Modelling the hydrology of the East African Rift System with PCR-GLOBWB2

Abstract

Water is widely recognized as the most vital natural resource; however, human activities and climate change threaten these freshwater resources. Climate change and subsequent global warming, driven by greenhouse gas emissions, particularly impacts Africa, where rising temperatures and economic dependency on climate-sensitive agriculture increase vulnerability. The East African Rift System (EARS) exemplifies these impacts, with its unique geomorphology influencing lake formations. The EARS’ eastern branch features endorheic amplifier lakes, which are highly sensitive to climatic variations, leading to rapid changes in their surface area and water levels. Observed trends in the water levels of these lakes vary, as exemplified by Lake Awassa (rising) and Lake Manyara (declining), due to different contributing factors documented in the literature. This study focuses on the endorheic lakes in the eastern branch of the EARS. The EARS lakes provide essential ecosystem services, but their water levels are under receivable pressure from surrounding human activities and climate change. Hydrological processes in lake catchments are sensitive to climate shifts, necessitating accurate spatiotemporal water budget information over the entire basin for a holistic view. Therefore, the global hydrological model (GHM) PCR-GLOBWB, developed at Utrecht University, was employed to simulate hydrological processes in the EARS. However, limited in situ data complicates hydrological modelling in East Africa, therefore, the remotely sensed precipitation products W5E5 and CHIRPS were utilized, since they provide a unique opportunity as input for accurate hydrological modelling. The objective of this study was to assess changes in the EARS amplifier lakes by reconstructing and quantifying their sensitive hydrology using two remotely sensed precipitation datasets in PCR-GLOBWB for the period 1981-2019. This study documents and analyses the water balance, lake characteristics (actual evapotranspiration (ETa), discharge and lake water level dynamics), precipitation dynamics and unexplained variances. Simulated results for lake characteristics were compared with the remotely sensed observational datasets GLEAM and DAHITI, as well as limited in situ observations from GRDC. Additionally, the model runs of each precipitation datasets, i.e. W5E5 and CHIRPS, were inter-compared to evaluate the effect of increasing resolution of the input data on the model. Both model runs for the EARS basin were performed at 5-arc min resolution, with both precipitation datasets scaled to match the model’s resolution, since the underlying model structure was kept unchanged. The findings reveal that while the PCR-GLOBWB model generally performs well in quantifying various hydrological output parameters, it fails to accurately capture the actual behaviour of the hydrodynamics of the endorheic lakes in the EARS. The model’s limitations have surfaced due to its incomplete representation of the unique characteristics and responses of these closed systems. Inconsistencies were found between simulated water balance components and subsequent trends (MK test) with reported values in the literature. Large biases, inconsistencies and discrepancies were found between simulations and observations for the studied lake characteristics, i.e., actual evaporation, discharge and lake water levels, as well. This highlights the model's limitations in capturing complex hydro-climatic and anthropogenic factors influencing the hydrodynamics of the endorheic lakes in the EARS. The CHIRPS dataset captures precipitation with greater accuracy due to its higher resolution, consideration of topographic effects, and integration of multiple data sources despite underestimating overall precipitation amounts compared to W5E5. The implementation of higher resolution CHIRPS precipitation product showed slight improvem