Grantee Research Project Results
Final Report: Chemical-Transport Modeling of the Global Atmosphere Environmental Problems: Evaluations, Comparisons and Initial Studies
EPA Grant Number: R826384Title: Chemical-Transport Modeling of the Global Atmosphere Environmental Problems: Evaluations, Comparisons and Initial Studies
Investigators: Wuebbles, Donald J. , Kotamarthi, V. Rao
Institution: University of Illinois Urbana-Champaign
EPA Project Officer: Hahn, Intaek
Project Period: April 1, 1998 through March 31, 2001
Project Amount: $312,334
RFA: Exploratory Research - Environmental Chemistry (1997) RFA Text | Recipients Lists
Research Category: Water , Land and Waste Management , Air , Safer Chemicals
Objective:
The main objectives of this research project were to: (1) develop an advanced three-dimensional modeling capability for the study of tropospheric processes; (2) test the model relative to available measurements, such as the field measurement campaigns conducted by the National Aeronautics and Space Administration (NASA) under the Global Tropospheric Experiment (GTE) program and the Climate Monitoring and Diagnostics Laboratory (CMDL) baseline data sets for ozone and CO; (3) use this model to analyze tropospheric loss of various halocarbons, including short-lived chlorofluorocarbon (CFC) replacement compounds emitted into the atmosphere; and (4) evaluate the predictability of major tropospheric oxidants in the model.
This project concerned the development, testing, and evaluation of a new three-dimensional chemical-transport model (3-D CTM) that will greatly increase the capabilities for dealing with global atmospheric issues of interest to the U.S. Environmental Protection Agency (EPA). This model was developed in coordination with the National Center for Atmospheric Research (NCAR). Initial project studies were directed at testing and evaluating the model in comparison with available observations. Initial research studies with the model focused primarily on the oxidizing capacity of the atmosphere. Many compounds, including most of the replacement compounds being used and/or considered for CFCs, halons, and other controlled chemicals, are designed to react in the troposphere, largely with OH. Improved understanding of the oxidative capacity of the atmosphere is essential for correctly estimating the lifetimes of these compounds and for predicting the fate of their degradation products in the troposphere. The sensitivity of model-calculated tropical oxidant concentrations to perturbations, including biomass burning, will be investigated. The calculated oxidant field will be tested with an offline simulation of CH3CCl3 and compared with observations. The sensitivity of model-calculated lifetimes of short-lived (~ months) replacements to temporal and spatial variabilities of the source will be investigated. We expect as a result of this project to reduce the uncertainties in calculating the tropospheric lifetimes of such replacement compounds and to provide an improved understanding of the oxidizing capacity of the atmosphere and the changes that may be occurring.
Summary/Accomplishments (Outputs/Outcomes):
The MOZART Chemical-Transport Model of the Global Atmosphere
A key aspect of this project was the continued development and testing of the global scale 3-D CTM of the troposphere and lower stratosphere Model for Ozone and Related Trace Gases (MOZART). We are teaming with NCAR in co-developing this model. At the inception of the project, we started with a preliminary version of MOZART (version 1) and now, partly as a result of our experiences with version 1, a new version of the model (version 2) is our primary modeling tool. Our initial studies under this project showed a number of problems with that version of the model, particularly in representing ozone and a number of other constituents in the middle to upper troposphere. The new version of the model, used in most of the studies discussed, is described below. This model is much improved over the earlier model and much more capable as a tool for applications of interest to the U.S. EPA in the understanding of tropospheric processes and their effects on environmental concerns.
