Long-lived and Robust Land-Atmosphere Flux Bias Estimation via SIF and NEE

The strength of sources and sinks of biospheric carbon dioxide represent one of the most critical, but least understood processes in carbon science. Since the 1990's, carbon dioxide inversion models have estimated the magnitude, location, and uncertainty of carbon sources and sinks. These inversions are underconstrained statistical problems that employ aggressive statistical regularizations in both space and time to estimate quantities like net ecosystem exchange (NEE) on weekly timescales over fine spatial scales. We used a new framework that leverages observational constraints on estimation of corrections toward only slowly varying biospheric processes, which control time-averaged sources and sinks. We estimated persistent multiplicative biases to time mean and several seasonal harmonics of gross primary production (GPP) and total respiration (RESP). The method is flexible enough to separately estimate component fluxes using additional observational constraints, including data with a high degree of uncertainty. We tested the capabilities of this method by estimating corrections to simulated component fluxes from the Simple Biosphere Model 4 (SiB4) at 8 diverse eddy-covariance towers by assimilating both Solar Induced Fluorescence (SIF) from the GOME-2 and OCO-2 satellites as a proxy seasonal signal of GPP and direct NEE measurements from the AMERIFLUX eddy-covariance datasets between 2007 to 2015. This methodology was able produce proper seasonality, seasonal onset, annual amplitudes, and interannual variability of both net and component fluxes. Furthermore, the time-filtering methodology also proved to be more robust to uncertainty in the observations when compared to a control inversion that represents current global inversions.