Assessing parameter transfer for runoff prediction in data-scarce Mid-Hill catchments of Nepal
- 1Department of Civil Engineering, Khwopa College of Engineering, Tribhuvan University, Libali-08, Bhaktapur 44800, Bagmati, Nepal
- 2Corresponding author.
Abstract
Study region: Kathmandu Valley, Nepal: the Bagmati (gauged donor), Bishnumati (gauged proxy) and Manohara (ungauged) river basins. Study focus: Streamflow estimation in ungauged basins is a major challenge for water resource and flood disaster management. This study compared two lumped hydrological models, HBV and HEC-HMS, and tested whether parameters calibrated in a gauged basin can be transferred without adjustment to a gauged proxy basin and to an ungauged basin. Continuous daily simulation, one-at-a-time sensitivity analysis, Taylor diagrams and Generalized Likelihood Uncertainty Estimation (GLUE) were used. New hydrological insights for the region: In the Bagmati basin, HBV outperformed HEC-HMS in validation (NSE 0.71 vs 0.57; R 2 0.79 vs 0.60; peak error 2.1% vs 43.6%) and had the higher NSE in five of seven individual years. HBV's GLUE bounds were narrower in both periods but enclosed fewer observations in calibration (67% vs 82%) and slightly more in validation (70% vs 68%). Direct transfer to the more urbanized Bishnumati basin was unsatisfactory for both models (NSE 0.39 for HBV and 0.44 for HEC-HMS), although HBV reproduced the observed peak more closely (error 10.5% vs 35.3%). In the ungauged Manohara basin, HBV simulations agreed with Catchment Area Ratio (CAR)-derived discharge (NSE 0.63; R 2 0.70), which is a consistency check rather than independent validation. Model structure affected performance in the donor basin, whereas catchment differences together with data limitations may explain the weak transfer to Bishnumati. With one donor and two recipient basins, these findings are case-specific, and local gauging remains necessary.
Related
- Editorial: Environmental management of headwater lakes and catchments
- A stochastic model of continuous hourly areal rainfall series applied to a wide range of French catchments
- The effects of climatic changes on plant physiological and catchment ecohydrological processes in the high-rainfall catchments of the Murray-Darling Basin: A scoping study
- cGRHyMoLAP: A compact continuous state-space formulation of GRHyMoLAP for rainfall–runoff modelling
- An interpretable model for efficient and robust runoff forecasting via causal-linear attention and adaptive KAN decoding
BibTeX
@article{bhattarai2026assessing,
title = {{Assessing parameter transfer for runoff prediction in data-scarce Mid-Hill catchments of Nepal}},
author = {Bhabuk Raj Bhattarai and Prabin Sunuwar and Shrija Guragain and Prayush Kumar Shrestha and Sunil Duwal},
journal = {Progress in Disaster Science},
year = {2026},
month = nov,
doi = {10.1016/j.pdisas.2026.100771},
url = {https://doi.org/10.1016/j.pdisas.2026.100771},
note = {CC BY},
}
RIS
TY - JOUR
TI - Assessing parameter transfer for runoff prediction in data-scarce Mid-Hill catchments of Nepal
AU - Bhabuk Raj Bhattarai
AU - Prabin Sunuwar
AU - Shrija Guragain
AU - Prayush Kumar Shrestha
AU - Sunil Duwal
PY - 2026
DA - 2026/11/05
JO - Progress in Disaster Science
DO - 10.1016/j.pdisas.2026.100771
UR - https://doi.org/10.1016/j.pdisas.2026.100771
AB - Study region: Kathmandu Valley, Nepal: the Bagmati (gauged donor), Bishnumati (gauged proxy) and Manohara (ungauged) river basins. Study focus: Streamflow estimation in ungauged basins is a major challenge for water resource and flood disaster management. This study compared two lumped hydrological models, HBV and HEC-HMS, and tested whether parameters calibrated in a gauged basin can be transferred without adjustment to a gauged proxy basin and to an ungauged basin. Continuous daily simulation, one-at-a-time sensitivity analysis, Taylor diagrams and Generalized Likelihood Uncertainty Estimation (GLUE) were used. New hydrological insights for the region: In the Bagmati basin, HBV outperformed HEC-HMS in validation (NSE 0.71 vs 0.57; R 2 0.79 vs 0.60; peak error 2.1% vs 43.6%) and had the higher NSE in five of seven individual years. HBV's GLUE bounds were narrower in both periods but enclosed fewer observations in calibration (67% vs 82%) and slightly more in validation (70% vs 68%). Direct transfer to the more urbanized Bishnumati basin was unsatisfactory for both models (NSE 0.39 for HBV and 0.44 for HEC-HMS), although HBV reproduced the observed peak more closely (error 10.5% vs 35.3%). In the ungauged Manohara basin, HBV simulations agreed with Catchment Area Ratio (CAR)-derived discharge (NSE 0.63; R 2 0.70), which is a consistency check rather than independent validation. Model structure affected performance in the donor basin, whereas catchment differences together with data limitations may explain the weak transfer to Bishnumati. With one donor and two recipient basins, these findings are case-specific, and local gauging remains necessary.
SN - 2590-0617
ER -