United States        Air and Radiation        EPA420-P-02-009
           Environmental Protection                  July 2002
           Agency                         NR-014b
svEPA    Geographic Allocation of
           State Level Nonroad Engine
           Population Data to the
           County Level
                                     > Printed on Recycled Paper

-------
                                                                 EPA420-P-02-009
                                                                         July 2002
                                     to the  County

                                 NR-014b
                        Assessment and Standards Division
                      Office of Transportation and Air Quality
                      U.S. Environmental Protection Agency
                                   NOTICE

   This technical report does not necessarily represent final EPA decisions or positions.
It is intended to present technical analysis of issues using data that are currently available.
        The purpose in the release of such reports is to facilitate the exchange of
     technical information and to inform the public of technical developments which
       may form the basis for a final EPA decision, position, or regulatory action.

-------
             Geographic Allocation of State Level Nonroad Engine
                       Population Data to the County Level

                                 Report No. NR-014b
                                 revised July 12, 2002

                           Assessment and Standards Division
                  EPA, Office of Transportation and Air Quality (OTAQ)
I.  Purpose
       The purpose of this report is to discuss the methodology and data that EPA decided to use
in the draft NONROAD2002 model to allocate equipment populations from the national to the
county level.

II. Background

       The EPA has developed a national nonroad air emissions inventory model called
NONROAD.  This model provides a tool for EPA, States, regional air pollution organizations,
and local air pollution control agencies to use in estimating pollution from nonroad vehicles and
equipment for State Implementation Plans (SIPs), as required by the 1990 Clean Air Act
Amendments, and other regulatory needs.

       With some exceptions, the model uses national engine population or sales data from
Power Systems Research (PSR), a company that tracks the sales and populations of all types of
engines sold in the U.S.  Since PSR also matches the engines to the equipment in which they are
used, the term "equipment populations" will be used for the purpose of this report to describe the
population of both nonroad equipment and the engines used to power that equipment.  The EPA
believes that PSR provides the most comprehensive national nonroad equipment population data
currently available.  PSR updates these data on a yearly basis.  However, in some cases, the EPA
has used population data from a source other than PSR when such a source is available and found
to be more accurate than the PSR data.  For some types of equipment (e.g., ATVs and
snowmobiles) the draft NONROAD2002 model uses equipment sales or population data from
industry sources or state registration data.

       The PSR database also geographically allocates equipment populations from the national
to the county level and then aggregates the county-level populations to generate state totals.
However, the  methods and data that PSR uses to perform these allocations have only been
explained in general terms, since PSR considers their methods to be proprietary information.
Since the EPA wants the methods that it uses to allocate equipment populations in NONROAD
to be fully understood by EPA and the public, we have decided to use publicly available data as
much as possible to serve as factors to allocate the national PSR equipment populations to the
county level. State/local users may elect to substitute their own equipment population data, where

                                           1

-------
such data is well-documented and specific to local conditions, for SIP purposes.  These data may
be derived from well-designed and executed surveys or other information sources. In order to be
used as input data for NONROAD, these surveys or alternative sources of information should
include the hours per year that the various types of equipment are used, as well as equipment
populations. Using only local population data in NONROAD without the corresponding local
activity data (or vice-versa) could result in distortions in the emission inventory estimates that
NONROAD calculates.

III. Allocating Activity Versus Engine Population

       One central feature of the NONROAD model is that it uses the same methods to allocate
equipment populations, equipment activity, and equipment emissions to specific geographic
areas. To the extent that a given equipment type is operated at the same power level (load factor),
operating cycle, and for the same number of hours in all areas, the distribution of the population
of this equipment type will match the distribution of its activity and engine emissions. In general,
population, activity, and emissions will tend to track one another, since emissions are a direct
function  of equipment activity and the conditions that stimulate increased engine activity are
likely to  stimulate increased engine populations.

       In reality, however, the geographic distribution of nonroad equipment may differ from the
geographic distribution of emissions from these equipment. Because an equipment type's
operating cycle and load factor most likely do not vary significantly by geographic area, the
difference between the distribution of population and emissions depends on variations in the
amount of activity. The amount of activity that each piece of nonroad equipment of a given type
experiences can vary from area to area as a result of variations in local economies, weather
patterns,  or other local conditions. For example, agricultural equipment and residential
lawnmowers may experience more use per year in areas with longer growing seasons;
construction equipment is likely to be used more intensively in areas experiencing an economic
boom and less intensively where the economy is not as robust.

       Currently, the draft NONROAD2002 model is capable of handling only one activity level
for each equipment type across all parts of the U.S. As a result, the model uses the same factors
to allocate engine populations and their associated activity. Wherever possible, the EPA has
sought indicators related to engine activity, since it is engine activity that results in emissions
(except for diurnal and hot soak emissions, which are more closely related to engine
populations). In some cases, however, the EPA was unable to find a suitable activity indicator
and had to rely on population-related indicators as a surrogate for engine activity. In this report,
the EPA  has attempted to be explicit as to whether each equipment type's geographic allocation
factor is an activity-oriented indicator, a population-oriented indicator, or an indicator that is
reasonable for both population and activity.

       For several equipment categories, the EPA has had particular difficulty finding either
activity or population indicators for several equipment categories. These  categories include

-------
recreational, railway maintenance, and AC/refrigeration equipment. The EPA welcomes
suggestions from the nonroad industry, state and local air quality agencies, and other interested
parties concerning improved methods to allocate these equipment categories to the county level.
The EPA also invites state and local air quality agencies to substitute adequately documented
local data for the national default allocation estimates, in accordance with EPA guidance, for
these categories and for other categories such as recreational marine engines where the national
default estimates may not adequately reflect local conditions.

       Through the EPA Office of Air and Radiation's (OAR) Section 103/105 Grant Program,
the EPA has sought to encourage regional, state, and local air pollution organizations to develop
and apply methods to collect local nonroad equipment population and activity data for categories
of equipment that contribute significantly to the total nonroad emissions inventory. The EPA
awarded a grant to the Northeast States Coordinated Air Use Management Association
(NESCAUM) in 1999 to conduct a multi-year effort that generally includes the following:

1) performing a review and analysis of known survey methods that have been used by state and
   local air agencies to collect local nonroad equipment activity data,
2) choosing one of the survey methods reviewed and analyzed in part 1 or developing a new
   method to be included in EPA guidance to state and local air agencies, and
3) applying the chosen method to selected areas to ensure that it works properly and produces
   reasonable and useful results.

