Evaluation of Entropy Theory Based on Random Forest in Quality Monitoring of Ground Water Network

The quality monitoring of groundwater networks is of great importance due to its importance in different sectors of agriculture, drinking, industry, etc., and the necessity of its periodic use leads to the recognition of quality changes of water resources in different periods. In this study, the entropy theory based on random forest was used to quality monitoring of electrical conductivity (EC) and total dissolved solids (TDS) in the groundwater of 12 wells in the Tasouj plain located in south of Lake Urmia from 2002 through 2019. In order to investigate the interaction of wells in the aquifer area, the conventional method (multivariate regression) and the random forest algorithm were used. By comparing the performance of the two mentioned models in simulating EC and TDS values in a 12-variable mode, the results showed that the random forest model has a better performance and a lower error rate than the multi-variable regression model. On average, the random forest algorithm reduced the error rate by 40% and 56% in simulation EC and TDS values, respectively, in the studied aquifer. The ranking results of the studied wells showed that the Qara Tape well has the highest rank and the Amestjan well has the least important rank among the studied wells, which indicates the importance of the information extracted from the Qara Tape well. According to the zoning of the information transformation index in the aquifer area, the results showed that there is no limitation in monitoring of EC values in the aquifer, and the scattering of the wells is the best. There is no shortage of wells in terms of exchange of salinity information in the study area. Regarding the TDS values, a lack of wells was observed in the central areas and the eastern and western border areas.