United States
pv&A Environmental Protection
# % Agency
EPA/600/R-15/191 | September 2015
www.epa.gov/ord
Benefit Indicators for
Flood Regulation Services
of Wetlands:
A Modeling Approach
Photos by: Kenneth Bent (top left and bottom right)
Bruce G. Hooke (large photo, bottom left and bottom middle)
I
Office of Research and Development
National Health and Environmental Effects Research Laboratory, Atlantic Ecology Division
-------
-------
SEPA
EPA/600/R-15/191
September 2015
www.epa.gov/ord
Benefit Indicators for Flood Regulation
Services of Wetlands:
A Modeling Approach
Justin Bousquin
Kristen Hychka
(Oak Ridge Institute for Science and Education Participants)
U.S. EPA, Office of Research and Development,
National Health and Environmental Effects Research Laboratory
Atlantic Ecology Division, Narragansett, Rl 02882
Marisa Mazzotta
U.S. EPA, Office of Research and Development
National Health and Environmental Effects Research Laboratory
Atlantic Ecology Division, Narragansett, Rl 02882
U.S. Environmental Protection Agency
Office of Research and Development
National Health and Environmental Effects Research Laboratory
Atlantic Ecology Division
Narragansett, Rl 02882
-------
N'il' O s. Ti J I'"! i ! i! !'i s»T
This project was supported in part by appointments to the Oak Ridge Institute for Science and
Education (ORISE) participant research program supported by an interagency agreement
between EPA and DOE. Although the research described in this report has been funded by the
U.S. Environmental Protection Agency, it has not been subjected to Agency review. Therefore, it
does not necessarily reflect the views of the Agency. This is contribution number ORD-013568
of the Atlantic Ecology Division, National Health and Environmental Effects Research
Laboratory, Office of Research and Development, U. S. Environmental Protection Agency.
h'i -v-orcl
This report builds on research reported in Bousquin et al. (2014). Whereas the previous report
focused on developing an event-based flood modeling process that would explicitly account for
flood water retention in freshwater wetlands, this report improves the modeling and details
how that modeling process was used to develop indicators to assess increases in flood
protection benefits from potential wetlands restoration. The assessment of flood protection
benefits follows a non-monetary benefit indicators framework outlined in Mazzotta and
Wainger (in preparation). The intensive modeling performed for the watershed case study
presented in this report may be too extensive for some watersheds and flood benefits
assessments. In light of this, we used the results to develop a set of indicators that could be
collected without the modeling. In this report we outline a three-tiered approach, present
indicators for Tiers II and III, and describe the process one would follow to apply such indicator
sets. Future work will further investigate the transferability of these flood benefit indicators and
the modeling process to other watersheds, as well as ways of making the assessment and tools
more accessible to a wider audience of users.
ii
Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
Table of Contents
Notice and Disclaimer ii
Foreword ii
List of Figures v
List of Tables vii
Acknowledgements vii
Executive Summary viii
CHAPTER 1: INTRODUCTION 1
The Woonasquatucket River Watershed 3
Indicator Development 6
Reading this Report 7
CHAPTER 2: FLOOD MODELING 9
Introduction 9
Model Selection and Inclusion of Wetlands 10
Model Selection 10
The Hydrologic Model 11
The Hydraulic Model 12
Flood Impact Modeling 12
Treatment of Wetlands in the Models 13
Estimating Flow Diverted by Wetlands 13
Estimating Maximum Outflow Diverted 14
Wetland Flood Protection Functional Assessment 15
Opportunity 15
Effectiveness 15
Social Significance 15
Hydrologic Modeling with HEC-HMS 16
Parameterization 17
Subbasin Elements 17
Reach Elements 17
Reservoir Elements 17
Diversions 18
Baseflow 18
Precipitation and Runoff Data 18
Storm Selection for Calibration and Validation 18
Calibration 21
Validation 22
Sensitivity Analysis 25
Lag Method 25
Precipitation 25
Wetland Percent Diverted 26
Wetland Max Volume Method 27
Hydraulic Modeling with HEC-RAS 28
Wetland Restoration Scenarios 29
Table of Contents
-------
Size of Restoration in Scenarios 29
Location of Restoration in Scenarios 30
Storm Events 30
CHAPTER 3: INDICATORS 33
The Indicators Framework 33
Tier III Assessment 34
I. Assess existence of an ecosystem service 36
II. Assess temporal reliability of the service 38
III. Assess who benefits 39
IV. Assess the magnitude of benefits to individuals or households 41
Tier II Indicators 44
Generalizing from Tier III Model Results 44
Methods 44
Results 45
Interpretation of Results for Indicators 52
Applying Tier II Indicators 52
I. Assess existence of an ecosystem service 52
II. Assess temporal reliability of services 53
III. Assess who benefits 53
IV. Assess the magnitude of benefits to individuals or households 54
CHAPTER 4: DISCUSSION AND NEXT STEPS 55
Modeling Summary 55
Tier III Indicators 57
Tier II Indicators 57
Future Directions 58
Summary 59
References 61
APPENDICES 67
Appendix A - Wetland Flood Protection Functional Assessment 67
Appendix B-Wetland Parameters 68
Appendix C - PeakFQ Analysis 75
Appendix D - HMS Parameters 77
Curve Number Loss Method Parameters 77
Muskingum Routing Method Parameters 79
Parameter Calibration 81
Appendix E - Reservoirs 83
Appendix F - HMS Calibration and Validation Data 85
Appendix G - Synthetic Storms 89
Appendix H - Maps 90
Appendix I - Tier II Indicator Development 95
Appendix J - R Script to read in DSS Peak Flows and save as csv 97
Appendix K - Data Quality and Limitations 101
Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
List of Figures
Figure 1-1. The ecosystem service cascade (adapted from Potschin and Haines-Young 2011) 2
Figure 1-2. The Woonasquatucket River watershed within Rhode Island, USA showing land cover
and the location of the stream gage and its contributing area 4
Figure 1-3. Flood frequency curves for pre- and post-urban development (circa 1952- and 1972,
respectively) in the Woonasquatucket River (from Doehring and Smith 1978). The vertical scale
is flow normalized to the mean annual flood (CFS) and the horizontal scale is the recurrence
interval in years 5
Figure 1-4. Average annual precipitation for Rhode Island for the period 1930 to 2013
(Valle and Giuliano 2014) 5
Figure 1-5. Required resources and results uncertainty across the three tiers of indicators 7
Figure 2-1. Modeling Approach 9
Figure 2-2- Figure showing wetlands in subbasin based on different assumptions:
(a) how HMS implements diversions, (b) bio-physical reality, and (c) the wetland contributing
area assumption we used 14
Figure 2-3. Hydrograph of 15 min interval flow (USGS Station 01114500; black line in m3/s (CMS)
on left axis) and hourly precipitation (NOAA Coop 376698; gray line in cm on right axis) for the
"Fall '05" calibration storm (October 15, 2005). Raw flows are shown without baseflow removed 20
Figure 2-4. Hydrograph of 15 min interval flow (USGS Station 01114500; black line in m3/s (CMS)
on left axis) and hourly precipitation (NOAA Coop 376698; gray line in cm on right axis) for the
"Spring '01" calibration storm (March 22, 2001). Raw flows are shown without baseflow removed.... 20
Figure 2-5. Two sets of optimizations were performed in sequence for the Spring '01 storm.
The first (a) had 200 iterations the second (b) had 320. The objective function is defined
by how close the model matches the observed peak. The area circled in red had the optimal
parameters used by the model (Calibration 2) resulting in lowest objective function 21
Figure 2-6. Fall '05 storm (Oct '05) hydrographs for Observed Flow (Black), Calibrated Flow (Blue),
Calibrated Flow from Bousquin et al. (2014) (Light Blue), and Un-calibrated Flow (Red) 23
Figure 2-7. Spring '01 storm (Mar '01) hydrographs for Observed Flow (Black), Calibrated Flow (Blue),
and Un-calibrated Flow (Red) 23
Figure 2-8. Results of altering lag method in comparison to observed flow (Blue). The TR-55 method
(Green) was used in the model, where the Curve Number method (Red) was not 25
Figure 2-9. Precipitation for the different sensitivity analysis runs 26
Figure 2-10. Sensitivity analysis hydrographs based on varied precipitation 26
Figure 2-11. Resulting hydrographs from sensitivity analysis model runs with different wetland
percent diverted assumptions ranging from the model assumption (Blue) to a 75% increase in
wetland catchment area contributing to diversion (Pink) 27
Figure 2-12. Sensitivity analysis results hydrographs with different wetland max volume assumptions:
water retention height derived from average (blue) or minimum (red) elevation of wetland
perimeter vertices 28
Figure 3-1. Subbasins selected for the example application 35
Figure 3-2. Indicator application process 36
Figure 3-3. 2025 Projected Landuse for Subbasin 1540 (left) and Subbasin 2370 (right) 39
List of Figures
-------
Figure 3-4. Building locations in the altered flood zone. Subbasin 2370 is shown in purple on map
a, with its flood zone magnified in map b. Subbasin 1540 is shown in orange on map a,
with its flood zone magnified in map c 40
Figure 3-5. Census tract Social Vulnerability Index for Subbasin 1540 (left) and Subbasin 2370 (right)... 42
Figure 3-6. Buildings where flooding is decreased by restorations in Subbasin 2370 and 1540
(black points) compared to DFIRM flood zones (100-year and 500-year) in the FEMA modeled area
(orange) and outside the FEMA modeled area (white). Our model boundary (black outline)
is provided as a reference for building locations 43
Figure 3-7. The change in flow (%) with distance downstream (m) from the simulated restoration
for all subbasins (n=139) for each restoration scenario (wetland change scenarios 1%, 5%, 10%,
and 25%) and synthetic storm event (1-year, 5-year, and 25-year). Note that the y-axis scale
varies between sub-plots for each of the wetland change scenarios 46
Figure 3-8. Boxplots of downstream distance (m) to no change in flow (<-0.2% change) by
restoration scenarios: (a) storm recurrence interval, (b) wetland change, and (c) both wetland
and storm recurrence interval 47
Figure 3-9. Boxplots of downstream distance (m) to no change in flow (<1% change) by
restoration scenarios: (a) storm recurrence interval, (b) wetland change, and (c) both wetland
and storm recurrence interval 49
Figure 3-10. Downstream distance to no change in flow (<-0.2% change) plotted against
subbasin characteristics: distance (m) to the outlet (Dist_USGS_GAGE), subbasin area
(Basin_Area_km2), % wetlands (Pct_Wetlands), mean basin slope (Basin_Slope),
% impervious cover (Pctlmp), and the longest flowpath distance (m) (Longest_Flowpath) 51
Figure 3-11. Boxplots of downstream distance to no change in flow (<-0.2% change) plotted
by subbasin characteristics (distance to the outlet, subbasin area, % wetlands, mean basin slope,
% impervious cover, and longest flowpath) split into below (0) and above (1) the mean values
for the basin characteristics (Table 3-7) 51
Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
List of Tables
Table 2-1. HEC-HMS parameters 16
Table 2-2. Top 10 Streamflow peaks for Woonasquatucket River gage for 2000-2010 in order
of peak flow 19
Table 2-3. Validation Results for Spring '01 and Fall '05 models validated and fit across four
observed storms 24
Table 2-4. Approximate recurrence intervals for synthetic storms and their peak flows using
different models 31
Table 2-5. Number of simulations generated based on three synthetic storms, four restoration sizes
and 139 individual basins 31
Table 3-1. Assessment of supply and demand 37
Table 3-2: Subbasin and restoration characteristics 37
Table 3-3. Assessment of reliability 38
Table 3-4. Assessment of quality of service and beneficiaries' vulnerability 40
Table 3-5. Downstream distance (m) where change in flow became negligible (-0.2%) 48
Table 3-6. Downstream distance (m) where change in flow became negligible (-1%) 50
Acknoi
We would like to thank Caroline Druschke, Walter Berry, Anne Kuhn, Nathaniel Merrill,
Tim Gleason, and Wayne Munns for helpful comments on this manuscript.
