A ''. United States Environmental Protection Agency Atmospheric Sciences •*« Research Laboratory "^ Research Triangle Park, NC 27711 ' Research and Development EPA/600/S3-88/017 July 1988 &EPA Project Summary Simulated Effects of Hydrocarbon Emissions Controls on Seasonal Ozone Levels in the Northeastern United States: A Preliminary Study Robert G. Lamb The Regional Oxidant Model (ROM) was developed to assist the formulation of emissions control strategies for attaining the primary ozone standard. The ROM was designed to simulate hourly con- centrations under meteorological conditions associated with the highest ozone levels during a year. Typically, these conditions span a period of one to two weeks in the summer months. Recently, efforts to promulgate a secondary ozone standard, which would protect crops and forests, have created the need for a seasonal ozone model; specif- ically, a model capable of estimating the effects of emissions changes on the frequency of multi-hour aver- aged ozone concentrations during the ozone "season," the seven- month period from 1 April to 31 October of each year. Using a method developed here, the pre- dicted ozone changes that result from emissions changes are extrap- olated from three, 2-week periods to the entire ozone season in 1980. The implementation of the extrapolation method is preliminary, but the details of the technical foundation of the method are described fully. This Project Summary was developed by EPA's Atmospheric Sciences Research Laboratory, Re- search Triangle Park, NC, to announce key findings of the research project that is fully documented in a separate report of the same title (see Project Report ordering information at back). Introduction The Regional Oxidant Model (ROM) was developed to assist the formulation of emissions control strategies for attaining the primary ozone standard Since the primary standard is expressed in terms of a maximum one-hour averaged ozone value that is not to be exceeded more than once per year, the ROM was designed to simulate hourly concentrations under meteorological conditions associated with the highest ozone levels during a year Typically, these conditions span a period of one to two weeks in the summer months A model capable of simulating concen- trations with hourly resolution is often referred to as an episodic model Recently, efforts to promulgate a secondary ozone standard, which would protect crops and forests, have created the need for a seasonal ozone model, specifically, a model capable of estimating the effects of emissions ------- changes on the frequency of multi-hour averaged ozone concentrations during the seven-month period from 1 April to 31 October of each year. The most accurate way of generating seasonal concentration sta- tistics is simply to run an episodic model for the entire season. This brute force approach is presently impractical in the case of a model like the ROM that requires 2 hours of CPU time on an IBM 3090 mainframe computer to simulate a single day. However, this technique will eventually become feasible as computing power continues to increase and com- puting costs continue to decline. A second method of modeling seasonal concentrations is to run an episodic model for selected short period intervals within the season, and then "extrapolate" the results to the entire season. This is the approach that we will develop in this paper. This method also requires the adoption of basic assump- tions, but their impact on the accuracy of the model overall can be controlled to a large extent through proper design of the extrapolation scheme and judicious selection of the subinterval episodes. The principal objective of this project is to describe estimated effects of VOC emissions changes on seasonal ozone levels in the Northeastern United States. The study is based on ROM simulations of three, 2-week episodes in April, July and August of 1980, extrapolated to the ozone season, 1 April - 31 October, using a scheme developed here Each of the three episodes was simulated (and the results extrapolated to the season) for each of three emissions distributions: the base, or reference, case; and two "control strategies." The latter consist of different percentage changes in VOC emissions that could be achieved under control plans contained in the 1982 SIPs. The model results are regarded only as interim estimates, partly because they are based on a very simplified imple- mentation of the seasonal extrapolation scheme; and partly because the 1980 emissions inventory, which is used as the base state in these simulations, is not regarded to be of high enough quality to serve as a basis for control strategy development. Furthermore, the verifi- cation analyses of the ROM are not yet complete, and therefore the accuracy of the model has not yet been established. Both the ROM verification studies and the development of a significantly im- proved emissions inventory (for 1985) are expected to be completed by late 1988 Extrapolating the Results of Simulated Emissions Control Strategy Effects from an Episode to a Season Next, we consider the problem of extrapolating concentration changes, or delta concentrations as we shall refer to them, simulated by an episodic model over a period of 3 to 4 weeks to a season consisting of 7 months. First we define basic terms: Season = 7-month period from the beginning of April to the end of October; Episode = the period, usually 3 to 4 weeks in length and not necessarily continuous, for which model simulations of emissions controls are performed. Considering only species A for the purpose of discussion, let AA(x) be some average concentration difference or some net benefit of controls predicted by the model for a given point x in the episode. Our task is to estimate the value AA(x,t) that the model would predict at the same point averaged over the season. Strictly speaking we are not concerned at this point with estimating what the true value of AA(x) actually is. This question is related to the issues of model accuracy, predictability, and so on which are not topics of concern in this analysis. The principal objective is the development of a method of extrapolating model results outside the simulation period irrespective of the accuracy of the model. There are three basic problems that obstruct the extrapolation of the model results from the episode simulation to the entire season (1) differences in the meteorology in the episode and season, (2) differences in the chemistry of the delta concentrations in the episode and season, and (3) differences in the emissions changes AS within the season. The last problem is generally not significant because any changes in emissions that are required under a control plan apply equally throughout the year. When one considers the problems created by differences in the meteorol- ogy and differences in the chemistry, it appears that the largest extrapolation errors will occur when the receptor in question is influenced strongly during the season by emissions sources other than those that impacted it during the episode. This suggests that if the estimation of seasonal effects is the principal goal of the model simulations, the episodes should be selected so that they contain at least one sample of each of the meteorological states that cause the major source areas in the vicinity of a given receptor to affect the concen- trations that receptor observes. A Proposed Extrapolation Procedure Consider three inert tracer species, T-), T2, and T$ emitted by three different source areas in the model domain. For any receptor site x in the model domain and any hour t in the season we define a binary number B(x,t) = (b-|,b2,b3) whose elements bn are binary numbers defined as follows: bn = 1 if the concentration of Tn at (x,t) is nonzero; bn = 0 otherwise. Thus, B(x,t) = 010 if TT = T3 = 0, T2*0 at (x,t), = 000 if TT =T2 = T3 = Oat (x,t); = 111 if T^Ta, T3*0at (x,t); etc Keep in mind that the function B is derived from the output of the model produced by a simulation of the three inert tracer sources over the entire season with no chemical reactions Consider a receptor, R, at which we are interested in extrapolating the results of simulated control strategy effects from an