   The EPA also awarded a Section 103/105 grant to the California Air Resources Board (ARB)
to conduct a lawn and garden equipment survey. This survey targets potential residential,
commercial, and municipal/institutional users of lawn and garden equipment.  This project
includes a questionnaire about the number and type of equipment owned and used. In an attempt
to develop better activity data, the ARB also is distributing data loggers to willing volunteers. A
volunteer attaches the data logger to his or her equipment in order to record the time  and date of
when the piece of equipment is started and shut off.  The results of the work performed under
this grant are expected to be available in late 2002.  Partial preliminary results can be viewed at
the following website.

                  http://arbis.arb.ca.gov/msei/projectl03/proj 103_Stats.htm

       The EPA welcomes suggestions and comments about these efforts, as well as information
about surveys of nonroad equipment that have been conducted in the past or are presently being
conducted, from stakeholders and other interested parties.

IV. Methodology

       The EPA decided to design NONROAD to use information related to equipment
population or activity to distribute national equipment populations and their associated activity to
the county level. This surrogate information constitutes a geographic allocation factor. The model

-------
can use a single allocation factor for entire categories of nonroad equipment, or it can use
separate factors for one or more equipment types within a category. Mechanically, the model
assumes that each piece of equipment of a given type experiences the same annual activity (i.e.,
hours/year). This reduces the allocation problem to one of allocating engine populations. In
essence, the allocation factor serves as a surrogate for population and activity.

       Since the NONROAD model only has the ability to allocate equipment populations from
the state to the county level, the national population data for each equipment type must be
processed outside of the model to obtain state-level populations used in the model. To obtain
state level populations of most equipment types for input into NONROAD, the county-based
allocation factors discussed in detail below are applied outside of the model to the national
equipment populations to obtain county-level equipment populations, as shown in the equation
below.

           (Equip. Population)county =  (Equip. Population)national  x Surrogatecounty
                                                              Surrogatenationai

These county-level populations are then aggregated into state totals which are input into
NONROAD as the default state populations. The county-based allocation factors are also
aggregated into state totals for use in the model. The model then uses the aggregated state
equipment population and allocation totals to calculate the county-level equipment populations it
outputs, as shown below.

(Equip. Population)comm, = (Calculated Equip. Population)state  X     Surrogatecount,	
                                                              Calculated Surrogatestate

Due to the large amount of data that needs to be manipulated, modifying the model so that it
could do all of this processing internally would increase the time it takes NONROAD to run a
scenario beyond reasonable limits.

       There are a few equipment types that use a modified form of the above method. As
explained in more detail below in the sections covering each equipment type, snowmobiles,
ATV's, offroad motorcycles, and recreational marine equipment use an equipment-specific
method to allocate from national to state (done outside of the model) and then use the above
method to allocate from state to county within the model.

V.  Sources and Types of Data

       There are three basic types of data that are potentially useful as allocation factors: human
population and  its associated income and housing data, business activity, and geographic data.
Most of these data are available from the U.S. Census Bureau or other federal agencies except for
data concerning construction activity, which is discussed separately below. Information from the
U.S. Census Bureau is especially attractive for use in the NONROAD model because census data

-------
undergo rigorous statistical analyses and quality assurance reviews.

Population Data

       The U.S. Census Bureau conducts a nationwide census on a decennial basis. The census
includes data on population, housing (e.g., number of homes by type, number of occupants per
home), and income. The most recent census available at the time these allocations were updated
occurred in 1990. The Census Bureau produces updated human population estimates on an
annual basis, but does not produce an annual update for housing or income data.

Business Activity Data

       The U.S. Census Bureau publishes an annual report called County Business Patterns
(CBP), which tracks the number of establishments and employees for various types of businesses
and industries at the national and county level using Standard Industrial Codes (SICs). The draft
NONROAD2002 model uses 1996 CBP data. EPA used County Business Pattern indicators to
allocate state-level populations to the county level for the 1991 Nonroad Engine and Vehicle
Emissions Study1 (NEVES).

Geographic Data

       Geographic data include factors related to an area's location or physical characteristics.
Such factors  include water or land surface area, weather data, and land use data. Such data are
available from government agencies such as the U.S. Census Bureau, the National
Oceanographic and Atmospheric Administration, and the U.S. Geological Survey.

VI. Allocation of Specific Populations of Equipment Categories/Types

       The allocation indicators that the EPA has examined and selected for use in the various
draft versions of the NONROAD model are discussed below. The indicators used in previous
draft versions of the model and those being used in the draft NONROAD2002 model  are
summarized in Table 1 at the end of this report.

Residential Lawn and Garden Equipment (except snowblowers)

       To allocate lawn and garden equipment used by private households, the EPA decided to
use U.S. Census data on one and two unit housing (i.e., single family homes and duplexes) by
county from the USA Counties database. Structures containing more than two units tend to be
condominiums or apartments that use commercial lawn care services. One and two unit housing
information is used as an allocation factor in the NEVES, and an analysis of this set of data
during the writing of the NEVES showed that it provided a good predictor of lawn and garden
equipment populations. In addition, the EPA has not been able to find an alternative type of data
to use as an allocation factor for residential lawn and garden equipment that offers the high

-------
quality, the necessary county-level detail, and the predictive strength of one and two unit housing
data from the U.S. Census Bureau.

       The one and two unit housing data found in the USA Counties database and used in the
June 1998 and April 1999 (Tier 2) versions of the draft model come from the 1990 census. As
stated in section V, housing information is only updated during each decennial census. However,
using the 1990 one and two unit housing data will cause the model to underestimate lawn and
garden emissions in more urbanized parts of the country because of the continuing creation and
growth of suburbs in counties that were still partially or largely rural in 1990. To address this
underestimation the draft NONROAD2002 model accounts for this growth by adjusting the
housing data by a ratio of 1997 human population estimates to human population data by county
from the 1990  Census. It should be noted that if the  1997 human population estimate for a county
is less than the 1990 census value, then the 1990 single and double housing unit value for that
county will remain unadjusted. In such cases, the relative longevity of most housing structures
suggests that the number of one and two-unit housing sites would not decline as quickly as
human population. For counties which experienced an increase in population between 1990 and
1997, the equation to  adjust the housing unit data is  as follows:

       1997 Adjusted Housing = 1990 Housing x  1997 human population estimate
                                               1990 census human population

This adjustment could result in an overestimation of lawn and garden emissions to the extent that
a disproportionate share of the population growth in the county is housed in multi-unit housing.
Any such overestimate would be partially offset by a concomitant underestimate in the growth in
commercial lawn and garden activity levels, and by the absence of any adjustment to account for
the ongoing reduction in the average number of people per household.