List of Tables
-------
Executive .Sum in a ry
This report describes a method for developing indicators of the benefits of flood regulation
services of freshwater wetlands and presents a companion case study. The critical role of
wetlands in flood protection, the difficulty in modeling that role, the high value that
communities place on flood protection, and the fact that many ecosystem services assessment
tools do not address flooding led to our focus on indicators of the benefits of flood regulation
services provided by wetlands. We demonstrate our approach through an application to the
Woonasquatucket River watershed in northern Rhode Island.
The benefits indicators approach developed in this report attempts to provide decision makers
with a more accessible alternative to monetary valuation, with an approach that explicitly links
functions to benefits. We propose criteria to facilitate comparison of freshwater wetland
restoration scenarios based not only on production of ecosystem services, but also on how
those services reach and benefit people. These benefit indicators are intended to augment
existing assessment methods, to which they can add critical information about benefits to
functional and service assessments already in use. With additional effort, our method can be
extended to incorporate location-specific dollar values or can be used to improve benefit
transfer of dollar values by calibrating those transfers.
The purpose of the indicators we present is to provide metrics of the factors that influence the
spatial flow of services from production to benefits, by assessing how flood waters flow across
the geographic area between wetlands that retain or slow storm flows and the important
structures and other resources that might be protected from flood risks. We based indicators
on sophisticated flood modeling results to ensure the rigor of results, but we recognize that in
some contexts a rapid assessment that can be readily applied to inform a decision may be more
valuable than a complex approach that is not feasible to apply. In an attempt to satisfy the
requirements of different decision contexts, we created a 3-tiered approach. Based on the
appropriate level of rigor for the decision context, each tier becomes more resource intensive
to implement and requires more site-specific data. Certainty and robustness increase with each
indicator tier.
Our modeling approach includes several linked models, applied in sequence. First, we used
hydrologic models to estimate baseline hydrology in each basin and translate restoration
scenarios—or changes in wetland coverage—to changes in hydrology and resulting stream
hydrographs. Second, we incorporated these hydrographs into hydraulic models, which
produced flood depth profiles. Third, we applied the flood depth profiles, using flood impact
models, to map flood depth and extent and estimate the impacts of these floods on potential
beneficiaries. We identified potential beneficiaries using counts of structures in the flood-prone
area and estimated benefits as reductions in flooding depth, based on the flood maps produced
from simulations.
Vlll Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
In order to run scenarios for our case study watershed using these models, we first modeled
flow in the Woonasquatucket watershed under current conditions. We then calibrated that
model using observed flows at a USGS gage in the watershed from two separate storms. These
two calibrated models were validated against flows observed for two additional storms, and we
then chose the calibrated model that performed best based on this validation. We then
conducted sensitivity analysis to test the effects of varying important model parameters.
We developed 12 scenarios to investigate the influence of different levels of wetland
restoration across a range of storm events on stream flow and resulting flood extent. The range
of storm events were based on three synthetic storms of known recurrence intervals (1-year, 5-
year, and 25-year) that resulted in modeled flows of known recurrence intervals (10-year, 100-
year, and 500-year). We chose different levels of wetland restoration based on the range of
sizes of potential restoration sites previously identified in the watershed (Golet et al. 2003), and
modeled the different levels of wetland restoration in each individual subbasin by converting
1%, 5%, 10%, or 25% of subbasin non-wetland area to wetlands in our calibrated and validated
hydrologic model. We then simulated flow hydrographs for each subbasin, wetland restoration
level and synthetic storm event using these updated models. We extracted peak flows from
these hydrographs to estimate the extent of flooding under different conditions using the
hydraulic model. Based on the changes in flow and flooding under each of these simulations,
we developed indicators of wetland flood regulation services and benefits to people.
These ecosystem service benefit indicators are based on a framework developed by Mazzotta
and Wainger (in preparation). The framework uses four questions to guide the process of
indicator selection and measurement: (1) Is an ecosystem service supplied?; (2) How likely is it
that the service will continue to be provided over the long run?; (3) How many people benefit?;
and (4) By how much do people benefit? These questions are ordered so that answering each
question in turn contributes additional information to an ecosystem service benefits
assessment. Each of these questions may in turn be answered in lesser or greater detail.
Our Tier III indicators are based on direct results of the watershed-specific models. We
developed Tier II indicators by generalizing the model outputs, through finding those factors
that are the most critical determinants of who will benefit and where. Tier I indicators are
based on results in the literature. We do not present Tier I indicators in this report, which
focuses on describing the modeling to develop Tier II and III indicators and the process for
applying the Tier II and III approaches. The Tier I indicators will be included in our forthcoming
guide to applying this approach (Mazzotta et al. in preparation).
The Tier III indicators are based on modeled peak flows and flood maps for the
Woonasquatucket Watershed. Using these model results, we present an example of how one
would compare wetland restoration scenarios using our set of benefit indicators. Using our
models, a user could compare scenarios for other subbasins within the Woonasquatucket
Watershed. To apply the Tier III approach to another watershed, a user would need to perform
extensive modeling similar to that described here. For those who want to apply this approach
Executive Summary
-------
to a different watershed, we describe the models and process involved in our application,
including caveats and factors that may differ for other locations.
We developed the Tier II indicators from trends and sensitivity observed in our modeling
results, where we generalized the model results to determine the factors that relate differences
in potential restoration sites to differences in flood reduction services and the benefits of
subsequent flood damage reduction. The primary consideration for generalizing to Tier II
indicators is the need to determine the relevant benefits area for flood regulation services.
Flood regulation occurs at the site of a wetland, but benefits people and structures
downstream of the wetland site. Therefore, our primary aim in generalizing from our detailed
modeling was to determine the "benefits area" and the factors that may influence how this
varies. Based on our analysis, a reasonable and conservative distance for delineating the area
where people could benefit is 4 km (2.5 mi) downstream of the restoration, based on the
average over all scenarios. A reasonable upper estimate, based on the average distance for a
25% restoration during a 1-year storm event, includes beneficiaries within 7.4 km (4.6 miles) of
the restoration site.
Although Tier II benefit indicators are expected to be less rigorous than the Tier III indicators,
they should be able to inform decisions without intensive modeling, based instead on existing
datasets. Because the Tier II results are based on the model outputs rather than simple
judgment, we expect that they will be applicable to watersheds similar in hydrological
characteristics to our case study watershed. While Tier II indicators are more easily applied, in
many cases they still require some analysis, often requiring knowledge of Geographic
Information Systems (GIS). We demonstrate how the Tier II indicators could be applied in other
locations.
What is most useful about the approach presented here is that it directly incorporates people
and the benefits they receive from ecosystem restoration. Further, it provides a framework that
can be used to compare potential wetland restoration scenarios based on these benefits
without the need for estimating dollar values. Using an approach to assessing non-dollar
benefits that is grounded in economic principles allows for more robust discussion of
alternatives through a disaggregated and transparent presentation of the various factors that
are likely to affect the level of benefits to people. This can inform many decision contexts
where a strict benefit-cost framework is either not appropriate or not necessary. This indicators
approach can also easily be extended to incorporate conceptions of value beyond the economic
definition of value, and can be transferred to other decision settings and benefit types. We
demonstrate how it can be used to complement an existing functional assessment approach,
and propose that it can also be used to inform benefit transfers to add more insight into
variations in restoration benefits across locations, and who might receive those benefits.
Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
* hi.-I'M I ! II M l 1 >i 11 i i« ! 1
This report describes a method for developing indicators of the benefits of flood regulation
services of freshwater wetlands and presents a companion case study. We demonstrate our
approach through an application to the Woonasquatucket River watershed in northern Rhode
Island. The critical role of wetlands in flood protection, the difficulty in modeling that role, the
high value that communities place on flood protection, and the fact that many ecosystem
services assessment tools do not address flooding led to our focus on evaluating the benefits of
flood regulation services provided by wetlands.
We have chosen to focus on non-monetary benefit indicators rather than dollar values for
several reasons. Because of the spatial variations in provision of wetland services and the
people who benefit, dollar values are highly context-dependent, making benefit transfer
difficult; our aim is to develop an approach that can be applied in different locations with
minimal effort and resources. For local decision makers trying to choose rapidly where to invest
resources, applying primary valuation studies will usually take too long, require too much
expertise, and in many cases exceed their needs. This often leads to local decisions based solely
on either supply side functional assessments or benefit transfer of somewhat generic wetland
values. Although both functional assessments and benefit transfers can be rigorous and provide
useful information, often these assessments rely on wetland area, with little consideration of
location-specific aspects and spatial flows of services and their benefits.
