EPA/600/R-07/033A
A Screening Assessment of the Potential
Impacts of Climate Change on Combined
Sewer Overflow (CSO) Mitigation in the
Great Lakes and New England Regions
External Review Draft Report
U.S. Environmental Protection Agency
Office of Research and Development
National Center for Environmental Assessment
Global Change Research Program
John Furlow, Thomas Johnson, Britta Bierwagen
and
ICF International
1725 Eye St, NW
Washington, DC 20006
J. Randall Freed, Jeremy Sharfenberg, Sarah Shapiro
September, 2006
-------
i PREFACE
3 The Environmental Protection Agency's Global Change Research Program (GCRP) is an
4 assessment-oriented program within the Office of Research and Development that focuses on
5 assessing how potential changes in climate and other global environmental stressors may impact
6 water quality, air quality, aquatic ecosystems, and human health in the United States. The
7 Program's focus on water quality is consistent with the Research Strategy of the U.S. Climate
8 Change Research Program - the federal umbrella organization for climate change science in the
9 U.S. government - and is responsive to EPA's mission and responsibilities as defined by the
10 Clean Water Act and the Safe Drinking Water Act. The GCRP's water quality assessments also
11 address an important research gap. In the 2001 National Assessment of the Potential
12 Consequences of Climate Change in the United States (Gleick and Adams, 2000), water quality
13 was addressed only in the context of the health risks associated with contaminated drinking
14 water. A comprehensive assessment of the potential impacts of global change on water quality
15 was not included.
16 Since 1998, the National Center for Environmental Assessment's office of the GCRP has
17 assessed the consequences of global change on water quality. Through its assessment projects,
18 this Program has provided timely scientific information to stakeholders and policy makers to
19 support them as they decide whether and how to respond to the risks and opportunities presented
20 by global change. This report assesses the potential impacts of climate change on combined
21 sewer overflow mitigation efforts in New England and the Great Lakes Region. Water treatment
22 infrastructure was identified as a priority concern because water treatment is an essential service
23 necessary to protect public health and ecosystems. Investments in water treatment infrastructure
24 are also capital-intensive, long-term in nature, and irreversible in the short- to medium term.
25 Today's decisions will thus influence the ability of treatment facilities to accommodate changes
26 in climate for many years into the future.
27 The report is a screening level analysis intended to determine the scope and magnitude of global
28 change impacts rather than a detailed assessment of specific impacts and adaptation measures.
29 Together with a companion report addressing the potential effects of climate change on treatment
30 costs at wastewater treatment facilities in the Great Lakes Region, this report fulfills a GCRP
31 2006 Annual Performance Measure to complete "two external review draft reports detailing the
32 possible impacts of global change on combined sewer overflows in key regions, and the possible
33 effects of climate change and variability on operations and management of publicly operated
34 treatment works (wastewater facilities) for OW and EPA Regions. "
35
36 Peter Preuss, Ph.D.
37 Director
38 National Center for Environmental Assessment
39 Office of Research and Development
40 U.S. Environmental Research Program
41
-------
1
2
3 We are very grateful for the many thoughtful comments and suggestions made by members of
4 staff from the EPA Global Change Research Program and ICF International in developing and
5 conducting this work. In particular, we thank Susan Julius from the EPA Global Change
6 Research Program. We are also very grateful for the many excellent comments and suggestions
7 provided by four peer reviewers of this report: Christopher Impellitteri (US EPA,
8 ORD/NRMRL/WSWRD/WQMB), Mohammed Billah (US EPA, OW/WPD), Donald Brown
9 (US EPA, ORD/NRMRL/WSWRD/WQMB), and Tom Faber (US EPA, Region 1, OEME).
10
-------
3 Climate is a major factor influencing the amount, timing, and quality of water available meet
4 human needs (Gleick 2000). During the last century, much of the U.S. experienced increased
5 ambient air temperatures and altered precipitation patterns (NAST 2000). Projections of future
6 climate suggest these trends are likely to continue, and potentially accelerate during the next
7 century. Future changes in climate could thus impact water quality management.
8 Streamflow is strongly influenced by precipitation intensity and frequency, air temperature, and
9 various natural and anthropogenic factors affecting watershed hydrologic processes. Projected
10 impacts of climate change on streamflow include changes in both the total amount and temporal
11 variability of flow. Climate change is expected to increase the proportion of rainfall occurring in
12 high intensity events (US GCRP 2000), resulting in increased stormwater runoff and high flow
13 events. At the same time, a shift towards more intense storms could also decrease infiltration and
14 groundwater recharge, resulting in reduced low flow periods between events.
15 Combined sewer systems (CSSs) collect and co-treat storm water and municipal wastewater.
16 During high intensity rainfall events, the capacity of CSSs can be exceeded resulting in the
17 discharge of untreated storm water and wastewater directly into receiving streams. These
18 combined sewer overflow events (CSOs) can result in high concentrations of microbial
19 pathogens, biochemical oxygen demand, suspended solids, and other pollutants in receiving
20 waters.
21 The EPA's Office of Water has established the CSO Control Policy, which requires CSS
22 communities to implement various mitigation measures as a component of the National Pollutant
23 Discharge Elimination System (NPDES) permitting process. Mitigation measures are typically
24 engineered to handle storm or flow events of a given intensity, duration, and frequency, and there
25 is an implicit assumption that precipitation and hydrology are constant over time. The rules
26 guiding the mitigation measures are also based in part on an understanding of how precipitation
27 intensity affects sewer performance. In many regions, climate change could increase the
28 frequency of high intensity rainfall events, resulting in an increased risk of CSO events and
29 associated water quality impairment.
30 The objective of this research was to characterize the nature and extent of climate change
31 impacts on CSO mitigation efforts in the Great Lakes Region (GLR) and New England Region
32 (NER). The study examined the extent to which CSO long-term control plans may be under-
33 designed if planners assume that past precipitation conditions are representative of future
34 conditions. The primary areas of focus were on potential changes related to CSO frequency and
35 design characteristics of mitigation efforts in response to climate change.
36 The two regions of the United States that were selected for study, the GLR and the NER, both
37 have many CSSs, which are generally found in older cities and towns. The GLR and NER
38 communities contain 182 and 135 CSSs, respectively. These communities account for nearly half
39 of the 746 CSS communities in the United States (US EPA 2004).
40 The analysis compared historical and projected precipitation characteristics using data sets
41 developed for the Vegetation-Ecosystem Modeling and Analysis Project (VEMAP). The Long
42 Term Control Plan (LTCP) "Presumption Approach" (keyed to reducing CSO events to no more
43 than four per year) was used as the mitigation target to assess impacts on CSOs. Using this
-------
1 mitigation target, a benchmark storm event was determined for each CSS community in each
2 region, based on historical precipitation data from VEMAP.
3 Climate change is expected to affect the frequency and intensity of precipitation events. In the
4 VEMAP analysis, two Global Circulation Models (GCMs) were used to project future
5 conditions: the Hadley Centre Model and the Canadian Climate Centre Model. Both models
6 provide projections on a grid with intervals of 1 degree latitude and longitude. The VEMAP
7 GCM runs, although several years old, were used because (1) they provided a set of historical
8 (weather station) data and projected (GCM) data on the same geographic footing (the 1-degree
9 grid), (2) they had undergone considerable manipulation to convert monthly temperature and
10 precipitation estimates (the raw output of the GCMs) to daily temperature and precipitation
11 estimates, and (3) the data sets had been thoroughly peer reviewed. By comparing projected
12 storm intensities against the benchmark storm event (based on historical storm intensities, the
13 event that corresponding to four CSOs per year), the impacts of climate change were
14 characterized in terms of (1) the extent to which systems may be "under-designed" and (2) the
15 additional system capacity required to meet the mitigation target in the future.
16 The benchmarking analysis matched the locations of CSS communities with historical and
17 projected precipitation datasets, using the portion of the VEMAP grid located in the GLR and
18 NER. For the GLR, historical precipitation data for a 40-year period (1954-1993) were compared
19 to projected precipitation data for a future 40-year period (2060-2099). For the NER, historical
20 precipitation data for a 25-year period (1968-1993) were compared to projected precipitation
21 data for a future 25-year period (2025-2050). The NER study was done subsequent to the GLR
22 study, and the 25-year time period was selected for the NER to obtain information on climate
23 change impacts that are more immediate in nature.
24 Results of this study suggest that for many communities in the GLR and a few in the NER, if
25 engineers design CSO abatement measures based on historical precipitation characteristics,
26 projected climate change will reduce the effectiveness of those measures in meeting the four
27 event per year benchmark. In the GLR, systems designed to meet the benchmark based on past
28 conditions would exceed the benchmark by an average (across communities and GCM
29 projections) of 38 percent (i.e., about 1.5 excess overflows per year). Across all 182
30 communities, this translates to 273 events per year above the objectives of EPA's LTCP.