       One and two unit housing is most properly thought of as a population allocation factor for
residential lawn and garden equipment. The population of such equipment in an area should be
roughly proportional to the number of single and double housing units in the area, since the
average household occupying such units would have the average probability  of owning any given
type of lawn and garden equipment. But the amount of use such equipment experiences may vary
considerably from area to area based on such variables as the average size of yards, length of
growing season, and amount of rainfall. Allocation factors based on residential lawn and garden
equipment gasoline consumption, tons of yard waste removed, or the land area occupied by
single and double housing units could, in principle, provide a more direct measure of activity.
However, the information regarding such potential activity allocation factors are either not
available, of questionable quality, or subject to confounding influences that make the potential
allocation factor even less reliable than the one  currently used in NONROAD. Therefore, the
EPA expects to use one and two unit single family housing data modified by the 1997 county
population estimates from the Census Bureau for the draft NONROAD2002  model.  EPA plans
to update this allocation factor using 2000 Census data for the final release of the model.

-------
Commercial Lawn and Garden Equipment (except snowblowers)

       To allocate commercial lawn and garden equipment, the draft 2002 version of
NONROAD uses the number of employees in landscape and horticultural services (SIC 78) from
the CBP database. The 1991 NEVES study uses this factor to allocate commercial lawn and
garden equipment. An analysis performed during the writing of the NEVES shows that the
number of employees in landscape and horticultural services is a good predictor of commercial
lawn and garden equipment populations. In addition, the EPA does not know of any other
sources to adequately serve as a geographic allocation factor for commercial lawn and garden
equipment.

       The number of employees in landscape and horticultural services is better suited for
allocating the population of commercial lawn and garden equipment than the activity associated
with this type of equipment. The level of mechanization in the landscape services industry is
likely to be relatively constant from county to county. By contrast, the number of hours per year
that the average piece of commercial lawn and garden equipment operates is likely to vary
considerably from county to county as a result of different growing seasons and rainfall patterns.
Reliable information on allocation factors more directly related to activity levels,  such as  gallons
of fuel consumed per county and year by commercial lawn and garden equipment, are not
available.

       One caveat for using the number of employees in landscape and horticultural services to
geographically allocate commercial lawn and garden equipment populations is that this factor
does not include municipal employees that perform landscape maintenance duties for schools,
parks, and other properties owned and maintained by local governments. The implicit assumption
used in NONROAD is that the population and activity level of such equipment is proportional to
commercial lawn and garden equipment population and activity levels. The EPA welcomes
comments from interested parties concerning methods or sources of data that can better account
for lawn and garden equipment used by municipal landscape employees.

Snowblowers

       Allocating snowblower populations and activity levels requires the use of allocation
factors that account for the impact of climatic differences among regions, in addition to the
factors used to allocate residential and commercial lawn and garden equipment. Put simply,
snowblower populations and activity levels depend on snowfall.  Snowblower populations in
warm-weather states like Florida, Louisiana, and Hawaii should be zero. Snowblowers may be
present in parts of states such as Texas and California because part of their territories receive
snow (e.g., Texas Panhandle, Sierra Nevada Mountains in California), while snowblower
populations in other parts of the state should be zero.

       Due to time and resource constraints, snowblower populations at the state and county
levels were set to zero in the June 1998 draft version of the model to avoid misallocation

-------
problems. However, this version of the model could calculate national annual snowblower
emissions. The April 1999 (Tier 2) draft version of the model has snowblower populations
assigned to states based on snowmobile registration data supplied by the International
Snowmobile Manufacturer's Association (ISMA). The model then allocated the snowblowers to
the county level using the same factors used to allocate other types of lawn and garden
equipment.

       The approach chosen by the EPA to allocate snowblowers in the draft NONROAD2002
model involves estimating which counties in the U.S. receive enough snowfall to necessitate the
presence and use of snowblowers. This was accomplished by overlaying a map of the U.S. from
the National Oceanographic and Atmospheric Agency (NOAA) showing ranges of long-term
average snowfall amounts on top of a map of U.S. counties and making an informed judgement
about the minimum amount of annual snowfall requiring the use of snowblowers. The EPA has
chosen a minimum snowfall limit of fifteen inches, based on discussions with a snowblower
manufacturer and by the mapping process mentioned above. The same allocation factors applied
to other lawn and garden equipment types (i.e., the number of single and duplex family housing
units for residential snowblowers and the number of employees in landscaping and horticultural
services for commercial snowblowers) will be used to determine the state snowblower population
inputs to NONROAD (as discussed in Section IV: Methodology) and as inputs  to the model to
allocate the state snowblower population inputs to counties receiving at least 15 inches of snow.
For counties that do not receive at least 15 inches of snow, the allocation factors will be set to
zero so that no snowblowers will be allocated to those counties or erroneously included in the
total state population.

Construction Equipment

       Initially, the EPA planned to use the number of employees engaged in construction by
county (CBP SIC 15) to geographically allocate construction equipment. However, early
comments from some stakeholders correctly pointed out that using this indicator could lead to
errors in estimating construction equipment population and activity in a county because
construction employees and equipment move from project to project, often crossing county lines.
In some parts of the country, such as the Northeast, construction employees and equipment may
cross state lines quite frequently. The CBP data only reflect where construction employees and
establishments are headquartered, not where they work.

