We also investigated the vertical cross section of the vertical pressure velocity (dP/dt) across 70°W to 10°E averaged over 20°S-5°S from December to March in 2003-2010. As shown in Figure 4a, it is clear that air sinks over 35°W-10°E and rises over 70°W-35°W. As a result of vertical motions, AIRS mid-tropospheric CO2 concentrations are relatively low over 35°W-10°E and relatively high over 70°W-35°W (Figure 4b). The difference of mid-tropospheric CO2 between South Atlantic Ocean and South America areas is about 1 ppm from December to March.

Variability of CO2 From Satellite Retrievals and Model Simulations Xun Jiang1, David Crisp2, Edward T. Olsen2, Susan S. Kulawik2, Charles E. Miller2, Thomas S. Pagano2, and Yuk L. Yung3

1Department of Earth & Atmospheric Sciences, University of Houston, Houston, TX 77204; 2Jet Propulsion Laboratory, California Institute of Technology, 4800 Oak Grove Drive, Pasadena, CA 91109;3Division of Geological and Planetary Sciences, California Institute of Technology, Pasadena, CA 91125

Satellite CO2 retrievals from the Greenhouse gases Observing SATellite (GOSAT), Atmospheric Infrared Sounder (AIRS), and Tropospheric Emission Spectrometer (TES) and in-situ measurements from the Earth System Research Laboratory (NOAA-ESRL) Surface CO2 and Total Carbon Column Observing Network (TCCON) are utilized to explore the CO2 variability at different altitudes. A multiple regression method is used to calculate the CO2 annual cycle and semiannual cycle amplitudes from different data sets. The CO2 annual cycle and semiannual cycle amplitudes for GOSAT XCO2 and TCCON XCO2 are consistent, but smaller than those seen in the NOAA-ESRL surface data. The CO2 annual and semiannual cycles are smallest in the AIRS mid-tropospheric CO2 compared with other data sets in the northern hemisphere. Similar regression analysis is applied to the Model for OZone And Related chemicl Tracers-2 (MOZART-2) and CarbonTracker model CO2. The convolved model CO2 annual cycle and semiannual cycle amplitudes are similar to those from the satellite CO2 retrievals, although the model tends to underestimate the CO2 annual cycle amplitudes in the northern hemisphere mid-latitudes and underestimate the CO2 semi-annual cycle amplitudes in the high latitudes.

AIRS mid-tropospheric CO2 data are also used to explore the variability of CO2 over the South Atlantic Ocean. It was found that the CO2 difference is ~1 ppm between the South Atlantic Ocean and South America during December to March. During December to March, there is sinking motion over the South Atlantic Ocean. The sinking air brings high altitude air with low concentrations of CO2 to the mid-troposphere. Meanwhile, air rises over South America, which brings surface high concentrations of CO2 to the mid-troposphere over South America. As a result, the mid-tropospheric CO2 concentrations are lower over the South Atlantic Ocean than over South America during December to March. It is also found that the detrended AIRS mid-tropospheric CO2 difference correlates well with the inverted and detrended 400 hPa vertical pressure velocity difference between South Atlantic and South America. Results obtained from this study demonstrate the strong modulation of large-scale circulation on the mid-tropospheric CO2 and suggest that mid-tropospheric CO2 measurements can be used as an innovative observational constraint on the simulation of large-scale circulations in climate models.

CO2 Annual Cycle & Semiannual Cycle

Figure 1: (a) Latitudinal distributions of CO2 annual cycle amplitudes. (b) Latitudinal distributions of CO2 semiannual cycle amplitudes. Blue lines are results from AIRS mid-tropospheric CO2. Green lines are results from GOSAT XCO2. Purple dots are results from NOAA-ESRL surface CO2. Orange triangles are results from TCCON XCO2. Error bars are the uncertainties of CO2 annual cycle and semiannual cycle amplitudes derived from the multiple regressions. Figure is from Jiang et al. [2014a].

Figure 3: AIRS mid-tropospheric CO2 averaged from December to March in 2003-2010. Units for CO2 are ppm. Color represents AIRS mid-tropospheric CO2. White contours are the NCEP2 400 hPa vertical pressure velocity (dP/dt). Solid white contours refer to the sinking air. Dashed white contours refer to the rising air. Figure is from Jiang et al. [2014b].

Conclusions

Results for the annual cycle and semiannual cycle amplitudes from the model convolved CO2 are plotted against satellite and surface CO2 in Fig. 2. The model convolved CO2 annual cycle amplitudes are similar to those from satellite CO2 and NOAA-ESRL surface CO2. The values obtained by convolving the model CO2 by the GOSAT averaging kernel are larger than the values obtained by convolving the model CO2 by the AIRS CO2 averaging kernel, because the GOSAT XCO2 averaging kernel’s maximum is closer to the surface than that for the AIRS CO2 averaging kernel.