This work is intended to address an important component of the assessment of ecosystem
services: the development of metrics that clearly show how ecosystem functioning benefits
people. This approach contributes in three ways to the assessment of flood regulation services
provided by wetlands. First, it goes beyond standard ecological assessments of wetland
functioning by linking functioning with how, where, and how many people benefit from
wetlands. Second, it provides a means of estimating defensible metrics using a tiered approach
ranging from metrics that are more easily estimated but with greater uncertainty, to metrics
that require detailed modeling and provide reduced uncertainty. Third, it works as an add-on to
existing functional assessment tools, in order to extend their applicability to assessing
ecosystem services, or augments economic benefit transfer approaches by providing metrics
that can be used to determine the "extent of the market" for flood reduction benefits or to
adjust values to better reflect local conditions.1
Our approach is consistent with the recently released Best Practices for Integrating Ecosystem
Services into Federal Decision Making (Olander et al. 2015). A primary recommendation of that
guide is to "extend assessments beyond purely ecological measures that are not explicitly tied
to people's values to measures of ecosystem services that are directly relevant to people"
(Olander et al. 2015, p. 2). To address that need, we adopt a general conceptual framework
1 Benefit transfer is an economic valuation method that applies existing values or value functions, from studies
conducted at particular locations, to estimate values at a different location (Richardson et al. 2015).
Introduction
-------
(Figure 1-1) based on the ecosystem service cascade (Potschin and Haines-Young 2011; also see
Turner et al. 2000 for a similar model and discussion of integration across disciplines). Figure 1-
1 shows how supply and demand interact to produce a valued ecosystem service and includes
the types of assessments that are relevant to each part of the cascade. Our work focuses on the
assessment of ecosystem services and indicators of their benefits.
Supply
Demand
Biophysical
structure:
Wetland
characteristics -
size, depth,
location
V J
Ecological
Assessment
Function:
Retention or
slowing of water
Functional
Assessment
Service:
Flood regulation
that reduces risk
of flooding
Ecosystem
Service
Assessment
Benefit
Reduction in
damages to
property and
infrastructure
Benefit
Assessment
Value
Value of lost uses;
avoided costs of
repair or
replacement
Monetary
Valuation
Figure 1-1: The ecosystem service cascade (adapted from Potschin and Haines-Young 2011). The cascade
shows how supply of and demand for ecosystem services are related, and illustrates how the
ecosystem's structure and processes affect its functioning, and lead to the provision of ecosystem
services to people who benefit from and value those services. Different types of assessments focus on
different parts of the cascade. Our analysis focuses on ecosystem service assessment and benefit
assessment.
Despite the importance of the ecosystem services wetlands provide, the extent of wetlands in
North America has declined substantially since colonization, particularly in urban areas (DahI
and Allord 1996). Wetland restoration is one way to try to recover some of the benefits that
have been lost. However, with many potential restoration sites and limited funding for
restoration, managers need to prioritize sites that have the highest chances for success and the
highest potential benefits. To make use of available funds, these decisions often must be made
rapidly and opportunistically, and few metrics exist to easily compare the ecosystem services
and benefits to people from wetland restoration projects under such circumstances. Other
decision contexts important to wetlands include choices about whether to re-construct a
wetland that has been filled rather than restoring another type of ecosystem; how to prioritize
conservation efforts to protect existing wetlands; and whether to invest in wetland restoration
Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
or construction as an alternative to gray infrastructure for stormwater or waste water
management.
There are many existing methods for evaluating aspects of the ecosystem service cascade for
fresh water wetlands, ranging from relatively easily applied functional assessments (e.g., and
Golet 2001) to complex and data-intensive spatial models of the full ecosystem service cascade
such as the ARIES model (Villa et al. 2014). The majority of existing wetland assessment
methods focus on wetland functions, with cursory attention to benefits, typically in the form of
a judgment regarding the "social significance" of each function (see King et al. 2000 and King
and Price 2004 for a review and listing of many of these approaches). At the other end of the
ecosystem service cascade, many economic valuation studies address dollar values of wetlands,
some of which value specific services of wetlands (see Brander et al. 2006 and Ghermandi et al.
2010 for summaries). These studies are context and location-specific and therefore require
additional information to be useful for benefit transfer and to be able to distinguish among
sites. Our work is notable in its linking of functions to benefits and values for flood regulating
services of fresh water wetlands and providing an approach that may be rapidly applied, in light
of existing methods that largely address either functions or dollar values or require data-
intensive modeling to apply.
In contrast to some other ecosystem services, the service of flood regulation is particularly
dependent on the spatial and hydrological characteristics of the landscape because sites
providing flood regulation typically benefit people at a distance downstream. Also, the value of
the flood regulation service is highly dependent on the number of people who benefit and the
value and level of vulnerability of structures protected from flooding. Thus, our analysis focuses
strongly on estimating the spatial flow of the flood regulation service, defining the area where
people are likely to benefit, and identifying the assets protected.
The Woonasquatucket River Watershed
The Woonasquatucket watershed is a 132 km2 (82 miles2) basin in northern Rhode Island
(Figure 1-2). The basin contributes to the Woonasquatucket River, a river with a long history
of cultural and industrial development. The Woonasquatucket joins the Moshassuck River to
become the Providence River, which flows through Rhode Island's capital, Providence, and into
upper Narragansett Bay. As was typical in the early industrialization of New England, the
Woonasquatucket River was used to generate power and transport goods, and development
occurred immediately adjacent to and even over the Woonasquatucket, often destroying
wetlands and filling floodplains (Hardmeyer and Spencer 2007). Currently, urbanization in the
watershed follows a gradient of increasing urbanization from north to south, where the
northern portion is less urban than the city of Providence near the southern river outlet.
Growth projections suggest the watershed will continue to urbanize in years to come
(Rhode Island Statewide Planning Program 2006).
Introduction
-------
Kilometer
NORTH
SMITHFIELD
BURRILLVILLE
SMITHFIEl
PAWTUCKET
NORTH
PROVIDENCE
PROVIDENCE
JOHNSTON
CRANSTON
I Kilometers
• Gage Location
Contributing Area to Gage
Woonasquatucket Watershed
National Land Cover Dataset 2011
| | Open Water
| Developed, Open Space
~ Developed. Low Intensity
Developed, Medium Intensity
] Developed, High Intensity
| Barren Land (Rock/Sand/Clay)
| Deciduous Forest
~ Evergreen Forest
~ Mixed Forest
| Shrub/Scrub
| Grassland/Herbaceous
| Pasture/Hay
~ Cultivated Crops
Woody Wetlands
~ Emergent Herbaceous Wetlands
Figure 1-2. The Woonasquatucket River watershed within Rhode Island, USA showing land cover
and the location of the stream gage and its contributing area.
As in many urbanizing watersheds in the northeast (Collins 2009; Hodgkins 2010; Villarini and
Smith 2010; Smith et al. 2010; Hirsch and Ryberg 2012; Peterson et al. 2014), flood magnitudes
are increasing in the Woonasquatucket, This is driven both by increased runoff due to loss of
permeability associated with urbanization and increasing duration and intensity of precipitation
events. In recent years, flood frequency curves have shifted up (Figure 1-3). Eight of the ten
largest recorded flood events have occurred since 1970, and the largest recorded flood event
occurred in March of 2010 and was greater than a 200-year event (Corcoran 2007).
In addition to increasing flow magnitudes due to urbanization, rainfall events causing these
flows are expected to continue to increase. Long-term data suggest that annual average rainfall
has been increasing by 1" each decade for the last 80 years (Figure 1-4). Depending on the
conditions, increased rainfall does not necessarily correlate with increased stream flow.
However, the annual peak stream flows observed have shown a trend of about 46 ft3/sec (cfs)
more each decade (Appendix C-l).
Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
RECURRENCE INTERVAL
1.2 2 5 10 20 50 100
post-urban
pre-urban
Figure 1-3. Flood frequency curves for pre- and post-urban development (circa 1952- and 1972,
respectively) in the Woonasquatucket River (from Doehring and Smith 1978). The vertical scale is flow
normalized to the mean annual flood (CFS) and the horizontal scale is the recurrence interval in years.
1200®
Rhode Island, Precipitation, January-December
1930-2013 Trend 1930-2000 T p -
-H .02".'Decade Avg: 45.40" Krce,p
Figure 1-4. Average annual precipitation for Rhode Island for the period 1930 to 2013
(Valle and Giuliano 2014)
With more people arid infrastructure located near the river, future floods will likely result in
increased damages. Therefore there is a growing need to understand how wetlands and other
green infrastructure may alleviate flooding, and to help facilitate land use decision making
around these critical resources in order to maintain or increase flood protection to
urbanizing areas.
Introduction
-------
Indicator Development
Use of indicators, measurable metrics that represent more complex phenomena, is a viable way
to assess ecosystem services and their benefits while potentially decreasing the necessary time
and expertise required for implementation. In developing the indicators presented here, we
had two major objectives: (1) providing a way to extend existing functional assessments and
valuation methods, and (2) developing a defensible approach to the use of indicators, based on
sound theory and science. This requires that our indicators be sensitive to and able to connect
with measurable information in other parts of the ecosystem service cascade. For flood
regulation, this means that flood modeling must incorporate parameters of wetland function,
and must also have a strong spatial component that accounts for important aspects of the flow
of services from wetlands (production of a service) to people who benefit. And, to be
compatible with economic theory and methods of valuation, the models must address factors
that are important in determining differences in value across people and locations. To this end,
the set of indicators chosen fit into a benefits assessment framework structured around the
economic theory of value (Mazzotta and Wainger in preparation).
Using indicators in place of directly measuring the phenomena of interest may result in poor
decisions if the indicators do not provide valid measures of the phenomena of interest. As a
result, a major criticism of indicators approaches is that correlations between indicators and
the actual supply and demand of services are often assumed rather than demonstrated
quantitatively (Anderson et al. 2009; Duelli and Obrist 2003). Where correlations cannot be
demonstrated it becomes hard to discern whether indicators are rigorous enough for a given
decision context. Although empirical demonstration of indicator robustness exceeds the scope
of this report, our aim is that our indicators be defensible metrics for the types of decision
contexts often encountered when investing in wetland restoration.
