United States          Office of Water (4503F)     EPA 841 -D-01 -005
          Environmental Protection     Washington, DC 20460     August 1, 2001
          Agency             www.epa.gov/ow


v>EPA    The  National Costs to Implement


          TMDLs (Draft Report):




           Support Document # 2

           for "The National Costs of the Total Maximum Daily Load Program
           (Draft Report)," August 2001,  USEPA

-------
Support Document #2
for "The National Costs of the Total Maximum Daily Load Program"
Draft report, July 2001, USEPA
                                      DRAFT
         THE NATIONAL COSTS TO IMPLEMENT TMDLS
                                   August 1, 2001
                                    Prepared for:

                           U.S. Environmental Protection Agency
                         Office of Wetlands, Oceans and Watersheds
                               1200 Pennsylvania Ave., NW
                                 Washington, D.C. 20460

                                     John Wilson,
                                    Richard lovanna
                             EPA Work Assignment Managers
                                     Prepared by:

                                   Environomics, Inc.
                            4405 East-West Highway, Suite 307
                                  Bethesda, MD 20814

                                         and

                                    Tetra Tech, Inc.
                               10306 Eaton Place, Suite 340
                                   Fairfax, VA 22030

                                         under

                              EPA contract No. 68-C-99-249

-------
                                 EXECUTIVE SUMMARY

                                       A. INTRODUCTION

        This analysis estimates the potential costs for point and nonpoint pollutant sources that are likely to
result from implementation of TMDLs nationwide.  The States have currently listed nearly 22,000 impaired
waters under Section 303(d) of the Clean Water Act (CWA) as waters that will not meet applicable water
quality standards even after the application of technology-based effluent limitations.  Nearly all of these
303(d) listed waters will need to have TMDLs established.1 TMDL requirements specify that when a
TMDL is developed for an impaired water body, maximum loads must be assigned for the specific point
and nonpoint sources that discharge the impairment pollutant(s) affecting the water body.  The assigned
maximum loads must be sufficient to  achieve applicable water quality standards with a margin of safety.

        In a TMDL, some point or nonpoint sources will be assigned allowable loads that are less than the
loads they currently discharge. These sources will presumably incur some costs to reduce their loads from
current levels to the lower levels assigned by the TMDL. It is these costs to reduce pollutant source loads
that we attribute to the TMDL program and that we estimate in this report.2

        More specifically, we estimate the costs that TMDLs will engender for point and nonpoint sources
in order to meet water quality standards for the set of impaired waters included in the States'  1998 303(d)
lists. For these impaired waters, we attribute to the TMDL program the costs of the additional controls
that pollutant sources will need to implement beyond a baseline that includes: 1) Whatever controls were in
place at point and nonpoint sources as of when the 1998 303(d) lists were developed; and 2) Assumed
compliance with all applicable technology-based requirements. Viewed in another way, the analysis
estimates the incremental costs to pollutant sources  of achieving water quality standards relative to a
baseline of their controls in place in 1998, but excluding the costs of whatever amount  of this further
progress will be achieved through meeting technology-based requirements that were unmet as of 1998.

B. THREE TMDL PROGRAM SCENARIOS

        We estimate the costs for pollutant sources  to meet TMDL allocations under each of three broad
scenarios: a "Least Flexible TMDL Program" scenario; a "Moderately Cost-effective TMDL Program"
scenario; and a "More Cost-Effective TMDL Program" scenario.

        I. The "Least Flexible TMDL Program" scenario.  This scenario explores what costs to pollutant
sources would result if the nation chose to restore the currently impaired waters under a TMDL program in
        A 303(d) listed water will not require a TMDL to be developed if no pollutant can be identified that is
        responsible for the impairment.  Such waters are said to be impaired by "pollution" but not pollutants.
        Causes of impairment for such waters might include flow modification, habitat alteration, and the like.

        Note that some of these same costs could be incurred in the absence of a TMDL, either before the water
        body is listed or during the period between listing and the development of the TMDL.  Section
        301(b)(l)(c) of the Clean Water Act requires water quality-based effluent limits (WQBELs) for NPDES
        permittees if it is determined that these discharges would "cause or contribute to a violation of a water
        quality standard."

                                              ES-  1

-------
which every source affecting an impaired water would be required to implement further control measures,
rather than a more calibrated approach.  We assume that States would continue the water quality-based
approaches that have commonly been used to date in situations where a TMDL has not been developed.
Under CWA section 301(b)(l)(c), NPDES permits for point source dischargers must include limits
necessary to meet water quality standards.  Similarly,  under section 319 of the CWA, States must address
the nonpoint sources that contribute to impairment of water bodies  (though not necessarily through
regulatory mechanisms, as are mandated for point sources through  the NPDES program). As States carry
out this "Least Flexible TMDL Program" approach for an impaired water body, we assume they would
need to address every point and nonpoint source that appears to contribute to impairment of the water body.

       2. The "Moderately Cost-effective TMDL Program" scenario.  The "Moderately Cost-effective
TMDL Program" scenario differs from the first scenario in that it presumes (consistent with current
implementing regulations) that the TMDL will: 1) Start with a holistic assessment of the impaired water
body and all the sources that affect it; and 2) Require carefully chosen load reductions from pollutant
sources that together will be just sufficient to achieve water quality standards with a margin of safety. The
TMDL determines how much load from all the sources together can be tolerated,  and allocates this
allowable load in some manner among the responsible sources.  Without this more flexible TMDL, every
source that discharges the impairment pollutant will presumably need to implement measures to abate its
discharge (the previous "Least Flexible TMDL Program"  scenario). With a more moderately cost-effective
TMDL, a much finer calculation is made, and often not every source will need to abate its discharge. The
TMDL determines exactly which sources will need to  reduce their loads, and by how much.  Depending  on
the severity of the impairment, with a moderately cost-effective TMDL somewhere between a few and
many of the sources discharging the impairment pollutant may not have to reduce their discharge at all.  In
addressing each source in isolation and requiring further controls from all of them individually, the previous
"Least Flexible TMDL Program" scenario is likely to substantially overshoot the load reduction needed to
attain water quality standards. Under this second scenario, the number of pollutant sources that have to
take any action, should, in most cases, be reduced.

       3. The "More Cost-Effective TMDL Program" scenario.  Neither the Clean Water Act nor EPA's
implementing regulations prescribe how a total maximum  daily load is to be allocated among the sources
that discharge the impairment pollutant.  The State may assign responsibilities among sources for load
reductions as the state wishes.  Different allocations will result in different total costs of achieving the
desired total load reduction, as a function of the differing costs per  pound for the various pollutant sources
to reduce their loads. In general, the total costs of achieving the target load reduction will be lower if the
sources with lower per unit control costs are assigned  responsibility for achieving the bulk of the desired
total load reduction. We use the term "cost-effective wasteload allocation" to denote a situation in which
the state attempts  to reduce aggregate costs by assigning responsibility for achieving most of the total
desired load reduction to sources that have relatively low costs of achieving load reductions. Alternatively,
the same economically efficient result (achieving a desired total load reduction in a lower cost manner) can
be achieved, in theory, given any initial allocation of control responsibilities, if "trading"  is allowed. With
trading, any source that is assigned responsibility for a load reduction is free to achieve that load reduction
itself, or to buy the equivalent load reduction from another source that might be able to provide it at lesser
cost. Whatever the initial allocation, trading will tend ultimately to elicit load reductions  from the lowest
cost sources.

       The "More Cost-Effective TMDL Program" scenario recognizes the possibility of reducing TMDL
costs to dischargers through either "cost-effective wasteload allocations" or through trading, or both.

                                              ES-2

-------
Either of these approaches would reduce the eventual costs to dischargers well below what they would be if
TMDLs assigned load reductions on a cost-neutral basis (e.g., if load reductions were determined on a
simple proportional rollback basis).

C. COSTS FOR POLLUTANT SOURCES UNDER THE THREE SCENARIOS

       For each of the three scenarios, we estimate the costs for controlling the point and nonpoint sources
that affect each of the currently listed 303(d) waters. These cost estimates are for the incremental controls
- relative to those that existed in 1998 or so when the impaired waters were listed - that will be needed to
achieve water quality standards in the listed waters. To the extent that some of the needed progress will be
achieved through compliance with as-yet-unmet technology-based standards, we do not count such costs, as
they are attributable to sections of the Clean Water Act other than §303. To the extent that some of these
costs have already been incurred since the waters were listed, the cost estimates we present here overstate
the costs that remain for dischargers.

       We estimate the costs to pollutant sources in first quarter 2000 dollars.  All costs include capital
and operating and maintenance costs, combined into a single annualized cost figure. The cost estimates
that we present represent levelized annual amounts beginning in the year 2000 that will continue each year,
forever. TMDLs are assumed to be developed at an even pace over the 15 years from now through 2015,
consistent with the deadline for TMDL development established by the new regulations. The average
source is assumed to begin incurring its costs to implement TMDL allocations five years after the TMDL
affecting that source is developed. The timing of compliance investments by sources is assumed to be
identical under each of the three TMDL program scenarios.  A real discount rate of 7  % is used.3

       Under a least flexible TMDL program, pollutant sources will incur costs estimated at $1.9 - $4.3
billion per year to implement controls for the nearly 20,000 waters for which TMDLs will be developed
(among the nearly 22,000 impaired waters). This is equivalent to approximately $95,000 -  $215,000
annually in implementation costs per water body. In addition to these costs, nonpoint sources may realize
cost savings of up to perhaps $1 billion per year from the management measures we project that they are
likely to adopt.

       In contrast, to achieve the same results in the same time frame, but with a moderately cost-effective
TMDL program, pollutant sources will need to spend only $1.0 - $3.4 billion per year. This is a cost
reduction of 21 -44%.  A moderately cost-effective TMDL program saves pollutant sources money
because TMDLs will involve careful calculations to determine the load reduction that will be sufficient to
achieve water quality standards. In the absence of a moderately cost-effective TMDL, though, a State is
likely to require further controls from all sources that discharge the impairment pollutant.

       Under the "More Cost-Effective TMDL Program" scenario, costs to dischargers may decline by
roughly 7 to 13 % ($140 - $235 million annually) from those that would occur under the second scenario if
each TMDL were to require equivalent control efforts from all sources needing to implement controls.
These savings that  we estimate from cost-effective WLAs or trading represent only the savings available
from shifting some point source control responsibilities to nonpoint sources (i.e., "point/nonpoint trading").
3       For the final report, costs will also be estimated assuming a real discount rate of 3% per year.

                                              ES-3

-------
We have not been able to estimate additional savings that might occur from other sorts of trading (e.g.,
"point/point" trading, pretreatment trading, trading among nonpoint sources).

                                          Exhibit ES -1
                   Estimated Costs for Pollutant Sources to Implement TMDLs
Type of Source
Point sources
Nonpoint sources
Total implementation costs
Potential savings for nonpoint sources
Annual Costs
(2000 $ in millions)
Least Flexible
TMDL Program
1,082-2,178
783 - 2,162
1,865 - 4,340
undetermined
Moderately Cost-
effective TMDL
Program
812- 1,634
234-1,791
1,046 - 3,425
undetermined
More Cost-
Effective TMDL
Program
625 - 1,321
281 - 1,869
906 - 3,190
undetermined
1. Costs for point source dischargers

        Under the first two scenarios, half or more of the costs will be incurred by point sources. This is
despite the fact that point sources affect only about 1/4 of the impaired waters while nonpoint sources
affect more than 90 % of them. Under the third scenario, some point source control responsibilities are
presumed to be shifted to nonpoint sources because of the expected ability of nonpoint sources to abate
loads at lower costs per pound. Even so, point sources may still incur the majority of the implementation
costs.
                                          Exhibit ES - 2
                    Estimated Costs for Point Sources -- Least Flexible TMDLs
Type of Source
Industrial dischargers
Indirect dischargers (metals)
POTWs
Total
Annual Costs
(2000 $ in millions)
Low Est.
676
10
396
1,082
High Est.
1,465
16
697
2,178
Number of Affected Facilities
Low Est.
3052
at 148 POTWs
1094
4,146
High Est.
8557
at 3 12 POTWs
3335
11,893
                                             ES-4

-------
                                         Exhibit ES - 3
                              Estimated Costs for Point Sources for
               Moderately Cost-effective and More Cost-Effective TMDL Programs
Type of Source
Industrial dischargers
Indirect dischargers (metals)
POTWs
Total
Potential savings from cost-
effective wasteload allocations
Annual Costs
(2000 $ in millions)
Low Est.
507
8
297
812
(187)
High Est.
1,099
12
523
1,634
(313)
Number of Affected Facilities
Low Est.
2289
at 111
POTWs
821
3,110
1,251
High Est.
6418
at 234 POTWs
2502
8,919
2,066
       The low and the high estimates shown above reflect differing judgments about how far upstream of
an impaired water can there typically be point sources that contribute to the water body's impairment. The
lower estimate assumes that only point sources  discharging the impairment pollutant directly into the
impaired water contribute to impairment.  The upper estimate assumes that point sources can contribute to
impairment from as far away as 25 miles upstream (if the impairment pollutant is BOD, ammonia or toxic
organic chemicals) or 50 miles upstream (if the impairment pollutant is nutrients or metals). We believe
these two estimates provide reasonable lower and upper estimates for the geographic extent of point
sources that will be judged as relevant in TMDLs.

       There are roughly 70,000 individually permitted point source dischargers in the nation.
Somewhere between 6 and 17 percent of them appear to contribute to the impairment of a 303(d) water,
and would likely be addressed by efforts to restore the nation's impaired waters under a least flexible
TMDL program.  Based on the experience from a sample of recently developed TMDLs, only about 3/4 of
these sources  (about 3,000 - 9,000 point source dischargers) will likely incur costs under a moderately
cost-effective TMDL program.  Of these point sources likely to be affected by the TMDL program,
perhaps 20 to 40 percent of them will incur no or reduced costs if the TMDL program proceeds in more a
cost-effective manner.

2. Costs for nonpoint source pollutant sources

       Costs were estimated for four types of nonpoint sources: agricultural land (including crop,  pasture
and range land), animal feeding operations (AFOs), silviculture, and on-site wastewater treatment systems
(septic tanks,  etc.). Some of the measures that nonpoint sources are likely to implement to achieve  TMDL-
mandated load reductions will yield partly offsetting cost savings (e.g., agricultural nutrient management
planning implemented pursuant to a TMDL can reduce farmers' costs for chemical fertilizers).  Exhibit ES
- 4 shows only the costs and not the cost savings from the management measures that may be implemented
by the four types of nonpoint sources that we analyze.
                                             ES-5

-------
                                          Exhibit ES - 4
                               Estimated Costs for Nonpoint Sources
Type
Agricultural land
crop land
pasture land
range land
Potential savings
AFOs
Potential savings
Silviculture
On-site wastewater treatment systems
Total
Potential savings
Annual Costs (2000 $ in millions)
Least Flexible
TMDL Program
645 - 1,956
5- 11
2-16
(not estimated)
76- 110
(not estimated)
30-42
24-28
783 - 2,162
(not estimated)
Moderately Cost-
effective TMDL
Program
183- 1,632
5- 11
2-16
(not estimated)
13-73
(not estimated)
7-31
24-28
234 - 1,791
(not estimated)
More Cost-
Effective TMDL
Program
Additional costs of
47 - 78 relative to
those incurred under
moderately cost-
effective TMDL
program. (Note that
in the "more cost-
effective" scenario,
point sources control
less and nonpoint
sources control more
by an equivalent
amount)
        There is a very wide range of uncertainty regarding the estimates of potential savings.  They could
range up to perhaps a billion dollars per year.  Additional work is being conducted to narrow this range.

        In the limited time available for this study, costs were not estimated for the further controls that
may be needed for several other potentially important categories of nonpoint sources, including abandoned
mines, contaminated sediments, air deposition, and more.

D. ANALYTICAL METHODOLOGY

        The task of projecting the costs for pollutant sources to meet the requirements of TMDLs for
nearly 22,000 impaired waters  is particularly difficult because the background information necessary for
developing TMDLs has been generated for only a very few of these waters.  We do not know, at this point,
how far out of attainment most of the impaired waters will be found to be, what sources will be found to be
responsible for each impairment, and what degree of load reduction will be required of each responsible
source. In this analysis we must estimate each of these elements now, before the background studies and
actual TMDLs have been developed.  Our analysis necessarily involves many assumptions that we apply to
the relatively limited data that now exists on these impaired waters and the sources that may contribute to
their impairment.

        The analysis is further  complicated by our interest in estimating costs under each of the three
scenarios. In addition to a series of technical assumptions involving the likely content of the eventual
                                              ES-6

-------
TMDLs for the impaired water bodies, we also make assumptions about what will happen differently under
each of the three scenarios.

       In general, our analysis proceeds by first estimating the costs if all pollutant sources contributing to
impairment of an impaired water body were required to implement reasonable measures to reduce their
discharge of the impairment pollutant(s). We assume that this reflects the most that might be required
under a TMDL — Scenario 1, the "Least Flexible TMDL Program". We then estimate the costs for
Scenario 2  by adjusting downward the costs estimated for Scenario  1, assuming that TMDLs in practice
will result in a more precise calculation of how much load reduction is needed from pollutant sources in
order to meet water quality standards.  The total load reduction required of pollutant sources under
Scenario 2  is  less than that which would be obtained if all pollutant  sources contributing to impairment
were to implement abatement measures, as under Scenario 1. The costs of Scenario 3 are then estimated
by estimating the savings compared with Scenario 2 that more cost-effective waste load allocations might
provide relative to the costs under a cost-neutral TMDL program.

       We began by identifying the universe of point and nonpoint sources potentially contributing to
impairment(s) of each of the 303(d)-listed water bodies in the nation. For many listed water bodies, States
identify whether the sources contributing to impairment are point sources, nonpoint sources, or both. For
each water  body cited as impaired by point sources (as well as perhaps other source types), we identified
the specific point sources that might potentially be contributing to the impairment. Similarly, for each
water body cited as impaired by nonpoint sources, we identified the  potentially responsible specific
nonpoint sources. For 303(d) waters for which States have not provided information on the sources of
impairment (e.g., when States cite "unknown" sources of impairment, or simply do not report any source of
impairment),  we identified all potentially relevant point and nonpoint sources and then extrapolated cost
information to them based on relationships we established for waters for which impairment sources were
reported.

       We assumed that the set of point sources affecting a point-source impaired water is somewhere
between two cases:

       •       Case 1,  "within and up stream".  We assume that a point source contributes to impairment
               if it discharges the pollutant of concern within 25 miles upstream of a water body impaired
               by BOD, ammonia or toxic organic chemicals, and within 50 miles upstream of a water
               body impaired by nutrients or metals.

       •       Case 2,  "within only". We assume that a point source contributes to impairment if it
               discharges the  pollutant of concern directly into the impaired water body.

The "within and upstream" case thus identifies a larger set of point sources as contributing to impairment
than does the  "within only" case. We use these two cases to establish upper and lower estimates for point
source costs for each of the three scenarios.  For the "Least Flexible TMDL Program" scenario, we assume
that States  will require further control of all the point sources that contribute to impairment,  under the
within and  upstream case as an upper estimate and under the within-only case as a lower estimate. For the
other two scenarios, we use data from a sample of recently completed TMDLs to estimate the fraction of
the point sources contributing to impairment that are typically required to reduce their loads in actual
TMDLs. In fact, some impaired waters are only moderately impaired, and for these waters TMDLs will
require load reductions from only some of all the point sources that discharge the impairment pollutant.

                                             ES-7

-------
The likelihood that many TMDLs will not require load reductions from all point sources contributing to
impairment makes the costs for the "Moderately Cost-effective TMDL Program" scenario less than for the
"Least Flexible" scenario.  Again, we use the two cases to establish upper and lower estimates for both the
"Moderately Cost-effective TMDL Program" scenario and the "More Cost-Effective TMDL Program"
scenario.

        For nonpoint sources, we assumed that States under the "Least Flexible" scenario would require
further controls of all the nonpoint source activity of the relevant variety that occurs within the same county
or counties as the impaired water body.  For example, if a State identifies a 303 (d) water body as impaired
by animal feeding operations (AFOs) and silviculture, we assumed that the State would require further
controls for all AFOs and all silviculture within the county(s) in which the impaired water body is located.
For the "Moderately Cost-effective TMDL Program" and "More Cost-Effective TMDL Program"
scenarios, again based on the results from a sample of actual TMDLs, we assumed that the State would
make a finer calculation regarding the geographic extent of nonpoint source activity from which load
reductions must be obtained. In the great majority of cases, actual TMDLs have required nonpoint source
controls from watershed areas much smaller than the entire county(s) surrounding the impaired water body.

        As a baseline for cost analysis, we assumed that all these identified affected point and nonpoint
sources have control measures in place equal to the greater of: 1) Their current controls in place;  and 2)
Controls necessary to meet applicable technology-based standards.  We assumed that the load allocations
established under TMDLs  would require all relevant sources to implement the "next treatment step" beyond
their assumed baseline controls in place:

        •      For industrial point sources: The next treatment step consisted of a further treatment
               technology, depending on the specific pollutant, beyond the technologies assumed to be in
               place to meet effluent guideline requirements.

        •      For POTWs: The next treatment step for most pollutants was assumed to be advanced
               secondary treatment, the next increment beyond secondary treatment assumed to be in
               place to meet secondary treatment requirements. When a POTW appeared to need further
               controls specifically for metals, the next treatment step was  assumed to be an enhanced
               local pretreatment program, requiring further controls of the POTW's indirect dischargers
               beyond applicable pretreatment standards in effluent guidelines.4

        •      For agricultural, AFO, silvicultural, and  on-site wastewater system nonpoint sources: We
               assumed there were no Federal technology-based requirements applicable in the baseline
               for these sources. The next treatment step beyond this baseline of no controls was
               assumed to be implementation of a basic  set of best management practices (BMPs) for the
               particular  nonpoint source type as suggested in EPA guidance documents. In estimating
       We made two further important assumptions for POTWs. First, we assumed that TMDLs will require no
       further controls for those POTWs that already provide advanced secondary treatment or better.  Second,
       we assumed that any costs for POTW treatment upgrades that have progressed sufficiently in planning to
       be included in the 1996 Clean Water Needs Survey (CWNS) should not be viewed as incrementally
       attributable to the TMDL program.  In essence, we consider POTW upgrade projects that were already far
       along in planning as of 1996 as predating and not deriving from the programs we analyze in this report.
       Later, we discuss the impact of these two particular assumptions for POTWs on the cost estimates.

                                              ES-8

-------
               the costs of implementing the set of BMPs representing the next treatment step we did,
               where data were available, reflect the fact that some nonpoint sources have already (in the
               baseline) adopted some of the BMPs.

        •       For urban wet weather sources: We did not project any further controls to be needed for
               urban wet weather sources beyond existing technology-based requirements addressing
               CSOs, SSOs and storm water phase I and II. To the extent that TMDLs do ultimately
               require further controls for some urban wet weather sources, we have not estimated these
               costs.

        Our estimates for the costs to pollutant sources under the three scenarios consist of the aggregated
costs for these "next treatment steps" for all the point and nonpoint sources identified as affecting 303(d)
waters.  The two sets of key assumptions underlying the analysis include:

        1.      The assumptions made in identifying the specific point and nonpoint sources that will need
               further controls beyond current levels and technology-based standards in order to achieve
               water quality standards; and

        2.      The assumption that the further  control needed from every identified source is the "next
               treatment step" beyond applicable technology-based requirements.

For many water bodies and many TMDLs, these assumptions may be substantially inaccurate.  For any
given water body, the sources a State might identify as needing further controls may be more or less than
the point sources within the water body and/or 25 or 50 miles upstream and the surrounding county's worth
of nonpoint sources. For any given water body, the additional control efforts needed from the affected
sources may also be more or less than the assumed next treatment step.

        In projecting what future TMDLs are likely to require for the impaired water bodies, we based
several key assumptions on our findings from reviewing the content of a sample of fifteen recently
completed TMDLs. This review is summarized in Appendix A - "Ground-Truthing the Implementation
Cost Analysis Assumptions". This sample of fifteen is smaller than we would like, and it will be increased
for the final version of this report. The major findings from this review of sample TMDLs are:

        •       TMDLs commonly, but not always,  address upstream point sources in addition to those
               point sources discharging the impairment pollutant directly into the impaired water. The
               average situation seems somewhere between the "within only" and the "upstream and
               within" cases;

        •       The aggregate load reduction needed from point sources is often obtained without requiring
               further controls from all of the point sources discharging the impairment pollutant;

        •       The geographic extent of nonpoint sources from which further controls are required is
               typically much less than the entire county(s) surrounding the impaired water;

        •       For both point and nonpoint sources, the degree of load reduction that is required is very
               often less than that which would be achieved if all relevant point and nonpoint  sources
               were to implement "the next treatment step".

                                             ES-9

-------
       More specifically, in estimating the costs of the "Moderately Cost-effective TMDL Program" and
"Cost-Effective TMDL Program" scenarios, we drew the following quantitative relationships from the
results of the fifteen TMDLs:

       •       For point sources. In about half the TMDLs, the aggregate load reduction actually
               required of point sources was roughly equivalent to what would be achieved if all point
               sources contributing to impairment of the water body were to implement "the next
               treatment step". In the other half of the TMDLs, "the next treatment step" for all point
               sources would result in about twice as much aggregate load reduction as was actually
               needed.

       •       For nonpoint sources.  The size of the watershed from which most TMDLs required
               nonpoint source load reductions was far smaller than the size of a typical county. The
               acreage of most nonpoint source TMDL watersheds ranged from about 5% to about 40%
               as large as the acreage of the county(s) within which the impaired water body was located.

These quantitative relationships should be regarded as tentative pending the evaluation of more completed
TMDLs.
                                            ES- 10

-------
                         TABLE OF CONTENTS
EXECUTIVE SUMMARY	ES - 1
     A. INTRODUCTION	ES - 1
     B. THREE TMDL PROGRAM SCENARIOS  	ES - 1
     C. COSTS FOR POLLUTANT SOURCES UNDER THE THREE SCENARIOS 	ES - 3
     D. ANALYTICAL METHODOLOGY	ES - 6

I. METHODOLOGY	I - 1
     A. THREE TMDL PROGRAM SCENARIOS  	I - 1
     B. OVERVIEW OF THE ANALYTICAL APPROACH	1-3
     C. COSTS THAT ARE NOT ESTIMATED	1-8
     D. CROSS-CUTTING ANALYTICAL ISSUES	I - 10

II. IMPLEMENTATION COSTS FOR POINT SOURCES  	II - 1
     A. OVERVIEW  	II - 1
     B. POLLUTANTS FOR ANALYSIS 	II - 2
     C. NEARBY FACILITIES	II - 2
     D. FACILITIES THAT NEED TO REDUCE LOADS	II - 4
     E. COST FUNCTIONS FOR THE "NEXT TREATMENT STEP"  	II - 8
     F. FLOW DATAFOR USE IN COST FUNCTIONS	11-11
     G. COSTS FOR SOURCES AFFECTING WATERS IMPAIRED BY NONPOINT
     SOURCES ONLY 	11-12
     H. COSTS FOR SCENARIO 2 (Moderately Cost-effective TMDL PROGRAM) 	11-15
     I. SAVINGS WITH "COST-EFFECTIVE WASTE LOAD ALLOCATIONS"	11-17
     J. SUMMARY COST ESTIMATES FOR POINT SOURCES 	11-23

III. IMPLEMENTATION COSTS FOR NONPOINT SOURCES	 Ill - 1
     A. COVERAGE OF THE NONPOINT SOURCE ANALYSIS	 Ill - 1
     B. WATERS IMPAIRED BY AGRICULTURE, AFOS, SILVICULTURE AND
     ON-SITE SYSTEMS 	 Ill - 4
     C. THE AMOUNT OF NONPOINT SOURCE ACTIVITY NEEDING FURTHER
     CONTROL FOR AN IMPAIRED WATER BODY 	 Ill - 7
     D. BMPS FOR REDUCING NONPOINT SOURCE LOADS	 Ill - 11
     E. UNIT COSTS FOR THESE BMPS 	 Ill - 14
     F. SUMMARY COST ESTIMATES FOR NONPOINT SOURCES	 Ill - 17

IV. MAJOR ASSUMPTIONS, BIASES, AND UNCERTAINTIES	 IV - 1

APPENDICES
APPENDIX A: GROUNDTRUTHING THE ASSUMPTIONS	A-l
APPENDIX B: TMDL PACE AND TIME LAG SCALE FACTOR 	B-l
APPENDIX C: POLLUTANTS FOR ANALYSIS  	C-l
APPENDIX D: APPROACHES FOR DETERMINING WHETHER A POINT SOURCE IS
     LIKELY TO BE CONSIDERED FOR FURTHER CONTROLS 	D-l
APPENDIX E: DETAIL ON COST FUNCTIONS FOR POINT SOURCES	E-l
APPENDIX F: PROCEDURES FOR ESTIMATING FLOW NEEDING TREATMENT	F-l

-------
APPENDIX G: EXCLUDING COSTS FOR POINT SOURCES AFFECTING
     WATERS THAT ARE IMPAIRED BY NONPOINT SOURCES ONLY	G-l
APPENDIX H: DETAIL ON SIMULATING "COST-EFFECTIVE WASTE LOAD
     ALLOCATIONS"	H-l
APPENDIX I: DETAIL ON ESTIMATED COSTS FOR NONPOINT SOURCES	1-1

-------
                                     I.  METHODOLOGY

A. THREE TMDL PROGRAM SCENARIOS

        We estimate the costs for pollutant sources to achieve the load reductions to be required by
TMDLs under each of three broad scenarios: a "Least Flexible TMDL Program" scenario; a "Moderately
Cost-effective TMDL Program" scenario;  and a "More Cost-Effective TMDL Program" scenario.

        I.  The "Least Flexible TMDL Pro gram" scenario. This scenario explores what costs to
pollutant sources would result if the nation chose to to restore the currently impaired waters under a TMDL
program in which every source  affecting an impaired water would be required to implement further control
measures, rather than a more calibrated approach. We see two possibilities for how the nation might meet
the Clean Water Act goal of restoring impaired waters under this scenario:

               Progressively tighten the nationally uniform technology-based requirements for relevant
               classes of dischargers until all water bodies eventually meet standards.  Although a
               theoretical possibility, this approach would be unreasonably inefficient.  A relatively small
               fraction of all point and nonpoint sources affect impaired water bodies. It would be
               exceedingly costly to require further  controls from the entire nation's worth of some
               category of point or nonpoint sources (e.g., all POTWs in the country, or every
               silvicultural operation) in  order to reduce loads from only the fraction of such sources
               affecting impaired waters.

               Continue the water quality-based approaches that have commonly been used to date in
               situations where a TMDL has not been developed.  Under CWA § 301(b)(l)(c), all
               NPDES permits for point source dischargers must include limits necessary to meet water
               quality standards. Similarly, under § 319 of the CWA, states must address the nonpoint
               sources that contribute to  impairment of water bodies (though not necessarily through
               regulatory mechanisms, as are mandated for point sources through the NPDES program).5

        The second approach to the "Least Flexible TMDL Program" scenario is far more realistic, and we
assume this is what would occur if more flexible TMDLs were not developed.
        Some observers have postulated a different scenario in the absence of TMDLs. These observers
        emphasize that point sources are subject to regulatory controls under the NPDES program, while nonpoint
        sources are generally not subject to federal regulatory controls.  If achieving water quality standards were
        to depend solely on federal regulatory authorities available under the Clean Water Act, states or EPA
        would be able to require further control efforts only from point sources. NPDES permit limits for point
        sources would be progressively tightened as necessary to make up for uncontrolled nonpoint sources.
        Under this least flexible scenario for point sources, many point sources would ultimately need to meet
        exceedingly costly "zero discharge" limits in an attempt to compensate for growing nonpoint source loads.

        We regard this scenario as unrealistic and will not analyze it. For many impaired water bodies, the
        contribution from point sources is minimal or non-existent. Any realistic program to achieve water
        quality standards in all impaired waters must seriously address nonpoint sources as well as point sources.

                                                I- 1

-------
       Our Methodology for Estimating Costs to Pollutant Sources - the"Worst- Case TMDL Program" Scenario

  We follow three steps:

  1.  Identify all the point and nonpoint sources in the nation that appear to discharge an impairment pollutant to one of the
  impaired waters on the 1998 303(d) lists.

  2.  Assume that every such relevant source will be required under the NPDES program or somehow induced under the 319
  program to implement additional measures (beyond those assumed to be in place already to meet existing technology-
  based standards) to abate this discharge.

  3.  Estimate the costs for each source to implement an appropriate "next treatment step" that will presumably sufficiently
  reduce the source's discharge.
        2. The "Moderately Cost-effective TMDL Program" scenario. The "Moderately Cost-effective
TMDL Program" scenario differs from the first scenario in that it presumes (consistent with current
implementing regulations) that the TMDL will:  1) Start with a holistic assessment of the impaired water
body and all the sources that affect it; and 2) Require carefully chosen load reductions from pollutant
sources that together will  be just sufficient to achieve water quality standards with a margin of safety.  The
TMDL determines how much load from all the sources together can be tolerated, and allocates this
allowable load in some manner among the responsible sources.  Without this more flexible TMDL, every
source that discharges the impairment pollutant will presumably need to implement measures to abate its
discharge (the previous "Least Flexible TMDL Program" scenario). With a more moderately cost-effective
TMDL, a much finer calculation is made, and often not every source will need to abate its discharge. The
TMDL determines exactly which sources will need to reduce their loads, and by how much.  Depending on
the severity of the impairment, with a moderately cost-effective TMDL somewhere between a few and
many of the sources discharging the impairment pollutant may not have to reduce their discharge at all.  In
addressing each source in isolation and requiring further controls from all of them individually, the previous
"Least Flexible TMDL Program" scenario is likely to substantially overshoot the load reduction needed to
attain water quality standards. Under this second  scenario, the number of pollutant sources that have to
take any action, should, in most cases, be reduced.
    Our Methodology for Estimating Costs to Pollutant Sources - the "Moderately Cost-effective TMDL Program"
                                               Scenario

  We start with the steps for the "Least Flexible TMDL Program" scenario: a) Identify all sources responsible for impairments;
  2) Estimate costs for all of them to implement an appropriate "next treatment step".  We then:

         Scale these costs down to reflect the average percentage load reduction identified in typical TMDLs relative to the
         load reduction that would be obtained if all sources were to implement the "next treatment step".
        3. The "More Cost-Effective TMDL Program" scenario.  Neither the Clean Water Act nor EPA's
implementing regulations prescribe how a total maximum daily load is to be allocated among the sources
that discharge the impairment pollutant. The state may assign responsibilities among sources for load
reductions as the state wishes.  Different allocations will result in different total costs of achieving the

                                                 1-2

-------
desired total load reduction, as a function of the differing costs per pound for the various pollutant sources
to reduce their loads.  In general, the total costs of achieving the target load reduction will be lower if the
sources with lower per unit control costs are assigned responsibility for achieving the bulk of the desired
total load reduction.  We use the term "more cost-effective wasteload allocation" to denote a situation in
which the state attempts to reduce aggregate costs by assigning responsibility for achieving most of the
total desired load reduction to sources that have relatively low costs of achieving load reductions.
Alternatively, the same economically efficient result (achieving a desired total load reduction in a lower
cost manner) can be achieved, in theory, given any initial allocation of control responsibilities, if "trading"
is allowed.  With trading, any source that is assigned responsibility for a load reduction is free to achieve
that load reduction itself, or to buy the equivalent load reduction from another source that might be able to
provide it at lesser cost. Whatever the initial allocation, trading will tend ultimately to elicit load reductions
from the lowest cost sources.

        The "More Cost-Effective TMDL Program" scenario recognizes the possibility of reducing TMDL
costs to dischargers through either "more cost-effective wasteload allocations" or through trading, or both.
Either of these approaches would reduce the eventual costs to dischargers well below what they would be  if
TMDLs assigned load reductions on a cost-neutral basis (e.g., if load reductions were determined on a
simple proportional rollback basis). We expect that pressure to adopt cost-minimizing approaches will
build, and more TMDLs will tend toward this "more cost-effective" model. Note, though, that there may
be some instances where other concerns (e.g., equity, concern about implementation and enforcement
complexities attendant to trading) prevent use of these cost-minimizing approaches.
   Our Methodology for Estimating Costs to Pollutant Sources - the "More Cost-Effective TMDL Program" Scenario

  We start by estimating the costs for the "Moderately Cost-effective TMDL Program" scenario. We then:

         Scale these costs down to reflect the typical percentage cost savings that might be realized through additional
         "cost-effective wasteload allocations" or trading.
B. OVERVIEW OF THE ANALYTICAL APPROACH

        In this analysis, we estimate the costs as of the spring of 2000 for pollutant sources to reduce their
loads as necessary to meet water quality standards for all waters on States'  approved 1998 303(d) lists
(21,845 listed waters and 41,318 causes of impairment cited for these waters). We estimate these costs
under each of the three scenarios we have described.  Implementation costs  are estimated for each of the
listed water bodies and then aggregated nationally. Limited site-specific information was available and
could be processed for all these water bodies and the sources that affect them.  It was not possible to
perform what would in effect be an initial TMDL for each of these water bodies - determine how far each
water body was from attainment, estimate the amount of load reductions necessary to achieve attainment,
and estimate the costs for the relevant pollutant sources to accomplish these load reductions. Instead, we
made several broad assumptions that limited the amount of site-specific information we needed to obtain
and analyze.

        In general, our analysis proceeds by first estimating the costs if all pollutant sources contributing to
impairment of an impaired water body were required to implement reasonable measures to reduce their

                                                1-3

-------
discharge of the impairment pollutant(s).  We assume that this reflects what would happen under an
inefficient or "Least Flexible TMDL Program". This is our Scenario 1.  We then estimate the costs for
Scenario 2 by adjusting downward the costs estimated for Scenario 1, assuming that moderately cost-
effective TMDLs will result in a more precise calculation of how much load reduction is needed from
pollutant sources in order to meet water quality standards.  The total load reduction required of pollutant
sources under Scenario 2 is less than that which would be obtained if all pollutant sources contributing to
impairment were to implement abatement measures.  The costs of Scenario 3 are then estimated by
estimating the savings that cost-effective waste load allocations might provide relative to the costs of a
cost-neutral TMDL program.

        We began by identifying the universe of point and nonpoint sources potentially contributing to
impairment(s) of each of the 303(d) listed water bodies in the nation.  For many listed water bodies, states
identify whether the sources contributing to impairment are point sources, nonpoint sources, or both. For
each water body cited as impaired by point sources (as well as perhaps other source types), we identified
the specific point sources that  might potentially be contributing to the impairment.  Similarly, for each
water body cited as impaired by nonpoint sources, we identified the potentially responsible specific
nonpoint sources. For 303(d)  waters for which States have not provided information on the sources of
impairment (e.g., when States  cite "unknown" sources of impairment, or simply do not report any source of
impairment), we identified all potentially relevant point and nonpoint sources and then extrapolated cost
information to them based on relationships we established for waters for which impairment  sources were
reported.

        We assumed that the set of point  sources affecting a point-source impaired water is somewhere
between two cases:

        •      Case 1, "within and upstream ".  We assume that a point source contributes to impairment
               if it discharges the pollutant of concern within 25 miles upstream of a water body impaired
               by BOD, ammonia or toxic organic chemicals, and within 50 miles upstream of a water
               body impaired by nutrients or metals.

        •      Case 2, "within only". We assume that a point source contributes to impairment if it
               discharges the pollutant of concern directly into the impaired water body.

The  "within and upstream" case thus identifies a larger set of point sources as contributing to impairment
than does the  "within only" case. We use these two cases to establish upper and lower estimates for point
source costs for each of the three scenarios. For the "Least Flexible TMDL Program"  scenario, we assume
that States will require further control of all the point sources that contribute to impairment, under the
within and upstream case as an upper estimate and under the within only case as a lower estimate.  For the
other two scenarios, we use data from a sample of recently completed TMDLs to estimate the fraction of
the point sources contributing  to impairment that are typically required to reduce their loads in actual
TMDLs. In fact, some impaired waters are only moderately impaired, and for these waters TMDLs will
require load reductions from only some of all the point sources that discharge the impairment pollutant.
The likelihood that many TMDLs will not require load reductions from all point sources contributing to
impairment makes the costs for the "Moderately Cost-effective TMDL Program" scenario less than for
the"Least Flexible TMDL Program" scenario. Again, we use the two cases to establish upper and lower
estimates for both the "Moderately Cost-effective TMDL Program" scenario and the "More Cost-Effective
TMDL Program" scenario.

                                              1-4

-------
       For nonpoint sources, we assumed that States under the "Least Flexible TMDL Program" scenario
would require further controls of all the nonpoint source activity of the relevant variety that occurs within
the same county as the impaired water body.  For example, if a State identifies a 303 (d) water body as
impaired by animal feeding operations (AFOs) and silviculture, we assumed that the State would require
further controls for all AFOs and all silviculture within the county(s) in which the impaired water body is
located. For the "Moderately Cost-effective TMDL Program" and "More Cost-Effective TMDL Program"
scenarios, again based on the results from a sample of actual TMDLs, we assumed that the State would
make a finer calculation regarding the geographic extent of nonpoint source activity from which load
reductions must be obtained. In the great majority  of cases, actual TMDLs have required nonpoint source
controls from watershed areas that are much smaller than the area of the entire county(s) surrounding the
impaired water  body.

       As a baseline for  cost analysis, we assumed that all these identified affected point and nonpoint
sources have control measures in place equal to the greater of: 1) Their current controls in place; and 2)
Controls  necessary to meet applicable technology-based standards.  We assumed that the load allocations
established under TMDLs would require some or all of these sources to implement a "next treatment step"
beyond their assumed baseline controls in place:

       •       For industrial point sources: The next treatment step consisted of a further treatment
               technology, depending on the specific pollutant, beyond the technologies assumed to be in
               place to meet effluent guideline requirements.

       •       For POTWs: The next treatment step for most pollutants was assumed to be advanced
               secondary treatment, the next increment beyond secondary treatment assumed to be in
               place to meet secondary treatment requirements.  When a POTW appeared to need further
               controls specifically for metals, the next treatment step was  assumed to be an enhanced
               local pretreatment program, requiring further controls of the POTW's indirect dischargers
               beyond applicable pretreatment standards in effluent guidelines.6

       •       For agricultural, AFO, silvicultural, and on-site wastewater system nonpoint sources: We
               assumed  there were no Federal technology-based requirements applicable in the baseline
               for these  sources.  The next treatment step beyond this baseline of no controls was
               assumed  to be implementation of a basic set of best management practices (BMPs) for the
               particular nonpoint source type as suggested in relevant EPA guidance documents. In
               estimating the costs of implementing the set of BMPs representing the next treatment step
               we did, where data were available, reflect the fact that some nonpoint sources have already
               (in the baseline) adopted some of the BMPs.
       We made two further important assumptions for POTWs. First, we assumed that TMDLs will require no
       further controls for those POTWs that already provide advanced secondary treatment or better.  Second,
       we assumed that any costs for POTW treatment upgrades that have progressed sufficiently in planning to
       be included in the 1996 Clean Water Needs Survey (CWNS) should not be viewed as incrementally
       attributable to the TMDL program .  In essence, we consider POTW upgrade projects that were already far
       along in planning as of 1996 as predating and not deriving from the TMDL program.  Later, we discuss
       the impact of these two particular assumptions for POTWs on the cost estimates.

                                               1-5

-------
        •       For urban wet weather sources:  We did not project any further controls to be needed for
               urban wet weather sources beyond existing technology-based requirements addressing
               CSOs, SSOs and storm water phase I and II. To the extent that TMDLs do ultimately
               require further controls for some urban wet weather sources, we have not estimated these
               costs.

        Our estimates for the costs to pollutant sources under the three scenarios consist of the aggregated
costs for these "next treatment steps" for all the point and nonpoint sources identified as affecting 303(d)
waters.  The two sets of key assumptions underlying the analysis include:

        1.      The assumptions made in identifying the specific point and nonpoint sources that will need
               further controls beyond current levels and technology-based standards in order to achieve
               water quality standards; and

        2.      The assumption that the further control needed from every identified source is the "next
               treatment step" beyond applicable technology-based requirements.

For many water bodies and many TMDLs, these assumptions may be substantially inaccurate.  For any
given water body, the sources a State might identify as needing further controls may be more or less than
the point sources within the water body and/or 25 or 50 miles upstream and the surrounding county's worth
of nonpoint sources. For any given water body, the additional control efforts needed from the affected
sources may also be more or less than the next treatment step we have assumed.

        In projecting what future TMDLs are likely to require for the impaired water bodies, we based
several key assumptions on our findings from reviewing the content of a sample of fifteen recently
completed TMDLs. This review is summarized in Appendix A - "Ground-Truthing the Implementation
Cost Analysis Assumptions". This sample of fifteen is smaller than we would like, and it will be increased
for the final version of this report.  The major findings from this review are:

        •       TMDLs commonly, but not always,  address upstream point sources in addition to those
               point sources discharging the impairment pollutant directly into the impaired water. The
               average situation seems somewhere between the "within only" and the "upstream and
               within" cases;

        •       The aggregate load reduction needed from point sources is often obtained without requiring
               further controls from all of the point sources discharging the impairment pollutant;

        •       The geographic extent of nonpoint sources from which further controls are required is
               typically much less than the entire county(s) surrounding the impaired water;

        •       For both point and nonpoint sources, the degree of load reduction that is required is very
               often less than that which would be achieved if all relevant point and nonpoint  sources
               were to implement "the next treatment step".

        More specifically, in estimating the costs of the "Moderately Cost-effective TMDL Program" and
"More Cost-Effective TMDL Program" scenarios, we drew the following quantitative relationships from
the results of the fifteen TMDLs:

                                               1-6

-------
        •       For point sources. In about half the TMDLs, the aggregate load reduction actually
                required of point sources was roughly equivalent to what would be achieved if all point
                sources contributing to impairment of the water body were to implement "the next
                treatment step". In the other half of the TMDLs, "the next treatment step" for all point
                sources would result in about twice as much aggregate load reduction as was actually
                needed.

        •       For nonpoint sources.  The size of the watershed from which most TMDLs required
                nonpoint source load reductions was far smaller than the size of a typical county. The
                acreage of most nonpoint source TMDL watersheds ranged from about 5 %  to about 40 %
                as large as the acreage of the county(s) within which the impaired water body was located.

These quantitative relationships should be regarded as tentative pending the evaluation of more completed
TMDLs.7
        Note that we are not presuming to use this small sample of 15 cases as a basis for projecting the national
        costs of TMDLs.  We have not estimated implementation costs for sources in each of the 15 cases and
        then extrapolated or scaled up from these cases to the nation as a whole. In fact, early in the history of
        this project to estimate TMDL implementation costs, we considered such an approach.  We decided
        quickly, though, that it would not be possible to select a set of 15, or 50, or 100, or perhaps even 200
        TMDLs that could serve as a representative sample from which to extrapolate to the nation. There is so
        much diversity across TMDLs --  in the size and type of impaired waterbodies, the degree to which they
        are impaired, in the pollutants and source types involved, in the geographic settings, etc. -- that we
        believed any sample of less than several hundred TMDLs would likely misrepresent in some important
        ways the universe of all of them.  We decided then that a "sample and extrapolate" approach for
        estimating national implementation costs was infeasible. A substantial effort would be required to
        perform a mini-TMDL in advance for a single sample impaired water body and then estimate the costs for
        sources to implement this TMDL: determining how far the water body is from attainment, modeling the
        load reduction needed,  identifying relevant sources, developing a load allocation, and estimating how
        much if would cost each source to achieve its load reduction.  We could not afford to do such analysis for
        several hundred sample impaired water bodies.

        Instead, we adopted the approach described here of estimating the implementation costs for all  of the
        impaired  water bodies in the nation by employing a series of simplifying assumptions about what typical
        TMDLs will require of relevant sources.  Under this approach, we use the 15 case studies not as the
        fundamental basis from which to extrapolate, but instead in a more limited way to shed light on the
        reasonableness of our assumptions. The case studies suggest that actual TMDLs only very rarely require
        load reductions as large as those presumed under our Scenario 1 (all relevant point sources implement the
        next treatment step, and all relevant nonpoint sources in the entire county implement the next treatment
        step).  The case studies thus suggest that Scenario 1  really is something like a worst case.  The case
        studies also suggest what some assumptions more typical of most TMDLs might be. We use some rough
        averages drawn from the case studies in defining our more  cost-effective Scenarios 2 and 3.

        This rather lengthy discussion is  intended as a reply to potential reviewers of this draft to the effect that
        the 15 case studies are unrepresentative in one or another way. EPA recognizes that 15 case studies
        cannot be representative of all conditions that may be found in potential TMDLs. For instance, none of
        the 15 is a water body impaired by agricultural chemicals or sediment in a major crop producing area. We
        believe this is acceptable because we use the case studies in a manner such that this sort of
        representativeness is not critical.  We use the case studies primarily to elucidate several much broader

                                                 1-7

-------
C. COSTS THAT ARE NOT ESTIMATED

        This study attempts comprehensively to estimate the costs that pollutant sources will incur to
achieve the load reductions that will likely be required by the eventual TMDLs for waters listed on the
States' 1998 303(d) lists.  Given this objective, we want to be clear that the study explicitly does not
estimate several sorts of costs:

        •       Costs for activities other than abating loads from pollutant sources.  We do not estimate
               the costs to develop TMDLs. These costs,  which will be borne by States and EPA (and
               perhaps also by local governments and other Federal agencies) are estimated in the
               companion report titled "The National Costs to Develop TMDLs".8 Nor do we estimate
               the costs that States, EPA, and other government agencies will incur to implement
               TMDLs.  Thus we do not estimate the costs that State permit authorities will incur in
               reissuing NPDES permits for point sources to incorporate the load reductions required by
               TMDLs, nor the costs that USDA and other agencies might incur in providing information
               and technical assistance to farmers who need to reduce their loads.9  However, to the extent
               that Federal, State or other agencies themselves are the owners of facilities or lands that
               are pollutant sources  (e.g., military bases, Federal forest and range lands), we do estimate
               these costs.

