Abstract. The objective of this study is to assess the impacts of land cover change on
the hydrological responses of the Mahanadi river basin, a large river basin
in India. Commonly, such assessments are accomplished by using distributed
hydrological models in conjunction with different land use scenarios.
However, these models, through their complex interactions among the model
parameters to generate hydrological processes, can introduce significant
uncertainties to the hydrological projections. Therefore, we seek to further
understand the uncertainties associated with model parameterization in those
simulated hydrological responses due to different land cover scenarios. We
performed a sensitivity-guided model calibration of a physically
semi-distributed model, the Variable Infiltration Capacity (VIC) model, within a
Monte Carlo framework to generate behavioural models that can yield
equally good or acceptable model performances for subcatchments of the
Mahanadi river basin. These behavioural models are then used in conjunction
with historical and future land cover scenarios from the recently released
Land-Use Harmonization version 2 (LUH2) dataset to generate hydrological
predictions and related uncertainties from behavioural model
parameterization. The LUH2 dataset indicates a noticeable increase in the
cropland (23.3 % cover) at the expense of forest (22.65 % cover) by the
end of year 2100 compared to the baseline year, 2005. As a response,
simulation results indicate a median percent increase in the extreme flows
(defined as the 95th percentile or higher river flow magnitude) and mean
annual flows in the range of 1.8 % to 11.3 % across the subcatchments. The
direct conversion of forested areas to agriculture (of the order of 30 000 km2) reduces the leaf area index, which subsequently reduces the
evapotranspiration (ET) and increases surface runoff. Further, the range of
behavioural hydrological predictions indicated variation in the magnitudes
of extreme flows simulated for the different land cover scenarios; for
instance, uncertainty in scenario labelled “Far Future” ranges from 17 to 210 m3 s−1 across subcatchments. This study indicates that the
recurrent flood events occurring in the Mahanadi river basin might be
influenced by the changes in land use/land cover (LULC) at the catchment scale and suggests that
model parameterization represents an uncertainty which should be accounted
for in the land use change impact assessment.