       The EPA believes that the dollar value of construction offers the best means available to
allocate construction equipment activity to counties. The dollar value of construction provides a
good reflection of activity, since there is a proportional relationship between the dollar value of
construction and the amount of construction activity in a given area. Also, using the dollar value
of construction by county as an allocation factor will distribute construction equipment to where
it is actually being used, as opposed to where it is headquartered. Furthermore, this indicator
provides a reasonable allocation factor for construction equipment populations: competitive
forces encourage construction companies to obtain the maximum return on their investments in

-------
costly pieces of construction equipment by maximizing their use as much as possible, thereby
strengthening the correlation between construction activity and construction equipment
population.

       Data on the dollar value of construction is collected and maintained by the Census Bureau
only for metropolitan statistical areas (MSAs) instead of counties. Construction valuation data by
county for 1997 is available from the F.W. Dodge Company. For the June 1998 and April 1999
(Tier 2) draft versions of NONROAD, the EPA decided to use the F.W. Dodge total dollar value
of construction data to allocate construction equipment populations to the county level.

       For the draft NONROAD2002 model, further refinements to this method have been
made. The Dodge data includes the dollar value of residential, commercial, and industrial
building construction, as well as road and other public works-related heavy construction. The
construction of the various types of buildings accounts for a large portion of the total dollar value
of construction. However, according to a survey of construction activity in Houston conducted by
Environ in 1998, road and other types of heavy construction constitute a much larger share of
actual equipment activity per dollar valuation compared to the construction of residential,
commercial, and industrial buildings. This apparent discrepancy can be explained by the fact that,
once the land is cleared and graded, heavy construction equipment is not used much in the
construction of the actual building, which also usually accounts for the majority of the project's
cost. The trends in the Environ survey compare favorably to a study conducted by Sierra
Research in  1993 that estimated the relative activity of construction equipment based on fuel cost
per project dollar valuation derived from the 1987 Census of Construction Industries.

       Therefore, the EPA has weighted the various categories of the Dodge construction dollar
value data based on  the Environ survey of Houston construction activity. The equation to
accomplish this is shown below.
                Allocation Factorj = (SFHj + 3*OBLDGj + 18.4*R&Bj + 8.5*PWj) /
                                       (SFH + 3*OBLDG + 18.4*R&B + 8.5*PW)

       Where the variables are the dollar valuation for either the county (j) or national total

          SFH for single/double-family housing  construction,
          OBLDG for other building construction,
          R&B for road and bridge construction, and
          PW is public works (sewer, water, and drainage) construction.

The heavier weighting given to road and other types of infrastructure construction generally tends
to decrease the allocation of construction equipment to urban counties and increase the allocation
of this equipment to rural counties. This stands to reason, since the road and other infrastructure
systems in urban counties tend to be largely established but are still being developed in more
rural counties where suburban sprawl continues to take place.

-------
Agricultural Equipment

       For this category, the EPA considered using the number of employees involved in
agricultural work by county as an allocation factor (CBP SIC 78), as was used in the NEVES.
However, this allocation indicator may not correlate well with either agricultural activity or
agricultural equipment populations. A small number of agricultural employees in a county could
cause the model to underestimate the population and activity of agricultural equipment if the
predominant type of farming in that county is highly automated or relies on migrant labor that is
recorded as being based in a different county. Conversely, a large number of agricultural
employees in a county could cause the model to overestimate the population and emissions of
agricultural equipment if the predominant type of farming is labor intensive or if migrant labor is
recorded as being based in the county. Instead, the EPA chose to use the acreage of cropland
harvested by county to allocate agricultural equipment populations in the June 1998 and April
1999 (Tier 2) draft versions of NONROAD. The EPA expects to use these data in the final
version of NONROAD as well. This information is available from the U.S. Census Bureau's
USA Counties database. It should be noted that in some cases where a county only contains a few
small farms the Census Bureau does not publish crop acreage data out of a concern for
confidentiality. However, agricultural equipment emissions in a county that only has a few small
farms are likely to be small relative to other nonroad sources within this county and to
agricultural equipment emissions in other counties within the state where farms  and agricultural
equipment are more numerous.

       Using the amount of harvested cropland as an allocation factor provides  a good predictor
of agricultural equipment activity, since a proportional relationship generally exists between the
amount of cropland harvested and how much equipment activity is needed to prepare the land
and plant, maintain, and harvest the crops. However, the amount of cropland harvested does not
necessarily provide as accurate a predictor of agricultural equipment population as it does for
activity for several reasons. First, the same amount of cropland in a county can be plowed,
planted and harvested by a few pieces of large equipment or several smaller ones. Second, the
amount of equipment present in a county may be more dependent on the number of farms than on
the amount of acreage harvested  (although this source of inaccuracy in estimating populations
may be mitigated by the presence of equipment-sharing arrangements in areas with smaller
farms).  Since the purpose of NONROAD is to estimate emission levels, and since emissions are
more directly associated with activity levels than with equipment populations, the EPA believes
that the amount of harvested cropland is an appropriate allocation factor for the draft
NONROAD2002 model.

Recreational Marine Equipment

       Because the county in which the equipment is purchased, registered, and/or stored may
not be the same county where the equipment is used, the geographic allocation of recreational
marine equipment presents a significant challenge. An urban or suburban county where a boat is
sold, registered, and/or stored may not contain a body of water that can support recreational
                                           10

-------
marine traffic, or water bodies near where a boat owner lives may be overcrowded. Small and
medium recreational marine craft, which constitute most of the recreational marine fleet, can be
transported by trailer over a wide area, further complicating matters. Thus, sales and registration
data are not sufficient to accurately allocate recreational marine equipment to the county level.

       The NEVES report estimated the population of boats actually being used in each of the 24
urban nonattainment areas covered by the NEVES through the use of data from a 1991 National
Marine Manufacturers Association survey of boat owners.2 These data included where the boat
owners said they primarily operated their boats, where the boats were registered, the boat owners'
estimates of the amount of hours they used their boats per boating season, and their estimates of
the amount of fuel their boats consumed per boating season. In general terms, the data from these
surveys were used to adjust registration-based boat populations so that only the boats actually
operating within the nonattainment area (as opposed to boats registered in the nonattainment area
but used elsewhere) were included when calculating the recreational marine emissions for each
area addressed in the NEVES. In order to check the reasonableness of the NMMA-based results,
the total square miles of water surface area in a nonattainment area, the estimated square miles of
water surface area needed for a typical boat to operate,  and the maximum number of possible
hours per boating season in each nonattainment area were used to ensure that the use of NMMA
survey data did not result in a boat population estimate that exceeded the maximum number of
boats that could theoretically operate during the boating season of a particular area. If the
NMMA-based boat population estimate exceeded the theoretical  maximum, then this population
was adjusted downward.