episode to the full season. The influence of the three source areas on concentrations at R over the season can be summarized in the form of a Karnaugh table as shown in Table 1 Here we are Table 1. Representation of the Function B Evaluated at Receptor R Inte- grated Over the Season. Numbers Indicate the Percentage of Hours in the Season When B(x=R,t) Had the Value Indicated. A Dash (-) in the Table Indicates Less than 1 Percent Occurrence of that Value bi 0 t 00 01 55 24 11 15 2 10 4 - using the Karnaugh table (which is borrowed from logic analysis) as a basis ------- for expressing the potential influence that each major source and combinations of major sources can have on a given receptor. The numbers in Table 1 are the percentages of time in the season that the function B, as determined by the model predictions, had the value specified by the row and column coordinates. In order to display B, which has three elements bt, bg and b$, on a 2-D chart, element b-| is split out to form the row entry to the table and the pair b2b3 make up the column entry. Thus, the number 55 in the upper left corner of Table 1 indicates that during 55% of the season all three tracers T-| ,T2 and TS were simultaneously absent at R, i.e., b-|b2b3 = 000. The number 24 just below the 55 indicates that 24% of the season, T-| was predicted at R but at those times neither T2 nor T3 was present (B = 100). Similarly, T2 and T3 were present at R with T-| absent (011) 15% of the time; T2 was present with T-| and T3 absent (010) 4% of the time; and on 2% of the hours in the season all three tracers were predicted at R simultaneously. The total of the entries in the table is 100 percent. We can construct tables like this for any number of sources, although for numbers larger than n = 8 the tables become quite unwieldy. Note that the number of elements, or slots, in a table is 2n, where n is the total number of source areas represented. We define the dominant entries of a table as the first M members of the rank-ordered set where M is the number such that the sum of the first M entries is closer to some limit, say 90%, than the sum of the first M-1 or m +1 entries. Using the example shown in Table 1 as an illustration, the rank ordered entries are 55, 24, 15, 4, 2. The sum of the first three of these is 94, which is closer to 90 than the sum of the first two (55 + 24 = 79) or the first four. Hence, 55, 24 and 15 are the dominant entries. We define the dominant elements of the table to be the elements associated with each dominant entry, and we indicate the dominant elements by checks in a dominant element table. Table 2 shows the dominant elements corresponding to Table 1. The Karnaugh tables that we have discussed up to this point were compiled using the inert tracer concentrations simulated by the model for the entire season of interest. Thus, we will refer to these as seasonal tables and to their elements as B = (bib2ba). By applying the same procedures to any subset of Table 2. Dominant Elements, Indicated by an X of the Karnaugh Table, Table 1, for Receptor R bi 0 1 00 01 X X 11 X 10 the simulated tracer concentrations over the season, we can produce episode tables. We will refer to the elements of an episode table as B' = (b'ib'2b'3).We emphasize that the subset of the season that is used as the episode does not have to be continuous in time. One can select one time interval at one point in the season and intervals of different lengths at different times and treat the combined set as a single subset or episode. We also want to emphasize that the seasonal and episode tables pertain to specific receptor sites, and in general the entries in either table will vary from receptor to receptor We now advance the following hypothesis: The simulated effects of emissions controls can be extrapolated from an episode to a season for any receptor site whose episode table covers its seasonal table. We say that Table A covers Table B if all dominant elements of B are also dominant elements of A. Thus, in simple terms this hypothesis states that the extrapolation process is possible if the dominant element table of