| Summary: | Abstract This study aims to evaluate the performance of the Climate Hazards Group Infrared Precipitation with Station observation Version 2 (CHIRPS v2.0), the Precipitation Estimation from Remotely Sensed Information using Artificial Neural Network‐Climate Data Record (PERSIANN‐CDR) and the Climate Research Unit Time Series Version 4 (CRU) products to identify which product delivers reliable rainfall caption over Burundi. The station data with long‐term records have been used as a reference to evaluate the performance of the three products for 34 years ranging from 1983 to 2016. Statistical metrics and precipitation detection capability have been used to measure the accuracy of each product over spatial and time scales. The result analysis carried out that the CHIRPS product has good performance compared to CRU and PERSIANN‐CDR products. It performs well over annual, monthly, and seasonal scales, and it strongly agrees with ground observation over the study domain with the lowest correlation coefficient (CC) = 0.78. The PERSIANN‐CDR product performs poorly over the study domain, though it seems to correlate with in situ rain gauge data in some regions where the CC > 0.7. It highly underestimates the rainfall amount for both short rains and long rains. The CRU shows a good performance in most of the regions with CC ≥ 0.74; however, it slightly overestimates rainfall amount over the less wet area including the north and the east regions and highly underestimates the heavy rain over mountainous region. The evaluation at daily scale shows that both CHIRPS and PERSIANN‐CDR moderately perform in most of the regions except the southern and western parts where CC < 0.3. The two products exhibit a good detection ability of light rainfall (<1 mm/day) but poorly detect heavy rainfall (>20 mm/day). The CHIRPS product can deliver reliable and useful information for monitoring meteorological hazards over Burundi.
|
|---|
