This study analyzes the Tropical Rainfall Measuring Mission (TRMM) Multi-Satellite Precipitation Analysis (TMPA) precipitation products for assessment of meteorological drought. Two versions of the TMPA research datasets (3B42V6 and 3B42V7) and one real-time dataset (3B42RTV7) are considered. The TMPA datasets are evaluated against a merged precipitation product which is estimated by merging four non-TMPA global satellite-gauge based datasets (non-TMPA merged). Comparisons are made over global land areas between 50 degrees S and 50 degrees N at monthly and 0.25 degrees spatial resolution from 2000 to 2009 (ten years). All the TMPA precipitation datasets show similar spatial patterns; however quantitatively they disagree considerably, especially over tropical regions. 3B42V7 and 3B42RTV7 show the lowest and highest differences with the non-TMPA merged product, respectively. The Standardized Precipitation Index (SPI) at various time scales (1 month to 12 months) is calculated for each dataset for detecting drought events, with drought defined as when monthly SPI < - 1.0 and severe drought when monthly SPI < -1.5. The SPI results complement the spatial patterns found in the precipitation statistics. The non-TMPA merged and the 3B42V7 precipitation datasets simultaneously identify months under drought more frequently than any other pair (i.e., non-TMPA merged 3B42V6 and non-TMPA merged 3B42RTV7) of precipitation datasets. We consider four severe drought events: (a) 2007 southeastern US drought, (b) 2003 western European heat wave and drought, (c) 2005 Amazon drought and (d) 2006 Kenyan drought as case studies. All precipitation products are able to identify the drought events in time and space except a few cases. The spatial correlation of drought area is the highest (>0.8) for the 2007 southeastern US drought and the lowest (<0.62) for the 2006 Kenyan drought. For severe drought (SPI < -1.5), all three TMPA products and the non-TMPA merged product show more than 50% area under severe drought for the four drought events with few exceptions. Our results show that major differences among datasets are found over many sparse gauge density regions which suggests that the skill of the datasets primarily depends on the differential performance of the respective processing algorithms in different geographic and climatic regions, density of the underlying rain-gauge station networks and the quality of the input data used from non-gauge data sources. Even though the 3B42V7 product performs the best, the 3B42V6 product also performs reasonably well during our study period and domain. The 3B42RTV7 real-time data perform the worst and are not comparable with the two TMPA research products, due to lack of corrections from gauge observations. Therefore, caution should be applied when using this product for real-time monitoring of the drought conditions. Our evaluation of the TMPA research products indicates that they can provide useful information for drought monitoring and as input to hydrological modeling applications for assessment of land surface conditions. (C) 2014 Elsevier Inc. All rights reserved.