The purpose of the indicators we present here is to provide metrics of the factors that influence
the spatial flow of services from production to benefits, by assessing how flood waters flow
across the geographic area between wetlands that retain or slow storm water flows and the
important structures and other resources that might be protected from flood risks. We based
indicators on sophisticated flood modeling results to ensure the rigor of results, but we
recognize that in some contexts a rapid assessment that can be readily applied to inform a
decision may be more valuable than a complex approach that is not feasible to apply. In an
attempt to satisfy the requirements of different decision contexts, we created a tiered
approach. Based on the appropriate level of rigor for the decision context, each tier becomes
more resource intensive to implement and requires more specific data. Certainty and
robustness increase with each indicator tier (from I to III; Figure 1-5).
To apply the Tier III approach, an end user would need to perform extensive modeling similar to
that described here, which is often too resource intensive and technical for most decision
makers and many decision contexts. We developed the Tier II indicators from trends and
sensitivity observed in our modeling results. Because they are based on the model outputs
Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
rather than simple judgment, we expect that they will be applicable to watersheds similar in
hydrological characteristics to our case study watershed. We intend to test this in future work.
We expect the Tier I indicators to be the least robust but most accessible and more broadly
applicable. We did not derive these indicators from model outputs; they are based on results in
the literature. We do not present Tier I indicators in this report, but they are included in our
guide for decision makers, which presents the overall benefit indicators approach and provides
a spreadsheet tool for applying Tier I indicators (Mazzotta et al. in preparation).
ro
ASSESSMENT TIERS
u
J TienJII
o
Qualitative/Uncertain
Uncertainty/Robustness
3
Figure 1-5. Required resources and results uncertainty across the three tiers of indicators.
Reading this Report
In Chapter 2, we describe the modeling process, assumptions, parameter values, model
validation and model sensitivity. Our intention is that users will be able to replicate this
modeling process to conduct a similar analysis in other watersheds. Results for both observed
and synthetic storms with known recurrence intervals are given. Chapter 3 shows how to fit
results from the modeling into the Tier III benefits indicators framework, and how these results,
along with the model sensitivity analysis from Chapter 2, are used to develop Tier II benefits
indicators. Chapter 3 also demonstrates how the indicators might be used, by presenting an
example implementation using Tier III indicators developed here to make a hypothetical
decision between two restoration sites. Chapter 4 provides a summary and ideas for future
research. The modeling details and other specifics, including data sources, are included in
appendices.
Introduction
-------
-------
CHAPTER 2: FLOOD MODELING
Introduction
This chapter describes the detailed modeling that we conducted in order to run scenarios using
simulated wetland restorations to evaluate their influence on downstream flooding and flow.
Ultimately, we developed indicators for flood regulation services and benefits from wetlands
based on the downstream influence of these simulated wetland restorations. Our modeling
approach includes several linked models, applied in sequence (Figure 2-1). First, we used
hydrologic models to estimate baseline hydrology in each basin and translate restoration
scenarios—or changes in wetland coverage—to changes in hydrology and resulting stream
hydrographs. Second, we incorporated these hydrographs into hydraulic models, which
produced flood depth profiles. Third, we applied the flood depth profiles, using flood impact
models, to map flood depth and extent and estimate the impacts of these floods on potential
beneficiaries.
Data
Inputs
Hydrographs
Other
Data
Inputs
Flood Maps
Other
Data
Inputs
Impact/Benefit
Estimates
Hydrologic Model
Hydraulic Model
Flood Impact Model
Wetland
Restoration
Scenarios
Figure 2-1. Modeling Approach
Chapter 2. Flood Modeling
-------
In order to explore scenarios using these models, we first modeled flow in the
Woonasquatucket watershed under current conditions, and then calibrated and validated the
models using observed flows at a USGS gage in the watershed. We applied these models to a
99.2 km2 (61.6 mi2) basin area of the watershed (Figure 1-2 area outlined in red). We did not
model the entire watershed because the area below the USGS gage used for model calibration
is highly urbanized and lacks information on stormwater infrastructure, which is likely to lead to
errors in predicting flood extent. The methods and assumptions for incorporating wetlands into
the models are critical to predicting how wetland restorations can potentially provide flood
regulation and the resulting potential to reduce flood risks to people and valued assets
(structures and infrastructure). To provide transparency regarding these critical topics we
emphasize them in Chapter 2 in our discussion of how we parameterized the hydrologic model.
To evaluate the impact of wetland restoration on flooding and to develop indicators of the
benefits of flood regulation, we modeled four wetland restoration scenarios. These scenarios
increased wetland surface area in each subbasin by converting 1%, 5%, 10%, or 25% of non-
wetland area to wetlands. We ran these wetland scenarios with three different synthetic
storms (24-hour precipitation events with recurrence intervals of 1-year, 5-year, and 25-year;
Appendix G). The resulting peak flows (approximately 10-year, 100-year, and 500-year
corresponding flow events) were each the result of a different synthetic storm. Based on the
changes in flow and flooding under these scenarios, we developed indicators of wetland flood
regulation services and benefits to people.
Our Tier III indicators are based on direct results of these watershed-specific models drawing
heavily on the sensitivity analysis to understand which variables most strongly influence results,
which is detailed in the Sensitivity Analysis section. We developed Tier II indicators by
generalizing the model outputs, through finding those factors that are the most critical
determinants of who will benefit and where. Defensible Tier II indicators depend on model
selection, assumptions, parameterization and validation.
An additional objective of our modeling approach was that both Tier II and Tier III indicators
be compatible with existing functional assessments. Such compatibility allows the indicators to
build on those assessments, adding information regarding benefits to people and how those
benefits differ across locations. To support the compatibility of our approach with a functional
assessment, we selected model inputs and developed benefit indicators to be consistent with
the Miller and Golet (2001) functional indicators assessment (Appendix A).
Mo(' ! ! :11. !fi !¦ I lh> i" i !, -\ 1 i! u!
Model Selection
We considered a number of existing flood models before selecting the models we deemed
most appropriate for our purposes. Available models differ in terms of the characteristics and
processes that they parameterize and their data requirements. In choosing models, we needed
to balance competing objectives: obtaining realistic results to produce the most defensible
10 Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
indicators and choosing an approach that other practitioners can reproduce in other
watersheds. Therefore, we opted to forgo the most sophisticated models with the most
specific and detailed data in favor of more flexible models.
In selecting models, we consulted previous studies where wetlands have been integrated
into modeling to estimate their downstream influence, and evaluated model advantages,
disadvantages, and standards for parameterization and optimization (Bengtson and
Padmanabhan 1999; Yuan and Qaiser 2011; Ogawa et al. 1986). In addition, some modeling
and analysis had been performed previously in the Woonasquatucket watershed, and we
selected models that are compatible with this previous work (Zarriello et al. 2013; Zarriello
et al. 2010; ACOE 2000).
The Hydrologic Model
The first criterion in selecting a hydrologic model was that the model reasonably account
for retention of storm water in wetlands. More specifically, the model needed to account for
current wetlands under baseline conditions, but also accommodate scenarios with added or
removed wetlands in different parts of the watershed. This requirement is not met by some
models that integrate wetlands. For example, wetlands are a land cover type used in the Curve
Number (CN) method (Williams and LaSeur 1976) that is used in many hydrologic models to
parameterize runoff characteristics of a surface. The CN method models wetland runoff loss
as a function of permeability, not accounting for surface water detention, and thus does not
meet our criterion. Additionally, we also selected a model that would be compatible with the
functional indicators used for flood retention by Miller and Golet (2001) (Appendix A), which is
discussed in greater detail in the Treatment of Wetlands section of this chapter.
The second criterion for the hydrologic model was the need to accommodate rainfall from
storms of various sizes, while differentiating between a baseline scenario and scenarios with
the same moisture and stream conditions but with a change in wetlands. The model needed to
reflect how a change in wetlands under the restoration scenario would lead to changes in
where runoff is stored (loss) and resulting changes in river flow. Hydrologic models with this
capability vary in complexity; for example, they may be spatially distributed, semi-distributed,
or lumped, so that input parameters are either characterized at a fine-scale resolution, within
sub-regions, or generalized to one number for the entire study area, respectively.
Based on these criteria, we selected the U.S. Army Corps of Engineers (US ACOE) Hydrologic
Engineering Center's Hydrologic Modeling System (HEC-HMS) for the hydrologic modeling
phase (Flemming 2013). HEC-HMS has the ability to model event-based (single-storm)
simulations based either on observed rainfall and stream flow or, once calibrated to
observations, simulations based on synthetic storms with known probabilities. Both observed
and synthetic rainfall data can be associated with a single gauged point or distributed across
the watershed (grid format). HEC-HMS is a semi-distributed model, meaning it aggregates
distributed data up to the subbasin. Although we might prefer a fully distributed model such as
the Hydrological Simulation Model-Fortran (HSPF) (Bicknell et al. 1997), we were able to
Chapter 2. Flood Modeling
-------
increase the distribution of the model by dividing the basin into many small subbasins. HEC-
HMS and its earlier version, HEC-1, have been used in previous hydrologic assessments of
wetlands on flood hydrographs (Bengston and Padmanabhan 1999; Qaiser et al. 2012).
Although HEC-HMS is not explicitly designed to parameterize flood water retention in wetlands,
these previous studies were able to account for retention in wetlands by treating them as
diversions.
The compatibility of HEC-HMS with other models was another critical factor. The Geospatial
Hydrologic Modeling extension (HEC-GeoHMS; Flemming and Doan 2013) ArcGIS toolkit
produced by the same US ACOE center allows for streamlined generation of model basin and
subbasin parameters from existing spatial data. The data management system (HEC-DSS) used
to store output hydrographs from HEC-HMS is used throughout HEC software, so HEC-HMS can
easily interface with HEC hydraulic software.
