Accuracy assessment of the near-real-time rainfall data from the Integrated Multisatellite Retrievals for GPM (IMERG) in the Roraima State

Favorite this paper

The strong El Niño Southern Oscillation event that occurred in 2015/2016 caused extreme drought in the northern Brazilian Amazon, especially in the Roraima state, increasing fire occurrence. In order to assess the adequacy of the recently released IMERG (Integrated Multi-satellite Retrievals for GPM) precipitation data to be used in fire modelling, we estimate the accuracy of the late-run near-real-time IMERG product compared to the 3B43 product of the Tropical Rainfall Measuring Mission and to rain gauge data in Roraima. The IMERG precipitation data showed high accuracy when compared with rain gauge measurements. During the dry season, monthly precipitation estimated using IMERG showed a difference from -10 mm up to +30 mm compared to 3B43 data over ~75% of the Roraima state. During the rainy season, 49.9% of the grid cells (186) showed from -30 to +30 mm difference between IMERG and 3B43 monthly precipitation and negative differences (IMERG < 3B43) were found in a larger area of the state compared to the dry season. We found a systematic underestimation of precipitation up to 120 mm month-1 in the late-run near-real-time IMERG data compared to 3B43 in the region of Roraima state located south of equator in both dry and rainy seasons. We conclude that the late-run near-real-time IMERG product is suitable for our purpose of modelling fire probability in Roraima, although careful should be taken in interpreting the results from south of the equator. As more IMERG data become available, further studies should investigate whether the patterns we observed are consistent over time.