Multiple Satellite Data Assimilation in Carbon Cycle Analysis using a Local Ensemble Transform Kalman Filter (LETKF)

Impacts of CO2 concentration data obtained from multiple satellite (GOSAT and OCO-2) measurements on the estimation of global surface CO2 fluxes have been investigated using an ensemble-based four-dimensional data assimilation system (LETKF). An online atmospheric transport model (MJ98-CDTM) is employed in the data assimilation system to optimize surface CO2 fluxes from real observations at spatial and temporal resolutions of 6 days and 2.8°, respectively. The features of OCO-2 data are their larger data number and coverage than that of GOSAT. We have tested 4 types of satellite bias correction experiments (w/o bias correction, monthly mean bias correction, all data bias correction and globally constant bias correction) using independent CO2 concentration analysis (JMA CO2 distribution) in our data assimilation system to avoid inconsistency of satellite data. Our results showed that estimated CO2 concentration and fluxes are significantly sensitive to bias correction scheme. In w/o bias correction case, estimated CO2 fluxes show noisy pattern due to inconsistency between GOSAT and OCO-2 data. In conclusion, suitable satellite data bias correction allows obtaining realistic CO2 concentration and flux field and modifying surface CO2 flux almost entire land surface after satellite bias correction. We could confirm that this satellite bias correction scheme makes it possible to use multiple satellite observation data simultaneously in CO2 data assimilation.