        •       Broader social consequences that might occur as pollutant sources meet TMDL
               requirements. We estimate the costs for sources to reduce their loads to meet TMDL
               requirements, but these costs may have further consequences for society.  We have not
               attempted to describe the social consequences of these actions or to assigne monetary
               values to these changes. For example, higher water and sewer rates as a result of
               increased costs for POTW treatment may increase the number of households facing high
               rates and place greater economic stress on these households.  In addition, increased
               production costs for farmers implementing  agricultural BMPs will likely result in reduced
        questions: Is Scenario 1 really something like a worst case? For a moderately cost-effective Scenario 2,
        how much less should we assume than "the next treatment step will be implemented by all relevant point
        sources within a relevant distance and all relevant nonpoint sources in the entire county"? For questions
        at this greater level of generality, we believe that our sample of 15 is reasonably representative.
        Nevertheless, we agree that the sample to be used for "groundtruthing" our assumptions should be
        expanded, and we will do so for the final version of this report.

        Environomics and Tetra Tech, Inc., National  Costs to Develop TMDLs, prepared for the U.S. EPA, Office
        of Wetlands, Oceans and Watersheds, draft, July 2001.

        We attempt to estimate the costs of the measures and practices that sources will need to implement to meet
        TMDL load reduction requirements, whoever  pays these costs.  In some cases, Federal and State
        governments contribute substantially in paying the costs for sources to implement these measures. For
        example, USDA and other agencies provide: 1) Cost-share funds to assist farmers in implementing BMPs;
        and 2) Technical assistance contributing to the planning and design tasks involved in implementing the
        BMPs.  With respect to agricultural nonpoint  source load reductions, then, we attempt to estimate the total
        costs of planning, designing and implementing the needed farm BMPs, whether the costs are paid for by
        the farmers, by USDA, or by others.

                                                1-8

-------
               agricultural output and/or higher agricultural commodity prices.  We also do not estimate
               the distributional impacts of the costs to pollutant sources.  If pollutant sources are
               particularly concentrated geographically, there may be economic dislocations in the local
               or regional areas surrounding the concentrations.  On the other hand, farming economies in
               other areas not affected by TMDLs will likely see increased activity triggered by reduced
               production in the areas affected by TMDLs and resulting higher commodity prices.  These
               sorts of secondary and ultimate impacts cannot be assessed without broad economic
               modeling of the sectors within which the TMDL-affected pollutant sources operate.  This
               sort of modeling is beyond the scope of this analysis. However, we expect that these
               broader consequences of TMDL costs will not be large,  as TMDL costs represent only a
               small fractional increase in the current costs of the activities the pollutant sources conduct.

        •       Costs for pollutant sources affecting waters that will be found in the future to need
               TMDLs developed for them. This analysis addresses costs relating to currently impaired
               waters on States' 1998 303(d) lists.  More waters may be listed in the future as needing
               TMDLs.  This might be because monitoring at some time in the future finds a currently
               unassessed water to be impaired, or because economic growth or something else causes a
               currently unimpaired water to become impaired. In either case, though, we have no basis
               for projecting where these as-yet-unlisted waters will be found and which pollutant sources
               might need to be addressed because of them.10

        Other sorts of costs are omitted from this analysis not because we define them as outside of our
analytical scope, but because we have been unsuccessful in finding a way to estimate them within the time
and data constraints for this study. The major sorts of costs that we have omitted because of analytical
resource limitations include:

        •       Costs for achieving load reductions from several difficult-to-analyze nonpoint source
               types. We estimate the costs for TMDL-prompted load  reductions from agriculture,
               confined livestock, silviculture and on-site wastewater treatment systems (septic tanks,
               etc.).  Likely important but omitted nonpoint source types include resource extraction
               (mines and oil and gas development),  atmospheric deposition, contaminated in-stream
               sediments, natural sources  (e.g., salt springs, natural mineral deposits) and land disposal
               (both formal and informal sites). We estimate that these omitted source types account for
               about 14 % of all 303(d) river miles and 22% of all 303(d) lake acres.  Some of these
               omitted source types can entail high costs for mitigation (e.g., some instances involving
               dredging and disposing of contaminated in-stream sediments).  On the other hand, if
               impairment from one of these source types cannot be remedied or will involve "widespread
               social and economic impacts", then the water quality standard giving rise to the TMDL
10      On the other side of the coin, some of the currently listed 303(d) waters will eventually achieve standards
        without a TMDL having been established for them. The 1998 303(d) lists represent water quality as of
        perhaps 1997 or so when the lists were compiled.  Since then there has been substantial progress in water
        pollution control (e.g., POTW upgrades, new effluent guidelines, wet weather requirements, voluntary and
        cost-shared nonpoint source programs) as well as shifts in population and economic activity also.  These
        processes will continue, and many of the 1998 303(d) waters will be found to have attained standards
        before their TMDLs are developed. Some of the currently listed waters may also be removed from the list
        for various reasons without a TMDL having been established for them.

                                                1-9

-------
               may be revised through a use attainability analysis. In general, we have not estimated
               costs for these source types because one or the other of the following sorts of information
               is not available: 1) Information at a relatively fine geographic level on the extent of the
               source activity that is occurring and that might need to be controlled in order to resolve
               local water quality problems; and 2) Information on the unit costs of broadly applicable
               measures to abate pollutant loads from the source.

        •       Costs for achieving any needed load reductions from point sources covered by general
               permits.  Our analysis of TMDL-related costs for point source dischargers covers all point
               sources for which individual NPDES permits have been issued. However, in addition to
               the roughly 60,000 active NPDES point source dischargers with individual permits,
               however, there are potentially 385,000 sources to be covered by stormwater general
               permits and approximately 52,000 point sources that are covered by non-stormwater
               NPDES  general permits.11 To the extent that TMDLs will require load reductions below
               currently allowed levels for these general permittees, we do not estimate these costs. We
               assume broadly that TMDLs are unlikely to require stormwater abatement beyond what is
               required  under recent stormwater  and construction regulations. We also believe that non-
               stormwater NPDES general permittees12 are typically much less environmentally
               significant than individual NPDES permittees, and that TMDLs will rarely require further
               controls  of them beyond what is already in their permits. Issuance of general permits has
               been discouraged for any sources  that discharge to impaired waters and may need water
               quality-based effluent limits;13 such sources have preferentially been addressed by
               individual permits even if they otherwise meet criteria for general permits. In sum, we
               believe that TMDLs  will rarely require further load reductions from point sources that are
               currently covered by general permits, and that our omission of these sources from the cost
               analysis  results in only modestly underestimating costs.

D. CROSS-CUTTING ANALYTICAL ISSUES

        A subsequent chapter describes in detail the methodology we use to estimate costs for point
sources, and another chapter addresses nonpoint sources.  This section discusses several general analytical
issues that pertain to the methodologies for both point and nonpoint sources.
11      Figures as of 10/2000. See: http://cfpubl.epa.gov/npdes/permitissuance/statistics.cfm?program_id=l

12      Some sorts of discharges that are often covered by non-stormwater general permits include: non-contact
        cooling water, oil and gas production facilities, pipelines, drinking water treatment facilities, aquaculture,
        mines, and lagoons. Most of these discharges are expected to be infrequent, and/or low volume and/or
        low impact.

13      See U.S. EPA, Office of Water.  "General Permit Program Guidance". February, 1988.
        Www.epa.gov/npdes/pubs/owm0465.pdf. Also, "Water Permitting 101", at
        www.epa.gov/npdes/pubs/101pcpe.htm.

                                               I- 10

-------
1. The baseline for the analysis

        We estimate the costs that TMDLs will engender for point and nonpoint sources in order to meet
water quality standards for the set of impaired waters included in the States' 1998 303(d) lists.  For these
impaired waters, we attribute to the TMDL program the costs of the additional controls that pollutant
sources will need to implement beyond a baseline that includes: 1) Whatever controls were in place at point
and nonpoint sources as of when the 1998 303(d) lists were developed; and 2) Assumed compliance with
all applicable technology-based requirements.  Viewed in another way, the analysis estimates the
incremental costs to pollutant sources of achieving water quality standards relative to a baseline of their
controls in place in 1998, but excluding the costs of whatever amount of this further progress will be
achieved through meeting technology-based requirements that were unmet as of 1998.  Several aspects of
how we define the baseline and costs for pollutant sources deserve more explanation:

        •      As of when the 1998 lists were developed, many pollutant sources had implemented
               control measures or BMPs beyond those required by technology-based standards.  Many
               point sources had implemented advanced treatment measures as required by water quality-
               based effluent limits (WQBELs) in their NPDES permits. Many nonpoint sources had
               implemented BMPs voluntarily or because of incentive programs or State requirements.
               Because of this progress beyond technology-based standards, our analysis thus does  not
               estimate the costs of achieving water quality standards over and above the costs of meeting
               technology-based standards. A  substantial amount of the progress beyond meeting
               technology-based standards that will be needed in order to attain water quality standards
               had already occurred by the time the 1998 lists were developed.

        •      By the same token, some of the progress needed to meet water quality  standards for the
               1998 303(d)-listed waters has already occurred since they were listed.  Since 1998 or so,
               many more point sources have installed advanced treatment measures as required by
               WQBELs,14 and many more nonpoint sources have implemented desirable BMPs. Some
               of the costs that we estimate in this report will need to be spent to meet water quality
               standards have already been spent. We estimate costs to meet water quality standards
               relative to a circa 1998 baseline. These costs that we estimate are therefore greater than
               would be necessary if we were to measure them relative to a current 2001 baseline.

        •      We do not count as costs of the  TMDL program those costs to dischargers that have yet to
               be incurred to meet existing technology-based standards.  Some of progress needed to meet
               water quality standards will come as sources meet as-yet-unmet technology-based
               requirements, most notably the requirements pertaining to storm water, CSOs and SSOs.
               The costs of meeting these technology-based standards will be substantial,  and for many
               waters these additional control efforts will be critical to attaining water quality standards.
14      EPA does not have full information on the fraction of all NPDES permittees who have permit limits
        reflecting only the minimum technology-based requirements and the fraction that have more stringent
        water quality-based limits. A recent analysis for POTWs specifically suggests that perhaps 68 % of major
        POTWs and 59 % of minor POTWs currently appear to have WQBELs. No data is available on industrial
        point sources.  It is clear, though, that extensive water quality-based permitting has already occurred, and
        the nation has a substantial head start on the task of achieving water quality standards in impaired waters.

                                              I- 11

-------
               However, technology-based requirements and their associated costs are pursuant to
               sections of the CWA other than §303 (TMDLs). Dischargers are and will be required to
               meet these technology-based standards regardless of whether a TMDL is established or
               not.  Consequently, we do not include these baseline costs in our study.

2. Important data bases

        For this analysis, we use data that is available at the national level on impaired waters and on the
point and nonpoint sources that may affect them.

        The set of 303(d) waters for which TMDLs must be developed is revised frequently as States
submit their new lists, EPA reviews them, and changes are made. Extensive information on the listed
waters (e.g., causes, pollutants, lengths, use impairments) is compiled and entered into a 303(d) data base
that is updated as the lists and associated information change. For this analysis, we used information in the
303(d) data base as of the spring of 2000.15  At that time, the 303(d) data base included some version of the
1998 303(d) list for every State; for most  States the data base included their fully approved 1998 303(d)
lists. As of then, the 303(d) data base included 21,845 waters and 41,318 listed causes of impairment for
these waters.

        For point source dischargers, the key data base that we used is the Permit Compliance System
(PCS).  PCS includes  information on every point source holding a NPDES permit, including data that we
used on each discharger's location, SIC code, flow, and monitoring results. Similarly as for the 303(d)
data base, we "froze"  the information from PCS as of mid-summer, 2000. At that time, there were
approximately 63,000 point source dischargers listed in PCS for the U.S. and territories, of which 58,977
were active. For POTWs, we also used the 1996  Clean Water Needs Survey to obtain more detailed
information on flow, industrial flow, and treatment equipment in place.

        The key nonpoint source data bases we used for this analysis included:

        •      For agricultural land (crop, pasture and range), the 1997 National Resources Inventory
               (NRI) (USDA) as it existed prior to very recently released corrections;

        •      For silviculture, the 1996 Timber Product Output Data File (U.S. Forest Service);

        •      For AFOs, the 1997 Census of Agriculture (USDA); and

        •      For on-site wastewater treatment systems, the 1992 Census of Housing (USDOC).

Each of these nonpoint source data bases are the most recent versions available.
15      Since the spring of 2000, the 303(d) data base has been updated several times to reflect new changes to
        States' lists, revised procedures for estimating the length of some impaired water bodies, and other
        developments.  Because we "froze" the data base as of the spring of 2000 for this analysis, information on
        impaired waters that is now available at the Agency reflects newer numbers and is slightly different from
        what is portrayed in this analysis. As of XXXX, 2001, the 303(d) data base includes XXXX waters and
        XXXX causes, a small change since last year.

                                              I- 12

-------
3. Scaling to adjust for limited geographic coverage

        The data used to estimate the implementation costs for point sources and for nonpoint sources do
not completely cover the United States.  Various sorts of desired data are unavailable for some States and
for  some sources. In general, we estimate implementation costs for the subset of States and sources for
which we have data, and then extrapolate the results to the remainder of the nation using a scaling factor
reflecting the portion of the nation for which we do have data.  We assume that the portion of the nation for
which we do not have data has impairments and will incur compliance costs at the same rate as the portion
of the nation for which we do have data. The scaling procedures we employ are different for point sources
and for nonpoint sources.

        a. Scaling in the point source analysis

        The point source analysis in effect depends on comparing the locations of georeferenced 303(d)
waters with the locations of georeferenced point sources.  We begin with an impaired water that we have
located geographically. We then search within that water body and upstream an appropriate distance along
a comprehensive network of linked water reaches to identify any point source dischargers that are
discharging the relevant impairment pollutant. We then simulate controls for these identified point sources
as necessary to address the point source contribution to impairment of the water body.  We repeat this
process  for all impaired waters. There are three reasons why, as we implement this analysis, we ultimately
do not cover all impaired waters and all potentially contributing point sources:

        •      Some 303 (d) waters have not been georeferenced by locating them with respect to Reach
               File 3 (RF3) reaches.  An impaired water that has not been georeferenced in this manner is
               not covered in our analysis because we cannot trace upstream from it to find the point
               sources that may be affecting it;

        •      Some point sources have not been georeferenced. A point source that has not been
               georeferenced is not covered in our analysis because it cannot be "found" via this process
               and cannot have any costs estimated for it;

        •      The procedure for matching water bodies against point sources is  not operable for a
               variety of reasons in several  States.  These entire States  are not covered by the analysis.

        We develop scaling factors to account for these shortfalls in analytical coverage. We estimate
point source costs for the portion of the universe that we do cover, assume that this portion is a sample
representative of the entire country, and then  extrapolate the costs estimated for our sample to the entire
country by using an appropriate scaling factor.

        The matching procedure is not currently operable for the Pacific Northwest States (AK, ID, OR,
WA), Massachusetts, Hawaii and the Territories. The matching procedure thus covers 44 States plus the
District of Columbia.  In these 44 States plus DC, there are 41,316 active point sources that have been
georeferenced.  In the nation as a whole plus territories, there are 58,977 active point sources. The scaling
factor to account for incomplete coverage involving point source georeferencing and the matching
procedure is thus 58,977/41,316, or 1.427.
                                              I- 13

-------
        In the 44 States plus DC, there are 18,162 303(d)-listed waters, of which 16,143 have been
georeferenced.  The scaling factor to account for incomplete georeferencing of impaired waters is thus
18,162/16,143, or  1.125.  Although we do apply this scaling factor, it is very likely an overestimate.
Adding more georeferenced waters to the analysis would likely lead to a less than proportional increase in
compliance costs, since many of the point sources implicated by the added waters would have already been
included in the analysis and would have already incurred compliance costs as a result of previously
georeferenced waters.  (We assume that a point source that already must incur a compliance cost because it
affects an impaired water will not incur increased costs if it is  found also to affect additional waters.)

        Combining the two point source scaling factors (1.427 x 1.125) yields a combined scaling factor of
1.605. The aggregate point source costs we estimate in our analysis are multiplied by 1.605 in order to
estimate the total national costs for point sources.  Again, we believe this is likely an overestimate.

        b. Scaling  in the nonpoint source analysis

        Scaling issues  are different for the nonpoint source cost analysis than for the point source analysis.
For the nonpoint source analysis, georeferencing of impaired waters is essentially complete, using
procedures that do not depend on RF3. All impaired waters have been located and the counties through
which they pass have been identified.  The coverage of our analysis is incomplete, though, because
identification of the nonpoint sources associated with these impaired waters is incomplete.

        States vary in the degree to which they provide information on the source types responsible for
impairment of impaired waters. Some States do not report source information at all, some report in modest
detail (e.g., using only broad identifiers of source types such as "point sources", "nonpoint sources",
"unknown sources", etc.),  and some report in great detail (e.g., using specific identifiers  of source types
such as "municipal point sources", "CSOs", "silviculture", "irrigated crop land", etc.).  Our procedure for
identifying water bodies impaired by various nonpoint  source categories relies on detailed source reporting
by those States that provide such information, either in their 303(d) submissions or in their 305(b)
submissions.  For example, in analyzing silviculture, we:

        •       Identify the 303(d) water bodies that States report as impaired by silviculture as a source.

        •       Identify the 305 (b) water bodies that States report as impaired by silviculture as a source.
               We then crosswalk from each of these silviculture-impaired 305 (b) waters and determine
               whether there is a corresponding 303(d) water body. (Some States report 305(b) source
               information but not 303(d). This second step effectively increases the set of States within
               which we can find silviculture-impaired 303(d) waters.)

        •       Add the results of the first and second steps, thus obtaining  a list of silviculture-impaired
               303(d) waters. We note the States in which these waters are located. We then assume this
               set of States to be the sample covered by our analysis, and we assume that this sample is
               representative of the nation as  a whole.

        •       States for which we can identify no silviculture-impaired water bodies may either: a)
               Actually have no silviculture-impaired water bodies; or b) Actually have them, but report
               source information in a manner that does not allow for identifying them. Conservatively,
               we assume the latter — we assume that a State that reports no silviculture-impaired water

                                               I- 14

-------
               bodies is effectively "non-reporting". We extrapolate the costs we estimate for controlling
               silviculture in the sample States (those that report silviculture impairments) to the assumed
               non-reporting States by scaling up based on the volume of silviculture occurring in the
               sample States relative to that occurring in the non-reporting States.

        We employ this approach for each of the different nonpoint source types we analyze.  Different
sets of States are considered to be "non-reporting" for the different nonpoint source types.  For each
nonpoint source type, we use a different volume-based scaling factor to extrapolate our cost estimates from
the "reporting" to the "non-reporting" States.  The volume measures that we use in scaling are as follows:

        •       For silviculture.  The annual volume of timber harvested.

        •       For agriculture.  Crop land-related costs are extrapolated based on the acreage of crop
               land. Similarly, pasture-related costs and range-related costs are extrapolated based on the
               acreage of pasture and range lands.

        •       For AFOs. The number of confined animal units (AUs).

        •       For on-site wastewater treatment systems.  The number of dwelling units served by septic
               systems.

        The specific scaling factors we develop (e.g., the ratio between the total national annual timber
harvest and the harvest volume in our "reporting" States) are described in the report sections providing cost
estimates for each of the nonpoint source categories.

        We also apply another sort of scaling factor in the nonpoint source cost analysis.  In our nonpoint
source analysis, we attempt to cover those States that report source information and scale to those  that do
not.  Among the States that report source information, though, we count a water body as impaired  by a
specific nonpoint source type only if the State affirmatively cites that specific nonpoint  source type as a
source. A State that reports source information sometimes reports the source as "nonpoint source  (not
classified)" or as "unknown source".  In some fraction of these cases, we would guess that silviculture or
AFOs or agricultural land or some other specific  nonpoint source type we are interested in will eventually
be found to be one of the responsible sources. Perhaps we should regard as a sample the instances in which
a specific source type is reported, and we  should extrapolate from that sample to the instances in which a
non-specific source type is reported (i.e., NFS not classified, or unknown source).  The following exhibit
provides information that can be used in developing a scaling factor reflecting this approach:
                                               I- 15

-------
                                            Exhibit 1-1
                       Information Used for Nonpoint Source Scaling Factors


Unknown source
NPS (not classified)
All other sources
303(d) information
# rivers listed
8.1 %
12.5%
79.4 %
# lakes listed
5.4 %
8.6 %
86.0 %
River miles
5.7 %
14.7%
79.6 %
Lake acres
6.8 %
8.8 %
84.4 %
305(b) information
River miles
1.9%
98.1 %
Lake acres
N.A.
N.A.
N.A.
        The nonpoint source categories we are interested in can be cited by a State only when the State
cites a specific source. Specific sources are cited in 79.4 - 98.1 % of the instances. The scale factor to
reflect extrapolation to instances where a State does not cite a specific source can range from 1.02
(100/98.1) to 1.26 (100/79.4).  About half of the waters identified in our analysis as impaired by specific
nonpoint source types derive from 303(d) and about half derive from 305(b) listings.  We thus adopt a
scale factor of 1.13, averaging between a typical scale factor of approximately 1.25 for 303(d) listings and
1.02 for 305(b) listings.

4. Cost estimating conventions

        We estimate costs for pollutant sources in first quarter 2000 dollars. All costs include capital and
operating and maintenance costs,  combined into a single annualized cost figure that is assumed to continue
forever.  We estimate the time at which each pollutant source will begin to incur this annualized cost
stream, sum the costs across pollutant sources, and then discount the costs incurred in future years back to
a present value cost in 2000. We then annualize this present value cost. The cost estimates that we present
thus represent levelized annual amounts that will continue each year, forever.16 A real discount rate of 7 %
per year is used. The assumed useful life of capital investments varies with the nature of the investment,
ranging from 3 years for some nonpoint source management measures (e.g., a nutrient management plan) to
20 years for capital equipment at  POTWs.

5. Assumptions regarding the time when costs will be incurred

        A pollutant source will presumably not incur costs resulting from a TMDL until sometime after the
TMDL affecting the pollutant source has been developed.  Given that the July 2000 TMDL regulations do
not require all TMDLs for the 1998 listed waters be developed until 2015, there may be many years until a
pollutant source needs to incur the costs of achieving any reduced load assigned by a TMDL.  To the extent
16      The cost estimates are expressed in this manner so that they can be compared with cost estimates for other
        regulations, policies, programs or initiatives that are also expressed as levelized, continuing annual
        amounts. The actual costs of any program typically fluctuate over time, with varying amounts of capital
        and operating/maintenance costs incurred in various future years. These costs may continue forever into
        the future, or they may end at some point. In summarizing the costs of any such program, analysts
        typically convert the fluctuating stream of actual cost payments into a single levelized annual amount by:
        1) Discounting all the actual costs in different future years back to a discounted present value in the base
        year; and 2) Annualizing this discounted present value, converting it into an equivalent stream of equal
        annual payments that continue forever.
                                               I- 16

-------
that costs are incurred far in the future, the present value of these costs is reduced.  The costs that pollutant
sources will incur as a result of the TMDL program (or alternatives to it) depend on when these costs will
be incurred.

       The pace at which TMDLs will be developed for the waters on the 1998 303(d) list is unknown.  A
companion report to this one -- the TMDL Development Cost Report17 -- discusses various possibilities
regarding the pace of TMDL development. For this analysis, we have chosen an "even pace" projection
from among those considered in the Development Cost Report.

                                            Exhibit 1-2
                   Projected Pace of TMDL Development Used for This Analysis
Year of
Completion
2000 and
before
2001
2002
2003
2004
2005 - 2014
2015
Total
# TMDLs
Completed
2,000
2,282
2,282
2,282
2,282
2,282 each year
2,282
36,225
The "even pace" projection reflects the number of TMDLs for 1998 303(d) waters that have already been
developed (an estimated 2,000 developed before 2001) and then assumes that the remaining TMDLs are
developed at an even pace through the deadline at the end of 2015.  Note that the estimated total number of
TMDLs to be developed is only 36,225, relative to a projected number of causes for the currently listed
303(d) waters totaling 41,318.  Roughly 5,000 of the causes (about 12 % of the total) associated with
currently listed waters involve "pollution" rather than pollutants (e.g., flow alteration, habitat alteration)
and TMDLs are not expected to be prepared for such causes. The costs we estimate for pollutant sources
will increase somewhat to the extent that further research ultimately identifies pollutants requiring TMDLs
for some of these "pollution"-impaired waters.

       For this  cost analysis, we assume that a point or nonpoint source that will need to implement
further control measures as a result of a TMDL will begin to incur the costs of doing so an average of five
years after the TMDL is developed. The capital  costs for the necessary control measures will be incurred
five years after the relevant TMDL is developed, and annual O&M costs will also begin at that point.
Thus, the schedule upon which sources will begin to incur TMDL compliance costs will be lagged five
years relative to the TMDL development schedule shown above.
17     Environomics and Tetra Tech, Inc., National Costs to Develop TMDLs, prepared for the U.S. EPA, Office
       of Wetlands, Oceans and Watersheds, draft, May 2001.
                                              I- 17

-------
       The cost models we use to estimate compliance costs for sources generate estimates of annualized
costs as of the year when compliance costs begin.  In effect, the numbers generated by the cost models are
estimates as if all the sources would need to begin incurring all their TMDL compliance costs immediately,
in the year 2000. To reflect the pace of TMDL development as shown above and the assumed 5-year time
lag between TMDL development and when compliance costs will begin, we must scale down the estimates
from the cost models. A spreadsheet in Appendix B shows the derivation of a scaling factor that reflects
the particular pace of TMDL development and compliance time lag that we have assumed. At a discount
rate of 7 %, the scale factor to convert the annualized cost once all sources are incurring their costs to an
equivalent annualized cost beginning in the year 2000s is 0.4484.

       This pace/lag scale factor is applied to the outputs of the point source and nonpoint source cost
models. The raw outputs of the cost models show total annualized costs for sources after the point in time
when all sources have begun incurring their TMDL compliance costs.  This will not be until 2020
(assuming the last TMDLs for the  1998 listed waters are developed in 2015 and compliance costs lag
TMDL development by another five years). Between now and 2020, annual costs to sources will slowly
rise until they reach the maximum in 2020, and they will then continue at this maximum level forever.
Many pollutant sources will not need to implement their projected control and management measures until
many years in the future. This  gradual development and implementation of TMDLs for currently  listed
waters will reduce the present value of the management and control costs for the TMDL program well
below what it would be if all TMDLs were completed immediately and sources  needed to comply
immediately. The impact of stretching out the compliance expenditures is captured in the "pace/lag" scale
factor.  If TMDL development were accelerated relative to the pace shown in the exhibit above, or if the
compliance time lag were assumed to be less than five years, the scale factor would increase and the
estimated present value TMDL implementation costs would increase.

       We assume that the timing of compliance investments by sources will be the same under Scenario 1
and Scenario 3 as what we have projected here for the Moderately Cost-effective TMDL Program
(Scenario 2). This is so that the estimated cost differences between the scenarios result only from
substantive differences between them rather than differences involving timing.

6. Assumptions regarding changing conditions  over time

       We estimate the costs for sources to achieve water quality standards based on current conditions.
For example, for a POTW  that will need to upgrade its treatment to meet the requirements of a TMDL, we
estimate its costs based on the POTW's current level of flow or population served and current treatment in
place.  For a county in which silvicultural activities will need to adopt improved management practices to
meet TMDL requirements,  we estimate costs based on the current volume of silvicultural activity in the
county. In reality, though,  if the TMDL affecting these two example sources will not be developed for
another decade or more, when it comes  time to comply with the TMDL's allocations, the POTW may have
a larger or smaller flow, and the volume of silvicultural activity taking place in the county and needing
control may have changed also. Or, conditions regarding other sources may have changed in a manner
affecting what these sources need to do - perhaps, for example, by the time the TMDL is eventually
developed, a large, poorly controlled industrial discharger affecting the same water body will have  gone out
of business, and the POTW and the silvicultural operations will not need to implement further controls.
We do not attempt to predict these  changes in source activities  over time that may affect what the sources
need to do in order to meet  the requirements of the eventual TMDLs.  We estimate costs for sources to
meet the likely requirements of TMDLs given current conditions.

                                              I-  18

-------
        We also do not evaluate the potential impact of another set of changing conditions over time --
changes in the list of waters for which TMDLs will be necessary.  As States assess more of their waters
and as source activities increase with a growing population and economy, additional water bodies will
likely be listed for which TMDLs will need to be developed. Also, though, as a wide variety of incentive
programs (e.g., agricultural cost share programs) and new regulations (e.g., State requirements for nutrient
management planning, new Federal effluent guidelines, and new rules for SSOs) continue to be developed
and implemented and sources respond, some currently impaired waters that need TMDLs will attain
standards and ultimately not need TMDLs. Again, in this analysis we estimate costs for the current set of
listed waters, assuming current conditions. This approach differs somewhat from the accompanying
TMDL development cost analysis, which estimates costs also for future waters that might be added to the
present 303(d) lists.18
18      This analysis diverges in several additional respects from the TMDL Development Cost Analysis. Most
        notably, costs are presented in this report as levelized annual amounts beginning now and continuing each
        year, forever. Whereas compliance costs continue indefinitely (we assume that a source's compliance
        obligation is a continuing one), the cost of developing a TMDL is a one-time cost - once a TMDL is
        developed, there is no further development cost.  Consistent with the fact that TMDL development will
        end at some point,  the Development Cost Analysis thus shows costs in several ways that are different from
        how costs are shown here: showing them as undiscounted costs in the years in which they will occur, and
        also as total undiscounted costs over this time period.

        The Development Cost Analysis also does not count the costs of developing the TMDLs for water bodies
        on the 1998 303(d) lists that have already been developed.  In this analysis, though, we include the costs
        for pollutant sources to meet the requirements of these TMDLs because the bulk of these costs have
        presumably not yet been incurred.

                                                I- 19

-------
              II.  IMPLEMENTATION COSTS FOR POINT SOURCES

A. OVERVIEW

       We identified the specific point source facilities that might discharge the pollutants responsible for
impairments in 303(d) listed waters.  For Scenario 1 ("Least Flexible TMDL Program"), we assumed that
all of the facilities that might reasonably be expected to control these discharges further will be required to
install the "next treatment step", and estimated the costs associated with those controls. For Scenario 2
("Moderately Cost-effective TMDL Program") we scaled down the costs estimated for Scenario 1 to reflect
the degree to which actual TMDLs have required aggregate load reductions from point sources less than
that which would be achieved if all point sources implemented the next treatment step. For Scenario 3
("Cost-Effective TMDL Program"),  we adjusted the costs estimated for Scenario 2 to reflect the potential
savings to point sources from more cost-effective waste load allocations or trading. We will describe this
analysis in detail as consisting of eight steps:

        1.      Select the pollutants that point sources are most likely to be required to control further -
               these included BOD, nutrients, toxic organics, ammonia and metals.

       2.      Identify the facilities that are within a relevant distance upstream of an impairment for any
               of these pollutants.

       3.      Determine which of these upstream facilities will likely be required to implement further
               controls for each impairment pollutant under a water quality-based permitting approach
               (Scenario 1).

       4.      Develop cost functions for the "next treatment step" for each relevant pollutant

       5.      Apply the cost functions for every pollutant source needing further control for an
               impairment pollutant.

       6.      Adjust the estimates to exclude costs for point sources affecting waters that States identify
               as impaired by nonpoint sources only.  At this point, costs for point sources under
               Scenario 1  are estimated.

       7.      Adjust the Scenario  1 cost estimates to reflect the results of a sample of actual TMDLs
               and estimate the costs for Scenario 2.

       8.      Adjust the Scenario  2 cost estimates to reflect opportunities for additional cost-effective
               waste load  allocations or trading in which lower-cost control of nonpoint sources might
               substitute for some further point source controls.

This chapter summarizes each of these steps and their results. Appendices C through H provide further
details.
                                              II- 1

-------
B. POLLUTANTS FOR ANALYSIS

        In order to limit the analytical workload (e.g., developing and applying cost functions for the next
treatment step for each pollutant), we chose to analyze only those particular pollutants for which TMDLs
or water quality-based permits are most likely to require further controls (beyond technology-based
standards) for point sources. We obtained data on the frequency with which different causes of impairment
are cited for waters impaired by point sources.  We considered the most frequent causes of impairment, and
asked whether each of these causes would commonly trigger the need for further controls of point sources
that are already meeting technology-based standards. We concluded that the following pollutant classes are
the most likely to prompt requirements for further point source controls: BOD, nutrients, toxic organic
chemicals, metals,  and ammonia. Some of our judgments in settling on this list of pollutants to trigger
point source controls were:

        •       We chose not to analyze causes of impairment when no pollutant was identified.
               Examples included flow or habitat alteration or fish consumption advisory when no
               pollutant was listed also.

        •       We chose not to analyze pollutants that are extremely unlikely to be discharged in
               sufficient quantity to be problematic by a point source meeting technology-based
               standards.  Examples included temperature or pH.

        •       We chose not to analyze pollutants for which only a very small fraction of point sources
               might require additional beyond-technology-based controls for process water  discharges.
               Examples include sediment or pathogens. We believe that the great majority  of instances
               where waters are impaired for these causes and point sources  are cited as a source involve
               wet weather discharges from point sources. We believe these problems will largely be
               remedied when existing technology-based standards for wet weather discharges — storm
               water, construction, CSO and SSO requirements - are complied with.

        •       We chose not to analyze in this study pollutants representing a very small fraction of
               causes that would require specialized treatment technologies as the"next step". Examples
               include chlorine and cyanide.  We could consider such infrequent causes of impairment in
               a future study.

Appendix C provides supporting data and further information on our judgments in arriving at the list of
five pollutants that we will consider as triggering point source controls under TMDLs or water quality-
based permits.

        The next step in the point source cost analysis is to identify the facilities that discharge any of these
five pollutant classes - BOD, nutrients, toxic organics, metals, or ammonia - in a manner so as to affect a
303(d) water that is impaired for one of these pollutants.

C. NEARBY FACILITIES

        These are the point sources that we consider potentially to contribute to impairment and perhaps to
be affected by TMDLs or water quality-based permits. We adopted two alternative scenarios for
identifying point sources and considering them for further controls:

                                               II-2

-------
        •       Case 1, "within and upstream".  We assume that a point source contributes to impairment
               if it discharges the pollutant of concern: a) Directly into the waterbody; or b) Within 25
               miles upstream of a water body impaired by BOD, ammonia or toxic organic chemicals, or
               within 50 miles upstream of a water body impaired by nutrients or metals.

        •       Case 2, "within only". We assume that a point source contributes to impairment if it
               discharges the pollutant of concern directly into the impaired water body.

        For each water body impaired by one of the five causes we selected as potentially triggering point
source controls (BOD, ammonia, toxic organic chemicals, nutrients,  metals), we identified all point sources
with outfalls discharging into the impaired segment or within the 25 or 50 mile distance upstream of the
upstream end of the impaired segment. This step involved a complex procedure to interlink three very large
electronic data bases: 1) Information on the location and causes of impairment for most of the more than
21,000 currently listed 303(d) water bodies; 2) Information on the location of the receiving water and other
data for the nearly 60,000 active point source dischargers; 3) A nationwide network model that links water
reaches to each other, allowing one to trace upstream any specified distance from a water body to upstream
reaches, tributaries, headwaters, etc.. This analysis was performed by Tetra Tech,  Inc.

        The two cases for analysis reflect different possible judgments about how far downstream the
pollutants  discharged by point sources will typically have an impact. Metals and nutrients -- conservative
pollutants  that do not degrade over time or distance -- have the longest range downstream effect.
Ammonia, BOD, and many toxic organics have a much shorter range downstream effect, strongly affecting
receiving waters shortly below the point of discharge and weakly affecting waters farther downstream.
Tetra Tech's engineers judged that 50 and 25 miles were reasonable  upper limits to the upstream distance
within which point sources discharging these pollutants would likely  have a substantial impact warranting
attention in TMDLs or water quality-based permitting.

        In the TMDL "groundtruthing" analysis reported in Appendix A, we have attempted to determine
whether actual TMDLs have typically focused only on point sources discharging to the impaired water
body or whether they have typically addressed upstream point sources also. And, if upstream dischargers
were considered, how far  upstream? The results from our sample of 15 TMDLs are not conclusive.
Sometimes only point sources discharging directly into the impaired water are considered, sometimes the
TMDL addresses upstream point sources also.

        In the absence of more extensive information,  we believe the "within and upstream" case (and the
25 and 50 mile assumed distances) is generally conservative. For most impaired water bodies, these
upstream distances will pull in more point sources than would likely be considered as contributing
importantly to impairment. However, for some large and significantly impaired water bodies, basin-wide
load reductions from point sources may be necessary that will extend well beyond these distances. On the
whole, though, we expect the total national costs for point sources that actually result from TMDLs and
water quality-based permitting will be somewhere between the  "within and upstream" case and the
"within only" case.

        It  should be noted that the number of point sources potentially affected by TMDLs or water
quality-based permitting increases much less than linearly with an increase in these assumed distances.
Many impaired waters are in headwaters, and 25 or 50 miles already reaches a stream's origin. Also,
                                              II-3

-------
many impaired waters are near other impaired waters and looking further upstream simply identifies point
sources already identified as affecting other impaired waters.

D. FACILITIES THAT NEED TO REDUCE LOADS

        The next step involved making a judgment about whether each point source discharging into or
within a 25/50 mile distance upstream of a water body that is impaired for one of these pollutants might be
required by the eventual TMDL or water quality-based permit to implement further controls. We sought to
determine whether an identified point source meeting applicable technology-based standards was likely to
discharge the impairment pollutant in sufficient quantity to warrant consideration for further controls.  This
judgment was particularly difficult to make because there is little data available at the national level on the
pollutants that individual facilities discharge and their amounts. For the majority of dischargers, the Permit
Compliance System (PCS) - EPA's major data base on point source dischargers that we used to identify
potentially relevant facilities - provides information only on which pollutants a facility is required to
monitor for. For only a very few facilities does PCS provide reliable information on the amount of the
monitored pollutants that are discharged.  PCS provides no information on whether a facility discharges
unmonitored pollutants or their amounts.

        We tested three alternative approaches for determining whether or not a specific point source (after
meeting applicable technology-based requirements) discharges the impairment pollutant in an amount
making the source likely to be addressed in the TMDL or water quality-based permit:

        •       The first approach was to use the information in PCS on whether monitoring is required
                for a pollutant as an indicator of whether or not a facility discharges the pollutant in a
                meaningful amount. We  applied this approach for each 4-digit SIC code - if at least 15 %
                of all the facilities in PCS in a particular 4-digit SIC code were required to monitor for a
                pollutant, then all the facilities in that SIC code were deemed to discharge the pollutant in a
                meaningful amount. If the facility was thus judged to discharge the impairment pollutant,
                we then assumed conservatively that the facility would be identified as contributing to
                impairment and would be addressed in the TMDL or water quality-based permitting
                process.

        •       The other approaches relied instead on an engineering judgment as to whether each 4-digit
                SIC is  likely, after meeting BPT/BAT/secondary treatment requirements, to discharge each
                pollutant at levels that could warrant potential further control.  Two EPA engineers
                experienced in industrial water pollution control made these judgments for each of more
                than 500 different SIC codes. We then implemented the engineering judgments in two
                alternative ways, resulting in high and low estimates for the number of point sources likely
                to be deemed as discharging  the impairment pollutant in sufficient quantity as to be
                contributing to impairment.  The high estimate adopts more liberal rules for matching
                sources and impairments  than does the low estimate. For the high estimate, when a water
                body is impaired by a specific metal (except mercury) or a specific toxic  organic (except
                PCBs or dioxin) and the facility (based on engineering judgment for its SIC) is expected
                generally to discharge metals or toxic organics, then the facility is  assumed to be a
                candidate for mandated further controls. For the low estimate, the match must be exact in
                order for further controls to be considered for the facility: if a water body is impaired for
                metals generally then facilities that discharge metals generally are  assumed to warrant

                                               II-4

-------
               consideration, but if a water body is impaired for a specific metal or toxic organic (e.g.,
               zinc, phenol), only those facilities discharging that specific metal or toxic organic are
               assumed to warrant consideration.

       We ultimately selected the engineering judgment approach using liberal matching rules.  The three
approaches yielded broadly similar results (within + or - 20 %) in terms of: a) the numbers of point sources
presumed likely to discharge an impairment pollutant in a quantity potentially warranting further control
under a TMDL or water quality-based permit; and b) the costs  of these controls. The engineering judgment
approach, however, often yielded more sensible results regarding individual SIC codes than did the
approach based on monitoring requirements. The monitoring requirement-based approach, for example,
projected that metal finishers contribute to impairments for nutrients, and "fabricated metal products"
dischargers do not contribute to impairments for metals.  The engineers, on the other hand, judged that
metal finishers meeting BPT/BAT requirements would typically not discharge significant quantities of
nutrients and that TMDLs  or water quality-based permits would be unlikely to address them when nutrients
were at issue.  The engineers believed the reverse was  likely true for "fabricated metal products" facilities.
We chose the approach using liberal matching rules in order to reduce the likelihood that we exclude from
our cost analysis some point sources that ultimately will end up being addressed by TMDLs or water
quality-based permits.

       A fuller description of the three alternate approaches and the results obtained under each is
provided in Appendix D.

       We made two further decisions that limited the set of point sources presumed likely to incur
additional costs as a result of TMDLs or water quality-based permits:

        1.      We assumed that POTWs that currently provide better-than-secondary treatment would
               not be required by TMDLs or water quality-based permits to further improve their
               treatment for BOD, nutrients, toxic organics and/or ammonia.

       2.      As a definitional matter, we decided to attribute any POTW treatment upgrade projects
               that were listed in the 1996 Clean Water Needs Survey to Clean Water Act requirements
               other than TMDLs. In effect, we believe that POTW upgrades that were far enough along
               to be included in the 1996 CWNS should not be attributed to the TMDL program and
               should instead be counted as part of the pre-TMDL baseline.

       These limitations on the set of POTWs that we consider likely to incur incremental costs
attributable to the TMDL program have an important  influence on the TMDL implementation costs that we
estimate. The following exhibit shows how the estimated implementation costs for POTWs would change
if one or  both of the limitations were not adopted.  Costs are shown for the "within and upstream" case.
                                              II-5

-------
                                          Exhibit II-l

    Incremental Implementation Costs for POTWs - Scenario 1 (Least Flexible TMDL Program)
                                   ("within+upstream" case)
Basis for Estimate
Preferred estimate: 1) assume POTWs that are already beyond 2°
will incur no costs, and 2) attribute no TMDL costs to projects
approved in 1 996 CWNS
Drop the 1st limitation: assume TMDLs will require "next treatment
step" of all POTWs, whether or not they already provide advanced
treatment
Drop the 2nd limitation: attribute costs of "next treatment step" for
POTWs not already beyond 2° to the TMDL program, whether or
not the upgrade was planned long ago
Drop both limitations
# of POTWs
Incurring Costs
3335
5262
3694
5622
Annualized Costs for
POTWs (2000 $ in
millions/yr)
697
1,869
836
2,009
   Incremental Implementation Costs for POTWs - Scenario 2 (Moderately Cost-effective TMDL
                                           Program)
                                   ("within+upstream" case)
Basis for Estimate
Preferred estimate: 1) assume POTWs that are already beyond 2°
will incur no costs, and 2) attribute no TMDL costs to projects
approved in 1996 CWNS
Drop the 1st limitation: assume TMDLs will require "next treatment
step" of all POTWs, whether or not they already provide advanced
treatment
Drop the 2nd limitation: attribute costs of "next treatment step" for
POTWs not already beyond 2° to the TMDL program, whether or
not the upgrade was planned long ago
Drop both limitations
# of POTWs
Incurring Costs
2502
3947
2770
4216
Annualized Costs for
POTWs (2000 $ in
millions/yr)
523
1,402
627
1,506
       Abandoning these two limitations would increase our implementation cost estimates for POTWs
sharply, by a factor of nearly three (from $697 million annually to $2,009 million annually for Scenario 1,
and from $523 million annually to $1,506 million annually for Scenario 2). The assumption that TMDLs
will not require further controls of POTWs that already provide better-than-secondary treatment is
particularly important, resulting in nearly $1.2 billion in estimated cost reductions for Scenario 1 and
nearly $900 million for Scenario 2. In essence, we project that there are many large POTWs that already
employ advanced treatment that nevertheless still appear to discharge meaningful amounts of an
                                             II-6

-------
impairment pollutant into or somewhat upstream of impaired waters.19 Are the eventual TMDLs likely to
require further controls of these already well-controlled POTWs?

        The sample of 15 TMDLs that we reviewed for "ground-truthing" purposes provide some
information relevant to this question. Across the 15 TMDLs, in some cases POTWs discharging the
impairment pollutant that now provide better than secondary treatment are being required to reduce their
loads further, and in some cases they are not. When they are not required to reduce loads further,
sometimes a rationale is offered that they have already reached their practical limits of treatment, and
sometimes no rationale is offered.  In no case does the TMDL submission explicitly cite an "equity"
rationale to the effect that responsibility for further load reductions should not focus on sources that have
already  gone beyond minimum technology-based requirements. On balance, we believe the sample size for
the groundtruthing case studies is small, and we cannot conclude anything  more than some TMDLs in
practice will adopt limitation #1 and some will not.

        The TMDL groundtruthing case studies do not shed any useful light on the second limitation.  It is
really a  policy question rather than an empirical one - what do we count as being in the baseline, and what
do we count as being incrementally attributable to the TMDL program? In our view, projects listed in the
1996 CWNS have been planned so far in advance of virtually all TMDLs that  they should be defined as
not attributable to the TMDL program.

        The rationales for these two potential limitations involving POTWs would seemingly apply to
industrial dischargers as well as to POTWs.

        •       Both industrial dischargers and POTWs that already provide better treatment than
               technology-based standards require may receive some special consideration from control
               authorities for their extra treatment efforts when wasteload allocations are developed under
               TMDLs. Control authorities may look for additional controls  first to sources that have not
               yet  gone beyond the required minimums.  In many cases, POTWs or industrial dischargers
               that already provide advanced treatment will not be required to do more in future TMDLs.
               In some cases, there may be no reasonably available treatment technologies that they could
               implement beyond the advanced treatment they have already adopted.

        •       Similarly, the cost of treatment upgrades that were planned and approved five or more
               years  ago, whether for POTWs or for industrial dischargers, should not be counted as
               prompted by the TMDL program.

        However, we have implemented these two limitations for POTWs  alone because we have data on
implementation of advanced treatment and long-planned projects only for POTWs, from the CWNS. There
is no parallel source of data for industrial dischargers. There is no ready source of information for
industrial dischargers on treatment-in-place (from which we could identify the facilities with treatment
beyond  BPT/BAT) or  on already-planned treatment upgrade projects.
19      This finding is quite consistent with the previous observation that most (68 %) major POTWs appear
        currently to have WQBELs that require better-than-secondary treatment. As noted previously, the nation
        has already made important progress beyond technology-based standards in addressing impaired water
        bodies.

                                              II-7

-------
E. COST FUNCTIONS FOR THE "NEXT TREATMENT STEP'

       We developed relationships predicting cost as a function of flow for all pollutant sources
potentially needing additional controls for BOD, nutrients, toxic organics, ammonia or metals. These cost
functions are as follows.

1. Treatment for metals from direct dischargers except POTWs

       Polishing multi-media filtration was assumed as the "next treatment step", assumed to be
incremental over the technologies assumed to be in place to meet BAT (flow reduction, chemical
precipitation, clarification).  The capital and O&M cost functions for polishing filtration are drawn from
EPA's development document for the centralized waste treatment industry.20 The equations are described
more fully in Appendix  E.

2. Treatment for metals from POTWs

       An enhanced pretreatment program with tighter local limits (tighter than PSES) for significant
metals indirect dischargers was assumed as the "next treatment step" when POTWs need to provide
enhanced control of metals.  The enhanced pretreatment program was assumed to be incremental over a
baseline pretreatment program in which local limits match effluent guideline requirements for indirect
dischargers.

       Our procedure for estimating the costs for such an enhanced pretreatment program at a POTW
involves calculating the number of the POTW's indirect dischargers that will need to improve their
treatment for metals, and then applying the cost functions for polishing filtration to the flows for these
indirect dischargers.  There are several steps in this procedure:

       •       We assumed that any major POTW within the relevant distance upstream of a metal-
               impaired water will need to implement an enhanced pretreatment program for  metals.  We
               assumed that no minor POTWs will need to implement such a program.. These
               assumptions likely result in some overestimate of the number of POTWs that will
               ultimately need to enhance their pretreatment programs as a result of TMDLs.

       •       For  a major POTW that is presumed to need an enhanced pretreatment program for
               metals, we obtained information from the CWNS on the POTW's industrial flow.  We
               then made assumptions and applied information from EPA's RIA for the Great Lakes
               Water Quality Guidance to estimate that significant industrial users totaling about 10 % of
               the POTW's industrial flow would need to improve their metals treatment. These
               significant industrial users were estimated to have an average flow of 0.1 mgd each.

       •       We then applied the cost functions for polishing filtration to the number of indirect
               dischargers that was calculated to need to implement this next treatment step.
20     U.S. EPA, Office of Water. Development Document for Proposed Effluent Limitations Guidelines and
       Standards for the Centralized Waste Treatment Industry. December, 1998.

                                              II-8

-------
3. Treatment for BOD, nutrients, ammonia and toxic organics

        Some form of advanced secondary treatment was assumed as the "next treatment step" for both
industrial dischargers and for POTWs. Advanced secondary treatment is the next treatment increment over
secondary biological treatment, which is assumed to be in place for industrial dischargers to meet
BPT/BAT and for POTWs to meet secondary treatment requirements. The capital cost functions for this
increment of control were drawn from the equations underlying EPA's 1996 Clean Water Needs Survey.21
The CWNS includes an extensive set of cost functions that are used as checks on State-submitted cost
estimates for POTW treatment upgrade projects, and as defaults for generating treatment cost estimates
when States have not developed them.  While the CWNS functions have been used previously by EPA
specifically to estimate costs for POTWs, we use them also to estimate costs for treatment upgrades at
industrial point sources.  The CWNS cost functions are based on underlying cost models for basic
wastewater treatment unit processes (e.g., screening, flow equalization, primary clarification, etc.) that
were developed to apply to both domestic and industrial wastewater.

