A global carbon assimilation system based on a dual optimization method

H Zheng,X Zheng,S P Zhang,Q Huang,F Jiang,L H Wang,W P Yuan,Z Q Chen,J M Chen,S M Li,T Wang,W X Huang,Y Li

doi:10.5194/bg-12-1131-2015

Abstract

Abstract. Ecological models are effective tools for simulating the distribution of global carbon sources and sinks. However, these models often suffer from substantial biases due to inaccurate simulations of complex ecological processes. We introduce a set of scaling factors (parameters) to an ecological model on the basis of plant functional type (PFT) and latitudes. A global carbon assimilation system (GCAS-DOM) is developed by employing a dual optimization method (DOM) to invert the time-dependent ecological model parameter state and the net carbon flux state simultaneously. We use GCAS-DOM to estimate the global distribution of the CO2 flux on 1° × 1° grid cells for the period from 2001 to 2007. Results show that land and ocean absorb −3.63 ± 0.50 and −1.82 ± 0.16 Pg C yr−1, respectively. North America, Europe and China contribute −0.98 ± 0.15, −0.42 ± 0.08 and −0.20 ± 0.29 Pg C yr−1, respectively. The uncertainties in the flux after optimization by GCAS-DOM have been remarkably reduced by more than 60%. Through parameter optimization, GCAS-DOM can provide improved estimates of the carbon flux for each PFT. Coniferous forest (−0.97 ± 0.27 Pg C yr−1) is the largest contributor to the global carbon sink. Fluxes of once-dominant deciduous forest generated by the Boreal Ecosystems Productivity Simulator (BEPS) are reduced to −0.78 ± 0.23 Pg C yr−1, the third largest carbon sink.

Highlights

The spatiotemporal distribution of carbon sources and sinks has drawn much attention in global carbon cycle research as carbon dioxide is a major greenhouse gas
Different from Lokupitiya et al (2008), which adjusts ecosystem respiration (ER) and gross primary productivity (GPP) using separate scaling factors, only the net ecosystem exchange (NEE) defined as the difference between ER and GPP is scaled. This is because both ER and GPP are much larger than the net ecosystem production (NEP) fluxes by approximately 1 order of magnitude; adjusting their separate influence could lead to spurious variations
We focus on the result of three large regions of North America, Europe and China

Summary

Introduction

The spatiotemporal distribution of carbon sources and sinks has drawn much attention in global carbon cycle research as carbon dioxide is a major greenhouse gas. Atmospheric inversion uses CO2 observations to infer the distribution of the carbon flux from global (Patra et al, 2005; Rödenbeck, 2005; Rayner et al, 2008; Maki et al, 2010) to regional scales (Gerbig et al, 2003; Peylin et al, 2005; Peters et al, 2007; Schuh et al, 2010). In the process of making inferences about flux from ecosystems, we need to exclude the contribution of other CO2 fluxes such as fire and fossil fuel emissions to observed concentrations. They are not perfectly known and and not the target of this study. We included an extra contribution of (0.175 ppm) in the observational error (see Eq 14)

Objectives

Methods

Results

Conclusion