Glycyrrhiza uralensis is an endangered medicinal plant and is mainly distributed in semiarid and arid areas in Northern China The conservation of this species and its communities is important and urgent. In the present work, we examined, by artificial neural network theory and methods, the ecological relationships of G. uralensis communities in Northern China, which is the basis for conservation. Data from 100 samples of 2 x 2 m were collected along a precipitation gradient from east to west in Northern China Species composition data and environmental data were measured and recorded for each sample. Self-organizing feature map (SOFM) is an important and superior network in neural network theory, and SOFM clustering and SOFM ordination was used to analyze ecological relations of these community data. The results showed that there were twelve communities dominated by G. uralensis in Northern China These communities represent almost all community types and distribution of G. uralensis in China They had different characteristics in composition, structure and environment. Precipitation was the key environmental factor affecting G. uralensis populations and communities. Water condition was a limited factor for plant community distribution in semiarid and arid areas in Northern China Topographical variables, such as elevation, slope and slope direction, were also important to the studied communities. Conservation for G. uralensis populations and communities must consider these relations. SOFM clustering and ordination were effective and useful techniques in the study of endangered medicinal plant community and should be applied more frequently.