1. Inter-comparison of retrieved CO 2 from TCCON, combining TCCON and TES to the overpass flight data Le Kuai 1, John Worden 1, Susan Kulawik 1, Kevin Bowman 1, Christian Frankenberg 1, Edward Olsen 1, Debra Wunch 3, Run-Lie Shia 3, Brian Connor 2, Charles Miller 1, and Yuk Yung 3 1.Jet Propulsion Laboratory, California Institute of Technology, 4800 Oak Grove Drive, Mail stop: 233-200, Pasadena, CA 91109 2.BC Consulting Ltd., 6 Fairway Dr, Alexandra 9320, New Zealand 3.California Institute of Technology, 1200 E. California Blvd., Pasadena, CA, 91125 The comparison of CO 2 from TCCON profile retrievals and aircraft overpass data 1.Comparison of dry mole fraction (DMF) profile: f CO2 (z) 2.Comparison of total column-averaged DMF: X CO2 3.Comparison of partial column-averaged DMF in boundary layer: pX CO2 BDL (combining TCCON and TES) Time window selection : Short enough to measure a same air parcel by flight and TCCON. Long enough for sufficient number of profiles from TCCON for a good statistical treatment. Overpass time windows: Lamont: Lat=36°, Lon=-97° Parkfalls: Lat=46°, Lon=-90° TCCON site DateFlight time window (UTC) Retrieval time window (UTC) Number of retrievals Parkfalls2004/07/1416:18:17 – 16:48:1915:00 – 17:3036 Parkfalls2004/07/1513:11:09 – 15:49:0013:00 – 16:0081 Parkfalls2004/08/1419:46:37 – 22:24:2919:00 – 23:0044 Parkfalls2008/05/1217:1918 – 17:55:2916:00 – 20:0036 Lamont2009/01/3019:43:18 – 20:46:3019:00 – 21:0042 Lamont2009/07/3114:37:00 – 17:31:0014:00 – 18:00111 Lamont2009/08/0215:05:00 – 17:57:0015:00 – 18:0085 Lamont2009/08/0315:14:00 – 18:00:0015:00 – 19:0090 Lamont2010/07/1816:15:39 – 20:27:5416:00 – 20:3079 Aircraft measured dry profile TCCON retrieved wet profile Covariance for CO 2 Covariance for temperature Convert to dry profile A priori: Retrieved: Parkfalls: Apply averaging kernel FLT_AK: 2004/07/12 2004/07/15 2004/08/02 2008/05/12 2009/01/30 2009/07/31 2009/08/02 2009/08/03 2010/07/18 Lamont: Abstract: The Total Carbon Column Observing Network (TCCON) provides measurements of column abundances of CO 2, CO, CH 4 and other molecules that absorb in the near infrared with high accuracy and high precision (e.g.< 0.25% for CO 2 ). Therefore, this dataset serves as a link between satellite measurements and ground- based in situ network. In this study, a retrieval algorithm is developed to retrieve the CO 2 profiles in addition to the column-averaged dry-air mole fractions (DMF) (X CO2 ). The inter-comparison between the TCCON retrieved CO 2 products and flight measurements are focus at Lamont as well as Park falls in this work. The retrieved profiles capture the first order vertical variability in the overpass aircraft measured profile at both sites. TCCON X CO2 have 1.24% negative bias to the integrated aircraft data with TCCON averaging kernel applied. The root-mean-square (RMS) of the current retrieved X CO2 is about 0.12%. Model values from the TES-assimilated GEOS-Chem model are used to subtract the partial column above 600 hPa from TCCON’s total column. The boundary layer CO 2 by combining TCCON and TES shows a good agreement with the integration of flight profile within the boundary layer (e.g. below 600 hPa). This method to disentangle the boundary layer CO 2 shows potential of joint retrieval of NIR and TIR, and combining any total column measurement (e.g. TCCON or GOSAT) with any free tropospheric CO 2 data (e.g. TES or AIRS). With a long term boundary layer CO 2, the CO 2 surface flux can be better estimated. Without O 2 Correction: With O 2 Correction 2. Total column-averaged DMF comparison 3. Determine boundary layer partial column by combining TCCON and TES data P cuf-off RMS P cut-off 800 hPa600 hPa Total column- averaged DMF TCCON0.42 ppm Partial column- averaged DMF in boundary layer TCCON – TES1.78 ppm0.70 ppm TCCON – A priori7.49 ppm2.61 ppm 800 hPa 600 hPa P cut-off = 600 hPaRMS TCCON0.46 ppm TCCON & TES1.46 ppm TCCON a priori3.2 ppm To determine dry X CO2, the normal way is to remove amount of water from the total amount of air. Our retrieval simultaneously retrieves H 2 O by shift its a priori profile. The derived dry X CO2 using retrieved H 2 O profile has small bias but low precision. Since O 2 can also be retrieved, the method by normalizing the retrieved O 2 not only provides dry X CO2 but also improves the precision. It is noticed that about 1% negative bias is induced because the limited knowledge of the spectroscopy in O 2 band. 1. Profile comparison: Applying the averaging kernel and a priori constraint vector to the aircraft data (Flt) ( which is on TCCON grid) yields Flt_AK, a profile which accounts for the TCCON sensitivity