Datta et al. (2025) Assessment of climate change impacts on runoff and hydrological drought risk using the VIC-3L model and four-variate D-vine copulas in the Upper Bhima basin, India

Identification

• Journal: Journal of Water and Climate Change
• Year: 2025
• Date: 2025-10-13
• Authors: Rajarshi Datta, M. Janga Reddy, Saswata Nandi
• DOI: 10.2166/wcc.2025.135

Research Groups

• Department of Civil Engineering, Indian Institute of Technology Bombay, Mumbai, India
• Center for Climate Studies, Indian Institute of Technology Bombay, Mumbai, India

Short Summary

This study developed a novel D-vine copula-based framework, integrated with the VIC-3L hydrological model, to assess climate change impacts on runoff and four-variate hydrological drought risk in the Upper Bhima basin, India. It projects significant seasonal shifts in runoff and an increased frequency of both mild and extreme droughts under future climate scenarios.

Objective

• To project future runoff variability in the Upper Bhima basin using the VIC model forced with multiple CMIP6 GCMs and compare changes across future climate scenarios (SSP2-4.5 and SSP5-8.5) relative to the historical period.
• To analyze seasonal and annual trends in precipitation, temperature, and runoff for both historical (1981–2019) and future (2020–2100) scenarios.
• To develop a four-variate joint distribution of hydrological drought characteristics (duration, severity, peak, and interarrival time) utilizing a D-vine copula-based methodology.
• To assess the influence of changing climate on hydrological drought properties and estimate four-variate return periods for mild and extreme drought events, evaluating changes compared to the historical period.

Study Configuration

• Spatial Scale: Upper Bhima basin, India (approximately 46,000 km², 73°–77°E and 16.5°–20°N), discretized into a 79-grid mesh with a 0.25° × 0.25° resolution.
• Temporal Scale: Historical period (1981–2019); Future periods (2020–2100), divided into mid-century (2020–2059) and late-century (2060–2100). Hydrological drought characterized at a 3-month time scale.

Methodology and Data

• Models used:
  • Variable Infiltration Capacity (VIC-3L) hydrological model, calibrated using the Differential Evolution (DE) algorithm.
  • D-vine copulas for constructing four-variate joint probability distributions and estimating joint return periods.
  • Run theory for identifying and characterizing hydrological drought events (duration, severity, peak, interarrival time).
  • Mann–Kendall (MK) test and Sen's slope estimator for trend analysis.
  • Univariate marginal distributions (Gamma, Lognormal, Weibull, Exponential) selected using the Kolmogorov–Smirnov (KS) test.
  • Bivariate copulas (Archimedean: Clayton, Frank, Gumbel, Joe; Elliptical: Student's t, Gaussian; Plackett) selected using Akaike Information Criteria (AIC) and Bayesian Information Criteria (BIC).
• Data sources:
  • Climate Projections: Bias-corrected and spatially downscaled outputs from five CMIP6 General Circulation Models (GCMs) (BCC-CSM2-MR, INM-CM5-0, IPSL-CM6A-LR, MPI-ESM1-2-HR, MIROC6) from NASA Earth Exchange Global Daily Downscaled Projections (NEX-GDDP) at 0.25° daily resolution, under SSP2-4.5 and SSP5-8.5 scenarios.
  • Historical Meteorological Data: India Meteorological Department (IMD) for gridded precipitation (0.25° × 0.25°) and maximum/minimum temperature (1° × 1°). National Centre for Environmental Prediction (NCEP) reanalysis for wind speed (2.5° × 2.5°).
  • Topographical Data: Shuttle Radar Topography Mission (SRTM) Digital Elevation Model (DEM) at 30 m spatial resolution.
  • Soil Data: Food and Agriculture Organization (FAO) global soil database at 30 arc-second spatial resolution.
  • Land Use Data: Advanced Very High-Resolution Radiometer (AVHRR) at 1 km spatial resolution from the University of Maryland.
  • Observed Streamflow: Historical streamflow data from the Takli River monitoring station (Water Resources Information System of India – WRIS-India) for VIC model calibration and validation.