The 3-D global chemical-transport model (GCTM), MOZART (version 2, details described in Horowitz, et al., 2001), has a horizontal resolution of 2.8 in latitude and longitude, and extends vertically from the surface to approximately 38 km. The GCTM has 34 vertical layers along sigma coordinates, and its vertical grid resolution varies from 60 meters near the surface to approximately 2 km at the top of the model domain. Transport is driven offline by NCAR Community Climate Model version 3 (CCM3) with dynamical variables (wind fields, pressure, temperature, and convection) saved every 3 hours. The CCM3-derived dynamical drivers for MOZART were extracted from a 10-year run of CCM3 forced at the lowered boundary with climatological values of sea surface temperatures. This is expected to yield a representative year for the current atmosphere. In this sense, the MOZART results described below are referred to as climatological. It should be noted, however, that the MOZART results would be representative of the current climate only if such factors as emissions also are of this nature. Because the emissions employed are from data sets built for a particular year, the simulation results presented here do not truly represent a chemical climatology.
This model accounts for surface emissions (including N2O, CH4, CO, NOx, NMHC, CH2O, isoprene, acetone, etc.), chemical and photochemical reactions, advection, convection, and wet and dry deposition. The emission fluxes for anthropogenic and biogenic sources were prescribed from the Emission Database for Global Atmospheric Research ([EDGAR], Olivia, et al., 1996) in the MOZART. A more detailed treatment of emissions used in MOZART version 2 is presented in Horowitz, et al. (2001). This version of MOZART provides spatial and temporal distributions for 52 chemical trace constituents. The chemistry scheme currently used in MOZART is similar to the one used in the Intermediate Model of Global Evolution of Species ([IMAGES], Muller and Brasseur, 1995), and incorporates 107 gas-phase, 5 heterogeneous, and 29 photochemical reactions. The NMHC chemistry is represented by the degradation mechanism as discussed in Brasseur, et al. (1998). The empirical first-order heterogeneous reactions involving N2O5 and NO3 on sulfate aerosols are implemented in the model (Muller and Brasseur, 1995). Advection of the trace gases is simulated using the flux-form semi-Lagrangian (FFSL) formulation of Lin and Rood (1996) and replaces the previous shape-preserving semi-Lagrangian scheme (Williamson and Rasch, 1989). The FFSL scheme is conservative and upstream biased. In addition, it contains monotonic constraints and conserves tracer correlations. The moist convection of trace gases was parameterized using the scheme developed by Hack (1994), and the dry convection in the boundary layer was solved according to Zhang and McFarlane (1995) and similar to CCM3. Vertical diffusion with the boundary layer is represented by the parameterization of Holtslag and Bonville (1993). Wet and dry deposition rates are calculated as first-order loss processes. Dry deposition is specified as a sum of species (e.g., O3, NOx, HNO3, PAN, organic nitrates, H2O2, organic peroxides, CH2O, CH3COCHO, CO, CH3COCH3, CH4, Ld-210), independent aerodynamic resistance, and a species-dependent surface resistance. The MOZART version 2 also contains both in-cloud (such as species CH3OOH, C3H7OOH, C3H6OHOOH, CH3COCH2OOH, CH3COOOH, C2H5OOH, HO2NO2, ONIT, CH2O) and below-cloud (such as HNO3 and H2O2) scavenging of trace gases. The in-cloud scavenging is parameterized according to Giorgi and Chameides (1985). For highly soluble gases, below-cloud scavenging by raindrops also is included. Several papers describing the earlier model and evaluating its performance are available, e.g., Brasseur, et al., 1998; Hauglustaine, et al., 1998.
Model Computational Time and Development of Parallel Versions
In the first year of the grant, we primarily had analyzed and tested the initial MOZART model in coordination with scientists at the NCAR. We also developed a version of the model for massively parallel computational platforms. These studies then led to the development of the next generation of the model, MOZART version 2. Initial studies focused again on testing and evaluating this model. For example, an analysis was made of the hydroxyl distribution in the model compared to other analyses of the tropospheric distribution of OH. Several areas of concern were found, but overall the model compares well with the available data analyses (Johnson, 2000). The model also compares well with available radon data in the troposphere. As discussed below, however, there remain some concerns about the ozone and NOx distributions that we still are trying to resolve.