31 In the NER, the analysis indicated inconsistent results with respect to the impacts of climate
32 change on CSO characteristics due to disparate projections of the Hadley and Canadian models,
33 and perhaps also due to the choice of a shorter, nearer-term time frame for the analysis. Under
34 Canadian model projections, climate related changes will result in a 17 percent decrease in the
35 annual frequency of CSO events on average, while the Hadley model predicts a 12 percent
36 increase in the frequency of CSO events. Across all 135 communities, this translated to 94 fewer
37 and 62 more events per year above the LTCP's objectives under the Canadian and Hadley
38 models, respectively.
39 In the GLR, the rainfall intensity corresponding to a recurrence interval of four events per year is
40 projected to increase in the 2060-2099 period by an average of 10 percent. Holding event
41 duration and infiltration constant, it can be assumed that the design capacity of a CSO storage
42 system is linearly proportional to precipitation intensity; thus, the average design capacity in the
43 GLR would need to increase by the same proportion. Considering that CSO mitigation projects
-------
1 can cost hundreds of millions of dollars, a 10 percent increase in cost to adapt to projected
2 climate change is significant.
3 The results for the NER in the shorter term (2025-2050) were ambiguous, given the differences
4 in direction in the results of the two GCMs.
5 Nonetheless, the results suggest that CSS planners are faced with an important decision on
6 whether to invest additional money now to build in an additional margin of safety to ensure the
7 mitigation effectiveness of CSO projects into the future, or to accept the risk of potentially
8 significant costs of retrofitting/refurbishing CSO projects in the future to maintain mitigation
9 effectiveness in the face of climate change. To the extent that climate change may involve more
10 intense precipitation events, if the engineers designing the LTCPs base their calculations of
11 system size (e.g., storage capacity) on current hydrology, in the future the mitigation actions
12 taken as part of the LTCPs may not be able to meet the objective of no more than four CSO
13 events annually. Given the improvements in the state-of-the-art in GCM modeling and analysis
14 of empirical trends in precipitation, it would be worthwhile to develop tools based on more
15 recent models and trend analysis to provide planners with heuristics for a margin of safety to
16 address climate change.
-------
4 1 Introduction 1
5 1.1 Combined Sewer Systems and Combined Sewer Overflows 1
6 1.2 CSO Controls 4
7 1.2.1 Nine Minimum Controls and Long Term Control Plans 4
8 1.2.2 CSO Control Policy Mitigation Requirements 5
9 1.3 Climate Change and Impacts on the Hydrologic Cycle 5
10 1.4 Objectives of this Research 6
11 2 Methods 7
12 2.1 CSS Selection 7
13 2.1.1 Great Lakes Region 7
14 2.1.2 New England Region 8
15 2.2 CSO Benchmarking Approach 8
16 2.2.1 Presumption Approach Basis 8
17 2.2.2 VEMAP data and GCMs 8
18 2.2.3 Benchmarking Analysis 9
19 3 Results and Discussion 11
20 3.1 Changes in CSO Event Frequency 11
21 3.1.1 Great Lakes Region 11
22 3.1.2 New England Region 13
23 3.2 CSO Benchmark Event Intensity 15
24 3.2.1 Great Lakes Region 15
25 3.2.2 New England Region 17
26 3.2.3 Costs of Adaptation Increased Mitigation Design Capacity 19
27 3.3 Regional and Temporal Variance 21
28 3.4 Limitations and Future Research 21
29 3.4.1 Limitations 21
30 3.4.2 Future Research 23
31 4 Conclusions 25
32 5 Literature Cited 27
-------
1
2
3 Table 1. GLR CSS Communities by State 7
4 Table 2. New England Region CSS Communities by State 8
5 Table 3. Weighted Average Percent Change in CSO Frequency in the Great Lakes Region, 2060-
6 2099 11
7 Table 4. Weighted Average Percent Change in CSO Event Frequency in the New England
8 Region, 2025-2050 13
9 Table 5. Average Change in Benchmark Event Intensity in the Great Lakes Region, 2060-2099
10 16
11 Table 6. Average Change in Benchmark Event Intensity in the New England Region, 2025-2050
12 18
13 Table 7. Projected Change in Runoff Volume, 1-day Averaging Period 20
14
15
16
17
18 Figure 1. Distribution of Combined Sewer Systems in the United States 2
19 Figure 2. Percent Change in Frequency of CSO Events in the Great Lakes Region, 2060-2099 (1-
20 Day) 12
21 Figure 3. Percent Change in Frequency of CSO Events in the Great Lakes Region, 2060-2099 (4-
22 Day) 12
23 Figure 4. Cumulative Distribution of % Change in CSO Frequency in the Great Lakes Region,
24 2060-2099 13
25 Figure 5. Percent Change in Frequency of CSO Events in the New England Region, 2025-2050
26 (1-Day) 14
27 Figure 6. Percent Change in Frequency of CSO Events in the New England Region, 2025-2050
28 (4-Day) 14
29 Figure 7. Cumulative Distribution of % Change in CSO Event Frequency in the New England
30 Region, 2025-2050 15
31 Figure 8. Percent Change in CSO Benchmark Event Intensity in the Great Lakes Region 2060-
32 2099 (1-Day) 16
33 Figure 9. Percent Change in CSO Benchmark Event Intensity in the Great Lakes Region, 2060-
34 2099 (4-Day) 17
35 Figure 10. Cumulative Distribution of % Change in Benchmark Event Intensity in the Great
36 Lakes Region, 2060-2099 17
July 2008 Draff Report: Do noi ate or quote. vii
-------
1 Figure 11. Percent Change in CSO Benchmark Intensity in the New England Region, 2025-2050
2 (1-Day) 18
3 Figure 12. Percent Change in CSO Benchmark Intensity in the New England Region, 2025-2050
4 (4-Day) 19
5 Figure 13. Cumulative Distribution of % Change in Benchmark Intensity in the New England
6 Region, 2025-2050 19
7
July 2006 ' " . .'. ' . y'te or quote. viii
-------
2 Climate is a major factor influencing the amount, timing, and quality of water available meet
3 human needs (Gleick 2000). During the last century, much of the U.S. experienced increased
4 ambient air temperatures and altered precipitation patterns (NAST, 2000). Projections of future
5 climate suggest these trends are likely to continue, and potentially accelerate during the next
6 century. Future changes in climate could thus impact water quality management.
7 Streamflow is strongly influenced by precipitation intensity and frequency, air temperature, and
8 various natural and anthropogenic factors affecting watershed hydrologic processes. Projected
9 impacts of climate change on streamflow include changes in both the total amount and temporal
10 variability of flow. Climate change is expected to increase the proportion of rainfall occurring in
11 high intensity events (US GCRP 2000), resulting in increased stormwater runoff and high flow
12 events. At the same time, a shift towards more intense storms could also decrease infiltration and
13 groundwater recharge, resulting in reduced low flow periods between events.
14 Combined sewer systems (CSSs) collect and co-treat storm water and municipal wastewater.
15 During high intensity rainfall events, the capacity of CSSs can be exceeded resulting in the
16 discharge of untreated storm water and wastewater directly into receiving streams. These
17 combined sewer overflow events (CSOs) can result in high concentrations of microbial
18 pathogens, biochemical oxygen demand, suspended solids, and other pollutants in receiving
19 waters.
20 The EPA's Office of Water has established the CSO Control Policy, which requires CSS
21 communities to implement various mitigation measures as a component of the National Pollutant
22 Discharge Elimination System (NPDES) permitting process. Mitigation measures are typically
23 engineered to handle storm or flow events of a given intensity, duration, and frequency, and there
24 is an implicit assumption that precipitation and hydrology are constant over time. The rules
25 guiding the mitigation measures are also based in part on an understanding of how precipitation
26 intensity affects sewer performance. In many regions, climate change could increase the
27 frequency of high intensity rainfall events, resulting in an increased risk of CSO events and
28 associated water quality impairment.
29 The objective of this research was to characterize the nature and extent of climate change
30 impacts on CSO mitigation efforts in the Great Lakes Region (GLR) and New England Region
31 (NER). The study is intended as a screening level analysis to determine the potential scope and
32 magnitude of climate change impacts rather than a detailed assessment of specific impacts and
33 adaptation measures. The GLR and NER regions account for nearly half of the 746 CSS
34 communities in the United States (US EPA 2004). The study examined the extent to which CSO
35 long-term control plans may be under-designed if planners assume that past precipitation
36 conditions are representative of future conditions. The primary areas of focus were on potential
37 changes related to CSO frequency and design characteristics of mitigation efforts in response to
38 climate change.
39 , :
40 A CSS collects storm water and sanitary wastewater in a common conveyance system and routes
41 them to a treatment plant (US EPA 2004). The storm water component fluctuates with the
42 weather; during rainfall events, the collection system and treatment plant must accommodate
43 more volume due to runoff entering the system directly (street catch basins and gutter
-------
Climate Change Impacts on Combined Sewer Overflows
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
downspouts) or indirectly through groundwater infiltration. By design, when the volume of water
in the collection pipes, storage facility, or at the treatment facility exceeds the system's capacity,
excess water is discharged untreated at different points around the system into surface waters
through CSO outfalls.
