January 16, 1990 MEMORANDUM SUBJECT: Recent Texas Air Control Board (TACB) Evaluation of the ISC Area Source Algorithm FROM: Joseph A. Tikvart, Chief Source Receptor Analysis Branch (MD-14) TO: James W. Yarbrough, Air Modeling Contact Region VI (6T-AN) As you are aware from telephone conversations you've had with the Model Clearinghouse over the last several months, we have been looking into several of the issues that the State of Texas has raised in conjunction with modeling area sources with ISC. Also we have been awaiting the results of a contractor study on area source algorithms for several models. That report is now complete and copies were sent to the Regional Modeling Contacts last month. In addition to the TRC report that you already have, I am attaching copies of two memoranda written by Russ Lee. These memoranda deal more directly with the issues that Texas has raised. Please feel free to distribute these materials to the State. The results of our analysis dispute the conclusion that the ISCST area source algorithm should not be used for modeling landfills or other area sources where two or more sources share a common boundary. Rather, the model is appropriate for landfills, when used properly and judiciously. The key is use of sources sufficiently small in size to minimize problems like those noted by TACB; use of multiple small areas is also recommended by the TRC report. Our guidance has always encouraged the use of such discretion, but it is frequently ignored by model users. As Russ notes, we have explored the various options suggested by TACB and found one of them to be inappropriate and two others to provide similar results to the current area source algorithm. We are also aware of the problems identified by TACB, and previously confirmed them by our own analyses. Issues like these have occurred in the past but we had not found a ready correction or a better modeling alternative. Thus the need for discretion. Resources preclude us from more vigorously pursuing the problems. But we believe that the TRC report now presents us with a firm enough basis to explore use of a PAL-like area source algorithm in ISC; we will consider this in our FY-91 plans. To summarize, we would like to make the following observations. Given the limitations of current area source algorithms in general, case-by-case professional judgment is necessary when determining the appropriate sizes of area sources to be modeled, thus making the results obtained from different modelers not necessarily replicable. While this is not an ideal situation from a regulatory point of view, it is a practical necessity when dealing with area sources and is an outcome that we have been aware of for a long time. Similarly, the occurrence of spurious spikes can be mitigated if sizes of the area source are made small compared to the distance to the nearest receptor. This latter mitigating measure may introduce some problems in practicality in cases where receptors need to be placed very close to the edge of an area source. Prudent judgment and perhaps some "microscale" remodeling may be useful in such cases. If you have any questions, please contact me. Attachments cc: J. Dicke R. Lee M. Smith J. Touma D. Wilson (Attachments Following) October 18, 1989 MEMORANDUM SUBJECT: Comments on TACB Area Source Discussion Which Was Dated 4/20/89, and its attached report, "Modeling Techniques for Municipal Solid Waste Landfills." FROM: Russell F. Lee Model Support Section, SRAB (MD-14) TO: Dean Wilson TES The Texas Air Control Board (TACB) has conducted some experiments using the ISCST area source algorithm with other approaches to modeling area sources. As a result of these experiments, TACB has concluded that the ISCST "area source algorithm, as currently programmed, should not be used for modeling landfills or any other area sources where two or more sources share a common boundary." Although their study raises some questions concerning the behavior of the ISCST area source algorithm, their conclusion is not justified by the results provided in their report. I believe that the ISCST area source algorithm is appropriate for modeling landfills, if it is used properly and judiciously. However, the TACB report does provide evidence of some anomalous results that should be investigated. The area source algorithm in ISCST is an algorithm that functions best for area sources that are small (relative to the source to receptor distance). This is consistent with its purpose of estimating air quality impact from area sources (e.g., gravel piles, tailings ponds, etc.) that are a part of an industrial source complex. Because of this characteristic of the algorithm, a relatively large area source must be represented as a group of small area sources to give appropriate results. Mr. Durrenberger's statement that "For an area source algorithm to appropriately function, it must be possible to take a given square and subdivide it into smaller squares and with each approach obtain the same results" is not correct in general, and is particularly incorrect in the case of the area source algorithm in ISCST. For a large area source, or a receptor very close to the area source, this algorithm will give more and more accurate results as the area source is subdivided into more and smaller sources. The results obtained by TACB in the area source scenarios show three notable characteristics. When the area source is represented as a single 50 x 50 meter source, the concentration profile is flat-topped. This is what one would expect intuitively. When the area source is represented by four 25 x 25 meter sources, a different peak concentration is found, which was expected, with a spike in the center, which was not expected. When the area source was represented as sixteen 12.5 x 12.5 meter sources, a similar result was found for the