The MOZART model is a prognostic model, which predicts the evolution of trace gas composition in the global atmosphere in the three spatial dimensions of longitude, latitude, and altitude as a function of time, given the wind, temperature, and other meteorological data. This is one of the most complete models currently available for the study of atmospheric composition and the associated chemical and physical processes of the troposphere and lower stratosphere. The meteorological data required to drive the model often is obtained from a general circulation model (GCM) such as CCM3, but the new model also can use assimilated meteorological fields.
MOZART integrates a set of partial differential equations describing the advective and diffusive transport of trace gas constituents and a number of physical removal processes, and several hundred stiff ordinary diffential equations (ODEs) describing the set of chemical reactions between the various gases constituting the atmosphere. Operator splitting is employed to solve numerically the large system of equations describing the multidimensional and multi-process physico-chemical model system. The partial differential equations (PDEs) describing the trace gas transport in the model are solved using a FFSL. The FFSL scheme was developed for implementation on spherical geometry and to conserve tracer mass, preserve gradients, and prevent negative concentration buildup. The set of stiff ODEs is solved using a fully implicit Euler scheme and a Newton-Raphson method for obtaining iterative convergent solution that requires a Jacobian. This particular set of ODEs produces a sparse Jacobian matrix, and several steps are implemented to optimize the matrix inversion for multitasking. The multitasking is achieved using OpenMP parallelization directives for compilation on the Silicon Graphics, Inc. Origin2000.
We have run several tests with of MOZART version 1, mostly of 1 to 5 model days, but including one test of a full model year, on single and multiple processors of the Origin2000. These runs permitted us to replace several features specific to the NCAR Cray C90, with coding that will work on the National Center for Supercomputing Applications (NCSA) Origin2000 array and Mass Storage System, in collaboration with NCAR and with NCSA advice. We also have tested NCAR-supplied analysis tools, which work with NCAR graphics, specifically a plotting package developed at NCAR for viewing 3-D CTM results. Several of our suggestions and improvements have been adopted by NCAR in designing MOZART version 2, which was released in November 1999 to our group for further testing and evaluation. The new version of MOZART directly generates net common data format (CDF) output files, which permits considerably easier analysis of MOZART output on our local workstations.
Figure 1 compares MOZART version 2 performance and parallelization speedup with those of MOZART version 1 for up to 16 processors. The model performance in terms of wall-clock time and number of processors used is shown in Figure 1. This figure was generated by running the model for a representative day of the model year repeatedly on the listed model configurations. All the runs were performed on 250 mHz/cpu machines in the NCSA cluster. Version 1 of the model was overall faster than version 2. This reduction in speed results from increased complexity in the model chemistry and the choice of fully implicit Euler scheme for solving the stiff ODE in version 2 compared to version 1. The relative efficiency (defined as time on one processor [T1]/number of processors (n)*time on n processors [Tp]) however, increased from 0.38 for version 1 to 0.67 for version 2. As a result, it is expected that the wall-clock time required for a 1-year model run is between 4 and 5 days with version 2. Although the time required for a version 2 model year is greater because of more complex physical processes, as discussed above, Figure 1 demonstrates that the decrease in wall-clock time with increasing number of processors is greater in version 2. Most of the calculations described herein were performed on the NCSA SGI Origin2000 cluster using 16 processors.
Evaluation of the Oxidative Capacity in the Model Troposphere
An important task in the project was to evaluate the representation of the oxidative capacity of the troposphere in the model. The oxidative capacity of both versions 1 and 2 was extensively analyzed using previous box-model based data-constrained calculations by Spivakovsky, et al., 1989 and 2000. The MOZART version 1 was found to produce up to 20 percent lower oxidative capacity when compared to CH3Cl lifetimes computed by Prinn, et al., 1996. MOZART version 2 corrects for several of these deficiencies. The oxidative capacity calculated with version 2 of the model is in much closer agreement with both the observational evaluated CH3CCCl3 lifetime and the spatial distribution patterns for OH calculated by Spivakovsky, et al., 2000.
