Lei et al. (2026) Process diagnostics of snowmelt runoff in global hydrological and land surface models – Part 2: Are more complex models better?
Identification
- Journal: Hydrology and earth system sciences
- Year: 2026
- Date: 2026-11-18
- Authors: Xiangyong Lei, Haomei Lin, K. Zheng, Peirong Lin
- DOI: 10.5194/hess-30-7121-2026
Research Groups
- Institute of Remote Sensing and Geographic Information Systems, School of Earth and Space Sciences, Peking University, Beijing, China
Short Summary
This study develops a Tree-Based Model Complexity Scoring (TBMCS) method to quantify snow-related process complexity across 13 global hydrological and land surface models. It finds that greater snow-process complexity is positively associated with improved simulation of snowmelt runoff timing, but not runoff magnitude, especially in complex basins and with higher-quality forcing data.
Objective
- To develop a Tree-Based Model Complexity Scoring (TBMCS) method to quantify snow-related process complexity in global hydrological and land surface models and examine its relationship with model performance and robustness for snowmelt runoff volume (Qsum), peak flow (Qmax), and centroid timing (CTQ) across 1455 snowmelt-dominated basins.
Study Configuration
- Spatial Scale: 1455 natural basins in the mid- to high-latitude Northern Hemisphere; models run at a spatial resolution of 0.5°.
- Temporal Scale: Daily time steps; observational runoff records from 1979–2019; snowmelt period defined by annual maximum SWE until SWE decreases below 1 millimeter.
Methodology and Data
- Models used: PCR-GLOBWB, DBH, VIC, MATSIRO, CLM40, LPJML, CWATM, H08, HYDROPY, JULES-W2, MIROC-INTEG-LAND, ORCHIDEE-MICT, WATERGAP2-2E.
- Data sources: ISIMIP2a/3a model outputs, ERA5 reanalysis (for Snow Water Equivalent - SWE), Global Streamflow Characteristics, Hydrometeorology, and Catchment Attributes (GSHA) dataset (for basin attributes), GSWP3 and GSWP3-W5E5 meteorological forcing datasets, observational runoff records.
Main Results
- Models exhibit substantial differences in physical process representation, with the largest divergence in snowmelt process treatments (complexity scores 1.8–6.1), followed by sublimation (0–4.7), interception (0–4.1), and rainfall-snowfall partitioning (0.7–1.9).
- Global hydrological models (GHMs) and dynamic global vegetation models (DGVMs) generally show lower complexity than land surface models (LSMs), primarily due to LSMs' detailed energy-balance formulations for melt.
- Model complexity shows weak or inconsistent associations with performance for total runoff (Qsum) and peak flow (Qmax).
- A positive correlation (r = 0.32) exists between snow-process complexity and model performance for centroid timing (CTQ), which strengthens under high basin complexity (r = 0.35) and becomes statistically significant (r = 0.81, P < 0.05) when models are driven by the GSWP3-W5E5 forcing dataset.
- For model robustness, complexity is negatively correlated with Qsum and Qmax robustness (r = -0.28 for both), but positively correlated with CTQ robustness (r = 0.33).
- Model performance is more influenced by systematic and balanced representations of key processes than by complexity alone; for example, the absence of key physical processes (e.g., canopy interception in H08) severely constrains performance.
Contributions
- Development of the Tree-Based Model Complexity Scoring (TBMCS) method for quantitative and process-traceable assessment of snow-related process complexity.
- First systematic quantification and comparison of physical complexity across 13 global hydrological and land surface models.
- Comprehensive assessment of the relationship between model complexity and performance/robustness for snowmelt runoff characteristics across 1455 global basins.
- Identification of characteristic-dependent patterns, showing that increased snow-process complexity is more consistently associated with improved simulation of runoff timing than runoff magnitude.
- Provides insights into where and to what extent added model complexity yields substantive performance gains, guiding future model development towards process-oriented and uncertainty-aware strategies.
Funding
- National Key Research and Development Program of China (grant no. 2022YFF0801303)
- Beijing Nova Program (grant no. 20230484302)
- Beijing Nova Interdisciplinary Program (grant no. 20240484647)
- National Natural Science Foundation of China (grant no. 42371481)
- Yunnan Provincial Science and Technology Project at Southwest United Graduate School (grant no. 202302AO370012)
Citation
@article{Lei2026Process,
author = {Lei, Xiangyong and Lin, Haomei and Zheng, K. and Lin, Peirong},
title = {Process diagnostics of snowmelt runoff in global hydrological and land surface models – Part 2: Are more complex models better?},
journal = {Hydrology and earth system sciences},
year = {2026},
doi = {10.5194/hess-30-7121-2026},
url = {https://doi.org/10.5194/hess-30-7121-2026}
}
Original Source: https://doi.org/10.5194/hess-30-7121-2026