Quantification of uncertainties associated with lumped and spatially explicit representations of small and medium reservoirs in a hydrological model for flood simulations

Small and medium reservoirs (SMRs) play a crucial role in regulating flood flows. However, the uncertainty associated with SMR representation for flood simulations has not been investigated. More importantly, the differentiation of model structure and parameter uncertainty associated with their representation has not been widely studied. Therefore, this research proposes a methodology to quantify the model structure and parameter uncertainties associated with the lumped and spatial explicit representation of SMRs in flood simulations. Four different model configurations were considered – without SMRs (S1), representing both cascade SMRs (CAS) and catchment outlet SMRs (COS) as lumped “surface storage” (S2), representing COS as spatially explicit “reservoirs” (S3), and CAS as “surface storage” and COS as “reservoirs” (S4). The Soil Moisture Accounting (SMA) method in Hydrologic Engineering Center – Hydrologic Modeling System (HEC-HMS) was used to simulate using a set of parameters generated by Latin hypercube sampling (LHC). The variance decomposition technique combined with the bootstrapping method was used to quantify the model structure and parametric uncertainties. The developed methodology was applied to the upstream of the Chembarambakkam reservoir in southern India. The results indicated that lumped representation (S2) contributes to the parametric uncertainty. However, when all four configurations were considered, the proportion of model variance was \(86.86\pm2.37\%\) very much higher than the parametric variance of \(11.88\pm2.27\%\). Further, the representation of COS as “reservoirs” ensured an increase in the p-factor from 0.078 for S1 to 0.449 for S3. Overall, the representation of COS as “reservoirs” is necessary for reducing the uncertainty of hydrological flood simulations.