The water that is discharged to local rivers and streams during a CSO event is typically a mixture
of raw sewage, other industrial wastewaters, and storm water. The sewage component is
typically of greatest concern from a water quality standpoint due to bacterial and/or viral
contamination. These discharges are known as CSOs. The term CSO refers to any discharge
from a CSS prior to the treatment plant, but most occur in response to wet weather (US EPA
2004). EPA estimates that 850 billion gallons of overflow are discharged into the nation's waters
each year.
According to EPA's 2004 Report to Congress on the Impacts and Control of CSOs and Sanitary
Sewer Overflows (SSOs), there are 746 communities with combined systems and a total of 9,348
CSO outfalls identified and regulated by the NPDES permits (US EPA 2004). Combined sewer
systems are found in 31 states and the District of Columbia, with the majority located in older
cities found in the Great Lakes and New England Regions (see Figure 1).
Figure 1. Distribution of Combined Sewer Systems in the United States.
Source: US EPA web site, http://cfpub.epa.gov/npdes/cso/demo.cfm7program id=5.
Construction of municipal sewer systems began in the 1880s. Prior to then, sewage was collected
in cesspools and privy vaults (Burian et al. 2000). Populations were sparse enough that aesthetic
concerns were not great, and the health risks were not well understood. As population densities
grew during the 19* century, the need to remove waste became more critical. Cities had two
goals in constructing sewers: first, as populations grew larger and more concentrated, privies and
cesspools were no longer sanitary or aesthetically acceptable. Sewers were constructed to
remove wastes from population areas. Second, heavy rains could render unpaved streets
impassable, so storm waters needed to be quickly conveyed to rivers or lakes. The two designs
that were common in this period were dedicated sanitary sewers and combined sewers to convey
sewage and storm water. A third option, separate sanitary and storm sewers, was viewed as too
costly for most communities (Burian et al. 2000).
Treatment was not a part of the process; the purpose was to convey the water away from the
population and into some receiving body. The design choice depended on the needs of the
community. In general, smaller communities opted for sanitary systems only. Larger cities used
July 2006
Draft Report: Do not cite or quote.
-------
1 combined systems because they efficiently and effectively removed both storm and wastewater
2 (US EPA 2004). In some cases, a combined system was chosen because storm water would flush
3 the sewage out of the system and into receiving waters (Burian et al. 2000;Schladweiler 2005).
4 At the turn of the twentieth century, these systems resulted in significant health benefits within
5 the cities they served; there were however, to be consequences with respect to water quality in
6 the receiving water bodies.
7 As the first sewer systems were being built, CSSs were a dramatic improvement over cesspools
8 and open sewers. However, as populations grew, impacts of discharges into receiving waters
9 grew as well. Now, of the 850 billion gallons discharged annually, 75 percent is discharged into
10 rivers, streams, or creeks; 13 percent into unclassified or other waters such as canals; 10 percent
11 into oceans, bays, or estuaries; and 2 percent into ponds, lakes, or reservoirs (US EPA 2004).
12 CSOs result mainly from two different events associated with CSSs: (1) insufficient conveyance
13 capacity within a portion of the sewer system and resultant surcharging and overflow through
14 manholes or designed outfalls and (2) insufficient capacity at the water treatment plant. In the
15 latter case the excess combined sewage must bypass the facility and be discharged at a specified
16 outfall. This effluent is often given primary treatment before discharging.
17 CSOs pose a threat to water quality and human health. The pollutants found in CSOs include
18 microbial pathogens, suspended solids, nutrients, toxics, and debris. Pollutant concentrations
19 vary from place to place and storm to storm, but they can be high enough to cause violations of
20 water quality standards. It is common for local rivers and streams to be considered dangerous to
21 human health after heavy rains due to CSO pollution. It is difficult to attribute the violation of
22 water quality standards exclusively to CSO discharges, because CSOs occur during storm events
23 when some of the same pollutants are washed directly into waterbodies by storm water runoff.
24 Despite the difficulty of attributing causality to particular sources, EPA has compared data on
25 CSO locations with data on 305(b) assessed water segments and 303(d) impaired waters in 19
26 states.l The study found that of a total of 59,335 assessed segments, 25 percent were impaired.
27 For 733 segments that were within a mile downstream of a CSO outfall, 75 percent were
28 impaired. Though it is difficult to determine how much of the impairment is due to the CSO, the
29 high percentage of impairment associated with CSOs suggests some correlation (US EPA 2004).
30 CSOs should be considered as a potential source of pollution during Total Maximum Daily Load
31 (TMDL) development, and in some communities substantial load reductions have been assigned
32 to CSOs as a result of the TMDL process (US EPA 2004).
33 CSOs also pose risks to human health and the environment. Humans can become sick by
34 drinking contaminated water, eating contaminated shellfish, or coming in direct contact with
35 contaminated water. The most common symptoms of pathogenic illness are diarrhea and nausea,
36 but respiratory and other problems can occur as well. Toxics present in CSO discharges include
37 metals and synthetic organic chemicals. Less is known about the risks of biologically active
38 chemicals such as antibiotics, hormones, and steroids (US EPA 2004).
1 305(b) and 303(d) are references to sections of the Clean Water Act that mandate assessment of water bodies, and
identification of impaired waters, respectively.
-------
1 1.2
2 Efforts to manage the risks of CSOs have evolved over several decades. Following the passage
3 of the Clean Water Act in 1972, publicly owned wastewater treatment works (POTWs) were
4 required to incorporate secondary treatment into their wastewater treatment processes. Effluent
5 from the treatment plant had to be treated, but it was unclear how CSOs - discharges from the
6 collection system, not the treatment plant - would be treated by the law. A 1980 court ruling
7 declared that CSO outfalls did not have to be subjected to the secondary treatment required of
8 discharges from a POTW. However, the discharges do fall under the National Pollutant
9 Discharge Elimination System (NPDES) permit program (US EPA 2004). Under NPDES, all
10 facilities which discharge pollutants from any point source into waters of the United States are
11 required to obtain a permit. The permit holder must provide treatment based on technology
12 accessible to all permittees in a particular industrial category (US EPA 2006).
13 In 1989, EPA issued the National CSO Control Strategy, which encouraged states to develop
14 statewide permitting strategies to ensure that all CSSs were subject to a discharge permit. The
15 strategy also recommended six minimum measures for controlling CSOs. As the control strategy
16 was being implemented, environmental groups pushed for further action against CSOs, while
17 many municipalities also called for greater clarity and a national approach (US EPA 2004). In
18 1994 EPA published the CSO Control Policy to establish objectives for CSS communities in
19 order to reduce the environmental impacts of CSOs. Four key elements of the CSO Control
20 Policy are meant to enable communities to cost effectively reduce overflows and meet the
21 obj ectives of the Clean Water Act:
22 1. Provide clear levels of control that would be presumed to meet appropriate health and
23 environmental objectives;
24 2. Provide sufficient flexibility to municipalities, especially financially disadvantaged
25 communities, to consider the site-specific nature of CSOs and to determine the most cost
26 effective means of reducing pollutants and meeting Clean Water Act objectives and
27 requirements;
28 3. Allow a phased approach to implementation of CSO controls considering a community's
29 financial capability; and,
30 4. Provide for review and revision, as appropriate, of water quality standards and their
31 implementation procedures when developing CSO control plans to reflect the site-
32 specific wet weather impacts of CSOs.
33 1.2.1
34 The national CSO Control Policy also requires communities to implement nine minimum
35 controls (referred to as NMC) and to develop a Long Term Control Plan (LTCP) to reduce the
36 frequency and adverse impact of CSOs. The NMC are expected to maximize the effectiveness of
37 existing systems. Among the controls are properly operating and maintaining the system;
38 maximizing the flow to the POTW from the collection system; eliminating overflows during dry
39 weather; and notifying the public of the occurrence and impacts of overflows. In addition to
40 implementing the NMC, communities are expected to develop LTCPs that will ultimately result
41 in compliance with the requirements of the Clean Water Act. The LTCPs are meant to provide
42 communities with as much flexibility as possible in meeting the requirements.
-------
1 The development and implementation of LTCPs are in various stages of completion, but all of
2 the 746 communities that have CSSs must develop plans to comply with the CSO Control Policy.
3 Permit holders designing modifications to their systems generally base their plans on historical
4 weather data. The infrastructure investments made to implement LTCPs are expected to have life
5 expectancies of several decades, and the costs will be considerable. There is no comprehensive
6 source of individual municipal expenditures for CSO control because there are multiple funding
7 sources for CSO projects. However, EPA compiled expenditures to date for 48 communities,
8 roughly 6 percent of the nation's total. Those expenditures totaled $6 billion and ranged from
9 $134,000 to $2.2 billion per community. EPA estimates that the capital costs of future CSO
10 control over the next 20 years will exceed $50 billion (US EPA 2004).
11
12 The EPA's CSO Control Policy allows for three basic approaches to be taken in order to meet
13 CSO mitigation requirements. First, a system may allow no more than four overflows per year
14 (though the permitting authority may allow an additional two). Second, a system may eliminate
15 or capture at least 85 percent by volume of the combined sewage collected in the system during a
16 precipitation event. These first two approaches are considered to be "presumptive" in nature.
