Assessment of uncertainties in global land cover products for hydro‐climate modeling in India

Abstract

Earth's land cover (LC) has significant influence on land‐atmospheric processes and affects the climate at multiple scales. There are multiple global LC (GLC) data sets which are yet to be evaluated for uncertainties and their propagation into the simulation of land surface fluxes (LSFs) in land surface/climate modeling. The present study assesses the uncertainties in seven GLC products with reference to a regional data set for the simulation of LSFs in India using a macro‐scale land surface model. There is considerable overestimation of the extent of croplands in most of the GLCs. The uncertainties in LCs exert significant bias in the simulation of the LSFs of actual evapotranspiration (ETa), latent heat (LE), and sensible heat (H) fluxes. Uncertainty propagation in LSFs is proportional to the bias in cropping intensity under rainfed condition. The high underrepresentation of cropland area in the UMd data set results in highest bias in LSFs whereas the least cropland bias in Globland30 leads to least bias. Irrigation has higher potential to alter the LSFs than uncertainties related to LC especially in regions with large area under irrigation like India. The changes in LSFs are higher in arid/semiarid regions with medium irrigation intensity than in subhumid regions with high irrigation intensity. This has significant implications for the country's future irrigation expansion plans in the arid/semiarid regions. The study also emphasizes the need for focused efforts to quantify the uncertainties from varying irrigation intensities in the next generation CMIP6 experiments.

Multi-model Hydroclimate Projections for the Alabama-Coosa-Tallapoosa River Basin in the Southeastern United States

“…An ideal but non-pragmatic way to characterize these uncertainties would encompass producing ensemble hydroclimate projections using a complete sample of uncertainty sources. However, given the limited resources, most impact assessment studies can focus only on a subset of these choices, thereby resulting in underestimation of the uncertainties in hydroclimate projections 7,13 .Given the plethora of choices in the above-mentioned modeling framework and their significance in hydroclimatic projections, many studies have investigated the effects of individual sources of uncertainties [14][15][16] , as well as combined uncertainties due to different methodological choices within the modeling framework 7,8,[17][18][19][20][21][22][23] . Several studies at global and regional scales indicate that the uncertainties from climate models are a more important source of uncertainty than other factors such as greenhouse gas emission scenarios and hydrologic model structures 8,12,19,24 .…”

A comparative study of conceptual rainfall-runoff models GR4J, AWBM and Sacramento at catchments in the upper Godavari river basin, India

“…However, in Indian basins, a model that simulates results based on distributed information is practised due to the heterogeneity in catchment topography and land uses (Gosain et al 2006(Gosain et al , 2011Garg et al 2013;Prabhanjan et al 2014;Narsimlu et al 2015). The literature reveals that distributed models such as VIC and SWAT are commonly used for hydrologic studies in the Godavari basin (Gosain et al 2006;Uniyal et al 2015;Hengade and Eldho 2016;Madhusoodhanan et al 2017;Hengade et al 2017). A study conducted in Tekra catchment of Godavari using VIC model received a daily NSE of 0.68 and monthly NSE of about 0.86 (Hengade and Eldho 2016;Hengade et al 2017).…”

“…Most of the river basins in India are heterogeneous in terms of land use, topography, climatology, etc. Therefore, distributed models such as SWAT (Soil Water Assessment Tool), Variable Infiltration Capacity (VIC) model, etc., are being used in these catchments for hydrologic modelling (Gosain et al 2006(Gosain et al , 2011Prabhanjan et al 2014;Narsimlu et al 2015;Hengade et al 2017;Madhusoodhanan et al 2017). However, hydrologic modelling experiments with these models are limited by the lack of ground data for model simulations (Refsgaard 1997).…”

Assessment of Multi-Source Evapotranspiration Products over China Using Eddy Covariance Observations

“…It should be noted, however, that these datasets were produced for specific purposes and applications, including analyses of LUC and vegetation changes and their impacts on the climate, hydrology, and ecosystem, and the developments of various geo-scientific models; thus, obvious discrepancies and even errors in these products have been reported, especially at the regional scale [115][116][117][118][119][120][121][122][123][124][125]. Therefore, without considering the suitability of LUC and NDVI/LAI/VOD products, biases originating from raw data and inconsistencies among the selected products and uncertainties owing to product selection and processing can be of the same magnitude as those from the representation of the processes under investigation [113,121,[126][127][128][129]. For example, Branger et al [126] investigated the impact of different LUC datasets on the long-term water balance of the Yzeron peri-urban catchment of France and stated that most water quantities (including ET) were sensitive to LUC selections.…”