United States Environmental Protection Agency •«k:' Atmospheric Sciences Research Laboratory x/ , Research Triangle Park NC 27711 ' Research and Development EPA/600/S3-86/038 Sept. 1986 ŁEPA Project Summary Numerical Simulations of Photochemical Air Pollution in the Northeastern United States: ROM1 Applications Robert G. Lamb The first-generation Regional Qx\- dant Model (ROM1) was used to simu- late pollutant concentrations during the nine-day period 23-31 July 1980. Two simulations were performed. The first, which is considered to be the base case, used the 1980 NAPAP 4.2 inven- tory for all hydrocarbon and NOX emis- sion rates. The second simulation, or control case, was identical in all re- spects except that the county-by- county hydrocarbon and NOX emissions rates were modified in accordance with baseline projections for 1987 contained in State Implementation Plans (SIPs). The one-hour and daily daylight (0900- 1600 1ST) averaged ozone concentra- tions produced in each simulation were compared to assess the effectiveness of the proposed emissions changes on air quality. Ozone concentrations in the control case were found to be everywhere lower than those in the base case, but the percentage reduction was not uni- form in space. In areas near the major VOC and NOX sources, the maximum one-hour averaged ozone levels were reduced by about 25% while in areas farther than 100 km from these sources peak values were only about 10% lower. Slightly smaller percentage re- ductions were found in the daily day- light average ozone concentrations, ft was also found that the emissions re- ductions lowered peak ozone concen- trations by considerably larger percent- ages than they reduced the median or mean concentration values. The analyses of the model results are prefaced by discussions of a number of basic issues on regional scale model- ing, including model initialization, se- lection of meteorological data, effects of grid size on model performance, esti- mating long-term concentration statis- tics from short-period simulations, probabilistic vs quasi-deterministic modes of model operation, uncertainty in emissions estimates, and the charac- teristics of VOC and NOX sources in the Northeast, among other topics. Prelimi- nary results of analyses of the SAROAD monitoring data, which reveal the char- acteristics of the ozone problem in the Northeastern United States, set the stage for the model simulations. This Project Summary was devel- oped by EPA's Atmospheric Sciences Research Laboratory, Research Triangle Park, NC, to announce key findings of the research project that is fully docu- mented in a separate report of the same title (see Project Report ordering infor- mation at back). Introduction The development of the Environmen- tal Protection Agency's Regional Oxi- dant Model (ROM) began in the late 1970's as a part of the Northeast Corri- dor Regional Modeling Project (NECRMP). The NECRMP was initiated out of the recognition that the adverse ozone concentrations observed in the Northeastern United States are due in large part to the regional transport of ozone and its precursor species. The ------- principal role envisioned for the ROM in this project was to assist the states in developing emissions control plans that would effect compliance with the Fed- eral ozone air quality standards in the most equitable and cost effective way. This report describes the second of a series of applications of the Regional Oxidant Model in this role. The report considers projected 1987 emissions, based on 1982 State Implementation Plans (SIPs), and compares the ozone concentrations simulated using these emissions with the corresponding con- centrations predicted using 1980 emis- sions data. Two questions are of pri- mary concern: (1) what impact will proposed VOC and NOX emissions con- trols, which were designed to attain the primary ozone standard in urban areas, have on ozone levels in rural and re- mote regions?; (2) what impact will these emissions changes have on longer period ozone averages, such as the 7-hour daily daylight (0900-1600 LST) average presently being consid- ered as a possible basis for a new sec- ondary ozone standard? The only way of obtaining answers to questions of this kind prior to implementing emis- sions controls is through models such as the ROM which simulate the meteo- rological and chemical processes that govern the evolution of air pollution episodes. Procedure The development and initial testing of the Regional Oxidant Model (ROM) were described in three earlier reports: A Regional Scale (1000 km) Model of Photochemical Air Pollution: Part 1. Theoretical Formulation, EPA-600/3-83- 035, May 1983; Part 2. Input Processor Network Design, EPA-600/3-84-085, Au- gust 1984; and Part 3. Tests of the Nu- merical Algorithms, EPA-600/S3-85-037, June 1985. The version of the model used in the present study is the first- generation ROM, referred to as ROM1, which differs from the ultimate second- generation model (ROM2) in several key respects. First, ROM1 utilizes the 23- species, 36-step Demerjian-Schere chemical mechanism whereas ROM2 will employ Carbon Bond-IV, which treats 