17 Finally, the system may eliminate or remove no less than the mass of the pollutants identified as
18 causing the water quality impairment for the volume that would be eliminated or captured by the
19 85 percent approach (US EPA 1994). The demonstration approach allows communities to
20 demonstrate that their system, though not meeting the criteria of the presumption approach, is
21 adequate to enable receiving waters to meet water quality standards and protect designated uses
22 (US EPA 2004).
23 Given the number of communities working to develop and implement LTCPs, the costs
24 associated with CSO control, and the weather's influence on overflow events, it is important to
25 assess the potential impact of climate change on the robustness of LTCPs 25 to 40 years from
26 now. To the extent that climate change may involve more intense precipitation events, if the
27 engineers designing the LTCPs base their calculations of system size (e.g., storage capacity) on
28 current hydrology, in the future the mitigation actions taken as part of the LTCPs may not be
29 able to meet the objective of no more than four CSO events annually.
30 1.3 on t'" ' ' :
31 Streamflow is strongly influenced by precipitation intensity and frequency, air temperature, and
32 various natural and anthropogenic factors affecting watershed hydrologic processes. Projected
33 changes in climate are thus likely to have a significant impact on the amount and seasonal
34 variability of Streamflow.
35 Climate change is expected to increase the proportion of rainfall occurring in high intensity
36 events (US GCRP 2000). This will tend to increase stormwater runoff during events, resulting in
37 increased high flow events without necessarily increasing the amount of water available. A shift
38 towards more intense storms will also decrease infiltration and groundwater recharge. Low flow
39 periods occur when there is little or no precipitation, and Streamflow is supplied by groundwater.
40 If a higher proportion of total precipitation is delivered in intense events, unless total
41 precipitation increases significantly, low flow periods between events will thus likely decline.
42 The Intergovernmental Panel on Climate Change (IPCC) has reported that over the past century
43 there has been a likely increase in the frequency of intense precipitation events in the mid-high
-------
1 latitudes of North America (Houghton et al. 2001). Recent observations and long term
2 projections suggest that climate change will significantly influence the volume of storm water
3 runoff and result in increased high-flow stream conditions (Watt et al. 2003). This has an
4 important influence on water quality. Extreme precipitation events tend to be correlated to
5 poorest water quality - due to runoff pollution and channel scouring during high flows - and are
6 a significant stress on aquatic ecosystems.
7 .: ;;:/<"' . "<" "/' -' s
8 This study sought to characterize the nature and extent of climate change impacts on the
9 effectiveness of CSO mitigation efforts in the Great Lakes and New England Regions, assuming
10 that mitigation will involve infrastructure that is designed based on historical precipitation
11 records. The study focuses on LTCP implementation, and uses a case study approach to examine
12 the extent to which CSO long-term control plans may be under-designed if planners assume that
13 past precipitation conditions are representative of future conditions. A limited, informal survey
14 of EPA regional staff involved in CSO implementation in the GLR and NER indicated that the
15 most common approach in the LTCPs was the presumption approach of controlling and
16 providing a minimum level of treatment to all but four overflow events per year. The study was
17 designed to analyze whether precipitation patterns in the future might result in more - or fewer -
18 CSO events (with respect to the target of four per year) and what the potential increase in design
19 capacity might be to adapt mitigation infrastructure to climate change.
20 The research questions to be evaluated were:
21 1. If CSSs meet the EPA's CSO Control Policy "presumption approach" of four
22 events per year based on historical precipitation, what will be the potential change
23 in CSO event frequency in the future as a result of climate change?
24 2. What would be the incremental change in the design capacity of mitigation
25 measures that would be needed to meet the presumption approach of four CSO
26 events per year under the proj ected precipitation regimes?
27
28 An improved understanding of the potential impacts of climate change on CSOs is important
29 because the occurrence and mitigation of CSO events is highly sensitive to climate, is one of the
30 highest-priority programs at the state and federal level, and will involve significant investment in
31 wastewater treatment. These characteristics have been identified as important criteria to help
32 focus decision support resources on activities that will have the greatest benefit towards adapting
33 to climate change (e.g., Purkey et al. in press; Freed and Sussman in press). In addition, CSSs are
34 typically managed by public agencies through decision-making processes that can be influenced
35 by EPA, and are the subject of strategic analysis at the Agency due to the large gap in funds
36 available versus funds needed for treatment system improvements. Thus, this study also
37 addressed the question of how climate change could influence the costs of complying with CSO
38 requirements. The Great Lakes (GLR) and New England (NER) Regions were selected as the
39 focus of this study because of the large number of CSSs in these areas, and because of EPA
40 GCRP's previous work with stakeholders in these regions as part of the regional assessment
41 program.
-------
28
i 2
2 The assessment methodology compares historical and future precipitation characteristics in the
3 Great Lakes and New England Regions using a benchmarking approach. The LTCP presumption
4 approach of reducing CSO events to no more than four per year was used as the mitigation
5 target. Using this mitigation target, a historical benchmark event was determined against which
6 projected precipitation events could be compared. Historical and projected precipitation data
7 were obtained from the Vegetation/Ecosystem Modeling and Analysis Project (VEMAP), which
8 uses projections of future precipitation from two global circulation models (GCMs): the Hadley
9 Centre Model and the Canadian Climate Centre Model. CSS communities in the Great Lakes and
10 New England Regions were mapped to VEMAP data points. The impacts of future climate
11 change on CSO frequency and mitigation requirements were then estimated by comparing the
12 benchmark event intensity representing four CSOs per year based on historical versus projected
13 future precipitation data. The following sections provide a more detailed discussion of the
14 assessment methodology.
15 2.1 G."".
16 A national list of CSS locations was obtained from a 2004 EPA Report to Congress (US EPA
17 2004). Latitude and longitude were determined for each CSS by cross referencing NPDES permit
18 numbers with location information in the Permit Compliance System (PCS), or based on city
19 location. CSS communities within the Great Lakes and New England Regions were selected as
20 described in the following sections.
21
22 The Great Lakes are part of the largest freshwater system in the world, and are bounded by eight
23 states: Minnesota, Wisconsin, Michigan, Illinois, Indiana, Ohio, Pennsylvania, and New York.
24 (US EPA 2004; GLRA 2006). 182 CSS communities with active CSO permits were identified
25 within the Great Lakes watershed; Table 1 presents a breakdown of CSS communities within the
26 Great Lakes watershed by state.
27 Table 1. GLR CSS Communities by State
State
Minnesota
Wisconsin
Michigan
Illinois
New York
Indiana
Pennsylvania
Ohio
Total
Number of CSS
Communities
3
2
46
34
23
24
3
47
182
-------
1 2.1.2
2 The New England Region was defined to include seven states: Maine, New Hampshire,
3 Vermont, Massachusetts, Connecticut, Rhode Island, and New York (excluding NYC). 135
4 active CSO permits were identified in the New England Region (US EPA 2004). Table 2 shows
5 the breakdown of CSS communities by state in the New England Region.
6 Table 2. New England Region CSS Communities by State
c. . Number of CSS
State
Communities
Connecticut
Maine
Massachusetts
New Hampshire
New York (upstate)
Rhode Island
Vermont
Total
5
39
22
6
53
3
7
135
7
8 2,2
9 There is considerable heterogeneity among CSSs in terms of baseline water quality conditions,
10 progress toward complying with EPA's national CSO Control Policy, and the site-specific
11 approaches that will be used to reduce the frequency of CSO events. It is required by EPA,
12 however, that all CSS communities develop a LTCP that includes an evaluation of alternatives to
13 meet CWA requirements by using either the "presumption approach" or the "demonstration
14 approach". One of the most common design objectives of LTCPs is to achieve EPA's
15 presumption approach standard of no more than an average of four CSO events per year. Under
16 this criterion, a CSO event is defined as any overflow from a CSS that does not receive the
17 minimum level of treatment defined in the CSO Control Policy.
18 2.2.1
19 In this study it was assumed that each CSO community in the GLR and NER will design their
20 system to achieve an average of four CSO events per year (i.e., the presumption approach
21 threshold), and will base their design on historical precipitation data. Given this standard, the
22 "benchmark" storm event that will need to be captured to meet the four-event per year average
23 was determined. By characterizing daily precipitation data in this way, a straightforward means
24 of comparing historical and projected precipitation characteristics was created. Historical
25 precipitation data were compared with precipitation projections from the two global circulation
26 models (GCMs) used in the VEMAP analysis.
