Abstract: Precipitation has significant implications for the regional hydrologic cycle. Studying the spatial and temporal distribution of precipitation is important to the ecological environment and people's livelihood in northwest arid zone in China. Remote sensing precipitation products have the ability to reflect the spatial and temporal distribution of precipitation, and help with solving the problem of sparse meteorological stations in this region. The TRMM （Tropical Rainfall Measurement Mission） precipitation products as a satellite data are widely used in the humid tropics in recent years, and it is proved that the performance is good. However, the effect of the products in the arid and semi-arid area remains to be unclear. It should do a qualitative assessment before applying the products. This paper chose the mid Tianshan Mountains region as the study area, which locates in the inland arid zone. The quality of TRMM3B42 precipitation products was mainly evaluated for different time scale and different elevation scale based on observed data of 15 stations around the study area. The result showed that: the accuracy of precipitation events estimated by the daily products is low，and the accuracy is better in warm season than that in cold season, also the accuracy is better in small amount than that in large amount; Bias analyses of daily precipitation show that the estimated value tends to be higher than that of observed data on small precipitation events ,and to be lower on large precipitation events in warm season, but it has an opposite situation in cold season that higher on large precipitation and lower on small precipitation. For monthly precipitation assessment, the products perform well [（R2>0.85）] at the two stations （Bayinbuluke and Baluntai） on the southern slope of the mid Tianshan Mountains, it is quite different for the stations on the northern slope. On the northern slope, compared to the performance of the low plain, the monthly precipitation estimations is better in the mountain area; both the [R2] and RMSE changes with the elevation show a parabolic distribution, and have a positive correlation with annual average precipitation; also the annual distribution of the deviation is different in each elevation ranges, large deviation could be found in the spring and the autumn in low altitude area but in the summer in high altitude. In terms of yearly scale evaluation, it is clear that the TRMM estimated precipitation is lower than that of observed data at all stations, but the estimated value still can reflect the tendency of spatial distribution of precipitation，as the elevation increasing the deviation becomes larger, and gets maximum at the altitudes between 2 000 m and 3 000 m. Overall, the TRMM 3B42 precipitation data have the limitation in quality for arid and semi-arid climate environment, but the deviations have certain regularity. Calibration is indispensable and feasible processing of TRMM in the arid region.

Key words: TRMM, arid area, precipitation, Tianshan Mountain