Draft 7/26/07

Prevention of Significant Deterioration/New Source Review: Refinement of Increment Modeling Procedures

Docket ID No. EPA-HQ-OAR-2006-0888

Comments prepared for MARAMA

Issue #1: What Meteorological models and data should be used in increment consuming modeling? (pgs 31391-31393)

General Comment: We believe this section is ambiguous. The US EPA should clarify if it is proposing procedures to use prognostic meteorological data for long-range transport (LRT) modeling only or for LRT and near-scale (local) dispersion modeling (e.g. AERMOD).

General Comment: The US EPA should develop detailed technical guidance before it allows permit-reviewing authorities to use prognostic meteorological data in near-scale modeling analyses. Several of these issues were outlined in a recently published article in the Journal of the Air & Waste Management Association (Touma et al, 2007)[1]

Specific Comments:

  • The US EPA is proposing to utilize existing guidance developed for ozone, PM-2.5 and regional haze SIP modeling. This guidance is designed for regional-scale modeling used in attainment demonstrations and may not be applicable for near-scale modeling purposes.
  • Some of the meteorological parameters needed for near-scale modeling may not be directly available from the prognostic meteorological model. Some variables may also depend on which planetary-boundary layer (PBL) scheme is employed in the prognostic meteorological model. The current SIP modeling guidance does not recommend any of the many PBL parameterization schemes that are available for the various prognostic meteorological models. Extensive analysis would have to be conducted to determine which PBL scheme most closely reproduces the needed meteorological fields.
  • Any statistical comparison should use meteorological data not incorporated into the prognostic meteorological model. Some of these models use measured meteorological data to “force” the model towards reality through the use of four-dimensional data assimilation (FDDA) often referred to as “nudging”.
  • The US EPA should distinguish if it prefers prognostic meteorological models run in “simulation” mode (i.e. FDDA) or operational forecasting mode. Simulation mode would theoretically provide the most accurate meteorological data.
  • The reviewing authority must consider the scale of the prognostic model in determining if modeled meteorological data will adequately capture important dispersion features. For example, if a source is heavily influenced by topographically induced flow or a sea or lake breeze the prognostic model must have adequate horizontal and vertical grid spacing to resolve these features. Generally, it takes 5-10 grid cells to adequately resolve a given meteorological feature; so a model with a 12 km grid spacing could simulate features from 60 to 120 km in scale (all other features are parameterized within the model)
  • The US EPA’s suggestion of maintaining a buffer of six (6) grid cells from the prognostic model’s lateral boundaries is inadequate for two reasons. The first is that it conflicts with general guidance used by the Federal Land Managers (FLMs) who maintain that a large enough buffer (100-150 km) must maintained between the model’s lateral boundary and model receptors/sources to prevent plumes in CALPUFF from crossing the model boundary that would later reenter the domain due to wind reversals. The second reason is prognostic models typically apply smoothing equations near the lateral model boundaries to prevent spurious waves from migrating back into the modeling domain. The application of these smoothing functions probably varies between the different prognostic models. These smoothing functions may unduly influence some of the meteorological variables in grid cells close to the lateral boundaries.
  • In near-scale modeling applications land-use categorizations would be radically different between the prognostic meteorological model (averaged land-use over a large grid cell) and the near-scale model (i.e. AERMOD). This could lead to radical differences between local land use categorizations between the two models.

Issue #2:How is a source with a Class I Area FLM variance treated in subsequent increment consumption modeling (Federal Registerpages 31,381-31,384).

EPA Proposal: If a source is permitted with a variance issued by a Federal Land Manager (FLM) for a Class I area, its emissions will be excluded from all future Class I increment consumption analyses for the Class I area the variance was granted. These variance emissions would only be included for the purposes of determining Class II increment consumption at the Class I area.

Comment:

Emission Variance and Future Class I Increment Analysis – Section 165(d)(2) of the CAA and CFR Part 51.166(p)(4) allow a source exceeding the Class I increments to obtain a permit through a variance. However, the rules and regulations do not imply that these emissions should be excluded in future increment analyses by other sources. Once an increment is exceeded at a Class I area, each successive increase in pollution concentrations at the Class I area should be modeled by the applicant to determine the magnitude and frequency of additional Class I increment violations at the Class I area. It should be the responsibility of each successive applicant, not the FLM, to conduct the modeling and make the demonstration of no adverse AQRV impact once the Class I increments have been violated. Transferring the burden back to the FLM makes little sense procedurally, will unnecessarily strain the FLM’s limited resources, and may in fact give an unfair economic advantage to the companies submitting applications after the variance source. Unlike the variance source, these later permit applicants would not need to devote the time and resources to the AQRV impact analysis until the Class I increments are once again exceeded without the variance emissions.