The Hydraulic Model
The second phase of the modeling approach uses hydraulic models for flow of water and how
water acts within the stream channel and floodplain, describing either how rapidly and with
how much loss water is conveyed or where it will overflow the channel and cause inundation.
Hydraulic models can use simple linear interpolation, can be one or two-dimensional, and can
simulate steady or unsteady flow, all with a range of detail and assumptions regarding the input
parameters. The main criteria for selection of a hydraulic model were compatibility with the
hydrologic model, Geographic Information Systems (GIS), and previous studies. Based on these
criteria, we selected the Hydrologic Engineering Center's River Analysis System (HEC-RAS;
Brunner 2010).
HEC-RAS performs one-dimensional steady or unsteady flow analysis. This analysis uses
geometric data that can be collected in an ArcGIS toolkit, HEC-GeoRAS (Ackerman 2009).
The geometric data account for elevations and roughness along stream profiles and cross
sections; any structures along the stream such as constricting bridges, culverts, levees, or dams;
and storage areas for use in unsteady flow analysis. Like HEC-HMS, HEC-RAS is also widely used.
Qaiser et al. (2012) used HEC-RAS effectively in their evaluation of wetlands for flood
mitigation. In the Woonasquatucket watershed, the U.S. Environmental Protection Agency
(EPA) used HEC-RAS for steady flow analysis to evaluate downstream flood impacts of removing
the dam at Centredale (US ACOE 2000). As part of FEMA's process of updating Flood Insurance
Rate Maps (FIRMs), U.S. Geological Survey (USGS) was contracted to create and validate a HEC-
RAS model for the entire Woonasquatucket River (Zarriello et al. 2013). We integrated the
geometric data used for this USGS study into our HEC-RAS model.
Flood Impact Modeling
Flood risk has two components: hazard and vulnerability. Flood hazard measures the exposure
or severity of the storm independent of its impact on people. To model flood hazard, we
overlaid flood depth profiles from the HEC-RAS model onto elevations in the watershed
to determine where flooding occurred for each scenario, and how deep that flooding was.
12 Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
Flood vulnerability measures who is impacted and the severity of that impact. For our
assessment we were interested in the impacts to the built environment and severity of those
impacts. We investigated several types of data to account for where vulnerable structures exist
in our projected floodplain, and found address information to be most useful at the resolution
of our analysis (Bousquin et al. 2014). We measured the severity of impacts on these addresses
in terms of areal extent and depth of flooding based on the flood maps produced from
simulations.
Treatment of Wetlands in the Models
Wetlands perform several hydrologic functions that support flood regulation. They can
desynchronize storm flows by providing short-term surface water storage, which can reduce
downstream flood peaks (Hubbard and Linder 1986; Hey and Philippi 1995) and they can
provide long-term storage and increase infiltration, which recharges groundwater and
maintains river and stream base flows (Winter 1999). The amount of flood regulation provided
varies with wetland type and landscape position (Brinson 1993). Therefore the provision of
flood regulation by restored wetlands is a function of both the type of restoration and the
location of the site such that it can receive flood water and is upstream of otherwise flood-
prone areas (Miller and Golet 2001). Below we discuss how wetlands are handled in the
hydrologic model and how we related this to the Miller and Golet (2001) wetland flood
protection functional assessment.
Wetland short-term surface water retention is represented in HEC-HMS as diversions—or
subbasin flow that is diverted from entering the main channel. The amount diverted from a
given subbasin in the model is a factor of (1) the rate at which subbasin outflow is diverted, and
(2) the maximum cumulative volume of water that can be diverted. Parameterizing these
diversion factors based on wetlands in a given subbasin required assumptions about the
function of wetlands.
Estimating Flow Diverted by Wetlands
To apply HEC-HMS, we divided the modeled basin (99.2 km2) into subbasins (n=139, mean
area=0.7 km2). The model cannot account for the location of a wetland within a subbasin.
Effectively, this means that the model calculates the water collected in a subbasin and then
diverts some portion of water based on all of the wetlands inside the subbasin before passing
the flow downstream (Figure 2-2A). In reality, each wetland captures some of the water from
its individual catchment area within the subbasin (Figure 2-2B). Aggregating wetlands and the
diversion of water to wetlands in this way could easily overestimate the percent of water
available for diversion to wetlands since all water from the subbasin, not just the wetland's
catchment, is subject to diversion. To alleviate this issue, we assumed that 100% of water
entering the wetland would be retained but that the catchment area of each wetland equaled
the wetland's areal extent, effectively assuming all wetlands were on the perimeter of their
subbasin where there was no conveyance of water from surrounding areas (Figure 2-2C). The
Chapter 2. Flood Modeling
-------
result, when wetlands' retention (Figure 2-2C) is considered on the subbasin scale (Figure 2-2A),
is that a subbasin's percent diverted equals the percent of the subbasin area in wetlands.
Figure 2-2- Figure showing wetlands in subbasin based on different assumptions: (a) how HMS
implements diversions, (b) bio-physical reality, and (c) the wetland contributing area assumption
we used.
This approach is more conservative than that of past studies. Wetlands in the watershed used
in Bengston and Padmanabhan (1999) occupied a surface area ranging from 0.58 to 5.14% of
the total watershed area, based on areas in the National Wetlands Inventory and their
estimated drained wetlands available for restoration. Divergence rates assumed for this
wetlands area were 25% to 50% (Bengston and Padmanabhan 1999). In Yuan and Qaiser (2011),
wetlands covered a maximum of 10% of the watershed area, and 25% of subbasin peak runoff
was set as the diversion rate.
Estimating Maximum Outflow Diverted
We used a Python toolbox in ArcMap 10.2 to estimate potential maximum volume retention,
following Lane and D'Amico (2010), where we used surface elevations from the Digital
Elevation Model (DEM; RIG IS 2013) as bottom contour and the minimum or average elevation
from perimeter vertices as height. The DEM was developed using LiDAR, a remote sensing
technology that uses pulses of light to detect variable distances to the ground to measure
ground elevations during a flyover, and was provided at a 1 m resolution with a vertical
precision of 0.01 m. We based the perimeter of wetlands on the same spatial dataset used to
define wetlands in the Miller and Golet (2001) functional assessment (RIGIS 1993). Details of
this process, including treatment of wetland types, adjacent wetlands and residual water
storage in wetlands before a precipitation event, are provided in Appendix B.
14 Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
Water diverted by wetlands is not released back into the modeled flow for a given storm event.
This is not an issue for our analysis since we are interested in the peak flows of water, which
cause the extent of flooding, rather than the total volume of water throughout a storm. We
assume that water causing peak flows occurs early in the storm, before wetlands are fully
saturated.
Wetland Flood Protection Functional Assessment
As mentioned above, we related wetland function in the model to the Miller and Golet (2001)
approach. The Miller and Golet (2001) functional assessment indicators are categorized as
opportunity, effectiveness, or social significance. In this section, we discuss how we
incorporated these criteria into our modeling effort.
Opportunity
Opportunity for a wetland to perform flood regulation functions relates to how much water
flows into or through the wetland. In the model, this is captured by the amount of water within
the subbasin available for divergence to wetlands. Some of the functional assessment
indicators for opportunity are landscape characteristics such as nearby impervious surfaces and
steep slopes, which are accounted for in a lumped fashion as parameters for the subbasin in the
hydrologic model. Although these characteristics were accounted for and increased water
available to be diverted by wetlands, more influential nearby characteristics were aggregated
with subbasin characteristics.
The model does not account for point-source inflow or flow into wetlands from streams. Most
point-source inflows in the Woonasquatucket are likely stormwater infrastructure, and basin-
wide spatial data for such infrastructure were not available. Excluding stormwater structures
and flows will influence model results more in urbanized areas. Calibrations of the hydrologic
model should correct for the role of stormwater infrastructure; however downstream of the
USGS gage is heavily urban and is uncalibrated, so we did not model that portion of the basin.
Effectiveness
The effectiveness of wetlands to perform flood regulation functions is related to how much
water can be retained in those wetlands. The Miller and Golet (2001) assessment tool includes
three factors to measure effectiveness: whether the wetland is a basin wetland, whether the
outlet is constricted, and whether the dominant vegetation is dense and persistent. Our model
accounts for the wetland type (basin or not) and whether the outlet is constricted in the
method used to calculate maximum volume diverted. If a wetland is not a basin, the maximum
volume will be zero. Our model does not account for the type of vegetation, because all water
diverted in wetlands was effectively removed from the system, making slowing by vegetation
irrelevant.
Social Significance
The functional assessment includes an indicator of social significance: whether there are
developed flood-prone areas within 5 miles (8 km) downstream of the site or to the nearest
dam, with a connection by stream or floodway. One of the main goals of our model is to go
Chapter 2. Flood Modeling
-------
beyond this simplistic type of social significance measure in order to better measure how much
people benefit from flood regulation services of wetlands. To this end, we directly modeled
flood extents and impacts. This is discussed in more detail in Chapter 3.
Hydrologic Modeling with HEC-HMS
HEC-HMS is a physically-based, semi-distributed model that can be used to simulate rainfall-
runoff for a single event (Bedient et al. 2007). We derived variables for the HEC-HMS model
parameters in several ways (Table 2-1). Many of the parameter variables were generated from
spatial datasets (Appendix D) using the HEC-GeoHMS toolbox (Flemming and Doan 2013) in
ArcGIS 10.1 (ESRI 2012). Other model parameters, including initial abstraction, were initially set
to standard defaults (e.g., 0.2 initial abstraction ratio from Soil Conservation Service 1985).
Once the model was fully parameterized, we calibrated it to observed precipitation and
observed stream flow. After calibration, we validated the HEC-HMS model against other
observed storms. After validation, we conducted sensitivity analysis on the calibrated model to
determine sensitivity of the results to changes in selected parameters and assumptions. Once
the hydrologic model validation and sensitivity analysis were complete, we assembled synthetic
storms of known recurrence intervals to use for evaluating wetland restoration scenarios with
the model.
