadAbad River in Golestan province)

River water temperature has both economic and ecological significance when considering issues such as water quality and biotic conditions in rivers. This parameter affects directly other water quality parameters and plays a major role in the quality of aquatic life and habitats. Consequently, with wide application of water temperature for conducting environmental impact assessments as well as for effective fisheries management, it is important to understand the thermal behavior of rivers and related heat exchange processes. Numerous deterministic and statistical models have been used for prediction of river water temperature by researchers. These modeling were generally based on the air temperature, yet. However, the river hydraulics and metrological parameters may have their special effects on river water temperature. Furthermore, there are limited researches undertaken by novel and intelligent algorithms. Hence, in this study, gene expression programming has been used for estimation of the water temperature of the MohammadAbad river located in Golestan province. In addition to the air temperature, the river hydraulics and metrological parameters were also accounted for modeling river temperature.

Gene Expression Programming (GEP) is an evolutionary algorithm that uses a population of individuals and selects individuals according to fitness, and introduces genetic variation using one or more genetic operators. For the water temperature modeling, the river hydraulics and meteorological parameters including river flow discharge, flow velocity, air temperature, humidity, wind speed and cloud cover during 7-year statistical period (2006-2012) were considered as input parameters and river water temperature was selected as output parameter.

Based on the comparison of the results of different GEP models with 1 to 6 input variables, it was concluded that the GEP model with 6-parameters has the highest accuracy in terms of the coefficient of determination and the root-mean-square error. These values were obtained 0.92 and 1.8˚C for training data and 0.90 and 2.3˚C for the testing data. The mean absolute errors of this model were obtained as 14.67% and 12.80% for training and testing phases, respectively, while the error of linear regression model was obtained greater than 38%. Results showed that in comparison with the multiple-linear regression model, the GEP model has better performance for river water temperature estimation.

According to the results obtained in this paper, one can use the GEP model for prediction of river water temperature with acceptable accuracy. It is concluded that in addition to the air temperature which has the highest impact on the river temperature, the river flow discharge also has considerable impact.