27 2.2.2 data and
28 Historical and projected daily precipitation data were obtained from the Vegetation/Ecosystem
29 Modeling and Analysis Project (VEMAP), administered by the National Center for Atmospheric
30 Research (NCAR) data group. Projected future precipitation datasets were created using two
31 global circulation models (GCMs): the Canadian Centre for Climate Modeling and Analysis and
-------
1 the Hadley Centre models created in 1995. The GCM outputs were converted to daily projections
2 based on a modified version of the stochastic weather generator WGEN, (Richardson and Wright
3 1984).2 The GCM outputs and the daily scaling incorporate many limitations and a considerable
4 degree of uncertainty, as described exhaustively in the VEMAP documentation
5 (http://www.cgd.ucar.edu/vemap/index.html). Although VEMAP simulations are several years
6 old and subject to limitations, the VEMAP results were deemed to be suitable for a screening
7 level assessment determining the potential magnitude and direction of climate change impacts.
8 2.2.3
9 The locations of CSS communities were matched with historical and projected precipitation
10 datasets from the VEMAP grid located in the GLR and NER. For each region, the CSS
11 communities were assigned to the nearest U.S. VEMAP site for which data were available. A
12 historical dataset compiled by VEMAP was selected for each CSS community, based on the grid
13 point to which each was mapped.
14 For the GLR, historical precipitation data for a 40-year period (1954-1993) were compared to
15 projected precipitation data for a future 40-year period (2060-2099). For the NER, historical
16 precipitation data for a 25-year period (1968-1993) were compared to projected precipitation
17 data for a future 25-year period (2025-2050). The NER study was done subsequent to the GLR
18 study, and the 25-year time period starting in 2025 was selected to provide information on
19 climate change impacts that are more immediate in nature.
20 The historical and projected precipitation data were analyzed based on both a 1-day and a 4-day
21 moving average. The moving average approach was used to allow the temporal characteristics of
22 storage in CSO mitigation measures to be simulated, i.e., the time lag between onset of
23 precipitation and peak flows (and demand for storage) within a CSS. The choice of 1-day and 4-
24 day timeframes provided lower and upper bounds on both (a) time of travel within the sewershed
25 or area draining to the treatment plant (from "upstream" boundaries to the treatment plant) and
26 (b) the effects of multiple rain events in quick succession.
27 The precipitation intensity value that corresponds to the theoretical benchmark event was
28 determined based on the assumption that CSSs in the GLR and NER will design their systems to
29 meet the four-event per year standard based on historical precipitation intensity. This served as
30 the historical benchmark event for each system. In theory there will be only four rain events (in
31 an average year) that exceed this benchmark if a CSS community is meeting the objectives of the
32 CSO Control Policy. In the case of the GLR, the benchmark event was identified as the 160*
33 largest precipitation intensity values (in inches/day) in each of the 40-year aggregated 1-day and
34 4-day average precipitation datasets (4 events per year * 40 years of data). For the NER, the
35 benchmark event was identified as the 100* largest precipitation intensity values in each of the
36 25-year aggregated 1-day and 4-day precipitation datasets (4 events per year * 25 years of data).
WGEN is a weather simulation model developed at the USDA-ARS Grassland, Soil and Water Research
Laboratory that is used to scale down GCM outputs to a daily time-step. The model uses a probability function
(first-order Markov chain) where the chance of precipitation is conditioned on the wet or dry status of the previous
day, and the intensity is based on a gamma distribution where small events occur more frequently than large events.
July 2008
-------
1 Note that in each region, this corresponds to the precipitation event (1-day or 4-day) with a
2 recurrence interval of 3 months, or a probability of occurrence of 0.011 on any given day. This
3 approach was used to identify the precipitation event intensity that would have to be mitigated to
4 meet the CSO Control Policy goal of an average of four events per year over the selected time
5 periods.
6 The future four-event per year benchmark event was determined for each VEMAP grid location
7 using both the Canadian and Hadley models. Each VEMAP grid location was also weighted
8 according to the number of CSSs represented by that grid location. By evaluating the projected
9 precipitation data in this manner, the theoretical number of excess events and the incremental
10 increase in the benchmark event intensity (as an indicator of required CSO storage capacity) was
11 determined.
July 2006 ' " . .<' ' . tie or quote. 10
-------
3 3.1
4 The results of this analysis provide an estimate of the percent change in the frequency of CSO
5 events in the future under climate change relative to the four-event per year threshold based on
6 historical data. It is assumed that municipalities will design mitigation measures (e.g., a deep
7 storage tunnel) to meet this standard, and if historical precipitation is used as the design standard,
8 the effectiveness of those mitigation measures could be reduced in the future. The basic metric
9 presented in this section is the percent change in CSO event frequency relative to the four-event
10 per year presumption approach benchmark event. For example, a 50 percent change in CSO
11 event frequency would equate to two additional CSO events per year, on average, or a total of six
12 CSO events per year if the mitigation measures were designed using historical precipitation data.
13 3.1.1
14 Results show that relative to an assumed four event per year benchmark event, the average
15 annual CSO frequency in the GLR would increase between 13 percent (Canadian model, 1-day
16 averaging period) and 70 percent (Hadley Centre model, 4-day averaging period). In other
17 words, the average number of CSO events per year would increase to 4.5 using the lowest
18 projected average change, and 7.1 using the highest average projected change (see Table 3).
19 Table 3. Weighted Average Percent Change in CSO Frequency in the Great Lakes Region, 2060-2099. The
20 percent change is expressed relative to the four events/year standard, e.g., a 50% change equals two
21 additional CSO events/year.
Moving Average Canadian Hadley
1-Day 13.4% 49.4%
4-Day 18.8 % 70.0 %
22
23 Figure 2 shows the distribution of the percentage change in CSO frequency (number of
24 communities) in the GLR based on a 1-day moving average of daily precipitation. The Hadley
25 model projects an increase in precipitation intensity for all locations, while the Canadian model
26 projects decreases at 10 of the communities (5 percent). Accordingly, based on the Canadian
27 model, there are 10 communities that are projected to experience a decrease in CSO frequency of
28 10 to 0.1 percent. At the other end of the spectrum, the Hadley model predicts an 80 to 90
29 percent increase in CSO frequency for 10 communities. For these 10 CSS communities, this
30 would mean an additional 32 CSO events per year, on average. Figure 3 shows a similar plot
31 using the 4-day averaging period. These results generally indicate wider distributions and greater
32 impacts (though for the Canadian model, there is an increase in the number of communities with
33 fewer CSO events). Figure 34 shows the cumulative distribution for the Hadley and Canadian
34 models for both averaging periods.
35
July 2008 Draft Report: Do not ate or quote. 11
-------
Climate Change Impacts on Combined Sewer Overflows
Figure 2. Percent Change in Frequency of CSO Events in the Great Lakes Region, 2060-2099 (1-Day)
2
3
4
0)
E
o
O
O
8
0)
.a
E
~.Q>
~.Q> ~
~.Q>
<£>"
j£
°>
' ' Q>' Q>' Qt' Qt'
a q> 0) 0)
Change Interval (percent)
July 2006
Draff Report: Do not cite or quote.
12
-------
Climate Change Impacts on Combined Sewer Overflows
Figure 4. Cumulative Distribution of % Change in CSO Frequency in the Great Lakes Region, 2060-2099
50 100
Percent Change
150
200
Canadian 1-day
Hadley 1-day Canadian 4-day
Hadley 4-day
3
4
5
6
7
10
11
12
13
14
15
16
17
18
19
20
21
22
23
3.1.2 New England Region
Results for the NER show that relative to an assumed four event per year benchmark, the average
annual CSO frequency in the NER during the 2025-2050 period would change within a range of
-24 and 14 percent (Table 4). In other words, the average number of CSO events per year would
decrease to 3.0 using the lowest projected average change, and 4.6 using the highest average
projected change.
Table 4. Weighted Average Percent Change in CSO Event Frequency in the New England Region, 2025-2050
Moving Average
Canadian
Hadley
1-Day
4-Day
-10.4 %
-24.5 %
8.8 %
14.4 %
Figure 5 shows the distribution of the percentage change in CSO frequency (number of
communities) in the NER based on a 1-day moving average of daily precipitation. The Hadley
model projects an increase in precipitation intensity for the majority of locations, while the
Canadian model projects decreases for the majority of the communities. Based on the Canadian
model, there are 10 communities that are projected to experience an increase in CSO frequency
of more than 10 percent, 76 communities that will have less than a +10 percent change, and 49
communities that are projected to experience a decrease in CSO frequency of more than -10
percent. Alternatively, the Hadley model predicts 48 communities to have an increase in CSO
frequency of more than 10 percent, 64 communities with less than a +10 percent change, and 23
communities with decreases in CSO frequency exceeding -10 percent.
Figure 6 shows a similar plot using the 4-day averaging data. As shown in Figure 7, the 4-day
results generally indicate a greater decrease in frequency than the 1-day results based on the
Canadian model. There is little difference in these distributions based on the Hadley model.
July 2006
Draft Report: Do not cite or quote.
13
-------
Climate Change Impacts on Combined Sewer Overflows
Figure 5. Percent Change in Frequency of CSO Events in the New England Region, 2025-2050 (1-Day)
2
3
4
5
6
I/)
0)
'E
3
E
o
O
8
0)
.a
t^N ^X.
