Vikram Khade
Department of Physics, University of Toronto, Ontarion, Canada
Saroja Polavarapu;
Michael Neish;
Pieter Houtekamer
Environment and Climate Change Canada
Dylan Jones
Department of Physics, University of Toronto, Ontarion, Canada
The 8th EnKF Data Assimilation Workshop
State estimation of CO with EC-GHG_EnKF by ingesting surface observations in an OSSE framework
Poster:
EnKF_CO_Poster_Khade_EnKF_2018_version_final.pdf
The ECCC Carbon Assimilation System (EC-CAS) is being developed to study the global carbon cycle and to monitor greenhouse gas emissions over Canada. By assimilating atmospheric measurements of greenhouse gases, exchanges of carbon (CO2, CH4 and CO) between the surface and the atmosphere can be estimated. While the traditional approach to flux estimation is through inverse modelling using analysis winds, EC-CAS uses a coupled state flux estimation with the low resolution configuration of our operational Ensemble Kalman Filter (EnKF) (Houtekamer et al. 2014) and our operational weather forecast model (GEM). GEM was adapted to simulate greenhouse gases and is run on a global 400x200 grid with 81 vertical levels (Polavarapu et al. 2016). In the EnKF approach the uncertainty in winds contributes to the uncertainty in the tracer field.
In this work, we test and tune the EnKF for CO state estimation using OSSEs (Observation System Simulation Experiment). An ensemble of size 64 is used to represent the uncertainty in the meteorological and CO state. An IAU (Incremental Analysis Update) is used within the EnKF to supress high frequency oscillations. Simulated observations are assimilated every 6 hours and consist of radiosonde, scatterometer, satellite winds, gps, aircraft and surface observations of CO. Variable localization (Kang et al. 2011) is implemented which results in the CO observations updating the CO state only. Similarly the meteorological observations update the meteorological state only.
The analysis RMSE for meteorological variables averaged over the 3d model domain tends to saturate in about 15 days. At saturation, the spread is under-dispersive (by 5%) for all the variables owing to the moderate ensemble size. Interestingly, without assimilating any CO observations, the RMSE in CO averaged over the bottom 3 km nevertheless decreases from about 6 ppb to about 3 ppb in 4 days simply due to the concurrent improvement in the wind fields. Results from the OSSEs used to tune the localization radius for the assimilation of surface observations of CO will be reported. The system will then be tested for the assimilation of MOPITT satellite data.
References :
Houtekamer, P., B. He, and H. Mitchell, 2014 : Parallel Implementation of an Ensemble Kalman Filter. Mon. Wea.
Rev., 142, 11631182, doi: 10.1175/MWR-D-13-00011.1.
Kang, J.-S., E. Kalnay, J. Liu, I. Fung, T. Miyoshi, and K. Ide (2011) : Variable localization in an ensemble Kalman
Filter: Application to the carbon cycle data assimilation, J. Geophys. Res., 116, D09110, doi:10.1029/2010JD014673.
Polavarapu, S. M., Neish, M., Tanguay, M., Girard, C., de Grandpr, J., Semeniuk, K., Gravel, S., Ren, S., Roche,
S., Chan, D., and Strong, K.: Greenhouse gas simulations with a coupled meteorological and transport model: the
predictability of CO2. Atmos. Chem. Phys., 16, 12005-12038, doi:10.5194/acp-16 12005-2016, 2016
The ECCC Carbon Assimilation System (EC-CAS) is being developed to study the global carbon cycle and to monitor greenhouse gas emissions over Canada. By assimilating atmospheric measurements of greenhouse gases, exchanges of carbon (CO2, CH4 and CO) between the surface and the atmosphere can be estimated. While the traditional approach to flux estimation is through inverse modelling using analysis winds, EC-CAS uses a coupled state flux estimation with the low resolution configuration of our operational Ensemble Kalman Filter (EnKF) (Houtekamer et al. 2014) and our operational weather forecast model (GEM). GEM was adapted to simulate greenhouse gases and is run on a global 400x200 grid with 81 vertical levels (Polavarapu et al. 2016). In the EnKF approach the uncertainty in winds contributes to the uncertainty in the tracer field.
In this work, we test and tune the EnKF for CO state estimation using OSSEs (Observation System Simulation Experiment). An ensemble of size 64 is used to represent the uncertainty in the meteorological and CO state. An IAU (Incremental Analysis Update) is used within the EnKF to supress high frequency oscillations. Simulated observations are assimilated every 6 hours and consist of radiosonde, scatterometer, satellite winds, gps, aircraft and surface observations of CO. Variable localization (Kang et al. 2011) is implemented which results in the CO observations updating the CO state only. Similarly the meteorological observations update the meteorological state only.
The analysis RMSE for meteorological variables averaged over the 3d model domain tends to saturate in about 15 days. At saturation, the spread is under-dispersive (by 5%) for all the variables owing to the moderate ensemble size. Interestingly, without assimilating any CO observations, the RMSE in CO averaged over the bottom 3 km nevertheless decreases from about 6 ppb to about 3 ppb in 4 days simply due to the concurrent improvement in the wind fields. Results from the OSSEs used to tune the localization radius for the assimilation of surface observations of CO will be reported. The system will then be tested for the assimilation of MOPITT satellite data.
References :
Houtekamer, P., B. He, and H. Mitchell, 2014 : Parallel Implementation of an Ensemble Kalman Filter. Mon. Wea.
Rev., 142, 11631182, doi: 10.1175/MWR-D-13-00011.1.
Kang, J.-S., E. Kalnay, J. Liu, I. Fung, T. Miyoshi, and K. Ide (2011) : Variable localization in an ensemble Kalman
Filter: Application to the carbon cycle data assimilation, J. Geophys. Res., 116, D09110, doi:10.1029/2010JD014673.
Polavarapu, S. M., Neish, M., Tanguay, M., Girard, C., de Grandpr, J., Semeniuk, K., Gravel, S., Ren, S., Roche,
S., Chan, D., and Strong, K.: Greenhouse gas simulations with a coupled meteorological and transport model: the
predictability of CO2. Atmos. Chem. Phys., 16, 12005-12038, doi:10.5194/acp-16 12005-2016, 2016
Thanks to McGill's Atmospheric and Oceanic Sciences department for hosting this web-site.