Road network is a critical component of public infrastructure, and the supporting system of social and economic development. Based on a modified kernel density estimate (KDE) algorithm, this study evaluated the road service capacity provided by a road network composed of multi-level roads (i.e. national, provincial, county and rural roads), by taking account of the differences of effect extent and intensity for roads of different levels. Summarized at town scale, the population burden and the annual rural economic income of unit road service capacity were used as the surrogates of social and economic demands for road service. This method was applied to the road network of the Three Parallel River Region, the northwestern Yunnan Province, China to evaluate the development of road network in this region. In results, the total road length of this region in 2005 was 3.70 × 104 km, and the length ratio between national, provincial, county and rural roads was 1:2:8:47. From 1989 to 2005, the regional road service capacity increased by 13.1%, of which the contributions from the national, provincial, county and rural roads were 11.1%, 19.4%, 22.6%, and 67.8%, respectively, revealing the effect of‘All Village Accessible’policy of road development in the mountainous regions in the last decade. The spatial patterns of population burden and economic requirement of unit road service suggested that the areas farther away from the national and provincial roads have higher road development priority (RDP). Based on the modified KDE model and the framework of RDP evaluation, this study provided a useful approach for developing an optimal plan of road development at regional scale.