Change Interval (percent)
Figure 6. Percent Change in Frequency of CSO Events in the New England Region, 2025-2050 (4-Day)
100
c
3
O
O
8
0)
.a
E
80-
70-
60
40-
D Canandian Hadley
0)0)0)0)
?vvv
Change Interval (percent)
July 2006
Draff Report: Do not cite or quote.
14
-------
Climate Change Impacts on Combined Sewer Overflows
1
2
3
4
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
Figure 7. Cumulative Distribution of % Change in CSO Event Frequency in the New England Region, 2025-
2050
-50
-25
25 50 75
Percent Change
100
125
150
175
Canadian 1-day
Hadley 1-day Canadian 4-day
Hadley 4-day
3.2 CSO Benchmark Event Intensity
A second objective of this analysis was to estimate the projected rainfall event intensity that
would need to be captured to meet the four-event per year threshold under future climate
conditions. Event intensity values provide an indication of how CSO mitigation design
parameters would need to be modified to account for future climate change. This information
could be useful to municipalities interested in implementing CSO mitigation measures (e.g., a
storage basin) that are robust to future changes in climate. The basic metric presented in this
section is the percent change in future benchmark event intensity relative to the historical
benchmark under the four-event per year presumption approach. In this case, a 10 percent
increase in benchmark event intensity would imply that the design of the system would need to
be sized for a roughly 10 percent larger storm to account for climate change. Although the
relationship between precipitation event intensity and CSO volume generated is not perfectly
linear, this analysis provides an approximation of the change in design capacity that would be
needed to adapt to climate change. An evaluation of potential costs associated with increasing
design capacity to adapt to climate change is also presented.
3.2.1 Great Lakes Region
The average rainfall intensity corresponding to a recurrence interval of 4 events per year in the
GLR is projected to increase by between 5 and 16 percent (see Table 5).
July 2006
Draft Report: Do not cite or quote.
15
-------
Climate Change Impacts on Combined Sewer Overflows
Table 5. Average Percent Change in Benchmark Event Intensity in the Great Lakes Region, 2060-2099
Moving Average
Canadian
Hadley
1-Day
4-Day
4.8
5.1
16.2
14.9
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
Individual CSS communities in the GLR, however, are projected to have future reductions as
well as increases in the CSO benchmark event intensity. Figure 8 shows the distribution of
changes in CSO benchmark event intensity based on a 1-day moving average of daily
precipitation. Based on the Canadian model, there are four communities that are projected to
experience a decrease in CSO benchmark event intensity between -5 to -0.1 percent. At the other
end of the spectrum, the Hadley model predicts a 25 to 29.9 percent increase in CSO benchmark
intensity for two communities. Where an increase in rainfall benchmark event intensity is
projected, the design capacity of a CSS storage system would need to increase in order to
maintain the presumption approach standard of no more than four events per year. Where there is
a projected decrease in benchmark event intensity, mitigation measures designed using historical
precipitation characteristics would be more effective in reducing CSOs than their design
objective.
Figure 9 shows the distribution of changes based on the 4-day averaging period. As with the
change in event frequency, for the Canadian model, the distributions for the 4-day period are
shifted to the right compared to those with the 1-day period. However, the distribution for the 4-
day period is shifted to the left slightly for the Hadley model. Figure 10 presents the cumulative
distributions of the 1-day and 4-day averaging periods for percent change in benchmark intensity
under the Hadley and Canadian models.
Figure 8. Percent Change in CSO Benchmark Event Intensity in the Great Lakes Region 2060-2099 (1-Day)
23
24
~<2>
..
^
Change Interval (percent)
July 2006
Draff Report: Do not cite or quote.
-------
Climate Change Impacts on Combined Sewer Overflows
4
5
6
7
9
10
Figure 9. Percent Change in CSO Benchmark Event Intensity in the Great Lakes Region, 2060-2099 (4-Day)
120
110
JlOO
| 90
E 80
o 70
O
OT en
O °0
| 20
z 10
o-l-
D Canadian Hadley
,
-
Change Interval (percent)
Figure 10. Cumulative Distribution of % Change in Benchmark Event Intensity in the Great Lakes Region,
2060-2099
10 15 20
Percent Change
25
30
35
Canadian 1-day
Hadley 1-day Canadian 4-day
Hadley 4-day
3.2.2 New England Region
In the NER, the average rainfall benchmark intensity corresponding to a recurrence interval of
four events per year is projected to change within the range of-6 to +3 percent (Table 6).
July 2006
Draff Report: Do not cite or quote.
17
-------
Climate Change Impacts on Combined Sewer Overflows
Table 6. Average Percent Change in Benchmark Event Intensity in the New England Region, 2025-2050
Moving Average
Canadian
Hadley
1-Day
4-Day
-3.2%
-5.7%
2.8 %
3.1%
4
5
6
7
10
11
12
13
14
15
Some CSS communities are projected to have reductions, and others to have increases in the
CSO benchmark event intensity due to differences in the GCM model projections. Figure 11
shows the distribution of changes in CSO benchmark intensity based on a 1-day moving average
of daily precipitation. Based on the Canadian model, only three communities are projected to
experience an increase in CSO benchmark event intensity exceeding 10 percent, while the
Hadley model projections imply that 21 communities would have increases in event intensity
exceeding 10 percent.
Figure 12 shows the distribution for the 4-day averaging period. As with the change in frequency
of events exceeding the historical benchmark, the distributions for the 4-day period under the
Canadian model are shifted slightly to the left, indicating a decrease compared to those with the
1-day period. Figure 12 presents the cumulative distributions of the 1-day and 4-day averaging
periods for percent change in benchmark event intensity under the Hadley and Canadian models.
Figure 11. Percent Change in CSO Benchmark Intensity in the New England Region, 2025-2050 (1-Day)
16
17
120
Change Interval (percent)
July 2006
Draft Report: Do not cite or quote.
-------
Climate Change Impacts on Combined Sewer Overflows
Figure 12. Percent Change in CSO Benchmark Intensity in the New England Region, 2025-2050 (4-Day)
2
3
4
5
7
8
9
10
11
12
13
14
mn
~ 80 -i
3
E
0 60
O
W40J
M
O
* 20 -I
3
^ n
/*«/*
V? Q
N
Vi
D Canadian Hadley |
1
-*
Change Interval (percent)
Figure 13. Cumulative Distribution of % Change in Benchmark Intensity in the New England Region,
2025-2050
-20
-10 0 10 20
Percent Change
30
40
Canadian 1-day
Hadley 1-day Canadian 4-day
Hadley 4-day
3.2.3 Costs of Adaptation Increased Mitigation Design Capacity
It is ultimately of concern to water managers how to adapt to future climatic conditions. In
considering how system capacity might have to change to accommodate the altered weather
patterns of the future, several considerations must be kept in mind. CSO volume is a function of
many factors, including sewer system conveyance and storage capacity, wastewater treatment
plant capacity, and inflow and infiltration rates. To provide a rough bounding analysis of the
costs of adaptation, two runoff scaling factors were used to estimate the change in runoff
compared with a change in precipitation intensity. Chiew et al. (1993) suggest that the percent
July 2006
Draff Report: Do not cite or quote.
-------
1 change in runoff is about twice the percent change in precipitation in wet and temperate areas.
2 The United Nations Food and Agriculture Organization (FAO) suggests that runoff increases at
3 half the rate of the increase in precipitation (FAO 1991). We used these marginal runoff
4 production rates as lower and upper bounds on the change in runoff per unit change in
5 precipitation.3
6 A future decrease in the intensity of the benchmark storm would improve the effectiveness of
7 control measures at reducing CSO frequency and volume. Therefore, four examples were
8 selected from the NER in which both models project a significant increase in the magnitude of
9 the benchmark storm for the 1-day average. Table 7 shows the projected change in runoff for
10 these examples using the Chiew and FAO scaling factors.
11 Table 7. Projected Change in Runoff Volume, 1-day Averaging Period
VEMAP Location
Warren Co., PA*
Allegany Co., NY
St. Lawrence Co., NY
Wyoming Co., NY
% Change,
Canadian
6.8
4.0
10.3
8.3
0.5x
scaling
3.4
2.0
5.2
4.2
2x
scaling
13.6
8.0
20.6
16.6
% Change,
Hadley
30.3
19.3
5.3
16.2
0.5x
scaling
15.2
9.7
2.7
8.1
2x
scaling
60.6
38.6
10.6
32.3
12
13 A CSS engineer or planner designing a system to accommodate increased runoff would have to
14 weigh the increased system costs against the risks of more intense storm events. The most
15 common approach to controlling greater runoff is to increase storage capacity. Capital cost
16 functions reported by EPA for constructing storm water mitigation systems suggest that costs
17 have a scaling factor, with respect to flow, on the order of 0.7 (US EPA 2002a). As an example,
18 Wyoming County, NY, is projected to have an increase in runoff of between 4 percent and 32
19 percent, using the 1-day Hadley model projections and the two runoff scaling factors. This
20 suggests a cost increase of between 3 and 21 percent, depending on the scaling factor.