the episode contains an X in every box that the seasonal table has one. The episode table may contain more checks (dom- inant elements) than the seasonal table, but this is not necessary to achieve coverage. Suppose that we have found an episode whose dominant elements cover those of the season shown in Table 2 Since there is a unique value of B' = (b'-ib^b's) at R for each hour of the episode, we can identify all those hours in the episode associated with each of the three dominant elements (b'-|b' b's) = (000), (011), and (100). Let Epoq be the predicted total benefit of emissions controls at R for all hours associated with (000) in the episode, and let E0n and EIQO be the corresponding benefits predicted for all hours associated with (011) and (100), respectively. The benefit might be the average delta concentration, e.g., where the summation is over all K hours t|< in the episode for which (b'ib'2 b'3) = 011, and AA is the predicted change in the concentration of a particular species; or the benefit might be some dollar value of crop yields. In any event, the corresponding benefit E at R for the season is extrapolated to be i(R)=eoooEooo+eoiiEon + eiooEioo where e is the percentage occurrence of (bib2D3) = (000) in the season, as given in the present example by the entries m Table 1 (6000= 55). Similar meanings apply to 6011 and e-ioo (eioo = -24. eon =.15). This is the essence of the proposed extrapolation technique. In practice we usually do not have the luxury of being able to change the modeling episode at will. Often the episode has already been selected and modeled, and our task is to extrapolate the results of that simulation to the season for any or all receptor sites in the model domain. In general we will find that the episode table does not cover the seasonal table at every receptor At those locations we can attempt to achieve coverage by one of two methods (1) we can ascertain whether those dominant elements in the seasonal table that are not covered by elements in the episode table are irrelevant insofar as the eval- uation of the control strategy of interest is concerned, and therefore are "remov- able" elements; or (2) we can attempt to "fill in" necessary elements in the episode table by borrowing from the episode tables of other receptors. Interim Implementation of the Extrapolation Scheme A rigorous implementation of the seasonal extrapolation method just described would require that the ROM generate the inert tracer concentration fields, used to define the states B, for the entire season. This is not possible at the present time, so we present here an interim implementation of the scheme that utilizes the ATAD trajectory model. We also adopt several strong assump- tions to simplify the analysis. These include the neglect of vertical wind shear, vertical turbulent diffusion, and vertical air motion including the vertical motion associated with convective clouds We also resolve the spatial and temporal variations in the tracer concentration fields with limited resolution. We treat blocks of 5x5 ROM cells as single "receptors"-this has the effect of smoothing the tracer concentrations over ------- roughly 100 km subdomains-and tracer concentrations are sampled every 6 hours rather than every 30 minutes as the ROM would provide. In order to minimize the data handling requirements of the interim scheme, we consider only 27 source sites, selected because they are the locations where the largest absolute reductions in VOC would occur under the control strategy identified as #1 in the on-going rural ozone studies. Beginning at hour 0000 Z on 1 April, and at every subsequent six-hour interval 0006, 0012, etc., until 2400 Z 31 October, a particle is released from each of the 27 source sites and tracked by the ATAD model for a period of 5 days, or until it passes outside the model domain. Every six hours a count is made of the particles in each receptor. These consist of 98 rectangular cells arranged in a 12x9 array. When the particle counts are performed, a separate running tally is maintained for each of the 27 sources. At the end of the season, the source