Table 2-1. HEC-HMS parameters
Parameter-Variable
Element Type
Processing Method
Appendix
Routing -X
Reach parameter
Variable default, calibrated
Table D-3
Routing - K
Reach parameter
Calculated from basin lag, calibrated
Table D-3
Routing -SubReaches
Reach parameter
Calculated
Equation D-3
Loss - CN
Basin parameter
GeoHMS, Calibrated
Table D-2
Loss - % Impervious
Basin parameter
GeoHMS
Table L-l
Loss - Initial Abstraction
Basin parameter
Calculated using abstraction ratio
Equation
Transform - Lag Time
Basin parameter
GeoHMS, TR-55
not provided
Diversion Rate
Diversion parameter
Calculated, Inflow-Diversion paired data
Table B-l
Max Volume
Diversion parameter
Calculated, Python Toolbox
Table B-l
Storage-Discharge
Reservoir parameter
Storage-Discharge paired data
E
Initial Discharge
Reservoir parameter
0.5 m3/s from Appendix E
E
Precipitation (in/hr)
Time series data
NOAA
C
Stream Flow (m3/s)
Time series data
USGS gage
C
Parameterization
The HEC-HMS model allows for the use of several loss, routing, and transform methods. Of the
available methods, we chose methods that could be parameterized using available geospatial
data. Initial parameterization of the HEC-HMS model used here matches that used in Bousquin
et al. (2014). Here we briefly summarize the methods used and draw attention to any revisions.
A list of geospatial data used, a summary of parameterization, and values for those parameters
are provided in Appendix D.
16 Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
The watershed—or basin in HEC-HMS—includes the upper Woonasquatucket watershed with
the USGS gage at its outflow. We delineated the basin into 139 subbasins using a subbasin
contributing area threshold of 0.5 km2 and adjustment for reservoirs. This threshold is less than
the default, which is 1% of total watershed area or 1 km2 in our basin, and therefore results in
smaller subbasins. This produces a more distributed model and therefore comes closer to being
able to differentiate single-site restorations. Because our models are differentiated at the
subbasin level, to compare two sites within a single subbasin, a decision maker could evaluate
sites based on their size and specific spatial characteristics such as slope. This is discussed
further in Chapter 3. The HEC-HMS basin is divided into elements for subbasins, stream
reaches, junctions, and other special features, including reservoirs and the subbasin diversions
we used to represent wetlands (Appendix D).
Subbasin Elements
We used the loss method in HEC-HMS to generate runoff volume based on the Soil
Conservation Service (SCS) CN approach (Williams and LaSeur 1976) and the percent impervious
surface in each subbasin. The transform method, the rate runoff moves through subbasins to a
junction with a downstream stream reach, used the SCS Unit Hydrograph with the lag time
calculated using the TR-55 method (McCuen 1982).
Reach Elements
Stream reaches carry water from one junction to the next. We used the Muskingum routing
method (McCarthy 1938) to parameterize the rate of this conveyance.
Reservoir Elements
We used Dams with a maximum storage of 12 acre-feet or more to identify reservoirs
(Appendix D). In the model, reservoirs store and release water based on a storage-discharge
curve and have an initial storage and discharge at the start of an observed storm event. The
storage-discharge curves were based on a linear relationship between normal and maximum
storage in Bousquin et al. (2014). We have since updated the tables used for the storage-
discharge curves to add data from fitted exponential equations for each reservoir (see example
in Appendix E). Modeled storm peaks improved slightly using these updated equations, despite
a slight decrease in the overall fit of the model (Appendix E). As in Bousquin et al. (2014), if the
maximum storage was exceeded, the discharge increased to approximately infinity to simulate
unrestricted streamflow (see example in Appendix Figure E-3).
Reservoir discharges likely help sustain observed baseflow in the watershed. Without direct
measurements of water storage or discharge before a storm event there is no way to know the
actual initial reservoir discharge. Bousquin et al. (2014) used a default initial discharge of inflow
= outflow, but we updated this to 0.5 m3/s for the model presented here. This better represents
reality and helps "pre-wet" the system, putting water into the hydrologic system before
observed storm events (see Appendix E for comparisons).
Chapter 2. Flood Modeling
-------
Diversions
Diversions divert water from subbasins at a rate based on curves from inflow-diversion tables
and the diversion's maximum volume. We used percent diverted, described in the Treatment of
Wetlands section, to parameterize the inflow-diversion tables, where the diversion rate for
each subbasin was a percentage of inflow proportionate to the percent of the subbasin that is
wetland (Appendix B). Once a diversion reaches its maximum—the maximum outflow diverted,
described in the Treatment of Wetlands section—it stops diverting water. No initial storage of
water in wetlands was assumed beyond that accounted for in the volume calculation
assumptions.
Base flow
Baseflow observed in the system is thought to predominately stem from reservoirs, but the
model requires parameterized baseflow to be distributed throughout the subbasins. Therefore
we computed and accounted for total baseflow outside the model. To remove baseflow from
the model, we assumed the minimum flow from the modeled time period to be baseflow and
removed for the entire time period, following the constant discharge method (Hall 1968).
Baseflow removed from each of the calibration and validation storms is provided in Appendix F.
Precipitation and Runoff Data
HEC-HMS uses rainfall event precipitation as an input and predicts runoff, in the form of a
hydrograph. The model does this by parameterizing basin characteristics in terms of the
amount of water, where it travels, and how much time it takes. The model is refined through
calibration to observed rainfall and runoff.
Our observed time series data for both precipitation and runoff came from stationary gages.
We used hourly precipitation data from the Providence Airport Station (COOP: 376698; NOAA
2014), and stream flow data from the USGS gage station at Centerdale (USGS# 01114500; USGS
2014). The precipitation station is outside the modeled watershed, 15.9 km (9.9 mi) south of
the USGS gage station.
Storm Selection for Calibration and Validation
We calibrated the model to observed storms with the largest flood magnitudes to ensure that
our models would capture flooding impacts. We considered this a reasonable choice, as these
storms are becoming more common. Since some of the increase in flooding in recent years may
be caused by changes in land cover, we chose modeled storms that occurred between 2000 and
2010 for calibration, since spatial data used to set initial parameter values were gathered
during this time frame. The ten largest stream flow peaks within this time period were
identified for initial consideration (Table 2-2).
The storm of record occurred in March of 2010; however, much of the data around the peak
of the event were not available. We chose to calibrate to the second largest event (October 15,
2005; Figure 2-3) and the third largest event (March 22, 2001; Figure 2-4). We selected these
storms to calibrate to both spring conditions with very wet antecedent conditions, and fall
conditions where there was more time between events. We used the fourth and sixth largest
18 Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
storms, both spring storms, for validation. We could not use the fifth largest storm because it
dropped below freezing temperatures during the duration of the storm and our model was not
set up to handle snow. Details on the four storm event start and stop times, durations, and
baseflow precipitation and hydrographs are provided in Appendix F. Probabilities and estimated
recurrence intervals from analysis in PeakFQ (Flynn et al. 2006) for the storm event discharges
are provided in Appendix C.
Table 2-2. Top 10 Streamflow peaks for Woonasquatucket River gage for
2000-2010 in order of peak flow
Rank
Date
Height (m)
Flow (m3/s)
Use
1
Mar. 30,
2010
2.80
51.25
Not used: missing data
2
Oct. 15,
2005
2.52
43.32
Fall calibration storm
3
Mar. 22,
2001
2.00
30.30
Spring calibration storm
4
Apr. 03,
2005
1.86
26.70
Validation storm 1
5
Dec. 12,
2008
1.85
26.56
Not used: below freezing
6
Apr. 16,
2007
1.74
24.10
Validation storm 2
7
Feb. 13,
2008
1.67
22.46
Not used
8
Dec. 13,
2010
1.59
20.70
Not used
9
Apr. 14,
2004
1.55
19.91
Not used
10
Apr. 22,
2000
1.28
14.07
Not used
Chapter 2. Flood Modeling
-------
LO
LO
LO
LO
LO
LO
LO
LO
LO
LO
LO
o
o
o
o
o
o
o
o
o
o
o
4-J
4-J
4-J
4-J
4-J
4-J
4-J
4-J
4-J
4-J
4-J
o
o
O
O
U
u
u
u
o
U
u
O
o
o
o
o
o
o
o
o
o
O
fN
fN
on
on
LO
LO
*—1
*h
*h
*h
*h
*—i
Figure 2-3. Hydrograph of 15 min interval flow (USGS Station 01114500; black line in m3/s (CMS) on
left axis) and hourly precipitation (NOAA Coop 376698; gray line in cm on right axis) for the "Fall '05"
calibration storm (October 15, 2005). Raw flows are shown without baseflow removed.
i
i
i
i
i
i
T—1
T—1
t—i
t—1
t—1
t—i
t—1
t—i
t—i
t—i
t—1
t—1
t—i
t—i
o
O
O
o
o
o
o
o
O
O
O
O
O
o
o
o
O
O
o
o
i_
i_
i_
i_
i_
i_
i_
i_
i_
ro
ro
ro
ro
ro
ro
ro
ro
ro
ro
ro
ro
ro
ro
ro
ro
ro
ro
ro
Q.
S
S
S
<
i
ro
^|-
i_n
ub
00
CTt
o
t—i
ro
^|-
ln
ub
r^>
00
CTt
o
t—i
r—1
i
i
i
i
i
T—1
T—1
r\j
f\J
f\J
f\J
f\|
f\J
f\J
r\j
f\J
f\J
m
m
O
Figure 2-4. Hydrograph of 15 min interval flow (USGS Station 01114500; black line in m3/s (CMS) on left
axis) and hourly precipitation (NOAA Coop 376698; gray line in cm on right axis) for the "Spring '01"
calibration storm (March 22, 2001). Raw flows are shown without baseflow removed.
Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
ITERATION
ITERATION
USGS_GAGE OPT:OPT7_FULL 0
Figure 2-5. Two sets of optimizations were performed in sequence for the Spring '01 storm. The first (a)
had 200 iterations the second (b) had 320. The objective function is defined by how close the model
matches the observed peak. The area circled in red had the optimal parameters used by the model
(Calibration 2) resulting in lowest objective function.
Calibration
We began by machine calibrating to the Fall '05 storm (Oct '05; 1530 cfs) and Spring '01 storm
(Mar '01; 1070 cfs). Precipitation occurred both before and after the Spring '01 storm. The
model run included some minor preceding precipitation to pre-wet the model, but ended
before precipitation for a second peak to ensure calibration and metrics for model fit were for
the correct peak flow.