21 The actual costs for a CSS would vary depending on any economies of scale related to estimated
22 construction costs for mitigation measures associated with Long Term Control Plans based on
23 historical precipitation data. Planners would have to consider the cost of constructing now versus
24 the cost of doing so, if necessary, in the future, as well as the cost of borrowing additional money
25 now for construction versus the cost of borrowing in the future.
3 Note that these estimates may be conservative; recent research conducted in the United Kingdom estimates that a
20 percent increase in rainfall due to climate change would require a seven-fold increase in CSO storage volume to
maintain performance (Wilkinson and Balmforth 2004), implying a rate of 35% increase in runoff per 1% increase
in precipitation.
July 2006 ' " . .<< ' . y'te or quote. 20
-------
1 To the extent that planners modify their LTCPs to account for an additional margin of safety to
2 account for climate change, it would require additional capital investment. This would place
3 additional burden on already limited funds for CSO mitigation projects at all levels of
4 government. As is indicated in the 2000 Clean Water Needs Survey, which was based on LTCPs
5 that had been submitted as of 2000, CSO mitigation efforts in the surveyed communities are
6 projected to require $50 billion over the next 20 years (US EPA 2003). This is in addition to
7 other drinking water and wastewater treatment needs that are in the tens of billions of dollars.
9 This analysis is strongly influenced by variability in climate change projections in time and
10 space. In the case of the long-term (2060-2099) projections for the Great Lakes, the Canadian
11 and Hadley GCMs both predict an increase in CSO event frequency (with respect to a historical
12 benchmark) and the benchmark for the 4-event per year storm. However, with the near-term New
13 England Region projections, the models produced inconsistent results; the Hadley GCM
14 predicted an increase in CSO frequency and the benchmark event, and the Canadian GCM
15 predicted a decrease in frequency and benchmark event.
16 This variability has important implications for how results should be interpreted. For example,
17 based on the Canadian model, if control measures in the NER are designed based on historical
18 precipitation characteristics, climate-related changes could result in a 17 percent decrease in
19 CSO events per year across averaging periods, i.e., an increase in the effectiveness of mitigation
20 measures. A similar analysis based on the Hadley model predicts a 12 percent increase in
21 overflows. Across all 135 communities, this translates to 94 fewer and 62 more events per year
22 above the control policy's objectives under the Canadian and Hadley models, respectively.
23 The contrasting GCM results found in the NER highlight the difficulty of making clear
24 predictions for specific communities. The Canadian and Hadley models both project that region-
25 wide in the NER, annual average precipitation will increase as of 2100 by about 10 and 25
26 percent, respectively, compared to 2000 levels. However, making projections over the shorter
27 term, and downscaling both in space (regions to grid cells to communities) and time (monthly
28 projections to daily values), the picture becomes more opaque.
29 Moreover, our methodology focused on values at the high end of the precipitation distribution -
30 the 100th largest out of 9,130 values (25 years) for the NER and the 160th largest out of 14,600
31 values (40 years) for the GLR. The validity of these values is a function both of the validity of
32 the GCM projections and the validity of the downscaling approach used to manipulate those
33 projections. Most models are more valid in their estimation of central tendencies than in the tails
34 of the output distribution, and to the extent that we are concerned here with the high-end tail, the
35 results should be used with caution.
36 3,4
37 There are several limitations to consider when interpreting the results of this research. These
38 limitations are related to the analytic framework and the data. These limitations, along with the
39 screening-level findings, point to the potential for future research in this area.
40 3,4,1 Limitations
41 Key limitations include the following:
-------
1 Presumption Approach - The CSO Control Policy "Presumption Approach" standard of
2 no more than four events per year is assumed to be the main objective of CSS managers,
3 and increasing system storage is the primary mitigation measure used to meet this
4 standard. In practice there are many site-specific issues that make development of long-
5 term control plans very complicated, and not all planners choose to increase storage as
6 their primary mitigation measure. Moreover, the Control Policy also allows use of a
7 "demonstration approach" based on levels of effluent treatment and water quality
8 standards of the receiving waterbody. Thus, the analysis takes a relatively simple
9 approach to a very complex and heterogeneous system, and it lacks the technical
10 robustness that would result from the use of sewer system models such as the Storm
11 Water Management Model (SWMM).
12 Benchmarking - Our analysis is based on a benchmarking comparison of historical and
13 future daily precipitation based on weather observations and GCM projections. The
14 statistical techniques used to determine benchmarks are often more sophisticated and
15 complicated than the simple ranking technique used in this study, and may yield slightly
16 different results. Moreover, the historical datasets used to establish storm intensity,
17 duration, and frequency may go back further in time than the 25-year period used for the
18 NER or the 40-year period used for the GLR. To the extent that there are underlying
19 trends in precipitation4 toward increasing intensity of storm events, using a thin temporal
20 slice of the historical dataset tends to understate the difference between design objectives
21 based on observed weather and future conditions.
22 General Circulation Models - The GCM results EPA used (Canadian and Hadley) are
23 projected on a monthly time step. They were then converted to daily projections based on
24 the stochastic weather generator, WGEN. The initial GCM outputs and the daily scaling
25 incorporate many limitations and a considerable degree of uncertainty, as described in
26 section 3.3 and more exhaustively in the VEMAP documentation. All of those limitations
27 and uncertainties apply directly to these results.
28 Averaging Period - Although the analysis uses 1-day and 4-day averaging periods to
29 bracket the effects of short intense storms and longer or multiple precipitation events in
30 sequence, it is possible that by choosing shorter or longer periods the results would
31 change. For example, the approach does not distinguish between intense precipitation
32 "pop-up" events (e.g., a thunderstorm yielding two inches of rain in one hour) versus a
33 steady rain that accumulates 2 inches over a full 24-hour period. These short but intense
34 events can quickly overwhelm a CSS and result in an overflow or surcharged sewer
35 condition. A higher resolution evaluation of hydraulic effects on an hourly time-step
36 could only be analyzed using more sophisticated models (e.g., SWMM) and more fme-
37 grained precipitation data.
4 For example, the observed 20th century values for annual precipitation were up to 25 percent greater than pre-20*
century records for the eastern coastline of the NER, where many of the CSS communities are located (US EPA
2002b).
; " . . ' ' - yte or quote. 22
-------
1 Snowmelt - Our analysis of total daily precipitation does not distinguish between rain and
2 snow. Snowfall is less likely than rainfall to result in elevated peak flows in a CSS at the
3 time of the precipitation event, but snowmelt at a later date can have significant impacts
4 on sewer flows, especially if the snowmelt occurs during a rain event. Moreover, rain on
5 snow events can result in significant runoff, with virtually all of the rainfall converted to
6 runoff, supplemented by melting snow. The evaluation of total daily precipitation does
7 not differentiate between rain and snow. This is a significant limitation for our analyses
8 of the GLR and NER, given the generally cold temperatures throughout these regions
9 during the winter months.
10 3,4.2
11 This study was intended as a screening level analysis to determine the potential scope and
12 magnitude of global change impacts rather than a detailed assessment of specific impacts and
13 adaptation measures. The results of this analysis suggest that more detailed analysis would be
14 worthwhile in several areas:
15 Align Time Periods of Study - Additional studies of this nature would benefit from an
16 alignment of time periods selected for determining historical and projected climate
17 conditions in different regions. Rerunning the CSO benchmarking analysis in the GLR
18 and NER to extend and synchronize the time periods utilized for each might help resolve
19 the disparate results presented in the study. If the longer-term results are consistent, it
20 would help dispel some of the ambiguity evident in the near-term NER results.
21 Utilize More Recent Climate Models - Newer GCM runs have become available since the
22 VEMAP II dataset was created in the mid 1990s. There are also additional approaches for
23 obtaining daily time-step projections through updated weather generators or GCMs that
24 generate daily precipitation data. More up-to-date GCM results would reflect advances in
25 the state of the science, and regional models (e.g., the UK Hadley Centre's PRECIS
26 model) can provide better resolution on a regional basis.
27 Case Studies - A smaller scale case study approach looking in detail at a few
28 communities might provide more accurate data on system responses on an hourly basis.
29 These case studies would likely involve the use of detailed sewershed runoff models such
30 as SWMM. These models typically utilize continuous precipitation data, so a method for
31 applying GCM outputs to modify historical continuous precipitation data would need to
32 be created. This would provide a more robust analysis as event intensity on an hourly (or
33 shorter time-step) basis for predicting CSO events would provide a more accurate basis
34 than the daily precipitation data utilized for this screening study.
35 Determine Best Practices for Characterizing Design Storms - One opportunity to provide
36 useful guidance would be to establish a straightforward approach for modifying the
37 Intensity-Duration-Frequency (DDF) curves and design storms commonly utilized for
38 water resource engineering design and planning. To develop guidance, it would be useful
39 to review (1) the current practices associated with creating IDF curves (e.g., how far back
40 in the historical record, what statistical techniques are used) and (2) the extent to which
41 recent trends in increasing storm intensity are already embedded in the IDF curves. One
42 possible approach for modifying the way that design storms are calculated would be to
43 utilize research examining trends in precipitation intensity over the last century. For
, -' i art: Do not ate or quote. 23
-------
1 example, research by Groisman et al. (2005) evaluated historical precipitation data for
2 the US and determined that there were statistically significant trends indicating an
3 increase in the intensity of the heavy (upper 5 percent of precipitation) events.
