MODEL-DATA COMPARISON OF MID-CONTINENTAL INTENSIVE FIELD CAMPAIGN ATMOSPHERIC CO2 MIXING RATIOS
Open Access
- Author:
- Diaz Isaac, Liza Ivelisse
- Graduate Program:
- Meteorology
- Degree:
- Master of Science
- Document Type:
- Master Thesis
- Date of Defense:
- May 10, 2010
- Committee Members:
- Kenneth James Davis, Thesis Advisor/Co-Advisor
Kenneth James Davis, Thesis Advisor/Co-Advisor
Natasha Lynn Miles, Thesis Advisor/Co-Advisor - Keywords:
- atmospheric inversion
carbon cycle
carbon dioxide
MCI field campaign - Abstract:
- In-situ tower-based CO2 observations from the Mid-Continental Intensive (MCI) field campaign are compared to Carbon Tracker simulated CO2 concentrations. This study uses simulated CO2 concentrations from the third level, which is the most representative of the convective boundary layer (CBL). Daily daytime average from both observed and simulated CO2 concentrations are utilized in this study to ensure well-mixed, boundary conditions. This model-data comparison is applied using both time and spatial statistical analysis for two different periods: June through December 2007 and the 2007 growing season. The comparison shows that the model tends to predict 10 to 15 ppm higher mid-summer or growing season concentrations at three sites located in the “corn belt”. Carbon Tracker tends to be highly correlated to the observations for the period of June through December (≥ 0.8), but this correlation is lower for the growing season period (≤ 0.8). Residuals are not Gaussian, raising questions regarding the assumption of Gaussian residuals typically used in atmospheric inversions. The time scale for autocorrelation of the residuals is 30 to 80 days, the result of Carbon Tracker’s persistent overestimate of the growing season mixing ratio. Spatial correlations are largest between sites that are closer and share the same type of vegetation, but some other high spatial correlations diagnosed are not explained by these factors. The residuals do not show any clear correlations with synoptic weather. Overall, the model results for all sites are more inconsistent with the observations during the growing season, with the “corn belt” sites showing the least skill of all.