with the highest total count at a given receptor is the source that affects that receptor most frequently. The source with the next highest count is the second most significant source, and so on. In this way we identify for each receptor (I,J) the eight most significant sources. This information is then used to define the state variable (B(l,J,tn) = bsbybgbsb^babi described earlier. Recall that the bk are binary numbers with b|< = 1 if at time tn, where tn is any hour in the season, tracer material is present at receptor (I,J) from the k-th significant source. Otherwise, bk = 0. For example, if at a given time, tracer is present from the most significant source (k = 1) simultaneously with material from the second and eighth most significant sources, then B(l,J,tn) = 10000011. If there were no tracer material present at tn from any of the top eight sources, B(l,J,tn) = 00000000. Note the following important aspects of B: (1) it is a discrete valued function with 28 = 256 possible values; (2) the specific sources (of the set of 27) represented by each bk may differ from one receptor to another; (3) the most significant source is represented by the rightmost bit of B while the 8th most significant source is represented by the leftmost bit; and (4) at each receptor there is a unique value of B associated with each hour of the year. As we noted earlier the ozone season is currently defined as 1 April through 31 October In the present study the episode consists of the following three disconnected periods: 14-29 April, 12-26 July, 20-31 August. Table 3 contains the seasonal/episode state tables for two receptors, (I,J) = (8,6) and (8,7), computed by the procedure described above using meteorological Table 3. Sample Episode/Seasonal State Table for Two Receptor Sites (I.J) See Text for Explanations of the Table Entries SEASONAL MID EPISODE KARIAU SEASON* 11 AFRIL-11 SEASON SUM- 156 TABLE FOR RECEPTOR OCTOBER 1»IO) ROW LABEL (1.1,1,1) REPRE 111. EENTS SOURCES 5, 1, 7) / oooo 0000 / 17:419 19 «001 / 2:11 0010 / 4:13 0011 / 3 0100 / 5:19 0101 / 1:5 0110 / 1 0111 / 1 1000 / 1:20 1001 / 1: 1010 / 1: 1011 / 1100 / 1101 / 1110 / 1111 / 0001 0010 0011 :102 1:31 2:19 2 2 1:5 1:1 1:1 1 ( I- 1 J- « ) EPISODE-I 12-24 EPISODE stm- is» COLUMN LABEL (1,1,1 0100 0101 0110 (:41 1 1 2:7 1:5 2 1 1:1 1 1 1:5 1 1 1 JULI , 20- ,1 IKE PRESENTS 0111 1000 2:25 1:2 1:2 1 1:1 1 1:1 1 1 AUGUST, 1 SOimCES(10.17, 1001 1010 2:10 7 2 1 1:2 1 1 4-29 APRIL, If 10) 12.1O 1011 1100 1101 4:19 1:3 1 1 1 1 1:1 1 1110 1111 1 1 1 1 SEASONAL AND EPISODE KARNAU TABLE FOR RECEPTOR (I- • J- 7 ) SEASON.(1 APRIL-11 OCTOBER 1910) EPISODE-I 12-24 JULY, SEASON SUM- 156 EPISODE SUN- 159 ROW LABEL (1,1,1,1) REPRESENTS SOURCES ( 2,17, 5, 9) / 0000 / 95 0001 / 0010 / 0011 / 0100 / 0101 / 0110 / 0111 / 1000 / 1001 / 1010 / 1011 / 1100 / 1101 / 1110 / 1111 / 0000 :500 4:27 1:4 1:1 1:24 4 1:1 1:4 1:1 1 0001 (:66 1:5 1:4 1:2 1:1 1 COLUMN 0010 0011 0100 1:26 5:20 3:21 1:4 1 2 1 2 1 1 111 1:2 LABEL 0101 7:11 1:2 1 1 (1,1,1 0110 3:10 1:5 1 1:2 1 1 ,1 (REPRESENTS SOURCES! 7,10,12,16) 0111 1000 1001 1010 1011 1100 10 2:14 1:3 2:9 2 2:2 1:6 1 2 1 2 1 1 2 1 1 1:2 1 1 1:1 1:1 1101 1110 1111 1 1 1 1:1 1:1 1 1 ------- data for 1980. The column entry of each table is the rightmost four bits of B(I,J), i.e., b4D3b2bi, which represent the four most significant sources at receptor (I,J); and the row entry of each table is the leftmost four bits, bsbybebg. The specific sources represented by each bit are shown just above the column and row headings. Each table has three types of entries: (1) two numbers separated by a colon, (2) a single number, or (3) a blank. A blank signifies that the state B represented by that particular column/ row entry in the table did not occur in the season, nor the episode. A single num- ber signifies that the state occurred in the season but not the episode. In this case the integer shown is the number of occurrences, in units of quarter days (6 hours), of that state in the season. Fin- ally, an entry of the form n:m indicates the number of occurrences of the state found in the season (m) and the episode (n). The state