Evaluation of extreme rainfall indices from CHIRPS precipitation estimates over the Brazilian Amazonia

• Published in: Atmospheric research Vol. 238; p. 104879
• Main Authors: Cavalcante, Rosane Barbosa Lopes, Ferreira, Douglas Batista da Silva, Pontes, Paulo Rógenes Monteiro, Tedeschi, Renata Gonçalves, da Costa, Cláudia Priscila Wanzeler, de Souza, Everaldo Barreiros
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.07.2020
• Subjects: Amazon; CHIRPS; Extreme rainfall; Rainfall trends; Rainfall trends; CHIRPS; Amazon; Extreme rainfall
• ISSN: 0169-8095; 1873-2895
• DOI: 10.1016/j.atmosres.2020.104879

Since several datasets are available with marked differences, the assessment of precipitation data is a key aspect to support the choice of the most adequate precipitation product for a certain research or operational application. In the present study, we evaluated the use of the daily rainfall dataset from CHIRPS with spatial resolution of 0.05° for different purposes in the states of the Brazilian Legal Amazon. We compared monthly rainfall, annual rainfall indices and their trends calculated using CHIRPS data and rain gauge observations with a point-to-pixel analysis. The use of daily CHIRPS data provided mean monthly rainfall similar to that obtained using data from the rain gauge stations, but CHIRPS data tend to underestimate the values for the rainiest months. The correlation was usually lower in the western Amazon, especially during its rainy season. The same underestimation was observed for extreme rainfall indices. CHIRPS product produces more similar results to rain gauge data for the indices PRCPTOT, nP, and R95pad, while strong underestimate the most extreme rainfall indices (R50mm, Rx1day, Rx5days). For the 45 stations and 15 rainfall indices analysed, 63 significant trends were detected using rain gauge data, of which only 13 were detected using CHIRPS product. Therefore, the use of CHIRPS data does not well represent the trends in rainfall indices. •Use of the rainfall dataset from CHIRPS for different purposes in Brazilian Amazon.•CHIRPS data tend to underestimate the rainfall for the rainiest months.•The correlation was greater for the rainy season in eastern Amazonia.•CHIRPS data underestimate the extreme rainfall indices.