This study evaluated Tropical Rainfall Measuring Mission (TRMM) Multi-satellite Precipitation Analysis (TMPA) product i.e. TRMM 3B42 data, using data from 52 rain gauge stations around the Meichuan watershed, which is a representative watershed of Poyang Lake basin in China. Both the latest Version 7 (V7) and previous Version 6 (V6) of TRMM 3B42 data were compared and evaluated for a 9-year period covering 2001–2005 and 2007–2010. The evaluations were conducted at different spatial (grid and watershed) and temporal (daily, monthly and annual) scales. For evaluation at grid scale, the Thiessen polygon method was used to transform pointed-based rain gauge data to areal precipitation at the same grid scale (0.25°) as TRMM 3B42 data. The results showed that there was little difference in performances of V6 and V7 TRMM 3B42 products. Overall, both V6 and V7 products slightly overestimated precipitation with a bias of 0.04. At daily scale, both V6 and V7 data were considered to be unreliable with large relative RMSE (135%–199%) at the two spatial scales, and they were deficient in capturing large storms. These results suggest that local calibration with rain gauge data should be conducted before V6 and V7 TRMM 3B42 data are used at daily scale. At monthly and annual scales, V6 and V7 TRMM 3B42 data match the rain gauge data well (R2 = 0.91–0.99, relative RMSE = 4%–23%) at both grid and watershed scale and thus have good potential for hydrological applications.