Main Results

• The VIC-3L model demonstrated satisfactory performance with NSE values of 0.80 (calibration) and 0.72 (validation), R² of 0.83 (calibration) and 0.73 (validation), PBIAS of -5.9% (calibration) and 6.3% (validation), and IoA of 0.93 (calibration) and 0.91 (validation).
• Future runoff projections indicate a decline of 15–37.5% in early monsoon (June–August) runoff during the mid-century under both SSP2-4.5 and SSP5-8.5 scenarios, while post-monsoon (September–December) runoff is projected to increase. Late-century SSP5-8.5 shows a net increase in monsoon runoff.
• Trend analysis revealed consistent and significant increases in future temperatures across all scenarios. Annual precipitation showed significant increasing trends under SSP2-4.5 (mid-century) and SSP5-8.5 (mid- and late-century), while winter precipitation displayed declining trends. Annual and spring runoff showed significant upward trends under SSP2-4.5 (mid-century) and SSP5-8.5 (late-century).
• The average number of drought events per grid is projected to increase from 30 (historical) to 48 (mid-century) and 53 (late-century) under future scenarios.
• Spatiotemporal analysis of extreme hydrological droughts projects a decrease in drought duration (43–51%) and severity (22–37%), but an increase in drought peak (35–48%) and a reduction in interarrival time (42–44%) under future scenarios, suggesting more frequent, shorter, and more intense events.
• Kendall's τ correlation coefficients showed very strong associations between duration–severity and severity–peak pairs (0.75–0.98), and moderate to strong correlations for duration–interarrival, duration–peak, peak–interarrival, and severity–interarrival pairs.
• The P-S-D-I (Peak-Severity-Duration-Interarrival time) sequence was identified as the most suitable four-variate D-vine copula structure for the majority of grids.
• Four-variate joint return period analysis (AND condition) for mild droughts projects critically decreasing return periods (increased risk) across approximately 70% of the basin under both SSP2-4.5 and SSP5-8.5 scenarios.
• For extreme droughts, critically decreasing return periods are projected over 17.7% (mid-century) to 19% (late-century) of the basin under SSP2-4.5, and significantly higher at 47% (mid-century) to 38% (late-century) under SSP5-8.5, indicating a substantially greater increase in extreme drought risk under the high-emission scenario.
• Univariate return period assessments consistently yielded lower values compared to the four-variate joint return periods, highlighting the potential for underestimation of drought risk by traditional methods.

Contributions

• Developed a novel, spatially comprehensive grid-based D-vine copula framework for four-variate hydrological drought frequency analysis, specifically using the 'AND' condition, which was previously underexplored for basin-wide hydrological drought risk assessment.
• Integrated the VIC-3L hydrological model with bias-corrected CMIP6 GCM projections to simulate runoff and assess climate change impacts on drought characteristics and risk at a fine spatial resolution across the Upper Bhima basin.
• Provided detailed spatiotemporal projections of future runoff variability and changes in hydrological drought properties (duration, severity, peak, interarrival time) under different Shared Socioeconomic Pathway (SSP) scenarios.
• Demonstrated the significant underestimation of drought risk by traditional univariate return period analyses compared to the more comprehensive four-variate joint return period approach, emphasizing the importance of multivariate dependency structures.

Funding

• Department of Science and Technology, India, DST-GISE Hub (Project Code: RD/0123-GISIR00-007).

Citation

@article{Datta2025Assessment,
  author = {Datta, Rajarshi and Reddy, M. Janga and Nandi, Saswata},
  title = {Assessment of climate change impacts on runoff and hydrological drought risk using the VIC-3L model and four-variate D-vine copulas in the Upper Bhima basin, India},
  journal = {Journal of Water and Climate Change},
  year = {2025},
  doi = {10.2166/wcc.2025.135},
  url = {https://doi.org/10.2166/wcc.2025.135}
}