简体中文


404 Not Found

404 Not Found


nginx
CO2 Time Series
 
Site

We check the accuracy of the optimized model simulation using the surface CO2 observations and the CONTRAIL aircraft CO2 measurements. Figure 4a compares modeled CO2 concentration with measurements at the surface site of MNM (Minamitorishima, Japan; [24.29°N, 153.98°E, 8 m]). The comparison of the surface CO2 time series shows that the simulated CO2 concentration is in good agreement with observed data during the period 2006-2010. The observed amplitudes in the CO2 concentration are reproduced well, and the seasonal variations show accurate timing in spring (March-April-May) and autumn (September-October-November) but sometimes with weaker amplitudes in winter (December-January-February) and summer (June-July-August). This mismatch in CO2 seasonal amplitude suggests that our estimated CO2 surface fluxes do not catch the peak of terrestrial carbon uptake and emissions in summer and winter. Previous studies have also found this seasonal mismatch, which may correlate with atmospheric transport, and has already been identified as a shortcoming in most inversions.

Figure 4b and 4c show the histogram of residual distribution (simulated minus observed) in summer and winter. At MNM, the annual mean bias is 0.15±0.87 ppm, with a relatively large model overestimate of 0.51±0.93 ppm in summer and a small bias of −0.05±0.89 ppm in winter. When considering all sites, the simulated mole fractions exhibit good agreement with the observed CO2 time series and the change in inferred mixing ratios and flux are within the specified uncertainties in our inversion system, an important prerequisite for a good flux estimate.

We also check the model performance in the free troposphere in addition to the surface CO2. Figure 4d shows the comparison between CO2 measurements in the free troposphere during the period 2006-2010 in the region covering 136-144°N, 32-40°E, 375-425 hPa. The simulated troposphere CO2 concentration also matches the CONTRAIL measurements well. Similar to the surface CO2, we find seasonal mismatches in the summer and winter (Figures 4e and 4f). The relatively small annual mean of model bias (model-minus-observed) in the free troposphere (−0.01±1.18 ppm) is an integrated result of an overestimate (0.35±1.2 ppm) in summer and an underestimate (−0.29±0.99 ppm) in winter. Overall, the agreement between model and measurement for both the surface and CONTRAIL observations is fairly good and consistent with previously known behavior in the CarbonTracker systems, derived then mostly from North American and European continuous sites. Most importantly, residuals and innovation statistics (not shown) suggests that our inverted fluxes, 26,470 CO2 observations (7,957 surface observations; 10,467 CONTRAIL observations), and simulated horizontal and vertical transport in CTDAS are to a high degree consistent with each other over Asia.

Figure 4. (a) CO2 observed and simulated time series; (b) Summer histograms of the residuals (residuals = simulated-minus-observed); (c) Winter histograms of the residuals at Surface site of NMN during 2006-2010. (d) CO2 observed and simulated time series; (e) Summer histograms of the residuals (residuals = simulated-minus-observed); (f) winter histograms of the residuals of CONTRAIL CO2 during the period 2006-2010 in the region covering 136-144°N, 32-40°E, 375-425 hPa. The blue column in 3b, 3c, 3e and 3f shows the histogram of the residuals themselves, and the blue lines and statistics shown in blue text (indicated mean, standard deviation and observed number, respectively) are a summary of the residuals interpreted as a normal distribution. The vertical scales are determined by the number of observations and how tightly they are grouped, with the area under the histogram forced to unity.s




京公网安备 11010502044510号