tables provide an easy means of assessing how well the chosen episode represents the season, at least insofar as the source impacts are concerned. Since there is a unique value of B associated with each hour, regardless of whether the hour is inside or outside the episode, for extrapolation purposes we can equate unmodeled hours in the season to modeled hours in the episode using the B values as the basis. Another way of saying this is that for any receptor (I,J) we can map the simulated effects of emissions controls at a given hour tn of the episode into any hour tm of the season as long as B(l,J,tn) = B(l,J,tm). Table 4 shows the mapping for the first 12 days of the season at 12 receptors (I.J), 1 si s12, J = 2, derived from the interim implementation of the extrapolation scheme that we have described to this point. The second column from the left contains 5 digit numbers of the form DDDHH where ODD is the Julian date in the season (1980), and HH is the ending hour (EST) of a 6-hour block. Columns 3-14 contain the corresponding day and hour in the episode from which model results can be used to estimate benefits at time ODDHH at the receptor identified at the top of the column. The mapping relationships shown in Table 4 were generated under the following constraints: (1) The 6-hour morning period ending at 1300 EST can be extrapolated Table 4. Episode-to-Season Extrapolation Matrix for the 12 Receptors Indicated by the Column Headings for the First 12 Days of the Season. See Text for Further Explanation NJWIM MCTOU rot uccms 1 M n tcu- | l_ 10 11 12 1) 14 IS 1C n 11 i» 20 21 22 ZJ 24 JS 2( 27 21 2* 10 11 32 1) 14 IS 1C 17 11 it 40 41 42 41 44 45 4( 47 41 4* XME-( ODOUR) O-JULIAX DATI -(-IOU*. rtXIOD (CST) UDIin *201 *207 *211 1219 »}01 *307 *313 1119 >401 9407 »413 Ml* 9S01 9S07 9S13 9519 9(01 9607 9613 9(19 9701 9707 9711 9719 9101 9107 9111 9119 9901 9907 9911 9919 10001 10007 10011 10019 10101 10107 10113 10119 10201 10207 10211 10219 10101 10107 10111 10119 10401 1 1 | ( 1,2) 10*01 10(07 10(11 10(19 10701 10707 1(713 10719 10101 11001 11013 1111* 11201 11301 11411 11419 11501 11(01 11(11 11(19 11701 12001 19(11 1971* 19101 19(07 19111 19119 19*01 19907 1*91) 19919 20001 20007 2001) 2001* 20101 20107 2011) 2011* 20201 20207 2*21) 20219 20)01 20)07 20111 2011* 20401 I 2,2) 11207 11401 11411 1141* 11(01 11(07 11*1) 1*41* 19(01 20107 * 2)31* 2)401 21407 21411 0 21*07 24001 24011 24019 24101 24107 24111 2411* 24101 24107 24111 24119 24401 10(07 10(11 10(19 10701 10787 10713 10919 11107 11201 11111 11319 11401 11407 11411 11419 11(01 11(07 11(11 1*41* 20001 ( 1,2) 10(07 11(07 20S1) 21119 23401 23*01 21*13 24219 23(01 0 0 2)71* 23101 23107 23113 2421* 24301 10707 10713 1111* 11407 12007 20013 2021* 20301 10107 20313 2041* 23101 21*07 21111 2421* 24101 24107 2411) 24)19 24401 10(07 10(13 10(1* 10701 10*01 11313 1171* 1*701 1*707 1*713 1*719 19101 ( 4,2) 10(01 19(01 20111 24119 24201 24207 11113 11519 11(07 1 0 1*41* 19501 19S07 20111 20119 21401 23701 21111 21119 21901 11(07 11(11 11(19 1*501 19S07 20111 2011* 20201 20207 20211 2021* 2*101 20307 20314 2041* 20 SOI 20107 23513 23S1* 24101 24201 11113 1141* 1*501 19507 20113 2011* 20201 1 S.2) 10(01 11707 1*513 1*519 19707 20107 20213 24119 1*707 0 1141) 11*1* 12001 1*501 1*(13 19(19 19701 1*101 11713 0 12001 1*501 1X13 11(19 20107 20201 20213 20219 20301 20307 2*311 2031* 20407 20501 20713 2341* 23S01 23S07 23S13 23S1* 23(07 24201 24213 2421* 24307 24401 10(13 10(11 10701 I (,2) 12001 19(01 20(13 23111 0 24107 24113 1111* 11701 12001 2341) 0 23S07 23(01 24013 2411* 24401 10707 10713 1071* 10101 20(01 20713 0 24301 24401 10(13 10(1* 10701 10001 10*11 1101* 11107 11101 1*71) lill* 19907 20107 20211 2021* 20101 20307 20313 20S1) 0 24301 24313 2431* 24401 ( 7,2) 0 0 23(11 2141* 19(07 0 1*111 1**19 20701 21407 21711 21519 24101 24107 24111 24119 24401 10(01 10(11 10719 11101 11107 0 24119 20501 20101 24111 24119 24401 10(01 10(11 10(1* 10701 10707 11111 11*1* 11*01 11*07 20111 2041* 20(01 24001 20111 2041* 20(01 24201 20111 2041* 20101 ( *,2I 10(07 10(11 10(1* 10101 10*01 11313 « 0 1*(01 19(11 11119 11401 11(01 11(11 11(19 11707 0 11111 11(1* 0 10*07 10*11 0 0 20107 20111 2031* 20401 20407 20413 2041* 20501 20507 20S1) 20519 20*07 23407 2)413 1101* 11101 0 « 1»«1» 1*901 1*907 20113 2031* 20401 I »,2) 10(07 10(13 10(11 10701 11001 11013 2071* 20107 23401 23*13 !