The Impact of Model Parameter Calibration on Future Extreme Event Predictions in the Ganges Basin

The increasing intensity and frequency of Extreme Hydrological Events (EHEs), such as floods and droughts, underscores the urgent need to understand how the water cycle is changing. To project what lies ahead, it is essential to investigate the past by identifying when, where, and how EHEs occur and how they shape current land–atmosphere interactions. Land surface models (LSMs) are widely used for this purpose, yet their accuracy is often limited by region-specific observations, structural simplifications, and challenges in model calibration. In many data-scarce regions, such as the Ganges River Basin, uncalibrated LSMs are frequently applied to reconstruct past hydrological conditions and to estimate potential future changes. However, the extent to which model calibration improves such projections remains poorly understood, particularly given strong spatial and seasonal variability.

In this study, we investigate how model calibration influences projections of EHEs using the Variable Infiltration Capacity (VIC) model. We compare three VIC model setups: an uncalibrated model to establish the baseline for our investigations; a single-site calibration against monthly in-situ streamflow observations at the basin outlet, Farakka; and a sequential multi-site calibration using available monthly in-situ streamflow observations at 12 stations across the basin to constrain basin-scale water balance, seasonal streamflow patterns, and interannual variability. The three model versions are then forced with precipitation and temperature from multiple Global Climate Models (GCMs) under different Shared Socioeconomic Pathway (SSP) scenarios. Historical simulations are used to define high- and low-flow thresholds and baseline variability, providing a reference for interpreting future projections. The resulting simulated streamflow is analyzed at monthly, seasonal, and annual scales to identify dominant drivers and regional contrasts, providing a basis for interpreting future changes relative to historical conditions.

The findings aim to support the development of hydrological monitoring frameworks with improved representation of regional- to large-scale processes, particularly in data-scarce basins, to strengthen disaster mitigation and climate risk assessment strategies.