and vertical resolution. Flt_AK also represents the profile that would be retrieved from TCCON measurements in the absence of other errors. The comparison should performed between the TCCON profile (Ret) and aircraft data have had the TCCON operator applied (Flt_AK). These profile comparisons give a good overview of the variability and bias in TCCON profile. Combining TCCON and TES assimilated data, the boundary layer partial column CO 2 is determined by subtracting the partial column amount within and above free troposphere from the total column amount by TCCON. The remained partial column amount in boundary layer is weighted by the partial column amount of dry air in the boundary layer for. The comparison of to those by integral the flight profile within boundary layer shows small bias and high precision. The knowledge of boundary layer CO 2 was greatly improved by combining TCCON and TES assimilated CO 2 data compared to the climatology a priori. Inter-comparison of TCCON combining TES to SGP flight measurements TES SGP FLT There are more flight measurements at Lamont in 2009 but these CO 2 profiles only go up to 5 to 6 Km. However, it still allow us to compare the boundary layer CO 2 from combining TCCON and TES to the flight data. For the comparison of total column, a priori CO 2 is replace above the ceiling of the flight measurements. Here are sixteen days’ comparison in 2009 from January to December when flight measurements are available. The bias and precision of X CO2 are both consist with previous results. The bias in boundary layer CO 2 stay small but root mean square (RMS) is increased due to two outliers. These two outliers are because flight boundary layer CO 2 are outside the a priori constrain region. Units (ppm)BiasStd Measureme nt error Systematic error 2009010 8-5.150.280.230.64 2009011 6-5.180.30.280.64 2009012 9-5.30.30.290.66 2009020 4-5.050.260.280.64 2009021 1-4.90.540.280.65 2009021 9-5.391.080.280.66 2009022 1-4.970.360.290.67 2009030 4-4.393.490.270.59 2009040 7-5.490.260.280.62 2009051 7-5.870.320.270.6 2009060 4-5.650.380.270.59 2009070 1-5.790.860.270.56 2009082 2-5.571.240.270.56 2009090 1 -4.810.450.270.57 2009110 1-6.110.360.270.59 2009121 8-5.860.470.280.65 Error analysis This table lists the bias of the mean retrieved X CO2 in each of sixteen time windows relative to the flight data. The true variability of the retrievals during most time window is consist with the expected measurement error. It suggests that the instrument random noise account for the variability of retrievals within the time window scale. Some of days have unusual large variability because of the cloud coverage (e.g. 2009/03/04, 2009/08/22, etc.). The standard deviation of these bias is 0.46 ppm. The expected systematic error due to the temperature covariance matrix show in fig 1 is about 0.6 ppm. It suggests that the variability of the bias is mostly due to the systematic error such as temperature uncertainties. Conclusions: The profile retrieval of TCCON measurement can capital most of the CO 2 vertical variability in the atmosphere. It integrated column-averaged CO 2 abundance has 1.24% negative bias to flight integrated column-averaged abundance. The precision is 0.12%. By subtracting free tropospheric partial column amount from TES-GEOS assimilated data from the total column amount from TCCON data, the partial column- averaged abundance in boundary layer can be estimated. The comparison to the integral of flight measurements at SGP in the boundary layer has small bias and 1.46 ppm for RMS. It improves the RMS for the X CO2 of the a priori to the flight column CO 2. There are about 50 similar SGP flight CO 2 profile near Lamont in 2009 in total. More points will be added to the future similar analysis to provide more robust conclusion. More total column and partial column in boundary layer data will be provided at other TCCON sites with different latitudinal and longer temporal coverage. They will allow us to study the spatial and temporal variability of both column and boundary layer CO 2. Furthermore, the boundary layer CO 2 could help constrain and improve the estimation of the surface flux. CALIFORNIA INSTITUTE OF TECHNOLOGY Profile retrieval set up The CO 2 profile is retrieved at 10 levels. Other interference gas, such as H 2 O, HDO and CH 4 are simultaneously retrieved by shift their a priori profiles to minimize the residual of the observed and model calculated spectra. The signal to noise ratio used in retrieval is 200 for the covariance for measurement random noise. Figure 1. Figure 2. Table 1. Figure 3. Figure 4. Figure 5. Table 2. Table 3. Table 4. Figure 6.