Prediction at Ungauged Catchments through Parameter Optimization and Uncertainty Estimation to Quantify the Regional Water Balance of the Ethiopian Rift Valley Lake Basin

Research output: Contribution to journalResearch articleContributedpeer-review



Quantifying uncertainties in water resource prediction in data-scarce regions is essential for resource development. We use globally available datasets of precipitation and potential evapotranspiration for the regionalization of model parameters in the data-scarce regions of Ethiopia. A regional model was developed based on 14 gauged catchments. Three possible parameter sets were tested for regionalization: (1) the best calibration parameters, (2) the best validation parameter set derived from behavioral parameters during the validation period, and (3) the stable parameter sets. Weighted multiple linear regression was applied by assigning more weight to identifiable parameters, using a novel leave-one-out cross-validation technique for evaluation and uncertainty quantification. The regionalized parameter sets were applied to the remaining 35 ungauged catchments in the Ethiopian Rift Valley Lake Basin (RVLB) to provide regional water balance estimations. The monthly calibration of the gauged catchments resulted in Nash Sutcliffe Efficiencies (NSE) ranging from 0.53 to 0.86. The regionalization approach provides acceptable regional model performances with a median NSE of 0.63. The results showed that, other than the commonly used best-calibrated parameters, the stable parameter sets provide the most robust estimates of regionalized parameters. As this approach is model-independent and the input data used are available globally, it can be applied to any other data-scarce region.


Original languageEnglish
Issue number8
Publication statusPublished - 19 Aug 2022

External IDs

Scopus 85136612905
Mendeley fe38edf9-72cb-35ce-93fe-b1a82739b053
unpaywall 10.3390/hydrology9080150



  • data-scarce region, parameter estimation, uncertainties, ungauged catchment, water balance, weighted regression