Possibility of global gridded streamflow dataset correction: applications of large-scale watersheds with different climates

Previous studies depicted that the global gridded hydroclimatic products mostly lack precision and are inconsistent throughout the real-world water cycle. The current study evaluates the efficiency of the eight streamflow datasets (FLDAS, GLDAS 2.1, GLDAS 2.0, ERA5, TERRA, ERA5_Land, MERRA2, and GRUN) in two large-scale watersheds with different climate conditions (Karkhe River basin in Iran and Rio Itapecuru River basin in Brazil). To correct the products via different statistical metrics, two tuning procedures (including scale factors and the network-based Muskingum method) have been used. Based on the results showing the paramount impacts on the correction procedures of the products, GRUN, ERA_land, and GLDAS 2 perform the best accuracy in the Karkhe River basin; however, the worst product in the watershed is GLDAS 2.1. In the Rio Itapecuru River basin, MERRA2 and TERRA have the best and worst performance, respectively. In the first watershed, the KGE and TaylorS in GLDAS 2.1 improved by 0.85 and 0.23 in the case of correction using scale factor, and these statistics also significantly increased using coupled scale factor and Muskingum routing methods by 0.84 and 0.21, respectively. In the second watershed, these statistics increased by 9.16 and 0.56 in the worst case using scale factors; these metrics also levelled up by 9.32 and 0.62 via coupled scale factor and Muskingum routing methods. In addition, it appeared that the corrected products could better simulate the streamflow in terms of oscillation time series and the extreme temporal values of the watersheds.