RIVPACS models were widely used for river health assessment.We used the data of 48 benthic assemblages and environmental samples collected in February and September, 2008 at Lijiang River, Guiling, Guangxi Autonomous Region, and developed, validated and tested a RIVPACS model.From this dataset, 32 samples were indentified as reference, 27 of which were used to calibrate the model, 5 remain reference samples, 10 slightly to moderately disturbed and 6 heavily disturbed samples were used to validate and test the model.Firstly, 27 reference samples were clustered into 3 groups using Bray-Curtis coefficient.Then, 5 environmental variables(mean stream velocity, width/depth, water temperature, substrate size Ⅰ(<2mm) and Ⅱ(2-8mm)) were determined as the best discriminated variables among 3 sample groups through a stepwise discriminant function analysts.The observed(O) and expected(E) biota, and O/E ratio of all 48 samples were obtained by the model, The O/E values of 27 reference samples used in model development had no significant difference with 5 remain reference samples for validation, but had significant difference with 10 slightly to moderately and 6 heavily disturbed samples, and 10 slightly to moderately distuibed samples were alao significantly different from 6 heavily disturbed samples.The validation and test results suggested that our model was robust and the potential for development and application of RIVPACS model and its role in the management and protection of water quality and water resource in China.