TransCom Continuous Experiment CSU NASA PCTM Scott Denning, John Kleist
PCTM Simulations 1° x 1.25° lat/lon grid, 55 levels (we use 25) NASA GEOS4 reanalyzed winds, turbulence, and convective mass fluxes for Partial results only: –FF –Takahashi –SiB monthly –CASA monthly Doesn’t balance budget
Seasonal Cycles BRW MLO SMO LEF green = obs red = model
April, 2002 BRW MLO SMO LEF green = obs red = model
May, 2002 BRW MLO SMO LEF green = obs red = model
July, 2002 BRW MLO SMO LEF green = obs red = model
September, 2002 BRW MLO SMO LEF green = obs red = model
October, 2002 BRW MLO SMO LEF green = obs red = model
Source/Sink Inversions of Synthetic Satellite CO 2 with Errors Scott Denning Kevin Gurney Kathy Corbin Mick Christi TransCom3 Modelers
Inversions of Monthly Synthetic Data Generate global CO 2 from TransCom models Background fields plus G m post Interpolate all models onto common 4 x 5 grid Apply ISCCP cloud climatology to mask grid Invert for fluxes using all models’ response functions (standard T3 cyclostationary method) All results using flasks (with T3 error) plus satellite
Synthetic Satellite Retrievals TransCom models provide monthly mean 3D arrays of CO 2 mole fraction over 9 pressure-bounded layers To simulate satellite products from these fields, we need to treat –Vertical “weighting” of 9 layers corresponding to instrument retrieval (not the same as retrieval averaging kernel!) –An estimate of uncertainty in the retrieval –Averaging of many retrievals in each model grid column, with appropriate treatment of error reduction –Effects of clouds on number of retrievals in each model grid column, and therefore on aggregate uncertainty –Sampling biases (?) due to 1 PM Equatorial crossing time and measurement of only clearsky conditions
Vertical Weighting Sample synthetic data using two different vertical weighting: –Thermal IR (AIRS-like) sees mostly mid-to-upper troposphere –near-IR (OCO-like) sees column mean, with information all the way to surface
Monthly Column Uncertainty = 0 exp(1.5 f c 3 ) Implemented by deweighting sat retrievals ppmv
Inversions of Flasks Plus AIRS Perfect transport on perfect data returns perfect fluxes Transport error makes satellite data pretty ineffective Even perfect model needs retrieval error about 1 ppm for significant improvement over flasks
Inversions of Flasks Plus OCO Significant improvements over flasks Fairly robust against terrible transport error! Caveats: monthly mean inversion! severe transport error!
Tropical America Flask constraint very weak Even poor transport beats flasks
Temperate North America Flask constraint already pretty good Need better transport to beat flask-only inversion aggregate uncertainty (ppmv)