Improving the representation of major Indian crops in the Community Land Model version 5.0 (CLM5) using site-scale crop data

Abstract. Accurate representation of croplands is essential for simulating terrestrial water, energy, and carbon fluxes over India because croplands constitute more than 50 % of the Indian land mass. Spring wheat and rice are the two major crops grown in India, covering more than 80 % of the agricultural land. The Community Land Model version 5 (CLM5) has significant errors in simulating the crop phenology, yield, and growing season lengths due to errors in the parameterizations of the crop module, leading to errors in carbon, water, and energy fluxes over these croplands. Our study aimed to improve the representation of these crops in CLM5. Unfortunately, the crop data necessary to calibrate and evaluate the models over the Indian region is not readily available. In this study, we used a comprehensive spring wheat and rice database that is the first of its kind for India and was created by digitizing historical observations. We used eight spring wheat sites and eight rice sites, and many of the sites have multiple growing seasons, bringing the tally up to nearly 20 growing seasons for each crop. We used this data to calibrate and improve the representation of the sowing dates, growing season, growth parameters, and base temperature in the CLM5 model. The modified CLM5 performed much better than the default model in simulating the crop phenology, yield, carbon, water, and energy fluxes when compared with the site-scale data and remote sensing observations. For instance, Pearson’s r for monthly LAI improved from 0.35 to 0.92, and monthly GPP improved from -0.46 to 0.79 compared to MODIS monthly data. The r values of the monthly sensible and latent heat fluxes improved from 0.76 and 0.52 to 0.9 and 0.88, respectively. Moreover, because of the corrected representation of the growing seasons, the seasonality of the simulated irrigation now matches the observations. This study demonstrates that global land models must use region-specific parameters rather than global parameters for accurately simulating vegetation processes and, eventually, land surface processes. Such improved land models will be a great asset in investigating global and regional-scale land-atmosphere interactions and developing future climate scenarios.