At present

it has become one of the vital research topics to estimate the distribution of urban impervious surface（UIS） precisely and effectively using remote sensing technologies.In this paper

while Xiamen Island was taken as the study area

the selective endmember linear spectral mixture model（LSMM） was applied to model the UIS distribution using China Brazil Earth Resources Satellite-02B（CBERS-02B） CCD data

and its advantage was discussed.Four typical endmembers were selected including high albedo impervious surface（High Imp）

low albedo impervious surface（Low Imp）

high albedo soil（High Soil） and vegetation for this region

and the UIS fraction was estimated from the fractions of High Imp and Low Imp consequently.Validation results indicated that the selective endmember method was superior to the generally fully constrained（GFC） method in view of several error measurements

and the UIS estimation obtained by taking into account the panchromatic band（band5） was more precise compared to which obtained without it.Consequently

at pixel scale the accuracy of the modeled UIS fraction was parallel with others using Landsat images

while unbiased estimation was obtained at land use polygon scale.Accordingly

our experiments suggest that UIS can be accurately estimated by using CBERS-02B data through proper processing and effective approaches

in spite of its imperfections related to radiometric calibration and geographical registration

which need to be improved further.