Hydrological responses to various land use, soil, and meteorological inputs in a large-scale river basin in India

Abstract:

Lack of in-situ data for weather inputs, soil and land use in developing countries, urges the need for integrating global datasets into the hydrological models. On that account, it is essential to test the available input datasets of unknown relative quality prior to using the model for any hydrometeorological applications. This study uses the Variable Infiltration Capacity (VIC) model to evaluate the impact of available national (or local) versus global datasets on hydrological processes of Mahanadi river basin in India. Meteorological data includes gauge rainfall and temperature data from Indian Meteorological department (IMD) (spatial resolution, 25 km), reanalysis rainfall and temperature products from ERA5-Land (resolution, ~10 km) and satellite rainfall data from the GPM (spatial resolution, ~10 km). Local and global soil maps are procured from National Bureau of Soil Survey and Land Use Planning (NBSSLUP), (spatial resolution, ~550 metres) and SoilGrids (spatial resolution, ~250 metres) respectively. Land use maps from the National Remote Sensing Centre (NRSC) (spatial resolution, 56 meters) and ESA CCI (spatial resolution, 300 meters) respectively. First, an ensemble of models is obtained from calibration, performed with the local datasets, for the period (1990-2000) within a Monte Carlo framework. Next, 11 experiments combining the land use, soil and meteorological datasets are designed to drive these ensemble VIC models at a common grid resolution of 5km for the period (2014-2016). Modelled discharge of all the experiments is assessed at the sub-catchment level using performance indices (NSE, R, MSE and P-Bias). Furthermore, changes in the water balance components, runoff, baseflow and ET are also analysed. As expected, the results indicate that experiment using ‘all local input datasets’ outperformed all other experiments. However, our work also evaluates the overall gain or loss of model performance with the different combination of local and global datasets in the region. In addition, experiments using ‘all local and GPM rainfall ‘, ‘all local and global soil’, and ‘all local and ERA5-Land temperature’ are also comparable. Experiments using ‘all local and Global LULC’ and ‘all global datasets’ resulted in a poor performance. This indicates that finer resolution global soil, ERA5-Land temperature and GPM rainfall could be the potential alternative to the locally available coarse resolution datasets.