        Capital cost functions were drawn from the CWNS for upgrades to the "next treatment step"
beyond secondary treatment, as follows:

        •       All dischargers presumed to need additional treatment for nutrients and/or ammonia are
               assumed to incur the costs to upgrade to "secondary treatment with nutrient removal";

        •        All dischargers presumed to need additional treatment for BOD and/or toxic organics are
               assumed to incur the costs to upgrade to "advanced treatment I"; and

        •       All dischargers presumed to need additional treatment for nutrients and/or ammonia and
               BOD and/or toxic organics are assumed to incur the costs to upgrade to "advanced
               treatment I with nutrient removal".

        The specific cost functions that we used are shown and described further in Appendix E. This
Appendix also provides further detail on all the other point  source cost functions discussed in this section.

        It should be noted that the capital costs for upgrades estimated using the CWNS cost functions are
substantially higher than the costs that would be projected using equations derived from POTW cost data
for advanced secondary treatment that underlie EPA's final Best Conventional Technology (BCT) rule.22
The BCT data had been suggested to us as an alternate source for cost information on treatment upgrades.
Upon examination, we concluded that the CWNS equations were more appropriate for our purposes
because:

        •       The treatment upgrades considered in the BCT analysis focused exclusively on increased
               chemical addition and included virtually no capital equipment.
21     Tetra Tech, Inc. Software Requirements Document for the 1996 Clean Water Needs Survey Treatment
       Plant Cost Curves, Revision 2. March 20, 1998.

22     51 F.R. 24974, July 9, 1986. Also: Science Applications International Corp. BCT Benchmarks:
       Methodology, Analysis and Results.  May, 1986.

                                              II-9

-------
        •       The BCT treatment upgrades were intended specifically to provide additional removals of
               BOD and suspended solids.  They likely would not have accomplished the additional
               removals of nutrients, ammonia and toxic organics that can be addressed with the CWNS
               cost functions.

        •       The BCT costs were estimated specifically and only for POTWs, and there is no reason to
               believe they would be appropriate for industrial dischargers also.  We believe the CWNS
               cost functions are reasonably applicable to both POTWs and industrial dischargers.

        •       The CWNS cost functions have been developed and used more recently and more widely
               than the BCT costs.

However, the CWNS cost functions have one major disadvantage.  They are designed in a manner such that
the cost of upgrading an existing treatment plant is estimated as the cost of a new plant providing the
desired level of treatment less the salvage value of the existing plant.  In reality, upgrades can generally be
accomplished for less than is projected by this "new plant less salvage value" approach. We believe the
CWNS cost equations thus probably overestimate the costs of treatment upgrades. Unfortunately, we are
aware of no alternative, broadly applicable, equations for estimating upgrade costs.

        The CWNS - since it is concerned with capital costs only - does not provide functions for
estimating the increased operating and maintenance (O&M) costs associated with upgrading wastewater
treatment facilities. We developed an O&M cost function for upgrades from survey data assembled by the
Association of Metropolitan Sewerage Authorities (AMSA) on O&M costs for POTWs operated by
member institutions.23 AMSA's survey provides  annual O&M costs for each of 119 agencies serving
nearly half of the nation's sewered population.  The data on O&M  costs and flow for these agencies are
broken down by level of treatment provided (primary, secondary, and tertiary).  We segregated the data by
level of treatment, and developed two regression equations, one explaining annual O&M costs as a function
of flow for secondary treatment POTWs and another explaining O&M costs as a function of flow for
tertiary POTWs. We then assumed that the increase in O&M cost for an upgrade from secondary to
advanced treatment was given by the difference between the estimated secondary and tertiary O&M cost
equations.

        We have two reservations about this approach to estimating O&M costs associated with the "next
treatment step" beyond  secondary treatment/BPT/BAT:

        •       The AMSA data is drawn specifically from POTWs, and there may be some reasons why
               O&M costs for industrial facilities would differ systematically from those for POTWs.
               Labor costs for wastewater treatment plant operators, for example, might differ between
               the private and public sectors.

        •       The AMSA data distinguished only between primary, secondary and tertiary levels of
               treatment.  AMSA's data did not distinguish between plants that provided "secondary with
               nutrient removal", "advanced treatment I", "advanced treatment I with nutrient removal",
               "advanced treatment II", and "advanced treatment II with nutrient removal".  Based on
23     Association of Metropolitan Sewerage Authorities. AMSA Financial Survey. 1999.

                                             II- 10

-------
               discussions with officials responsible for the AMSA survey, respondents would likely
               include plants at all of these levels within AMSA's "tertiary" category. The likely result is
               that the O&M cost differential we estimate between AMSA's secondary and tertiary
               categories somewhat overstates the real O&M cost differentials that we are interested in
               resulting from the smaller increments from secondary to  "secondary with nutrient
               removal",  to "advanced treatment I", and to "advanced treatment I with nutrient removal".

       We have used the equation derived from the AMSA data despite these reservations. We are not
aware of any better alternative.

       A further issue in costing the "next treatment step" for point sources is that all of our cost
equations presume that the "next treatment step" begins from the point at which the source just meets
applicable technology-based requirements. For POTWs, the "next treatment step" is assumed to be an
increment beyond secondary treatment, while for industrial point sources the increment is beyond BPT and
BAT. In fact, though, as noted previously, many POTWs and industrial dischargers currently have
treatment in place that exceeds technology-based requirements. For dischargers with advanced treatment in
place, the "next treatment step" should be different from the "next treatment step" for dischargers without
advanced treatment in place. Assuming increasing marginal costs of control, one might surmise that the
"next treatment step" for a discharger already having advanced treatment would be more costly than that
for a discharger just meeting technology-based standards.

       There are several reasons why we have not been able to improve our costing procedure and address
this issue:

       •       For POTWs.  Note that this issue does not arise when we assume that the TMDL program
               will not require a "next treatment step" of POTWs that already provide better-than-
               secondary treatment. Under this assumption, we do not cost out any additional treatment
               for POTWs that already have advanced treatment.  However, in the sensitivity analysis we
               showed in the previous section (e.g., Exhibit II - 1) where we investigate the impact if
               beyond-secondary POTWs are required to take the "next treatment step", we have
               assumed and costed "next treatment steps" that start from the wrong point for these
               POTWs. We would face a major difficulty in trying to remedy this. The CWNS capital
               cost functions for POTW upgrades involve projecting the cost of building a plant at the
               new treatment level less the salvage value of the existing plant at the current treatment
               level.  The salvage value equations exist only for secondary plants, not for beyond-
               secondary plants.  Upgrading a beyond-secondary plant to provide further improvements
               would, following the CWNS approach, get a salvage value credit for the existing plant as
               if it were only at secondary. The projected net costs for the upgrade from beyond-
               secondary to well-beyond-secondary would be unreasonably large.

       •       For industrial dischargers. There is no reasonable way to obtain information on the extent
               of treatment in place for each industrial discharger, and hence the starting point for each
               discharger from which the "next treatment step" will begin.
                                             II-  11

-------
F. FLOW DATA FOR USE IN COST FUNCTIONS

        Costs were estimated on a facility-by-facility basis using the previously discussed cost functions
and facility flow information from PCS or CWNS.24  Costs were then summed for all dischargers.

        We encountered significant problems resulting from missing or obviously inaccurate facility flow
information in PCS/CWNS. The set of assumptions and procedures adopted to address the issues that
arose are outlined in Appendix F.

        The set of assumptions and procedures to deal with flow issues has a substantial impact on the
estimated costs for the affected sources to install the next treatment step.  If the assumed limits on the flows
to be treated were much higher, the estimated costs could as much as double. We believe, however, that
our assumed upper limits on the flows presumed to be treated are conservative — the great majority of
minor and major dischargers will have flows needing additional treatment that are actually well below the
upper limits that were assumed.

G. COSTS FOR SOURCES AFFECTING WATERS IMPAIRED BY NONPOINT SOURCES
ONLY

        Many States provide an assessment of the source types that impair each of their 303(d) waters.
For the States that provide this information, the data can be compiled to determine whether each water is
impaired by point sources only, by nonpoint sources only, by mixed point and nonpoint sources,  by
unknown sources, or by other categories of sources. Other States do not provide such information on their
impaired waters. In our analysis to this point, we have identified all point sources that presumably
discharge an impairment pollutant within a relevant distance upstream of or into an impaired water.25  This
set of point sources is likely the maximum that potentially might be addressed in water quality-based
permitting or TMDLs.  However, we believe we can use the information that States provide on source
types responsible for the impairments in these waters to judge which of the maximum set of potentially
relevant point sources are really likely to be addressed in the eventual TMDLs or water quality-based
permits.  As the most obvious examples:

        •       When a State indicates that point sources are the source of impairment in a water body, we
               expect that the TMDL or alternative approaches would likely address all the point sources
               discharging the impairment pollutant within a relevant distance upstream (i.e., all of the
               point sources we have identified in this analysis); but

        •       When a State indicates that nonpoint sources - and not point sources - are the source of
               impairment in a water body, we expect that the TMDL or water quality-based permitting
               process most likely would not address point sources, even if there are some point sources
24     For POTWs, flow information from CWNS was used preferentially over flow information from PCS.

25     We say "presumably" because we do not have information particular to each point source on whether it
       actually does or does not discharge any given pollutant. Instead, we make a judgment for each SIC as to
       whether all dischargers in that SIC presumably discharge each class of pollutants. These broad judgments
       may be inaccurate with respect to any particular point source discharger.

                                              II-  12

-------
                that apparently (according to our engineering judgment approach) discharge the
                impairment pollutant within a relevant distance.

        In this step of the analysis, we use the information provided by States on the source types
responsible for impairment of each 303(d) water body to reduce the set of point sources that could
potentially be considered in TMDLs or water quality-based permitting (the maximum set) down to a
smaller set of point sources that are likely to be considered in these processes.

        In the "within and upstream" case, there are 4,234 impaired water bodies that we identify as
impaired by one or more of the five pollutant classes that we analyze (BOD, nutrients, metals, toxic
organics, and ammonia) and that have one or more point source dischargers within a relevant distance
upstream that presumably discharge the impairment pollutant.26 The following exhibit shows what States
report as the sources of impairment for these water bodies, and also shows,  in the final column, our
judgment as to whether point sources  affecting these water bodies are likely to be addressed in TMDLs or
water quality-based permitting.

        In essence, we assume that a State will consider requiring further controls for relevant point
sources via TMDLs when dealing with water bodies that States say are impaired by point sources, and a
State will not consider requiring further controls for point sources when dealing with water bodies that are
not impaired by  point sources. This assumption is supported by findings  among our "groundtruthing"
sample of actual TMDLs (see Appendix A). For a final category of waters  - those for which States have
not reported information about the types of sources that are responsible for impairment - we will
extrapolate information derived from the waters for which States have reported information on sources of
impairment.

        Our procedure for using the information from States on the source types responsible for
impairment is described fully in Appendix G. In essence, we assume that a  State will not require further
controls for point sources if the water body is cited by the State as impaired by source types other than
point sources. This assumption has the effect of reducing estimated point source costs by roughly 35 %
(varying slightly across Scenarios and cases) relative to the costs that would be incurred if further controls
were to be required of all point sources discharging the impairment pollutant within a relevant distance of
any impaired water.

        At this point, costs for Scenario 1 ("Least Flexible  TMDL Program") are estimated. The major
assumptions in estimating costs for Scenario 1 are summarized in Exhibit II-3.
26      These 4,234 water bodies are for the "within and upstream" case. They represent a little more than one-
        quarter of the 16,143 impaired water bodies that we cover in our analysis.  (Our procedure for matching
        impaired water bodies against point sources is operable in 44 States plus the District of Columbia.  In
        these 44 States plus DC, there are 18,162  303(d)-listed waters, of which 16,143 have been
        georeferenced.) Interpreting the "within and upstream" case as an upper bound, we thus believe that at
        most about 1/4 of all TMDLs have the potential to trigger additional controls for point sources. In
        contrast, the "within only" case provides a reasonable lower bound. In the "within only" case, there are
        XXX water bodies that are impaired by one or more of the five pollutants and have one or more point
        sources discharging the impairment pollutant directly into the impaired water.  This suggests that there
        may be as few as XXX percent of all TMDLs that have the potential to trigger additional controls for
        point sources.

                                               II- 13

-------
                                               Exhibit 11-2
Sources of Impairment Reported for Water Bodies That Have Point Sources Within/Upstream That
                           Presumably Discharge the Impairment Pollutant
Sources of Impairment Reported by
States
PS only
NPS only
other only
unknown only
not reported
PS + NPS only
PS + NPS + other only
PS + NPS + unknown only
PS + other only
PS + unknown only
PS + other + unknown only
NPS + other only
NPS + unknown only
NPS + other + unknown only
other + unknown only
PS + NPS + other + unknown
Total
Number of Water Bodies
(Within and Upstream Case)
141
829
71
110
1,727
368
339
15
53
6
5
425
48
35
25
37
4,234
For These Water Bodies, are TMDLs or Water
Quality-Based Permits Likely to Address Point
Sources?
Yes
No
No
Unclear - scale to these waters
Unclear - scale to these waters
Yes
Yes
Yes
Yes
Yes
Yes
No
No
No
No
Yes

                                                Exhibit II - 3
              Assumptions in Estimating Costs for Scenario 1 - Least Flexible TMDL Program

 1.  Under a least flexible TMDL program, States would write water quality-based permits for every point sources identified
 as discharging an impairment pollutant into ("Within only" case) or into or near upstream of ("Within and upstream case") an
 impaired water body, except:

             This applies only to water bodies cited as impaired in part by point sources. WQBELs will not be developed
             for point sources affecting impaired waters cited as not impaired by point sources.

 2.  The WQBEL will require each such point source to implement an appropriate "next treatment step" beyond whatever
 controls the source now has in place, except:

             We assume that POTWs already providing better-than-secondary treatment will not be required to implement
             the "next treatment step";

             The costs for POTW upgrades that were sufficiently far along in planning as of 1996 to be included in the
             1996 CWNS are counted as part of the baseline and should not be attributed to Scenario 1.

             We lack information on treatment in place at industrial dischargers, and assume that the cost of the "next
             treatment step" for an industrial discharger is the same whether treatment in place is just sufficient to meet
             technology-based standards, or exceeds this minimum.

 3.  The sum of the "next treatment steps" that get implemented consistent with these assumptions is sufficient to abate the
 point source contributions to impairment of all waters that are impaired at least in part by point sources.
                                                  II- 14

-------
        Using these major assumptions, we estimate the costs to point sources under the "Least Flexible
TMDL Program" scenario to be:

                                           Exhibit II - 4
                Estimated Costs for Point Sources -- Least Flexible TMDL Program
Type of Source
Industrial dischargers
Indirect dischargers (metals)
POTWs
Total
Annual Costs
(2000 $ in millions)
Low Est.
676
10
396
1,082
High Est.
1,465
16
697
2,178
Number of Affected Facilities
Low Est.
3052
at 148 POTWs
1094
4,146
High Est.
8557
at 3 12 POTWs
3335
11,893
H. COSTS FOR SCENARIO 2 (Moderately Cost-effective TMDL PROGRAM)

        In Scenario 1 ("Least Flexible TMDL Program"), we assume that States will require load
reductions for every point source identified as contributing to impairment of a 303 (d) water body.  In
estimating the costs of Scenario 2 ("Moderately Cost-effective TMDL Program"), the costs estimated for
Scenario 1 are scaled down to reflect the load reductions typically required of point sources in actual
TMDLs relative to the load reductions simulated in Scenario 1. Based on a  sample of 15 recent TMDLs
(see Appendix A),27 we project that:

•       About half of all TMDLs will require an aggregate load reduction from point sources
        approximately equal to the load reduction that would be obtained if  all contributing point sources
        were to implement the "next treatment step" (roughly 50 - 85 %).

•       The remaining half of all TMDLs will require an aggregate load reduction from point sources
        (roughly 10 - 40 %) that is only about half as much as the load reduction that would be obtained if
        all contributing point sources were to implement the  "next treatment step" (roughly 50 - 85 %). In
        these TMDLs, the cost for  point source dischargers to achieve their  aggregate load reduction might
        be only about half of that if all contributing point sources were to have to implement the "next
        treatment step".28
27      The sample of 15 actual TMDLs is used to inform and test the assumptions applied in this analysis. The
        sample will be expanded and our assumptions will be revised accordingly for the final version of this
        analysis.

28      Perhaps only half of all the contributing point sources would need to implement the "next treatment step",
        perhaps all the contributing point sources would need to implement something substantially less than the
        "next treatment step", or perhaps the outcome would be somewhere in between. In any case, the costs for
        point source dischargers to achieve a 10 - 40 % aggregate load reduction would be roughly half of the
                                              II- 15

-------
This information suggests that total national costs for point sources if TMDLs are developed would be only
about 3/4 as much as their costs would be if States proceeded to develop water quality-based permits for all
point sources without making more careful TMDL calculations. For about half of the waters where point
sources contribute to impairment, the holistic assessment conducted during a TMDL will likely determine
that the aggregate load reduction needed from point sources is much less than the reduction that would be
obtained if each point source were addressed individually through water quality-based permitting. For
point sources, we estimate the costs under the "Moderately Cost-effective TMDL Program" scenario to be
3/4 of the costs that would prevail under the "Least Flexible TMDL Program" scenario.29 The major
assumptions in estimating costs for the Moderately Cost-effective TMDL Program scenario are
summarized in Exhibit II-4.
                                               Exhibit II - 4
     Major Assumptions in Estimating Costs for Scenario 2 - Moderately Cost-effective TMDL Program

  1. A State will develop a TMDL for an impaired water in a careful, holistic manner. The State will determine the maximum
  pollutant load the water body can tolerate while meeting water quality standards (with a margin of safety), and will then
  allocate this allowable load among the sources contributing the pollutant. The totaled allowable loads assigned to all
  sources will be less than but close to the maximum that the water body can tolerate. In contrast, the water quality-based
  permitting approach will address each point source in isolation, and States will often require load reductions that in total
  overshoot the aggregate load reduction needed to meet standards.

  2. Costs for the WQBEL approach (Scenario 1) are estimated based on the assumption that the State will require all point
  sources that contribute to impairment to implement an appropriate "next treatment step" (subject to the assumptions
  previously noted regarding waters impaired by nonpoint sources rather than point sources, POTWs that already provide
  better-than-secondary treatment,  etc.).

  3. For about half of all TMDLs involving point sources, the aggregate load reduction required of point sources will be
  approximately equal to the load reduction that would be obtained if all contributing point sources were to implement the
  "next treatment step".  For the remaining half of the TMDLs involving point sources, the aggregate load  reduction required of
  point sources will be approximately half of the load reduction that would be obtained if all point sources implemented the
  "next treatment step".

                  Costs of  Scenario 2 for point sources are thus estimated at 3/4 the costs of Scenario 1.
        Using these major assumptions, we estimate the costs to point sources under the "Moderately Cost-
effective TMDL Program" scenario to be:
        costs to achieve a 50 - 85 % aggregate load reduction.

29      Note again that we have not defined the "Least Flexible TMDL Program" scenario or the "Moderately
        Cost-effective TMDL Program" scenario to involve the worst case for point sources in which States pursue
        point sources for load reductions in an attempt to eliminate impairments in waters that are affected
        primarily or entirely by nonpoint sources.

                                                  II- 16

-------
                                           Exhibit II - 5
      Estimated Costs for Point Sources Under the Moderately Cost-effective TMDL Program
Type of Source
Industrial dischargers
Indirect dischargers (metals)
POTWs
Total
Annual Costs
(2000 $ in millions)
Low Est.
507
8
297
812
High Est.
1,099
12
523
1,634
Number of Affected Facilities
Low Est.
2289
at 111
POTWs
821
3,110
High Est.
6418
at 234 POTWs
2502
8,919
I. SAVINGS WITH "COST-EFFECTIVE WASTE LOAD ALLOCATIONS"

        Neither the Clean Water Act nor EPA's implementing regulations prescribe how a total maximum
daily load is to be allocated among the sources that discharge the impairment pollutant. The State may
assign responsibilities among sources for load reductions as the State wishes.  Different allocations will
result in different total costs of achieving the desired total load reduction, as a function of the differing
costs per pound for the various pollutant sources to reduce their loads.  In general, the total costs of
achieving the target load reduction will be lower if the sources with lower per unit control costs are
assigned responsibility for achieving the bulk of the desired total load reduction. We use the term "cost-
effective wasteload allocation" to denote a situation in which the State attempts to minimize aggregate costs
by assigning responsibility for achieving most of the  total desired load reduction to sources that have
relatively low costs of achieving load reductions. Alternatively, the same economically efficient result
(achieving a desired total load reduction in the least-cost manner) can be achieved, in theory, given any
initial allocation of control responsibilities, if "trading" is allowed.  With trading, any source that is
assigned responsibility for a load reduction is free to achieve that load reduction itself, or to buy the
equivalent load reduction from another source that might be able to provide it at lesser cost. Whatever the
initial allocation, trading will tend ultimately to elicit load reductions from the lowest cost sources.

        The "More Cost-Effective TMDL Program" scenario recognizes the possibility of reducing TMDL
costs to dischargers through either additional "cost-effective wasteload allocations" or through trading, or
both.  Either of these approaches would reduce the eventual costs to dischargers well below what they
would be if TMDLs assigned load reductions on a cost-neutral basis (e.g., if sources were assigned load
reductions that required each of them to implement the "next treatment step"; or if load reductions were
assigned on a simple proportional rollback basis).  We expect that pressure to adopt cost-minimizing
approaches will build as the TMDL program grows in the future, and more TMDLs will tend toward this
"more cost-effective" model. Note, though, that there may be some instances where other concerns (e.g.,
equity, concern about implementation and enforcement complexities attendant to trading) prevent use of
these cost-minimizing approaches.
                                              II- 17

-------
        Many factors account for differences across sources in costs to abate additional pounds of a
pollutant:

        •      General differences between point sources and nonpoint sources in their typical costs of
               controlling pollutants.  In many cases, nonpoint sources have been found able to abate
               additional pounds of nutrients, BOD and sediment at much lower costs than can point
               sources.30

        •      Idiosyncratic differences between dischargers that reflect differing pollution abatement
               options available to different dischargers. These differences may have some component
               that reflects systematic differences across industries (e.g., metal finishers are likely able to
               abate metal discharges at lower cost per pound than food processors, while food
               processors are likely able to abate BOD dischargers at lower cost per pound than metal
               finishers).  However, there are also very often substantial additional differences across
               dischargers in cost per pound of control that are essentially unpredictable.

        •      Systematic differences between pollutant sources  involving economies of scale (e.g., other
               things being equal, a large AFO will be able to abate additional nutrient loadings at lower
               cost per pound than a small AFO) and increasing  marginal costs of control (e.g., other
               things being equal, a POTW currently providing secondary treatment will likely be able to
               abate additional nutrient loadings at lower cost per pound than a POTW providing tertiary
               treatment).

        Such differences in per pound abatement costs provide opportunities for many sorts of cost-
effective waste load allocations: in allocations among several point sources, in allocations between point
and nonpoint sources, in allocations among several nonpoint sources, in allocations (developed by a
POTW) among several indirect dischargers, etc..  Or, restating this in terms of trading, there are many
ways that trading may save money: point/point trading, point/nonpoint trading, nonpoint/nonpoint trading,
pretreatment trading, etc..  Because so many of these trading or cost-effective waste load allocation
opportunities are idiosyncratic and not predictable, we are unable to simulate them in this analysis.  In this
analysis, we have been able to simulate savings only from cost-effective waste load allocations in which
additional control of nonpoint sources is substituted for additional control of point sources (e.g.,
point/nonpoint trading). This form of trading is the most common form that has occurred to date in the
water program.31  However, experience in other programs suggests that other forms of trading may also
yield important cost savings (e.g., air program trading among point sources involving SOx, NOx, VOCs,
greenhouse gases, etc.). Ultimately, the savings achievable in restoring impaired waters via cost-effective
waste load allocations are likely to be much greater than those we  can estimate here.

        The cost-effective waste load allocation activity that we simulate here as representative of Scenario
3 involves shifting some responsibilities for additional control responsibilities from point sources to
nonpoint sources.  This shift of some responsibilities from point sources to nonpoint sources may occur
30      See: Environomics, Inc. A Summary of U.S. Effluent Trading and Offset Projects.  November, 1999.
        www.epa.gov/owow/watershed/trading.

31      Ibid.

                                               II- 18

-------
before the waste load allocation (WLA) is developed under the TMDL (e.g., the State considers the relative
cost-effectiveness of controlling point and nonpoint sources and then allocates loads among the source
types in a manner that reflects this and minimizes total compliance costs) or after the WLA is developed
(e.g., the WLA is established and then point and nonpoint sources are allowed to trade -- typically a point
source substitutes a lower cost reduction that it achieves by controlling nonpoint sources or by paying
nonpoint sources to control for some of the load reduction that the point  source is required to accomplish).

        We believe the most likely circumstances in which there will be cost-effective opportunities for
control of nonpoint sources to substitute for further control of point sources are as follows:

        •      For water bodies that are impaired by both point and nonpoint sources. Obviously both
               sorts of sources must contribute meaningful amounts of the impairment pollutant if there is
               to be any opportunity for control of one to substitute for control of the other.

        •      For impairments involving nutrients or BOD.  Virtually all of the instances thus far in
               which point-nonpoint tradeoffs have occurred involve nutrients or BOD.32 It is rare that a
               water body impaired by a toxic pollutant faces significant discharges of the toxic pollutant
               by both point and nonpoint sources that can allow for tradeoffs.

        •      For non-flowing as opposed to flowing water bodies. Impairment of lakes and estuaries
               typically results from an accumulated load of a pollutant from an entire watershed over a
               substantial period of time (ranging from  a season to many years). In these circumstances,
               loads from different locations in the watershed and loads that occur at different times can
               relatively safely be traded off against each other - a pound of the pollutant  has a roughly
               similar impact largely independent of whether it comes from a point or nonpoint source
               and where and when it is discharged. In contrast, impairment of a flowing water body is
               often more localized (e.g., a particular river reach rather than an entire lake) and more
               episodic (e.g., the impairment occurs at particular times, such as during wet weather or
               during low flow conditions). In flowing waters, discharges of the same pollutant by
               various point and nonpoint sources often have differing  impacts depending on the location,
               time, and nature of the discharge. It is much more difficult to implement an
               environmentally protective trading program under such circumstances.

        We evaluated all the point sources affecting impaired waters with respect to these considerations,
and assigned the point sources (and the costs we estimated they would incur under Scenario  2, the
Moderately Cost-effective TMDL Program) to one of three groups reflecting their suitability for cost-
effective waste load allocations involving point/nonpoint tradeoffs:

        •      Group 1. Point sources highly suited to cost-effective WLAs/trading. These point
               sources discharge BOD and/or nutrients  to non-flowing  waters that are impaired for these
               pollutants by both point and nonpoint sources.  We estimate that this group comprises 9 %
               (upstream and within case) or 17 % (within only case) of all point sources that affect
               impaired waters. The potentially tradeable costs for these point sources amount to 8 %
32      Ibid.

                                              II- 19

-------
                (upstream and within case) or 10 % (within only case) of the total costs that we estimate
                point sources will incur under the Moderately Cost-effective TMDL Program (Scenario 2).

        •        Group 2.  Point sources moderately suited to cost-effective WLAs/trading. These point
                sources discharge BOD and/or nutrients to flowing waters that are impaired for these
                pollutants by both point and nonpoint sources. We estimate that this group comprises 29
                % (upstream and within case) or 47 % (within-only case) of all point sources that affect
                impaired waters.  The potentially tradeable costs for these point sources amount to 21  %
                (upstream and within case) or 27 % (within only case) of the total costs that we estimate
                point sources will incur under the Moderately Cost-effective TMDL Program (Scenario
                2).33

        •        Group 3.  Point sources not suited to cost-effective WLAs/trading.  These point sources
                fail one or more of the conditions that we presume necessary for cost-effective
                WLAs/trading. We estimate that this group comprises 62 % (upstream and within case) or
                36 % (within only case) of all point sources that affect impaired waters. Non-tradeable
                costs for the point sources in this and the other groups amount toll % (upstream and
                within case) or 63 % (within only case) of the total costs that we estimate point sources
                will incur under the Moderately Cost-effective TMDL Program (Scenario 2).

Note that we have assigned point sources and costs to these three groups in a conservative manner:

        •        A point source is assumed to be able to save money from participating in a more cost-
                effective WLA only if all the impaired water bodies that the point source affects are
                amenable to more cost-effective WLAs.  Thus, if as few as one impaired water body
                affected by a point source is not amenable to cost-effective WLA (i.e., if at least one water
                body does not have both point and nonpoint sources discharging the impairment pollutant),
                the source will be precluded by that water body from avoiding the next treatment step.34

        •        Even though a point source might seem able to avoid the cost of a "next treatment step" by
                participating in a more cost-effective WLA for nutrients or BOD, the point source will not
                be able to do so if it nevertheless is presumed to need the next treatment step to abate
33      Note that the percentage of point sources apparently suited to participating in cost-effective WLAs (in
        either Group 1 or Group 2) is substantially greater than the percentage of costs that are amenable to cost-
        effective WLAs. This is because we have assumed that metals, toxic organics and ammonia are not
        amenable to cost-effective WLAs, and costs for any "next treatment step" controls needed to abate these
        pollutant classes are "off the table". Many point sources may be able to participate in cost-effective WLAs
        for BOD or nutrients ("tradeable" pollutants) while nevertheless incurring some control costs (for non-
        "tradeable" pollutants) that are not eligible for cost-effective WLAs.

34      It is important to note that many point sources affect multiple impaired waters, particularly in the
        upstream and within  case, in which a point source's influence is presumed to be felt far downstream.
        Considering only non-metal pollutants (in our view, metals require wholly different  treatment
        technologies than non-metals and have no impact on suitability or non-suitability for a point source to
        participate in a more cost-effective WLA), 44 %  of the point sources contribute to impairment of multiple
        water bodies in the upstream and within case, and 12 % of the point sources do so in the within only case.

                                               11-20

-------
               either toxic organics (advanced secondary treatment, the same upgrade as is needed for
               BOD) or ammonia (secondary treatment with nutrient removal, the same upgrade as is
               needed for nutrients). In effect, the presumed need to abate toxic organic loads precludes
               participating in a cost-effective WLA for BOD, while the need to abate ammonia precludes
               participating for nutrients.

        •       Any point source that affects multiple impaired waters and appears able to participate in a
               cost-effective WLA for each of them is assigned to Group 2 rather than Group 1 if as few
               as one of the impaired waters is non-flowing.

        In estimating the potential savings to point sources from participating in more cost-effective WLAs
(Scenario 3), we assume that: 1) All of the Group  1 (highly suited) point sources will participate; 2) Half of
the Group 2 (moderately suited) point sources will participate; and 3)  None of the Group 3 (unsuited) point
sources will participate.  We have no empirical basis for this particular assumption regarding participation
rates. We are confident only that very few of the Group 3 sources will be able to participate in
point/nonpoint sorts of cost-effective WLAs, while participation should be substantially higher in Group 2
and substantially higher again in Group 1.  Different assumptions may be substituted for the participation
rates we assume, and the effects will be proportional (if half of Group 1  sources were assumed to
participate rather than all of them, the savings for them would be half as much).

        With these assumptions, we can estimate the cost savings for point sources that could result from
more cost-effective WLAs that shift some control responsibilities to nonpoint sources.

        How much might it cost for nonpoint sources to control the quantity of pollutants that are no longer
to be controlled by point sources? The answer  is highly uncertain. In Appendix H  we summarize the
results from nine instances in which the costs for nonpoint source controls have been compared with the
costs for point source controls.  The nine cases  involve TMDLs or water quality standard attainment
studies that are very similar to the situations for which we are postulating more cost-effective WLAs -
instances in which nonpoint source controls  for nutrients or BOD are being considered as alternatives to
additional treatment efforts at point sources  (typically advanced treatment at POTWs).  The results of
shifting control responsibilities from point to nonpoint sources are quite variable, ranging from more than
90 % savings in one case to an increase in cost  in another case. The median figure  across the seven
instances where percentage cost comparisons have been quantified is a 75 % savings for nonpoint source
controls relative to point source controls.  On this basis, we will assume for our analysis that nonpoint
sources can provide additional control of nutrients and BOD at a cost  per pound equal to 25 % of that for
additional point source controls.  The additional nonpoint source controls that are needed to offset the
increased  point source loads allowed in a more  cost-effective WLA will cost 25 % as much as the savings
in point source costs.

        Based on these assumptions, Exhibit II - 6 shows the estimated costs for Scenario 3, the "More
Cost-Effective TMDL Program".
                                              11-21

-------
                                 Exhibit II - 6
Estimated Costs for More Cost-Effective TMDL Program (Scenario 3) ($ in millions)

Upstrean and within case
Hiojly suted to costeffedive VMAs
Moderately suted to costeffedive VMAs
Not suted to costeffedive WAs
Tctel
Wthin only case
Hiojly suted to costeffedive VMAs
Modardsly suted to ccstefedive WAs
Not suted to costeffedive WAs
Tctel
TWO. Pro-am
#cfPS




8919




3110
Cosb




$1,634




$312
Cost-Elective 1ML Program
%cfPS

881
23.72
6247


16.79
46.89
33.32

% of Costs

846
21.40
70.14


9.33
23.98
63.52

% Parfdpding

100
33
0


103
33
0

# PS PaHdpafr^

783
1231
0
2037

522
729
0
1251
PSCostSa/ings

$133
$175
$0
$313

$77
$110

-------
        The gross savings to point sources from more cost-effective WLAs are estimated at $187 - 313
million/yr, constituting 19 - 22 % of point source costs under the Moderately Cost-effective TMDL
Program (Scenario 2). 23 - 40 % of all point sources contributing to impairments are likely to participate
in the more cost-effective WLAs. The net savings from more cost-effective WLAs are estimated at $140 -
235 million/yr (savings to point sources partly offset by increased costs of nonpoint source controls).
These savings are estimated only for cost-effective WLAs involving shifting control responsibilities from
point to nonpoint sources.  Further savings are possible from other sorts of reallocations, but these have not
been estimated.
                                             Exhibit II - 7
        Major Assumptions in Estimating Costs for Scenario 3 - More Cost-Effective TMDL Program

  1. The estimates address potential savings from shifting control responsibilities from point to nonpoint sources.  Savings
  have not been estimated for other sorts of reallocations.

  2. Cost-effective WLA opportunities will focus on non-flowing waters impaired by nutrients or BOD from both point and
  nonpoint sources.  Lesser opportunities will be available for similarly impaired flowing waters.

  3. The following percentages of point sources are assumed to participate in more cost-effective WLAs: 100 % of highly
  suited point sources, 50 % of moderately suited point sources, and no unsuited point sources.

  4. The cost per pound for nonpoint sources to abate nutrient or BOD loads is about 1/4 of that for point sources.
        A more detailed description of some of our assumptions and procedures in simulating more cost-
effective WLAs is provided in Appendix H.

J. SUMMARY COST ESTIMATES FOR POINT SOURCES

        For Scenario 1  (Least Flexible TMDL Program), costs are estimated for the next treatment step for
each point source identified as discharging the impairment pollutant into or within a relevant distance
upstream of each point source-impaired water body. Costs are then summed across pollutant sources. We
assume that a pollutant source affecting multiple impaired water bodies that are impaired for the same
pollutant will need to implement the next treatment step only once - we assume that the next treatment step
will suffice to achieve this pollutant source's desired load reductions whether the pollutant source
contributes to impairment of one or many water bodies.

        The raw cost estimates calculated in this manner are then scaled to reflect two factors:

        •      Incomplete coverage in the analysis.  The analysis is incomplete, in that it does not cover
               all impaired water bodies (some are not georeferenced with respect to Reach File 3
               reaches) and all point sources (some point sources  also have not been georeferenced), and
               the geographic procedure for matching water bodies against point sources is not operable
               in several States.  As described in section I.C.3, the scaling factor to  account for
               incomplete coverage in the point source analysis is 1.605.

        •      The time at which pollutant sources will begin incurring implementation costs.  Costs are
               estimated initially as the annualized amounts that pollutant sources will pay once their

                                               11-23

-------
               implementation costs begin.  In order to focus any comparison across Scenarios on
               technical differences rather than timing issues, we assume that pollutant sources will incur
               implementation costs at the same time under each of the three Scenarios.  We assume that
               implementation costs for a source will not to begin until five years after the TMDL is
               developed that requires further controls for the source. We assume that TMDLs will be
               developed at an even pace over the years from now through 2015, the deadline for
               completing TMDLs for the 1998 303(d) list.  As a result, the median source is not
               expected to begin incurring its implementation costs until 2013.35 As described in section
               I.C.5, the scaling factor to reflect the assumed pace of TMDL development and
               compliance lag time is 0.4484.

        Combining these two scaling factors, the raw implementation costs estimated for point sources
from the cost equations are multiplied by 0.7197 (1.605 x 0.4484) to develop national estimates of
annualized costs beginning in the year 2000.

        Cost estimates for Scenario 2 (Moderately Cost-effective TMDL Program) are developed by
adjusting downward the costs estimated for Scenario 1, assuming that  TMDLs will result in a more precise
calculation of how much load reduction is needed from point sources in order to meet water quality
standards. More specifically, the costs estimated for Scenario 2 reflect the degree of aggregate load
reduction that actual TMDLs have required of point sources, relative to the load reduction that would be
achieved if all point sources were to implement the "next treatment step" as in Scenario 1.

        Cost estimates for Scenario 3 (More Cost-Effective  TMDL Program) are developed by applying to
the cost estimates for Scenario 2: a) projections regarding how many point sources will participate in more
cost-effective WLAs; and b) an estimate for the percentage cost savings when nonpoint source controls for
nutrients and BOD are substituted for point source controls.

        The resulting estimated annual costs for point source pollutant sources under each of the three
scenarios are as follows:
35      There are an estimated 36,225 TMDLs to be developed for the 1998 303(d) waters.  The mid-point in the
        projected schedule over which these TMDLs will be developed will occur in 2008 - half of the TMDLs
        will be developed before then, half afterwards (see Exhibit 1-2).  A point source required to implement
        further controls by a TMDL developed in 2008 is assumed to begin incurring costs five years later, in
        2013.

                                              11-24

-------
                         Exhibit II - 8
Point Source Costs Under Scenario 1: Least Flexible TMDL Program
             Annualized Costs by Type of Discharger

                  Within and Upstream Case

# Facilities
Total Cost/yr
Average Cost/Facility/yr
SI Us -Metals
Treatment
at312POTWs
$16,493,652
$52,864
POTWs (non-
Metals)
3,335
$696,968,954
$208,986
Industrial
Dischargers
8,557
$1,464,694,468
$171,169
Total - All Sources
11,893
$2,178,157,075
$183,146
                      Within Only Case

# Facilities
Total Cost/yr
Average Cost/Facility/yr
SI Us -Metals
Treatment
148 POTWs
$10,120,959
$68,385
POTWs (non-
Metals)
1,094
$395,814,376
$361,805
Industrial
Dischargers
3,052
$676,345,333
$221,607
Total - All Sources
4,146
$1,082,280,668
$261,042
                         Exhibit II - 9
Point Source Costs Under Scenario 1: Least Flexible TMDL Program
             Annualized Costs by Type of Discharger

                  Within and Upstream Case

# Facilities
Total Cost/yr
Average Cost/Facility/yr
Metals
4,688
$209,419,393
$44,671
All Other
Pollutants
9,865
$1,968,737,682
$199,568
Total -All
Pollutants
11,893
$2,178,157,075
$183,146
                      Within Only Case

# Facilities
Total Cost/yr
Average Cost/Facility/yr
Metals
1,707
$93,827,202
$54,966
All Other
Pollutants
3,289
$988,453,467
$300,533
Total - All
Pollutants
4,146
$1,082,280,668
$261,042
                           11-25

-------
                                        Exhibit II -10
         Point Source Costs Under Scenario 2: Moderately Cost-effective TMDL Program
                            Annualized Costs by Type of Discharger

                                  Within and Upstream Case

# Facilities
Total Cost/yr
Average Cost/Facility/yr
SI Us -Metals
Treatment
at 234 POTWs
$12,370,239
$52,864
POTWs (non-
Metals)
2,502
$522,726,716
$208,924
Industrial
Dischargers
6,418
$1,098,520,851
$171,162
Total - All Sources
8,919
$1,633,617,806
$183,162
                                       Within Only Case

# Facilities
Total Cost/yr
Average Cost/Facility/yr
SI Us -Metals
Treatment
111 POTWs
$7,590,719
$68,385
POTWs (non-
Metals)
821
$296,860,782
$361,584
Industrial
Dischargers
2,289
$507,259,000
$221,607
Total - All Sources
3,110
$811,710,501
$261,000
                                        Exhibit 11-11
         Point Source Costs Under Scenario 2: Moderately Cost-effective TMDL Program
                            Annualized Costs by Type of Discharger

                                  Within and Upstream Case

# Facilities
Total Cost/yr
Average Cost/Facility/yr
Metals
3,516
$157,064,544
$44,671
All Other
Pollutants
7,398
$1,476,553,261
$199,588
Total - All
Pollutants
8,919
$1,633,617,806
$183,162
                                       Within Only Case

# Facilities
Total Cost/yr
Average Cost/Facility/yr
Metals
1,280
$70,370,401
$54,977
All Other
Pollutants
2,467
$741,340,100
$300,503
Total -All
Pollutants
3,110
$811,710,501
$261,000
       We project that total costs for point sources under the Moderately Cost-effective TMDL Program
(Scenario 2) will be 25 % lower than they would be with the Least Flexible TMDL program (Scenario 1).
Although we show in the exhibits a similar 25 % reduction in the number of point sources affected under
the Moderately Cost-effective TMDL Program compared with the Least Flexible TMDL Program, we do
not intend this to be a conclusion of the analysis. The Moderately Cost-effective TMDL Program will
require 25 % less aggregate load reduction from point sources than will be required under the Least
                                            11-26

-------
Flexible TMDL Program. This lesser load reduction can be accomplished by: 1) Requiring the same
load reduction per source but requiring such load reductions from 25 % fewer sources (as we show);
2) Requiring a 25  % lower load reduction per source, but requiring these load reductions from all sources;
or 3) Any combination of load reduction per source and number of sources addressed that results in 25 %
less aggregate load reduction.

       Costs for point sources under the More Cost-Effective TMDL program will be an estimated 19 -23
% lower than the costs under the Moderately Cost-effective TMDL Program.

                                          Exhibit 11-12
          Projected Savings to Point Sources from Cost-Effective WLAs ($ in millions/yr)

Into + upstream
case
Into only case
Moderately Cost-effective
Costs # PS w/ costs
$1,634 8919
$812 3110
More Cost-Effective TMDL Program
Costs Savings
$1,321 $313
% relative to TMDL 19.2%
program:
$625 $187
% relative to TMDL 23.0%
program:
#PS w/savings
2067
23.2%
1251
40.2%
       Finally, Exhibit 11-13 shows the costs for point sources estimated under each of the three
scenarios.

                                          Exhibit 11-13
       Costs for Point Sources Under All Three TMDL Program Scenarios ($ in millions/yr)

Into + upstream case
Into only case
Least Flexible TMDL
Program
$2,178
$1,082
Moderately Cost-effective
TMDL Program
$1,634
25 % savings
$812
25 % savings
More Cost-Effective TMDL Program
$1,321
39 % savings
$625
42 % savings
       Relative to the "Least Flexible TMDL Program" scenario, having the Moderately Cost-effective
TMDL Program will save point source dischargers an estimated 25 %. A more cost-effective TMDL
program could save 39 - 42 %.
                                             11-27

-------
           III. IMPLEMENTATION COSTS FOR NONPOINT SOURCES

       The implementation costs that nonpoint sources might incur under the three TMDL program
scenarios are estimated for the following nonpoint source types:

       •       Agricultural land, including crop land, pasture land, and range land;
       •       Animal feeding operations (AFOs);
       •       Silviculture; and
       •       On-site wastewater treatment systems (including septic systems, cesspools, etc.)

       Our approach for estimating implementation costs for these nonpoint sources is similar to our
approach for point sources.  For each 303(d) water impaired by one of these nonpoint source types, we
identify the volume of nonpoint source activity contributing to the impairment, assume that some fraction
(depending on the scenario) of this nonpoint source activity will need to implement the "next treatment
step", and then estimate the costs associated with this amount of "next treatment step" controls. The Least
Flexible TMDL Program scenario and the Moderately Cost-effective TMDL Program scenario are
differentiated based on the degree to which actual TMDLs have required load reductions from nonpoint
sources relative to the load reductions that we simulate for the Least Flexible TMDL program scenario.
The Cost-Effective TMDL Program scenario adds the additional  costs for nonpoint source controls that we
estimated in the previous chapter as some point source control responsibilities are assumed shifted to
nonpoint sources.

A. COVERAGE OF THE NONPOINT SOURCE ANALYSIS

       There are many other nonpoint source types that contribute to impairment of 303 (d) waters for
which we were not able to estimate implementation costs, such as:

       •       abandoned mines,
       •       contaminated in-stream sediments,
       •       air deposition,
       •       natural sources,
       •       land disposal,
       •       many leaks  and spills,
       •       hydrological modification,
       •       habitat alteration,
       •       dispersed petroleum and other resource extraction activities; and
       •       CSOs, SSOs and storm water (urban, construction, industrial) to the extent they are not
               addressed by existing technology-based requirements.

       The nonpoint source categories that we do cover likely comprise the bulk of nonpoint source
pollutant loadings to impaired waters in total,  but there remain many waters that are impaired only by one
or more of these omitted categories for which we do not estimate  costs. For each of the omitted nonpoint
source categories, at least one of the data sets  required for estimating costs is not available:
                                             III- 1

-------
        •       Information on the volume and location of the nonpoint source activity, at sufficiently fine
                geographic resolution so that the amount of the nonpoint source activity contributing to
                impairment of a specific water body can be identified; and

        •       Information on the level of water pollution-abating practices that are currently applied by
                the nonpoint source activity, and on unit costs for the "next treatment step" for the activity.

        The following exhibit suggests the relative importance of the source types we cover in the
implementation cost analysis and those we omit.

                                            Exhibit III-l
                Coverage of Different Source Types in Implementation Cost Analysis
Source
Agriculture
Hydromodification/Habitat Alteration
Nonpoint Source (No Further
Information)
Municipal Point Sources
Resource Extraction
Other Source
Urban Runoff/Storm Sewers
Source Unknown
Silviculture
Construction
Septic Systems
Industrial Point Sources
Combined Sewer Overflow
Atmospheric Deposition
Marinas
Total36
303(d)
River Miles
108,284
48,319
30,223
23,136
22,971
20,821
18,732
16,206
10,350
9,805
7,009
6,319
3,918
1,696
378
328,168
303(d) Lake
Acres
(thousands)
2,548
1,879
1269
930
293
1,899
1,908
992
289
409
1,394
292
436
1,080
151
15,769
Now Covered
in Cost
Analysis
X


X




X

X
X




Pollution,
not
Pollutants

X














Will be
Mitigated by
Tech-Based
Standards






X


X


X



Omitted


scaled to

X
X

scaled to





X
X

Source: Tetra Tech, Inc. analysis of 303(d) data base, with sources of impairment aggregated into these 15 groups.

        The exhibit shows the different source types listed by States as responsible for impairment of
303(d) waters and the river miles and lake acres of such impaired waters. Hydromodification and habitat
36      The totals shown should not be interpreted as the total extent of impaired waters in the nation.  Only about
        half of the States report source information for their 303(d) lists. This causes the totals shown in the
        exhibit to fall far short of the national total. On the other hand, for many impaired waters States cite
        multiple sources of impairment, and this will tend to make the figures in the exhibit exceed the national
        total.  It is only by chance that the total river miles shown in the exhibit is close to the roughly 300,000
        mile aggregate length of all impaired river segments on the 1998 303(d) lists.
                                               Ill-2

-------
alteration often involve physical changes to water bodies or their drainage areas without explicit pollutants,
and these sources will often not be addressed by TMDLs. Most impairments due to urban runoff, storm
sewers, construction and CSOs will be mitigated by full compliance with recently adopted technology-
based standards for these sources, and implementation costs in these areas can be attributed to the baseline
rather than to the TMDL program. This leaves several important source categories as potentially affected
by the TMDL program but omitted from this implementation cost analysis:

        •       "Source Unknown" and "Nonpoint Source (No Further Information)" are categories that
                we cannot analyze presently because there is insufficient information available as to
                exactly what the responsible sources are. Our scaling process  described earlier (Section
                I.C.B.b), however, extrapolates costs to these waters by scaling from costs estimated for
                waters with known nonpoint source types. In effect, TMDL implementation costs are
                estimated for waters impaired by such sources.

        •       "Resource extraction" includes mostly impacts from mining and, to a lesser degree, oil and
                gas development. Some mining impacts will be addressed (and have  been costed in the
                point source analysis) through NPDES permits for mining discharges that involve discrete
                conveyances. Most impacts, though, involve abandoned mines or diffuse runoff and
                seepage from mined areas generally.  Implementation costs  will occur for these sources,
                but have not been estimated here.  These costs could be significant.37
37      Here are two estimates that suggest the magnitude of the costs that might be involved for control of some
        sorts of abandoned mines.