The pre-calibrated model peak flows were overestimated for the Fall '05 storm and
underestimated for the Spring '01 storm (Figures 2-6 and 2-7). Machine calibrations varied
three parameters: Basin Loss CN, Routing Muskingum X, and Routing Muskingum K. The
machine calibration procedure followed for the Fall '05 storm is detailed in Bousquin et al.
(2014). We ran the calibration for the Spring '01 storm as two consecutive optimizations. The
first Spring '01 optimization included 200 iterations (Figure 2-5a); the second included another
320 iterations (Figure 2-5b), using the results of the first optimization. The later iterations in the
second optimization were suboptimal, so we used the best calibration from earlier iterations
(Figure 2-5b iterations circled in red). Appendix D presents all pre- and post-calibration
parameter values.
In these initial calibrations, Basin Loss CN calibrations diverged for the two models, but both
routing parameters were calibrated in the same direction (Appendix D). This is understandable
considering the differences between the types of storms that take place in the fall compared to
the spring, in terms of antecedent moisture conditions, intensity, and other factors. The
Chapter 2. Flood Modeling 21
-------
conditions for the Spring '01 storm were generally wetter, with larger preceding baseflow
(2.5 m3/s compared to 0.88 m3/s) and more rainfall observed in the time period leading up to
the storm. To address this, we developed two models, with the same optimized routing
parameters but different loss parameters, depending on whether the storm occurred during
wet or dry conditions.
Subbasin loss is expected to be most influenced by wet and dry conditions. The HEC-HMS
model uses a default initial abstraction ratio (which is the ratio of the amount of water before
runoff to the potential maximum soil moisture retention after runoff begins) of 0.2 in loss
calculations, based on the National Engineering Handbook (SCS 1985, Figure 10.2). Other
studies have found that most basins have a lower abstraction ratio, closer to 0.05 (Woodward
et al. 2003). We maintained the 0.2 ratio for the dry condition storm, but reduced the initial
abstraction ratio to 0.02 for the wet condition storm to emphasize the wet conditions. CN can
also be adjusted to better represent antecedent soil moisture conditions. Antecedent Soil
Moisture Condition II (AMC II) definitions of CN used for pre-calibration were converted to AMC
I for dry conditions (Fall '05) and to AMC III for wet conditions (Spring '01) following Ward and
Trimble (2003) and NRCS (1984). This method for calibrating CN was preferable since it
weighted increases and decreases in CN based on original CN, where the machine calibration
altered all CNs uniformly. Without further calibration, these changes to initial abstraction and
CN improved performance of both models (Table D-4; Appendix D).
Although both machine calibrations increased routing parameters, increases were not the same
for both calibrations. Having altered the loss calculations for both models after calibration may
have also altered the optimal routing parameters. To address this, we calculated a blended
routing parameter value for each reach based on the average increase for all reaches (Appendix
D-3). The three routing parameter values (Fall '05 calibrated, Spring '01 calibrated, and blended
calibration) showed the least peak error across both storms using the wet, Spring '01 calibrated
values.
Validation
The two calibrated models, represented by blue lines in Figures 2-6 and 2-7, outperformed the
uncalibrated models, represented by red lines, and showed peak flows very close to those
observed (Table 2-3). Modeled volume of both calibrated models was less than the observed
volume. Water diverted to wetlands may account for some of this missing volume.
22 Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
Uncalibrated Flow
Observed Flow
Updated Calibrated Flow
Old Calibrated Flow
12:00
130ct2005
12:00
140ct2005
12:00
150ct2005
12:00
160ct2005
Figure 2-6. Fall '05 storm (Oct '05) hydrographs for Observed Flow (Black), Calibrated Flow (Blue),
Calibrated Flow from Bousquin et al. (2014) (Light Blue), and Un-calibrated Flow (Red).
Observed Flow
— Calibrated Flow
Un-calibrated Flow
30-
25"
> 15'
Mar2001
Figure 2-7. Spring '01 storm (Mar '01) hydrographs for Observed Flow (Black), Calibrated Flow (Blue),
and Un-calibrated Flow (Red).
Chapter 2. Flood Modeling 23
-------
Table 2-3. Validation Results for Spring '01 and Fall '05 models validated and fit across four
observed storms
Observed Storms
Mar '01 Apr '05 Oct '05 Apr '07
(Spring Calibration) (Validation 1) (Fall Calibration) (Validation 2)
Peak
SD
Peak
SD
Peak
SD
Peak
SD
Observed (m3/s)
34.3
7.6
26.1
6.2
42.4
13.5
20.0
5.4
Peak
RMSE
N-S
Peak
RMSE
N-S
Peak
RMSE
N-S
Peak
RMSE N-S
Fall model ('05)
5.0
9.4
-0.54
4.9
8.6
-0.94
45.2
4.2
0.91
3.9
7.2 -0.69
Spring model ('01)
31.5
5.0
0.56
14.0
4.2
0.52
168.7
42.3
-8.77
30.9
5.5 -0.01
We evaluated model fit using Nash-Sutcliffe (N-S) efficiency and Root Mean Square Error
(RMSE) statistics (Legates and McCabe 1999; Table 2-3). The N-S efficiency (Nash and Sutcliffe
1970) measures the relative difference between the model residuals (difference between
modeled values and observed values) and the data's variance (difference between observed
values and the observed mean). The index can range from -« to 1, with 1 representing a
perfect fit between the model and observation and 0 indicating that the model is as good as
using the mean of the observed data. N-S values between 0 and 1 are generally considered
acceptable values; however for stream flow model evaluation, an N-S efficiency greater than
0.5 is considered satisfactory (Moriasi et al. 2007). We consider the 0.5 N-S standard for
satisfactory model fit to be overly restrictive for our model application, because N-S is based on
all discharges and in our model the volume of water diverted to wetlands was removed from
the system rather than rereleased, artificially reducing discharges after the peak flow.
Therefore, we judged an N-S value greater than 0 as acceptable for our evaluation. The Spring
'01 calibration met the > 0.5 criterion for the April 2005 storm, and came very close to the > 0
criterion for the April 2007 storm. The Fall '05 calibration did not meet the > 0.5 criterion for
either of the validation storms (which were both spring storms). This reinforces the importance
of antecedent conditions and using a wetter spring-calibrated model for spring storms.
RMSE is the un-standardized sum of differences between observations and the model output;
therefore this index is in the units of the modeled parameter (in this case flow, or m3/s). A zero
RMSE indicates a perfect fit between the modeled and observed data, with lower RMSE
indicating a better model fit. An RMSE less than 70% of the standard deviation is considered
satisfactory (Moriasi et al. 2007). The RMSE results were similar to those of the N-S efficiency
tests. The Spring '01 model RMSE was < 70% of the standard deviation observed during the
March 2001 calibration storm (7.6 m3/s; RMSE < 5.3) and the April 2005 validation storm (6.2
m3/s; RMSE < 4.33). The Spring '01 model RMSE was > 70% of the standard deviation observed
for the April 2007 storm (13.5 m3/s; RMSE < 9.5) and the October 2005 storm (5.4 m3/s; RMSE <
3.8). While the Fall '05 model RMSE met the RMSE validation criterion for the storm used to
calibrate it, it did not meet the criterion for any of the validation storms.
Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
Sensitivity Analysis
We conducted sensitivity analysis using the Spring '01 calibrated model. We excluded
parameters from the sensitivity analysis based on 3 criteria: those that were used during
calibration (routing Muskingum X and K, basin loss CN, and initial abstraction ratio), those that
were calculated based on those calibrated parameters (Muskingum subreaches), and those that
were used to parameterize other HEC-HMS parameters externally from the model (percent
imperviousness and slope). The following sections describe each of the sensitivity analyses.
Lag Method
The basin lag method can be calculated using either CN or the TR-55 method. We chose the TR-
55 method because it integrates a greater number of basin characteristics. Comparing results
from the two methods suggested that the model was not sensitive to the choice of loss method
(Figure 2-8).
Precipitation
We examined five scenarios by increasing or decreasing all rainfall during the modeled duration
by an equal factor (Figure 2-9). As expected, the model was very sensitive to changes in
precipitation (Figure 2-10). The results showed a shift in hydrograph shape and peak flow when
precipitation was increased by 50% or more. Small increases or decreases in precipitation (+/-
10%) resulted in proportional changes in peak flow (+/-10%). However, more extreme
increases or decreases in precipitation (+/- 50%) resulted in disproportionately larger changes
in peak flow (+100% and -30%).
TR-S5 Lag
Curve Number Lag
Figure 2-8. Results of altering lag method in comparison to observed flow (Blue). The TR-55 method
(Green) was used in the model, where the Curve Number method (Red) was not.
Chapter 2. Flood Modeling
-------
0.8-
0.7-
0.6-
•c 0.5-
-C
c
c" 0.4-
o
"ro
0.3-
a>
IX
0.2-
0.1 H
+
li
Precipitation Gage
Precipitation Gage
Precipitation Gage
¦ Precipitation Gage
Precipitation Gage
' Precipitation Gage
x 2.0
x 1.5
x 1.1
Observed
x 1.5
x 0.5
K
13 '14'15
1 16
Mar2001
18 '19 '20 ^ 21
'25* 26
Figure 2-9. Precipitation for the different sensitivity analysis runs.
100
90
E 50
I
^ 40
30
20
Flow resultingfrom Precipitation x 0.5
Flow resultingfrom Precipitation x 0.9
— Flow resultingfrom Precipitation Actual
— Flow resultingfrom Precipitation x 1.1
— Flow resultingfrom Precipitation x 1.5
Flow resultingfrom Precipitation x 2.0
Observed Flow
14 ' 15 ' 16 ' 17 ' 18~
Mar2001
^9 ' 20 ' 21 ' 22 ' 23 ' 24 ' 25 ' 26 ' 27 ^ 28 1 29
Figure 2-10. Sensitivity analysis hydrographs based on varied precipitation.