Inclusion of Variance Emissions in AQRV Impact Analyses - Once an increment is exceeded at a Class I area, each successive increase in pollution concentrations at the Class I area should be examined to determine if the threshold for adverse impacts on the AQRV has been crossed. In a cumulative analysis for an AQRV such as deposition, a new source will need to include not only its emissions, but the emissions from all existing sources and all sources permitted but not yet operating. As proposed, it is not clear that the variance emissions would be included when determining whether AQRVs are adversely impacted in future modeling. These variance emissions should be included in cumulative AQRV analyses.

Complying with 40 CFR 51.166(a)(3) and the Need for a SIP Amendment

The need for a SIP amendment when the Class I increment is exceeded could be eliminated by adding language to 40 CFR 51.166(a)(3) that states “The SIP need not be amended for a Class I increment violation if a variance is obtained from the permitting authority for the source whose emissions are causing or contributing to the violation.”

Another possibility is for the State to submit a source-specific SIP action for the variance emissions. In either case, the emissions variance in the permit should have the following restrictions: apply only to the pollutant predicted to exceed the Class I area increment, apply only to the location where the Class I area increments are predicted to be exceeded, and apply only for the time periods where exceedances of the Class I increment are predicted.

Issue #3: How are emissions estimated for sources that consume increment? - (3)Actual Emission Rates Used to Model Short-Term Increment Compliance (Federal Register pages 31389 – 31390)

Proposal:

EPA is proposing to replace the current approaches for estimating emissions when modeling of short-term (24-hours or less) PSD increment consumption. EPA believes these current practices are too rigid and do not allow the flexibility to states to decide on the proper approach. EPA seeks comments on recommendations submitted by WESTAR that includes a menu of equally acceptable options for estimating short-term emissions.

Comment:

We do not agree with the proposed menu of options. The menu should have a hierarchy based on the availability of reliable data concerning source operations.

The proposal contradicts guidance given in the preamble of the 1980 PSD rules. In determining baseline emission rates, the 1980 preamble {45FR at 52718, col. 3} states that,

“EPA believes it is generally appropriate to presume the source will operate and emit at the allowed levels” and that, “When EPA or a state devotes the resources necessary to develop source-specific emissions limitations, EPA believes it is reasonable to presume those limitations closely reflect actual source operation. EPA, states, and sources should then be able to rely on those emissions limitations when modeling increment consumption.”

In this discussion, EPA also cautions that,

“The presumption that federally enforceable source-specific requirements correctly reflect actual operating conditions should be rejected by EPA or a state, if reliable evidence is available which shows that actual emissions differ from the level established in the permit.”

This preamble clearly indicates a hierarchy must be used when developing a PSD inventory. Unless there is reliable data to show actual emissions differ from allowable; the allowable emissions must be used in the PSD increment. Therefore, allowable emissions must rank first in the hierarchy.

The data needed for calculating representative annual average actual emission rates is usually readily available for point sources (annual fuel use, annual production, etc). The annual average actual emission rates used in NSR applicability tests also seem appropriate for modeling annual PSD increment consumption.

Reliable data needed to establish short-term emission rates of SO2 and PM-10 that are less than the allowable rates may not be available. Fuel use or production rates on a daily or hourly basis would be needed calculating PM-10 emissions. If there are no CEMs (which is the case for most sources), the same information would be needed for calculating SO2 emissions. In most cases, use of weekly, monthly, or annual fuel use or production data would not constitute reliable evidence that actual emissions differed from allowable emissions on a 3 or 24-hour basis. When a source lacks the data to estimate a reasonably accurate short-term emission rate, it should not be rewarded by being allowed to use its 2-year annual average emission rate when determining compliance with the short-term PSD increments.

EPA proposes two separate menu approaches depending on whether or not CEM data is available for the source. Both menus contain a set of choices that can be used without a preferred hierarchy. An approach without a hierarchy is only appropriate when each of the methods is equally acceptable and technically sound. Unfortunately, this is not the case with EPA’s proposal. Some of these will clearly lead to an underestimation of the consumption of short-term PSD increments.

Of the two lists of choices, those under the “CEM data available” list can generally be supported. Those under the “non-CEM data available” list are for the most part not technically sound or acceptable. As mentioned earlier and reiterated in EPA’s proposal, SO2 CEM data are usually not available for a source. There are currently no CEM data available for PM10 (nor will there be any for PM2.5 emissions when PSD increments are developed for that pollutant). As a result, in most cases WESTAR’s approach and EPA’s proposal will limit sources and the reviewing authorities to the four choices under the menu with no-CEM data available.

