Watershed Health Prediction based on Surface Water Quality Variables (Case Study: Taleghan Watershed)

Quality of Surface water is one of the important criteria to determine the health of watershed. So its changes determine the state of health and sustainability of environment and human societies. Since in the last decade, due to the use of water resources and its pollution, this vital source has changed, the information of its status in the coming years can play an important role in environmental planning and sustainability of human geography. Taleghan watershed in Alborz province is one of the areas that are affected by these changes. In this research, using the Galinak hydrometric station data and the values of 10 parameters of water quality K, Na, mg, Ca, So₄, Cl, Hco₃, PH, Ec, TDS in the years (1990-2016), the health status of this area was evaluated using gene expression planning model. Data for years (1990-2006), (2007-2014), (2015-2016) were considered as training, test and error data respectively, and at least one year (from September 2016 until September 2017) as predictive data fitted using an algorithm with R² 0.87 and RMSE 3.003. In this study, the data were normalized and actual data. The results showed that by training in the model, the values of surface water quality variables of Taleghan watershed can be predicted with acceptable accuracy by investigating the data pattern. In general, when the data are normalized, the accuracy of the model is relatively 10% lower than accuracy of model with real data.
Extended Abstract
1-Introduction
The agricultural and industrial activities within the catchment area, as well as the disposal of urban sewage and industrial wastewater, affect the quality of water resources and, in most cases, reduce water quality. The set of changes and behavior that humans create in the watershed changes the state of normal life. One of the manifestations of these changes is to reduce or increase the quality of water. In recent years, we have generally encountered a decline in water quality and quantity. One of the criteria to assess the health and sustainability of a watershed is the study of surface water quality changes. Considering this criterion, the process of watershed changes can be determined. Using normalization of data or data in normal mode plays an important role in research results. We should use models that have the highest accuracy with the least computations and the ability to teach other people to use in all watersheds in Iran. In this study, the gene expression algorithm for modeling water quality data and its changes has been investigated. The current health status of the watershed is determined using the reference table. The proposed algorithm predicts the health status of the watershed until the end of September 2018.
2-Materials and Methods
One of the methods for assessing the existing status of watersheds is to review the water quality and its changes. Considering that Taleghan watershed has different land use changes, these changes have reduced the quality of water resources in this watershed. In order to investigate these changes in different years, the gene expression has been used in data modeling and prediction. Normal mode and normalized data usage are two methods to check data on entering the model. First, the variables that had the most effect on the model were determined and then the gene expression algorithm was created. In order to evaluate the generated model, three factors of determination coefficient, mean error of absolute and root mean square error were used.
3-Results and Discussion
Using the obtained results, it was found that during investigated time, the use of normalized data provides different results than normal data. As well as, variables whose trend is close to the linear trend shows higher modeling accuracy. Among the variables studied in the model training section, the highest correlation between natural data is related to the amount of PH and the lowest amount is magnesium. In the test section, electrical conductivity has the highest correlation, while magnesium has the lowest correlation. During the period of this study, the EC and TDS are 483.33 and 271.50, which are categorized as "healthy". Recorded data does not show any better value in this watershed. Water quality values, used from variables related to the watershed potential, show a relatively stable trend over time. The results show that data normalization has reduced the accuracy of the model and the impact rank in the model is different. Environmental data often has a nonlinear process which reduces the modeling accuracy. According to the average coefficient of determination 0.87, it is expected that by the end of September 2018, the health status of the watershed will have been reduced to 27 and fall within the scope of recurrent cancer.
4-Conclusion
This study was carried out using water quality data at Galinak station in Alborz province. The results of this study indicated that normalizing data to reduce the amount of data for increase the speed of their analysis, can affect the true results of modeling. The reference values of variables in each vary watershed according to their potential, and it is not possible to provide the same reference points for all watersheds in Iran. Changes in the natural data process, especially in annual scale, are slow, and it is recommended to pay attention to the causes of these changes. The results of this study showed that data modeling using the gene expression in terms of speed in data processing and accuracy in the results is acceptable and gives accuracy of over 70% in all variables. It is suggested that modeling of water quality data using normalization of numbers and natural conditions in other watersheds should be done in order to evaluate the accuracy of the model in both modes.