Figure 1. Wall-Clock Time Required for a 1-Year Simulation With MOZART Model on the NCSA SGI Origin Supercluster for Various Processor Configurations
Model Evaluation Using Observational Data
The surface ozone mixing ratios in MOZART version 1 tend to be lower than the measured values at several of the CMDL surface monitoring sites in the winter months. Figure 2a shows the observed and modeled ozone at a CMDL site located in Bermuda for the summer and winter months. This points to probable shortcomings in the model transport regime as wintertime ozone levels are modulated by long-range transport and subsidence from the stratosphere.
The ozone in the model now produces higher ozone levels in the winter months at the CMDL baseline stations at the surface. Figure 2b shows a comparison of ozone calculated by the two versions of the model at selected CMDL sites for the entire year. In general, it is apparent that version 2 of the model now has more ozone than version 1 for most of the locations. At continental locations like Niwot Ridge, CO, where version 1 tended to have higher ozone, version 2 calculates lower values and is in better agreement with measurements. An extensive evaluation of the model Southern Hemisphere ozone budgets and mixing ratios using the Pacific Exploration Mission in the Tropics (PEM-TROPICS) A and B data sets also was performed. A new set of ozonesonde data was used to evaluate the vertical profiles of ozone in the model. Figure 3 shows the model-calculated version 2 and measured ozone from the ozonesonde releases during the Southern Hemisphere Additional Ozonesondes (SHADOZ) measurement campaign at Fiji for the months of April and October. The model shows similar profiles to measurements and tends to be lower than measurements during October in the mid-tropospheric region (Wei, et al., 2001).
A parallel effort was conducted to evaluate the model by comparing the model calculated NOx, CO, and NMHC mixing ratios to those measured during several past GTE field measurement campaigns. Modeled NOx was compared with the currently available NOx climatology for the troposphere. A statistical sampling space was devised to give the greatest number of measurements for each of the model grid points for this comparison. This has been used for comparing the rest of the trace gases for which significant measurements are available in the troposphere.
Figure 2. (a) MOZART Version 1 Comparison With Ozone Data at Bermuda Station; (b) MOZART Version 1 and Version 2.
Figure 3. Modeled Ozone Compared to Ozonesonde Data During the Southern Hemisphere Additional Ozone Sondes (SHADOZ) Measurement Campaign at Fiji for the Months of April and October
Single Tracer Version of the Model
A single tracer version of the model with offline OH values stored from the full chemistry simulation of the 3-D GCTM was developed for analyzing short-lived CFC replacement compounds. This model was used in a number of model evaluation studies, for example, to calculate the lifetime, distribution, and fluxes into the lower stratosphere of nPB and its degradation products.
MOZART version 2 was used to evaluate the tropical Southern Pacific using field experiments conducted by NASA, PEM-TROPICS A and B. Conducted over two different periods of the year corresponding to fall (September/October, 1996) and spring (March/April, 1999) in the Southern Hemisphere under the GTE program, these experiments were used for evaluating the model results. Other data sets for this period, including ozone profiles from SHADOZ and measurements of CO and ozone made by the National Oceanic and Atmospheric Administration (NOAA) CMDL also were used in this evaluation. The primary focus of the study was to address the following key scientific questions: What controls the ozone budget in the southern tropics? What is the significance of advected fluxes of ozone and precursors into the Northeast Pacific box and the Southwest Pacific box of the Southern Hemisphere in determining the local ozone levels? What is the magnitude of ozone in situ production/loss? What is the significance of transport from stratosphere in relation to in situ production and long-range transport from the Northern Hemisphere/South America/South Africa/Southeast Asia? What is the role of the Intertropical Convergence Zone and South Pacific Convergence Zone in controlling the inter-hemispheric transport? This study has resulted in one extensive paper (Wei et al., 2001) and several others in are process.
