Abstract: National park is a major institutional innovation to promote the construction of ecological civilization in China. How to scientifically classify types and identify spaces is a fundamental task in the layout and construction of national parks, which is critically needed in practice. Based on the national conditions of China and related international experience, we classified national parks into wilderness oriented, ecological priority, recreation oriented, and heritage oriented types, and constructed a relatively complete national park classification scheme. With Yunnan Province as a case, which has a high degree of natural and human diversity, we established a set of index and zoning rules based on “dual evaluation”. The artificial neural networks were used to establish a land use evolution learning algorithm. The meta-cellular automata incorporating an adaptive inertia mechanism was used for spatio-temporal simulation. Spatial identification of different types of national parks was performed for the whole province under high resolution. The contraction-expansion principle was applied to compare, correct, and optimize the identified areas. A comprehensive plan for the future layout of Yunnan National Park was proposed. The results showed that national parks in Yunnan Province were mainly concentrated in the Sanjiang region and the Hengduan Mountains, the west and southwest Yunnan. Those three types of areas could be used as key areas for future natio-nal park planning and protection. The general and worth popularizing research paradigm for national park typology and spatial identification established here could be served as a reference for national application.

Key words: national park, classification, spatial identification, artificial neural network-cellular automaton, Yunnan Province.