        The U.S. Forest Service estimates that there are approximately 38,000 abandoned mines on National
Forest lands, among which perhaps 10,000 pose important water quality problems.  The 1,800 sites that have toxic
discharges are being pursued under CERCLA and the point source program. The remaining 8,200 sites do not
involve toxic pollutants, but are thought to cause significant resource damage.  It is not known how many of these
sites affect impaired waters specifically. The average cost of addressing these sites is estimated at $150,000 -
$280,000 per site.  USFS estimated that it will require $60 - $115 million per year to resolve water quality
problems from these sites by the year 2022. (Note that these cost estimates are not directly comparable with the
levelized annual costs continuing forever that we estimate in this report.) Source: USDA review of EPA's draft
TMDL cost analyses, June/July, 2001.

        EPA's Draft Nonpoint Source Gap Analysis (Tetra Tech, Inc., February 7, 2001, op cit.) estimates the cost
to reclaim all coal-related abandoned mined lands at $5.7 billion. This estimate includes mines on all lands,
including Federal lands. It excludes costs for non-coal mines, which are thought to be less numerous and costly
than are coal mines. The estimate therefore overlaps to some degree with the USFS estimate. If this $5.7billion
cost were incurred over the period from 2005 through 2020 and then converted into a levelized annual amount
continuing forever, thus making these costs comparable to those estimated in this report, the annual cost would be
$173 million/yr.

        It is now known how many of either the USFS sites  or the coal-related sites affect 303(d)-listed impaired
waters specifically. For other nonpoint source types that we analyze, the fraction of all nonpoint source activity
that affects 303(d) waters specifically (under the Least Flexible Scenario) ranges from 1% (for on-site wastewater
treatment systems) to 53% (crop land).
                                               Ill-3

-------
        •       "Other sources" include a wide variety of activities that we did not break out individually,
               such as contaminated sediments, land disposal, and natural sources (e.g., salt springs, acid
               bogs).  Most, if not all, of these sources will be addressed by TMDLs and implementation
               costs will often occur. In some cases, the costs to implement appropriate treatment or
               remediation will be substantial (e.g., dredging and disposing of contaminated sediments).

        •       "Atmospheric deposition" is a difficult issue for the TMDL program. Hopefully most
               atmospheric deposition problems will be resolved as air pollution control programs
               continue  to progress (e.g., NESHAPs, NOx and SOx programs, controls on utility
               emissions of mercury, etc.). It is unclear how a TMDL developed by a State will be able
               to address atmospheric deposition originating from beyond the State's boundaries.  The
               first TMDL for atmospheric deposition was recently approved by EPA (for the Savannah
               River, GA), and it was projected to add no costs for sources beyond what they will need to
               spend to  meet existing air regulations.

        In sum, the source types completely omitted from the implementation cost analysis (excepting the
unknown and "nonpoint source (no further information)" categories that we scale to) account for about 14
% of the 303(d) river miles and 22 % of the 303(d) lake acres.  The important omitted sorts of nonpoint
sources that we would like to include in the analysis if the necessary data were available are:  resource
extraction (mining, petroleum activities), land disposal, and perhaps some sorts  of contaminated sediments
and natural sources. Developing methods and data for costing these omitted nonpoint source activities
should be high priority in future efforts to improve our estimates of TMDL implementation costs.

B. WATERS IMPAIRED BY AGRICULTURE, AFOS, SILVICULTURE AND ON-SITE
SYSTEMS

        When one of these nonpoint source types is cited by a State as a source of impairment for a 303(d)
water,  we assume that the State will require or induce this nonpoint source type to implement further
control measures, whether under the moderate and more cost-effective TMDL program scenarios
(Scenarios 2 and 3) or under the least flexible TMDL program (Scenario 1).  Our first step in estimating
the implementation costs for these nonpoint sources is to identify the 303 (d) waters that are impaired by
each of the nonpoint source types that we cover.

        States vary in the degree to which they provide information on the source types responsible for
impairment of impaired waters.  Some States do not report source information at all for their 303(d)
waters, some report in modest detail (e.g., using only broad identifiers of source types such as "point
sources", "nonpoint sources", "unknown sources", etc.), and some report in great detail (e.g., using specific
identifiers of source types such as "municipal point sources", "CSOs", "silviculture", "irrigated crop land",
etc.).  Our procedure for identifying water bodies impaired by various nonpoint  source categories relies on
detailed source reporting by those States that provide such information, either in their 303(d) submissions
or in their 305(b) submissions. For example, in analyzing silviculture, we:

        •       Identify the 303(d) water bodies that States report as impaired by silviculture as a source.

        •       Identify the 305 (b) water bodies that States report as impaired by silviculture as a source.
               We then  crosswalk from each of these silviculture-impaired 305 (b) waters and determine
               whether there is a corresponding 303(d) water body.  (Some States report 305(b) source

                                              III-4

-------
               information but not 303(d).  This second step effectively increases the set of States within
               which we can find silviculture-impaired 303(d) waters.)

        •       Add the results of the first and second steps, thus obtaining a list of silviculture-impaired
               303(d) waters.  We note the States in which these waters are located.  We then regard this
               set of States as a sample that we analyze. We assume that this sample is representative of
               the nation as a whole.

        •       States for which we can identify no silviculture-impaired water bodies may either:

               a)      Actually have no silviculture-impaired water bodies; or
               b)      Actually have them, but report source information in a manner that does not allow
                       for identifying them.

               Conservatively, we assume the latter - we assume that a State that reports  no silviculture-
               impaired water bodies is effectively "non-reporting". We extrapolate the costs we estimate
               for controlling silviculture in the sample States (those that report silviculture impairments)
               to the assumed non-reporting States by scaling up based on the volume of silviculture
               occurring in the sample  States relative to that occurring in the non-reporting States. The
               scaling factor we use for this extrapolation is the ratio between the total national annual
               timber harvest and the harvest volume in our "reporting" States. In this case,  10.91
               million cubic feet of timber was harvested in the 30 "reporting" States and  16.35 million
               cubic feet was harvested in the nation as a whole. The scaling factor is thus 16.35/10.91,
               or 1.499.

        We employ this approach for each of the different nonpoint source types we analyze; for crop land,
pasture land, range land, AFOs and on-site wastewater systems as well as for silviculture. Different sets of
States are considered to be "non-reporting" for the different nonpoint source types. For each nonpoint
source type, we use a different volume-based scaling factor to extrapolate our cost estimates from the
"reporting" States to the entire nation. The volume measures that we use in scaling are:

        •       For silviculture.  The annual volume of timber harvested.

        •       For agriculture.  Crop land-related costs are extrapolated based on the acreage of crop
               land.  Similarly, pasture-related costs and range-related costs are extrapolated based on the
               acreage of pasture and range lands.

        •       For AFOs.  The number of confined animal units (AUs).

        •       For on-site  wastewater  systems. The number of dwelling units served by on-site
               wastewater  systems.

        The exhibit below shows the States that reported (in either 303(d) or 305(b) reporting) 303(d)
water bodies impaired by each nonpoint  source type that we analyze. At the bottom of the exhibit we show
the fraction of the total national volume of activity occurring in the "reporting" States, and the resulting
scaling factor that we apply in extrapolating TMDL implementation  costs from the "reporting" States  to
the nation as a whole.

                                              Ill-5

-------
                       Exhibit III-2
States Using NFS Categories in Reporting Sources of Impairment
STATE
Alabama
Alaska
Arizona
Arkansas
California
Colorado
Connecticut
Delaware
District of Columbia
Florida
Georgia
Hawaii
Idaho
Illinois
Indiana
Iowa
Kansas
Kentucky
Louisiana
Maine
Maryland
Massachusetts
Michigan
Minnesota
Mississippi
Missouri
Montana
Nebraska
Nevada
New Hampshire
New Jersey
New Mexico
New York
North Carolina
North Dakota
Ohio
Oklahoma
Oregon
Pennsylvania
Rhode Island
South Carolina
South Dakota
Tennessee
Texas
Utah
Vermont
Virginia
Washington
West Virginia
Wisconsin
Wyoming
Number of States using this category
Fraction of total in "reporting" States
Scale Factor
Crop Land
X

X

X








X


X
X
X
X


X

X
X
X
X



X

X
X
X
X




X
X
X

X


X


23
.64
1.56
Pasture or Range
X

X

X








X



X
X
X


X

X

X
X
X


X

X
X
X
X


X

X
X
X

X
X

X
X

25
.59 .70
1.69 1.43
AFOs
X



X

X
X





X

X
X
X
X
X


X

X
X
X
X

X
X
X

X
X
X
X


X
X
X
X
X

X
X

X
X
X
32
.65
1.55
Silviculture
X
X
X
X
X




X



X



X
X
X

X
X
X
X

X




X
X
X
X
X
X

X
X
X
X
X


X
X

X

X
30
.67
1.499
On-Site Wastewater
X

X
X


X
X





X



X
X


X
X

X

X




X
X
X

X
X

X


X
X
X

X
X

X


24
.58
1.732
                          III-6

-------
        The exhibit shows that the number of States citing (in either their 303(d) or 305(b) lists) these
nonpoint source types as responsible for impairment of waters ranged from 23 (for crop land) to 32 (for
AFOs). We conservatively assumed that any State not citing one of these nonpoint source types as
responsible for impairments actually had such impairments, but simply did not report sources at all or did
not use an appropriate code in its system of reporting sources of impairments. The States that we classified
as "reporting" accounted for a percentage of the total national volume of activity ranging from 58 % for
on-site wastewater treatment systems land to 70 % for range land.

        We noted each water body that was cited by a "reporting" State as impaired by one of the nonpoint
source categories that we address. This set of water bodies impaired by each nonpoint  source category was
the starting point for the cost analysis.  After estimating the costs that nonpoint sources affecting these
water bodies in "reporting" States would incur, we scaled up these implementation costs to extend the cost
estimate from the "reporting" States to the entire nation.

C. THE AMOUNT OF NONPOINT SOURCE ACTIVITY NEEDING FURTHER CONTROL
FOR AN IMPAIRED WATER BODY

        When a State identifies a water body as impaired by a nonpoint source type, we assume in the
"Least Flexible TMDL Program" scenario that the State will require further control for the entire volume
of that nonpoint source activity that occurs within the county or counties in which the impaired water body
is located. Thus, for example, in addressing an AFO-impaired water body, we assume  in this scenario that
the State would require or  somehow elicit load reductions from all the AFOs in the county  or counties
surrounding the impaired water body.

        This assumption is based on the general belief that most nonpoint source management programs
are targeted at a finer geographic level than an entire county. States, USDA and others typically direct
nonpoint source management and assistance efforts at particular watersheds with water quality problems.
Practices are eligible for cost-share in some watersheds but not others, technical assistance is focused on
some watersheds and not others, etc.. Commonly the watershed is defined at the scale of roughly an 11-
digit catalog unit (substantially smaller than a county),  but sometimes the watershed is  defined at a more
gross scale approximating  an 8-digit catalog unit (approximately similar in size to a county).38 In
reviewing a draft of this report, USDA staff believed that the "entire county" assumption represented a
substantial overestimate of what typically occurs in practice: "While technically feasible, it is highly
unlikely that Section 319 efforts will cover 100 percent of the county.  This is not the level of activity being
observed.."39

        We believe the "entire county" assumption underlying the "Least Flexible TMDL Program"
scenario represents for most impaired water bodies a substantial overestimate of the volume of nonpoint
source activity that will actually be addressed in a TMDL as needing further controls.   Most of the
identified nonpoint source-impaired water bodies are relatively small (typical length of the impaired river
segments is 5 to 25 miles). For most impaired water bodies, the upstream watershed from which the
impairment pollutants derive is much smaller than the size of the entire county or counties within which the
water body is located. The assumption that the typical TMDL will seek nonpoint source load reductions
38      There are roughly 2200 8-digit watersheds (USGS hydrologic unit codes) in the U.S., 3000 counties, and
        XXX 11-digit watersheds.

39      USDA staff, personal communication, June 22, 2001.

                                              Ill-7

-------
from an area as large as the entire county or counties surrounding the impaired water seems likely to
generate something close to a  worst case.

        Ideally we might like to delineate the watershed boundaries for each impaired water body and
determine what volume of relevant nonpoint source activity occurs within each exact watershed.  This was
not possible - both because nonpoint source data is not available at the watershed or sub-county level, and
also because the watershed boundaries corresponding to 303(d) water bodies have not yet been identified at
a sufficiently fine level of detail.40

        We have generated some limited information on the relationship in practice between the size of the
geographic area from which actual TMDLs have required nonpoint source load reductions and the size of
the county or counties surrounding a water body. Among the sample of 15  TMDLs we reviewed for
"ground-truthing" purposes (see Appendix A), 13 required load reductions from nonpoint sources.  In 7 of
the 13 cases, the acreage needing controls is less than 10% as large as the acreage of the county or counties
in which the water bodies are located.  In 2 more cases, the acreage needing control is 10 - 50% as large as
the acreage of the surrounding counties.  In 2 cases, the acreage needing control is greater than the size of
the surrounding  counties, but both of these "TMDLs" is actually a single submission covering multiple
impaired waters. Across our "ground-truthing"  sample, the median acreage needing control is somewhat
less than 10% as large as the acreage of the county(s) surrounding the water body.41

        In Appendix A, we rank order the 13 TMDLs that require nonpoint source load reductions
according to the size of the watershed area from which nonpoint source load reductions are required.  The
25th percentile of this distribution is 3.2 % of the surrounding county (i.e., 25  % of this sample of TMDLs
require nonpoint source controls for a watershed area with a number of acres  less than 3.2% of the number
of acres in the surrounding county).  The 75th percentile of this distribution  is  32.8  % of the number of
acres in the surrounding county (i.e., 75 % of this sample of TMDLs require nonpoint source controls for a
geographic area smaller than 32.8 % of the surrounding county).  We use these 25th and 75th percentile
figures to establish a rough range in estimating the size of the area for which a typical TMDL will require
nonpoint source controls.42 In simulating the costs of the Scenario 2 (Moderately Cost-
40      It is currently possible to locate the boundaries of the 8-digit watershed within which each impaired water
        is located, but for most impaired waters this will include very large amounts of land that drain into waters
        downstream of the impaired water itself, and which thus will not be implicated in the TMDL for the
        impaired water.  At present, it is not possible to locate the 11-digit watershed within which each impaired
        water is located. This level of detail will be necessary before watershed areas corresponding closely to
        impaired waters can be identified on a nationwide basis using a GIS.

41      As noted previously, we are expanding this sample of completed TMDLs that we analyze, and we will
        reflect the results from the expanded sample in the final report.

42      We seek comment on this approach. Although a range established by the 25th and 75th percentiles would
        seem to encompass the typical TMDL, our sample of 15 TMDLs includes an outlier on the high end.  The
        TMDL for the Neuse River Estuary will require nonpoint source controls from a very large watershed
        encompassing most of 19 counties. (This TMDL will actually address another impaired water body
        within the watershed, so this TMDL seeks nonpoint source controls from nearly 9.5 counties per impaired
        water body.)  This large geographic area for this one TMDL skews the distribution such that the mean
        (not median) area addressed per impaired water body across our sample is nearly 86 % of a county. For
        this sample of TMDLs, the mean area addressed (86 % of a county) is substantially greater than the
        median and greater than even the 75th percentile. This sort of pattern that we see in this sample of 15
        TMDLs may be typical for all TMDLs. We believe it likely that the great majority of impaired waters

                                               III-8

-------
effective TMDL Program), we assume that the typical TMDL will require controls for acreage equal to 5%
(lower estimate) to 40 % (upper estimate) of the number of acres in the county or counties within which the
impaired water body is located.

        In sum, for the "Least Flexible TMDL Program" scenario (Scenario 1), we assumed that load
reductions would be required for all of the relevant nonpoint source activity that occurs within the entire
county or counties within which the nonpoint source-impaired water body is located. We also assumed that
it makes no difference how many impaired water bodies there are in  single county. One impaired water
body is  sufficient to trigger the need for controls for all of the relevant nonpoint source type activity in the
county;  but once all the nonpoint source activity in the county is controlled we assume this will be
sufficient to address the nonpoint source contribution to impairment for all the impaired water bodies in the
county.

        For the Moderately Cost-effective TMDL Program scenario, we assumed instead (based on the
results from our sample of 15 actual TMDLs) that load reductions would be required for a geographic area
ranging in acreage from 5  % to 40 % of the number of acres in the surrounding county or counties. Absent
any data on how the nonpoint source activity is distributed within a county, we assumed that 5 % to 40 %
of the area of the county would comprise 5 % to 40 % of the relevant sort of nonpoint source activity
within the county.  We assumed for the  Moderately Cost-effective TMDL Program scenario that this need
for nonpoint source controls was cumulative with multiple impaired water bodies within a county until 100
% of the county's nonpoint source activity had been controlled.

        The following exhibit provides numerical examples demonstrating how much nonpoint source
activity we assumed would need control under the two scenarios.  For the purpose of these examples, the
nonpoint source type under consideration is silviculture.

                                           Exhibit III-3
                      Examples Showing Amount of Nonpoint Source Activity
                           Needing Further Control for Scenarios 1 and 2
# of Silviculture-Impaired
Water Bodies in the
County
1
2
3
4
Amount of Silviculture Needing Further Control in the County
Least Flexible TMDL Program (Scenario
1)
100 % of silviculture in the county
100 % of silviculture in the county
100 % of silviculture in the county
100 % of silviculture in the county
Moderately Cost-effective TMDL Program
(Scenario 2)
5 - 40 % of silviculture in the county
10-80 % of silviculture in the county
15 - 100 % of silviculture in the county
20-100 % of silviculture in the county
        (and the great majority of TMDLs) are for streams or lakes with small to modest watersheds.  However, a
        few impaired waters may be major estuaries or large lakes whose eventual TMDLs will require reduction
        of nonpoint source loads from very large upstream watersheds. Our procedure looking at the 25th and 75th
        percentiles of a range ignores the impact of such rare TMDLs requiring controls for large areas.
        However, these rare TMDLs may, in fact, account for an important fraction of the total land area for
        which controls will ultimately be required across all TMDLs.  Perhaps we should develop a cost
        estimation procedure that explicitly accounts for the impact of these outliers.  We welcome suggestions.
                                              Ill-9

-------
        We thus employed the following procedure to determine the volume of nonpoint source activity that
will likely be required to implement further controls under the three TMDL program scenarios that we
analyze:

        •       The Research Triangle Institute (RTI) identified all 303(d) waters that States had cited (in
               either their 303(d) or 305(b) submissions) as impaired by any of the nonpoint source types
               that we cover in this analysis;

        •       RTI provided us with further information on each of these nonpoint source-impaired water
               bodies, including the county or counties in which it is located, its size, and its type (e.g.,
               lake, river, estuary, coastal shoreline); and

        •       Environomics and Tetra Tech  extracted information from various national data bases on
               the volume of nonpoint source activities in each of the counties with nonpoint source-
               impaired waters.

        County-level nonpoint source information was extracted from the most current national data bases
relevant to each nonpoint source type. The information that was obtained included data on the volume of
the nonpoint source activity occurring in each impairment county and further data on characteristics of that
activity in each county that affect the potential  costs of management measures for the nonpoint sources.
We used the following data bases:

        •       For agricultural land (crop, pasture and range), the 1997 National Resources Inventory
               (NRI)43 as it existed prior to some corrections released in early 2001.  Data obtained from
               NRI included the acreage of crop, pasture and range land in each county, the average slope
               of such land, and the acreage of crop land eroding at greater than 15 tons/acre/year.

        •       For AFOs, the 1997 Census of Agriculture.44  From the Census of Agriculture, we
               obtained information by county on the number of farms with confined beef, dairy cows,
               swine, broilers and layers, and on the size distribution of these farms.

        •       For silviculture, the Timber Product Output Data File.45 This provides information on the
               amount of timber harvested and on characteristics of the land from which it was harvested
               (e.g., slope).  Regional timber  yield estimates were applied to convert the harvest data to
               county-by-county estimates of the timber acreage harvested and the acreage undergoing
               artificial regeneration.
43      USDA. December 1999 (unrevised version). 1997 National Resources Inventory.  Natural Resources
        Conservation Service.  http://www.nhg.nrcs.usda.gov/NRI/1997.

44      USDA. 1997. 7997 Census of Agriculture. National Agricultural Statistics Service.
        http://www.nass.usda.gov/census.

45      USDA Forest Service.  Timber Product Output Datafile. Timber Product Output (TPO) Database Retrieval
        System, http://srsfia.usfs.msstate.edu/rpa/tpo/. The TPO provides timber harvest data that is updated at
        different times for different States.  At any point in time, the data in TPO derive from different years for
        different States.

                                              Ill - 10

-------
        •       For on-site wastewater systems, the 1992 Census of Housing (USDOC cite...).  This
               provides information on the number of dwelling units served by on-site wastewater
               systems.

        As noted, for most nonpoint source BMPs, we assumed under Scenarios 1 and 2 that differing
fractions of the nonpoint source activity in a county would be required to reduce their loads and implement
the next treatment step. However, for those BMPs that are intended to be applied specifically in riparian
zones -- for example streamside buffer strips to arrest runoff and trap sediment, or fencing to keep grazing
animals out of streams -- we assumed they would be applied in a band alongside impaired waters rather
than throughout some percentage of the surrounding county. For these riparian BMPs, we did not vary the
amount of the BMP needed with the scenario (i.e., we assumed the same amount of riparian BMPs would
be necessary under the Moderately Cost-effective TMDL Program scenario as for the Least Flexible
TMDL Program scenario.  The riparian BMPs include:

        •       For crop land: a riparian forest buffer comprising a 75-foot wide corridor on each side of
               every crop-impaired water body.

        •       For pasture land: a similar 50-foot wide riparian forest buffer on each  side of every
               pasture-impaired water body.46

        •       For range land: fencing along the entire bank length of every range-impaired water body
               or, alternatively, conservation practices encompassing "stream protection"  and/or
               "streambank stabilization" applied to a riparian zone of 100 feet on each bank of every
               range-impaired water body.

        •       For AFOs: a riparian forest buffer similar to that for crop land, on each side of every
               AFO-impaired water body.

        •       For on-site wastewater systems: a zone extending for 100 yards from the shoreline of every
               septic system-impaired water body within which all malfunctioning septic systems must be
               rehabilitated.

        For all of these riparian BMPs, the key data elements we used in estimating the volume of nonpoint
source activities that will need to implement the BMP include: the length of the impaired water body and
the width of the riparian zone within which the BMP is to be applied.

D. BMPS FOR REDUCING NONPOINT SOURCE LOADS

        We assume that the nonpoint sources addressed under either the Least Flexible TMDL Program
Scenario or the Moderately Cost-effective TMDL Program will be required to achieve load reductions  that
can be accomplished by implementing a set of basic BMPs specific to each source type. For each nonpoint
source type, we chose a set of basic BMPs that would constitute the "next treatment step" for cost-
estimating purposes for that nonpoint source type.  We did this in two steps.  First, we chose a set of broad
practice groups that we assumed the nonpoint source type would need to implement.  For crop land, for
example, we assumed that farms contributing to water quality impairments would implement five groups of
practices:

        1.      Conservation tillage;
46     USDA recommends a wider buffer strip when adjoining crop land than when adjoining pasture land.

                                             Ill- 11

-------
        2.      Nutrient management;
        3.      Practices to reduce sediment transport within or at the edge of the field;
        4.      Practices to protect and restore riparian areas; and
        5.      Management of highly erosive crop land.

For each chosen practice group, we then selected a single BMP that we would use to represent what the
costs of implementing the practice group might be. Each practice group includes a range of possible BMPs
(e.g., in-field or edge-of-field measures to reduce sediment transport can include many different contouring,
buffer and runoff management measures). For costing purposes, though, we chose a single, relatively
expensive BMP to represent what it might cost to implement whatever specific BMP among the practice
group is appropriate in each individual circumstance.  To represent costs for riparian practice groups, for
example, we chose riparian forest buffers, a BMP which is generally more expensive than other sorts of
buffer, filter strip or stream protection measures.  To represent costs for managing highly erosive crop
land, we chose retirement with establishment of permanent vegetative cover, a BMP that again is generally
more costly than alternatives such as conservation strip cropping, contouring, etc..

        We believe our selected BMPs yield a conservative cost estimate in two respects:

        •       We have chosen to simulate the implementation of each of a broad set of several practice
               groups, even though many impairment situations may not require all of the practice
               groups. In most circumstances, the entire package of practices that we assume and then
               estimate costs for will be more than enough to achieve the desired load reduction from the
               nonpoint source.

        •       For costing purposes, we represent each practice group with a relatively expensive specific
               BMP.

        Note that we have selected these practice  groups and specific BMPs only for the purpose of
estimating costs. We do not mean to imply that our particular selected measures should or must always be
implemented by nonpoint sources in order to mitigate water quality problems.  The selection of appropriate
BMPs in practice must be highly site-specific.  One or more of our particular selected BMPs may be poor
choices in many circumstances.  Riparian forest buffers, for example, are not cost-effective for areas where
trees are very difficult to plant and grow.

        In sum, the set of practice groups we chose are intended to represent the initial, broadly applicable
measures that each nonpoint source type can be expected to implement to minimize impacts on water
quality.  Relatively more expensive BMPs were selected to represent the costs  of each practice group.  Our
selected practice groups and BMPs are broadly consistent with guidance published by EPA (national
nonpoint source management measures developed pursuant to CZARA and subsequently updated) and
USDA (National Handbook of Conservation Practices).  The BMPs chosen for agriculture and AFOs
derive largely from discussions with EPA and USDA experts.47 USDA does not, however, in any way
endorse the specific selection of BMPs,  or,  more broadly, the assumptions or findings of this report. In
fact, USDA staff have suggested many changes or improvements in this analysis that we have not yet
accomplished.
47      Personal communications with: Clay Ogg (EPA, 9/00, 10/00), Matt Moore (ARS, 10/00), Seth Dabney
        (ARS, 10/00), Jim Fouss (ARS, 10/00), Glen Weesies (NRCS, 11/00), and Wayne Skaggs (North Carolina
        State University, 11/00). Also, comments from numerous USDA staff during USDA's review of the draft
        TMDL cost studies in June and July, 2001.
                                             Ill - 12

-------
        The basic sets of BMPs that we chose for each nonpoint source type are as follows:
                                                Exhibit III-4
                         Basic BMPs Assumed for Each Nonpoint Source Type
 Nonpoint Source
      Type
      Practice Group and BMP
                       Purpose
Crop land
1. Conservation tillage

2. Nutrient management planning
3. In-field, edge-of-field measures:
vegetative barrier
4. Riparian measures: riparian forest
buffers
5. Management of erosive land:
retirement and cover
1. Reduce sheet and rill erosion and sediment transport
from tilled land
2. Avoid over-application of nutrients
3. Prevent concentrated (gully) erosion in upland fields

4. Remove sediment and nutrients from runoff before
reaching water bodies
5. Avoid disturbing highly erosive land
Pasture land
1. Riparian measures: riparian forest
buffers
2. Upland measures: rotational stocking
1. Remove sediment and nutrients from runoff before
reaching water bodies
2. Rotate stock sequentially through fenced pastures to
prevent overgrazing and erosion of any one area
Range land
Riparian measures:
1. Use exclusion
  or
1A. Stream protection/bank stabilization
                                                         1.  Fence stock away from watersides to avoid damage to
                                                         stream banks and bottoms, avoid manure in water
                                                         1 A. Repair and prevent further damage by animals to
                                                         riparian zone
AFOs
1. On-farm manure collection and
management (facilities and equipment
to collect, store, manage and use
manure on nearby land, control runoff,
and compost dead animals)
2. Manure transport from nutrient-
surplus to nutrient-deficient areas

3. Nutrient management planning
(plans, soil and manure testing, training,
record-keeping)
1. Prevent nutrients, pathogens and BOD from reaching
waters through runoff, spills from lagoons, and groundwater
                                                         2. Avoid over- application of manure nutrients relative to
                                                         local crop needs, with resulting buildup of N and P and
                                                         release to local waters
                                                         3. Avoid over-application of nutrients (both chemical
                                                         fertilizers and manure)
Silviculture
Various BMPs addressing:
 1) Preharvest planning;
 2) Streamside management areas;
 3) Road construction/reconstruction;
 4) Road management;
 5) Timber harvesting;
 6) Site preparation and forest
regeneration;
 7) Fire management;
 8) Revegetation of disturbed areas;
 9) Chemical management;
10) Wetlands  forest management
Minimize erosion, siltation, and bank destabilization. Avoid
loss of vegetative cover and increase in water temperature.
Avoid slash and other material in streams. Avoid runoff of
forest management chemicals
On-site wastewater
systems
Rehabilitate all failing septic systems
within a wide riparian zone
Prevent nutrients, pathogens and BOD from reaching water
bodies through runoff and infiltration
                                                   III -  13

-------
These BMPs are assumed to be applied to:

        •       For the Least Flexible TMDL Program scenario, the entire volume of the nonpoint source
               activity that occurs in the county or counties within which the nonpoint source-impaired
               water body is located; or

        •       For the Moderately Cost-effective TMDL Program scenario, per impaired water body in
               the county, 5 - 40 % of the volume of the nonpoint source activity that occurs in the
               county; and

        •       (For riparian BMPs) the entire riparian zone surrounding the nonpoint source-impaired
               water body.

E. UNIT COSTS FOR THESE BMPS

        Many sources of unit cost information are used in estimating the cost of implementing these
nonpoint source BMPs as a function of the volume and characteristics of the nonpoint sources. The costing
relationships and the underlying sources are described fully in Appendix I. The more important cost
references we used include:

        •       For all nonpoint source categories.  EPA's recent " "Draft Nonpoint Source Gap
               Analysis".48

        •       For agriculture (crop, pasture and range land).  Numerous studies supported by EPA
               and USDA on costs or savings from various BMPs.  Information summarized in EPA's
               recent Draft National Management Measures to Control Nonpoint Source Pollution from
               Agriculture (cite..). Information provided by USDA staff in the course of the USDA
               review during June and July, 2001, of EPA's draft TMDL cost studies.49

        •       For AFOs. Costing studies supporting EPA's promulgation of recommended management
               measures for AFOs in the "Management Measures Guidance for Coastal Zone Nonpoint
               Source Pollution" required by the Coastal Zone Management Act Reauthorization
               Amendments of 1990.50  Also, to a much lesser degree, some information has been used
               from preliminary costing studies supporting EPA's proposed new effluent guidelines for
               feedlots.51 Note here that EPA's cost studies supporting the effluent guidelines are much
48     Tetra Tech, Inc. Draft Nonpoint Source Gap Analysis.  Drafts of July, 2000; February 7, 2001. Some of
       the unit cost information used in developing the cost estimates in this report will be updated and revised
       for the final report to bring it into better conformity with unit cost information cited in the most recent
       draft of the Nonpoint Source Gap Analysis.

49     USDA. "Comments on EPA documents 'Cost of Restoring the Nation's Impaired Waters' dated June 13,
       2001 and 'Total National Costs for Pollutant Sources to Implement TMDLs' dated June 13, 2001". June
       22, 2001

50     DPRA, Incorporated. Economic Impact Analysis of National Nonpoint Source Management Measures
       Affecting Confined Animal Facilities. May 17, 1995.

51     U.S. EPA, Office of Water, Office of Science and Technology.  Final Cost Methodology Report for Beef
       and Dairy AFOs.  Final Cost Methodology Report for Swine and Poultry Sectors.  January, 2001.

                                             Ill -  14

-------
               newer and more comprehensive than EPA's CZARA costing effort, and they differ from
               the CZARA cost estimates in some important respects. However, EPA's feedlot effluent
               guideline cost studies are currently being revised, and we did not believe that we could
               effectively use them while they were still in flux.  As a result, our AFO cost estimates in
               this report rely on unit cost information that we know in some instances is outdated and
               has been supreseded. The AFO cost estimates in this report are inconsistent in important
               ways with EPA's AFO cost estimates published thus far in support of the feedlots effluent
               guidelines.  We intend to conform our estimates to the effluent guideline  cost information
               before this report is finalized. Finally, important information relevant to needs for
               transporting manure from AFOs was obtained from a series of USDA studies  by Lander,
               et al.52

        •       For silviculture. Costing studies supporting  EPA's promulgation of the management
               measures for silviculture under CZARA.53

        •       For on-site wastewater systems. Information drawn from the National Census of
               Housing54 on the proportion of septic systems found to be failing was combined with
               information from EPA's "nonpoint source gap analysis" on the costs of rehabilitating a
               failed septic system.

These references provided both information on the unit costs of implementing the chosen  BMPs for these
nonpoint source categories and information on the extent to which these BMPs have already been
implemented.

        In some cases, BMPs anticipated to be adopted by nonpoint sources in order to meet load reduction
targets may yield savings that partly or perhaps even completely offset the costs  of the BMPs.  BMPs that
likely involve such cost savings include:

        •       Nutrient management planning for crop farmers and AFOs.  Nutrient management
               planning includes as a major objective balancing the amount of nutrients applied to crops
               to crop needs. Most farmers over apply nutrients relative to crop needs; better planning
               can save the cost of the no longer applied excess nutrients.  Many studies have found that
               the costs of developing nutrient management  plans (including costs for soil and manure
               testing, training, and record-keeping) are outweighed by the cost savings that result as less
               chemical fertilizer is purchased and applied.  Few studies, though, provide a
               comprehensive before/after comparison that also considers effects on crop yield and
               overall farm profitability.
52      Lander et al. Manure Nutrients Relative to the Capacity of Cropland and Pastureland to Assimilate
        Nutrients: Spatial and Temporal Trends for the United States. USDA. Publication No. psOO-0579. 2000.
        Also, Lander et al. Nutrients Available from Livestock Manure Relative to Crop Growth Requirements.
        USDA. Resource Assessment and Strategic Planning Working Paper 98-1.1998.

53      Research Triangle Institute. Economic Analysis of Coastal Nonpoint Source Pollution Controls:
        Forestry.  1992.

54      U.S. Census Bureau. American Housing Survey for the United States. 1997.
        Www.census.gov/prod/99pubs/hl50-97.pdf

                                             III - 15

-------
        •        Conservation tillage.  Conservation tillage is widely agreed to reduce costs for labor, fuel
                and equipment maintenance as less tillage is performed. On the other hand, conservation
                tillage involves purchase of new equipment and likely increased costs for herbicides and
                seed. It is also perceived as being more risky. Effects on yields are variable.55

        •        Transporting manure from a nutrient-surplus area to a nutrient-deficient area.. Manure
                increases in value when it is shipped from a location where manure nutrients are in excess
                to a location where the manure's full nutrient content is needed. The increase in value is
                likely not sufficient to pay for the shipping cost (otherwise one would expect that market
                forces would have already prompted the shipping to occur). Nevertheless, there will likely
                be some increase in value that will partly offset the shipping cost.

        •        Repairing a  malfunctioning septic system. In many cases, repairing a poorly functioning
                septic system forestalls more expensive responses that will be needed if the system is
                allowed to deteriorate to the point where it fails completely.

        It is very difficult to project what these sorts of savings might amount to if these BMPs were
applied on a widespread basis, as we simulate in this analysis.  There are both conceptual and empirical
difficulties:

        •        In theory, farmers, livestock operators and homeowners as economically rational actors
                would adopt these BMPs in the circumstances in which the BMPs would pay for
                themselves.  These individuals do not need a TMDL to induce them to act in their self-
                interest. We might reasonably assume that, in all instances where these BMPs have not
                yet been adopted, the cost savings  would not be large enough to offset the costs.

        •        On the  other hand, lack of information and adherence to traditional ways are significant
                barriers to adoption of newer BMPs, even in instances where they would appear to be
                profitable.

        •        On the  other hand again, farmers and others are risk averse.  Over application of nutrients,
                for example, may cost a little more than is necessary but it also provides a relatively
                inexpensive sort of insurance against disasters that could occur as a result of weather,
                pests, or miscalculations.

        •        Analytically, the cost savings from adoption of these BMPs tend to have been studied in
                the circumstances for which the BMPs are best suited.  It is questionable whether the
                savings found in instances where the BMPs have been adopted can be extrapolated to other
                settings where the BMPs have not yet been adopted.  One might expect the circumstances
55      USDA concludes in one analysis: "Generally speaking, no-till systems offer a slight to fairly significant
        reduction in input costs.  If proper management of conservation tillage is used, yields will most likely be
        maintained and costs will decrease.  An overall improvement in the efficiency of a farm operation will
        result and thus enhance profitability. In areas where moisture retention  is improved and soil productivity
        rises, yield increases can be expected together with improved profits." USDA,  Natural Resource
        Conservation Service. CORE4 Conservation Practices, the Common Sense Approach to Natural
        Resource Conservation.  1999.

                                               Ill -  16

-------
               in which the BMPs have been adopted to differ systematically from the circumstances
               where they have not yet been adopted, and extrapolation may be inappropriate.56

        In view of these very large uncertainties in estimating the national savings that might result from
widespread implementation of TMDL-prompted BMPs, we decided to treat the potential savings from the
BMPs in a different manner from the costs. We explicitly estimate the costs of the BMPs and display them
as our estimate of TMDL implementation costs. We also develop some very rough quantified estimates of
the potential cost savings from the BMPs, but, — in an effort both to be conservative and to recognize the
much greater uncertainty of the savings estimates -- we choose not to display these savings estimates and
not to net them out in the tables summarizing TMDL implementation costs. Note that this decision not to
display the savings estimates does not mean that we believe the potential savings to be unimportant or
nonexistent. To the contrary, we  believe that in many circumstances these BMPs will engender substantial
savings  that offset some portion of the BMP costs.  We are unable, however, to  estimate these savings with
much confidence. For  the final version of this  analysis, we intend to gather additional data that will allow
us to narrow the range of uncertainty in our national savings estimates.

        Unit costs and cost savings drawn from these references were applied to the volume of nonpoint
source activity assumed to need further controls as estimated in the previous step.  To the extent that some
portion of the nonpoint sources have already implemented some of the BMPs, the implementation cost and
savings  estimates were reduced to reflect the practices that are already in place.

F. SUMMARY COST ESTIMATES FOR NONPOINT SOURCES

        For each type of nonpoint source activity, we sum the costs across impaired water bodies and scale
appropriately to reflect missing data and obtain a nationwide estimate.

        We have previously discussed the three scaling steps involved in developing a national
implementation cost estimate for nonpoint sources:

        •       Scaling to reflect the portion of the national total of each nonpoint source activity that
               occurs in States that report on impairment of their waters by the particular nonpoint source
               activity. (Different scaling factor for each nonpoint source type.)

        •       Scaling to reflect the likelihood that the nonpoint source types we analyze will eventually
               be found responsible for a  share of the waters reported by States as impaired by "unknown
               sources" or by "nonpoint sources (no further information)". (Scaling factor of 1.13.)
56      For example, in the "CORE4" study cited above, USDA estimates for a sample farm that adopting a no-till
        system plus nutrient management reduces net costs for corn and wheat but increases costs for soybeans.
        Considering positive yield changes for all three crops, the total impact of adopting no-till and nutrient
        management appears to be positive for all three crops.

        Should we use this information as a basis for estimating the cost savings that might occur if no-till and
nutrient management were adopted more generally as a result of TMDLs? We think not. Recent figures indicate
that conservation tillage is used for 33% of all wheat (small grains) acreage, 54% of soybean acreage, and 40% of
corn acreage.  There must be reasons why farmers have not adopted conservation tillage for most of these three
crops' acreage, and we are wary of projecting that acreage that doesn't use this measure will benefit from it to the
same degree as the acreage that already does use it.  Furthermore, what should we surmise about the likely impact
of conservation tillage on other crops, particularly ones where it is currently used much less than for these three
crops (e.g., cotton, for which conservation tillage is used on only 12% of the acreage)?

                                              Ill -  17

-------
•      Scaling to reflect the assumed pace of TMDL development for the 1998 303(d)-listed
       waters and the assumed 5-year lag between TMDL development and when affected
       sources begin to incur implementation costs. (Scaling factor of .4484.)

The estimated national nonpoint source implementation costs are as follows:

                                 Exhibit III-5a
     Implementation Costs for Nonpoint Sources -- Crop Land ($ in millions/yr)
Number of States analyzed
Number of counties in these States with crop-impaired waters
Fraction of crop acreage in these States that is in impairment counties
Scale factor from these States to the nation

National total annualized implementation costs (millions of 2000 dollars/yr):
Conservation tillage
Nutrient management planning
Riparian forest buffers
Vegetative barriers
Retirement of highly erosive crop land
Total costs

National total potential savings from implementing these BMPs (millions of 2000
dollars/yr):
Conservation tillage
Nutrient management planning
Total potential savings
23
710
.53
1.56
Scenario 1
Least Flexible
TMDL
85 - 785
317-781
41 -104
49-108
154-177

645-1956
Scenario 1
Least Flexible
TMDL
0-414
0-804
0-1218
Scenario 2
Reas. TMDL
19-644
70 - 641
41 -104
12-88
42-154
183-1632

Scenario 2
Reas. TMDL
0-340
0-660
0-999
                                    III - 18

-------
                            Exhibit III-5b
Implementation Costs for Nonpoint Sources -Pasture Land ($ in millions/yr)
Number of States analyzed
# of counties with pasture and/or rangeland-impaired waters in these
States
Fraction of pasture acreage in these States that is in impairment
counties
Scale factor from these States to the nation
National total annualized implementation costs (millions of 2000
dollars/yr):
Riparian forest buffers
Rotational stocking
Total costs
25
511
.363
1.69
Scenario 1
Least Flexible TMDL
5.0-10.7
0
5.0-10.7
Scenario 2
Reas. TMDL
5.0-10.7
0
5.0-10.7
                            Exhibit III-5c
 Implementation Costs for Nonpoint Sources -Range Land ($ in millions/yr)
Number of States analyzed
# of counties with pasture and/or rangeland-impaired waters in these
States
Fraction of rangeland acreage in these States that is in impairment
counties
Scale factor from these States to the nation

National total annualized implementation costs (millions of 2000 dollars/yr):
Riparian fencing only; OR
Stream protection/bank stabilization
Total costs

25
511
.459
1.43
Scenario 1
Least Flexible
TMDL
5.1 -16.4
2.3-5.1

2.3-16.4
Scenario 2
Reas. TMDL
5.1 -16.4
2.3-5.1
2.3-16.4

                               III - 19

-------
                                            Exhibit III-5d
               Implementation Costs for Nonpoint Sources -- AFOs ($ in millions/yr)
Number of States analyzed
                  32
Number of counties with AFO-impaired waters in these States
                  451
Fraction of animal units in these States that is in impairment counties
                  .31
Scale factor from these States to the nation
                  1.55
National total annualized implementation costs (millions of 2000
dollars/yr):

   Riparian forest buffers
   Facilities upgrades
   Additional manure hauling
   Nutrient management planning

   Total costs
    Scenario 1
Least Flexible TMDL

     4.0-10.2
       28.3
    41.0-68.2
        3.0

    76.4-109.8
 Scenario 2
Reas. TMDL

 4.0-10.2
 4.3-19.9
 3.8-41.2
  0.4-2.1

 12.5-73.4
National total potential savings from implementing these BMPs (millions
of 2000 dollars/yr):

   Value of additional collected manure
   Increased valued from additional manure hauling
   Nutrient management planning

   Total potential savings
    Scenario 1
Least Flexible TMDL

      0-17.3
     7.5-18.2
      0-31.2

     7.5-66.7
 Scenario 2
Reas. TMDL

  0-11.3
 0.2-11.9
  0-19.6

 0.2-42.7
                                            Exhibit III-4e
            Implementation Costs for Nonpoint Sources -- Silviculture ($ in millions/yr)
Number of States covered
Number of counties with silviculture-impaired waters in these States
Fraction of timber harvest in these States that is in impairment
counties
Scale factor from these States to the nation
National total annualized BMP costs (millions of 2000 dollars/yr)
30
294
.2106
1.499
Scenario 1
Least Flexible TMDL
29.7-41.7
Scenario 2
Reas. TMDL
7.2 - 30.6
                                                III - 20

-------
                                          Exhibit III-5f
     Implementation Costs for Nonpoint Sources - On-Site Wastewater Treatment Systems ($ in
                                           millions/yr)
Number of States covered
Number of counties with OWTS-impaired 303(d) waters in these States
Fraction of OWTS-served dwelling units in these States that is in riparian
zones and will be addressed by TMDLs
Scale factor from these States to the nation
National total annualized implementation costs (millions of 2000
dollars/yr)
24
318
0.013
1.732
Scenario 1
Least Flexible TMDL
24.1 -27.7
Scenario 2
Reas. TMDL
24.1 -27.7
Total nonpoint source costs are estimated as follows:
                                          Exhibit III-6
                         Total Implementation Costs for Nonpoint Sources
Type
Agricultural land
crop land
pasture land
range land
Potential savings
AFOs
Potential savings
Silviculture
On-site wastewater treatment systems
Total
Potential Savings
ADDITIONAL COSTS TO NONPOINT SOURCES
UNDER SCENARIO 3 (MORE COST-EFFECTIVE
TMDL PROGRAM)
Scenario 1
Least Flexible TMDL
Program
645-1,159
5-11
2-16
(not estimated)
76-110
(not estimated)
30-42
24-28
783-2,162
(not estimated)
Scenario 2
Moderately Cost-
effective TMDL
Program
183-1,632
5-11
2-16
(not estimated)
13-73
(not estimated)
7-31
24-28
234-1,791
(not estimated)
47-78
       In the final row of this exhibit showing total nonpoint source costs, we show the estimated costs for
additional nonpoint source control measures under our simulated "More Cost-Effective TMDL Program"
Scenario. These are the costs for nonpoint sources to provide increased load reductions to offset the
additional loads to be discharged by point sources under this scenario (see section II -1). We have not
estimated which specific nonpoint source types will incur these costs for additional nonpoint source
controls.
                                             Ill - 21

-------
IV. MAJOR ASSUMPTIONS, BIASES, AND UNCERTAINTIES

       Any estimate of the nationwide costs to pollutant sources of meeting water quality standards
through the TMDL program or other means will necessarily be imprecise. The analysis must cover nearly
22,000 impaired waters, at least 60,000 point sources, and millions of acres of diverse nonpoint source
activities.  Accurate estimation of the amounts by which loads from these sources to these waters exceed
desired levels, and identification of the specific  sources that are responsible and the load reductions that
they will need to achieve must await development of the actual TMDLs for these waters. Projecting these
elements in the absence of TMDLs, years in advance and on a nationwide basis, is a difficult task.  In
developing our estimates of the implementation costs likely to arise in meeting water quality standards, we
have made many simplifying analytical assumptions. In this section of the report, we note these
assumptions and discuss the potential biases and uncertainties that likely result from them. We first
discuss each major assumption in some detail, and we then provide a exhibit summarizing the implications
of all the major assumptions.

A. HOW THE "LEAST FLEXIBLE TMDL PROGRAM" SCENARIO IS DEFINED

Assumption

       For point sources, we assume that a reasonable least flexible case is that States will continue
writing WQBELs as required by the Clean Water Act. We assume that each point source that contributes
to impairment will have a WQBEL written that will require the source to implement an appropriate "next
treatment step". (An even-worse case might be envisioned, in which nonpoint sources are left out of the
TMDL and States attempt to make up for not including nonpoint sources by developing ever-tighter limits
for point sources.  We consider this option unrealistic.)

       For nonpoint sources, we assume that States will either induce or require all nonpoint sources that
contribute to impairment similarly to implement an appropriate "next treatment step". We assume,  based
on how many nonpoint source programs are currently managed, that a least flexible case would be that
these further nonpoint source controls will be sought for all of the relevant nonpoint sources within the
county or counties in which the nonpoint source-impaired water body is located.

Impact of Assumption

       For both point and nonpoint sources, these assumptions will usually cause the "Least Flexible
TMDL Program" to overshoot the load reduction that would actually be required to meet water quality
standards. One might question whether States would be so short-sighted as to require controls of individual
point and nonpoint sources that in the aggregate far overshoot the overall load reduction needed. Despite
the high cost of this approach, though, we find it plausible that this could occur in some  TMDLs.  Water
quality-based permitting often does proceed for each point source discharger in isolation, without
considering the aggregate load reduction that is being obtained from all dischargers. Nonpoint source
management programs, particularly when they confer some benefit on participating nonpoint sources (e.g.,
agricultural cost-share programs) sometimes find it difficult to direct resources at one part of a county and
exclude the remainder of the county from participation.
                                             IV- 1

-------
B. HOW MUCH LOAD REDUCTION FROM POINT SOURCES WILL THE "MODERATELY
COST-EFFECTIVE TMDL PROGRAM" SEEK?

Assumption

       Based on the results from a sample of 15 completed TMDLs, we assume that about half of the
TMDLs in practice will require an aggregate load reduction from point sources approximately equal to that
which would be obtained if all point sources that contribute to impairment were to implement an
appropriate "next treatment step". The remaining half of the TMDLs will require an aggregate load
reduction from point sources that is about half of that which would be obtained if all point sources that
contribute to impairment were to implement the "next treatment step".

Impact of Assumption

       This specific assumption results in point source costs for the "Moderately Cost-effective TMDL
Program" scenario being only about 3/4 of the costs that would prevail if all point sources that contribute
to impairment were to implement the "next treatment step".  Basing such an important assumption on a
relatively small sample of only 15 TMDLs is unfortunate. We will expand the sample and revise this
assumption accordingly for the final report.

C. HOW MUCH LOAD REDUCTION FROM NONPOINT SOURCES WILL THE
"MODERATELY COST-EFFECTIVE TMDL PROGRAM" SEEK?

Assumption

       First, we assume that TMDLs are legally required for waters impaired by nonpoint sources only.
Second, based on the results (25th and 75th percentile figures) from a sample of 15 completed TMDLs, we
assume that a typical TMDL for a single impaired water will require load reductions from a watershed
acreage amounting to an area ranging in size from about 5 % to 40 % of the surrounding county.  Third,
we assume that the percentage load reduction that a typical moderately cost-effective TMDL will require of
nonpoint sources within this geographic area is roughly similar to the percentage load reduction that would
be achieved by our chosen "next treatment steps" for nonpoint sources.