A New Approach to Ozone Depletion Potentials for Short-Lived Compounds
As part of this project, a new approach involving use of the 3-D GCTM was developed for redefining the concept of, and determining the Ozone Depletion Potentials (ODPs) for, short-lived replacement compounds, in particular, halocarbons with atmospheric lifetimes of less than 6 months (Wuebbles, et al., 2000a, b; 2001). Part of this work was included in a Masters of Science Thesis by Matthew Johnson. Don Wuebbles served as co-leader at a meeting sponsored by EPA and NASA on the issues relating to short-lived replacement compounds. The new approach outlined below is being adopted by the U.S. EPA in further policy considerations of short-lived compounds and also is under consideration by the Montreal Protocol an international agreement aimed at protecting global stratospheric ozone.
A number of the compounds proposed as replacements for substances controlled under the Montreal Protocol have extremely short atmospheric lifetimes, on the order of days to a few months. An important example is n-propyl bromide (also referred to as 1-bromopropane, CH2BrCH2CH3, or simplified as 1-C3H7Br or nPB). This compound, useful as a solvent, has an atmospheric lifetime of less than 20 days because of its reaction with hydroxyl. Because nPB contains bromine, any amount reaching the stratosphere has the potential to affect concentrations of stratospheric ozone. The definition of ODP needed to be modified for such short-lived compounds to account for the location and timing of emissions. It is not adequate to treat these chemicals as if they were uniformly emitted at all latitudes and longitudes, as is normally done for longer-lived gases. Thus, for short-lived compounds, policymakers will need a table of ODP values instead of the single value generally provided in past studies. Our work used the MOZART version 2 3-D CTM in combination with studies with our less computationally expensive two-dimensional model to examine the potential effects of nPB on stratospheric ozone. The 3-D CTM was used to determine the amount of the short-lived compound reaching the stratosphere as a function of where emissions of the compound might be expected to occur. Multiple facets of this study examine key questions regarding the amount of bromine reaching the stratosphere following emission of nPB.
Our most significant findings from this study for the purposes of short-lived replacement compound ozone effects are summarized as follows. The degradation of nPB produces a significant quantity of bromoacetone, which increases the amount of bromine transported to the stratosphere in nPB. Much of that effect, however, is not because of bromoacetone itself, but instead because of inorganic bromine, which is produced from tropospheric oxidation of nPB, bromoacetone, and other degradation products and is transported above the dry and wet deposition processes of the model. MOZART version 2 nPB results indicate a minimal correction of the two-dimensional results to derive our final results?an nPB chemical lifetime of 19 days and an ODP range of 0.033 to 0.040 for assumed global emissions over land masses, 19 days and 0.021 to 0.028, respectively, for assumed emissions in the industrialized regions of the Northern Hemisphere, and 9 days and 0.087 to 0.105, respectively, for assumed emission in tropical Southeast Asia.
We would have been able to do a much more complete analysis using only the 3-D CTM if it covered the entire stratosphere, but the addition of a complete stratosphere representation was beyond the scope of this project.
The following research objectives were achieved during the 3 years of funding from this project:
(1) A comprehensive global-scale model for the study of tropospheric and lower stratospheric chemistry was tested and evaluated with field observational data for a number of important trace gases of consequence to the oxidative capacity of the troposphere. The oxidative capacity relates to the ability of the atmosphere to remove pollutants by oxidation chemistry.
(2) Insights gained from our preliminary evaluation of the model with field data contributed to the development of an advanced version of this model.
(3) Important contributions were made for the development of a parallel version of the model. As a result of our analysis, the model computational performance on distributed computing machines is much faster than on single processor computers.
(4) A single tracer global scale model developed based on the full 3-D CTM was used for developing a new approach for assessing the ODP (an important concept used in policymaking to protect stratospheric ozone) of short-lived halocarbon compounds. Not only are these compounds of considerable interest to EPA, but the new assessment approach is being integrated into international policy recommendations in the Montreal Protocol to protect stratospheric ozone.
(5) The model was used to analyze ozone and oxidant seasonal variability on the tropical central and southern Pacific, a significant region providing almost a quarter of Earth’s oxidative capacity. In the process, the model demonstrated the importance of long-range transport of pollutants on the atmospheric gases in this region.