The first three choices under the “CEM data not available” menu are related in one way or another to the calculation of annual average actual emissions. The last choice, that of using the allowable emissions rate, is the only acceptable method in this portion of the menu. For the set of choices under the CEM availability data menu, the approaches using the maximum or upper percentile short-term emissions for each source are technically acceptable. The use of hourly CEM data (last entry) might be acceptable.

Instead of the proposed menu, we suggest an alternative that is based on a hierarchy of choices. For determining emissions to be used for short-term PSD increment calculations, we propose the following hierarchy:

1) Use allowable emissions if data to properly quantify actual maximum emissions are not available or cannot be substantiated by the reviewing agency. Using the allowable rate as the first entry meets the PSD regulations 40 CFR 51.166(b)(21)(iii) and 51.166(k), but can be bypassed if valid short-term actual data is available.

2) Use the maximum actual emission rate or a upper percentile value that is based on Page C.49 of EPA’s draft NSR Workshop Manual or existing EPA guidance memos. A summary of this guidance is given below.

Draft 1990 NSR Workshop Manual - page C.49. For short-term averaging periods (24-hours or less), the maximum actual emissions rate is the highest occurrence for that averaging period during the previous two years of operation.

May 3, 1985 memo from Gary McCutchen to Mike Trutna (NSR Section Chief) entitled “NSR Advisory Memorandum #1: TSP PSD Increment Consumption in North Carolina” – Use the maximum 24-hour emission rate over a two year time span that represented normal operations. To avoid abnormal operations, the maximum emission rate must have occurred at least 5 percent of the total 24-hour operating time periods.

July 31, 1981 memo from Thomas Devine (Director, AHMD) to State/Local Directors entitled “PSD Policy Determinations by Region IV” #4 - use the maximum 24-hour emission rate over a two year time span for short-term increment consumption.

May 2003 EPA Region 8 report “Dispersion Modeling Analysis of PSD Class I Increment Consumption in North Dakota and Eastern Montana” pages 19-20 – use the highest 90th percentile actual emission rate for short-term increment consumption.

3) Use hourly CEM data when representative of normal source operations.

To justify the use of annual average emission rates in short-term increment consumption modeling, EPA’s proposal states that it is unreasonable to assume that a multitude of sources included in an emission inventory will all operate simultaneously at their maximum emission rates throughout the year. Such an assumption leads to overly conservative impact calculations. Although it is true that continuous operations of many different sources at their maximum emission rates is very unlikely, it has been our experience that most problems with PSD increment consumption involved very few sources. The likelihood of two or three sources operating near their maximum allowable emission rates is much higher than a scenario involving 15 to 20 sources. Use of the annual average emission rates when modeling the short-term PSD increment consumption of a few sources would most likely under-estimate actual impacts.

It should be noted that with most agencies, when a source submits a multisource PSD increment analysis there is only one inventory, not two. The inventory attempts to document all emission changes since the baseline date. All new emission sources will be represented as positive emissions, emission reductions such as source shutdowns are represented as negative emissions. Due to the existence of better source data in recent years, there are normally more sources with emission increases than decreases.

Issue #4: How are emissions estimated for sources that consume increment? - (5)Emissions from a New or Modified Source (Federal Register page 31391)

Proposal: Projected actual emissions can be used in the PSD increment analysis conducted for a proposed modified source. Currently, the modified source’s potential to emit is used.

Comment: The proposal is justified by referencing the December 2002 NSR rule revisions. In these revisions, EPA allowed modified sources to use its projected actual emissions in calculating whether the change resulted in a significant net increase in emissions. Five years later EPA has now concluded that projected actual emissions should also be used in determining PSD increment consumption when a source undergoes a major modification.

Use of the projected actual emissions as described in the December 2002 NSR rule revisions for determining PSD increment consumption by the modified source is inappropriate for the following reasons.

  1. The methodology in the December 2002 NSR rule revisions concerns the calculation of annual average projected actual emissions, not short-term (24-hours and less) actual emissions. Modified sources would use their annual average projected actual emissions for determining short-term increment. As discussed in the previous comment, we find that unacceptable.
  2. This issue has already been addressed in the Federal Register Final Notice for the December 2002 NSR rule revisions (December 31, 2002). The following statement is made on Page 80196 of the Federal Register Final Notice,

“when you must determine your source’s compliance with the PSD increments following a major modification, you must still use the allowable emissions from each emission unit that is modified, or is affected by the modification.”