Generation of a global, daily, spatiotemporally seamless solar-induced chlorophyll fluorescence dataset from 2018 to 2021

Solar-induced chlorophyll fluorescence (SIF) data with fine spatial and temporal resolutions are valuable for monitoring global terrestrial ecosystems. However, the TROPOspheric Monitoring Instrument (TROPOMI) SIF dataset contains significant gaps, including entire days of missing data, which severely limits its use for detailed vegetation monitoring. To address this issue, we developed a Categorical Boosting reconstructed SIF model using SIF data samples in high-revisit areas, vegetation photosynthesis variables and spatiotemporal features. The model filled localized gaps and full-day missing SIF data, producing a global, daily, spatiotemporally seamless SIF dataset (S2DSIF) from 2018 to 2021. Through validation with training and test datasets (training: correlation coefficient (CC) = 0.885 and root mean square error (RMSE) = 0.114 mW m−2 sr−1 nm−1; testing: CC = 0.854 and RMSE = 0.128 mW m−2 sr−1 nm−1), the model effectively learned the relationships between SIF and key variables. By referring to the in situ SIF data, S2DSIF achieved a CC of 0.791. Compared to other SIF products, S2DSIF demonstrated superior accuracy. Additionally, validation with the in situ gross primary productivity (GPP) data produced a CC of 0.743. Therefore, the S2DSIF product holds great promise for global GPP estimation and global land surface ecological monitoring.