Impact of Assumption

       The latter two assumptions make nonpoint source costs under the "Moderately Cost-effective
TMDL Program" scenario much lower than under the "Least Flexible TMDL Program" scenario. This
seems sensible - typically the watershed for a 303(d) water is much smaller than a county.  However, a few
TMDLs will address water bodies (particularly  estuaries) that will require nonpoint source load reductions
throughout a very large upstream watershed.  These few instances where very large areas need further
controls may make the average size of the geographic area needing  controls larger than even the 75th
percentile figure. We seek suggestions from reviewers about how to deal with this issue.

       Again, the sample of 15 completed TMDLs is undesirably  small.

       We have very little data on the load reduction effectiveness of the BMPs we have selected as the
"next treatment step" for the various nonpoint source types.  We believe that the BMPs we selected may
generally abate a greater proportion of nonpoint source loads than is typically required in actual TMDLs,
particularly because for several nonpoint source types we include several BMPs simultaneously as


                                            IV-2

-------
constituting the next treatment step. If so, somewhat less than 5 - 40 % of all the nonpoint sources in the
county would need to implement our "next treatment step" in order to obtain the desired aggregate load
reduction. If so, this assumption leads us to overestimate nonpoint source costs.

D. WHAT COST SAVINGS OPPORTUNITIES ARE AVAILABLE UNDER THE "MORE
COST-EFFECTIVE  TMDL PROGRAM?

Assumption

       We simulate only one sort of more cost-effective WLA opportunity - instances where some point
source control responsibilities may be shifted to nonpoint sources.  We do not simulate any of several other
potential cost-saving approaches, including point/point tradeoffs and nonpoint/nonpoint tradeoffs. For the
point/nonpoint allocations that we do simulate,  we adopt a rather restrictive set of rules as to when such
allocations or trading is possible (e.g., for nutrients or BOD only, for nonflowing waters preferentially).
We assume (based on nine case examples of point/nonpoint trading) that nonpoint sources can abate
nutrients and BOD at roughly 1/4 the cost per pound as  can point sources.

Impact of Assumption

       The potential cost savings associated with More Cost-Effective TMDLs are sharply
underestimated because we have not been able to simulate many of the available mechanisms. On the other
hand, the administrative costs of investigating, implementing and overseeing more cost-effective WLAs or
trades have not been estimated, and they could consume  some of the projected savings and/or deter these
allocations from being implemented.

E. UNDER WHICH OF THE THREE SCENARIOS WILL THE IMPAIRED WATERS BE
RESTORED SOONEST?

Assumption

       We do not address this issue.  We assume that the timing of compliance spending by pollutant
sources will be identical across the three scenarios.

Impact of Assumption

       We have made this assumption so that the estimated cost differences between the  scenarios result
only from differences between them in how much load reduction they will require and how efficiently they
will achieve this load reduction. We do not want to complicate this comparison by introducing differences
in the timing when load reduction efforts will be made.

F. THE DISTANCE UPSTREAM OF AN IMPAIRED WATER WITHIN  WHICH POINT
SOURCES CONTRIBUTE TO IMPAIRMENT

Assumption

       We simulate two cases. In the "within only" case, only point sources discharging the impairment
pollutant directly into the impaired water are assumed to contribute to impairment. In the "within and
upstream" case, the point sources contributing to impairment are those that discharge the impairment
                                            IV-3

-------
pollutant: 1) Within 25 miles upstream of impairments for BOD, ammonia or toxic organics; and 2) Within
50 miles upstream of impairments for nutrients or metals.

Impact of assumption

       The "within and upstream" case likely "pulls in" too many point sources. Most impaired stream
segments are relatively short (more than 80 % are less than 20 miles).  Unless the stream segments
immediately upstream and immediately downstream of these short impaired segments are impaired also, we
can conclude that most impairments are localized - they begin and end within a short distance.  This
suggests that the source of impairment is also local. If pollution from a more distant source (up to 25 or 50
miles away) were the cause of impairment, the impaired stretch of river would tend to be much longer.

       On the other hand, the "within only" case likely pulls  in too few point sources.  Several of the
"ground-truthing" TMDLs that we reviewed address point sources located upstream of the impaired
segment itself. Several address all the point sources in the watershed upstream of the impairment.

       The two cases likely bracket the true average.

G. DETERMINING WHETHER OR NOT A POINT SOURCE IS LIKELY TO DISCHARGE THE
IMPAIRMENT POLLUTANT IN A QUANTITY WARRANTING FURTHER CONTROLS
BEYOND TECHNOLOGY-BASED STANDARDS

Assumption

       We test several different decision rules that involve the SIC code in which the point source is
classified and what we know about the typical nature of discharges from facilities in that SIC. A first
alternative assumption is that if at least 15 % of the facilities in an SIC monitor for a pollutant, then all
facilities in the SIC discharge the pollutant in quantities  warranting further treatment. An alternative set of
assumptions involves SIC-by-SIC engineering judgments as to whether or not a facility that meets
applicable technology-based standards is likely to discharge a sufficient remaining quantity of the pollutant
to warrant serious consideration for further control.

Impact of assumption

       Sensitivity analysis  suggests that alternative assumptions have relatively little impact (+ or -  15 %
or so) on estimated implementation costs.

H. POINT SOURCES THAT ARE LEFT OUT OF THE ANALYSIS

Assumption

       The need for further control measures is considered for every point source discharger in PCS, as
well as for all indirect industrial dischargers connected to major POTWs.  Point sources that do not have
individual NPDES permits are thus omitted, including, for example,  CAFOs that are not yet permitted,  and
the many small point sources covered by general permits. Point source discharges from inactive and
abandoned mines (lAMs)  also are not estimated here.  (Note that costs for further controls  needed for urban
wet weather point sources are viewed as attributable to existing technology-based standards and are
counted as part of the baseline. These include costs associated with CSOs, SSOs., municipal and industrial
storm water dischargers.)


                                             IV-4

-------
Impact of assumption

       Some of the omitted point source dischargers are not likely to be significant targets in TMDLs, and
their omission probably results in only a small underestimate of point source compliance costs (e.g many of
the small dischargers covered under general permits).  Other of the omitted point source pollutant sources
are effectively covered in our nonpoint source cost analysis (e.g., CAFOs).  Discharges from lAMs,
however, could be costly to mitigate.  Currently we do not have accurate estimates of the proportion of
listed waters that are impaired by lAMs. Omitting this category from the cost assessment underestimates
point source implementation costs,  but it is difficult to predict the exact magnitude of this underestimation
without further detailed analysis.

I. WILL TMDLs FOR WATERS CITED BY STATES AS IMPAIRED BY NONPOINT
SOURCES ONLY REQUIRE CONTROLS ALSO FOR POINT SOURCES?

Assumption

       We assume not. If the State indicates that a water body is impaired by nonpoint sources only, we
presume that the TMDL will not require controls for point sources even if, in our analysis, there appear to
be point sources within  a relevant distance that discharge the impairment pollutant.

Impact of the assumption

       This assumption reduces our cost estimates by roughly 35 %. We believe the assumption is
reasonable. We suspect that many of the point sources we identify as apparently discharging the
impairment pollutant into or near a nonpoint source-impaired water body are "false positives". Presumably
the State is well aware of any point sources that may be contributing to an impairment, and will note any
such point sources when listing the sources of impairment. We probably often incorrectly attribute the
discharge of an impairment pollutant to a point source as a result of our crude SIC-by-SIC engineering
judgments.

J. SOURCE TYPES COVERED BY THE NONPOINT SOURCE ANALYSIS AND SOURCE
TYPES THAT ARE LEFT OUT

Assumption

       The need for further control measures is considered for four sorts of nonpoint sources: agricultural
land (including crop, pasture and range), AFOs,  silviculture,  and on-site wastewater treatment systems.
Many potentially important types of nonpoint and other sources are omitted: abandoned mines,
contaminated sediments, unintended stream modification, atmospheric deposition, etc..

Impact of assumption

       In omitting these categories of nonpoint sources, we underestimate total costs to nonpoint sources
of meeting water quality standards. The categories of nonpoint source activities that we cover are
responsible for the majority of nonpoint source impairments; the categories that we omit account for about
14 % of the 303(d) river miles and  22 % of the 303(d) lake acres. However, some of the omitted nonpoint
source categories, although they are occur less frequently, can be costly to mitigate: contaminated
sediments and abandoned mines, for example.
                                             IV-5

-------
K. IDENTIFYING THE WATER BODIES THAT ARE IMPAIRED BY EACH OF THESE
NONPOINT SOURCE ACTIVITIES

Assumption

       We rely on the data provided by States on the sources responsible for impairment of each listed
water. If a State does report some impairments due to a particular nonpoint source type, we assume that the
State has reported all such impairments (except that we assume some fraction of the "source unknown" and
"nonpoint source - no further information" waters will also eventually prove to be impaired by the source
type of interest).  We also assume that a State that does not report any impairments due to a particular
nonpoint source type is, in fact, not reporting at all -- the State may really have impairments due to the
nonpoint source type. We scale from the "reporting" States to the "non-reporting" States, presuming that
there are impairments dues to the nonpoint source type in the "non-reporting" States also.

Impact of assumption

       This likely results in a modest overestimate of the number of water bodies actually impaired by the
nonpoint source type in question.  Some of the States that do not report any impairments  due to the
nonpoint source type in fact have no such impairments.

L. WILL STATES REQUIRE FURTHER CONTROLS OF POTWs  THAT ALREADY
PROVIDE BETTER-THAN-SECONDARY TREATMENT?

Assumption

       We assume not. POTWs that now provide better-than-secondary treatment are identified using
CWNS data, and we assume that TMDLs as a matter of equity will focus requirements for load reductions
on other sources that have not yet implemented control measures beyond the applicable technology-based
requirements.

Impact of assumption

       This assumption has a large impact.  It reduces estimated costs for POTWs by approximately 63
% (Moderately Cost-effective TMDL Program scenario, "upstream and within" case), from $1.40
billion/yr to $523 million/yr. The ground-truthing information suggests that some TMDLs are requiring
further controls from POTWs that already provide advanced treatment, while other TMDLs are not
requiring further controls from such POTWs.  This assumption represents one of  the most important
respects in which we may underestimate the costs of the TMDL program.

M. THE SPECIFIC FURTHER CONTROLS THAT POINT OR NONPOINT SOURCES WILL
NEED TO IMPLEMENT

Assumption

       We assume that each source that needs to reduce its load will do so by implementing "the next
treatment step" beyond whatever technology-based requirements are currently applicable to that source. If
the source has already implemented all or part of "the next treatment step", we reduce the estimated costs
accordingly (e.g., no costs are estimated for POTWs that already provide better-than-secondary treatment,
and nonpoint source costs are reduced to the extent that the nonpoint source already has some of the "next
treatment step" BMPs in place). However, we are unable for industrial dischargers to apply this credit
                                            IV-6

-------
when treatment in place exceeds technology-based requirements because we have no means of identifying
which or how many industrial dischargers have such treatment in place.

Impact of assumption

        Uncertain. Some water bodies are so seriously impaired that all contributing sources will need to
implement controls well beyond the next treatment step.  Other water bodies are only slightly impaired, and
they will need only modest further source controls, well short of the next treatment step for all sources. For
the Moderately Cost-effective TMDL Program Scenario, we have attempted to match the extent of point
and nonpoint sources that we assume will need to implement the next treatment step against the typical
aggregate load reductions from point and nonpoint sources that have been required in a sample of actual
TMDLs. Hopefully this procedure calibrates our cost estimates so that, although we may be far from
accurate about costs for any particular water body, our total national estimate is reasonably accurate.

        Costs for industrial dischargers are overestimated because we do not have the data that would
allow us to reduce their estimated costs by recognizing the extent to which they have already implemented
the next treatment step.

        Many point and nonpoint pollutant  sources currently have more treatment in place than is needed
to meet technology-based requirements.  If we were to assume that all pollutant sources need to implement
the "next treatment step", even those who currently surpass technology-based requirements, we would face
the difficult task of developing cost functions for many additional treatment technologies that represent
"next steps" beyond the wide variety of technologies currently in place.

N. THE "NEXT TREATMENT STEP' A POINT OR NONPOINT SOURCE WILL NEED TO
IMPLEMENT

Assumption

        The assumed next treatment step varies  with the type of point or nonpoint source and (for point
sources) the pollutant needing further control. All sources are assumed to have treatment in place to meet
applicable current technology-based-standards.  For point sources, this means  treatment measures
sufficient to meet effluent guideline (BPT, BAT, PSES, etc.) and secondary treatment requirements.  For
nonpoint sources, no treatment is assumed to be in place to meet Federal regulatory requirements (i.e., for
most nonpoint source types, there are no technology-based standards currently applicable), but we attempt
to reflect the degree to which treatment is in place for other reasons.  For urban wet weather sources, we
assume that controls are in place or are being implemented to meet CSO, SSO and storm water
requirements under current CWA programs. The following exhibit shows the treatment assumed to be in
place and the assumed next treatment step for all types of sources.

Impact of assumption

        These assumptions regarding treatment in place and what the next treatment step would be are
undoubtedly inaccurate in many specific instances.  To improve on these assumptions, though, we would
need to acquire source-specific information  on treatment actually in place  and potential further control
options for tens of thousands of point and nonpoint sources - a formidable task. We do not believe there is
any systematic bias resulting from the assumptions we have made.
                                             IV-7

-------
          Exhibit IV-1
Treatment Technology Assumptions
Type of Source
Industrial point
sources
POTW


Agricultural
land (nonpoint
source)



AFOs (nonpoint
source)

Silviculture
(nonpoint
source)
On-site
wastewater
treatment
Pollutant or Activity
Metals
BOD, nutrients, ammonia,
or toxic organics
Metals
BOD, nutrients, ammonia,
or toxic organics
Tillage
Fertilizer use
Riparian zones
Concentrated runoff on
sloped fields
Farming on highly erosive
land
Collect manure and spread
it on nearby fields
Manure hauling
Manure and fertilizer use
Harvesting-related
activities
Artificial forest
regeneration
Failing systems
Treatment in Place
Chemical precipitation
Biological treatment
(secondary)
Pretreatment program
requiring metals dischargers
to meet effluent guidelines
(chemical precipitation)
Biological treatment
(secondary)
Partial adoption of
conservation tillage
Partial adoption of nutrient
management planning
Minimal
None
Some land already retired in
Conservation Reserve
Most necessary facilities are
commonly in place
Virtually none
Partial adoption of nutrient
management planning
As prevailed in roughly
1990
As prevailed in roughly
1990
Actual rate of failure
Next Treatment Step
Polishing filtration
Advanced secondary
treatment
Enhanced pretreatment
program with tighter limits
(polishing filtration)
Advanced secondary
treatment, secondary
treatment with nutrient
removal, or both
Conservation tillage
Nutrient management
planning
Riparian forest buffers for
crop and pasture land; use
exclusion or stream protection
for range land
Vegetative barriers
Retire and cover remaining
highly erosive cropped land
Comprehensive set of manure
management measures
Transport manure to locations
where it can be spread at
agronomic (P-based) rates
Nutrient management
planning
Comprehensive set of
measures to minimize
impacts
Comprehensive set of
measures to minimize
impacts
Repair and failing systems
in riparian zone
            IV-8

-------
O. ESTIMATING COSTS FOR SOURCES TO IMPLEMENT THESE CONTROL MEASURES

Assumption

       For each type of source and type of control, we estimate costs using a control cost function drawn
from an appropriate EPA reference. Each cost function projects costs as a function of the size (e.g.,
wastewater flow, number of animals at an AFO, number of acres cut for silviculture) of the source.

Impact of assumption

       Unknown.  Perhaps these cost models overestimate the ultimate costs for meeting LAs and WLAs.
Ex post analyses of the actual costs that  sources have incurred to meet pollution control standards have
often found that ex ante engineering cost models overestimated costs — process modifications or material
substitutions were implemented that reduced emissions at costs lower than the add-on control equipment
that provided the basis for the original cost estimate.

       The capital cost functions drawn from the CWNS for the next treatment step for point sources for
BOD, nutrients, toxic organics and ammonia seem potentially likely to overestimate costs of upgrades
because they simulate building a new plant and receiving a credit for the salvage value of the existing plant.

       For nonpoint sources, we generally selecte a more expensive specific BMP to represent the costs of
the entire variety of possible BMPs within any given practice group.

P. FLOW OR VOLUME INFORMATION USED FOR EACH SOURCE IN TREATMENT COST
FUNCTIONS

Assumption

       For point sources, wastewater flow information is obtained from PCS. This is despite our
impression that flow information reported in PCS is poorly quality-controlled and often in error.  We use
the flow information reported in PCS for any facility for which such information is available. For the
many dischargers with no flow information reported,  we assign the source the average flow reported for
other sources in that SIC or group. We assume that dischargers with large flows will segregate their waste
streams such that advanced treatment will be applied to only the small, most contaminated portions of their
total flow. We thus limit in various ways the maximum flows that we assume each point source discharger
will need to treat.  For nonpoint sources, we use the following measures  of volume in the cost functions:
acreage (for agricultural land), acreage and volume cut (for silviculture), number of animals or animal
units (for AFOs), and number of dwelling units served by on-site wastewater treatment systems (for
OWTS).

Impact of assumption

       The implementation costs we estimate for point sources are substantially affected by the set of
assumptions  and procedures to deal with flow issues. On the whole, we believe that our assumed upper
limits on the flows that will be treated are conservative — that is, the great majority of point sources will
have flows to be treated that are actually well  below the upper limits that we assumed, and costs therefore
are likely overestimated. For nonpoint sources, we believe the volume information used for costing is
reasonably accurate.
                                             IV-9

-------
Q. ADJUSTMENTS OF THE IMPLEMENTATION COST ESTIMATES TO REFLECT
MISSING DATA

Assumption

        The data that we use to estimate costs are not complete. Some impaired water bodies are not
georeferenced. Without location information for these water bodies, they can not trigger any
implementation costs for nearby point or nonpoint sources.  Similarly, some point sources are not
georeferenced; we cannot estimate whether or not they will incur costs because we do not know whether or
not they are within or upstream of impaired waters.  We use a scaling procedure to adjust our cost
estimates to reflect this missing data.  Implementation costs are estimated for the subset of States and
sources for which we have data, and we then extrapolate the results to the remainder of the Nation using a
scaling factor reflecting the portion of the Nation for which we do have data. Our key assumption is that
the portion of the Nation for which we do not have data has impairments and will incur compliance costs at
the same rate as a function of source activities as does the portion of the Nation for which we do have data.

Impact of assumption

        This scaling adjustment for missing  data adds some uncertainty, but probably entails little bias.
Our extrapolation from the areas where we have data to the areas where data are missing is problematic
only if there is something systematically different in the relationships between sources and impairments in
the areas with data compared with these relationships in the areas without data. We do not expect any such
systematic differences  to arise because of missing georeference information for some point sources, as it
appears to be generally random whether or not point sources are georeferenced. However, there are
systematic differences  in whether or not water body location information is  available. Georeference
information is not available for about 35 % of all impaired waters, including all waters in CA and Region
10 (which include about 17 % of all impaired waters). Our extrapolation procedure may be inaccurate if
source/impairment relationships are generally different in these far-western  States than in the rest of the
country. If, for example, silviculture is either more or less likely to cause water quality impairments in the
far-western States than in the rest of the country, then extrapolating silviculture implementation costs based
on volume cut from the remainder  of the Nation to these States is inaccurate. We believe, for example, that
silviculture is significantly more tightly regulated in these States than elsewhere, suggesting that it may
cause fewer water quality problems in the far West. On the other hand, the generally steep slopes in the
areas  where logging is conducted in the far West suggest perhaps greater problems there than elsewhere.
On the whole, we believe that these sorts of possibilities for bias in our extrapolation procedure are not
substantial.

        USDA has suggested in reviewing a draft of this report that using national  average figures for the
unit costs  of agricultural BMPs may result in underestimating national costs if costs are typically higher in
the specific areas where TMDLs will most frequently require load  reductions.  For  example, although we
have estimated the costs to retire crop land based on national averages for this practice, USDA surmises
that much of the crop land retirement that may be prompted by TMDLs will occur in the midwestern
agricultural heartland, where crop  land values are particularly high. We will investigate this sort of
possibility before finalizing this report.

R. TIMING ASSUMPTIONS CHOSEN FOR THE IMPLEMENTATION COST ANALYSIS

Assumption

        We have estimated the costs for point and nonpoint sources to achieve  water quality standards for
the currently listed 303(d) waters.  We estimate these costs assuming a steady pace of TMDL development

                                             IV- 10

-------
from now through 2015 (and we further assume that costs will be incurred under the other scenarios at the
same time as under the TMDL Program scenario).  We assume that the average source will begin incurring
its compliance costs five years after the TMDL is developed. We discount costs that will be incurred in the
future back to the present, and calculate an equivalent level, annualized amount beginning in the year 2000
and continuing forever.

Impact of assumption

       The implementation costs we estimate would be substantially different if different analytical
ground rules were chosen.  Our chosen TMDL pace/compliance time lag scenario reduces the costs we
show as annualized costs beginning now to an amount that is 44.84 % of what annual costs will be after all
affected sources have had to meet TMDL requirements.

       On the other hand, the costs for a pollutant source to achieve its allowed load under a TMDL
developed far in the future may be greater than if the TMDL were developed today.  Growth in the
population served by a POTW or growth in production by an industrial discharger may make achieving the
same discharge limit in, say, 2010 more costly than achieving that limit in 1999. In addition, as States
assess more of their waters, additional water bodies will be listed for which TMDLs  will need to be
developed.  In this analysis, we estimate the costs associated with implementing TMDLs only for waters
that are currently listed.
                                             IV- 11

-------
APPENDICES

-------
             APPENDIX A: GROUNDTRUTHING THE ASSUMPTIONS

       In estimating the costs to pollutant sources from TMDLs or alternatives to them, this study makes
many assumptions that simplify the analysis and reduce the extent of data needed. To assess the validity of
these assumptions, we would like to compare them against actual experience among the TMDLs that have
been completed thus far.  We have only begun to do this. We have reviewed 15 completed TMDLs and
compiled information from them that is relevant to the analytical assumptions we have made. In this
appendix, we "groundtruth" the assumptions by comparing them against what has happened among this
sample of completed TMDLs.  Given the wide variation in the sorts of waters and impairments that actual
TMDLs will address, a sample of 15 completed TMDLs is obviously far short of an ideal data base from
which to draw conclusions. Nevertheless, this initial groundtruthing effort provides some preliminary
indications of patterns across actual TMDLs, and the results are sufficiently interesting to warrant a
substantially increased effort to develop a larger and more representative database of completed TMDLs
for groundtruthing.

       Tetra Tech selected 15 recently completed TMDL submissions for analysis.  No formal statistical
sampling or selection procedure was employed.  Tetra Tech simply selected TMDLs that the firm was
familiar with, and which would cover a substantial range of impairment pollutants, source types (both point
and nonpoint sources) and geographic locations. Tetra Tech then abstracted information from the selected
TMDLs to fill in a common data collection template. This information was further condensed to develop a
one-page summary for each TMDL (attached at the end of this appendix) and a summary exhibit for the
entire set of 15 TMDLs (appearing on the next page). On the following pages, we review what can be
learned from the 15 TMDLs that is relevant to the major assumptions adopted for this analysis.
                                             A-l

-------
                   Exhibit A -1
Summary of Groundtruthing Information From 15 TMDLs
TMDL Submission
Anderson Run, WV
Appoquinimink River,
DE
Bayou Nezpique, LA
Buckhannon River,
WV
Coeur d'Alene River
Basin, ID
Duck Creek, AK
Elk Creek, OK
# of 303(d)
Waterbodies
Covered
1
1
1
(or more,
unclear)
1
28 plus
1
1
Pollutants
Fecal
coliform
bacteria
CBOD
NBOD
P
CBOD
NH3
Al
Fe
Mn
Dissolved
Metals (Pb,
Cd.Zn)
Sediment/
Turbidity
BOD
NH3
% Reduction
42
56
52
68
For both pollutants,
0-50% reduction
from PS, 85-90%
reduction from
NPS
8
8
3
Unknown. Target
load is assigned
but current loads
unknown
42-62
50-67
Unknown
Source Types
Assigned
Reductions
NPS
PS
NPS
PS
NPS
NPS
PS
NPS
NPS
PS
Are All Relevant
PS Assigned
Reductions?
N/A
Yes
No
No
Yes
N/A
Yes
Geographic Extent of NPS
Assigned Reductions
Entire watershed, 7% of county
Entire watershed, 11% of county
Entire watershed, 22.3% of four
counties
Loads were addressed in the entire
watershed, which encompasses
1 8% of total land area of three
counties. Reductions were
assigned to 3 of 14 subwatersheds
within this land area.
Entire watershed, comprising most
of 3 counties
Entire watershed, 0.03% of county
N/A
                       A-2

-------
TMDL Submission
Garcia River
Watershed, CA
Hurricane Lake, WV

Lake Thonotosassa,
FL


Muddy Creek, VA

Nanticoke River and
Broad Creek, DE

Neuse River Estuary,
NC
Noyo River
Watershed, CA
Sanders Branch/
Coosawhatchie
River, SC
# of 303(d)
Waterbodies
Covered
1
1

1


1

2

1 (large
estuary)
1
2


Pollutants
Sediment
Sediment
P
Fe
N

P
Fecal
coliform
bacteria
BODS
N
P
N
Sediment
CBOD
NBOD

% Reduction
60
30
45
30
Unknown. Target
load is assigned
but current loads
unknown
86
99

29
23
32
30
67
Unknown


Source Types
Assigned
Reductions
NPS
NPS

NPS


NPS

PS
NPS

PS
NPS
NPS
PS


Are All Relevant
PS Assigned
Reductions?
N/A
No

No


No

No

Yes
N/A
No


Geographic Extent of NPS
Assigned Reductions
Entire watershed, 3.2% of county
Entire watershed, 0.7% of county

Entire watershed, 5.2% of county


Entire watershed, 3.7% of county

Entire watershed, 42% of county

Entire watershed, includes most of
19 counties
Entire watershed, 3.2% of county
N/A


A-3

-------
I. COMPARISON OF KEY ANALYTICAL ASSUMPTIONS WITH INFORMATION FROM 15
TMDLs

A. ASSUMPTIONS RELEVANT TO POINT SOURCE COSTS

        •       Are controls for point sources triggered by pollutants other than BOD, nutrients, metals,
               toxic organics, or ammonia?
                      No. Among this set of TMDLs, the pollutants triggering point source controls
                      include: BOD (5 instances), nutrients (3), ammonia (2), and metals (1).

        •       Are the only point sources assigned reductions those that discharge directly to the impaired
               segment,  or are upstream point sources addressed also?
                      In many instances, point sources upstream of the impaired segment are addressed
                      also. In some instances, some upstream point sources are believed to be too
                      distant to warrant further controls.

        •       At what distance upstream of the impairment are point sources that discharge the
               impairment pollutant still considered to need controls? (e.g. 25/50 miles?)
                      We can't tell from the available information how far upstream each point source is
                      located (whether the point source is addressed or not addressed). Further
                      investigation would be necessary to ascertain distances.

        •       Are all point sources thought to  be discharging the impairment pollutant within a relevant
               distance of the impairment assigned a load reduction?
                      No. In 4  of 11 cases, all relevant point sources are assigned reductions.  In 4
                      cases the  point sources contributing minimally to impairment  are assigned no
                      reductions and all important point source pollutant sources are assigned
                      reductions.  In 2 cases explicit decisions were made that some significant pollutant
                      sources would need reductions while others would not. In one case an aggregate
                      load reduction was assigned to all point sources and trading was expected.

        •       Do the TMDLs often require further load reductions from point sources that already
               provide treatment in excess of technology-based standards?
                      Yes, in at least two cases point sources that already provide advanced treatment
                      are being required to do more. In many cases, it is difficult to tell from the limited
                      information available.

                      In two other cases, particular point sources are not required to reduce their loads
                      further because they have already adopted some form of advanced waste treatment
                      (even though they could reduce loads further  anc achieve zero discharge).

        •       How does the percentage load reduction required of point sources in the aggregate compare
               with the load reduction that would be achieved if all relevant point sources implemented
               "the next treatment step"?
                      See the exhibit and detailed conclusions that follow below.
                                               A-4

-------
                                      Exhibit A-2
Information From 11 Groundtruthing TMDLs Regarding % Reductions Sought from Point Sources
Water Body
Appoquinimink River, DE
Bayou Nezpique, LA
Buckhannon River, WV
Coeurd'Alene River Basin, ID
Elk Creek, OK
Hurricane Lake, WV
Lake Thonotosassa, FL
Muddy Creek, VA
Nanticoke River & Broad Creek, DE
Overall Approach to Point Sources in This TMDL
River is affected by only one point source; it is assigned load reductions
Is significantly affected by 9 PS and insignificantly affected by 16 PS.
The 9 are assigned varying load reductions ranging from 0 - 50 % for
BOD and NH3. The 16 are assigned no load reductions
Very small contributions from numerous PS in the watershed. No load
reductions assigned to PS
Numerous PS identified as contributing metals, the impairment
pollutants. Allowable loads were assigned to PS in the aggregate, but
current loads are unknown and hence % reduction is unknown
Creek is affected by only one point source; it is assigned load reductions
Lake may be affected trivially by only one small PS. It is not assigned a
load reduction
Lake may be minimally affected by numerous PS that already employ
land treatment/zero discharge
Very small contributions (of pathogens) by PS. They are not assigned
load reductions
Significant contributions from many point sources. Explicit decision
made to require load reductions from some and not from others
Our Judgment About Overall % Reduction That
This TMDL Requires for Total Load from Point
Sources
CBOD: 57%
NBOD: 83%
P: 85%
BOD, NH3 both likely 10 -40%
Waterbody is affected by NPS only. This TMDL is not
relevant in assessing typical % reduction required of
PS.
Unknown
BOD: 50 - 67 %, depending on season
Waterbody is affected by NPS only. This TMDL is not
relevant in assessing typical % reduction required of
PS.
Waterbody is affected by NPS only. This TMDL is not
relevant in assessing typical % reduction required of
PS.
Waterbody is affected by NPS only. This TMDL is not
relevant in assessing typical % reduction required of
PS.
BOD: roughly 30 %
Nutrients: roughly 30 %
                                         A-5

-------
Neuse River Estuary, NC
Sanders Branch/Coosawhatchie R,
SC
Significant contributions from many, many point sources. Uniform
permit limits assigned to each that will achieve the desired aggregate
load reduction. Trading expected
Several important, several lesser PS identified. Significant reductions
assigned to the important PS, while no reductions required of lesser PS
N: 30 %
BOD: unclear,
but likely significant (50 - 80 %?)
A-6

-------
B. CONCLUSIONS REGARDING % REDUCTIONS SOUGHT FROM POINT SOURCES:

       •       Among the sample of 15 TMDLs, 4 do not involve point sources at all, leaving 11 to be
               reviewed in this analysis.

       •       Many waters identified as impaired essentially by nonpoint sources nevertheless have point
               sources nearby that contribute small or insignificant amounts of the impairment pollutant.
               We have 4 examples of this sort among the 11 groundtruthing TMDLs.  In each of these 4
               cases, the TMDL requires no load reductions from the point sources. These cases support
               our assumption in the implementation cost analysis that TMDLs for NPS-only waters will
               not require further controls for point sources.

       •       Among the 7 cases where point sources are seriously addressed by the TMDL, 3 involve
               an aggregate load reduction across all point sources in the 10 - 40 % range, 3 involve
               aggregate load reductions in the 50-85 % range, and one is unknown.

       •       The "next treatment steps" that we simulate for point sources will yield very roughly the
               following percentage load reductions:  (Note, further work is proceeding on this issue)

                      metals from industrial dischargers                     90 %
                      metals from POTWs  (enhanced pretreatment)            70 %
                      BOD                                               50 %
                      nitrogen                                             75 %
                      phosphorus                                          50 %
                      ammonia                                            75 %
                      toxic organics                                       ? (not addressed in this
                                                                          sample)

       •       For perhaps half of the TMDLs, the aggregate load reduction that will be obtained by
               implementing the next treatment step at all relevant point sources (50 to 90 %) is much
               greater than the aggregate load reduction needed  (10-40 %). For the other half of the
               TMDLs, the aggregate load reduction that will be obtained by the next treatment step for
               all relevant point sources will yield approximately the same aggregate load reduction as
               that called for by the TMDL.

       •       Based on this, in analyzing the "Moderately Cost-effective TMDL Program" Scenario, we
               will assume that:

                      In half of the TMDLs, all the point sources will need to take the next treatment
                      step;

                      In the other half of the TMDLs the next treatment step  for all point sources will
                      provide twice the load reduction that is needed. In this  half of the TMDLs, in
                      effect, only half the point sources will need to implement the next treatment step
                      and half of the point sources will likely be able to avoid the next treatment step.

                      We will thus assume that Moderately Cost-effective  TMDL Program costs
                      can be reduced by 1/4 from Scenario 1 costs.


                                              A-7

-------
C. ASSUMPTIONS RELEVANT TO NONPOINT SOURCE COSTS

•      What is the geographic extent of the nonpoint source watersheds from which load reductions are
       sought?

               In 12 of 13 cases, nonpoint sources throughout the entire upstream watershed will be
               affected. In one case, nonpoint sources were singled out in only a few of the key
               sub water sheds.

•      Relative to the acreage in the county or counties within which the impaired waters is located, how
       many acres are in the zone within which nonpoint sources will need to be controlled?

               For Scenario 1 ("Least Flexible TMDL Program"), we assumed that the acreage to be
               controlled was equal to the acreage of the surrounding county or counties.  This appears
               nearly always to be a substantial overestimate relative to what actual TMDLs have
               required. See the exhibit on the next page.  Across the 13 TMDLs requiring nonpoint
               source controls, the median watershed acreage needing controls per waterbody amounted
               to 7.0 % of the acreage of the county or counties. The 25th percentile was controls for
               acreage amounting to 3.2 % of the county's acreage per waterbody, while the 75th
               percentile was controls for acreage amounting to 32.8% of the acreage of the county per
               waterbody.  The mean figure is substantially higher, due to one TMDL covering two
               waterbodies that will require nonpoint source controls for acreage amounting to that for 19
               surrounding counties. See the exhibit on the following page.

               For this analysis, for Scenario 2 ("TMDL Program"), we will assume that the TMDL
               for a nonpoint source-impaired water body will controls for acreage equal to 5 % of
               the number of acres in the surrounding county (lower estimate, slightly larger than
               the 25th percentile figure) or 40 % of the number of acres in the surrounding county
               (upper estimate, slightly larger than the 75th percentile figure).

•      How does the percentage load reduction required of NFS within the area needing controls compare
       with the percentage likely to result from our assumed "next treatment step" packages for nonpoint
       sources?

               The percentage reduction required by the TMDLs from nonpoint sources within the
               targeted areas is: 0 - 25%,  1 case;  25 - 50%, 4 cases; 50 -  75%, 4 cases; 75%+, 2 cases.

               We need to do further research to determine what percentage reduction is likely achieved
               by each of our NFS "next treatment step" packages. As a very rough guess, if our NFS
               packages were to reduce loadings by 75%, we would be overestimating costs somewhat.
                                              A-8

-------
Exhibit A-3

Extent of Nonpoint Source
Controls
Required
by TMDLs


Comparison of Number of Acres in County with Number of Acres of Watershed Controlled
TMDL

Duck Creek, AK
Hurricane Lake, WV
Garcia River, CA
Noyo River, CA
Muddy Creek, VA
Lake Thonotosassa, FL
Anderson Run, WV
Coeurd'Alene, ID
Appoquinimink, DE
Nanticoke River, Broad
Creek, DE
Bayou Nezpique, LA
Buckhannon, WV
Neuse River, NC
Elk Creek, OK
Sanders Branch/
Coosawhatchie, SC






Additional
# of 303d 303d
waterbodi waterbodies
es subsumed
addresse in
din TMDL controlled
area
1
1
1
1
1
1
1
28
1

2
1 1
1
1 1
1

2






Total #
waterbodi
es

1
1
1
1
1
1
1
28
1

2
2
1
2
1

2






#
Counties
within
which
NFS load
reduction
sare
sought
1
1
1
1
1
1
1
3
1

1
4
3
19



mean
median
25th
percentile
75th
percentile
% of these
counties'
area
affected

0.03
0.7
3.2
3.2
3.7
5.2
7.0
100.0
11.0

42.0
22.3
18.0
100.0









% of one
typical
county's
area
affected

0.03
0.7
3.2
3.2
3.7
5.2
7.0
300.0
11.0

42.0
89.2
54.0
1900.0









% of a
typical
county's
area
affected
per
r
waterbody
0.03
0.7
3.2
3.2
3.7
5.2
7.0
10.7
11.0

21.0
44.6
54.0
950.0



85.7
7.0

3.2

32.8
   A-9

-------
APPENDIX B: TMDL PACE AND TIME LAG SCALE FACTOR
            Scale Factor for Pace of TMDL  Development and Compliance Time Lag
   Assume, once all sources have begun incurring costs, that the total annualized compliance costs for all these so
   is $1.00 per year, continuing forever

                                                    Discount Rote:
                                                     After 5-Year Lag, %
                                                         of Sources
                                                          Incurring
                                                     Compliance Costs
Year
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
4> Tinm o/ ntni Cumulative
# TMDLs % TMDLs „/ i-mm
*~ , i , *~ , i , » TMDLs
Completed Completed Comp|eted
2000
2281.666667
2281.666667
2281.666667
2281.666667
2281.666667
2281.666667
2281.666667
2281.666667
2281.666667
2281.666667
2281.666667
2281.666667
2281.666667
2281.666667
2281.666667

36225



5.52%
6.30%
6.30%
6.30%
6.30%
6.30%
6.30%
6.30%
6.30%
6.30%
6.30%
6.30%
6.30%
6.30%
6.30%
6.30%

1



5.52%
1 1 .82%
18.12%
24.42%
30.72%
37.01%
43.31%
49.61%
55.91%
62.21%
68.51%
74.81%
81.10%
87.40%
93.70%
100.00%





                                                                 5.52%
                                                                11.82%
                                                                18.12%
                                                                24.42%
                                                                30.72%
                                                                37.01%
                                                                43.31%
                                                                49.61%
                                                                55.91%
                                                                62.21%
                                                                68.51%
                                                                74.81%
                                                                81.10%
                                                                87.40%
                                                                93.70%
                                                               100.00%
                            Present value for years 2000 through 2020:

                            PV in 2020 for all years after that:
                            PV in 2000 for annual costs after 2020:

                            PV for  all years:
                            Annuity value:
    Scale factor, relative to cost of $l//r forever, beginning in 2000:
0.07
Present
Value of
Compliance
Costs
0.0000
0.0000
0.0000
0.0000
0.0000
0.0394'
0.0788
0.1128
0.1421
0.1671
0.1882
0.2058
0.2203
0.2320
0.2413
0.2483
0.2534
0.2568
0.2586
0.2591
0.2584
3.1622
14.2857
3.6917
6.8539
0.4798
0.4484
0.03
Present
Value of
Compliance
Costs
0.0000
0.0000
0.0000
0.0000
0.0000
0.0476
0.0990
0.1473
0.1927
0.2354
0.2754
0.3129
0.3480
0.3807
0.4113
0.4397
0.4662
0.4907
0.5134
0.5344
0.5537
5.4484
33.3333
18.4559
23.9042
0.7171
0.6962
               0.07        0.03

Baseline: $1 per year, beginning
  2000 and continuing forever
  .0000
  .0000
  .0000
  .0000'
  .0000
  .0000
  .0000
  .0000
  .0000
  .0000
  .0000
  .0000
  .0000
  .0000
  .0000
  .0000
  .0000
  .0000
  .0000
  .0000
  .0000
1.0000
0.9346
0.8734
0.8163
0.7629
0.7130
0.6663
0.6227
0.5820
0.5439
0.5083
0.4751
0.4440
0.4150
0.3878
0.3624
0.3387
0.3166
0.2959
0.2765
0.2584
             11.5940

             14.2857
              3.6917

             15.2857
              1.0700
              1.0000
1.0000'
0.9709
0.9426
0.9151
0.8885
0.8626
0.8375
0.8131
0.7894
0.7664
0.7441 :
0.7224
0.7014:
0.6810^
0.6611 :
0.6419
0.6232''
0.6050 :
0.5874
0.5703
0.5537
           15.8775

           33.3333
           18.4559

           34.3333
            1.0300
            1.0000 :
                                                                     B-l

-------
                    APPENDIX C: POLLUTANTS FOR ANALYSIS
       In order to limit the analytical workload (e.g., developing and applying cost functions for the next
treatment step for each pollutant), we chose to analyze only those particular pollutants for which TMDLs
are most likely to require further controls (beyond technology-based standards) for point sources.  We
considered the most frequent causes of impairment, and asked whether each of these causes would
commonly trigger the need for further controls of point sources that are already meeting technology-based
standards.

       Exhibit C - 1 summarizes our judgments and rationale.  The exhibit shows the number of point
sources upstream within 50 miles of 303(d) waters impaired for different pollutants. Pollutants are listed in
descending order of the frequency with which they implicate point sources for potential further control
needs.  For example, the exhibit shows that there are 4386 point sources upstream within 50 miles of water
bodies that are impaired for BOD/DO and that States have cited as impaired by both point and nonpoint
sources. There are 460 point sources upstream within 50 miles of water bodies that are impaired for
BOD/DO and that States have cited as impaired by point sources only.

                                         Exhibit C -1
      Decisions on Causes/Pollutants to Analyze as Triggers for Further Point Source Controls
Cause/pollutant
BOD/DO
Metals
Nutrients
Sediment
Pathogens
Habitat
Suspended solids
PCB
Flow alteration
Ammonia
# of point sources w/in
50 mi. upstream of
waters impaired by
Waters
Impaired
byPS +
NPS
4386
4277
3965
3656
3564
2420
2165
1599
1570
993
Waters
Impaired
by PS
only
460
776
293
86
131
37
130
126
NA
64
Include this
as a
pollutant
for
analysis?
Yes
Yes
Yes
No
No
No
No
No
No
Yes
Rationale



Probably rare that PS process water discharges are worth
controlling for sediment beyond BPT. PS sediment problems
in wet weather will likely be handled by eventual compliance
with storm water and construction technology-based
requirements.
Problems are likely CSOs, SSOs and other wet weather
discharges, to be controlled by compliance with technology-
based standards. Unlikely that other PS are worth controlling
for pathogens beyond BPT.
TMDL necessary only if a pollutant can be identified
Same as sediment
Problems are likely legacies, not ongoing, controllable process
water discharges needing control beyond BAT
No TMDL necessary
Which SICs might be worth controlling? POTWs? Others?
                                              C-l

-------
Pesticides
Dioxin
PH
Oil and grease
Salinity
Toxics
Chlorine
Thermal mod.
Unknown
Organics
Odor
Dissolved solids
Fish advisories
(w/no mention of
the pollutant)
Bacteria
Sulfates
(Next most
frequent)
Cyanide
Total (Note: this
exceeds the
number of point
sources upstream
of impairments.
Many PS are
counted for
multiple causes.)
971
794
622
560
532
517
293
288
275
274
150
123
84
78
48
24
4
43,747
24
27
8
21
4
35
1
17
43
30
0
15
NA
0
0
23
140
4,136
No
No
No
No
No
Yes
No
No
No
Yes
No
No
No
No
No
No
No
XXX
Very few PS likely worth controlling for pesticides beyond BAT.
Which SICs?
Most problems are likely due to air deposition or legacy, not
ongoing, controllable discharges. Any SICs worth controlling
beyond BAT? Pulp/paper?
Unlikely that PS at BPT could be causing a problem
Unlikely that PS process discharges are worth controlling
beyond BPT. Compliance with existing wet weather
requirements will help.
Highly likely to be NPS problem
Absent more specificity, though, it is unclear what controls to
implement.
Why is this a frequent problem in PS/NPS waters, but not in
PS only waters? Could use further investigation. Candidate
for addition to analysis.
Expect that any PS problems should be taken care of by BPT
Unclear what controls to implement. Scale up to cover this.
Assume this means toxic organics.
Interesting that this shows up for PS/NPS waters, but not for
PS waters. Unclear what treatment measures might be
considered.
Likely a NPS problem only
Usually due to NPS, legacies, air deposition. Candidate for
addition to analysis if can get more data on which pollutant
and whether any suspected contribution from PS process
discharges.
Same as pathogens
Likely a NPS problem only
Do not address very infrequent causes/pollutants
Surprisingly high frequency in PS-only waters. Why?
Candidate for addition to analysis.
XXX
C-2

-------
        We chose to analyze further those pollutants for which point sources are most frequently likely to
need additional controls beyond technology-based standards. Several sorts of considerations are reflected
in Exhibit C-  1:

        •       We chose not to analyze causes of impairment when no pollutant was identified.
               Examples included flow or habitat alteration or fish consumption advisory when no
               pollutant was listed also.

        •       We chose not to analyze pollutants that are extremely unlikely to be discharged in
               sufficient quantity to be problematic by a point source meeting technology-based
               standards.  Examples included temperature or pH.

        •       We chose not to analyze pollutants for which only a very small fraction of point  sources
               might require additional beyond-technology-based controls for process water discharges.
               Examples include sediment or pathogens. We believe that the great majority of instances
               where waters are impaired for these causes and point sources are cited as a source involve
               wet weather discharges from point sources.  We believe these problems will largely be
               remedied when existing technology-based standards for wet weather discharges — storm
               water, construction, CSO and SSO requirements - are complied with.

        •       We chose not to analyze in this study pollutants representing a very small fraction of
               causes that would require specialized treatment technologies as the"next step". Examples
               include chlorine and cyanide.  We could consider such infrequent causes of point source
               impairment in a future study.
                                               C-3

-------
    APPENDIX D: APPROACHES FOR DETERMINING WHETHER A POINT
    SOURCE IS LIKELY TO BE CONSIDERED FOR FURTHER CONTROLS

        The objective of this portion of the analysis was to make a judgment about whether each point
source within a relevant distance upstream of an impairment for a specific pollutant would likely be
considered by the eventual TMDL or water quality-based permit to implement further controls.  We sought
to determine whether an identified point source meeting applicable technology-based standards was likely to
discharge the impairment pollutant in sufficient quantity to warrant consideration for further controls. This
judgment was particularly difficult to make because there is little data available at the national level on the
pollutants that individual facilities discharge and their amounts.  For the majority of dischargers, the Permit
Compliance System (PCS) - EPA's major data base on point source dischargers that we used to identify
potentially relevant facilities — provides information only on which pollutants a facility is required to
monitor for. Only for a very few facilities does PCS provide reliable information on the amount of the
monitored pollutants that are discharged.  PCS provides no information on whether a facility discharges
unmonitored pollutants or their amounts.

        We tested three alternative approaches for determining whether or not a specific point source (after
meeting applicable technology-based requirements) discharges the impairment pollutant in an amount
sufficient to contribute to impairment and making the source likely to be considered in the TMDL or water
quality-based permit.

I. USE OF PCS INFORMATION: "SCREEN #1"

        Our first approach to making this judgment was to use the information in PCS on frequency of
monitoring to judge by 4-digit SIC code whether each upstream facility is likely to discharge each
impairment pollutant.  If the facility was judged to discharge the impairment pollutant, we would then
assume conservatively that the TMDL  would require the facility to provide further controls beyond
technology-based standards. This approach, which we term Screen #1 rests on three assumptions:

        •      Assume that if facilities in a SIC  generally monitor for a particular pollutant, all facilities
               in that SIC likely discharge that pollutant.  This assumption is probably nearly always
               accurate.

        •      Assume also the converse: if facilities in a SIC rarely monitor for a pollutant then all
               facilities do not discharge the pollutant. This assumption will often be untrue.

        •      Assume that if a facility  discharges the impairment pollutant, the TMDL will likely require
               further controls beyond technology-based standards.  This assumption is likely often
               untrue, but yields a conservative cost estimate.

Consistent with these assumptions, Screen #1 was implemented as follows:

        •      If no facilities in the SIC monitor for any pollutant, discard all facilities in the SIC. No
               such facilities will be required to  control further by the TMDL.

        •      If less than 15% of the facilities in the SIC monitor for the pollutant, assume that none of
               the facilities without monitoring information discharge that pollutant. Assume that any
               facility that does monitor for the pollutant does discharge it.

                                               D-l

-------
        •       If more than 15% of the facilities in the SIC monitor for the pollutant, assume that all of
               the facilities without monitoring information (as well as those with monitoring information)
               discharge that pollutant

        The result when Screen # 1 was implemented that the roughly 20,000 point sources identified as
located upstream within 25/50 miles of impaired waters declined to 10,838 facilities.57 This Screen
eliminated about half of the identified facilities as being unlikely to discharge the impairment pollutant, and
thus unlikely to be required to control further by the TMDL.

        The results  of Screen #1 included many apparent anomalies.  For example, metal finishers
appeared to need beyond-BPT/BAT controls for nutrients. Facilities in the "fabricated metal products"
industry did not appear to be candidates for further metals controls.

II. USE OF ENGINEERING JUDGEMENT: "SCREEN #2"

        Screen #2 involves  an engineering judgment as to whether each 4-digit SIC is likely, after meeting
BPT/BAT/secondary treatment requirements, to discharge each pollutant at levels that could warrant
potential further control. Two EPA engineers experienced in industrial water pollution control made these
judgments for each of more than 500 different SIC codes.

        An example page among the engineers' judgments follows. The page lists the facilities in SICs that
had passed Screen #1.  For  SIC 8062 (general hospitals), for example, 24 facilities had been identified
using Screen #1 as discharging the impairment pollutant within the relevant distance upstream of an
impaired water body. 4 hospitals discharged BOD upstream of a BOD/DO impairment, 22 discharged
metals upstream of a metals impairment, and 10 discharged a nutrient upstream of a nutrient impairment.
Screen #1 would lead to these hospitals being costed for the next treatment step for these pollutants.
Screen #1 did not identify any hospitals as discharging ammonia or toxic organics upstream of an
impairment for either of these pollutants.  In contrast, the engineers' judgment was that hospitals were
likely candidates for further controls for BOD, metals,  nutrients and toxic organics whenever they were
upstream  within the relevant distance of an impairment for any of these pollutants. The engineers believed
that hospitals were not  a likely candidate for further controls if ammonia was the impairment pollutant. In
Screen #2, engineering judgment was used to over-ride the conclusions drawn from the monitoring-based
Screen #1. Some pollutants were crossed out as being unlikely to warrant beyond-technology-based
controls for some SICs, and other pollutants were added as being likely candidates for further controls even
though the monitoring information for facilities in that SIC suggested that perhaps the pollutants were not
being discharged by the SIC.  Engineering judgments were also provided for several hundred SICs that
Screen #1 had eliminated from consideration altogether because no facilities in these SICs monitored for
any impairment pollutants.

