Constructing a carbon cycle analysis system with the local ensemble transform Kalman filter and online transport model

In the current carbon cycle analysis, inverse modeling plays an important role. However, it requires more computational resources when we deal with more flux regions and more observations. The local ensemble transform Kalman filter (LETKF) is expected to reduce such problems. We constructed a carbon cycle analysis system with the LETKF and MRI (Meteorological Research Institute) online transport model (MJ98-CDTM). In MJ98-CDTM, an off-line transport model (CDTM) is directly coupled with the MRI GCM (MJ98). In this study, we use CO2 observations of surface data (continuous and flask) and satellite data (GOSAT) obtained in 2009. We estimated 3-day-mean CO2 flux at a resolution of T42. Here, only CO2 concentrations and fluxes are analyzed whereas meteorological fields are nudged by JCDAS re-analysis. The horizontal localization length scale and assimilation window are chosen to be 1000 km and 3 days, respectively. We also introduce a GOSAT bias correction scheme, since there are low concentration biases in the current GOSAT data comparing with the CO2 distribution analyzed by JMA. The assimilation system works properly, better than free transport model run validating with independent CO2 concentration observational data. We further plan to compare our results with current inverse model results.