Investigating Representation Errors in Inversions of Satellite CO 2 Retrievals K.D. Corbin, A.S. Denning, N.C. Parazoo Department of Atmospheric Science Colorado State University Transcom Meeting - Purdue University April 24-27, 2007

Transcom Meeting - Purdue University April 24-27, 2007 Motivation  OCO and GOSAT will retrieve total column X CO2 measurements beginning late 2008  Inverse modelers will use these measurements to help identify CO 2 sources and sinks  Using space borne X CO2 to represent a transport model grid-cell may introduce sampling errors into inversions  X CO2 measurements will only represent clear conditions  Investigate sampling errors using continuous data, a regional cloud-resolving model (SiB-RAMS), and a global transport model (PCTM)

Transcom Meeting - Purdue University April 24-27, 2007  CO 2 continuous data from 3 towers:  WLEF:  Harvard Forest:  Tapajos (km67):  Mid-day means from LST  Created clear-sky subset  Ranked PAR measurements from all years at each site  Selected 20% of days per month with highest PAR  Fit separate harmonic functions to clear-sky subset and entire time-series  Subtracted fits: CO 2 fit CLEAR - CO 2 fit TOTAL Calculating Clear-Sky Errors at CO 2 Towers

Transcom Meeting - Purdue University April 24-27, 2007 Clear-Sky Bias at Continuous CO 2 Tower Sites  CO 2 lower on clear days than on average  Temperate sites have greatest bias in winter  Tropical site shows biggest difference in rainy season [Corbin and Denning, 2006]

Transcom Meeting - Purdue University April 24-27, 2007 NEE Clear-Sky Bias at Tower Sites [Corbin and Denning, 2006]  In mid-lats, enhanced uptake on clear days during summer, but negligible winter errors  In tropics, enhanced uptake year-round on clear days  NEE bias cannot account for CO 2 errors

Transcom Meeting - Purdue University April 24-27, 2007 SiB2-RAMS Case Descriptions North AmericaSouth America 97 KM 450 KM  August 11-21, 2001  3 frontal passages  August 1-16, 2001  Dry season - calm conditions  Bulk microphysical parameterization to simulate clouds and precipitation explicitly Emulated satellite track

Transcom Meeting - Purdue University April 24-27, 2007 Spatial Representation Errors using SiB-RAMS [Corbin et al., 2007]  Errors are unbiased and generally less than 0.5 ppm  Spatial errors increase with domain heterogeneity and size

Transcom Meeting - Purdue University April 24-27, 2007 Clear-Sky Temporal Errors using SiB-RAMS [Corbin et al., 2007]  Large errors at both sites  Biased errors at temperate site due to CO 2 anomalies associated with frontal systems that are masked by clouds

Transcom Meeting - Purdue University April 24-27, 2007 Clear-Sky Errors using PCTM  Global 2003 simulation  1.25 o longitude x 1.0 o latitude  Calculated clear-sky CO 2 error for each land grid-cell  Used daytime mean total column CO 2 concentrations  Created clear-sky subset using downward shortwave radiation  Fit 2 harmonics to clear-sky and total data  Subtracted Fit CLEAR - Fit TOTAL WLEF Tapajos

Transcom Meeting - Purdue University April 24-27, PCTM Comparisons to Observations Clear-sky bias from PCTM at tower locations match observed errors reasonably well

Transcom Meeting - Purdue University April 24-27, 2007 Annual Mean Clear-Sky Errors in PCTM  Errors vary regionally with spatially coherent patterns  Underestimation of CO 2 in South America and Alaska  Overestimation of CO 2 in Asia

Transcom Meeting - Purdue University April 24-27, 2007 Annual Mean Clear-Sky Errors by Latitude  Clear-sky errors larger in NH  Underestimation of mean in sub-tropics

Transcom Meeting - Purdue University April 24-27, 2007 Seasonal Clear-Sky Errors in PCTM  Magnitude of errors varies with season

Transcom Meeting - Purdue University April 24-27, 2007 Seasonal Clear-Sky Errors by Latitude  Large underestimation of CO 2 in NH land during winter  Overestimation of CO 2 in NH summer  Underestimation of CO 2 in SH spring

Transcom Meeting - Purdue University April 24-27, 2007 Conclusions  Spatial representation errors are small (< 0.5 ppm)  Clear-sky (temporal) sampling errors are large and vary seasonally and regionally  Satellite X CO2 cannot be used to represent temporal averages  Transport must be modeled accurately and sampled at same time/location as satellite

Transcom Meeting - Purdue University April 24-27, 2007 Acknowledgements  Thanks to Steve Wofsy for the Tapajos Forest (km67) and Harvard Forest tower data and Ken Davis for the data from WLEF  Funding by NASA Earth System Science Fellowship , NASA Contract NNG04GQ15H SUPP2, and NASA Subcontract (via Purdue University)