Optimal Determination of Global Tropospheric OH Concentrations Using Multiple Trace Gases

Huang, J.
CGCS Report Series, Ph.D. Thesis, Department of Earth, Atmospheric and Planetary Sciences, MIT, Report Nr. 65
2000

The hydroxyl radical (OH) plays a decisive role in tropospheric chemistry. Reactions with OH provide the dominant path of removal for a variety of greenhouse gases and trace species that contribute to the destruction of the ozone layer. Accurate determination of global tropospheric OH concentrations [OH] is therefore a very important issue. Previous research at the global scale has focused on scaling model-calculated OH concentration fields using a single so-called titrating species, either CH3CCl3 or 14CO, and the data usually come from one measurement network. Therefore, the estimation of [OH] relies heavily on the accuracy of the emission estimates and absolute calibration of the observed mixing ratios of the single species. The goal of this thesis is to reduce the dependence of estimating [OH] fields on a single species and thus to improve our knowledge of global OH concentrations and trends. To achieve this goal, we developed a multiple titrating gases scheme which combines all the possible available surface measurements of CH3CCl3, CHF2CL (HCFC-22), CH2FCF3 (HFC-134a), CH3CFCl2 (HCFC-141b) and CH3CF2Cl (HCFC-142b) from both AGAGE (Advanced Global Atmospheric Gases Experiment) and CMDL/NOAH Nitrous Oxide And Halocompounds) network.

The optimal estimation of the global OH concentration and its trend is accomplished through a Kalman filtering procedure by minimizing the weighted difference between the predicted mixing ratios from atmospheric chemical-transport models and, for the first time, all measurements of the various titrating gases simultaneously. A two-dimensional land-ocean-resolving (LO) statistical-dynamical model and a 12-box model are used to predict the concentrations of the titrating gases. These two models are computationally efficient, and suitable for repetitive runs and long-term integrations. The eddy-diffusive transports in the 12-box model and the 2D-LO model are tuned optimally by using the Kalman filtering and CFC-11 and CFC-12 data before the estimations of OH are carried out. Three different techniques (content method, trend method, and time-varying OH method) are used to perform the Kalman filtering. These three methods optimally fit different features of the measurements and have different sources of errors. Errors in the measurements, industrial emission estimates, and chemical-transport models are included in great detail for the OH estimation problem. The random measurement errors and mismatch errors are included in the noise matrix in the Kalman filter. For other random errors from the emission estimates and chemical-transport models, we use the Q-inclusion method which specifies the random model errors explicitly in the state error matrix Q inside the Kalman filtering. For systematic errors in the calibration, model, and emissions, we use the brute-force method which repeats the entire inverse method many times using different possible values of the measurement sensitivity matrix in the Kalman filtering.

Using multiple gases, both CMDL and AGAGE data, two chemical-transport models, and selected content and trend results, our best estimate of the global mean tropospheric OH concentrations is 9.4 (+2.7 / -1.7) x 105 radicals cm-3, and our best estimate of the linear OH trend is -0.5 trend +/- 1.0% per year over the 1978-1998 time period. Methyl chloroform data give the heaviest weight to the overall estimations. This is because there are more CH3CCl3 measurements than for any other titrating gases, and the industrial emission estimates of this gas are the most accurate. The derived OH estimations agree statistically with previous studies taking into account the fact that the global mean OH concentration of (9.7 +/- 0.6) x 105 radicals-3 and on OH trend of 0.0 +/-0.2% per year over the 1978-1993 are reported in Prinn et al. (1995).

As far as the major sources of error in the OH estimations are concerned, we find that, using individual gases separately, the uncertainties in absolute calibrations, rate constants, and industrial emissions estimates are important sources of error for all five titrating gases. The measurement errors and the initial a priori guesses in the Kalman filter are also important sources of error for the three newer titrating gases (HCFC-141b, HCFC-142b, and HFC-134a) because of their very low mole fractions as well as the short measurement records for these gases. Combining multiple OH titrating gases together, we find that errors in industrial emissions contribute the most to the uncertainty in the OH estimation problem.

We also find that incorporating random model errors (other than mismatch errors) using the Q-inclusion method generates satisfactory agreement for best guess estimates with the approach in which Q = 0 in the Kalman filter. However the Q-inclusion method provides an estimate of the effect of random model error. Newer titrating gases generally yield OH estimates comparable to those from CH3CCl3 but with larger uncertainties. One of the exceptions is using HCFC-142b data with the content method, which yields a physically impossible negative OH concentration because of the underestimates of emissions for this gas. However, the trend method using HCFC-142b data still delivers reasonable OH estimates, because the trend method is not sensitive to systematic errors. The measurements of the newer OH titrating gases can be used effectively with appropriate techniques to ultimately replace the use of CH3CCl3 (which is disappearing from the atmosphere), provided estimates of their emissions are improved. This is particularly true for HCFC-142b. In addition to the OH estimations, a time-varying adaptive-Kalman filter is also utilized in this thesis to deduce monthly emissions of HCFC-141b and HCFC-142b. We find that the current industrial estimates of HCFC-142b need to be at least doubled, and the emissions of HCFC-141b need to be increased by 20 to 30% to achieve the best agreement with observations.