Remote sensing of dissolved CO2 concentrations in meso-eutrophic lakes using Sentinel-3 imagery

Satellite observation can significantly reduce the uncertainties in CO2 emission estimations compared to insufficient field data. However, big challenges remain in developing remote sensing-based models for mapping concentrations of dissolved carbon dioxide (cCO2) in lakes at regional or global scales. We developed a cCO2 estimation model using Sentinel-3-derived lake environmental variables and near-synchronous field cCO2 data from 16 lakes in the middle and lower reaches of the Yangtze and Huai River (ML_YHR) basins in Eastern China (N = 248). Stepwise quadratic polynomial regressions of several combinations of chlorophyll-a (Chl-a), water temperature (Tw), Secchi disk depth (ZSD), and photosynthetic active radiation (PAR)-related variables were tested and validated to select the best approach. The final model showed high performance in calibration and validation (R2 > 0.72, RMSE < 6.35 μmol L−1, MAPE < 30.31%). The model sensitivity analysis, based on Monte Carlo simulations, showed the model's estimated bias as <25% based on uncertainties of all input variables. Spatial and temporal dynamics of dissolved CO2 concentrations in 113 lakes (≥ 10 km2) in the ML_YHR basins were mapped from 2016 to 2021 using the Sentinel-3 data. The result showed that CO2 concentrations were low in the summer and autumn but high in the winter and spring with dramatic variations (e.g., mean coefficient of variation: 52.59%). The annual mean CO2 concentrations of lakes revealed that about 28% of the lakes acted as weak atmospheric CO2 sinks (14.96 ± 1.13 μmol L−1) while the rest were sources (19.22 ± 2.02 μmol L−1), compared with a mean concentration of CO2 atmospheric equilibrium (16.29 μmol L−1). CO2 concentrations decreased with increasing eutrophication and decreasing lake size (p < 0.05). This study advances current knowledge about CO2 emissions from emerging meso-eutrophic lakes and shows how satellite remote sensing can expand the spatiotemporal coverage of lake CO2.