        Screen #2 reflecting the engineers' judgments was ultimately implemented in two alternative ways,
resulting in high and low estimates for the number of point sources likely to be required by TMDLs to
implement further controls.  The high estimate adopts more liberal rules for matching  sources and
impairments than does the low estimate.  For the high estimate, when a water body is impaired by a specific
metal (except mercury) or a specific toxic organic (except PCBs or dioxin) and the facility (based on
engineering judgment for its SIC) is  expected generally to discharge metals or toxic organics, then further
57      These are raw figures, not yet scaled to reflect the incomplete coverage of the analysis.

                                               D-2

-------
controls are assumed to be required for that facility. For the low estimate, the match must be exact in order
for controls to be required at a facility: if a water body is impaired for metals generally then facilities that
discharge metals generally are assumed to require controls, but if a water body is impaired for a specific
metal or toxic organic (e.g.,  zinc, phenol), only those facilities discharging that specific metal or toxic
organic are assumed to require controls.

       Screen #2 resulted in a variety of changes relative to Screen #1. Many SICs were dropped as
unlikely to discharge the impairment pollutant in sufficient quantity to make the SIC a likely target for
beyond-technology-based controls.  Some new SICs were added with many facilities that appear to be
mostly storm water (surface coal mining [600 facilities], crude petroleum & natural gas [46], oil & gas
field services [40]) and may not be reasonable targets for further controls under TMDLs.  Other SICs were
added that  may reasonably be targets for further controls: private households (11 facilities) and car washes
(23).
                                                D-3

-------
INSERT PHOTOCOPY PAGE OF ENGINEERS' JUDGMENT
                              D-4

-------
        Ultimately Screen #1 and Screen #2 (high and low estimates) gave surprisingly similar results, in
terms of the number of point sources likely to need further controls under TMDLs and in terms of
implementation costs for these sources:

                                           Exhibit D-l
                        Summary Results of Alternate Screening Processes
Screen
#1, based on PCS monitoring information
#2, engineering judgment, higher estimate
#2, engineering judgment, lower estimate
# of Point Sources
Incurring Costs
9,101
9,492
8,533
Annualized Projected Costs
(1999 $ in millions)
1,499
1,378
1,141
        On the whole, we believe that the pattern of pollutant/SIC combinations for which further controls
are projected to be necessary is most reasonable under Screen #2 (higher estimate).  We adopted this
screening approach for our final cost estimating procedure. Note that the three screens give very similar
results despite rather different underlying rationales, lending some credence to our conclusions.

        Also note that this comparison of alternate screening processes was developed relatively early in
the implementation cost analysis. We made several important changes in our analytical procedures
subsequent to choosing among the three screening approaches. As a result, the estimates shown in Exhibit
B-labove are not comparable to other estimates shown elsewhere in this report.  Most importantly, the cost
functions for next step treatment technologies that we ultimately use in the cost analysis are different from
the preliminary cost functions that underlie Exhibit D-l and that we were using at the time we made the
choice among screening processes.
                                               D-5

-------
     APPENDIX E: DETAIL ON COST FUNCTIONS FOR POINT SOURCES

       This appendix provides additional information on all the cost functions used in the point source
analysis.  The functions themselves are also shown.

I. CAPITAL COSTS FOR POTWs AND INDUSTRIAL DISCHARGERS FOR "NEXT STEP'
TREATMENT

       The relevant CWNS capital cost functions are as follows:

       Capital Cost = $1,000,000 x A x (flow/

where A and B are coefficients selected from the exhibit below based on the type of function requested, and
flow is expressed in million gallons per day.

                                         Exhibit E-l
                          CWNS Capital Cost Functions (1996 dollars)
Function
Secondary treatment
Secondary treatment w/ nutrient removal
Advanced treatment 1
Advanced treatment 1 w/ nutrient removal
Filtration
Secondary treatment salvage
flow < 0.35 mgd
A
1.188
NA
NA
B
0.396
NA
NA
flow 0.35 to 1 mgd
A
5.488
5.903
7.215
8.651
0.770
1.195
B
1.854
1.923
2.114
2.287
0.527
0.840
flow > or = 1 mgd
A
5.488
5.903
7.215
8.651
0.770
1.195
B
0.673
0.754
0.740
0.728
0.527
0.840
The CWNS applies these capital cost functions as follows:

       •      For most plants with flow greater than 0.35 mgd currently providing secondary treatment,
              the cost of an upgrade is estimated as the cost of a new plant at the desired level of
              treatment (e.g., advanced treatment I) as given by the correct function, less the salvage
              value of the former secondary treatment facility, as given by the "secondary treatment
              salvage" function.

       •      For plants employing lagoon systems with no mechanical process units (generally smaller
              plants), the cost to upgrade from secondary treatment is given by the filtration cost
              function.

       •      For plants of less than 0.35 mgd flow, the cost to upgrade from secondary treatment is
              0.35 x the cost of a new plant, as given by the function shown above for these small plants.

       The capital costs for treatment upgrades estimated using these functions are in January, 1996
dollars.  They are updated to March, 2000 dollars by multiplying by 1.0662 to represent the change over
this period in the producer price index for materials and components for construction.
                                             E-l

-------
II. O&M COSTS FOR POTWs AND INDUSTRIAL DISCHARGERS FOR "NEXT TREATMENT
STEP'

        Our cost estimate is developed by subtracting a regression equation estimating the O&M cost for
secondary treatment as a function of flow from another regression equation estimating the O&M cost for
tertiary treatment as a function of flow. The two equations are as follows:

                O&M cost for secondary treatment ($/yr) = 261,396 x flow0'855           (1)

                O&M cost for tertiary treatment ($/yr) = 399,749 x flow0 78°              (2)

where flow is measured in mgd.

        Costs are given in these equations in 1999 dollars.  We update to March, 2000 dollars based on the
producer price index for finished goods (factor of 1.0286).

        The O&M cost increment in upgrading from secondary to tertiary treatment is given by equation
(2) minus equation (1). Although this gives the cost difference in upgrading from secondary to tertiary
treatment as AMS A defines it, we use this relationship also to estimate the cost of upgrading from
secondary to advanced secondary treatment.

        There is an important quirk in this relationship.  At flows exceeding about 80 mgd, the relationship
predicts that the O&M cost differential will decline as flow increases.  The relationship predicts, for
example, that the O&M cost increase in upgrading a 100 mgd plant would substantially exceed the cost
increase in upgrading a 200 mgd plant. We believe this is unlikely to be true in reality.  We have adjusted
the O&M cost function for upgrades to linearize the relationship at flows in excess of 80 mgd.  At flows
greater than 80 mgd, we assume that the O&M cost increase varies linearly with flow. Thus, for example,
we assume that the O&M cost increase when upgrading a 120 mgd plant is 1 Vz times the O&M cost
increase that would be experienced in upgrading a 80 mgd plant.

        There are several complexities in developing the regression equations using AMSA's data.  AMS A
presents information by sewerage authority, not by plant.  Thus, AMSA might report that an authority
operates 3 plants providing secondary treatment and 2 plants providing tertiary treatment, and AMSA then
gives data on the total flow and total O&M costs for the 3 secondary plants and on total flow and total
O&M costs for the 2 tertiary plants. Lacking data on individual plants, we thus perform the regressions on
data aggregated at the authority level.  Assuming that there is some non-linearity in the relationship
between flow and O&M cost for  an individual plant, performing regressions with the data aggregated at the
authority level results in increased statistical imprecision.

        AMSA also excludes sludge management costs in reporting the authority's O&M costs.  Sludge
management costs are reported in total for the entire authority, and are not broken out and attributed
separately to the authority's primary, secondary and tertiary plants. We thus don't know the authority's
full O&M costs, including sludge costs, for its plants providing secondary treatment and for its plants
providing tertiary treatment. We address this  shortcoming in a rough manner by allocating sludge
management costs to the authority's primary,  secondary and tertiary plants in proportion to their flow.  If,
for example, sixty percent of an authority's  flow occurred in secondary plants and forty percent in tertiary
plants, we would allocate 60 % of the sludge costs to the secondary plants and 40 % to the tertiary plants.
                                                E-2

-------
This proportional allocation is inaccurate; a tertiary plant will likely generate significantly more sludge than
a secondary plant of the same flow.

       Nevertheless, the regression equations come out quite well despite these rough adjustments. The
regression equations explain a high fraction of the variance in the dependent variables (secondary and
tertiary O&M costs), and the coefficient estimates for the independent variables (secondary and tertiary
flow) are both significant at beyond the 99 % level.

III. TREATMENT FOR METALS FROM DIRECT DISCHARGERS EXCEPT POTWs

       Polishing multi-media filtration was assumed as the "next treatment step", assumed to be
incremental over the technologies assumed to be in place to meet BAT (flow reduction, chemical
precipitation, clarification).  The capital and O&M cost functions for polishing filtration are drawn from
EPA's development document for the centralized waste treatment industry.58  The equations are:

               Capital cost:    ln(Yl) = 12.0126 + 0.480251n(X) + 0.04623(ln(X))2

               O&M cost:     ln(Y2) = 11.5039 + 0.724581n(X) + 0.09535(ln(X))2

where: Yl = capital costs (1989 dollars)
       Y2 = O&M costs (1989 dollars/yr)
         X = flow rate (mgd)

       To update to March, 2000 dollars, we multiplied the capital costs by 1.2473 (PPI for materials and
components for construction) and O&M costs by 1.2042 (PPI for all finished goods).

       The applicable flow rate range for these equations is 0.023 - 1.0 mgd. For a discharger with flow
needing additional metals treatment of less than 0.023 mgd, we assumed costs as if flow was 0.023 mgd.
For a discharger with flow needing additional metals treatment of greater than 1.0 mgd, we assumed a
proportional increase in costs relative to costs for treating a flow of 1.0  mgd. Thus, for example, for a
discharger needing to treat 2.0 mgd, we assumed that costs would equal twice what the EPA cost equations
gave for a flow of 1.0 mgd.

IV. TREATMENT FOR METALS FROM POTWs

       An enhanced pretreatment program with tighter local limits (tighter than PSES) for significant
metals indirect dischargers was assumed as the "next treatment step" when POTWs need to provide
enhanced control of metals.  The enhanced pretreatment program was assumed to be incremental over a
baseline pretreatment program in which local limits match effluent guideline requirements for indirect
dischargers.

       The key steps in developing the cost function for this enhanced pretreatment program were as
follows:
58     U.S. EPA, Office of Water. Development Document for Proposed Effluent Limitations Guidelines and
       Standards for the Centralized Waste Treatment Industry. December, 1998.

                                              E-3

-------
       •       We assumed that any major POTW within the 50 mile distance upstream of a metal-
               impaired water will need to implement an enhanced pretreatment program.  We assumed
               that no minor POTWs will need to implement such a program..

       These two assumptions may overstate the number of POTWs that will ultimately need to enhance
their pretreatment programs as a result of TMDLs. There are approximately 1,600 POTWs required to
have pretreatment programs. All of them are majors.  They account for about 30 billion gpd of flow, or
approximately 80 % of total national POTW flows. There  are many additional major POTWs that are not
required to have pretreatment programs, but we could not in the time  available single out the pretreatment
POTWs and cost metals controls only for them.

       •       For each major POTW needing to implement an enhanced pretreatment program, we
               determined from the CWNS data base the POTW's average daily industrial flow.

       •       EPA's RIA for the Great Lakes Water Quality Guidance59 estimated that the typical
               POTW needing to meet more restrictive Guidance-based limits for toxic pollutants would
               focus its tightened pretreatment requirements on 30 % of its significant industrial users
               (SIUs).60 We adopted this estimate, and further assumed that 2/3 of such SIUs might be
               affected by tighter requirements primarily for metals, while 1/3 of the affected  SIUs would
               be affected primarily by tighter requirements for other toxic pollutants (organics, cyanide,
               pesticides, etc.)  Thus, we assumed that a typical POTW facing tighter metals discharge
               requirements will require further metals treatment for 20 % of its SIUs (30 % x 2/3) and
               20% of its SIU flow.

       •       We assume that half of total industrial flow to pretreatment POTWs is from SIUs, while
               half is from other industrial users. Across  the major  POTWs identified in our analysis as
               needing enhanced pretreatment for metals, industrial  flow averages about 20 % of total
               flow. Thus, we assume that SIU flow averages about 10 % of flow for these POTWs.

       •       If this 10 % average figure holds also for all pretreatment POTWs nationally, the average
               SIU has a flow of approximately 0.1 mgd.  (30 billion gpd in total pretreatment POTW
               flow) x (10 % of this flow that is attributable to SIUs) 4- (32,000 SIUs) = approximately
               0.1 mgd/SIU.

       •       For each POTW needing an enhanced pretreatment program for metals, we take the
               assumed 20 % of SIU flow that will need further metals treatment (equivalent to 10 % of
               the POTW's industrial flow) and divide this by the average flow for an SIU (0.1 mgd).
               This calculation yields the number of SIUs that will need to implement enhanced treatment
               for metals at each POTW.  For example, assume a POTW with 20 mgd total flow and 35
59     RCG/Hagler Bailly, Inc. Regulatory Impact Analysis of the Final Great Lakes Water Quality Guidance.
       Final Report, March, 1995.

60     The 1,600 pretreatment POTWs serve 270,000 industrial users, of which 31,842 are SIUs. The average
       pretreatment POTW thus serves about 169 industrial users, of which about 20 are SIUs.  SIUs are defined
       as those indirect industrial dischargers that are either: 1)  subject to categorical standards (14,914 SIUs) or
       2) greater than 25,000 gpd flow or greater than 5 % of POTW flow or have reasonable potential for pass-
       through or interference.

                                              E-4

-------
               % industrial flow (7 mgd). This POTW will have 3.5 mgd of SIU flow, 20 % of which
               (0.7 mgd) will need enhanced treatment for metals. At 0.1 mgd for the average SIU, 7
               indirect discharger SIUs will thus need to adopt enhanced treatment for metals.

        •       We assume that the enhanced treatment for metals that SIUs will implement is polishing
               filtration — similarly as for direct dischargers needing additional treatment for metals.  The
               cost function for polishing filtration for SIUs is the same as the cost function previously
               discussed for direct dischargers.

        •       In sum, the costs for an enhanced pretreatment program for a major POTW needing
               additional control of metals is calculated as follows:

               # of SIUs needing polishing filtration =
               POTW industrial flow (mgd) x 0.5 x 0.2 -f 0.1 (mgd)

               cost for each of these SIUs =
               polishing filtration cost functions given previously, at 0.1 mgd flow

        Such costs apply only for major POTWs. Minor POTWs are assumed unlikely to have important
indirect dischargers and unlikely to be targeted in TMDLs for additional metals treatment.
                                               E-5

-------
       APPENDIX F: PROCEDURES FOR ESTIMATING FLOW NEEDING
                                       TREATMENT
I. DEALING WITH MISSING FLOW INFORMATION

       What if information on flow was not available in PCS/CWNS for a facility for which we wanted to
cost additional controls? This was common. For example, for the facilities identified as discharging the
impairment pollutant within a relevant distance upstream in the "within or upstream" case:

               42.1 % had positive reported flows
                2.6 % had zero reported flow
               55.3 % had no flow reported (mostly industrial dischargers, which are addressed only in
                      PCS)

       We used the flow information in PCS/CWNS for any facility for which flow information was
reported. We assumed that a facility with zero reported flow was exactly that, a permitted facility with
zero discharge (perhaps land treatment or some such). For a facility with no flow information reported, we
assigned it the average flow across all facilities of its SIC code for which flow information was reported.
For facilities in SICs for which no facilities have reported flows, we assigned the average flow observed in
our data base for major industrials (2.505 mgd for metals, 6.313 mgd for non-metals), minor industrials
(0.186 mgd for metals, 0.309 mgd for non-metals), major POTWs (8.537 for metals, 8.137 mgd for non-
metals), or minor POTWs (0.189 mgd for non-metals), as appropriate.

II. DEALING WITH LIKELY INACCURATE OR INAPPROPRIATE FLOW INFORMATION

       Flow information in PCS has not been well quality-controlled, and some very high numbers
reported for facility flows undoubtedly represent errors.61 In other cases, the flow listed in PCS appears
likely to be cooling water or perhaps storm water, not process waste water.  It is extremely unlikely that
anyone would consider applying the advanced control technologies that we are costing to huge cooling
water or storm water flows.

       In order to avoid estimating costs for technologies to treat clearly unreasonable flows, we adopted
the following assumptions:

       •       Generally, dischargers with large flows will segregate their waste streams in order that
               advanced treatment (polishing filtration for metals, or advanced secondary for BOD,
               nutrients, ammonia or toxic organics) can be applied to only the smaller, more
               concentrated portions of their total flow.
61     Flow information that is reported in CWNS for the POTWs appearing in CWNS appears to have fewer
       problems than the flow information in PCS.

                                              F-l

-------
We assumed that a minor industrial NPDES permittee will need to apply advanced
treatment for the lesser of its flow reported in PCS and these flows:

       For metals                                  1 mgd
       For BOD, nutrients, etc.                      5 mgd

We assumed that a major industrial NPDES permittee will need to apply advanced
treatment for the lesser of its flow reported in PCS and these flows:

       For metals                                  5 mgd
       For BOD, nutrients, etc.                      25 mgd

However, for electric utilities specifically (SIC 4911 and related), whether they are major
or not, we assumed they will apply advanced treatment for the lesser of their flows
reported in PCS and these flows:

       For metals                                  1 mgd
       For BOD, nutrients, etc.                      0.5 mgd

We adjusted any flow shown in CWNS/PCS for a minor POTW that exceeded 1 mgd (the
maximum that a minor POTW can have) downward to the 1 mgd limit.  (Two minor
POTWs, for example, were shown in PCS as having flows of 55 mgd and 1003 mgd.)
                              F-2

-------
    APPENDIX G: EXCLUDING COSTS FOR POINT SOURCES AFFECTING
       WATERS  THAT ARE IMPAIRED BY NONPOINT SOURCES ONLY

       Many States provide an assessment of the source types that impair each of their 303(d) waters.
For the States that provide this information, the data can be compiled to determine whether each water is
impaired by point sources only, by nonpoint sources only, by mixed point and nonpoint sources, by
unknown sources, or by other categories of sources.  Other States do not provide such information on their
impaired waters. In our analysis to this point, we have identified all point sources that presumably
discharge an impairment pollutant upstream within a relevant distance upstream of or into an impaired
water.62 This set of point  sources is likely the maximum that potentially might be addressed in water
quality-based permitting or TMDLs.

I. USING STATE DATA ON SOURCE TYPES TO IDENTIFY LIKELY POINT SOURCES

       However, we believe we can use the information that States provide on source types responsible for
the impairments in these waters to judge which of the maximum set of potentially relevant point sources are
really likely to be addressed in the eventual TMDLs or water quality-based permits.  As the most obvious
examples:

       •       When a State indicates that point sources are the source of impairment in a water body, we
               expect that the TMDL or alternative approaches would likely consider all the point sources
               discharging the impairment pollutant within a relevant distance upstream (i.e., all of the
               point sources we have identified in this analysis); but

       •       When a State indicates that nonpoint sources - and not point sources - are the source of
               impairment in a water body, we expect that the TMDL or water  quality-based permitting
               process most likely would not address point sources, even if there are some point sources
               that apparently (according to our engineering judgment approach) discharge the
               impairment pollutant within a relevant distance.

       In this step of the analysis, we use the information provided by States on the source types
responsible for impairment of each 303(d) water body to reduce the set of point sources that could
potentially be considered in TMDLs or water quality-based permitting (the maximum set) down to a
smaller set of point sources that are likely to be considered in these processes.

       In the "within and upstream" case, there are 4,234 impaired water bodies that we identify as
impaired by one or more of the five pollutant classes that we analyze (BOD, nutrients, metals, toxic
organics, and ammonia) and that have one or more point source dischargers within a relevant distance
upstream that presumably discharge the impairment pollutant.63 The following exhibit shows what States
62     We say "presumably" because we do not have information particular to each point source on whether it
       actually does or does not discharge any given pollutant.  Instead, we make a judgment for each SIC as to
       whether all dischargers in that SIC presumably discharge each class of pollutants. These broad judgments
       may be inaccurate with respect to any particular point source discharger.

63     These 4,234 water bodies are for the "within and upstream" case. They represent a little more than one-
       quarter of the 16,143 impaired water bodies that we cover in our analysis. (Our procedure for matching
       impaired  water bodies against point sources is operable in 44 States plus the District of Columbia. In

                                              G-l

-------
report as the sources of impairment for these water bodies, and also shows, in the final column, our
judgment as to whether point sources affecting these water bodies are likely to be addressed in TMDLs and
water quality-based permitting.

                                            Exhibit G-l
 Sources of Impairment Reported for Water Bodies That Have Point Sources Within/Upstream That
                          Presumably Discharge the Impairment Pollutant
Sources of Impairment Reported by
States
PS only
NPS only
other only
unknown only
not reported
PS + NPS only
PS + NPS + other only
PS + NPS + unknown only
PS + other only
PS + unknown only
PS + other + unknown only
NPS + other only
NPS + unknown only
NPS + other + unknown only
other + unknown only
PS + NPS + other + unknown
Total
Number of Water Bodies
(Within and Upstream Case)
141
829
71
110
1,727
368
339
15
53
6
5
425
48
35
25
37
4,234
For These Water Bodies, are TMDLs or Water
Quality-Based Permits Likely to Address Point
Sources?
Yes
No
No
Unclear - scale to these waters
Unclear - scale to these waters
Yes
Yes
Yes
Yes
Yes
Yes
No
No
No
No
Yes

Several aspects of this exhibit need explanation:

        •      The first column provides a comprehensive list of categories into which we have grouped
               the information provided by the States on the sources of impairment for each of their
               303(d) waters.  States use hundreds of different terms or codes in describing sources of
               impairment, ranging from the general (e.g., "nonpoint sources") to highly specific (e.g.,
               "Duenweg Mines Area", "zebra mussels", "1-95 culvert has inadequate passage"). For
               this analysis, Tetra Tech grouped all the source terms and codes in a particular manner.
               "PS" includes any codes referring to discharges from point sources likely to be included in
               PCS. "NPS" includes any source codes involving nonpoint sources that produce
               pollutants that  could reasonably be addressed with controls  in TMDLs (e.g., agriculture,
               silviculture, urban runoff, abandoned mines, road maintenance).  "Other" includes any
               source codes that are unlikely to be addressed with controls in TMDLs, either because they
        these 44 States plus DC, there are 18,162 303(d)-listed waters, of which 16,143 have been georeferenced.)
        Interpreting the "within and upstream" case as an upper bound, we thus believe that at most about 1/4 of
        all TMDLs have the potential to trigger additional controls for point sources. In contrast, the "within
        only" case provides a reasonable lower bound. In the "within only" case, there are XXX water bodies that
        are impaired by one or more of the five pollutants and have one or more point sources discharging the
        impairment pollutant directly into the impaired water. This suggests that there may be as few as XXX
        percent of all TMDLs that have the potential to trigger additional controls for point sources.
                                                G-2

-------
               involve pollution rather than pollutants (e.g., habitat alteration, hydrologic modification),
               or because they are very unlikely to be controllable by a State (e.g., air deposition).
               "Unknown" is where a State reports that a source of an impairment is unknown. "Not
               reported" includes instances in which the State provides  no information of any sort about
               the sources responsible for an impairment.

        •       As shown in the exhibit, States may report several sources of impairment for a single
               water body. The various sources of impairment may all contribute to a single  cause of
               impairment (e.g., both point sources and nonpoint sources may contribute excessive
               nutrients), or one source of impairment may be responsible for one cause while another
               source is responsible for another cause (e.g., a water body is impaired by BOD from a
               point source and by undetermined toxicity from an unknown source), or there can be
               various other combinations.

        •       Unfortunately, there is no information available on the sources of impairment for many
               303(d) water bodies.  Among the 4,234 water bodies identified in the "within and
               upstream" case, for example, more than 43 % have either no source information reported
               or report unknown sources only.

        The final column of the exhibit presents our judgment as to whether point sources affecting these
water bodies are likely to be addressed in TMDLs. For example, if the State cites the water body as
impaired by point sources only, we presume that a TMDL is likely to consider further  control requirements
for all relevant point sources.  The opposite would be true for a water body that the State cites  as impaired
by nonpoint sources only — for this water body, we would not expect a TMDL to consider further controls
for the point sources that we have identified as presumably discharging the impairment pollutant. The
decision rules underlying the conclusions in the final column are:

        •       Point sources will be addressed in TMDLs for any water for which a State cites point
               sources as being among the sources of impairment;

        •       Point sources will not be addressed in TMDLs for any water for which a State: a) Cites
               one or more sources of impairment (except for "unknown"); and b) Does not include point
               sources among the cited sources; and

        •       For water bodies for which a State provides no information on sources of impairment ("not
               reported") or cites only "unknown sources", we have no information suggesting whether
               point sources will or will not likely be addressed in TMDLs.

        We regard the first two sets of waters -- those for which we have information suggesting whether
point sources are among the source types responsible for impairment -- as a sample, and extrapolate data
from them to the third set of waters for which we do not have such information. We assume that the degree
to which various source types are responsible for impairments in the third set of waters is the same as it is
in the first two sets of waters.64
64     We are not aware of any reason why the third set of waters should be systematically different from the first
       two sets of waters. Among the 44 States plus DC that we cover in the analysis, we do not find any
       important differences between the States that do not report 303(d) source information (which contribute
       the great majority of the third set of waters) and those that do.

                                               G-3

-------
II. EXTRAPOLATING INFORMATION FROM WATERS WITH KNOWN SOURCES OF
IMPAIRMENT TO WATERS WITHOUT KNOWN SOURCES OF IMPAIRMENT

        Our process for using information on these three sets of waters in order to estimate point source
implementation costs is rather complex.

        At this point, we introduce a new, "shorthand" terminology.  We refer to an impaired water as
"tagging" a point source if the point source presumably discharges the impairment pollutant directly to the
impaired water (for the "within only" case) or directly to the impaired water or within 25/50 miles
upstream (for the "within and upstream" case).  All the point sources in our analysis are tagged one or
more times by impaired waters.  Many point sources are tagged multiple times. If a water body is impaired
for several pollutants, a particular point source may be tagged multiple times by that water body for
different pollutants.  For example, according to the EPA engineers' judgments, a major POTW is presumed
to discharge all five pollutant classes that we analyze -- BOD, nutrients, metals, ammonia, and toxic
organics -- in sufficient quantity so as potentially to warrant further controls beyond applicable technology-
based standards.  A water body that is impaired for all five of these pollutants thus could "tag" a major
POTW that discharges  into this water body five times, once for each pollutant class. A point source can
also have multiple tags  because it is tagged by multiple water bodies. Particularly in the "within and
upstream" case, there can be several impaired water bodies within 25/50 miles downstream of a point
source. Even in the "within only" case, impaired water bodies may overlap and point sources are
sometimes tagged by multiple water bodies.

        When a point source is tagged multiple times for different pollutants by water bodies  of different
sorts (i.e., by some water bodies that are presumed to impose costs on the point sources they tag, by some
that are presumed not to impose costs on the point sources they tag, and by some that are in the
undetermined and "to be scaled to" category), it becomes complex to  determine whether the point source
should or should not be expected to incur costs for further controls. We take the following steps:

        •      Point sources that are tagged by one or more waters with reported source types become the
               set of facilities that we scale from. We estimate for this set of "scale from" facilities: A)
               What their costs would be if all impaired waters (not just those waters impaired by point
               sources, among other source types) were assumed to impose costs on relevant point
               sources; and B) What their costs are when we assume that only those waters  impaired by
               point sources impose costs on relevant point sources. Among the "scale from" facilities,
               our assumption that point sources do not incur costs  when they are tagged by waters not
               impaired by point sources results in reducing the total estimated implementation costs from
               A to B. Expressed another way, the fraction B/A represents the degree to which the
               potential total cost of all tags (among the "scale from" facilities) gets reduced by
               recognizing that some tags do impose costs on point  sources and that some do not.

        •      Point sources that are tagged exclusively by waters that are uncertain in whether or not
               they will impose costs on point sources ("source not  reported" or "source unknown"
               waters) become the set of facilities that we scale to.  We estimate the costs these facilities
               would bear if all the waters that tag them were of a sort that would impose costs on point
               sources (C). We then multiply this raw cost by the fraction B/A, representing the degree
               to which (based on the "scale from" facilities) the raw cost is reduced by recognizing that
               some tags do not impose costs on point sources.
                                               G-4

-------
               The resulting total costs for point sources is given by:

               B + [(B/A) x C].
               where   A = the costs among "scale from" facilities if all tags were to impose costs
                           on point sources
                       B = the costs among "scale from" facilities from tags only by those water bodies
                           that we assume do impose costs on point sources (i.e., water bodies that
                           states cite as impaired by PS, as well as perhaps additional source types)
                       C = the costs among "scale to" facilities if all tags were to impose costs on point
                          sources.

Sample calculations of this sort are shown in the exhibit below, for Scenario 1 ("No TMDL Program")

                                           Exhibit G-2
       Calculations Reflecting Information from States on Sources of Impairment (Scenario 1)
All tagged facilities, assuming all water bodies impose costs on point
sources
A. "Scale from" facilities (those tagged by water bodies with information
on responsible source types)
B. "Scale from" facilities that do incur costs (tagged by water
bodies impaired by PS)
"Scale from" facilities that don't incur costs (tagged by water
bodies not impaired by PS)
Scaling factor: B/A
C. "Scale to" facilities (those tagged exclusively by water bodies
without information on responsible source types)
"Scale to" facilities that will incur costs, after application of scaling
factor: (B/A)xC
Total estimated costs and number of facilities: B + (B/A)xC
Within Only Case
Costs (billion S/yr)
$1.684
$1.060
$0.681
$0.379
0.64
$0.624
$0.401
$1.082
#ofPS
7012
4324
2557
1767
0.59
2688
1590
4146
Within and Upstream Case
Costs (billion S/yr)
$3.369
$2.078
$1.343
$0.734
0.65
$1.291
$0.835
$2.178
#ofPS
19438
12086
7394
4691
0.61
7353
4498
11893
        Referring to the figures for the "within only" case, if we were to assume that water quality-based
permits would be developed for all tagged point source facilities, even those that discharge into waters
impaired by nonpoint sources only, 7012 point sources would incur costs of $1.684 billion/year. However,
we assume that tagged point source facilities that discharge only into waters impaired by nonpoint sources
will not have water quality-based permits issued for them, and costs thus decline to $1.082 billion/year for
4146 facilities.  Our assumption that waters impaired by nonpoint sources only will not give rise to control
obligations for point sources has the effect in this case of reducing costs by 36% and the number of point
sources affected by 41%.  This assumption has roughly similar impacts in reducing costs and the number
of sources affected for other cases (i.e., "upstream and within") and Scenarios (i.e., Scenarios 2 and 3).
                                               G-5

-------
 APPENDIX H: DETAIL ON SIMULATING "COST-EFFECTIVE WASTE LOAD
                                    ALLOCATIONS"

I. CIRCUMSTANCES WHERE COST-EFFECTIVE WLAs MAY BE POSSIBLE

       We assumed that a cost-effective WLA could occur only if both point sources and nonpoint
sources were believed by the State to be responsible for impairment of the water body. If point sources
alone were responsible for the impairment, there would presumably be no or few contributing nonpoint
sources to whom some of the point sources' responsibilities for further controls could be shifted.  The
following exhibit shows the source combinations for which we assumed cost-effective WLAs to be
possible:

                                         Exhibit H-l
      Impairment Source Combinations for Which Cost-Effective WLAs are Assumed Possible
Sources of Impairment Reported by
States
PS only
NPS only
other only
unknown only
not reported
PS + NPS only
PS + NPS + other only
PS + NPS + unknown only
PS + other only
PS + unknown only
PS + other + unknown only
NPS + other only
NPS + unknown only
NPS + other + unknown only
other + unknown only
PS + NPS + other + unknown
Total
Are TMDLs for These Water
Bodies Likely to Require Further
Controls for Point Sources?
Yes
No
No
Unclear - scale to these waters
Unclear - scale to these waters
Yes
Yes
Yes
Yes
Yes
Yes
No
No
No
No
Yes

Are Point Sources Affecting These
Waters Then Candidates for Cost-
Effective WLAs?
No
No
No
Unclear - scale to these waters
Unclear - scale to these waters
Yes
Yes
Yes
No
No*
No*
No
No
No
No
Yes

       The final column of the exhibit indicates that we assume cost-effective WLAs to be possible only
for waters for which the State indicates that both point and nonpoint sources are responsible for the
impairment.  Waters for which there is no information about the source types responsible for impairment
are scaled to.

       Note that point sources may be "tagged" as needing further controls by multiple water bodies.  We
assume that a point source can benefit from a more cost-effective WLA and avoid the next treatment step
for a particular pollutant if and only if:

       •      It is tagged by at least one "yes" water body for that pollutant (i.e., the point source needs
              further treatment for a pollutant as a result of a water body that also has nonpoint sources
              that contribute that pollutant)
                                            H-6

-------
        •       It is not tagged by any "no" water bodies for that pollutant. A single "no" water body that
               tags a point source will preclude the point source from avoiding the next treatment step for
               that pollutant.

        •       "No*" water bodies have no effect. They represent situations in which it is not clear
               whether or not cost-effective WLAs will be possible — point sources will incur costs (the
               first condition for possible trade offs), but it is uncertain whether or not there will be
               nonpoint sources also found to contribute to impairment.  Perhaps the "unknown" sources
               will eventually prove to be nonpoint sources and a cost-effective WLA will be possible,
               perhaps not.  In our simulation, a tag by a "no*" water body neither enables cost-effective
               WLAs nor precludes them.

        The assumption that a tag by a single "no" water body precludes a source from participating in a
cost-effective WLA is conservative and has an important effect in reducing the number of point sources
that we estimate as likely to be able to reduce their needs for additional treatment. Many point sources are
tagged by several water bodies, all but one  of which would be amenable to cost-effective WLAs.  We
assume, however, that the one non-amenable water body precludes such sources from avoiding the next
treatment step for the impairment pollutant.

        We assumed there would be an opportunity for a more cost-effective WLA only for water bodies
impaired by nutrients and/or BOD.  However, even if a water body is impaired for one of these pollutants
and is amenable to a cost-effective WLA, a point source discharging one of these pollutants may not be
able to participate in the more cost-effective WLA if the point source discharges another pollutant that
requires the point source to implement the next treatment step notwithstanding the cost-effective WLA.

        More specifically, the assumed next treatment step for nutrients (secondary treatment with nutrient
removal) is also the next treatment step that we assume is required if the point source needs to provide
further controls for toxic organic compounds. Thus, a point source that needs further control of both
nutrients and toxic organics will need to implement secondary treatment with nutrient removal to address
the toxic organics, whether or not there is a cost-effective WLA for nutrients.  In effect, a need to control
toxic organics prevents a point source from saving any money by participating in a cost-effective WLA for
nutrients.  Likewise, a need to control ammonia prevents a point source from saving any money (avoiding
the need for advanced secondary treatment) by participating in a cost-effective WLA for BOD. In sum, we
assume that a point source may avoid the need for a next treatment  step by participating in a more cost-
effective WLA if the point source needs additional control for:

        •       Nutrients and not toxic organics; and/or
        •       BOD and not ammonia.

II.  COSTS FOR CONTROLLING NONPOINT SOURCES RELATIVE TO POINT SOURCES

        The following are several examples suggesting the magnitude of costs for nonpoint sources to
abate a pollutant load in comparison to the  costs for point sources to abate a similar load.  Most of these
examples involve point/nonpoint effluent trading.  Nearly all involve nonpoint  source controls substituting
for additional point source control efforts for nutrients or BOD. Most of the following are prospective
studies of the savings if trading were to be implemented. Few are retrospective estimates of what actual
trades have saved.
                                               F-l

-------
Long Island Sound, CT - The State of Connecticut estimated that to reach the 15 year goal of nitrogen
reduction in the absence of trading would cost point sources (mostly POTWs) $960 million, while the
inclusion of trading would reduce costs to $760 million. Thus trading is expected to result in a $200 million
savings over 15 years.  The great majority of the trading simulated here is point/point trading, not
point/nonpoint.

Tar-Pamlico, NC - A Great Lakes Trading Network report cites an estimate by a member of the Tar-
Pamlico Basin Association that nutrient reduction for point sources would cost approximately $70 million
compared with $11 million for similar reductions through increased NFS  controls.

Saginaw Basin, MI - Modeling performed by World Resources Institute. Establishing a 0.5 mg/L
phosphorus limit for point sources would reduce loadings by 16% at a cost of $23.89/lb reduced.  A broad
subsidy program to non-point sources would cost $16.00/lb reduced. The least cost solution - a targeted
and performance-based subsidy to only some non-point sources - would reduce loadings 15% and cost
$1.87/lb. What was identified as a best solution (not  the least cost) would be a combination of point source
controls and point/non-point trading (including targeting the most cost-effective NFS reductions) using 2:1
ratios;  this would reduce loadings 26% and cost $4.37/lb reduced.

Minnesota River Basin, MN - Modeling performed by WRI, very similar to the Saginaw study. The Study
projects that the costs per Ib of controlling P could be reduced from $18 for point source controls only to
$4-5 for trading coupled with "targeted" BMPs.

Boulder Creek, CO - Estimated savings (thought 12/96) of $3 - $7 million as a result of in-stream
restoration efforts rather than plant upgrades alone. No information regarding what percentage savings this
represents.

Virginia Tech Study (Watershed'96) - Paper showed  a range of non-point source control costs typical for
Virginia compared with Biological Nitrogen Removal. BNR costs were listed as $20-$50/lb.  Most non-
point source costs listed were less expensive; some were more expensive.

Lower Boise River, ID  - Municipalities were asked to consider what their 20 year plan would be in the face
of a mandated 20%, 40% or 80% phosphorus reduction, given a low population growth scenario and a high
population growth scenario. The municipalities responded that the cost range for an 80% reduction would
be $12-$178/lb of P.  For non-point sources, the stakeholders looked at a few BMP studies that focused
primarily on sediment reductions at sites on the Boise River and the mid-Snake River, and factoring in what
is known about the sediment-phosphorus relationship, determined that the costs of phosphorus reductions
would be in the range of $2-$20/lb of P. The implied  cost savings are therefore $10-$158/lb. of P, but with
some uncertainty as to what will actually occur in the market.

Lake Chatfield, CO - A high level of further point source reductions would be necessary before NFS
reductions and trading become economical.

Stamford, CT (Watershed '96) - It was calculated that reductions from the treatment plant would be more
cost-effective than reductions from non-point sources.
                                               F-2

-------
                                          Exhibit H-2
                 Percent Savings in Cost per Pound for Nonpoint Source Controls
                                Relative to Point Source Controls
Long Island Sound, CT
Tar-Pamlico, NC
Saginaw Basin, MI
Minnesota River Basin, MN
Boulder Creek, CO
Virginia Tech study
Lower Boise River, ID
Lake Chatfield, CO
Stamford, CT
21
84
82-92
75
Unknown
Variable
83-89
Minimal
Negative
Among the seven instances where the percentage savings can be rank-ordered (all except for Boulder Creek
and the Virginia Tech study), the median figure is a 75 % savings.
                                             F-3

-------
  APPENDIX I: DETAIL ON ESTIMATED COSTS FOR NONPOINT SOURCES

       This Appendix provides detail on how the costs for nonpoint sources were estimated for the "Least
Flexible TMDL Program" scenario (Scenario 1). Under this Scenario, we assume when a State identifies a
water body as impaired by a nonpoint source type that the State will require further control for the entire
volume of that nonpoint source activity that occurs within the county or counties in which the impaired
water body is located.  Costs under Scenarios 2 and 3 are estimated by adjusting the information and
estimates developed for Scenario 1 in the manner described in Chapter III, Section C.

I. AGRICULTURE

A. CROP LAND

       We found 2,228 water bodies on States' 303(d) lists that States have identified as impaired by crop
land.  These water bodies were either cited directly in a State's 303(d) submission as impaired by crop
land, or they were cited as impaired by crop land in a State's 305(b) submission and corresponded to a
303(d)-listed water body. These water bodies are in  710 counties65 in 23 States. We assume conservatively
that the remaining States may have crop land-impaired water-bodies, but either did not report sources of
impairment at all or reported them in a manner that was insufficiently specific to identify crop land as a
source of impairment.  The 23 States reporting crop land-impaired 303(d) waters contain 240.3 million
acres of crop land.  In the Nation as a whole, there are 375 million acres of crop land.  We assume that the
degree to which crop land impairs waters in the 23 "reporting" States is replicated in the remaining States.
The scale factor to extrapolate from whatever crop land costs we estimate for the 23 States to the entire
nation is thus 1.56 (375 million divided by 240.3 million).

1. Extent of nonpoint source  activity requiring BMPs

       We determined the amount of crop land in these 710 counties, located in 23 States, using crop land
acreage data from USDA's 1997 National Resources Inventory (NRI) as it existed prior to recently
released corrections. We also  acquired data from NRI on the acreage of this crop land that is eroding at
greater than 15 tons/acre/year, and on the average slope of the crop land in each of these counties.

       There are 128.1  million acres of crop land in the 710 counties within which the crop land-impaired
water bodies are located.  We assume for the "Least  Flexible" scenario that the TMDLs for these impaired
water bodies will require the implementation of BMPs for the entire acreage of crop land in these counties.
The 128.1 million acres of crop land needing BMPs in these 23 States represent 53.3 percent of all the crop
land in the 23 States.  Applying a scale factor of 1.56 to extrapolate from the 23 States to the entire nation,
we estimate that there are 199.9 million acres  of crop land in the nation that will need to implement BMPs
under the Least Flexible  scenario.

2. Description of the BMPs required where controls are needed

       We assumed that owners of crop land identified as contributing to impairment of a 303(d) water
will achieve the load reductions required by TMDLs  by implementing BMPs that are widely available,
relatively low cost, effective, and easily put in place.   Based on these criteria and the ability to slow or
65      3 of the 710 counties have no crop land.

                                               1-1

-------
prevent soil erosion and phosphorus and nitrogen loss from crop lands, five broad groups of practices were
selected for use on crop lands contributing to impairment of 303(d) waters:

        1.       Conservation tillage
        2.       Nutrient management
        3.       Practices to reduce sediment transport within or at the edge of the field
        4.       Practices to protect and restore riparian areas
        5.       Management of highly erosive crop land

These conservation practices or groups of practices are described as follows:

        a)       Conservation tillage. Management of planting and tillage practices and crop residues so
                as to maintain  at least 30 % of the soil surface covered by residue after planting. Three
                varieties of conservation tillage include no-till, mulch till and ridge till.  Conservation
                tillage increases the water-holding capacity of the soil and reduces sheet and rill erosion.

        b)       Nutrient management. Balancing the amount of nutrients provided with crop nutrient
                requirements so as to prevent nutrient loss to the environment. Requires proper timing of
                nutrient application and control of the amount of nutrients applied.

        c)       Practices to reduce sediment transport within or at the edge of the field.  In general,
                physical measures to reduce slope length and steepness (e.g., contour farming, terraces,
                contour strip-cropping) reduce runoff velocity and increase infiltration. Vegetative
                measures (e.g., grassed waterways, filter strips, vegetative barriers, field borders) will also
                slow runoff, increase infiltration  and trap sediment.  A wide variety of different practices
                can be appropriate in different circumstances.  In order to develop a reasonably
                conservative (i.e., not likely to be too low) estimate for the costs of applying appropriate
                practices to reduce in-field or edge-of-field sediment transport, we will assume costs as if
                vegetative barriers were to be implemented.  Vegetative barriers are intended to reduce
                concentrated (gully) erosion on sloped crop land by planting a series of 4 - 6 feet wide
                barriers of stiff, tall grasses across a field at specified increments of elevation. Vegetative
                barriers entail installation and management costs that likely exceed the costs for most other
                in-field or edge-of-field measures to reduce  sediment transport.

        d)       Practices to protect and restore  riparian areas. These may include practices that
                intercept runoff before it reaches streams or lakes (e.g., riparian filter strips, riparian forest
                buffers) and practices that protect the waterway's banks and channel (e.g., streambank and
                shoreline protection, livestock exclusion, stream channel stabilization).  Again in order to
                develop a reasonably conservative estimate for the costs of such practices, we will assume
                costs as if riparian forest buffers were to be implemented.  Riparian forest buffers include:
                1) a permanent zone of trees and shrubs adjoining the water, 2) an additional adjoining
                zone of trees and shrubs from which modest harvesting may occur; and 3) an  additional
                vegetated filter strip when the buffer is adjacent to cropland or other erosive areas.  Such a
                corridor of permanent vegetation on the banks of a water body will reduce sediment,
                nutrients and other pollutants in surface runoff, and will also reduce excess nutrients and
                other chemicals in shallow ground-water flow. The riparian buffer will also create  shade
                to lower water temperatures, stabilize banks, provide wildlife habitat, and protect against
                                                 1-2

-------
               scour erosion within the floodplain. Establishment costs for riparian forest buffers are
               likely higher than other riparian measures to intercept runoff.

        e)      Management of highly erosive crop land.  A variety of more intensive measures can be
               targeted specifically for crop land that is eroding at high rates.  One possibility is to retire
               such crop land: take it out of production and establish a permanent vegetative cover on it.
               Retirement, conservation crop rotations, contour strip cropping, terracing and other
               measures will substantially reduce losses of soil and agricultural chemicals.  In estimating
               the costs of this practice, we will assume that all crop land eroding at greater than 15
               tons/acre/year (as estimated in NRI) will need one of these more intensive measures.
               Again, in order to be conservative in costing, we will estimate costs as if the selected
               measure is to retire the land and establish permanent vegetative cover on it.

        For most crop land-related water quality problems, the combination of all these selected practices
will be more than enough to mitigate the contribution of crop land to impairment. TMDLs requiring
reductions of sediment or nutrients from agricultural land have typically required reductions in the range of
20 - 70 %.66  EPA cites the following rough estimates of the effectiveness of some of the practice groups
we are assuming will be applied:67

                                            Exhibit 1-1
             Relative Gross Effectiveness of Sediment Control Measures (% reduction)
Practice Category
Reduced tillage systems (e.g., conservation
tillage)
Diversion systems (e.g., grassed waterways)
Terrace systems
Filter strips (vegetative control measures)
Total Phosphorus
45
30
70
75
Total Nitrogen
55
10
20
70
Sediment
75
35
85
65
        These effectiveness figures are for each of the practices applied individually. We assume,
however, that  several of these practices will be applied simultaneously (conservation tillage plus in/edge-
of-field measures plus riparian measures), as well as additional measures not among those shown in the
table (nutrient management planning plus management of highly erosive crop land). In most
circumstances, the entire package of practices that we assume and then estimate costs for will be more than
enough to achieve the desired reduction in load from crop land. This fits with our aim to be generally
conservative in costing.  In some instances, however, the water quality problem caused by crop land may
not be addressed by the group of practices we assume.  For example, water quality problems relating to use
of agricultural herbicides may require application of BMPs other than these (e.g., integrated pest
management).  We have also not considered any of the specific measures that might apply particularly to
66      We will update this range when we complete the review we plan to do of additional completed TMDLs for
        "ground-truthing" purposes.

67      U.S. EPA.  National Management Measures to Control Nonpoint Source Pollution from Agriculture.
        Page 101. (http://www.epa.gov/owow/nps/agmm/chapter_4)
                                                1-3

-------
irrigated agriculture.68 Note, though, that even in these instances where our set of practices might not be
sufficient (e.g., pesticides, irrigation), the set of practices that we do simulate will at least contribute
substantially to diminution of the water quality problem. For example, the practices we simulate that
reduce sediment losses and transport will also sharply reduce direct runoff of pesticides and transport of
pesticides adsorbed to soil particles.

        In assuming the application of this full set of practices in our cost analysis, we do not mean to
imply that all these particular practices should necessarily or always be chosen by farmers to meet the load
reduction requirements of TMDLs. Conservation tillage, for example, is obviously not appropriate for
crops that do not involve tillage.  The choice of appropriate practices will be highly site-specific, depending
on both the nature of the agricultural operation and the nature and severity of the water quality problem.
We intend this chosen set of practices as a reasonably comprehensive set of measures that will be more
than sufficient to address the great majority of water quality problems from crop land. We might attempt
to be more specific about additional tailored practices that could be appropriate in particular circumstances
(e.g., for irrigated land, for tile-drained land, for different sorts of crops), but we judged that our current
level of detail is reasonable for a nationwide analysis that involves broad uncertainties in virtually all the
other steps in developing cost estimates.

