“…Robust NEE gap-filling approaches are critical for quantifying the annual and interannual variability of carbon budgets (Falge et al, 2001;Irvin et al, 2021;Moffat et al, 2007;Pastorello et al, 2020;Richardson and Hollinger, 2007;Soloway et al, 2017;Wutzler et al, 2018). Previous studies have developed and evaluated a number of NEE gapfilling approaches including non-linear regressions (NLRs), look-up tables (e.g., marginal distribution sampling, MDS), machine learning (ML) algorithms (e.g., artificial neural networks), and process-based models (Falge et al, 2001;Huang and Hsieh, 2020;Moffat et al, 2007;Reichstein et al, 2005;Wutzler et al, 2018). NLR fills NEE gaps based on regression analyses between NEE and meteorological variables such as temperature (e.g., air or soil temperature) and light (e.g., photosynthetically active radiation), whereas MDS is based on look-up tables for similar meteorological conditions (i.e., global radiation, air temperature, and vapor pressure deficit) (Falge et al, 2001;Moffat et al, 2007;Reichstein et al, 2005).…”