*(!> 1*101 1*107 1991) 1*919 2*007 10707 11013 11019 19707 24001 0 10719 10*07 0 11*13 11*1* 11101 11201 11513 11(1* 11701 11707 11711 11*1* 11*01 11*07 11*11 11*1* 12001 1M01 0 2401* 24401 0 10(11 10(19 10701 11*, 21 10(07 10(11 10(1* 10701 10(07 10*11 0 11107 1*507 19513 19519 20(07 21401 21411 21419 21S01 0 19(11 0 19507 11001 0 0 0 24007 0 24219 24)01 24107 24313 24319 24401 10(01 10(13 10(19 10701 10*07 10113 10(1* 11007 11101 0 24019 24101 24301 24313 24319 24401 111,2) 10(07 10(13 10(1* 10701 10707 10113 1101* 0 11101 11111 1111* 11401 11507 11511 11519 0 19907 0 1*419 19701 19707 0 19719 10901 11)01 11411 11419 11507 11(01 11(13 11(1* 11701 11707 1171) 1171* 11101 11(07 11(13 11*1* 11*01 11*07 11*11 24019 11107 0 0 1111* 114*1 (12,2) 10(07 10(11 10(1* 10701 10707 10(11 10(1* 0 11007 0 !*(!» 1*701 1*707 1*713 1*719 19*01 19(07 24011 24119 24201 24207 24211 19(19 19701 19707 19711 19719 1M01 1M07 1*(11 1*11* 1*901 19907 19913 lt*li 20001 20007 20113 2011* 20201 20207 20211 0 20101 0 0 2011* 2*401 ------- only from 6-hour morning periods in the episode. (2) The 6-hour afternoon period ending at 1900 EST can be extrapolated only from 6-hour afternoon periods in the episode. (3) The nighttime period 1900-0700 can be extrapolated only from nighttime periods in the episode. These constraints help maintain accurate diurnal variations in concentration pat- terns. To help maintain realistic temporal variations over multiple days, the episode states B(l,J,tn) are examined sequentially in forward time order, i.e., tn increasing in unit step, from the point tno, say, that was extrapolated into the immediately preceding quarter-day of the season. To help maintain realistic seasonal variations in concentration only the April episode period is used to fill in the Spring (1 April - 20 June) and autumn (20 September - 31 October) months, and only the July and August episodes are used to fill in the Summer (20 June - 20 September). The "0" entries in Table 4 indicate that the state 6(l,J,tn) prevailing at the given receptor and time in the season did not occur in the episode, under the constraints described above. There are innumerable ways to "fill in" these gaps of missing values. Here we have chosen simply to march across the gap by continuing the episode time sequence that was last in effect. If in this process we reach the end of the episode period before we have completely crossed the gap, "we wrap around" and continue the interpolation from the beginning of the same episode period. In cases where the season begins with missing states, as is the case at receptor (7,2) as indicated in Table 4, the first mapping time is taken from any one of the closest receptors whose factor is not missing. In this way we obtain a complete mapping of episode into the season. The extrapolation sequence derived for each receptor and each hour in the quarter-day intervals in the episode is mapped one-to-one into the hours in the season. When the final values are applied to the concentration differences predicted by the model for given emissions changes, the outcome is hourly values of delta concentrations at each hour of the season in each ROM grid cell. These values may then be processed in any desired manner to arrive at estimates of the seasonal impact of given emissions changes. ------- RG. Lamb is on assignment to the U.S. Environmental Protection Agency from „ the National Oceanic and Atmospheric Administration, U.S. Department of Commerce. K.L Schere is the EPA Project Officer {see below). The complete report, entitled "Simulated Effects of Hydrocarbon Emissions Controls on Seasonal Ozone Levels in the Northeastern United States: A Preliminary Study," (Order No. PB 88-195 706/AS; Cost: $14.95, 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/S3-88/017 0000329 PS U S EHVIR PR0TSCTIQH AGENCY REGION $ LIBRARY 230 S DEARBORN STREET CHICAGO It 60604 ------- |