        In estimating the costs of applying this group of practices  to crop lands that contribute to water
quality impairment, we have further assumed particular BMPs that we use to represent the cost of each
practice group.  Each practice group includes a range of possible BMPs (e.g., in-field or edge-of-field
measures to reduce sediment transport can include many different contouring, buffer and runoff
management measures).  For costing purposes, though, we have chosen a single, relatively expensive BMP
to represent what it might cost to implement whatever specific BMP among the practice group is
appropriate  in each individual circumstance. The specific BMPs we selected for costing purposes are as
follows:
                                             Exhibit 1-2
             Specific BMPs Selected to Represent Practice Groups for Costing Purposes
Practice Group
Conservation tillage
Nutrient management
Practices to reduce sediment transport within or at the
edge of the field
Practices to protect and restore riparian areas
Management of highly erosive crop land
BMP Selected for Costing
Mix of no-till/ridge till/mulch till
Nutrient management planning: plans, soil testing
Vegetative barriers (discrete)
Riparian forest buffers
Retirement of land and establishment of permanent
cover
68      The great majority of States do not identify sources of agricultural impairment with sufficient specificity to
        indicate whether the crop land is irrigated or whether irrigation somehow contributes to the problem.
        Even if we wanted to simulate the application of a set of practices specifically for irrigated crop land, we
        would have difficulty determining where these practices might need to be applied and where they would
        not.
                                                 1-4

-------
        These specific BMPs will be described further in the next section. We have selected these specific
BMPs only for the purpose of estimating costs. We do not mean to imply that these particular BMPs
should or must be applied by farmers in order to mitigate water quality problems.  Again, the selection of
appropriate BMPs in practice should be site-specific.  One or more of our particular selected BMPs may be
poor choices in many circumstances. Riparian forest buffers are not reasonable for areas where trees are
very difficult to plant and grow.  Vegetative barriers are a relatively new and somewhat experimental (but
very promising) measure, and alternative in-field measures should also be considered. We have not
selected these particular BMPs with the intention of being prescriptive.  Instead, we believe these particular
BMPs are reasonably broadly applicable and they are among the more expensive BMPs included within
each of the practice groups, and they therefore provide a relatively conservative indication of what the costs
of abating loads from crop lands might be.

3. Unit costs for These BMPs

        For each BMP in each of our 710 counties, we (1) determined the number of crop land acres to
which the BMP would need to be applied; (2) estimated the cost per acre required to implement the BMP;
and (3) estimated the proportion of acreage that is assumed already to have the BMP in place.  Further
detail on these procedures for each of the five crop land BMPs is provided below.

                                              Exhibit 1-3
                                              Retirement
      Extent BMP is Needed
       Unit Cost for BMP
  Degree Already Implemented
 For all cropped crop land eroding at >
 15 tons/acre/yr in every county with at
 least one crop-impaired water body.
 These data were obtained from NRI.
Assume that Conservation Reserve
Program (CRP) rental payments for
retired crop land equal the actual social
cost of this BMP. Annual rental costs
were estimated at $52.76/acre/yr (the
avg. rate for latest CRP signup).
Establishment cost for cover estimated
at $67.20/acre (low, for non-native
grasses) to $98.86/acre (high, for
native grasses).69 Annual  maintenance
cost assumed as zero (low est.) to
$5/acre (high est. based on MD CREP
data).
Much highly erosive land is already
enrolled in CRP and retired. Such land
is not counted as crop land in NRI.
Thus, none of the additional crop land
we calculate as needing to be retired
(based on NRI figures) has yet been
retired.
69      USD A, comments provided to EPA upon review of draft version of this report, June/July, 2001.

                                                  1-5

-------
                                                 Exhibit 1-4
                                           Riparian Forest Buffer
      Extent BMP is Needed
       Unit Cost for BMP
   Degree Already Implemented
 75-foot wide corridor on each side of
 every crop-impaired water body.
 (NRCS conservation practice standard
 is for minimum of 55 feet plus)
For low est., assume that CRP rental
rate ($52.76/acre/yr for latest signup)
accurately represents social cost of
having land in this use rather than
crop production.  For high est,
assume that USDA CREP rental
payments needed to attract small
tracts represent the social cost
($130/acre).70 Add to this the cost of
establishing trees (assume  $300/acre
for poplar71) plus annual maintenance
cost (assumed at $5/acre/yr based on
MDCREP)	
About 2.2 of the needed 4.2 - 5.0 million
acres of all sorts of buffers are currently
in place. However, more than 90 % of
the buffer acreage in place is grass
waterways and shelterbelts.72  For low
cost estimate we assume 25% of the
needed riparian forest buffer is already
in place, for high estimate we assume
zero.
70      USDA.  Ibid.

71      See Turhollow, A. July 2000.  Costs of Producing Biomass from Riparian Buffer Strips.  Oak Ridge
        National Laboratory, Energy Division.

72       See TetraTech. July 2000.  Draft Nonpoint Source Gap Analysis.

                                                     1-6

-------
                                                Exhibit 1-5
                                            Vegetative Barrier
      Extent BMP is Needed
       Unit Cost for BMP
   Degree Already Implemented
 Barrier 4 feet wide constructed for
 every 5 feet of elevation difference in
 crop field.73 Applied to all crop land  in
 every county with at least one crop-
 impaired water body. The acreage
 that would be needed for continuous
 barriers in each county is calculated
 from crop acreage and slope data in
 NRI, then scaled by 5/14 to  represent
 the lesser acreage that would be
 needed for discrete barriers.74
Similar reasoning as for riparian forest
buffer.  Costs = CRP rental rate or
CREP payment rate for small plots +
establishment cost ($100/acre for
switchgrass75) + annual maintenance
cost. (Note: establishment cost might
be higher for very small plots as would
be in barriers than the figure in the
reference.) Annual maintenance
costs were assumed to range from
$5/acre  (lower estimate, as above) to
$10/acre (higher estimate; SEDLAB
suggests vegetative barriers could
need more costly maintenance than
other sorts of vegetated strips)
No data. Given the newness of this
BMP, we presume that very little is
already in place. We have assumed
that none is in place.
73      Based on SEDLAB briefing. See USDA. Vegetative Barriers: A New Upland Buffer Tool. National
        Sedimentation Laboratory. Upland Erosion Processes Research Unit.
        (http://www.sedlab.olemiss.edu/uep_unit/projects/Dab_veg/index.htm)

74      Ibid.  The SEDLAB briefing simulates both continuous (following the complete field contour line) and
        discrete (placed strategically on only those portions of the field contour line where gully erosion is most
        likely) barriers.

75      See Turhollow, A. July 2000. Costs of Producing Biomass from Riparian Buffer Strips. Oak Ridge
        National Laboratory, Energy Division.
                                                     1-7

-------
                                                 Exhibit 1-6
                                            Conservation Tillage
   Extent BMP is Needed
           Unit Cost for BMP
   Degree Already Implemented
 For all annually planted crop
 land (that does not go into
 retirement, buffer or barrier)
 in every county with at least
 one crop-impaired water body
Low estimate: $12.43/acre in capital costs and
$8.66/acre savings in annual costs.76 (Note:
these figures represent an average for the
corn/wheat/soybean acreage where CT is
already being used. One might expect that the
cost savings from CT would  be much less if it
were applied to the acreage  where it currently
isn't being used. Pending further research, we
will assume $8.66 as an  upper estimate of
savings and $0 as a lower estimate.)
High estimate is based on figures from EQIP
for ridge-till ($13.30/acre), mulch till
($10.60/acre) and no-till/strip-till ($23.70/acre),
assumed to be annual costs.77 These figures
were weighted by the percentage of
conservation tillage land  in each of these three
practices (3.2%, 53%, 43.8%) from CTIC,
1998, to derive an average cost per acre of
$16.43.
37.2% of all planted acres nationally in
1998.78
76      See TetraTech. July 2000. Draft Nonpoint Source Gap Analysis.

77      USDA, 2001, op cit.

78      CTIC. 1998.  1998 Crop Residue Management Survey Report, 1998 United States Summary.
        Conservation Technology Information Center, West Lafayette, Indiana.
                                                     1-8

-------
                                              Exhibit 1-7
                                    Nutrient Management Planning
    Extent BMP is
       Needed
               Unit Cost for BMP
Degree Already Implemented
 For all crop land (that
 does not go into
 retirement, barrier or
 buffer) in every county
 with at least one crop-
 impaired water body.
USEPA estimates $5 - $15/acre/yr for nutrient
management plan and soil testing.79 NPS Gap analysis
assumes $12/acre initial cost plus $7/acre for soil test
every three years.80 We assumed initial cost (for plan) of
$7 - $15/acre and annual costs (testing) of $2.33 - $4/acre.
Annual savings from NMP (at $.15 - $.25/lb of N) may
range from virtually nothing to $30/acre. A median figure
assuming farmers will save only through reduced use of
commercial N is $10/acre/yr.81  We use this as an upper
estimate for cost savings, and zero as a lower estimate.
(Note: These savings calculations could use more work.)
6 % of cultivated crop land is
managed under CCA NMPs (NPS
Gap analysis).  Other surveys
suggest more extensive use of
NMP techniques:  IL/IA/IN/WI-30
-37%,82PA-36%.83  We
assumed 35% (upper estimate) or
17.5% (lower estimate) of total
acreage already had nutrient
management plans in place.
4. Total national annual BMP costs for crop land

        To estimate the number of acres that will need each BMP, we used the following data from NRI
for each of the 710 counties: (1) acres of crop land in the county; (2) the average slope of these crop acres;
(3) the number of crop acres eroding at greater than 15 tons/acre/year.  From these data, we estimated the
nationwide acres needing BMPs as follows:
79      See USEPA. National Management Measures to Control Nonpoint Source Pollution from Agriculture.
        Chapter 4A, page 63. (http://www.epa.gov/owow/nps/agmm/index.html)

80      TetraTech, July 2000, op cit.

81      U.S. EPA. National Management Measures to Control Nonpoint Source Pollution from Agriculture. Op
        cit. This document cites the change in N and P application rates observed for 19 USDA Demonstration
        and Hydrologic Unit Area Projects where nutrient management planning was implemented. (Source:
        Meals, Sutton and Griggs.  1996.  Assessment of Progress of Selected Water Quality Projects of USDA
        and State Cooperators. USDA-NRCS.  1996)  The median change in N use across 15 N-oriented projects
        was a reduction of 43 Ib/acre. The median P reduction across 14 P-oriented projects was 21 Ib/acre. We
        valued reduced N application at $.23/lb and reduced P application (conservatively) at $0/lb. $.23/lb x 43
        Ib/acre = roughly $10/acre. The source does not report whether changes in crop yields, costs for nutrient
        application, or other factors were observed in the projects.  The absence of such data on other potential
        impacts of nutrient management planning means that our assessment of potential savings from
        nutrient management planning are incomplete.

82      See Khanna et al., 2000. Site Specific Crop Management: Adoption Patterns and Incentives. Review of
        Agricultural Economics. 21: 455-472.

83      See Shortle et al. 1993. Economic and Environmental Potential of the Pre-sidedressing Soil Nitrate Test.
        Draft  final report on the USEPA Contract No. CR-817379-01-0, Department of Agricultural Economics
        and Rural Sociology. The Pennsylvania State University.
                                                  1-9

-------
        a. Retirement acres

        In each of the 710 counties, the acreage to be retired was assumed to be all acres that are eroding
at a rate greater than 15 tons/acre/year. We then summed over all counties to find the total acres to be
retired in the 23 States, and multiplied this sum by the scaling factor of 1.56 to obtain the national estimate.
In the 710 counties, there are 3,118,400 acres of crop land eroding at greater than 15 tons/acre/year.
Nationally, we estimate 4,864,704 acres of crop land to be retired.

        b. Riparian forest buffer acres

        The number of acres requiring riparian buffers in each county depends on the length of the banks
of the impaired waters in the county and the width of the buffer that is assumed necessary.  For each crop
land-impaired water body, we obtained information from the 303 (d) data base on its size — its length (for
rivers and coastal shorelines) or area (for lakes and estuaries).  A 75 foot buffer84 was assumed to be
necessary on each side of the rivers, around the perimeter of lakes and estuaries, and along the length of
coastal shoreline. We calculated the total  number of bank-feet needing buffers as follows:

        •       For rivers.  We assumed that buffers would be required on each side of a river. The
               number of bank-feet needing buffers is thus double the length of all the impaired river
               segments.

        •       For lakes.  We assumed  that all lakes were circular in shape, and thus estimated each
               lake's circumference from its area.  A buffer was assumed to be required around the lake's
               perimeter.  The number of bank-feet needing buffers is therefore the circumference of the
               lake.

        •       For estuaries.  Similarly as for lakes, we assumed that estuaries are roughly circular and
               calculated their circumference given their area.  Estuaries are partly but not fully enclosed
               by land.; we assumed that the average estuary is enclosed by land around only two-thirds
               of its circumference. The number of bank-feet needing buffers is therefore the
               circumference multiplied by two-thirds.

        •       For coastal shoreline. A buffer is required on only the landward side, and the number of
               bank-feet needing buffers is thus equal to the length of crop-impaired shoreline.

We summed bank-feet across the crop-impaired water bodies in a county, and then estimated the number of
acres to be put into riparian buffers in the county through the following equation:

Number of acres needing riparian buffer = total number of bank-feet x 75 feet x .000022957 acres/sqft

We summed over all 710 counties and multiplied by 1.56 to scale up to the nation.
84      The NRCS conservation practice standard for a riparian forest buffer requires a Zone 1 of at least 15 feet
        immediately adjacent to the stream and a Zone 2 of at least 20 feet adjoining Zone 1. The total should be
        100 feet or 30 % of the flood plain, whichever is less, but at least 35 feet.  In addition, when the riparian
        forest buffer is to be next to cropland or other sparsely vegetated or highly eroded areas, NRCS
        recommends an additional filter strip (Zone 3) of at least 20 feet.  For cropland riparian buffers, we have
        assumed an average width of 75 feet, comprising Zones 1,2 and 3.

                                               1-10

-------
        Calculating the number of bank-feet for each type of impaired water body as described above, we
estimate 469,364,293 bank-feet needing buffers for the 2,228 crop-impaired water bodies in the 710
counties. Applying a buffer depth of 75 feet, converting to acres and scaling to the entire nation, we
estimate 1,260,698 acres of riparian forest buffer needed.  We assume that all this buffer will be land that
was formerly crop land.

        c. Vegetative barrier acres

        We developed assumptions based on a briefing describing this BMP by the USDA National
Sedimentation Laboratory (SEDLAB).85 SEDLAB considers the use of two types of vegetative barriers:
continuous contour barriers, and discrete barriers. Continuous vegetative barriers on contours across the
full width of a field would prevent both sheet and rill and concentrated erosion, whereas properly located
discrete barriers would be sufficient to prevent concentrated erosion alone. A SEDLAB  simulation
suggests that discrete barriers might require only 5/14 the acreage that continuous barriers would require.
We assumed that conservation tillage would satisfactorily reduce sheet and rill erosion, and hence
simulated the implementation of discrete vegetative barriers to address concentrated erosion specifically.
We simulated discrete buffers by first estimating the acreage of continuous barriers that would be needed
and then by scaling this acreage down by the 5/14 factor.

        Consistent with SEDLAB guidance, we assumed a 4-feet wide continuous vegetative barrier for
every 5 feet of field elevation change. We estimated the acreage of continuous barriers that would be
needed in counties with crop impairments for all crop land that will neither be retired nor put in riparian
forest buffers. The proportion of  crop land acres in each county that will need to go into barriers is given
by the following equation:

        Fraction of acreage needed in barrier strips = 0.8* [l/sqrt(l+l/(average slope)2)]

        We estimated the number of acres needing vegetative barriers by multiplying this fraction by the
total crop land acres in each county. We then scaled down this acreage estimate in each county for
continuous barriers by multiplying by 5/14. To calculate the national estimated acres required for
vegetative barriers, we summed over all counties and multiplied this sum by the scaling factor of 1.56.

        In the 710 counties, assuming that the highly erosive crop land is retired and using the information
on average slope,  we calculate 2,524,753 acres as needed if continuous vegetative barrier strips were to be
implemented. Scaling down by 5/14 to represent discrete barriers rather than continuous ones and scaling
up by 1.56 to obtain a national estimate, we project 1,406,648 acres needed for discrete vegetative barriers.

        d. Conservation tillage and nutrient management plan acres

        To estimate the number of acres needing conservation tillage and nutrient management plans in the
710 counties, we subtracted from  the total area of crop land in these counties the acreage that is to be
retired, devoted to vegetative barriers, and devoted to riparian buffers to determine the acres of crop land
remaining.  It was assumed that both conservation tillage and nutrient management would be applied to
these lands. We then scaled this estimate by 1.56 to extrapolate to the nation as a whole.
85      SEDLAB, op cit.

                                               1-11

-------
        In the 710 counties, there are:

                128,138,800 acres             of crop land
                -  3,118,400 acres             to be retired (eroding at > 15 tons/acre/year)
                    901,698 acres             needed for discrete vegetative barriers
                    808,140 acres             needed for riparian forest buffers
                123,310,562 acres             needing conservation tillage and nutrient management

After scaling to the nation, we estimate 192,364,478 crop land acres potentially needing conservation
tillage and nutrient management planning. However, not all of this acreage of crop land is planted in
annual crops for which conservation tillage could be appropriate.  Some is orchards, perennial crops,
vegetables, potatoes, etc.. NRI, from which we draw our information on crop land in each county,
estimates 375 million acres of crop land in the U.S.  CTIC estimates only 293 million acres of annually
planted crop land.86 Thus 21.9% of all crop land ([375-293J/375) is inappropriate for conservation tillage.
We scale down the estimated 123,310,562 acres potentially needing conservation tillage by a factor of .781
to account for the lesser acreage for which conservation tillage might by appropriate.

        e. Total  annualized national costs and savings from BMPs for crop land

        To estimate the nationwide costs for each BMP, we multiply the total number of acres needing
each BMP by the estimated costs per acre and by the percent of acreage needing BMPs. Summing across
all 5 BMP costs, we estimate the total annual costs for crop land BMPs at roughly $1.3 billion/year (lower
estimate) or $3.9 billion/year (higher estimate).  Applying two further scaling factors (0.4484 to reflect the
pace/lag scale factor and 1.13 to reflect unknown/not classified source information), we estimate the total
annualized national cost at $645 million/year (lower estimate) and $1.956 billion/year (higher estimate).

        For conservation tillage and nutrient management plans, we estimate the unsealed annual potential
savings at up to  $0.8 million and $1.6 billion, respectively.  Applying the two further scaling factors
(0.4484 to reflect the pace/lag scale factor and 1.13 to reflect unknown/not classified source information),
we estimate the annualized national potential savings at up to $414 million for conservation tillage and up
to $804 million for nutrient management plans.   The lower estimate for potential annual savings for both of
these practices is zero.

        A spread sheet following the discussion of pasture land and range land provides detail on these
calculations for crop land.

B. PASTURE AND RANGE  LAND

        We found 1,454 water  bodies on States' 303(d) lists that States have identified as impaired by
pasture and/or range land (either directly in their 303(d) submission or cited as impaired in their  305(b)
submissions and corresponding with a 303(d)-listed water body).87 These water bodies are in 511 counties
86      Conservation Tillage Information Center, 1998, op cit.

87      Sometimes States reported in a manner that identified whether the impairment was due specifically to
        pasture land or to range land or to both. In other instances, though, a State reported source information in
        a manner that did not allow us to determine whether pasture or range was responsible -- citing "animal
        damage to riparian zone", for example. Because we were unable to distinguish pasture from range
        impairments, we decided to estimate the amount of BMPs that are necessary to respond to both source

                                                1-12

-------
in 25 States. We assume conservatively that the remaining States may have pasture or range land-impaired
water-bodies, but either did not report sources of impairment at all or reported them in a manner that was
insufficiently specific to identify pasture or range land as a source of impairment.  The 25 States reporting
pasture or range land-impaired 303(d) waters contain 70.596 million acres of pasture land. In the Nation
as a whole, there are 119.573 million acres. We assume that the degree to which pasture land impairs
waters in the 25 "reporting" States is replicated in the remaining States.  The scale factor to extrapolate
from whatever pasture land costs we estimate for the 25 States to the entire nation is thus 1.69 (119.573
million divided by 70.596 million).

        Similarly, in the 25 States there are 387.143 million acres of range land.88 In the nation as a
whole, there are 555.278 million acres. The scale factor for range land is thus 1.43.

l.Extent of nonpoint source activity requiring BMPs

        Using NRI data, we determined that there are 25,628,300 acres of pasture land and 137,703,200
acres of private range land in these 511 counties. The pasture land in these counties constitutes 36.3% of
all the  pasture land in the 25 States (70.596 million acres).  The private range land in these counties
constitutes 45.9% of the private range land in the 25 States (299.986 million acres).  We assume that in
these counties there is also 45.9% of all the Federally owned grazing land in the 25 States (87.157 million
acres), or 40.005 million acres of Federal grazing land. In total, then, there are 177.708 million acres of
range land in the 511 counties.

        Applying the scale factors of 1.69 and 1.43 to extrapolate from the 25 States to the entire nation,
we estimate that 43.31  million acres of pasture land and 254.1 million acres of range land will need to
implement BMPs in the Least Flexible TMDL Program scenario.

2. Description of the BMPs required where controls are needed

        We assumed that owners of pasture land identified as contributing to impairment of a 303 (d) water
will achieve the load reductions required by TMDLs by implementing the following groups of BMPs:

        1.      Prescribed grazing practices.  Planting, field management and livestock use are planned
               and coordinated such that grazing needs are met and healthy vegetative cover is maintained
               to minimize soil erosion. Also, grazing is  managed in a manner to distribute the manure to
        types simultaneously, and then to split this total amount of BMPs required into portions attributable to
        each of pasture and range separately.

        This calculation is complicated by the fact that the NRI data on range acreage by county covers private
        land only. In many States, though, there is substantial additional non-private land (e.g., Federal, State,
        Tribal) that is used for range grazing. This non-private range land, in additional to the private range
        land, will need BMPs if range-related impairments are to be remedied.  We are unable to find complete
        data on non-private range land acreage. GSA provides figures (by State, not county) on Federal land
        leased for grazing purposes,  which is  likely the large majority of non-private range land. (Federal land
        leased for grazing amounts to roughly 38 % as much land as NRI reports for private range acreage.) We
        have added the amount of Federally leased range land to the private range land totals from NRI in
        estimating the extent of range land in each of the 25  "reporting" States and in the nation as a
        whole. Source of Federal range information: General Services Administration.  Summary Tables
        of Real Property Owned by the U.S. Throughout the World. Table 12.

                                                1-13

-------
               increase its rate of decomposition and nutrient cycling. In estimating the costs of such
               practices when applied to pasture land, we have assumed intensive rotational stocking as a
               specific practice. Intensive rotational stocking is defined as the rotation of grazing animals
               among several small pasture subunits (paddocks) rather than continuously grazing one
               large pasture. Each paddock is grazed quickly and then allowed to regrow for a period
               ungrazed until is it ready for another grazing.  The practice both maximizes forage and
               eliminates denuded areas.

        2.      Practices to protect and restore riparian areas. As for crop land, we will assume
               riparian forest buffers as a relatively higher cost measure within this general category.

        For range land, we assumed that conservation practices applied in riparian areas would be
sufficient to address most range land-related impairments, and that measures addressing the range acreage
itself would not be necessary.  For costing purposes we assumed two alternative riparian measures that will
establish two different cost estimates:

        I.      Use exclusion. Grazing animals are prevented from accessing the riparian zone,
               commonly by fencing it off.  The exclusion may be permanent, or only when streambanks
               are most vulnerable to damage (via fences with gates). This will allow riparian vegetation
               to re-establish itself, and will prevent livestock from degrading a stream's banks and
               channel.

        2.      Stream protection and/or bank stabilization.  These practices  can include a variety of
               measures to protect and/or restore riparian areas, including livestock exclusion, alternate
               livestock watering facilities, tree planting, bank stabilization, filter strips,  critical area
               plantings, channel vegetation,  mulching, stock trails and walkways, and more.

        Again, these various practices for pasture and range are selected for costing purposes.  We do not
mean to suggest that any of these specific practices must be chosen.  Selection of specific BMPs should
reflect the particular characteristics  of the water quality problem and the pasture or range grazing operation
that needs to be managed.

3. Unit costs for these BMPs

        For these BMPs in each of the 511 counties with pasture and/or range-impaired waters, we (1)
determined the number of acres to which the BMP would need to be  applied; (2) estimated  the cost per
acre required to implement the BMP ; and (3) estimated the proportion of acreage that is assumed already
to have the BMP in place. Further detail on unit costs for each BMP is provided below.
                                                1-14

-------
                                              Exhibit 1-7
                                Intensive Rotational Stocking (Pasture)
     Extent BMP is Needed
      Unit Cost for BMP
  Degree Already Implemented
 For all pasture land in every
 county with at least one pasture-
 impaired water body. These data
 were obtained from NRI.
Zero net costs. Several studies
considering net farm profit find
that intensive rotational stocking
(IRS) is profitable relative to
alternative approaches. IRS was
found more profitable than
continuous pasture, hay, or corn
silage for dairy operations in PA.
IRS was found to increase net
returns by 2 - 8 % relative to open
access grazing for TX dairy
farms.90
Because of uncertainty in
extrapolating these savings to all
pasture operations, we have
assumed zero savings as well as
zero costs.
                                                                 89
 15 % in PA. Unknown elsewhere.
 The degree to which this practice is
 already implemented makes no
 difference, however, when we
 assume that it entails zero net costs.
                                              Exhibit 1-8
                                   Riparian Forest Buffer (Pasture)
      Extent BMP is Needed
       Unit Cost for BMP
   Degree Already Implemented
 50-foot wide corridor on each side
 of every pasture-impaired water
 body.  (NRCS conservation
 practice standard is for minimum
 of 35 feet plus).  Note that this is
 different from the assumption for
 buffers adjoining crop land, which
 require an additional zone of 20
 feet on each side.
Same approach as for riparian
forest buffer for crop land.  The
average CRP rental rate for
pasture land is $43.29, and for the
low estimate we assumed that this
represents the social cost of
having land in this use rather than
pasture. For the high estimate, we
assumed $130/acre the average
cost incurred by CREP to enroll
small tracts, as discussed for crop
land buffers. The cost of
establishing trees ($300/acre) and
annual maintenance cost ($5/acre)
are assumed to be the same as for
crop land buffers.	
As cited previously, about 2.2 of the
needed 4.2 - 5.0 million acres of all
sorts of buffers are currently in
place. However, more than 90 % of
the buffer acreage in place is grass
waterways and shelterbelts.  We
assume that no riparian forest buffer
has been implemented for pasture
land.
89      USDA, Grazing Lands Technology Institute. Dairy Farmer Profitability Using Intensive Rotational
        Stocking.  September, 1996. Www.ftw.nrcs.usda.gov/pdf/Dairy.pdf

90      Jan McNitt, et al.  Livestock and the Environment: Precedents for Runoff Policy.  Policy Options -
        CEEOT-LP.  Prepared for the U.S. EPA's Office of Policy Development, contract No. CR 820374-02.
        October, 1999.
                                                 1-15

-------
                                              Exhibit 1-9
                                        Use Exclusion (Range)
     Extent BMP is Needed
      Unit Cost for BMP
 Degree Already Implemented
 Along both banks for the entire
 length of every range-impaired
 water body.  9 acres are enclosed
 by fence (lost to grazing) per
 bank-mile (18 acres per stream
 mile), making the exclusion zone
 equivalent to roughly 75 feet in
 depth on each bank..91
Low estimate is based on costs
from BLM/SCS studies.92 Fencing
installation costs $5261 per stream
mile (1978 $), or $3,006 per bank-
mile in current dollars.  Annual
O&M cost = 1 % of capital cost.
Average annual rental rate for
grazing land = $8/acre/yr or
$9.14/acre/yr in today's $. Assume
that riparian land lost to grazing is
worth twice the average, or
$18.28/acre/year. Assume 25 year
useful life for fencing.

High estimate. Fencing costs
$19,000 per stream mile ($9500
per bank-mile).93 Annual O&M
cost = 5% of capital cost.94 Value
of lost grazing in exclusion zone
calculated identically as for low
estimate.
Assumed not to be implemented at
all yet.
91      As summarized in RCG/Hagler, Bailly, Inc. Controlling Nonpoint Source Loadings from Federal Lands:
        an Analysis in support of Clean Water Act Reauthorization. July 1, 1992.

92      Ibid.

93      Robert Edwards and Travis Stoe, Susquehanna River Basin Commission. Nutrient Reduction Cost
        Effectiveness Analysis, 1996 Update. Publication No. 195, March, 1998.

94      RCG/Hagler Bailly, Inc.  Op cit.
                                                 1-16

-------
                                             Exhibit 1-10
                           Stream Protection and/or Streambank Stabilization
      Extent BMP is Needed
       Unit Cost for BMP
  Degree Already Implemented
 Along both banks for the entire
 length of every range-impaired
 water body.  Assume that these
 practices are applied to a riparian
 zone of 100 feet in depth on each
 bank.95
Capital cost is estimated from
NRCS data for 1997 on cost share
payments per acre for installation
of WP2 (stream protection) and
SP10 (streambank stabilization)
(avg. $47.88/acre, applied to the
riparian zone calculated as the
number of bank-feet of impaired
water multiplied by the assumed
100 foot depth). This cost share
amount is assumed to represent
40% (upper estimate) to 75% of
total practice cost.96  WP2 and
SP10 useful life are assumed to be
15 years.  Annual O&M costs are
calculated similarly as for use
exclusion - 1 % of capital cost for
low estimate, 5 % for high
estimate.  50 % (low estimate) to
100 % (high estimate) of the 100
foot depth riparian zone is
assumed lost to grazing use, and
the acreage lost is valued at
$18.28/acre, as for use exclusion.
Assumed not to be implemented at
all yet.
Note again that these two sets of practices - use exclusion and stream protection and/or streambank
stabilization - are regarded as alternatives for protecting and restoring range-impaired riparian areas.  The
costs estimated for the two sets of practices represent alternative cost estimates. We do not presume that
both sets of practices need to be applied, and the cost estimates are not to be added.

4. Total national annual BMP costs for pasture and range land

        Because we have assumed that the unit costs of implementing intensive rotational stocking are
zero, we estimate that there will be no cost to implement this practice on all the pasture land in the 511
counties with pasture-impaired waters in the 25  States. The cost after scaling to the nation is similarly
zero.

        Our procedure for estimating costs for pasture land-related riparian forest buffers is similar to that
for crop-land related buffers.  However, we believe there should be no costs for buffers for the pasture
land-impaired waters that are also crop land-impaired. Many pasture-impaired water bodies are also crop-
95      U.S. EPA, Agriculture Policy Branch, Office of Policy, Planning and Evaluation. Economic Achievability
        Analysis: Agricultural Management Measures. December 18, 1992.

96      Tony Esser, NRCS. Personal communication, March, 2000.

                                                1-17

-------
impaired, and 75-foot buffers have already been costed for them in the crop land section.  Adding further
buffer costs due to pasture land would constitute double-counting.

        We make a similar assumption regarding use exclusion or stream protection/streambank
stabilization for range-impaired water bodies. For any water body that is also impaired by crop land and
has had riparian forest buffers already costed for it, we assume that use exclusion or stream
protection/streambank stabilization is unnecessary.97

        Among the 1454 water bodies in the 25 States that are cited as pasture and/or rangeland impaired,
773 water bodies are also cropland-impaired and have already had buffers costed for them.  The remaining
681 pasture and/or rangeland-only water bodies are in 22 States. For three States (ME, NC, NE), all
pasture and/or rangeland-impaired water bodies are also impaired by crop land. No additional bank-feet of
buffer, use exclusion or stream protection are needed in these three States due to pasture and/or range land.

        For the 681 waterbodies (impaired by pasture and/or rangeland, but not by cropland) in 22 States,
there are 150,473,159 bank-feet of control measures needed.  However, four of these 22 States reported
pasture/range impaired waterbodies but did not report on crop impairments (NV, RI, VA, WI).  These
latter  four States pose some difficulty in our analysis.  In the crop land analysis, we assumed that these
"non-reporting" States for crop land may in fact have crop land-related impairments, and we scaled to these
States in estimating the need for crop land-related buffers, as well as the other crop land-related BMPs.
Thus, for four of the 19 States that report pasture or range land-related impairments, crop land-related
buffers have already been projected, but we do not know which (presumed) crop land-related water bodies
these buffers  apply to. We apply a scaling procedure to address this issue. Some of the pasture/range
impaired water bodies in these four States are also likely crop-impaired, and these water bodies have
already been scaled to in the crop land buffer analysis.

        •       147,714,887 bank-feet are in the 18 states that report for rangeland and/or pastureland but
               not for cropland. These bank-feet are completely incremental to the bank-feet needed for
               crop land in these  18 States.

               150,473,159 - 147,714,887 = 2,758,272 bank-feet are associated with pasture and/or
               rangeland water bodies in the four States that did not report on crop impairments. Some of
               these bank-feet are already scaled to in the crop land analysis; thus not all of these bank-
               feet are incrementally due to pasture and range.

        •       In the 21 States that report for both crop and pasture/range, there are  353,066,997 bank-
               feet associated with pasture and/or rangeland-impaired water bodies.  147,714,887 of
               these bank-feet are associated with waterbodies that are impaired only by pasture and/or
               range land, and not by crop land.  The portion of the total that is due exclusively to pasture
               and/or range is thus 41.83% (147,714,887/353,066,997 = .4183).

        •       We apply this percentage to the total pasture/range-impaired bank-feet in the 4 States that
               do not report for crop land: 2,758,272*41.83% = 1,153,996. Thus 1,153,996 bank-feet
               are needed for exclusively crop/pasture-impaired water bodies in the 4 States.
97      Note that the cost per acre for a crop land riparian buffer is higher than the cost per acre for either use
        exclusion or stream protection/streambank stabilization.

                                               1-18

-------
        •       The total number of incremental bank-feet attributable solely to pasture/range in the 25
               States is thusl,153,996 +147,714,887 = 148,868,883.

        We split this total amount of pasture/range bank-feet needed among pasture and range in each
State according to the fractions of pasture vs. range land in each State. The result for the 25 States is
66.548 million bank-feet needed for pasture lands, and 82.321 million bank-feet needed for range lands.

        The scale factor for pasture land is 1.69.  Therefore, the national number of bank-feet of riparian
buffers needed for water bodies impaired by pasture only is 1.69*66.548 = 112.466 million bank-feet.

        The scale factor for range land is  1.43. Therefore, the national number of bank-feet of use
exclusion or stream protection/streambank stabilization needed for water bodies impaired by range only is
1.43*82.321 = 117.719 million bank-feet.

        We convert these estimates of bank-feet needing BMPs to acreage and then to costs by applying
the factors shown previously in the unit cost tables. We also scale the costs to reflect the pace of TMDL
development/compliance time lag (factor of 0.4484) and to reflect unknown/not classified source
information (factor of 1.13).

        We estimate the costs for pasture land-related BMPs at $5.0 - $10.7 million/year.  These costs are
for buffers, as intensive rotational stocking is assumed to be costless.

        We estimate the total annual costs for range land BMPs at:

        •       For use exclusion: $5.1 -  $16.4 million/year.  OR

        •       For stream protection and/or streambank stabilization: $2.3 - $5.1 million/year.

Regarding these two sets of practices as alternative ways of estimating costs and taking the low and high
estimates, we project the costs for range land BMPs at $2.3 million to $16.4 million per year.

C. SUMMARY COSTS FOR AGRICULTURAL LAND

        Exhibits I-10 through 1-12 summarize the estimated costs for Scenario  1 (Least Flexible TMDL
Program) for crop, pasture and range land.
                                               1-19

-------
                                           Exhibit 1-10
                      Implementation Costs for Crop Land ($ in millions/yr)
Number of States analyzed
    23
Number of counties in these States with crop-impaired waters
   710
Fraction of crop acreage in these States that is in impairment counties
   .53
Scale factor from these States to the nation
   1.56
National total annualized implementation costs (millions of 2000 dollars/yr):

   Conservation tillage
   Nutrient management planning
   Riparian measures
   In-field or edge-of-field measures
   Management of highly erosive crop land

   Total costs
 85 - 785
 317-781
 41 - 104
 49 - 108
 154 - 177

645 - 1956
National total potential savings from implementing these BMPs (millions of 2000
dollars/yr):

   Conservation tillage
   Nutrient management planning

   Total potential savings
  0-414
  0-804

 0-1218
                                           Exhibit 1-11
                     Implementation Costs for Pasture Land ($ in millions/yr)
Number of States analyzed
Number of counties with pasture and/or rangeland-impaired waters in these States
Fraction of pasture acreage in these States that is in impairment counties
Scale factor from these States to the nation
National total annualized implementation costs (millions of 2000 dollars/yr):
Riparian measures
Intensive rotational stocking
Total costs
25
511
.363
1.69
5.0 - 10.7
0
5.0 - 10.7
                                               1-20

-------
                                          Exhibit 1-12
                     Implementation Costs for Range Land ($ in millions/yr)
Number of States analyzed
Number of counties with pasture and/or rangeland-impaired waters in these States
Fraction of rangeland acreage in these States that is in impairment counties
Scale factor from these States to the nation
National total annualized implementation costs (millions of 2000 dollars/yr):
Riparian measures:
Use exclusion (fencing); OR
Stream protection/bank stabilization
Total costs
25
511
.459
1.43
5.1 - 16.4
2.3-5.1
2.3 - 16.4
       On the following page is a spread sheet showing the derivation of these estimates for crop, pasture
and range land.
                                             1-21

-------
                                             Scenario 1 - Summary Costs for Nonpoint Sources; Detail on Agriculture and AFO BMPs ($/yr)
   Acres*   Capital Cost/Capital Cost/ Useful Life
Needing BMP Acre* - Low Acre*- High   (Years)
                                                                                  Annual O&M Annual O&M Annualized  Annualized
               BMP
            Crop Land
Conservation tillage
   Potential CT savings
Nutrient mgt plan
   Potential NMP savings
Riparian forest buffer
Vegetative barrier
Retirement
            Pasture Land
Riparian forest buffer
Rotational stocking
            Range Land
Riparian fencing only*; or                     22,295    $3,005.50
Stream protection/bank stabilization        '     270,248'      $63.84
* for range riparian fencing, units are miles rather than acres
                                                 Degree Not  Degree Not
                                                   Needed/     Needed/      Estimated      Estimated     Annualized   Annualized
Cost/Acre* - Cost/Acre* - Cost/Acre* - Cost/Acre* -   Already     Already   Annualized Cost Annualized Cost Cost w/Scale  Cost w/Scale
    Low        High        Low        High     Implemented Implemented     -Low         -High     Factors-Low Factors-High
                                                    - High       - Low
150,236,657
150,236,657
192,364,478
192,364,478
1,260,698
1,406,648
4,864,704


129,094'
43,180,906
$12.43
$0.00'
$7.00
$0.00
$300.00
$100.00
$67.20


$300.00

$0.00
$0.00
$15.00
$0.00
$300.00'
$100.00
$98.86


$300.00

10'
10
3
3'
20
15
10


20'

$0.00
($8.66)
$2.33
($10.00)
$57.76
$57.76'
$52.76


$48.29

$16.43
$0.00
$4.00
$0.00'
$135.00
$140.00'
$57.76


$135.00

$1.77'
($8.66)
$5.00
($10.00)
$86.08
$68.74
$62.33


$76.61
$0.00
$16.43 '
$0.00
$9.72
$0.00
$163.32'
$150.98
$71.84


$163.32
$0.00
37.20%
37.20% '
35.00%
35.00%
25.00%
0.00%
0.00%'


0.00%
0.00%
37.20%
37.20%
17.50%'
17.50%
0.00%
0.00%'
0.00%


0.00%
0.00%'
$166,973,694 $1,550,147,840
($817,059,056) $0
$624,854,664 $1,541,900,236'
($1,587,006,944) $0
$81,388,652' $205,894,512
$96,692,234 $212,374,974
$303,206,143 $349,458,152
Subtotal of costs:
Subtotal of potential savings:
$9,889,635' $21,083,397
$0' $0'
$84,604,235
($413,997,287)
$316,608,859
($804,123,722)
$41 ,238,979'
$48,993,182
$153,632,127
$645,077,382
($1,218,121,009)
$5,010,999
$0'
$785,447,509
$0
$781,268,515
$0
$104,325,102
$107,608,700
$177,067,650
$1,955,717,476
$0
$10,682,788
$0
                             $9,500.00         25      $194.58      $639.52      $452.48     $1,454.72       0.00%'       0.00%     $10,088,131     $32,433,374'    $5,111,575    $16,433,731
                              $119.70         15'        $9.78       $24.27       $16.79       $37.41        0.00%       0.00%      $4,536,832     $10,109,264     $2,298,777     $5,122,283
                                                                                                       1-22

-------
II. ANIMAL FEEDING OPERATIONS (AFOs)

       We found 750 water bodies on States' 303(d) lists that States have identified as impaired by
AFOs. These water bodies were either cited directly in a State's 303(d) submission as impaired by AFOs,
or they were cited as impaired by AFOs in a State's 305(b) submission and corresponded to a 303(d)-listed
water body. These water bodies are in 451 counties in 32 states. We assume conservatively that the
remaining 18 states may have AFO-impaired water bodies, but either did not report sources of impairment
at all or reported them in a manner that was insufficiently specific to identify AFOs as a source of
impairment. The 32 States reporting AFO-impaired waters have 44,790,402 animal units confined in
AFOs. In the nation as a whole, there are 69,312,402 animal units confined in AFOs.98  We assume that
the degree to which AFOs impair waters in the 32 "reporting" States is replicated in the remaining States.
The scale factor to extrapolate from whatever AFO costs we estimate for the 32 States to the entire nation
is thus 1.55 (69,312,402 divided by 44,790,224).

A. EXTENT OF AFO ACTIVITY REQUIRING CONTROLS

       To determine the amount of AFO activity occurring in these 451 counties, located in 32 States, we
used data from the 1997 Census of Agriculture on the number of farms in each county and their animal
inventories.

1. Types of animals covered in the analysis

       The types of animals covered in the analysis include:

       •       Beef cows"
       •       Milk cows
       •       Layers 3 weeks old and older
       •       Broilers and other meat-type chickens
       •       Swine.

       The Census of Agriculture also provides some, but not complete, information on other animal
species: turkeys,  sheep, goats, horses, ducks, mink, etc..  For  a variety of reasons, including incomplete
county-level inventory information, limited information on manure management costs for farms raising
these species, and a desire to keep the amount of data collection and analysis to a manageable level, we
decided to analyze in this study only AFOs raising beef and dairy cows, swine, layers and broilers.  This  set
of animals that we analyzed accounts for nearly 95 % of the volume of confined animal manure produced
98      Source: Data on confined animals of various sorts in each State, obtained from the 1997 Census of
        Agriculture, multiplied by the number of animal units (EPA definition) per head for each sort of animal.
        U.S. Department of Agriculture.  1997 Census of Agriculture. Www.nass.usda.gov/census.

99      We assume that AFOs where beef cattle are raised in confinement are represented by the Census category
        "Cattle and calves fattened on grain and concentrates for slaughter". This category should generally
        include beef feedlots and exclude pasture and range operations in which the cattle are not confined.

                                              1-23

-------
annually at U.S. livestock and poultry farms.100 We believe that omitting the numerous animal types other
than the five we have analyzed results in only a very small underestimate of the likely costs for AFOs to
achieve TMDL-mandated load reductions.

       For broilers as well as beef fattened on grain and concentrates, the Census provides information on
the number of animals sold over the course of the year rather than on the inventory on hand at the end of
the year.  We converted the sales information to approximate inventory by dividing sales by the number of
production cycles assumed to take place at a typical farm in a year (for broilers, 6.0 and for confined beef,
2.5).

2. Size of AFOs covered in the analysis

       We assumed that very small AFOs would be unlikely to be addressed in TMDLs.  USDA
estimated that in 1997 AFOs with less than 50  animal units represented roughly 40 % of all farms, but less
than 5 % of all animal units.101 For this analysis, we assume that AFOs  with less than approximately 20
animal units will not face TMDL requirements. This cutoff level defining a de minimis AFO is the same as
EPA used for its economic analyses of both coastal zone management measures for C AFOs and national
nonpoint source management measures for CAFOs. More specifically, AFOs with fewer than the
following numbers of animals  were assumed to be unlikely targets for the TMDL requirements:

       •       50 beef cattle sold (equivalent  to 20 beef cattle inventory, or 20 animal units);

       •       20 dairy cows (28 animal units; the Census provides data on dairies using a 20 cow break
               point, and provides no data on farms corresponding with 20 animal units or  approximately
               14 dairy cows);

       •       50 swine (20 animal units);

       •       3,200 layers (the equivalent of 32 animal units, the closest of the Census break points to
               20 animal units for layers); and

       •       60,000 broilers produced (10,000 broilers inventory, the equivalent of 100 animal units,
               but roughly the size of the smallest broiler house in common use).

       In the cost analysis, we also make some distinctions involving AFOs that are sufficiently large to
be CAFOs. We defined CAFOs with reference only to their size and not to their "method of discharge".102
In this analysis, CAFOs include all farms with  more than the following numbers of animals:
100     Source: U.S. EPA. Preliminary Study of the Livestock and Poultry Industry: Section 2, Economics.
        September 30, 1998.

101     U.S. Department of Agriculture, Economic Research Service. "Confined Animal Production Poses
        Manure Management Problems." Agricultural Outlook. September, 2000.  Pages 12-18.

102     Existing regulations also define some farms of smaller size as CAFOs because they discharge pollutants in
        specified undesirable ways.

                                               1-24

-------
        •       1,000 beef cattle (equivalent to 2,500 cattle sold);

        •       700 dairy cows;

        •       2,500 swine;

        •       30,000 layers or broilers (equivalent tol 80,000 annual production of broilers) at an AFO
               with a liquid manure system.  We assume that 10 % of layer farms use liquid manure
               systems, and no broiler farms use such systems.103

        Using data from the Census of Agriculture, we calculated the total number of animal units in the
451 AFO-impaired counties as 13,830,056. This number represents 31% of the total number of animal
units in the 32 States.

B. RELATIONSHIP BETWEEN THE AFO COST ANALYSIS AND EPA's PROPOSED CAFO REGULATIONS

        EPA has recently proposed substantially revised effluent guidelines for feedlots. The AFO cost
analysis in this report does not account for these proposed revisions. We assume that the existing feedlots
effluent guideline is part of the baseline for the cost analysis (as we assume for other promulgated
technology-based standards also), and that the new proposed-but-not-yet-finalized effluent guideline is not
part of the baseline.  Any not-yet-incurred costs for C AFOs (as defined under the existing regulations) to
meet existing requirements are part of the baseline, not attributable to the TMDL program.  To the extent
that TMDLs will likely require load reductions of AFOs or CAFOs that exceed what is required by the
existing effluent guideline, we attribute the ensuing costs to the TMDL program.  In fact, if the revised
feedlot effluent guideline is promulgated in roughly the same form as it has been proposed, much of these
costs that  we now attribute to the TMDL program will instead be required by the new effluent guideline.
Nevertheless, for this analysis, we do not anticipate the promulgation of the revised effluent guideline, and
we instead count as attributable to the TMDL program all costs for AFOs beyond the requirements of the
existing guideline.

        In support of the proposed effluent guideline, EPA has conducted extensive research and analysis
on potential management measures for AFOs and their costs.104 Most of the work for the TMDL cost
analysis was completed before this effluent guideline AFO costing work was performed. In many respects,
this TMDL cost analysis for AFOs does not reflect the newer and generally better data developed for the
effluent guideline. However, EPA's feedlot effluent guideline cost studies are currently being revised, and
we did not believe that we should make the substantial effort to change our TMDL costing to conform to
the effluent guidelines work while this work is still in flux.  We plan soon (likely during the period while the
draft TMDL cost analysis is undergoing public review) to revise the TMDL cost analysis for AFOs  to
conform our estimates to the effluent guideline cost analysis information.
103     Source: USDA.  "USDA Review of EPA's TMDL Cost Assessment" June 29, 2001.

104     U.S. EPA, Office of Water, Office of Science and Technology. Final Cost Methodology Report for Beef
        and Dairy AFOs. Final Cost Methodology Report for Swine and Poultry Sectors. January, 2001.

                                              1-25

-------
C. DESCRIPTION OF BMPs REQUIRED WHERE CONTROLS ARE NEEDED

       We assumed four sets of measures would be necessary for AFOs that contribute to impairment of
waters that States have identified as AFO-impaired:

       •       Upgrades to AFO facilities as necessary to collect, store and spread manure on nearby
               fields.  Necessary facilities and equipment include those to:l) Collect and store
               manure/process wastewater and runoff that comes in contact with manure; 2) Manage the
               manure by spreading it on nearby agricultural land; 3) Compost dead animals; and 4)
               Prevent pollutant movement to ground water.

       •       Additional manure hauling and/or composting.  Manure should be used consistent with
               agronomic needs rather than simply spread nearby. In areas where the amount of manure
               generated exceeds crop needs, excess manure will need to be transported either to nutrient-
               deficient areas or to alternative end uses such as composting. We assume that manure
               application rates should not exceed phosphorus-based limits.105

       •       Nutrient management planning. For AFOs, this will include: (1) Developing a nutrient
               management plan; 2) Manure and soil testing; 3) Training and certification of personnel
               who apply manure; and 4) Keeping records regarding manure generation and disposition.

       •       Practices  to protect and restore AFO-impaired riparian areas. As for crop and pasture
               land, for costing purposes we will assume riparian forest buffers will be implemented.
               Again, riparian forest buffers will tend to be more expensive than alternatives such as
               grass buffers, use exclusion, stream protection, etc..

       We estimate the costs for these four practice groups for all AFOs of greater than de minimis size in
counties within which there are AFO-impaired water bodies.  These practices are likely to generate some
savings that will offset the costs in part or in whole:

       •       As AFOs improve their manure management facilities, they will collect and manage a
               greater percentage of their manure. The additional manure collected and available for
               spreading may have some value.

       •       Manure increases in value when it is hauled to where its nutrient content can be more fully
               used rather than always being spread near the farm where it is generated.

       •       Nutrient management planning and careful use of manure nutrients may allow a farmer to
               reduce his purchases of commercial fertilizers.

We will also estimate the amount of these potential savings.
105     In general, P-based limits allow less manure to be spread than will N-based limits. P-based limits will
        thus require manure to be transported a greater distance in order to find a sufficient quantity of
        agricultural land on whcih the manure can be used. EPA's proposed feedlots effluent guideline would
        require application of manure consistent with  phosphorus needs.