4 Specifically, their research has found a 4.6 percent increase in event intensity per decade
5 for the largest 5 percent of precipitation events; a 7.2 percent increase in event intensity
6 per decade for the largest 1 percent of precipitation events; and a 14.1 percent increase in
7 event intensity per decade for the largest 0.1 percent of precipitation events.
8 Relationships like these could be used, along with assumptions on the design lifetime of
9 CSSs, to provide adjustment factors for characterizing future storm intensity.
10 Margin of Safety - Engineering design practices for CSS improvement projects typically
11 include a margin of safety (MOS) to account for uncertainties. It would be informative to
12 quantify the magnitude of the MOS typically utilized in CSO mitigation design.
13 Moreover, it would provide a useful perspective to put the increase in design storm
14 intensity (due to climate change) in the context of this margin of safety, as well as the
15 potential for increase due to future changes in impervious surface, population, and other
16 key design factors.
July 2006 ' " . .<< ' . y'te or quote. 24
-------
1 4 <
2 If climate change does indeed result in flashier precipitation events, changes in annual
3 precipitation, earlier snow melt, and rising sea levels, water resource managers will need to
4 adapt. Our analysis of long-term CSO mitigation in the GLR and NER indicates that if systems
5 are designed based on historical precipitation characteristics, their efficacy in the future could be
6 diminished due to climate-related changes. Specifically, it is likely that in the long term (2060-
7 2099), CSO mitigation projects in the GLR designed using historical precipitation characteristics
8 will experience an increase in the frequency of CSO events beyond their design capacity, and an
9 increase in overflow volume discharged to receiving waterbodies. This could result in
10 communities failing to meet the CSO mitigation goals established in their Long Term Control
11 Plans, which in turn would result in the need for further modifications to the CSSs.
12 The results are inconsistent in the near term (2025-2050) in the NER, with different
13 directionality in the results of the Hadley and Canadian models. The differences in projected
14 precipitation changes between the GLR and NER demonstrates the regional texture of climate
15 change, which translates into different decisions that planners in these regions may be making to
16 protect wastewater treatment infrastructure.
17 The uncertainty associated with GCM projections (and especially the ambiguity in the near-term
18 NER results) may deter CSS planners from considering climate change as a design factor in their
19 long-term planning, given that CSO mitigation efforts are already highly complex and costly.
20 However, it is important to consider this uncertainty in context:
21 Planners already include other factors with similar uncertainty and long-term effects in
22 their design. Although precipitation intensity is a key design factor, so is total impervious
23 area, sewered area, population served, and per capita water demand. The design is quite
24 sensitive to all of these factors, and each is subject to considerable uncertainty. Because
25 the uncertainty is more familiar, however, planners are much more comfortable
26 characterizing these factors with assumptions and estimates.
27 Throughout much of the US, there is empirical evidence for a trend to increasing
28 precipitation intensity (Karl and Knight 1998). Irrespective of the uncertainties in the
29 GCM projections, it would be prudent to consider the likelihood of increasing storm
30 intensity based on weather observations.
31 It is possible that if more recent GCM results and downscaling routines were used, the
32 results would be less ambiguous in terms of the direction of the effect.
33 A report published for the Toronto-Niagara Region Study on Atmospheric Change recommends
34 that design storms utilized for storm water infrastructure construction be made 15 percent larger
35 to account for expected climate change (Watt et al. 2003). In engineering terms, this
36 recommendation essentially incorporates a margin of safety to specifically address climate
37 change in the system design.
38 The threat of increased precipitation variability in the future could provide additional motivation
39 for rigorous inflow and infiltration (I&I) mitigation programs that maximize the capacity of the
40 existing CSS. By eliminating flow in sewers due to groundwater infiltration and runoff from
41 gutter downspouts and sump pumps, additional capacity can be made available to reduce the
42 frequency and severity of CSOs. Many municipalities already have aggressive I&I programs in
25
-------
1 place, and the potential for increased precipitation intensity in the future under climate change
2 makes this mitigation option even more important.
3 States and local municipalities are already struggling to meet the demands of water resource
4 management, and climate change would add an additional factor to the equation when weighing
5 the price of clean water. The New York City Department of Environmental Protection estimates
6 that the cost of increasing wet weather capture from 75 percent to 95 percent in response to
7 climate change would cost $12-$40 billion for New York City alone(NYC DEP 2005).5 This is a
8 significant sum, and is close to the total cost of CSO mitigation efforts indicated for those
9 communities responding to the 2000 Clean Water Needs Survey. As indicated by the research
10 described in this report, it is clear that the costs of adapting CSO mitigation plans to manage the
11 long-term risks associated with climate change could significantly increase funding
12 requirements, exacerbating the existing gap between funding that is available and funding that is
13 needed.
5 The basis for the increase in percent CSO volume capture from 75 percent to 95 percent (and how it relates to the
CSO Control Policy) was not explained by the NYC DEP in the reference cited.
1 or quote. 26
-------
1
2
4 Burian S. I, Stephan J.Nix, Robert E.Pitt, and S. Rockey Durrans. 2000. Urban Wastewater
5 Management in the United States: Past, Present, and Future. Journal of Urban Technology, 7:33-
6 62.
7
8 Chiew F., P.Whetton, A. McMahon, and A.Pittock. 1993. Simulation of the impacts of climate
9 change on runoff and soil moisture in Australian catchments. Journal of Hydrology.
10
11 FAO. Water Harvesting: A Manual for the Design and Construction of Water Harvesting
12 Schemes for Plant Production. Food and Agricultural Organization of the United Nations . 1991.
13
14 Gleick, P. H. Water: the potential consequences of climate variability and change for the water
15 resources of the United States. 2000. Pacific Institute for Studies in Development, Environment,
16 and Security.
17
18 GLRA. Great Lakes Summary Report on Current Climate. Great Lakes Regional Assessment.
19 2006.
20
21 Groisman P. Y., R. W. Knight, D. R. Easterling, T. R. Karl, G. C. Hegerl, and V. N. Razuvaev.
22 2005. Trends in Intense Precipitation in the Climate Record. Journal of Climate, 18:1326-1350.
23
24 Houghton, J., Y.Ding, Y, Griggs, D, Noguer, M, van der Linden, P, Da, X, Maskell, K, and
25 Johnson, C. Climate Change 2001: The Scientific Basis. 2001. Contribution of Working Group I
26 to the Third Edition of the Third Assessment Report of the Inter Governmental Panel on Climate
27 Change.
28
29 Karl T. R. and R. W. Knight. 1998. Secular Trends of Precipitation Amount, Frequency, and
30 Intensity in the United States. Bulletin of the American Meteorological Society, 79:231-41.
31
July 2008 Draff Report: Do mi ate or quote. 27
-------
1 NAST. US National Assessment of the Potential Consequences of Climate Variability and
2 Change. 2000.
3
4 NYC DEP. Adapting NYC's Water Supply and Wastewater Treatment Systems to Climate
5 Change. 2005.
6
7 Richardson C. W. and D. A. Wright 1984. WGEN: A Model for Generating Daily Weather
8 Variables, United States Department of Agriculture, Agriculture Research Service.
9
10 Schladweiler, Jon C. Tracking Down the Roots of Our Sanitary Sewers: Design Choices and
11 Philosophies, sewerhistory.org . 2005.
12
13 US EPA. Combined Sewer Overflow (CSO) Control Policy. 59 FR 18688. 1994. United States
14 Environmental Protection Agency.
15
16 US EPA. Cost of Urban Stormwater Control. EPA-600-R-02-021. 2002a. National Risk
17 Management Research Laboratory, Office of Research and Development, United States
18 Environmental Protection Agency.
19
20 US EPA. 2002b. Impacts and Adaptation. Pages 81-112 U.S. Climate Action Report 2002.
21
22 US EPA. Clean Water Needs Survey: 2000 Report to Congress. 2003. United States
23 Environmental Protection Agency.
24
25 US EPA. Report to Congress: Impacts and Control of CSOs and SSOs. 833-R-04-001. 2004.
26 United States Environmental Protection Agency.
27
28 US EPA. Water Permitting 101. 2006. US EPA Office of Wastewater Management, United
29 States Environmental Protection Agency.
30
July 2006 ' " . .<< ' . y'te or quote. 28
-------
1 Watt, W., Waters, D, and McLean, R. Climate Change and Urban Stormwater Infrastructure in
2 Canada: Context and Case Studies. 2003. Hydrology Research Group, Department of Civil
3 Engineering, Queen's University.
4
5 Wilkinson, B and Balmforth, D. Effects of Climate Change on Sewer System Performance.
6 2004. The United Kingdom Water Industry Research and MWH UK, Ltd.
7
July 2006 ' " . .<< ' . y'te or quote. 29
------- |