Both models show CO2 semiannual cycles that are larger in the NH than SH. The CO2 semiannual cycle amplitude obtained by convolving the models with the GOSAT averaging kernel is about 0.5-2 ppm in the NH, which is similar to the measured GOSAT CO2 semiannual cycle shown in Fig. 2f. The amplitude of the CO2 semiannual cycle obtained by convolving the models with the AIRS averaging kernel is about 0.5-1 ppm in the NH, which is weaker than that from AIRS CO2 semiannual cycle in the high latitudes and requires further exploration with in-situ CO2 profile data in the future.

Figure 4: (a) Vertical pressure velocity (dP/dt) averaged over 20°S-5°S from December to March in 2003-2010. Units are 10-2 Pa/s. Solid white contours refer to the sinking air. Dashed white contours refer to the rising air. (b) AIRS CO2 averaged over 20°S-5°S from December to March in 2003-2010. Units are ppm. Figure is from Jiang et al. [2014b].

We investigated the temporal correlation between the South Atlantic Walker Circulation and the mid-tropospheric CO2 difference between the South Atlantic Ocean (30°W-10°E; 20°S-5°S) and the South America (70°W-40°W, 20°S-5°S) in Figure 5. As shown in Figure 5a, the detrended AIRS CO2 difference correlates well with the inverted and detrended vertical pressure velocity differences. The correlation coefficient between the detrended AIRS CO2 difference (black solid line) and the inverted and detrended mean vertical pressure velocity difference (red dashed line) is 0.66. The corresponding significance level is 9%.

The correlation coefficients between the detrended AIRS CO2 difference and the inverted and detrended vertical pressure velocity differences derived from reanalysis datasets and CMIP5 model simulations range from 0.55 to 0.72 (Figure 5b). The correlation coefficients are 0.67 for NCEP2, 0.64 for ERA-Interim, 0.55 for MERRA, between 0.57 and 0.72 for CMIP5 model simulations. Given the importance of large-scale circulation in driving global energy and water cycles, improving model simulations of large-scale circulation is critical to reducing the model spread in climate sensitivity estimates (Su et al. 2014). Since there are limited direct observations of vertical velocity, the mid-tropospheric CO2 can be utilized as an indirect constraint on model representation of large-scale circulation, for example, the vertical velocity of the South Atlantic Walker Cell.

Influence of South Atlantic Walker Circulation on CO2

AbstractFigure 2: (a) Comparison of annual cycle amplitudes between AIRS mid-tropospheric CO2 and model-convolved CO2 from MOZART (dotted line) and CarbonTracker (dash-dot line) (b) Comparison of semiannual cycle amplitudes between AIRS mid-tropospheric CO2 and model-convolved CO2, (c) and (d) are the comparisons of annual and semiannual cycle amplitudes between TES mid-tropospheric CO2 and model convolved CO2. (e) and (f) are the comparisons of annual and semiannual cycle amplitudes between GOSAT XCO2 and model-convolved CO2. (g) and (h) are the comparisons of annual and semiannual cycle amplitudes between NOAA-ESRL surface CO2 and model surface CO2. Units are ppm. Figure is from Jiang et al. [2014a].

References:Jiang, X., D. Crisp, E. T. Olsen, S. S. Kulawik, C. E. Miller, T. S. Pagano, M. Liang, and Y. L. Yung, (2014a), CO2 annual and semiannual cycles from multiple

satellite retrievals and models, Submitted to ESS. Jiang, X., E. T. Olsen, T. S. Pagano, H. Su, and Y. L. Yung, (2014b), Modulation of mid-tropospheric CO2 by the South Atlantic Walker Circulation, Submitted to

JAS.

A41H-3164

Figure 5: (a) Difference of the detrended AIRS mid-tropospheric CO2 between the South Atlantic Ocean (30°W-10°E; 20°S-5°S) and South America (70°W-40°W, 20°S-5°S) (black solid line) and difference of the inverted and detrended 400 hPa vertical pressure velocity (dP/dt) between the South Atlantic Ocean and the South America from reanalysis data and CMIP5 models. Different colored dashed lines are from different reanalysis data and CMIP5 models. Bold red dashed line is the averaged vertical pressure velocity difference from all reanalysis data and model simulations. (b) Correlation coefficients between detrended CO2 difference and detrended and inverted vertical pressure velocity differences from reanalysis data and CMIP5 models. A 3-month running mean has been applied to all time series to remove the high frequency signals. Figure is from Jiang et al. [2014b].

CO2 annual cycle and semiannual cycle amplitudes decrease with altitudes. Model-convolved CO2 annual cycle and semiannual cycle amplitudes are similar to those from the satellite CO2 retrievals.

Low concentrations of CO2 are seen over the Southern Atlantic Ocean, which is related to the sinking branch in the Atlantic Walker Circulation.

AIRS mid-tropospheric CO2 difference correlates well with the inverted and detrended 400 hPa vertical pressure velocity difference between South Atlantic and South America. Satellite CO2 retrievals can be used as an innovative observational constraint on the simulation of large-scale circulation in climate models.