Satellite-inferred European carbon sink larger than expected

Abstract. Current knowledge about the European terrestrial biospheric carbon sink, from the Atlantic to the Urals, relies upon bottom-up inventory and surface flux inverse model estimates (e.g. 0.27±0.16 GtC a−1 for 2000–2005 (Schulze et al., 2009), 0.17±0.44 GtC a−1 for 2001–2007 (Peters et al., 2010), 0.45±0.40 GtC a−1 for 2010 (Chevallier et al., 2014), 0.40±0.42 GtC a−1 for 2001–2004 (Peylin et al., 2013)). Inverse models assimilate in situ CO2 atmospheric concentrations measured by surface-based air sampling networks. The intrinsic sparseness of these networks is one reason for the relatively large flux uncertainties (Peters et al., 2010; Bruhwiler et al., 2011). Satellite-based CO2 measurements have the potential to reduce these uncertainties (Miller et al., 2007; Chevallier et al., 2007). Global inversion experiments using independent models and independent GOSAT satellite data products consistently derived a considerably larger European sink (1.0–1.3 GtC a−1 for 09/2009–08/2010 (Basu et al., 2013), 1.2–1.8 GtC a−1 in 2010 (Chevallier et al., 2014)). However, these results have been considered unrealistic due to potential retrieval biases and/or transport errors (Chevallier et al., 2014) or have not been discussed at all (Basu et al., 2013; Takagi et al., 2014). Our analysis comprises a regional inversion approach using STILT (Gerbig et al., 2003; Lin et al., 2003) short-range (days) particle dispersion modelling, rendering it insensitive to large-scale retrieval biases and less sensitive to long-range transport errors. We show that the satellite-derived European terrestrial carbon sink is indeed much larger (1.02±0.30 GtC a−1 in 2010) than previously expected. This is qualitatively consistent among an ensemble of five different inversion set-ups and five independent satellite retrievals (BESD (Reuter et al., 2011) 2003–2010, ACOS (O’Dell et al., 2012) 2010, UoL-FP (Cogan et al., 2012) 2010, RemoTeC (Butz et al., 2011) 2010, and NIES (Yoshida et al., 2013) 2010) using data from two different instruments (SCIAMACHY (Bovensmann et al., 1999) and GOSAT (Kuze et al., 2009)). The difference to in situ based inversions (Peylin et al., 2013), whilst large with respect to the mean reported European carbon sink (0.4 GtC a−1 for 2001–2004), is similar in magnitude to the reported uncertainty (0.42 GtC a−1). The highest gain in information is obtained during the growing season when satellite observation conditions are advantageous, a priori uncertainties are largest, and the surface sink maximises; during the dormant season, the results are dominated by the a priori. Our results provide evidence that the current understanding of the European carbon sink has to be revisited.