       This method cannot be incorporated into the current design of the NONROAD model, but
at least some of its elements might be adapted to allocate recreational marine equipment
populations and the activity associated with them to specific counties within a state for direct
input into the NONROAD  model. A limitation of the NEVES method was that it focused solely
on urban areas. Since most boating tends to occur outside of urban areas, it is likely that NEVES
significantly underestimated recreational marine emissions in the U.S. In order to be usable for
the NONROAD model, the NEVES method would have to be modified to estimate the
population and accompanying activity of recreational marine equipment being used in both urban
and non-urban counties. Also, new survey data would need to be  collected,  since much of the
1991 NMMA survey data are most likely outdated. This would be especially true in regard to the
rapid population growth and usage of personal watercraft ("jet skis"), which also would have
different activity profiles from other types of boats.

       The EPA decided to use water surface area data by county from the U.S. Census Bureau
to allocate recreational marine populations in the June 1998 and April 1999 (Tier 2) draft
releases of the NONROAD model. However, using this type of data has some limitations. Water
surface area does not make a distinction between navigable bodies of water and those that are too
shallow for boating or have obstructions through which boats are unable  to pass. Also, water
surface area does not account for convenience of location (proximity to areas of significant
population) or the recreational quality of the water body (which includes  such factors as its
                                           11

-------
attractiveness for fishing, its visual appeal, and its water quality), both of which could be
expected to affect a body of water's recreational marine activity per unit area. Another limitation
is that water surface area alone does not account for access restrictions that may prevent boating
or limit the number of boats permitted to operate on a given body of water. Lastly, using water
surface area alone does not distinguish between the differences in usage patterns for small
outboard engine-driven boats, larger inboard boats, and personal watercraft.

       Even with these limitations, the EPA decided to use water surface area as a factor for the
allocation of equipment populations because either the other potential approaches described
above have even greater limitations or, as in the case of the NEVES approach, additional work is
needed to determine the best way to adapt the approach to be compatible with NONROAD (or
vice versa) and collect the data necessary to implement the approach.

       For the draft NONROAD2002 model, some additional work has been completed to refine
the  water surface area allocation method. The EPA has  decided to use a 1992 gasoline
consumption distribution developed by the Oak Ridge National Laboratory (ORNL) for use in its
1994 Nonhighway Gasoline Use Estimator Model to allocate the national recreational boat
population to the state level. The ORNL gasoline consumption distribution is also used by the
Federal Highway Administration (FHWA) to estimate annual fuel consumption for this category
in states for which no gasoline tax records are available. Because the fuel consumption
distribution data directly relate to activity, using this method to allocate boats to the state-level is
better than using water surface area alone to allocate boating equipment populations from the
national to the county level and then summing the county populations to get a state total. Using
the  latter method results in an over-allocation of boating equipment to some states that have a
long coastline, such as Michigan. In addition, the State of Alaska received a disproportionate
share  of boating  equipment since much of its coastline and bodies of water are either inaccessible
and/or inhospitable to recreational boating.

       To further refine the allocation of recreational marine equipment, the differences in how
far  each kind of boat (i.e., personal watercraft, outboards, inboards) tends to operate from the
shore are addressed by modifying the  water surface area allocation factors for each of these types.
In NEVES and the earlier (June 1998  and April 1999) draft releases of NONROAD, a general
assumption was made that all  boats operate within one mile of the coast. For the draft
NONROAD2002 model, it will be assumed that personal watercraft and boats with outboard
engines operate no more than  a quarter mile off the coast while boats with inboard engines
operate up to two miles off the coast.  This modification to the water surface area allocation
factors attempts to at least address one aspect of the different usage patterns of these types of
boats. The effect of this modification  is that a greater number of larger boats with inboard
engines will tend to be allocated to coastal counties while the allocation of smaller personal
watercraft and boats with outboard engines will tend to shift toward waterways in inland
counties.
                                           12

-------
       The EPA hopes to continue to investigate ways to improve upon the water surface area
allocation method currently used in the model and also explore whether there might be other
methods and data available to use in NONROAD to better allocate the population of recreational
marine equipment for all of the counties in the U.S.  Local surveys of recreational boating
activity focusing on things such as marina and boat ramp usage offer the most accurate means to
assess boat populations and activity at the county level. One early stakeholder comment
suggested using data  on boating violations and accidents to allocate recreational marine
equipment. Surveys better capture the actual activity on local lakes, rivers, and other waterways,
as well as account for boats registered in one county but used in another. If States, regional air
organizations, and local air pollution control districts have such types of data, then the EPA is
interested in learning about them. Furthermore, EPA encourages state, regional, and local air
organizations to use these local data in the NONROAD model for county-level boat populations,
subject to appropriate guidance.

Recreational Equipment (except for snowmobiles and golf carts)

       The allocation of recreational equipment, such as all terrain vehicles (ATVs) shares the
same challenge as the allocation of recreational boats, namely where the equipment is registered,
purchased, serviced, and stored is usually not the location where the equipment is actually used.
Because of convenience, people tend to purchase recreational equipment, like other products,
near where they live.  Hence, most recreational equipment is purchased in urban and suburban
areas, where the majority of the U.S. population lives, and this equipment also  is registered,
stored and serviced in these areas.  However, there are relatively few places in urban and
suburban areas where it is possible and legal to operate recreational equipment. Generally
speaking,  recreational equipment usage tends to be concentrated in rural and semi-rural areas
near a metropolitan area; such areas are conveniently accessible to the owners of most of the
recreational equipment, have more area that is attractive for recreational equipment use, and tend
to impose fewer restrictions on recreational equipment use than more densely populated areas.
Thus far, however, the EPA has not found an adequate allocation factor that can be used  with the
current allocation method contained in the model (discussed in Section IE, Methodology) to
reflect this population/activity distribution pattern.

