Abstract: Water quality index (WQI) overcomes the shortcomings of traditional water quality evaluation methods that cannot satisfy both qualitative and quantitative evaluation. However, WQI requires so many evaluation indexes that it cannot be widely used. Taking the upper reaches of Qiongjiang River Basin as the representative of the typical basin in the hilly area of central Sichuan Province, the monitoring data of water quality from 2018 to 2019 were analyzed through the multiple linear stepwise regression method. One improved model, the minimum water quality index (WQI_min) model, was established by choosing NH₃-N, COD_Cr, BOD₅ and DO from 25 water quality indexes. WQI_min model had better predictive performance, and its evaluation results were significantly correlated with WQI (R²=0.81, PE=11.9%). According to the results of WQI_min model, the average WQI_min value in the upper reaches of Qiongjiang River Basin were 63.1±17.2, and the water quality grade was "good". On the whole, the WQI_min value of the mainstream increased along the flow direction. When it flowed through Anju District, WQI_min decreased significantly, but it could return to its pre-affected level before entering the city quickly. The WQI_min values from high to low were as follows: Qiongjiang River mainstream > Huilong River > Yufeng River > Panlong River > Shidong river, and the water quality in tributaries was worse than that of main streams. Therefore, the priority of water pollution control in this area should be gradually put on the tributaries.