Abstract: Although the Contingent Valuation Method(CVM) is considered as an only method to evaluate the non-use value of biodiversity and other environmental resources by scholars abroad and at home,there still exist some undefinite factors such as the accuracy and availability of the valuing outcome and various error etc.To improve the accuracy and availability of the outcome of the method,the authors raise another statistical model,i.e.,stratified random sampling model,and compare it with simple random sampling model which is usually used.The case study was chosen in Yaoluoping National Nature Reserve which preserves plants and animals of sub-tropical zone in China.On the basis of the statistic data of questionaire in Jiangsu and Anhui provinces,the two statistical models are applied and compared in this paper.The mean WTP(willingness to pay) of Anhui and Jiangsu procince is 14.2 yuan(RMB) and 42 yuan acoording to the outcome of the questionaire.The mean WTP of the simple random sampling method is 47.8 yuan per person while that of stratified random sampling method is 28.6 yuan per person,so the total WTP of the two methods is 331 335?10⁶ yuan/a and 198 039 million yuan/a respectively.The result indicates that the outcome of the stratified random sampling method is 60% of that of the simple random sampling method,and the former can improve the accuracy and availability of the outcome comparing with that of simple random sampling method to some degree.The authors show that the stratified random sampling method is an alternative approach though it has some deficiencies initiated from CVM method itself.