       The NEVES report used CBP data set SIC 557 (number of motorcycle establishments) to
allocate recreational equipment to the county level. However, this data is not available for one or
more counties in some States. The NEVES report also used SIC 55  (number of employees in auto
dealerships and service stations), of which SIC 557 is a subset. Neither of these data sets provide
a reasonable allocation  factor for recreational equipment because most motorcycle
establishments, auto dealerships and service stations are located in urban and suburban areas
instead of rural and semi-rural areas where most recreational equipment activity occurs.

       Registration data also exist for ATVs in most States, but there may be some cases where
these data are not available or up to date. Sales data also are available from PSR or
manufacturers. However, using registration or sales data as allocation factors presents the same
                                           13

-------
drawback as using the CBP data: the location of population does not correlate well with the
location of recreational equipment activity outside of heavily urbanized or suburbanized areas.

       The EPA has considered using the simple alternative approach of allocating recreational
equipment population and activity based on the inverse of population or population density (i.e.,
the higher the population or population density of a county, the less recreational equipment
activity is allocated to that county). The EPA has  also considered allocating activity based on the
amount of non-urbanized land area per county within a State. While these approaches would
reduce the risk of overestimating urban activity, they would tend to overestimate activity in
remote rural areas, such as the North Slope of Alaska, Michigan's Upper Peninsula, the Mojave
Desert, or the Texas Panhandle. Yet another approach would use employment or Gross Domestic
Product (GDP) economic activity directly associated with recreational equipment usage to
allocate recreational equipment. However, the EPA has been unable to locate these data at the
county level. The CBP database does not include  recreational equipment GDP data, and the
Bureau of Economic Analysis (BEA) only tracks  GDP data down to the state level.

       Given these problems, the EPA chose to use the number of camps and recreational
vehicle park establishments (CBP SIC 7030) in the June 1998 and April 1999 (Tier 2) draft
versions of NONROAD. CBP SIC 7030 includes sporting and recreational camps, trailer parks,
and campsites. The data subset from CBP SIC 7030 containing the number of establishments
offers better coverage of the U.S. than the subset  containing the number of camp and recreational
vehicle park employees. The data subset containing the number of employees appears to be
missing information for areas known to contain national and state parks, near which camps and
recreational vehicle parks are likely to be located. The EPA acknowledges that this approach may
not adequately account for recreational equipment being used on private and public lands that are
not associated with and/or adjacent to camps and  recreational vehicle parks. In addition, using
the number of camps  and recreational vehicle parks as an allocation factor is only loosely
correlated to the level of recreational equipment activity occurring in a county. At the present
time, however, the EPA is not aware of other allocation methods that are both practical and
reasonably accurate to allocate from state to county population and  activity.

       For the April 2000 draft version of NONROAD, the EPA was able to use state population
estimates obtained from the Motorcycle Industry  Council (MIC) for offroad motorcycles and
ATV's to allocate population and activity of these vehicles from national to state levels. This
method resulted in allocation factors that were not greatly different  from those used in the earlier
draft versions of NONROAD. This hybrid method of allocating from national to state levels
using state population estimates, and allocating from state to county levels according to number
of RV parks is also used for the draft NONROAD2002 model.  The EPA welcomes suggestions
of alternative methods and requests reviewers and stakeholders with such suggestions to  share
them with the team.
                                           14

-------
Golf Carts

       Golf carts have a different pattern of usage from other types of equipment in the
recreational category. Unlike ATVs or snowmobiles, golf carts are predominantly used in a
central location (golf courses), which is usually within or close to an urban/suburban area. The
EPA initially planned to use public golf course employees (CBP SIC 7992) as an allocation
factor for this equipment type. However, the EPA has discovered that these data are incomplete
for many counties in the U.S. For example, no employees of golf courses are reported for the
state of Colorado. The EPA decided to use this factor as a temporary place holder in the June
1998 draft version of NONROAD. For the April  1999 (Tier 2) version of the model, the EPA
switched to using the number of golf courses by county for allocating golf carts.

       Using the number of golf courses to allocate golf carts and their emissions to the county
level does not provide a precise reflection  of golf cart population or activity. Like the allocation
factor presently being used to allocate the other types of recreational equipment (the number of
camps and recreational vehicle parks), the relationship between the number of golf courses on the
one hand and the population and activity level of golf carts and on the other is a loose one. The
population and activity of golf carts at a given golf course depends on the size, popularity, and
type of course.  A large, popular, 36-hole championship golf course will have more golf carts  that
are used more intensively than a small, less intensively used 9 hole course. The location of a golf
course also affects golf cart activity. A golf course adjacent to an urban area or in a suburban area
will tend to have more players than one located in a rural area, resulting in higher golf cart
activity at the urban or suburban course. An additional complication is that many golf courses use
electrically powered carts instead of carts using gasoline-powered engines. However, the EPA
does not know of any nationally applicable allocation factors that account for these influences.
Therefore, the EPA plans to continue using the number of golf courses as an  allocation factor in
the draft NONROAD2002 model. However, if available, the use of local activity and population
data may be the best alternative. The EPA welcomes comments from stakeholders and other
interested reviewers concerning the existence  of better factors or methods to allocate golf carts
from the state level to the county level, as well as sources of local data on golf cart activity and/or
population.

Snowmobiles

       The allocation of snowmobile activity presents the same challenges as for other
recreational equipment (except for golf carts), but it is additionally complicated by the need to
take snowfall into account. For the June 1998 draft version of the model, time and resource
constraints did not allow the EPA to properly  address how to properly allocate snowmobiles. As
a result, the EPA set state and county snowmobile populations to zero for this first version of the
draft model, but it could calculate national snowmobile emissions. For the April 1999 (Tier 2)
draft version of the model, the EPA decided to use registration data of snowmobiles for national
and state populations. Using these data avoids allocating snowmobiles to states without
significant or any snowfall. There are some states that do not require snowmobiles to be
                                           15

-------
registered, such as Connecticut, but data were not available to estimate snowmobile populations
in these states. The same allocation factor used for other types of recreational equipment, the
number of camps and recreational vehicle park establishments, is used to allocate snowmobile
populations to the county level in the April  1999 (Tier 2) version of NONROAD. However, the
EPA realizes that the use of this allocation factor is still inadequate to allocate snowmobiles to
the county level, since snowmobiles need to be restricted to rural counties within a state that
receive frequent snowfall that results in significant and persistent snow coverage.

