Global Sensitivity Analysis of the GEOS-Chem Chemical Transport Model for Six NASA Field Campaigns
Open Access
- Author:
- Christian, Kenneth Edward
- Graduate Program:
- Meteorology
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- August 11, 2017
- Committee Members:
- William Henry Brune, Dissertation Advisor/Co-Advisor
William Henry Brune, Committee Chair/Co-Chair
Kenneth James Davis, Committee Member
Anne Thompson, Committee Member
Sarma V Pisupati, Outside Member
Jingqiu Mao, Special Member - Keywords:
- global sensitivity analysis
atmospheric chemistry
chemical transport model
ozone
oxidant
sensitivity analysis - Abstract:
- Global chemical transport models are popular tools within the air chemistry community to both assess our understanding of air chemical processes and to provide a platform to predict how atmospheric composition may respond to future changes. However, there are errors and uncertainties in almost all model inputs, such as emissions, chemistry, and meteorology rendering modeled results likewise uncertain. Thus a detailed analysis of model uncertainties and sensitivity of modeled results to its inputs are required to qualify and improve these predictions. Using a global sensitivity method involving the simultaneous perturbation of many chemical transport model input factors, we find the model uncertainty for ozone (O3), hydroxyl radical (OH), and hydroperoxyl radical (HO2) mixing ratios and apportion this uncertainty to model inputs including emissions, meteorology, and chemistry using the Random Sampling-High Dimensional Model Representation (RS-HMDR) global sensitivity analysis method for the DC-8 flight tracks corresponding to six NASA aircraft field campaigns. These campaigns took place from 1999 to 2008 in locations stretching from the South Pacific to the North American Arctic and from East Asia to the North Atlantic. In general, we find modeled and measured oxidant mixing ratios to agree once uncertainties in both are taken into account with some notable exceptions. Notably, these exceptions include ozone during the Houston flights of the INTEX-B campaign, HO2 for the flights over the northernmost Pacific Ocean during INTEX-B, low altitude ozone and HO2 during TRACE-P, and ozone in a few of the vertical bins for PEM-Tropics B. For ozone and OH, modeled mixing ratios for most locations were most sensitive to a bevy of emissions, notably lightning NOx, various surface NOx sources, carbon monoxide, and isoprene. This means that lessening model uncertainties for these species in these regions will likely require improving emissions inventories. HO2 mixing ratios were most sensitive to CO and isoprene emissions as well as the aerosol uptake of HO2. In the case of some Arctic flights, we found HO2 and OH to be overwhelmingly sensitive to the aerosol uptake of HO2 with this one factor contributing upwards of 75 % of the total uncertainty in HO2. With the dominance of this one factor in the sensitivity of HOx mixing ratios for this region, improving the modeled characterization of HOx in this domain will require it being addressed. Across most campaigns, we find better model-measurement agreement with lower rates of aerosol uptake and lower lightning NOx emissions suggesting both may need to be reduced. For almost all campaigns the same seven factors were responsible for the majority of the model ensemble uncertainty for ozone, OH, and HO2: surface NOx emissions, carbon monoxide emissions, lightning NOx emissions, aerosol uptake of HO2, isoprene emissions, the reaction rate of NO2 + OH, and the photolysis rate of NO2. The frequency of these factors among the most sensitive factors in each campaign suggests that the applicability of these results may extend beyond the specific regions covered by the respective campaigns. In order to achieve appreciable future progress in decreasing modeled oxidant uncertainty, addressing these seven factors is likely required.