Quantification and evaluation of input data source induced uncertainty in the InVEST sediment exportmodelling framework for major Indian River basins

Soil erosion and sediment transport pose major challenges for river basin management in India, where
intense monsoon rainfall, diverse physiography, and rapid land-use change generate high and spatially
variable sediment fluxes causing significant challenges like reservoir siltation, soil degradation, and
downstream coastal impacts. However, sediment quantification through modeling at national and basin
scales in India is often constrained by data availability, input data selection and other uncertainties
associated with the choice of empirical options in the models. This study aims to explicitly quantify and
assess input data source–induced uncertainty in the InVEST Sediment Delivery Ratio (SDR) model driven
at 1km resolution for major Indian River basins (viz. Sabarmati, Narmada, Baitarani, and Tapi) for the
period 2005–2019. The adopted multi-input modeling framework utilized several datasets, including
topography from the HydroSHEDS digital elevation model, land use and land cover from HILDA+, rainfall
erosivity (R factor) derived from ERA5 hourly precipitation data using the EI60 formulation,
Furthermore, the rainfall erosivity was computed using five empirical kinetic energy relationships
(Wischmeier & Smith; Brown & Foster; McGregor et al.; Van Dijk et al.; Meshesha et al.) to capture
methodological uncertainty in rainfall intensity representation. Four soil erodibility (K factor)
combinations were generated based on two data sources and two estimation methods: (1) HWSD–EPIC,
(2) HWSD–Nomograph, (3) SoilGrids–EPIC, and (4) SoilGrids–Nomograph . In total, 20 rainfall
erosivity–soil erodibility input combinations were created by systematically varying the erosivity and
erodibility datasets and estimation methods within the InVEST SDR model, using its default
configuration settings. Results indicate strong basin-specific sensitivity to input data selection, with
rainfall erosivity emerging as the dominant control on sediment export, followed by soil erodibility and
then topographic controls (LS factor). Sediment export estimates showed comparatively lower
uncertainty for the Sabarmati and Narmada basins, followed by Baitarani and Tapi. The study highlights
that input data choice represents a major source of uncertainty in large-scale sediment modelling in
India river basins and underscores the need for transparent evaluation of data-driven variability prior to
calibration.