Abstract: Night roosting site is a vital habitat for diurnal birds. Birds can show obvious preference in the selection of night roosting habitat. With the acceleration of urbanization, urban green land has gradually become a kind of habitat shelter for urban birds. How to adapt to the disturbance of urbanization and roost safely at night in the urban green land is a vital problem faced by birds. To explore the night roosting habitat selection of urban birds, we collected data on night roosting trees and 17 ecological factors related to the roosting site of oriental magpie robin (Copsychus saularis) in breeding stage and non-breeding stage, from March of 2014 to February of 2015 in Yizhou City of Guangxi Province. The results showed that oriental magpie robin selected 15 night roosting tree species. There was significant difference in the selection of tree species between breeding and non-breeding period. In both periods, the most frequently used tree species were Osmanthus fragrans, Mangifera indica and Bambusa ventricosa among the five tree species selected by oriental magpie robin. The results of principal component analysis showed that the main factors of night roosting habitat selection in breeding and non-breeding period were the concealment of roosting tree and perch and surrounding safety factors, followed by the other perch factors, water source, and food factors, and the last factor was disturbance. For nine factors, there were significant differences between the two periods, including the height of roosting tree and perch, the coverage of tree and herb, the coverage under roosting perch. Oriental magpie robin tended to select suitable night roosting habitat at different stages of life history, reflecting the high adaptability to urban environment with strong disturbance. In the future configuration and construction of urban green land, it is necessary to increase the arbor species with variable perch, tree height and coverage characteristics, and improve the diversity of shrub and herb, so as to meet the needs of night roosting habitat selection from different bird species.

Key words: moisture content of forest surface soil, TDR-300, microwave remote sensing, Quad-pol SAR, BP neural network model, multivariate linear regression model.