Spatiotemporal variations in terrestrial biospheric CO₂ fluxes of India derived from MODIS, OCO-2 and TROPOMI satellite observations and a diagnostic terrestrial vegetation model

Abstract. Accurate quantification of regional terrestrial fluxes is essential for improving our knowledge of the carbon sequestration potential of ecosystems, ecosystem functioning, and emission reduction demand in the context of climate change mitigation. However, the quantification is challenging owing to methodological and observational constraints, especially for regions with severe gaps in the ground-based observational network, like India. This study examines the potential of recent satellite missions, such as TROPOMI and OCO-2 providing retrievals of Solar-Induced chlorophyll Fluorescence (SIF) to improve terrestrial biosphere CO₂ flux estimates over India. Here, we present high-resolution estimates of Gross Primary Productivity (GPP) and Net Ecosystem Exchange (NEE) over India on a 0.1°×0.1° grid at a temporal resolution of 1 hour from 2012 to 2020. These products can be used for various applications such as those related to the carbon cycle (e.g., inverse modelling of CO₂), benchmarking terrestrial biosphere models over the region, and understanding ecosystem responses to climate change. We follow a satellite-based diagnostic data-driven approach using a biosphere model, namely the Vegetation Photosynthesis and Respiration Model (VPRM) simulating both GPP and NEE, based on light use efficiency and satellite observations of the near-infrared radiance of vegetation (NIRv). We calibrate the standard VPRM GPP estimates using SIF-GPP relationship and investigate the model performance by comparing the simulations with eddy-covariance flux tower measurements. Our best model predictions are with a mean bias error (MBE) = 2.4 µmol m^-2 s^-1, root mean squared error (RMSE) = 3.8 µmol m^-2 s^-1 and squared correlation coefficient (R²) = 0.56 when evaluating with observations at a monthly scale over the period from 2012 to 2018. The observed seasonal anomalies in NEE and GPP range from -4.9 to 8.0 µmol m^-2 s^-1 and -7.0 to 17.0 µmol m^-2 s^-1, respectively, and are well captured by our model. The model simulations are highly correlated with observations during 2018, the only common year when both EC and SIF observations are available, with R² values of 0.68 and 0.74 for NEE and GPP, respectively. Incorporating the SIF signals in the vegetation model improves model performance in capturing the seasonality and magnitudes of GPP, thereby improving the estimates of NEE. We show the influence of soil temperature and soil moisture on ecosystem respiration and refined the VPRM's ecosystem respiration calculation to better constrain the fluxes, resulting in simulations closer to the observations. Ecosystem respiration fluxes are less well constrained than ecosystem productivity fluxes due to limited observations. Based on satellite observations and the refined model, the annual NEE and GPP estimates range from -0.38 Pg C yr^-1 to -0.53 Pg C yr^-1 (land C sink) and 3.39 Pg C yr^-1 to 3.88 Pg C yr^-1, respectively, over India for the years from 2012 to 2020. The biospheric flux distribution over the region is found to be associated with ecosystem heterogeneity, variations in precipitation, and soil characteristics at a regional scale. Overall, our results show that the satellite-based SIF data products can potentially inform ecosystem-scale vegetation responses across biomes in India. Future improvements in the terrestrial biosphere CO₂ flux estimates over India can be attained through the carbon cycle data assimilation with the availability of both flux and mixing ratio observations of CO₂.