70 reactions among 28 chemical species. The latter mechanism provides explicit treatment of the biogenic hydro- carbon species isoprene. To make use of this provision, ROM2 will use an emissions inventory in which anthropo- genic source data are supplemented with gridded estimates of the fluxes of hydrocarbons generated by natural sources. Another major difference be- tween the first- and second-generation models is that the layer thicknesses in ROM1 are constant in space and time and the winds are horizontally non- divergent. In ROM2, the layer thick- nesses will vary spatially and tempo- rally to keep track of meteorological phenomena, and the winds will be non- divergent, allowing for large scale verti- cal motion. Due to these basic differ- ences between ROM1 and ROM2, the principal objectives of the former are to provide experience in operating large, complex regional scale air pollution models, and to gain some preliminary insight into the effects proposed emis- sions changes are likely to have on ozone concentrations in the Northeast. These are the general topics of the present study. The primary objective of, ROM2 will be to provide a credible basis for formulating regional emissions con- trol policies for ozone attainment. In the course of applying ROM1 con- siderable knowledge was acquired in dealing with problems that are unique to regional scale photochemical mod- els. One of these is the problem of model initialization. In the case of ROM1, initialization means specifying the concentrations of each of 23 chemi- cal species at each of some 7500 grid cells at the hour the model simulation is to begin. Unfortunately, this requirement vastly exceeds the information content of presently available air monitoring data bases. Within the present ROM do- main, measured hourly ozone concen- trations are available at about 150 sur- face sites; lower quality data are available for NOX at fewer sites, and no measurements are available for any of the remaining 23 species. Moreover, no data are available on the concentrations of any species above ground level. Thus, in practice initialization requires reconstructing the 3-D spatial structure of the concentration fields of 23 com- pounds given the concentration of only two of the species at a few surface loca- tions. If the nature of the pollutant chemistry were such that the concentra- tions of all species were unique func- tions of the concentrations of the two given species, then initialization would not be a problem. But this is not the case. In fact, it appears that the ozone concentration that evolves from a given initial mix of species is quite sensitive to the initial levels assumed for hydrocar- bon and NOX, and as a consequence anomalous ozone concentrations may arise in the course of the simulation that are merely artifacts of the initialization. This problem is made all the more acute in a regional model by the prolonged residence time, 4-5 days or more, of the initially present species within the model domain. Similar problems arise in specifying concentrations at inflow boundaries. Model simulations per- formed in control strategy studies run long enough that species that enter the domain through the inflow boundaries eventually permeate the model area to nearly the same extent that the initial concentration field does. After consider- able study of the problem of recreating concentration fields from sparse data and of the sensitivity of model predic- tions to uncertainties in the initial and boundary concentrations, it was con- cluded that the simplest and perhaps most reasonable approach is to assume "clean" tropospheric values for both the initial and boundary concentrations, at least in model simulations whose only aim is to compare the effectiveness of given emissions controls. This method is viable only if the model do- main is large enough to encompass the majority of the sources that affect air quality in the areas of interest. Because in this case the influence of trans- boundary fluxes on species concentra- tions at receptors in the interior of the domain is at most a second-order effect and well below the level of tolerable error in the predictions. Applications of ROM1 also yielded in- formation on procedures for selecting meteorological data, interpreting model results, and other important aspects of regional model applications. This knowledge will be valuable in the de- sign and execution of ROM2 applica- tions that are planned to support emis- sions control policy formulation. Conclusions Following are some of the major find- ings of this study. (1) Based on the 1980 NAPAP version 4.2 emissions inventory, 70 counties in the Northeastern U.S. were iden- tified as major sources of VOC and NOX. These counties have the highest emissions densities (moles per area per day) of VOC and NOX of all counties in the region, and to- gether they produce about one-half of all VOC and NOX emissions. The areas of highest measured ozone concentrations are closely associ- ated with the locations of these 70 counties. ------- (2) The characteristic spatial scale of the major VOC and NOX sources in the Northeast is estimated