Abstract: In this paper, based on the panel data from 17 border cities of poverty-stricken counties in Xinjiang during 2006-2012, the following results by have been got compared with OLS and quantile regression method. It is indicated as follows:(1) With some certain high elevation environments, there exists obvious negative effects on farmers' income and rural poverty, while during the frost-free period, the effects on farmers' income and the rural poverty have not significant difference in proportion of the affected area. Per capital of forest land areas, mineral resources and other nature conditions is of obvious effects on farmers' income and rural poverty, but in the opposite, per capital of forest land area is of positive effect for the farmers' income and rural poverty. Mineral resources has a negative effect to the farmers' income, but a positive effect on rural poverty.(2)Per acre in the amount of water that is available reduce forest area, county road density affects farmers' income but no significant effect on rural poverty, At the same time, annual rainfall, per capital arable land area and the urban area only have significant effects on rural poverty.(3) The farmers' education level and the poverty alleviation funds can't significantly increase farmers' income per capital, but it will slow down the rural poverty. Other social and economic factors on farmers' income and poverty reduction have certain significant effects. The innovation of this paper is to combine poor natural geographical environment with the special social background of the frontier and thus analyze the natural geographical environment of poor effect and the social and economic development of poverty reduction effect under the conditions of different quantile levels.

Key words: poverty, natural geographical environment, Quantile regression, the border poverty counties and urban areas