       Snowmobile trail mileage by county presents an apparently logical method to allocate
snowmobiles, since they are only located where the annual snowfall would support snowmobile
use and the amount of activity would be reflected by how many miles of trails a county would
have. Most states where snowmobiles are used have trail maps, but these maps vary significantly
in quality and may or may not be  scaled accurately. In addition, the EPA found that states with
snowmobile trails do not keep track of the mileage of these trails by county. Trail mileage also
does not necessarily provide an accurate reflection of activity, since it does not capture how
intensively a given trail is used, and it does not account for off-trail snowmobiling.

       For the draft NONROAD2002 model, the state snowmobile registration data used in the
April 1999 (Tier 2) draft version has been updated and an attempt made to account for
unregistered snowmobiles, since anecdotal information suggests that these may account for a
significant portion of total snowmobile emissions. To allocate snowmobiles from the state level
to counties with sufficient snowfall, the model uses a minimal average annual snowfall limit of
forty inches  and the allocation factors for counties receiving less than forty inches of snow have
been set to zero, similar to the method used for snowblowers (discussed above). The annual
average snowfall data are available from NOAA. The model uses inverse human population as
the county allocation factor to ensure that snowmobiles are allocated to rural counties where
snowmobile trails would be located. An exception to using inverse human population has been
made for Alaska, since using inverse human population would allocate snowmobiles to the
numerous areas of Alaska that are uninhabited and inaccessible.  Since most of the populated
parts of Alaska are largely rural to begin with, human population is directly used in conjunction
with snowfall data.

Airport Ground Support Equipment

       For the June 1998, April 1999 (Tier 2), and 2002 draft versions of NONROAD, the EPA
decided to use the number of people employed in air transportation by county (CBP SIC 4500) to
allocate ground support equipment. However, this indicator may include employees that are not
directly connected to aircraft operations, such as airline reservation staff and ticket agents. Using
this factor may lead to an overestimation of aircraft ground support equipment population and
activity in a  county, especially in  counties that either have airports with one or more airline
"hubs" or that do not have a commercial airport but have branch ticket offices for various
airlines. EPA plans to develop a better method for allocating ground support equipment for the
final version of the model.
                                           16

-------
Light Commercial and Industrial Equipment

       For light commercial and industrial equipment, the EPA chose to use the same indicators
as those used in the 1991 NEVES report in the June 1998 and April 1999 (Tier 2) draft versions
of the model. For light commercial equipment, the EPA decided to use the number of wholesale
establishments by county (CBP SIC 50). For industrial equipment, the EPA chose to use the
number of employees in manufacturing (CBP SIC 20). Analyses done for the NEVES report
showed that these indicators provided reasonable predictors of light commercial and industrial
equipment populations. Because these types of equipment are expected to remain close to a fixed
central base of operations, as opposed to types of equipment that tend to move around a wide
area (e.g., construction equipment), one would expect that the number of establishments and
employees would be good factors with which to allocate light commercial and industrial
equipment populations, respectively. The EPA acknowledges that the number of establishments
may not be a good indicator of activity, which is a function of the size of the average
establishment and the mix of establishment types in addition to the absolute number of
establishments. The draft NONROAD2002 model uses these same allocation factors. The EPA
requests comment on possible sources of alternative activity-related allocation factors such as the
dollar value of commercial, wholesale, or industrial output, including their advantages and
disadvantages relative to the number of employees and establishments and their availability on a
county-by-county basis for the entire U.S.

Logging Equipment

       For logging equipment, the EPA chose to use the number of employees in logging (CBP
SIC 2410) combined with the number of employees in saw and planing mills (CBP  SIC 2420) for
the June 1998 And April 1999 (Tier 2) draft versions of NONROAD. However, the number of
employees in saw and planing mills allocated logging equipment populations to unlikely places
such as Southern California and various urban areas in Texas. For the final version of
NONROAD, the EPA plans to use only the number of employees in logging operations to
allocate logging equipment. The number of logging employees should provide a good reflection
of logging equipment activity. Generally speaking, each employee most likely uses one piece of
equipment at a time. Economic pressures would discourage logging companies from having
expensive logging equipment sit idle for long periods of time, which suggests that the number of
employees in logging operations can also serve as an allocation factor for logging equipment
populations.  The draft NONROAD2002 model uses this allocation factor as well. The EPA
requests comment on potentially more reliable activity and population allocation factors, such as
the number of acres of forest harvested per county, and on possible sources of data for those
alternatives.
                                          17

-------
Oil Field. Underground Mining. Railroad Maintenance, and AC/Refriaeration Equipment

       To allocate oil field equipment population and activity, the EPA decided to use the
number of employees in oil and gas extraction operations (CBP SIC 1300) for the June 1998
draft version of NONROAD. For the April 1999 (Tier 2) draft version of the model, the EPA
decided to continue using this factor. The draft 2002 version of NONROAD also uses this factor.
To allocate underground mining equipment population and activity in the June 1998 draft version
of NONROAD, the EPA decided to use the number of employees engaged in metal mining (CBP
SIC 1000),  and the April 1999 (Tier 2) version also contains this factor. However, the EPA no
longer believes that this indicator represents underground mining operations because most metal
mining is now performed above ground through the excavation of large open pits. As a result, the
EPA has decided to use CBP employment data on coal mines (CBP SIC 1200) to allocate
underground mining equipment in the draft NONROAD2002 model. Employment data provide
reasonable allocation factors for oil field and underground mining equipment activity because a
proportional relationship between the number of employees and the amount of equipment they
use is likely to exist for both of these categories. Furthermore, economic incentives to avoid
leaving expensive equipment idle suggests that activity and equipment populations will be
closely correlated. Finally, these types of equipment tend to not move around (unlike
construction equipment, for example), so the location of activity for oil field and underground
mining equipment usually coincides with the location where the employees are based. A
production-based allocation factor, such as gallons of oil pumped or tons of coal mined, might be
an even better allocation factor, but the EPA has been unable to find these types of activity-
related data at the county level.