nearest receptor distance. Again, this was not surprising. At other distances, the concentration profile is much more realistic, being nearly flat-topped, with concentrations tapering off near the edges of the area source, as one would expect. The best test to determine how realistic the representation of the area source is, is to compare these results to an area source represented as a very large number of ground level point sources. As the area source emissions are divided among a larger and larger number of point sources, the concentration calculations approach a "true" area source representation. This was not done. (As noted in the TACB report, the point source experiment was done with a different data set.) TACB recommends modeling landfills either 1) using an array of pseudo-point sources, 2) an arrangement of volume sources, or 3) an arrangement of area sources. In method 1, the array of point sources is designed for virtually no plume rise, and a height of release of 1 meter. This is reasonable, although a height of zero meters would probably be more appropriate. The TACB report described the volume source in ISCST as "a box with uniform emissions through the rectangular 'window' defined by the initial coefficients." This is incorrect. The emissions profile through the "window" is bivariate gaussian, not uniform. As a result, a single volume source is not an appropriate representation of an area source. As with point sources, an area source can be represented by a sufficiently large number of volume sources. Differences between the concentration estimates obtained from using volume sources versus using area sources in the model are more likely to be due to inappropriate use of the volume sources than a flaw in the area source algorithm. I recommend that we resolve concerns about possible inaccuracies of any of the area source methodologies. To do this, comparisons should be made between each methodology and a "correct" representation of an area source, namely, a sufficiently large number of point sources to define the limit of using an infinite number of point sources to represent the area source. The point sources should be at ground level, with no plume rise. I strongly suggest that we identify the cause of the spike in the concentration profile, to determine if this is normal for the algorithm or an error in the code. I also suggest that we derive a simple scheme to determine how much an area source should be subdivided to give acceptable results at receptor at a given distance. Attachment cc: Jerry Mersch Joe Tikvart November 17, 1989 MEMORANDUM SUBJECT: Additional Comments Related to TACB Area Source Discussion. FROM: Russell F. Lee Model Support Section, SRAB (MD-14) TO: Dean Wilson TES I recommend that we resolve concerns about possible inaccuracies of any of the area source methodologies. To do this, comparisons should be made between each methodology and a "correct" representation of an area source, namely, a sufficiently large number of point sources to define the limit of using an infinite number of point sources to represent the area source. The point sources should be at ground level, with no plume rise. I strongly suggest that we identify the cause of the spike in the concentration profile, to determine if this is normal for the algorithm or an error in the code. I also suggest that we derive a simple scheme to determine how much an area source should be subdivided to give acceptable results at receptor at a given distance. In my October 18 memo I suggested that we derive a simple scheme to determine how much an area source should be subdivided for ISCST to give acceptable results at a receptor at a given distance. I have located a study that provides enough information to get a rough idea of what such a scheme would be. I have located a copy of an SRAB report, "Sensitivity of ISCST Model Predictions to Area Source Size." In this study, ISCST was run with a year's worth of meteorological data, and each of three area sources of 100, 400, and 1000 meters on a side, respectively. Concentration calculations were made at two receptors, at 100 meters and 300 meters downwind of the center of the source, respectively. The area source was represented as both as area sources and point sources, singly and subdivided into 4, 16, 64, 256, 1024, 4096, and 16,384 sources. Note that two area source algorithms are referenced in the report. One, labeled "Current ISC Area Source Algorithm," refers to older versions of ISC which would not make concentration calculations closer than 100 meters from an area (or any other) source, and is therefore no longer relevant. The other algorithm, referenced as "Future ISC Area Source Algorithm," refers to the way area source calculations are now made in ISCST. Two results of the study are relevant to the TACB discussion. First, for both point and area sources, concentration calculations converge on a stable value as the source is subdivided into more and more sources. Second, as noted in the Summary of Results of the SRAB report, the calculations based on multiple area sources converge to a value 12 to 13% higher than those based on multiple point sources. I suspect an error in a constant in the area source algorithm is causing this. Third, from the tables in that report, it appears that area sources should probably be subdivided until the length of the side of a source is about one-fourth or less of the distance from the downwind edge of the source to the receptor. If point sources are used to represent the original area source, they should be approximately twice as dense as the area sources, i.e., a maximum separation between point sources of about one-eighth the distance to the receptor. These figures are rough, and should not be used until confirmed by a much more detailed study. Attachment cc: Jerry Mersch Joe Tikvart