Efficiency of global precipitation datasets in tropical and subtropical catchments revealed by large sampling hydrological modelling

• Published in: Journal of hydrology (Amsterdam) 2024-04, Vol.633, p.131016, Article 131016
• Main Authors: Andrade, João M., Ribeiro Neto, Alfredo, Nóbrega, Rodolfo L.B., Rico-Ramirez, Miguel A., Montenegro, Suzana M.G.L.
• Format: Article
• Language: English
• Subjects: Brazil; Rainfall-runoff modelling; Reanalysis; Remote sensing; Satellite precipitation products
• ISSN: 0022-1694; 1879-2707
• DOI: 10.1016/j.jhydrol.2024.131016

•Five precipitation datasets were evaluated on 714 basins across Brazil.•CHIRPS, MERRA-2 and IMERG excelled in hydrological simulations.•Hydrologic modelling presented KGE > 0.5 for 78 % of the catchments.•The worst hydrological performances were in the most arid hydrographic regions.•Parameter sensitivity varies with catchment climate and aridity. Satellite-based and reanalysis precipitation products are widely adopted as complementary information to in situ measurements for estimating river discharge using hydrological modelling. However, there is still a notable research gap in the literature associated with assessing the accuracy of satellite-based or reanalysis products in different tropical and sub-tropical catchments at large-sampling hydrological modelling with sensitivity analysis. We investigated the accuracy of precipitation, hydrological model performance and parameter sensitivity related to seven precipitation data sets based on satellite and reanalysis products, i.e., CHIRPS, TRMM, GLDAS, IMERG, MERRA-2, PERSIANN-CDR, and ERA5 over 714 contrasting tropical and subtropical catchments located in Brazil. We used the Génie Rural Journalier 4 (GR4J) hydrological model to simulate the hydrological processes of the different catchments with two approaches for the calibration: using measured ground-based precipitation data (approach I) or using each individual satellite/reanalysis precipitation products (approach II) to calibrate the models. The results showed that the precipitation products tend to overestimate precipitation, with the exception of ERA5 and MERRA-2. CHIRPS is the only product that produces unbiased precipitation estimates for most catchments. The model calibration using each precipitation product individually improved the hydrological model performance. CHIRPS, IMERG, and MERRA-2 showed good accuracy in terms of both, precipitation estimation and hydrological simulation performance in the calibration period. In the validation period, the best products in terms of KGE were CHIRPS, IMERG and TRMM (KGE > 0.64). The errors in precipitation products are better compensated via hydrological modelling in wet regions. The model parameter sensitivity varies according to precipitation input, climate, and catchment aridity. Overall, all seven products exhibited their worst hydrological performance in arid regions. This study helps to improve our understanding of the catchment response in tropical and subtropical regions while also providing key