Habitat selection is regarded as a crucial part of wildlife research nowadays. The Chinese water deer (Hydropotes inermis) was rediscovered in Jilin Province recently by us, but it was thought to have disappeared since 1950s. However, both the population and distribution of water deer in Jilin province were very limited. We therefore, firstly carried out field research on the habitat selection of water deer in Baishan Musk Deer Nature Reserve, which aimed to figure out how the deer interact with the environment to meet their requirements for survival and reproduction. Filed study was carried out from May to July 2018 and March to April 2019 and a total of 104 used plots (53 in spring and 51 in summer) and 85 control plots (46 in spring and 39 in summer) were recorded. The Chi-square test was adopted to analyze nonnumeric habitat factors, i.e. the vegetation type, dominant plant species, slope and slope direction. The results indicated that there were significant differences among these factors when the water deer selected habitats and they preferred grassland at sunny and middle slope with the dominant plant was Artemisia carvifolia. While the numeric habitats factors, i.e. the distance to human disturbance, distance to water, herbage coverage, hiding cover, elevation, height of dominant herbage and slope degree were analyzed by using the independent-samples T test and Mann-Whitney U test. The results revealed that habitats with higher hiding cover (30.189±14.609 in spring, 62.745±29.737 in summer), higher height of dominant herbage (87.359±16.190 in spring, 84.510±29.618 in summer), and less steep slope (14.245±3.721 in spring, 13.333±5.260 in summer) were preferentially chosen by the water deer both in spring and summer. Besides, a Resource Selection Function model was applied to study the effects of different habitat factors on the habitat selection of the water deer. Moreover, Spearman Correlation test was employed to analyze the correlation of numeric habitat factors and the highly correlated ones, whose absolute value of correlation coefficients was more than 0.5 would be removed before applied into Resource Selection Function model. Generally, Resource Selection Function model could predict the habitat selection of water deer in Baishan properly for the correctly predicted rate in our study were more than 90% both in spring and summer. The population size and distribution aera of the water deer in Baishan Musk Deer Nature Reserve were pretty limited and the majority of their habitats were closed to farmland. In order to protect the water deer, we should make efforts to protect their suitable habitats, enhance people's awareness of wildlife protection and reduce the impact of anthropogenic disturbance on the deer.