       To  allocate railroad maintenance equipment populations, no applicable CBP SIC exists
and no other reasonable alternatives have been found. The EPA is investigating the possibility of
using railroad track mileage for the final version of NONROAD, if such data are available by
county. The EPA has also been unable to find a targeted allocation factor for air
conditioning/refrigeration units used  on trucking trailers to keep food cold and fresh while it is
transported to restaurants and markets. In the absence of an allocation factor that is more directly
related to such activity, the EPA decided to use human population as the allocation factor for
railroad maintenance and AC/refrigeration equipment in the June 1998 and April 1999 (Tier 2)
versions of the draft NONROAD model. The draft NONROAD2002 model also uses human
population for these equipment categories.

       The EPA acknowledges that human population is unlikely to correspond well to the
location and usage of railroad maintenance equipment. The EPA requests comments and ideas
for a better allocation factor for this category of equipment. However, unless a better factor is
found, the EPA will continue to use human population to allocate railroad maintenance
equipment. Human population may be a sufficiently reliable indicator of AC/refrigeration unit
populations and activity levels, since the number of units being used to transport food into or
within a given county is likely to be directly related to the size of the population in the county.
                                           18

-------
The EPA welcomes comments concerning better allocation factors for these equipment
categories.
                                          19

-------
                                                                Table 1
                                             NONROAD Surrogate Allocation Factors
    Nonroad
   Equipment
    Category
           Allocation Factor
       June 1998 Draft Version
           Allocation Factor
    April 1999 Draft Tier 2 Version
           Allocation Factor
          2002 Draft Version
Lawn and Garden
Residential
Number of single and double (duplex)
family housing units from 1990 Census by
county.
Number of single and double (duplex)
family housing units from 1990 Census by
county adjusted by 1997 county human
population estimates from U.S. Census
Bureau.
Same as April 1999 draft version.
Lawn and Garden
Commercial
Number of employees in landscape and
horticultural services, County Bus.
Patterns (CBP), Standard Industrial Code
(SIC)78.
Same as June 1998 draft version.
Same as June 1998 draft version.
Residential
Snowblowers
Snowblowers set to zero pending
implementation of proper allocation based
on snowfall.
Populations allocated to states based on
snowmobile registration data by state, then
allocated to counties using same factor as
residential lawn and garden.
Same as residential lawn and garden, but
allocation factors for counties with
snowfall less than 15 inches set to zero.
Commercial
Snowblowers
Snowblowers set to zero pending
implementation of proper allocation based
on snowfall.
Populations allocated to states based on
snowmobile registration data by state, then
allocated to counties using same factor as
commercial lawn and garden
Same as commercial lawn and garden, but
allocation factors for counties with
snowfall less than 15 inches set to zero.
Construction
Total dollar value of construction by
county.
Same as June 1998
Categories (e.g., housing, commercial
buildings, public works construction) of
F. W. Dodge construction dollar value
data weighted by 1998 Environ survey of
construction equipment activity in
Houston, TX and then totaled.
Agricultural
Harvested cropland (U.S. Census Bureau,
USA Counties database).
Same as June 1998 draft version.
Same as June 1998 draft version.
                                               20

-------
    Nonroad
   Equipment
    Category
           Allocation Factor
       June 1998 Draft Version
           Allocation Factor
    April 1999 Draft Tier 2 Version
           Allocation Factor
          2002 Draft Version
Recreational
Marine
Ratio of county water surface area to total
national water surface area.
Same as June 1998 draft version.
Population will be allocated to states using
ORNL fuel consumption distribution.
Allocation to counties using water surface
area with different operating limits from
shore for personal watercraft, outboards,
and inboards.
Recreational
(except
snowmobiles and
golf carts)
Number of camps and recreational vehicle
park establishments (CBP SIC 7030).
Number of camps and recreational vehicle
park establishments (CBP SIC 7030).
State motorcycle/ATV population
estimates from MIC for national to state
allocation and number of camps and
recreational vehicle park establishments
(CBP SIC 7030) for state to county
allocation.
Snowmobiles
Snowmobiles set to zero pending
implementation of proper allocation based
on snowfall.
Populations allocated to states based on
ISMA snowmobile registration data by
state, then allocated to counties using the
same factor as other recreational
equipment, the number of RV park/camp
establishments.
Registration data from ISMA to be used
for state populations. Will allocate to
counties using snowfall limit of 40 inches
and inverse human population. For Alaska,
will use direct human population.
Golf Carts
Number of public golf course employees
(CBP SIC 7992).
Number of public golf courses (CBP SIC
7992).
Same as April 1999 draft version.
Aircraft Ground
Support
Equipment
Number of employees in air transportation
(CBP SIC 4500).
Same as June 1998 draft version.
Revised to be based on FAA number of
landings and takeoffs (LTOs) of
commercial aircraft.
Light Commercial
Number of wholesale establishments (CBP
SIC 50).
Same as June 1998 draft version.
Same as June 1998 draft version.
Industrial
Number of employees in manufacturing
(CBP SIC 20)
Same as June 1998 draft version.
Same as June 1998 draft version.
                                               21

-------
Nonroad
Equipment
Category
Logging
Oil Field
Equipment
Railroad
Maintenance
Equipment
Underground
Mining
Equipment
AC/Refrigeration
Equipment
Allocation Factor
June 1998 Draft Version
Number of employees in logging plus saw
and planing mills (CBP SIC 2410 and
2420).
Number of employees engaged in oil and
gas extraction (CBP SIC 1300).
1990 human population.
Number of employees engaged in metals
mining(CBP SIC 1000).
1990 human population.
Allocation Factor
April 1999 Draft Tier 2 Version
Same as June 1998 draft version.
Same as June 1998 draft version.
1990 and 1996 human population.
Same as June 1998 draft version.
1990 and 1996 human population
Allocation Factor
2002 Draft Version
Number of employees in logging
(CBP SIC 24 10).
Same as June 1998 draft version.
Might be revised using different allocation
factor for final version, pending review.
Number of Employees in coal mining
(CBP SIC 1200).
Will update with latest human population
data available.
22

-------
Endnotes

1.  Environmental Protection Agency, Office of Air and Radiation. Nonroad Engine and Vehicle
   Emission Study. 21A-2001, November 1991.

2.  Irwin Broh & Associates, Inc., NMMA Boat Usage Survey, prepared for the National Marine
   Manufacturers Association, Des Plaines, IL, August 1991.
                                          23

-------