EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-ND 4.0

Benchmarking and Selecting Optimal Hydrological Models for Large‐Sample Applications Considering Complexity and Uncertainty

Nikunj K. Mangukiya, Gopeshwar Sahu, Sagar Gupta, Ashutosh Sharma

Water Resources Research · 2026-07 · 62 (7)

DOI: 10.1029/2026wr044753 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Water Resources Research, online erschienen am 03.07.2026. Verfasst von Nikunj K. Mangukiya und 3 weiteren Autor:innen. Die Arbeit wurde bisher 3-mal zitiert (Crossref) und nennt 113 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0). Gefördert von: Ministry of Earth Sciences, India Meteorological Department, Indian Institute of Technology Roorkee.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Reliable large‐sample hydrological predictions require systematic benchmarking and careful model selection. However, this process is challenging due to structural uncertainty among models, high computational demand, and the climatic and physiographic diversity of a region. This study benchmarked 47 conceptual rainfall–runoff models across 159 watersheds in Peninsular India using the Modular Assessment of Rainfall–Runoff Models Toolbox (MARRMoT) and applied a consistent performance threshold to identify plausible simulations. The benchmarking framework enabled systematic evaluation of model adequacy, complexity, and uncertainty, followed by selection of a compact set of models. Results showed that 141 watersheds met the performance threshold, although model performance varied considerably, with semi‐arid regions posing the greatest challenge and tropical regions exhibiting high equifinality. A compact set of five models (HBV, GR4J, MODHYDROLOG, SMA, and Hillslope) was found sufficient to reproduce reliable streamflow in over 85% of watersheds, while an additional five models extended robust flood prediction to the same coverage. Storage analysis revealed that soil moisture was the dominant process representation, followed by routing and interception storages, and the most reliable models required no more than three storages. Parameter complexity also influenced performance: models with intermediate complexity (10–15 parameters) generally performed best, although some lower‐parameter models, such as GR4J, also achieved strong performance. Model skill was further associated with watershed characteristics, including soil porosity, forest cover, and leaf area index. Overall, the results demonstrate that a small, well‐chosen set of models with appropriate storage structures and moderate complexity can reliably represent hydrological processes across Peninsular India.

Förderung

  • Ministry of Earth Sciences
  • India Meteorological Department
  • Indian Institute of Technology Roorkee
  • Earth System Sciences Organization, Ministry of Earth Sciences

Zitieren

APA

Mangukiya, N. K., Sahu, G., Gupta, S., & Sharma, A. (2026). Benchmarking and Selecting Optimal Hydrological Models for Large‐Sample Applications Considering Complexity and Uncertainty. Water Resources Research, 62(7). https://doi.org/10.1029/2026wr044753

Vancouver

Mangukiya NK, Sahu G, Gupta S, Sharma A. Benchmarking and Selecting Optimal Hydrological Models for Large‐Sample Applications Considering Complexity and Uncertainty. Water Resources Research. 2026;62(7). doi:10.1029/2026wr044753

BibTeX
@article{Mangukiya2026,
  author = {Mangukiya, Nikunj K. and Sahu, Gopeshwar and Gupta, Sagar and Sharma, Ashutosh},
  title = {Benchmarking and Selecting Optimal Hydrological Models for Large‐Sample Applications Considering Complexity and Uncertainty},
  journal = {Water Resources Research},
  year = {2026},
  volume = {62},
  number = {7},
  doi = {10.1029/2026wr044753}
}
Kostenlos starten →