                                               1-26

-------
D. CALCULATING THE COSTS TO UPGRADE AFO FACILITIES

        In estimating the costs of upgrading manure management facilities at AFOs, we used a set of unit
cost estimates developed by DPRA for EPA in connection with the Agency's "Management Measures
Guidance for Coastal Zone Nonpoint Source Pollution" required by the Coastal Zone Management Act
Reauthorization Amendments of 1990.  Until the recent feedlots effluent guideline work, the set of CZAR A
measures promulgated for AFOs and the accompanying cost analyses represented EPA's most
comprehensive review of AFO facilities needs.  The 1995 Economic Impact Analysis supporting the
CZARA management measures  describes the measures recommended for AFOs and their costs in more
detail:106

        Management measures considered in this analysis focus on (1) facility runoff controls, (2) nutrient
        management practices with respect to the disposal of manure on agricultural land and (3) dead
        animal composting for poultry. Facility runoff is controlled primarily through diversions (diking)
        for eliminating run-on and channeling on-site effluents to the ultimate control mechanisms... [The
        measures also] included lined retention ponds and irrigation for ultimate disposal of effluents...

        Variable rainfall patterns were also considered in estimating management measures for controlling
        run-on and runoff.  For example, retention ponds were designed to hold runoff from a 25-year, 24-
        hour rainfall event, plus allowances for additional freeboard, process water, and storage for periods
        when water cannot be disposed.  Because a storm of this magnitude is variable, control designs
        varied to provide capacity for 4-, 6-, 8- and 10-inch rainfall events. Ultimately, facilities in different
        states were required to have management measure controls designed appropriately for the expected
        rainfall event (e.g., Louisiana facilities were designed on the basis of a 10-inch 25-year/24-hour
        storm, Michigan facilities for a 4-inch storm).

        Investment costs for effluent control mechanisms vary by size of facility and expected intensity of
        storm events.  The ranges of total investment costs are ...  [from $900 for a small swine operation to
        $60,500 for a large dairy]... (Effluent control costs for poultry operations, which are presumed to be
        entirely enclosed, are not significant and were not estimated.) ... Annual operating costs for these
        facilities, including farmer provided labor, range from a low of approximately four percent to a high
        of eight percent of total investment costs.

        Manure management includes provisions for disposing of manure on agricultural land.  Manure
        management consisted of spreading manure so that application of the nitrogen and phosphorus
        constituents do not exceed 60 and 20 pounds per acre, respectively. This requirement increased
        manure land application costs by requiring operators to apply manure over larger areas, consequently
        increasing transportation costs. Transport distances were estimated to increase up to five (5) miles
        (one-way) for the largest model facilities considered. Manure management cost estimates ranged
        from approximately $100 per year for small swine feedlots to over $12,000 per year for large dairies.

        Dead animal disposal was considered only for poultry (layers and broilers). The disposal method
        considered was composting dead birds with poultry litter.  Composting facility investment costs
        ranged from approximately $400 for small 5,000-bird facilities, to more than $6,700 for 80,000-bird
        facilities. (Source: DPRA, 1995. op cit., pp 7-8.)
106     DPRA Incorporated. Economic Impact Analysis of Coastal Zone Management Measures Affecting
        Confined Animal Facilities. October 7, 1992. Also, DPRA Incorporated. Economic Impact Analysis of
        National Nonpoint Source Management Measures Affecting Confined Animal Facilities. May 17, 1995.

                                                I-?'7
                                                 -z- /

-------
       DPRA estimated the costs for such facilities for model AFOs of different sizes for each animal
type that we address (confined beef, dairy, swine, layers, broilers). We made several adjustments in using
DPRA's unit costs for AFOs for this study. First, we deleted a cost reduction of $3,500 per farm assumed
by DPRA to be provided in environmental cost sharing funds by Federal and State agricultural agencies.
Our analysis aims to estimate the total implementation costs of TMDLs.  Cost sharing funds reduce costs
to farmers, but for our analysis they represent transfer payments from governments rather than any
reduction in total implementation costs. Second, we extend DPRA's cost estimates to several new model
AFOs that were sized smaller or larger than the range that DPRA had dealt with in its analysis.  We did
this by developing regression equations giving DPRA's estimated costs as a function of AFO size, and
using these estimated equations to extrapolate costs  for model AFO sizes outside of DPRA's range.  Third,
we inflate DPRA's costs to March, 2000 dollars, using the index of producer prices.

       There are both pros and cons in our using the DPRA costs to estimate the unit costs for AFOs that
will need to improve their environmental performance to meet the requirements of TMDLs:

•      Pro: The DPRA costs  cover a broad set of minimum management measures that are typical of the
       least that a representative AFO will need to do.  The costs reflect the effluent guideline requirement
       of no discharge except for a 25-year/24-hour storm, and add some modest manure spreading
       requirements.

•      Pro: The DPRA costs  represent the most comprehensive set of costs for AFO water quality
       management that EPA had developed until recently, and they have been widely used by the Agency
       and others.

•      Con: DPRA presents the costs as a package that is very difficult to break into component pieces
       and component costs.  One result is that it is difficult to adjust the DPRA costs  to reflect the fact
       that the various different items are currently in place at AFOs to varying degrees.  DPRA costed
       out construction of waste lagoons, retention basins, liquid waste irrigation equipment, dead animal
       composting facilities, manure hauling equipment, etc.,  assuming that no portion of any of these
       facilities or equipment was yet in place.  Some of these items are currently in place at the great
       majority of AFOs (virtually all AFOs have  some equipment for collecting and disposing of their
       manure), but other items are much less commonly in place (e.g., many AFOs already have waste
       lagoons, but many of the lagoons are not lined).

•      Con: DPRA does not assume that poultry farms need manure storage structures. In fact, virtually
       all poultry farms should have such structures  (perhaps only half of them now have them) in order
       to store dry litter appropriately after it is removed from the poultry house until it can be spread on
       fields.

•      Con: DPRA's assumptions seem generally low regarding the typical distance from an AFO that
       manure will need to be transported before it can be spread consistent with agronomic needs.
       DPRA assumes transport distances of 0 - 5  miles. In many areas of the country, however, there
       are large excesses of confined animal manure relative to local crop  needs and manure will need to
       be shipped much greater distances. We are able to estimate the amount of manure needing to be
       shipped out of the county of origin and the resulting costs for inter-county shipping (see below),
       but there may also be a substantial quantity of manure that needs to be shipped  some distance
                                              1-28

-------
       intermediate between DPRA's 0-5 miles and entirely out of the county (typically 15 miles or
       more). We likely miss the costs when such intermediate shipping is needed.

•      Con: Embedded within the DPRA estimate of annualized costs is an assumed 10 % discount rate.
       We would prefer to use a 7 % discount rate consistent with the remainder of this analysis.
       However, it would be very difficult to disentangle DPRA's annualized costs into capital and annual
       O&M costs and then substitute our preferred discount rate. We have chosen to retain DPRA's
       annualized costs despite their 10 % discount rate assumption.  Annualized compliance costs with a
       10 % discount rate may be, by our rough estimate, about 0 - 5 % higher than they would be if we
       were to recalculate them with a 7 % discount rate.  This error is small compared with other likely
       uncertainties in our estimates.

In sum, the DPRA cost estimates have some important shortcomings relative to our desires, but they are the
best available to us pending completion of the feedlots effluent guideline cost estimates.  We will convert
this analysis to use the effluent guideline cost estimates as soon as possible.

       We applied DPRA's unit costs to the numbers of farms of each animal type and size in each of the
counties with AFO-impaired waters. We also reflected the degree to which each type/size farm currently
has these facilities in place.

       C AFOs were presumed already to have all the needed facilities in place due to requirements of the
existing CAFO effluent guideline. Costs for facilities upgrades were presumed to be zero for all CAFOs.
We assumed that CAFOs included all farms with more than 1,000 AU (EPA definition) of a single animal
type, with the exception of broiler farms (all were assumed to use dry manure management techniques, and
hence none are CAFOs by reason of size) and 90 % of layer farms with more than 30,000 inventory (10%
of such large layer farms were assumed to use liquid manure systems and hence subject to CAFO status).

       Treatment-in-place estimates for AFOs having fewer than 300 animal units were developed from
EPA's Nonpoint Source Gap Analysis data. We assumed that the following percentage of AFOs
containing fewer than 300 AU already had the desired facilities in place:

                                          Exhibit 1-6
                  Percentage of AFOs < 300  Animal Units With Facilities in Place
Animal Type
Beef
Dairy
Swine
Layer
Broiler
Storm Water Diversion,
Management
90
90
50
73
11
               Source: TetraTech, Inc. Nonpoint Source Gap Analysis. Op cit.
                                              1-29

-------
       For AFOs containing 300 to 1000 AU, we assumed that the percentage needing facilities was half
of that for AFOs containing fewer than 300 AU, such that:

                                           Exhibit 1-7
                Percentage of AFOs 300-1000 Animal Units With Facilities in Place
Animal Type
Beef
Dairy
Swine
Layer
Broiler
Storm Water Diversion,
Management
95
95
75
87
56
Note that the information on facilities in place from EPA's Nonpoint Source Gap analysis covers only
some of the sorts of facilities presumed to be necessary at AFOs in EPA's CZARA analysis. We assume
that the fraction of farms with storm water diversion and management facilities in place is representative of
the fraction of farms with all of the osther sorts of facilities in place  also (e.g., dead animal composting,
lined lagoons). This assumption is likely inaccurate, and better information on facilities in place will be
obtained before this analysis is finalized.

       To estimate the costs for AFOs to upgrade manure management facilities, we multiply the number
of farms in each of the 451 counties of each animal type/size clas by the appropriate DPRA cost for that
sort of farm by the proportion of that sort of farm that is presumed not yet to have the needed facilities in
place.  Exhibit 1-8 shows the resulting estimated costs for AFOs to upgrade their facilities. These costs
include the various scaling  factors to extrapolate from the States that "report" regarding AFO impairments
and to reflect the incomplete coverage of the  nonpoint source analysis and the TMDL pace/lag factor.

                                           Exhibit 1-8
              National Facilities Costs by Animal Type (in millions of 2000 dollars/yr)
Animal Type
Broilers
Swine
Beef
Layers
Milk Cows
Total
Facilities Costs
$19.707
$3.916
$0.997
$0.498
$3.197
$28.316
                                              1-30

-------
 Our final estimate of the nationwide incremental cost of upgrading facilities at AFOs is $28.3 million/yr
for Scenario 1.

E. POTENTIAL COST SAVINGS FROM ADDITIONAL MANURE MANAGEMENT FACILITIES

       As some AFOs improve their facilities for collecting and managing manure, more manure will
become available for use.  Depending on the local relationship between the amount of nutrients available
through already-collected manure and crop needs, this additional collected manure may or may not have
economic value for local use. We estimate this potential value of the additional collected manure assuming,
for this step in the analysis, that it is not shipped. We assume that the additional collected manure has
value (without hauling) if local crop nutrient needs exceed local manure availability of nutrients through
manure.

       We estimate the volume of additional manure collected as a result of the AFO facilities upgrades,
the amount of nutrients recovered and available to crops if such manure were spread locally, and the value
of these nutrients as a function of the local balance between nutrient needs and availability. These
calculations involve the following data:

       •       Data from the 1997 Census of Agriculture on the number of confined animals in each
               county within which an AFO is located;

       •       Estimates from DPR A of the tons of manure generated per animal per year;

       •       Data from Lander et al107 on:

                      The proportion of generated manure from each animal type that is typically
                      recovered and then available for use (the "recoverable manure factor"),

                      The pounds of nutrients (phosphorus, nitrogen, and potassium) available per ton
                      of recoverable  manure generated by each animal type; and

                      The ratio of the amount of nutrients that will be available to plants after the
                      manure is spread to the total amount of nutrients in recovered manure.

       For each county and animal type, the total tons of manure generated and then recovered as a result
of upgraded facilities at AFOs are calculated by multiplying the number of animals at upgraded facility
farms by DPRA's estimated number of tons of manure generated per animal. The total tons of recoverable
manure is calculated by multiplying the total tons of manure generated by the "recoverable manure factor".
For phosphorus and potassium, the total pounds of nutrients available to crops in the additional manure are
calculated by multiplying the total tons of recoveraed manure by the pounds of nutrients per ton of
107    Lander et al. February 1998. Nutrients Available from Livestock Manure Relative to Crop
       Growth Requirements.

                                              1-31

-------
recoverable manure.  For nitrogen, based on Lander, it is assumed that 30 percent of nitrogen in recovered
manure is lost and unavailable to crops due to volatilization of ammonia.108

       The application of manure to crop land can have economic value if it allows for decreased
expenditures on commercial fertilizer. Exhibit 1-9 shows the average potential value per pound that we
assumed for the nutrients contained in the additional manure.  This value will be realized if all the manure
nutrients displace on a one-for-one basis an equivalent amount of purchased nutrients in commercial
fertilizer.
                                            Exhibit 1-9
                         Maximum Value Per Pound for Manure Nutrients
Nutrient (oxide form)
N
P205
K2O
Value per Ib.
$0.23
$0.237
$0.154
Nutrient
N
P
K
Value per Ib.109
$0.23
$0.543
$0.186
        The spreadsheet "sum of nutrients all animals" contains the final estimates of the total value of
        nutrients in each county. Those counties located in Great Plains states are assigned no value for
        Potassium because	EXPLAIN....The number of pounds of P, N, and K are summed up over all
        animal types to give the total pounds of nutrients contained within each county.

        To determine whether nutrients in the additional manure collected due to upgraded AFO facilities
has value at these maximum levels, or at some lower level, or has no value at all, we conducted a county-
by-county analysis. For each of the 451 counties, we used data from Lander et al. to compare the amount
of nutrients available in all manure generated in the county to the amount of nutrients that is needed for
non-legume, harvested crop land, hay land,  and pasture land in the county.  We assumed the following:

        •       Nitrogen.  Manure nitrogen has full value if it is generated in a county where total manure
               nitrogen amounts to 25% or less of total county crop and pasture nitrogen needs. It has
               half of full value if it is generated in a county where total manure nitrogen amounts to 25 -
               50% of crop/pasture needs. It has 25% of full value ... THE AFO DISCUSSION IN
               THIS APPENDIX IS NOT WRITTEN/EDITED BEYOND THIS POINT
108    Lander et al. Op cit.

109    Sources for value per pound and factors for converting from oxide form.  Jacobs, L.W. February 12, 1997.
       Resource Value of Dairy Manure. Crop and Soil Science Department. Michigan State University.
       http://ww.canr.msu.edu/fldcrp/manure2.htni.  Zublena, J.P. et al. March 1996. Swine Manure as a
       Fertilizer Source. North Carolina Cooperative Extension Service. Publication Number AG-439-4.  Hart, J.
       et al. August 1997.  Dairy Manure as a Fertilizer Source. Oregon State University Extension Service.
       http://eesc.orst.edu. Bennett, M.,  Fulhage, C., and Osburn, D. October 1993. Waste Management
       Systems for Dairy Herds.  Department of Agricultural Economics, University of Missouri-Columbia.
                                               1-32

-------
       b. Nutrient management planning costs (for both AFOs and CAFOs)

       AFO nutrient management planning costs include the following:

       •      Capital costs associated with training
       •      Costs to develop and periodically update nutrient management plans
       •      Annual soil testing
       •      Annual manure testing
       •      Annual record-keeping

       Due to the lack of nutrient management plan requirements in the existing effluent guidelines for
CAFOs, we calculated costs for both AFOs and CAFOs. For some AFOs and CAFOs, we assumed that
nutrient management planning was current practice, and hence did not require them to incur any
incremental costs. Exhibit 1-9 indicates the percentage of AFOs and CAFOs that are assumed currently to
conduct these nutrient management activities.

                                          Exhibit 1-9
                               Percentage of AFOs and CAFOs
                     Currently Conducting Nutrient Management Activities
Animal Type
Beef
Dairy
Swine
Layer
Broiler
Training
0
0
0
0
0
NMP
100
100
13
72
19
Soil Test
100
100
90
42
20
Manure Test
100
100
40
73
20
              Source: Tetra Tech GAP Analysis

       We assumed that AFOs and CAFOs would have the same performance, regardless of size.
However, CAFOs and larger AFOs were assumed to incur greater nutrient management planning costs than
smaller AFOs (that is, those with fewer than 300 animal units).  Exhibits 1-10 and 1-11 illustrate the
assumed cost per AFO for AFOs with fewer than 300 animal units, and the additional assumptions used for
CAFOs and larger AFOs, respectively.
                                             1-33

-------
                                          Exhibit 1-10
                Nutrient Management Cost per AFO—Fewer than 300 Animal Units
Animal
Type
Beef
Dairy
Swine
Layer
Broiler
Training1
(capital)
$117
$117
$117
$117
$117
NMP
(yearly)
$11
$50
$26
$9
$28
Soil Test
(yearly)
$12
$25
$17
$11
$18
Manure Test
(yearly)
$50
$50
$50
$50
$50
Record
Keeping111
(yearly)
$20
$20
$20
$20
$20
                                          Exhibit Ml
    Nutrient Management Cost per AFO and CAFO (Relative to AFO with Fewer than 300 AU)

Costs per AFO
<300 AU
Costs per AFO
300 - 1,000 AU
Costs per CAFO
Training
A
A
A
NMP
B
4B
8B
Soil Test
C
4C
8C
Manure Test
D
2D
4D
Record Keeping
E
4E
8E
       With these assumptions, we estimated the national nutrient management costs as follows:

       •       We obtain Census of Agriculture data for each county on the number of farms for each
               animal type and farm size category (1)
       •       We multiply (1) by the proportion of farms assumed to be currently conducting nutrient
               management plans by the estimated cost per farm. We sum each management activity for
               all counties and determine a total nutrient management plan cost for the 32 states.
       •       To obtain an estimate for the entire nation, we multiply by the scaling factor of 1.55 and
               obtain a national estimate of $ 5.9 million/yr.

       We next apply two further scaling factors (.4484 to reflect the pace of TMDL development at a 7
% discount rate, and 1.13 to reflect unknown/not classified source information).  Our final estimate of the
nationwide incremental cost of nutrient management plans for AFOs is $ 3.0 million/yr.
110    We assumed a 5-year useful life for the turnover of the $117 capital investment in training and certifying
       the assumed one employee per farm who is to supervise manure testing and application.

Ill    We assumed that record-keeping was required as an element of nutrient management planning. Record
       keeping entails keeping records of how the manure the AFO or CAFO generates is used and disposed of.
       We estimated this cost at $20/year (2 hours at $10/hour) for an AFO < 300 AU.
                                              1-34

-------
       c. Savings due to nutrient management plans

       d. Additional manure hauling and/or composting costs (only for AFOs and CAFOs in nutrient
access counties)

       Some 303(d) water impaired counties were considered "manure-excess counties," because the
nutrient content in the manure produced from concentrated animals exceeded plant uptake. We assumed
that both AFOs and CAFOs in these counties would need to transport their manure some extra distance
(over the minimal distance  presumed in the DPRA facility costs) in order to find land on which it could be
agronomically applied. We used 1997 data from a USDA/Lander et al study112 on nutrient balance in every
US county to identify which of the 451  303 (d) counties needed extra shipping. The USDA/Lander et al
study provides data on confined animal manure P and N production and crop P and N needs. We assumed
manure would need to be hauled a sufficient distance to permit its application consistent with phosphorus
needs for crop and pasture  land.

       In order to estimate the costs for additional manure hauling, we calculated the following:

       4.      Tons of manure needing shipping.  The tons of manure needing shipping were calculated
               for each animal type using several data sources: Agricultural Census data on the number
               of confined animals per farm, DPRA data on the tons of manure produced per head per
               day, and USDA/Lander et al data on the percentage of recoverable manure in each state.

       5.      Extra shipping distance.  In order to decide whether an extra shipping distance was
               required and how far the manure should be shipped, we used the following criteria:

                       If the county crop nutrient needs exceeded recoverable manure nutrient
                       availability. No additional shipping distance was assigned and the minimal
                       manure hauling included in the facilities costs is presumed sufficient.

                       If the recoverable manure P exceeds the crop P needs.
                              a.      A quantity of manure sufficient to meet crop P needs  can stay in-
                                     county, incurring no additional shipping cost; and
                              b.      The remainder of the manure was assumed to need to be shipped
                                     into another county.

               The shipping distance depended upon several factors:

                       The size of the county.  Each county was assumed to be circular in shape, and the
                       distance to the perimeter was assumed to be the radius taken from its known
                       area113.
112      Lander et al. Manure Nutrients Relative to the Capacity of Cropland and Pastureland to Assimilate
       Nutrients: Spatial and Temporal Trends for the United States. USDA. Publication No. psOO-0579. 2000.

113    County areas (in square miles) taken from U.S. Census Bureau.  Land Area, Population, and Density for
       States and Counties: 1990. Www.census.gov/population/censusdata/90den_stco.txt

                                              1-35

-------
                        Proportion of the county surrounded by other counties with excess P. If the
                       county was surrounded by another excess county, it was assumed that manure
                       would need to be shipped some distance beyond that county.
                        Proportion of the county where transport was impossible or uncertain. If a
                       portion of the county was surrounded by a large lake or ocean, it was assumed
                       that manure must be shipped elsewhere..In addition, if the county was  surrounded
                       by a state that is not included in our AFO analysis, hauling manure into that
                       county was considered uncertain.

        3.      Hauling cost per mile.  In general, hauling costs  per ton-mile vary widely, depending upon
               the assumed capacity of the transport vehicle and the nature of the waste. For the average
               cost per ton-mile, we used a range recommended by TetraTech, set at $. 10 to $.25 per ton
               mile.  To calculate these upper and lower hauling costs estimates, we multiplied the tons
               recoverable manure by the assigned shipping distance by the average $.10 and $.25 per ton
               mile.

With these assumptions in place, we estimated the national nutrient management costs as follows:

        •       To calculate the tons of manure needing shipping in each of the 451 counties, we collected
               data on (1) the number of animals for each animal type in each farm size category; (2) tons
               of manure produced per animal for each animal type; and (3) the proportion of  manure
               generated that is recoverable, and multiply (1) by (2) by (3) and summed across all animal
               types  for each county.
        •       To calculate the excess shipping distance, we collected data on (1) the radius of each of the
               451 counties, (2) the proportion of the county that is surrounded by other excess counties;
               (3) the proportion of the county that is surrounded by an ocean or large lake; and (4) the
               proportion of the county where transport is uncertain. We then multiplied:
                (1) x [l+[(2)+ (3)]/[l-(4)]] We capped the total cost estimated for excess shipping at
               some  figure corresponding to an assumed cost for a large regional composting facility.  We
               assumed the cost of a central composting facility to be $5.33/ton, based on Pratt,  Jones
               and Jones (TIAER, 1997).114 We assumed that such a central composting facility becomes
               economically feasible only if the volume of recoverable manure to be managed  exceeds
               150,000 tons/year within an assumed 20-mile radius catchment area.115
        •       We then multiplied the  tons of manure needing shipping by the excess shipping  distance by
               the $.10 (lower estimate) and $.25 (higher estimate) cost per ton mile estimates  for each
               county and estimated the total national manure hauling costs in 32 states. To obtain an
114     Pratt et al. Estimate that using a large central composting facility at 20 miles distance would cost a dairy
        farm $73/cow/year (sum of hauling cost and tipping fee). At 100 Ib manure generated/head/day for a dairy
        cow (DPRA), about 75% of which is recoverable (Lander), the $73 would pay for disposal of 75
        lbs/day*365 days/yr=13.7 tons/yr of manure. The composting cost would thus be $5.33/ton of manure
        managed.

115     Again based on Pratt, Jones and Jones. The large central composting facility they evaluate serves
        approximately 12,000 dairy cows. At 13.7 tons/yr of recoverable manure per dairy cow, the composting
        facility processes 164,400 tons of manure per year. We assumed that the composting facility becomes
        economical at a slightly lower figure of 150,000 tons of recoverable manure per year.

                                               1-36

-------
               estimate for the entire nation, we multiply by the scaling factor of 1.55  and obtain a
               national estimate of $ 81.0 -134.7 million/yr.

        We next apply two further scaling factors (.4484 to reflect the pace of TMDL development at a 7
% discount rate, and 1.13 to reflect unknown/not classified source information).  Our final estimate of the
nationwide incremental cost of hauling additional manure for AFOs is $41.0 - 68.2 million/yr.

        e. Increased economic value of the shipped/composted manure

        The increased value of manure when it is shipped may offset part of the shipping cost. Manure can
have economic value because its use on crops can reduce the amount of chemical fertilizer that needs to be
used.  The value of manure depends on its nutrient content (N, P and K), and the effective value of each
nutrient in the location to which the manure has been shipped (e.g., whether the manure nutrients serve to
displace purchased chemical nutrients).

        To estimate the increased value of the shipped manure, we first estimated the tons of recoverable
manure generated by multiplying the number of confined animals per animal type116 by the tons manure per
head per year by the percentage of recoverable manure for each animal type. We then multiplied the tons of
recoverable manure by the percentage content of N, P and K for each variety of manure.3.62 pounds P117
to calculate the P content of the manure.

        $7.5 -$18.2 million/year.

        f. Costs for buffers for AFO-impaired water bodies

        We estimated costs for a 75 foot buffer width on both sides of an impaired river, and that the
buffer would take crop land out of production. We estimated buffer needs only for those AFO-impaired
water bodies that have not yet had buffers costed because of crop, pasture or range impairments.

        $4.0 - $10.2 million/year.
116     For broilers and beef fattened off of grains and concentrates, the county data pertained to sales instead of
        inventories. To calculate the number of confined animals by inventory, we divided the number of animals
        sold by the production cycle (for broilers :6, for beef :2.5). For broilers and layers, county-level data was
        only provided for the number of farms. To estimate the number of animals, we assumed that each county
        had the same average number of animals per farm as the state.  We calculated the average number of
        animals per farm for each state in each farm size category  and multiplied it by the number of farms in
        each farm size county for each county. In cases where state data was lacking, we referred to national-level
        data. For the rest of the animal groups, there were instances where the data on the number of animals was
        not provided. In these cases, we used the mean number of animals in each farm size category as an
        approximation of the number of animals.

117     According  to XXX source, there are 3.62 pounds of P produced for every ton of recoverable manure.

                                                1-37

-------
        g.  Total costs and savings for AFOs

        Summing over all estimated costs for AFOs, we estimate the total national implementation cost for
AFOs at $76.4 - $109.8 million/year. The total potential savings are estimated at $7.5 - $66.7
million/year.

III. SILVICULTURE

        Silvicultural activities can contribute to water impairment through several means: erosion, siltation,
bank destabilization, runoff of forest management chemicals, loss of vegetative cover, and an increase in
water temperature.  The level of water quality impairment caused by silviculture depends not only on the
degree of site disturbance (given by the acres or volume of timber harvested), but also the terrain (whether
an area is hilly or flat).

        We found 646 water bodies  on States' 303(d) lists that States have identified as impaired by
silviculture (either directly in their 303(d) submission or cited as impaired in their 305(b) submissions and
corresponding with a 303(d)-listed water body). These water bodies are in 294 counties in 30 States. We
assume conservatively that the remaining States may have silviculture-impaired water bodies, but either did
not report sources of impairment at all or reported them in a manner that was insufficiently specific to
identify silviculture  as a source of impairment.  The 30 States reporting silviculture-impaired 303(d) waters
harvest 10.9 million cubic feet (mcf) of timber per year. In the Nation as a whole 16.3 mcf are harvested
per year. We assume that the degree to which silviculture impairs waters in the 30 "reporting" States is
replicated in the remaining States. The scale factor to extrapolate from whatever silviculture costs we
estimate for the 30 States to the entire nation is thus 1.498 (16.3 million divided by 10.9 million).

A. EXTENT OF NONPOINT SOURCE ACTIVITY REQUIRING BMPs

        We determined the amount of Silvicultural activity in these 294 counties, located in 30 states, using
timber harvest volume data from the U.S. Forest Service's Timber Product Output Data Retrieval System
(http://srsfia.usfs.msstate.edu/rpa/tpo/), developed in support of the 1997 Resources  Planning Act
Assessment. Timber harvest volume for  1996 is available by type of product at the regional, sub-regional,
state and county level, and can be broken out into volumes harvested on National Forest land, other public
land, forest industry land, and other private land.118 The timber harvest volume estimates for 1996 are
shown in the exhibit below.
118      Other private land is defined as private land owned by an entity that does not operate a wood-using
        facility.

                                               1-38

-------
                                           Exhibit I-12
                                National Timber Cut in 1996 (mcf)
From private lands
From National Forests
From other public lands
Total
14,330,530
912,060
1,025,628
16,268,218
        Using this harvest data, we calculated the total volume of timber harvested in the 294 counties: 2.3
million cubic feet (mcf). This volume represents 21% of the total timber harvest in the 30 States for which
we have data. 119  For our BMP unit costs, we used both mcf estimates and acreage estimates.

        In order to convert the USFS timber harvest data into acreage information, The Timber Data
Company developed for us regional estimates of the number of acres on which harvesting operations would
occur in order to obtain 1 mcf of timber.120 These estimates are as follows:
                                           Exhibit I-13
                                Timber volume per acre, by region
Region
Westside Oregon & Washington
California
Arizona & New Mexico
Inland & Rocky Mountain
Lake States
Midwest
Northeast
Westside Southeast
Eastside Southeast
mcf/Acre
3.0
2.0
1.5
1.5
1.3
0.6
0.6
1.0
1.0
119     In 1996, there were 10.9 mcf of timber in the 30 States for which we had data.

120      TDC estimated the acres harvested in order to produce a given log volume based on queries of Timber
        Data Company's database of USFS and selected State timber sale appraisal data for 1998 sales. TDC
        adjusted this data somewhat based on discussions with industry professionals regarding differences
        between public and private timberland management practices.  Adjustments also were made for known
        regional differences in land growth potential and harvest methods.
                                               1-39

-------
B. DESCRIPTION OF BMPs REQUIRED WHERE CONTROLS ARE NEEDED

        There exists a fairly typical set of BMPs that silviculture operations should employ to minimize
their impacts on water quality.  These practices, and the particular combination of practices used, will vary
somewhat according to the type of silviculture operation and the conditions (e.g. region, climate,
topography, soils) of the silviculture site. Examples of BMPs for silvicultural activities are available from
numerous sources, including State forest practices laws and State forestry boards. The most important
document in which EPA has indicated the Agency's judgment about the BMPs that are generally
appropriate is EPA's 1993 Guidance Specifying Management Measures for Sources ofNonpoint
Pollution in Coastal Waters. The Guidance specifies  10 management measures for use in coastal States
(including the Great Lakes States) to protect waters from silvicultural sources of nonpoint pollution, and
lists and describes management practices that can be applied successfully to achieve the management
measures.

        According to The Guidance, we assumed that the following BMPs would be required where
controls are needed:

        2.      Preharvest planning
        3.      Streamside management areas
        4.      Road construction/reconstruction
        5.      Road management
        6.      Timber harvesting
        7.      Site preparation and forest regeneration
        8.      Fire management
        9.      Revegetation of disturbed areas
        10.     Chemical management
        11.     Wetlands forest management

Each BMP is described as  follows:

        12.     Preharvest planning . Performing advance planning for forest harvesting to minimize
               negative water quality impacts. BMPs include laying out harvest units to minimize the
               number of stream crossings, systematically designing transportation systems to minimize
               total mileage; minimizing road and skid trail grades; and surfacing roads (with gravel,
               grass, crushed rocks,  etc.) where grades increase the potential for surface erosion.

        2.      Streamside Management Areas. Establishing and maintaining a Streamside management
               area (SMA) along surface waters, which is sufficiently wide and which includes a
               sufficient number of canopy species to buffer against detrimental changes in the
               temperature regime of the water body, to provide bank stability, and to withstand wind
               damage. BMPs include providing a minimum SMA width of 35 to 50 feet; avoiding
               operating skidders or other heavy machinery in the SMA;  and applying harvesting
               restrictions in the SMA to maintain its integrity.

        3.      Road Construction/Reconstruction.  Minimizing delivery of sediment to surface waters
               during road construction/reconstruction projects.  BMPs include following the design
               developed during preharvest planning to minimize erosion by properly timing  and limiting

                                               1-40

-------
        ground disturbance operations; using straw bales, grass seeding, and other erosion control
        and revegetation techniques to complete the construction project; and installing surface
        drainage controls to remove storm water from the roadbed before the flow gains enough
        volume and velocity to erode the surface.

4.      Road Management. Managing existing roads to maintain stability and utility and to
        minimize sedimentation and pollution from runoff-transported materials.  BMPs include
        maintaining road surfaces by mowing,  patching, or resurfacing as necessary; and
        revegetating to proving erosion control.

5.      Timber Harvesting. Minimizing sedimentation resulting from the siting and operation of
        timber harvesting, and managing petroleum products properly.  BMPs include felling trees
        away from watercourses; removing slash from the water body and placing it out of the
        SMA; minimizing the  size of landings; skidding uphill to landings whenever possible;
        avoiding cable yarding in or across watercourses; and taking precautions to prevent fuel
        leakage and spills.

6.      Site Preparation and Forest Management. Regenerating harvested forest lands.
        Examples of BMPs include avoiding mechanical site preparation on slopes greater than 30
        percent or in SMAs; distributing seedlings evenly across the site; and hand planting highly
        erodible sites, steep slopes, and SMAs.

7.      Fire Management. Minimizing potential NFS pollution and erosion resulting from
        prescribed fire for site preparation and from the methods used for wildfire control or
        suppression. BMPs include planning burning to achieve the desired results while
        minimizing impacts on water quality; avoiding intense prescribed fire or construction of
        firelines in SMAs; avoiding burning on steep slopes with high-erosion-hazard areas or
        highly erodible soils.

8.      Revegetation of Disturbed Areas. Preventing sediment and pollutants from entering water
        bodies by revegetating disturbed soil. BMPs include using seed mixtures adapted to the
        site; using native woody plants planted in rows, cordons, or wattles on steep slopes; and
        seeding during optimum periods for establishment, preferably just prior to fall rains.

9.      Chemical Management.  Minimizing the water quality impact of the use of pesticides and
        fertilizers. BMPs include maintaining  a buffer area around all water bodies for aerial
        spray applications; and applying pesticides and fertilizers during favorable atmospheric
        conditions, and during maximum plant uptake periods to minimize leaching.

10.     Wetlands Forest Management.  Taking special measures to protect beneficial wetlands
        functions and avoid water quality impacts in wetlands areas.  BMPs include providing
        adequate cross drainage  to maintain the natural surface and subsurface flow of the
        wetland; and establishing a SMA adjacent to natural perennial streams, lakes, ponds, and
        other standing water in the forested wetland.
                                        1-41

-------
C. UNIT COSTS FOR BMPs

        To cost out a typical set of BMPs that designated and permitted silvicultural operations might need
to employ, we used the set of BMPs and costs provided in EPA's 1992 report, "Economic Analysis of
Coastal Nonpoint Source Pollution Controls: Forestry", prepared by the Research Triangle Institute (RTI).
The report analyzes the economic achievability of the management measures cited above and contained in
the Guidance Specifying Management Measures for Sources of Nonpoint Pollution in Coastal Waters.
RTI estimated the incremental costs of these management measures by: 1) Specifying a set of key BMPs
that RTI judged the coastal states in each major forest region as likely to require in order to achieve each
management measure; and 2) Estimating the incremental costs of these BMPs relative to the costs of
following current silvicultural practices. In summarizing the results of the incremental cost analysis, RTI
combined the BMPs for the ten management measures into groups:

        •       Management measures for activities generally relating to harvesting.  RTI analyzed the
               incremental costs for seven management measures -- preharvest planning, streamside
               management areas, road construction/reconstruction, road management, timber harvesting,
               revegetation of disturbed areas, and wetlands forest management. Incremental costs for
               the management measures for these seven activities were added and then presented in two
               alternative ways — as the  summed incremental cost of these seven management measures
               per acre harvested, and as the summed  incremental cost of these seven management
               measures per volume of timber harvested.121

        •       Management measures for activities relating to artificial forest regeneration. Costs for
               the remaining four management measures -- site preparation prior to planting, forest
               regeneration (tree planting or seeding),  fire management (prescribed burns for clearing
               unwanted vegetation) and forest chemical management (herbicides, fertilizer) - were dealt
               with by RTI in a very different manner than the harvesting-related costs.

        The unit costs for harvesting-related activities in each of the 30 States and the unit costs for
artificial forest regeneration in the north and south regions, are described in the exhibits below.
                                               1-42

-------
                Exhibit 1-14
Estimated Harvest-Related BMP Costs (2000 $)

Alabama
Alaska
Arizona
California
rlorida
Ilinois

-------
                                           Exhibit 1-15
   Incremental Costs for Management Measures Associated With Artificial Regeneration (2000 $)
Region
South - hilly
South - flat
North - hilly
North - flat
West
Costs
($.63 + $15.05 x %S) x SH
$0.63 x SF
($.19 + $3.50x%S)xNH
$0.19xNF
Assumed identical to North
SH = acreage harvested in South/hilly
SF = acreage harvested in South/flat
NH = acreage harvested in North/hilly
NF = acreage harvested in North/flat
%S = percentage of hilly acreage harvested in each State that has slope > 30%
D. TOTAL NATIONAL ANNUAL BMP COSTS
To estimate total national annual BMP costs:
               We obtained data for each county on:(l) Total million cubic feet (mcf) of timber harvested
               on all lands; (2) The total number of acres harvested on all lands; (3) The percentage of
               all lands that have flat terrain; (4) The percentage of all lands that have hilly terrain; and
               (5) The percentage of hilly lands that have steep terrain (classified as having a greater than
               a 30% slope) The BMP costs per acre and costs per 100 cubic feet for each of the 30
               States are listed in the exhibit below.

               In every county of our 30 States, we multiplied the number of acres in all lands by: (1) The
               per-unit harvest BMP costs (State-by-State per-acre and per-volume cost estimates,
               weighted according to the flat and hilly terrain percentages); and (2) The per-unit artificial
               regeneration BMP costs (per-acre cost estimates for Southern and Northern/Western
               States, weighted according to the flat, hilly, and steep terrain percentages).  For each state,
               we then added (1) to (2) to estimate the higher estimate of the total BMP cost (based off of
               the BMP cost per mcf and the artificial regeneration BMP cost per acre) and the lower
               estimate of the total BMP cost per acre (based off of the BMP cost per acre and the
               artificial regeneration BMP cost per acre). Summing across all 30 states, we estimated the
               total annual cost for silviculture BMPs at $39.1 million (lower estimate) or $54.9 million
               (higher estimate).

               Applying two further scaling factors (.4484 to reflect the pace of TMDL development at 7
               % discount rate, and  1.13 to reflect unknown/not classified source information), we
               estimated $29.7 - 49.7 million/yr as the annualized national BMP cost for silviculture.
                                               1-44

-------
IV. ON-SITE WASTEWATER TREATMENT SYSTEMS (OWTS)

        On-site wastewater treatment systems (OWTS) may serve one or more homes or businesses,
typically in rural areas where public sewers and a central wastewater treatment plant are not available. An
OWTS typically includes a septic tank and a leach field, where partially treated wastewater from the septic
tank percolates into the soil which provides further removal before the effluent leaches to ground water.
OWTS can fail or perform poorly for a variety of reasons, including:

        •       Locations in inappropriate soils, or too close to surface water or ground water;

        •       Too many systems located in close proximity to each other;

        •       Hydraulic overloading;

        •       Insufficient maintenance (conventional septic tanks need to be pumped out every 3 - 5
               years);

        •       Mechanical failure with age.

        Approximately 23 % of the dwellings in the U.S. are served by OWTS. Their prevalence is
particularly high among second homes and rural resort areas.

        We found 596 water bodies on States' 303(d) lists that States have identified as impaired by
OWTS (either directly in their 303(d) submission or cited as impaired in their 305(b) submissions and
corresponding with a 303(d)-listed water body). These water bodies are in 318  counties in 24 States. We
assume conservatively that the remaining States may have OWTS-impaired water bodies, but either did not
report sources of impairment at all or reported them in a manner that was insufficiently specific to identify
OWTS as a source of impairment.  The 24 States reporting OWTS-impaired 303(d) waters include 14.2
million dwelling units served by OWTS.  In the Nation as a whole, 24.7 million dwelling units are served
by OWTS. We assume that the degree to which OWTS impair waters in the 24 "reporting" States is
replicated in the remaining States. The scale factor to extrapolate from whatever OWTS costs we estimate
for the 24 States to the entire nation is thus 1.732 (24.7 million divided by 14.2 million).

        We assume that a TMDL addressing OWTS  will require all OWTS within a riparian zone around
the impaired water to be functioning properly.  We further  assume that "functioning properly" means that
any OWTS that fails must promptly be repaired or replaced.

        Estimates of the rate at which OWTS are currently failing vary across studies. Reasons for
variation may include differences across studies in how "failing" is defined, and geographic differences in
construction practices and environmental settings that result in actual differences in failure rates. EPA's
Draft Nonpoint Source Gap Analysis122 summarizes the estimates developed in many studies. One of the
most broadly based estimates was developed by the National Small Flows Clearinghouse, which estimated
122     Tetra Tech, Inc. Draft Nonpoint Source Gap Analysis. Prepared for U.S. EPA, Office of Water.
        February 7, 2001.

                                               1-45

-------
10.2 % of OWTS as failing during the mid-1990s.123 A survey of State officials yielded failure estimates
that ranged from as low as 0.5 percent (Arizona and Utah) to 50 - 70 percent in Minnesota.124  In the
Nonpoint Source Gap Analysis, Tetra Tech reviews these and other studies and assigns 24 States as having
a "low" average failure rate (5 %), 19 States as having a "moderate" average failure rate (20 %), and 7
States  as having a "high" failure rate (30 %).  For this analysis, we will assume as a low estimate that 10
% of all OWTS are currently failing, and as a high estimate that 20 % of all OWTS  are currently failing.

        The cost of addressing a failed OWTS varies with the nature of the failure.  In the Nonpoint
Source Gap Analysis, Tetra Tech cites possible remedial responses ranging from pumping out an
overloaded septic tank and installing a tank filter (cost about $300) to installing an entirely new advanced
system with nitrogen removal and/or disinfection (cost about $12,000).  The most common likely
responses, according to Tetra Tech, are developing a new subsurface wastewater infiltration field (cost
about $1,500) or installing a complete new conventional septic tank/leach field (cost about $3,000). Tetra
Tech weights each remedial response by its likelihood and develops a weighted average cost of $3,040 per
failed OWTS.  We use this figure in this analysis.

        A properly operated and maintained OWTS can last 20 - 35 years or more.  Tetra Tech assumes
that the average system will function for 25 years.

        We thus estimate the cost to assure that the average OWTS is functioning correctly as $142.88 - $
164.16 per year, as  follows:
Exhibit 1-16
Average Cost Per OWTS For Assuring Pro

Cost to remediate the backlog
of failed systems
Recurring cost to remediate
the additional systems that
will fail each year
Total
Low Estimate
0.1 (low est. of fraction of failed
systems) x $3,040 = $304 capital
cost, or, at 7% interest, $21.28
annualized cost
0.04 (fraction of systems failing each
year) x $3,040 per system = $121.60
per year
$142.88 per system per year
>er Performance
High Estimate
0.2 (high est. of fraction of failed
systems) x
$3,040 = $608 capital cost, or, at 7%
interest, $42.56 annualized cost
0.04 (fraction of systems failing each
year) x $3,040 per system = $121.60
per year
$164.16 per system per year
This estimated average cost will be applied to each OWTS that is assumed to be required by a TMDL to be
functioning correctly.
123     National Small Flows Clearinghouse. Summary of On-Site Systems in the United States. 1996.

124     Nelson, Dix and Shepard. Advanced On-Site Wastewater Treatment and Management Scoping Study:
        Assessment ofShrot-Term Opprotunities and Long-Run Potential.  Prepared for the Electric Power
        Research Institute, the National Rural Electric Cooperative Association, and the Water Environment
        Research Federation. 1999.
                                                1-46

-------
       We assume that the riparian zone within which TMDLs will require all OWTS to be functioning
properly extends for 100 yards in all directions from an OWTS-impaired water.  The 100 yard distance
was based on the following information:

       •       We reviewed State recommendations or requirements regarding surface water setback
               distances - the minimum distance that an OWTS should be located away from a surface
               water body. We reasoned that the riparian zone within which all OWTS should be
               functioning correctly should be at least as large as these setback distances.  The maximum
               State setback distance in any State, as cited in EPA's National Nonpoint Source
               Management Measures Guidance, was 100 feet.125

       •       We reviewed the management measures for OWTS recommended by the Buzzards Bay
               National Estuary Project, an effort rather like a TMDL and one of the best known
               investigations of extensive pollution from OWTS. The Project found that existing OWTS
               setback requirements, while adequate to deal with indicator bacteria, were inadequate to
               deal with viruses from OWTS that may reach ground water and then travel great
               distances. The Project recommended a setback distance of 250 feet for all new OWTS,
               and improvements to system design and application rate for existing OWTS within this
               zone.126

       •       We reviewed a study examining transport of pollutants from large failed septic systems.
               The study showed evidence of increased pollutant concentrations for different pollutants at
               various downgradient distances, the greatest of which was nearly 100 meters.127

       Based on this information, we assume that a surface water is likely to be adequately protected from
impairment by OWTS if all OWTS within 100 yards of the waters are properly sited (i.e., no inappropriate
soils, no excessive OWTS density, no hydraulic overloading), operated and maintained.

       To determine the area of the riparian zone around each OWTS-impaired water, we then multiplied
the length of each impaired water body by the assumed 100-yard-deep zone, as adjusted for the type of
water body:

       •       Rivers. In the 24 States, there were 5440 miles of OWTS-impaired rivers, multiplied by
               300 feet on each of two sides. This results in 395,630 riparian acres needing OWTS
               controls.
125     U.S. EPA. Guidance Specifying Management Measures for Sources of Nonpoint Pollution in Coastal
        Waters. Www.epa.gov/owowwtrl/NPS/MMGI/Chapter4/ch4-5a.

126     Buzzards Bay Project. National Estuary Project Management Plan: Septic System Action Plan.
        Www.buzzardsbay.org/ccmp/septicac.

127     W.D. Robertson and J. Harman. Phosphate Plume Persistence at Two Decommissioned Septic System
        Sites. Ground Water. Volume 37, No. 2, 1999, pp. 228 - 236

                                               1-47

-------
        •       Lakes.  The 167 OWTS-impaired lakes average 5,103 acres per lake.  Approximating the
               lakes' shape as circular, a 300-foot riparian ring around each lake results in an estimated
               60,785  acres needing OWTS controls.

        •       Estuaries.  The 47 OWTS-impaired estuaries average 7,064 acres each. Approximating
               their shape as circular and assuming that land occupies 2/3 of their circumference, a 300-
               foot ring around their landed perimeter results in an estimated 13,424 acres needing
               OWTS controls.

        •       Coastal shorelines.  In the 24 States,  there were 249 miles of OWTS-impaired coastal
               shorelines, multiplied by 300 feet on the land side of each. This results in 9,067 riparian
               acres needing OWTS controls.

        Summing across all OWTS-impaired water bodies in the 24 States, we estimate there are 478,906
riparian acres within which OWTS controls are needed.

        How many OWTS are there within this acreage?  Absent a painstaking and impractical
investigation to delineate these riparian zones, count the dwelling units within them, and estimate the
fraction of the dwelling units that rely on OWTS, we cannot know the answer  with any precision. Instead,
we make a very rough simplifying assumption - that the density of dwellings served by OWTS in any
riparian zone is 10 times the density of dwellings served by  OWTS in the county surrounding the impaired
water. In essence, we assume that the riparian zone is much more intensively developed than is the county
as a whole.  This seems reasonable, as we would expect waterside home sites typically to be much more
attractive than more undistinguished locations. If, for  example,  an entire county has an average of one
OWTS-served dwelling unit per four acres of land area (e.g., as in Fairfield County, CT, one of the
counties with the greatest number of OWTS-served homes), then we would assume in this county that there
would be one dwelling per 0.4 acres specifically in the riparian zone around OWTS-impaired waters.128

        With this final assumption that the density of OWTS in the riparian zones surrounding OWTS-
impaired waters in a county is 10 times the density of OWTS in the county as  a whole, we are able to
estimate the national costs for control of OWTS:
128     This result seems at least plausible in this instance. Some of the OWTS-impaired waters in Fairfield
        County have shorelines fully developed with OWTS-served homes on lots averaging as little as 8,000
        square feet per lot (density of roughly one dwelling per 0.2 acres). Others of the impaired waters in
        Fairfield County have typically larger lots in the developed shoreline areas, and some (limited) portions of
        the shorelines that are relatively undeveloped. The average density throughout the entire riparian zones
        for Fairfield County's ten OWTS-impaired waters could well be approximately one OWTS-served
        dwelling per 0.4 acres.

                                                1-48

-------
        •       We obtained data for each county on: (1) The riparian acreage needing OWTS controls, as
               described previously; (2) The total acreage in each county;129 (3) The number of dwelling
               units served by OWTS.130

        •       [(l)/(2)] x 10 x (3) = # of dwelling units needing OWTS controls in each county.

        •       Summing across all 318 counties with OWTS-impaired waters in 24 States gives 191,921
               dwellings needing OWTS controls

               Total annual cost for OWTS controls in the 24 States = 191,921 x $142.88 (lower
               estimate) or $164.16 (higher estimate) per system per year = $27.4 million/yr (lower
               estimate) or $31.5 million/yr (higher estimate).

        •       To obtain an estimate for the entire nation, we multiply by the scaling factor of 1.732. and
               obtain a national estimate of $47.5 - $54.6 million/yr.

        We next apply two further scaling factors (.4484 to reflect the pace of TMDL development at a 7
% discount rate, and 1.13 to reflect unknown/not classified source information).  Our final estimate of the
nationwide incremental cost of TMDLs for OWTS is $24.1 - $27.7 million/yr.
129     U.S. Census Bureau. Land Area, Population, and Density for States and Counties: 1990.
        Www.census.gov/population/censusdata/90den_stco.txt

130     U.S. Census Bureau. American Housing Survey for the United States. 1997.
        Www.census.gov/prod/99pubs/hl50-97.pdf

                                               1-49

-------