(6) This study was one of the first to clearly show the connection to, and significance of, biomass-burning emissions in South America and South Africa on the oxidative capacity of the southern tropical Pacific.
(7) Methods for comparing field data collected from aircraft platforms with results from a global scale CTM were developed and applied in the studies for this project. This capability now can be applied to evaluating the scientific understanding to be gained from future measurement campaigns.
(8) The project provided a unique opportunity for graduate students to work with complex, large-scale, cutting-edge, numerical models and advanced computational resources for the study of issues of importance to the EPA.
In addition to the publications and presentations listed below, we are in the process of writing several additional journal articles.
With the MOZART version 2 model now available as a major modeling tool, we plan to continue the analyses on the long-range transport of pollutants and their effects on the oxidation capacity of the atmosphere. Studies will be expanded to include further analyses with PEM-TROPICS A and B data, while also bringing in other data sets such as the measurements from the BIBLE campaign being sponsored by Japan. Analyses with the model also continue to evaluate its capabilities in treating convective and boundary layer processes, and dry and wet deposition. Tropospheric chlorine and bromine chemistry is being added to the model. We also are continuing our analyses of how well the model can treat short-lived chemicals of interest to policy considerations at EPA.
Journal Articles on this Report : 4 Displayed | Download in RIS Format
Other project views: | All 19 publications | 5 publications in selected types | All 4 journal articles |
---|
Type | Citation | ||
---|---|---|---|
|
Horowitz LW, Walters S, Mauzerall DL, Emmons LK, Rasch PJ, Granier C, Tie X, Lamarque J-F, Schultz MG, Tyndall GS, Orlando JJ, Brasseur G. A global simulation of tropospheric ozone and related tracers: description and evaluation of MOZART, version 2. Journal of Geophysical Research: Atmospheres 2003;108(D24):4784, doi:10.1029/2002JD002853. |
R826384 (Final) |
Exit Exit |
|
Kotamarthi VR, Wuebbles DJ, Reck RA. Effects of nonmethane hydrocarbons on lower stratospheric and upper tropospheric chemical climatology in a two-dimensional zonal average model. Journal of Geophysical Research: Atmospheres 1999;104(D17):21537-21547. |
R826384 (1999) R826384 (Final) |
Exit Exit |
|
Wei C-F, Kotamarthi VR, Ogunsola OJ, Horowitz LW, Walters S, Wuebbles DJ, Avery MA, Blake DR, Browell EV, Sachse GW. Seasonal variability of ozone mixing ratios and budgets in the tropical southern Pacific: a GCTM perspective. Journal of Geophysical Research: Atmospheres 2002;107(D2):8235, doi:10.1029/2001JD000772. [printed 108(D2), 2003]. |
R826384 (Final) |
Exit Exit |
|
Wuebbles DJ, Patten KO, Johnson MT, Kotamarthi R. New methodology for Ozone Depletion Potentials of short-lived compounds:n-propyl bromide as an example. Journal of Geophysical Research:Atmospheres 2001;106(D13):14551-14572. |
R826384 (Final) |
Exit Exit |
Supplemental Keywords:
modeling of tropospheric processes, tropospheric chemistry, tropospheric transport, global warming, nitrogen budget, semi-volatile organochlorines, oxidizing capacity, long-range transport of pollutants, ozone depletion potentials., RFA, Scientific Discipline, Air, Toxics, Ecology, Environmental Chemistry, Chemistry, climate change, CFCs, Engineering, Chemistry, & Physics, environmental monitoring, fate, global scale, atmospheric particles, global warming calculations, three dimensional transport model, global change, chemical transport modeling, organochlorides, chemical kinetics, halons, troposphere, nitrogen removal, oxidant, chemical transport modelsProgress and Final Reports:
Original AbstractThe perspectives, information and conclusions conveyed in research project abstracts, progress reports, final reports, journal abstracts and journal publications convey the viewpoints of the principal investigator and may not represent the views and policies of ORD and EPA. Conclusions drawn by the principal investigators have not been reviewed by the Agency.