Assessing the performance of gridded precipitation products – a comparative analysis of the Black Sea and East Africa regions

Abstract

Hydrological and climatological studies require a good quality of precipitation data. It is difficult to assess the spatial and temporal variability of precipitation for a basin without sufficient gauging stations. Recently, several high-resolution public-domain reanalysis and gridded precipitation products (GPPs) have been produced. The performance of four GPPs in estimating the spatio-temporal distribution of precipitation was examined in this study. The chosen regions are the Rize province (Black Sea region) and the Blue Nile basin (East Africa region). Both basins have similar topograpy and high precipitation. In this study, the observed precipitation data were compared with the GPP datasets using statistical metrics and time-series plots. Then, the performance of these products was evaluated by assessing their capacity in estimating river flow with a calibrated physically based hydrological model- SWAT. On top of that, we have assessed the performance of the multi gauge calibration (MGC) technique to improve the SWAT model efficiency. We assessed four calibration methods to improve the model simulation outcome. The temporal and seasonal data performance of the GPPs was, then, assessed before testing the quality of prediction with various statistical analysis metrics. Finally, the climate change impact on the hydroclimatology of both basins was studied using GCM scenario for future climate predictions. Estimating the parameters that represent the various hydrological processes is one of the main issue hydrologists need to solve since it is impossible to measure all the hydrological parameters. We tested four methods where calibration was done using flow data: only from upstream area (US), only from downstream area (DC), using both upstream and downstream areas of the basin (MGC), and using first the upstream then downstream flow data (UCDC). The results showed that model calibration using the MGC and UCDC techniques proved to have improved the model performance, unlike the single gauge calibrations. The reason is the model obtains much wider information on the basin characteristics to adjust its parameters when more than one gauging station is used simultaneously for calibration. Our study has contributed to the validation of gridded precipitation products for the mountainous regions of the Black Sea and East Africa regions which have scarcity in weather gauging station. In addition to statistical and visual map assessment, we adopted a hydrological model (SWAT) parameterized for these basins to make original assessments on the hydrological responses for each GPP dataset precipitation inputs. The resulting graph showed that the annual cyclic behavior of all data is quite consistent with each other, like having a peak on May. The result for the Ikizdere basin indicated that the peak water yield magnitude estimattion in May is significantly greater than the observed water yield for CFSR and MSWEP datasets. However, the water yield value was lower for APHRODITE. Moreover, water yields were highly overestimated during the periods May-December by the MSWEP simulation, which is due to overestimated precipitation in the first four months (January-April) when most of the precipitation falls in the form of snow. The most comparable simulated flow regime with the observation was that of the ECMWF simulation flow. When we come to the Blue Nile basin, it is considered the most important river basin for the Nile River as it generated much of the annual flow. Hence, having a good accuracy of summer precipitation (rainfall) estimate is essential for Blue Nile basin since much of the annual rainfall falls in the highland region during the summer season (Jun – Sep). The MSWEP precipitation dataset has overestimated the peak flows which is obviously caused by the overestimation of rainfall in the summer season (Jun-Sep). The CFSR dataset performance was poor with negative deviations specifically for the wet season of the year (Jun-Sep). It resulted in underestimation of flow, especially for the peak flows which could be explained by the underestimation of precipitation over the study area. Among the GPPs, ECMWF captures the annual cycle of the measured flow cycle with little deviations. Like the Ikizdere basin, it has proved to perform better for the Blue Nile basin, where both the total water yield and surface flows for the wet has been estimated with good accuracy. The results allign with the previous studies outcomes which concluded that the ECMWF weather data could not only be successfully used in place of surface weather observation records, but also improve hydrological modelling performance. For both study areas, the ECMWF gridded precipitation estimates have been proven to be the most comparable with the observed precipitation suggesting that the dataset could be implimented for mountaneous areas of the world with very scarce ground weather observation stations. Finally, the seasonal average simulated flow of Ikizdere basin was forced with the GCM climate change scenarios and compared with the observed flow at basin outlet. There is consist trend in the forecasted flow for the various GCM projections. There is a dipole of climate projection underperformance in the sense of underestimation for winter & autumn seasons, whereas, overestimating for the spring season. For winter (Dec – Feb) and autumn (Sep-Nov), all the climate models forecast decrease in flow for the river, whereas for the spring season (Mar-May) all the GCMs forecasted an increase in flow. Hence, towards the end of the century, the climate projections indicate a decreasing trend for the winter season and an increasing trend for spring season. This finding agrees with the previous study results where the long-term historical precipitation data analysis demonstrated a decreasing trend in winter season for the Black Sea region. The case is difference for summer season where the GCM scenarios didn't agree on the precipitation projection trend. Generally, the impact of climate change on the seasonal precipitation of the Rize province project a decreasing trend for the winter season and increasing trend for spring season towards the end of the century. When we come to the Blue Nile basin, the GCM scenarios did not agree on the future climate projections. An increase in precipitation as well as river flow from Rosieres is projected by some of the emission scenarios (MIMR and INCM3) throughout the year except in March and April, which increases outflow in the river. Contrary to that, the other scenarios (BCM2 and CSMK3) estimate a decrease in precipitation and outflow. When we have a closer look into the emission scenarios, the A1B scenarios generally predict increased river flow for the Blue Nile river, while B1 predicts decrease in flow. A low flow, especially in summer (Jul-Sep) mean significant reduction in the heads of reservoirs in Ethiopia, Sudan and Egypt which will have a severe impact on the livelihood of people.