to be a few tens of kilometers. This means that if a model with a grid size much larger than this is used to simulate photochemical air pollution in this region, significant systematic errors can occur not only in the predicted peak concentrations of secondary species, such as ozone, but also in the predicted response of the con- centrations of these species to changes in VOC and NOX emissions. The latter point is of critical impor- tance in the use of models in regula- tory studies and requires further de- tailed analysis. (3) A comparison of the 1980 NAPAP version 4.2 emissions inventory with the earlier 1979 NECRMP in- ventory revealed differences of 300 percent and more in the gridded (-18 x 18 km) VOC and NOX emis- sion rates. This represents a level of uncertainty in the base emissions that is roughly ten times the magni- tude of the changes in emissions that are contemplated in present control strategies. (4) Air quality models that treat re- gional scale and larger areas, i.e., domains > 1000 km in extent, can be operated in either of two modes, which are called the probabilistic mode and the quasi-deterministic mode. In the former, the model predicts the probabilities, expec- tations and other statistical prop- erties of concentrations at specific sites at specific times. In the quasi- deterministic mode, the model pro- vides statistics of concentrations at given times or integrated over given periods within given receptor classes rather than at specific sites. In the present study, the model is run only in the quasi-deterministic mode, yielding information on the concentrations of 23 different chem- ical species in four receptor classes —Urban, Suburban, Rural, and Wilderness. (5) Two criteria were tentatively pro- posed for selecting historical me- teorological data for use in re- gional scale modeling studies of photochemical oxidant: (1) the meteorological scenario should begin on a day when the median value of the maximum hourly ozone concentrations observed at all measuring sites in the model domain is near the seasonal mini- mum value; (2)the scenario should be long enough that the frequency distribution of mea- sured hourly ozone values during the scenario period approximates the corresponding seasonal distri- bution closely enough to give the results of the model simulations broad applicability. Moreover, the scenario must be more than about 5 days long, to minimize the ef- fects of the initialization procedure on predicted concentrations in re- mote areas. A 9-day scenario is ap- parently not long enough to model the processes that control concen- trations above the 90-th percentile level at any site. (6) Two 9-day simulations were per- formed with ROM1 using meteoro- logical data from 23-31 July, 1980. One simulation, the base case, used the 1980 NAPAP version 4.2 emis- sions data as input. The other simu- lation, the control case, use pro- jected 1987 baseline emissions. Comparisons of the predicted ozone concentrations in the base case with corresponding values in the control case revealed the following data: (a) In general, at any given loca- tion and hour the concentra- tion in the control case is less than or equal to that in the base case; (b) Within each receptor group, peak concentrations are re- duced by larger percentages than the median values are reduced. Reductions of 0 to 50% occur in the peak com- pared to 0 to 25% reductions in the median. This is true of all concentration averaging times. (c) Ozone is reduced more at sites near the major VOC and NOX sources than at locations far away. For example, me- dian ozone levels at suburban locations (defined to be within 50 km of major source centers) were reduced up to 25% whereas in Wilderness areas (greater than 100 km from major sources) reduc- tions were less than 15%. (d) The maximum ozone concen- trations in rural and wilder- ness areas occurred on differ- ent days in the control case than in the base case, even though meteorological condi- tions were identical in both simulations. This suggests that the source-receptor rela- tionship between VOC/NOX sources and remote sites is strongly nonlinear. (e) Overall, the 1987 emissions reductions appear to have two basic effects on ozone: they cause a delay in ozone formation, and they reduce the total quantity produced. ------- The EPA author, Robert G. Lamb (also the EPA Project Officer, see below) is with Atmospheric Sciences Research Laboratory, Research Triangle Park, NC 27711. The complete report, entitled "Numerical Simulations of Photochemical Air Pollution in the Northeastern United States: ROM 1 Applications," (Order No. PB 86-219 201/AS; Cost: $16.95. subject to change) will be available only from: National Technical Information Service 5285 Port Royal Road Springfield, VA 22161 Telephone: 703-487-4650 The EPA Project Officer can be contacted at: Atmospheric Sciences Research Laboratory U.S. Environmental Protection Agency Research Triangle Park, NC 27711 United States Environmental Protection Agency Center for Environmental Research Information Cincinnati OH 45268 Official Business Penalty for Private Use $300 EPA/600/S3-86/038 ps ------- |