Wetland Percent Diverted
Important assumptions and uncertainties surrounding parameters dictating the percent
of subbasin outflow diverted to wetlands have already been detailed in the Treatment of
Wetlands section of this chapter. Results of the modeling effort rely on those assumptions,
so we performed sensitivity analysis on those parameters to test our assumptions. Using the
most conservative assumption—where percent diverted is equal to the percent of the subbasin
surface area in wetlands—as a baseline, we increased the area contributing to wetlands in
the remaining catchment area by 10%, 25%, 50%, and 75%. For example, for the 50% run,
represented as a black line (Figure 2-11), half of the non-wetland area was assumed to
contribute flow into the wetlands. Although this may seem like an extreme increase in
contributing area, a conservative 100-m buffer around wetlands adds an area comparable
Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
to the 50% run used in sensitivity analysis (Appendix B). The results showed a strong relation
between peak flow and assumptions about the subbasin area that contributes flow to wetlands,
which we used in making generalizations to develop Tier II indicators (as described in
Chapter 3).
Wetland Max Volume Method
During initial parameterization of wetlands, we calculated wetland volume following Lane and
D'Amico (2010). This method derives volume from the area of a wetland times its height—the
average elevation of wetland perimeter vertices. A more conservative estimate of height was
based on the minimum elevation of wetland perimeter vertices. We tested model sensitivity
to this assumption, and found a small difference (Figure 2-12). During this storm only one
subbasin divergence (D_W2980) reached its average-based maximum volume. In a scenario
where more of the subbasins reach maximum volume, such as when there is a simultaneous
increase in percent diverted to wetlands (Figure 2-11) or precipitation is increased (Figure
2-10), we suspect the model sensitivity to wetland maximum volume would increase.
— 0 Additional Wetland Catchment
— 10% Additional Wetland Catchment
— 25% Additional Wetland Catchment
— 50% Additional Wetland Catchment
— 75% Additional Wetland Catchment
25-
20-
E 15-
10-
Mar2001
Figure 2-11. Resulting hydrographs from sensitivity analysis model runs with different wetland percent
diverted assumptions ranging from the model assumption (Blue) to a 75% increase in wetland
catchment area contributing to diversion (Pink).
Chapter 2. Flood Modeling
-------
Wetland Volume (Average Method)
Wetland Volume (Minimum Method)
Observed Flow
40"
35"
25-
§ 20-
15"
10-
Figure 2-12. Sensitivity analysis results hydrographs with different wetland max volume assumptions:
water retention height derived from average (blue) or minimum (red) elevation of wetland perimeter
vertices.
Hydraulic Modeling with HEC-RAS
We estimated the extent of flooding using the hydraulic model Hydrologic Engineering Center
River Analysis System (HEC-RAS) of the Army Corps of Engineers (Brunner 2010). HEC-RAS
generates water surface elevations at profiles along a channel for steady, gradually varying flow
using a standard step method for finding solutions to the 1-dimensional energy equation
(Bedient et al. 2007).
Surface profiles for HEC-RAS use cross sections along the river reaches and are part of the HEC-
RAS geometry dataset. HEC-RAS geometry includes river reaches (the synthetic streams
generated from HEC-HMS, not including first order stream reaches), cross sections, and other
constructed elements along the river such as bridges, culverts, levees, and dams. We used the
geometry dataset from Zarriello et al. (2013), supplemented by data generated by Bousquin et
al. (2014). The added HEC-RAS geometry was characterized with elevations, from the DEM, and
roughness, from landuse data (Appendix D). We developed an R script (Appendix J) to extract
peak flows from HEC-HMS to use as inputs into HEC-RAS to perform steady flow analysis.
USGS validated their hydraulic model against observed storm elevations (Zarriello et al. 2013).
The additional geometric features from Bousquin et al. (2014) were not as intensively validated,
but were compared to Digital Flood Insurance Rate Maps (DFIRMs) based on flood extent and
input from the Town Engineer of Smithfield, Rl.
The final modeling step was flood impact modeling, where we exported flood depth profiles
from HEC-RAS into ArcGIS and compared to other data to estimate the impact of the flood.
28 Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
¥ Scenarios
We developed 12 scenarios to investigate the influence of different levels of wetland
restoration, across a range of storm events, on stream flow and resulting flood extent. We used
the resulting peak flows and flood maps to demonstrate the development of Tier III benefit
indicators to compare wetland restoration scenarios (see Chapter 3 Tier III Assessment Section).
We also relied on the wetland restoration scenario results to develop Tier II benefit indicators,
generalizing the results to develop indicators of differences in potential restoration sites that
correlate with differences in flood reduction services and subsequent benefits from flood
damage reduction.
Size of Restoration in Scenarios
To simulate various wetland restoration scenarios we needed to vary the volume and/or the
area of wetlands by subbasin. The sensitivity analysis suggested the model was not sensitive
to changes in wetland volume because most subbasins did not exceed their maximum
volume-when we ran the model for the Spring '01 storm (wet conditions), only HEC-HMS
diversions in 2 subbasins reached their maximum volume. Also, altering the wetland volume
calculation method showed only a minor response in flow (Figure 2-12). However, sensitivity
analysis of wetland percent divergence, which is equivalent in the model to increasing wetland
area in a subbasin, showed that model results were very sensitive to this parameter, where
a 25% increase in wetland divergence resulted in greater than 25% decrease in peak flow
(Figure 2-11). Based on these findings, we simulated wetland restoration using changes
in wetland surface area.
The subbasins differ in size and in proportion of land that is currently wetlands. Representing
wetland restoration scenarios in the hydrologic model with across-the-board increases by a
given area (e.g., adding 1000 m2 of wetland area to each subbasin) would result in unequal
changes in subbasin percent diverted, which would make it difficult to make generalizations for
Tier II indicators. Alternatively, representing wetland restoration scenarios as across-the-board
increases by a given percentage (e.g., adding 10% to wetland area in each subbasin) would
increase the proportion of wetlands in some subbasins to greater than 100%. Instead, we
reduced the non-wetland area of the subbasin by a set percent (e.g., for a 10% conversion
of non-wetland area, a subbasin with 20% wetlands would increase to 28% wetlands). This
method ensured a consistent change in subbasin outflow while never exceeding 100% wetlands
in a subbasin.
We selected a range of wetland change scenarios that were realistic enough to inform Tier III
indicators, but also large enough to show detectable changes in flow downstream for
developing Tier II indicators. To determine what might constitute realistic wetland scenarios,
we used the average and range in surface area of restorable wetland fill sites identified in the
watershed by Golet et al. (2003). The average area was 6,026 m2, which is approximately the
area that would be added by a 1% conversion of non-wetland area (5,922 m2). The maximum
area restorable to wetlands identified by Golet et al. (2003) was 88,383 m2, which is
Chapter 2. Flood Modeling
-------
comparable to the area increased with a 10% conversion (59,216 m2). Based on our sensitivity
analysis, where contributing area was defined by the 100-m buffer around a wetland and was
comparable to a 50% conversion, we also modeled a 25% restoration scenario to account for
larger restorations; and, to represent restoration across a range of potential scenarios, we
added a 5% scenario. So, the four restoration scenarios we ran were 1%, 5%, 10% and 25%
conversions of non-wetland area to wetlands.
Location of Restoration in Scenarios
Because the hydrologic model is spatially lumped to the subbasin, the location of simulated
restorations can only be varied by subbasin. However, we expected the wetland scenarios
applied across different subbasins to have differing effects on flow and flooding. For example,
a subbasin that is directly upstream of a dam might alter downstream flow differently than one
with no dam downstream. Since there were many subbasin characteristics (e.g., distance to
outlet, downstream dams, and basin area) that might correlate with changes in flow, we ran all
four restoration size scenarios for each of the 139 subbasins individually, resulting in 556 "basin
scenarios."
Storm Events
We developed synthetic storms based on recently updated gridded storm recurrence period
data (DeGaetano and Zarrow 2011). We interpolated the gridded data over the entire
watershed rather than measuring at a single point such as the meteorological station data
used for calibration. To integrate these data, we aggregated the data to subbasins using their
centroids. We imported three storms, with 100%, 20% and 4% probabilities (1-year, 5-year,
and 25-year recurrence intervals) into HEC-HMS. Each of these three storms had 6 duration
intervals (1-hour, 2-hour, 3-hour, 6-hour, 12-hour, and 24-hour) used to create the synthetic
storm and define peak rainfall intensity (DeGaetano and Zarrow 2011). We excluded shorter
duration intervals because they exceeded the resolution of the hourly precipitation data used
for calibration and, because the model was run on a 1-minute time interval, shorter duration
intervals resulted in rainfall with unrealisticaMy high peaks (Appendix G).
The recurrence intervals for precipitation-based synthetic storms (1-year, 5-year, and 25-year)
did not result in a peak flow or flood with the same recurrence interval. This is primarily
because the evapotranspiration and antecedent moisture conditions at the time of the event
greatly influence how much precipitation actually ends up as runoff. We demonstrated this by
estimating the peak flow recurrence intervals for the three synthetic storms using both the
spring and fall calibrated models in peakFQ analysis (Table 2-4).
For the basin scenarios, we modeled the range of synthetic storms with the model calibrated to
wet conditions (Spring '01), because it validated better than the Fall '05 model and because wet
conditions typically lead to more flooding from the same size precipitation event. We ran the
four restoration sizes and three synthetic storms for each subbasin. A total of 1668 simulations
were run using the hydrologic and hydraulic models, plus a baseline for each synthetic storm
30 Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
(Table 2-5). These hydrologic and hydraulic simulations are the root of the process used to
derive Tier II and Tier III indicators presented in Chapter 3.
Table 2-4. Approximate recurrence intervals for synthetic storms and their peak flows using
different models
Recurrence
Spring
Fall
Precipitation
1 Year
5 Year
25 Year
1 Year
5 Year
25 Year
Flow
10 Year
100 Year
500+ Year
1 Year
3 Year
5 Year
Table 2-5. Number of simulations generated based on three synthetic storms, four restoration
sizes and 139 individual basins
1-Year 5-Year 25-Year
Restoration
1%
139
139
139
5%
139
139
139
10%
139
139
139
25%
139
139
139
Baseline
1
1
1
Chapter 2. Flood Modeling
-------
-------
CHAPTER 3: INDICATORS
In Chapter 2 we introduced and presented models that we used to generate subbasin-specific
scenarios for four different wetland restoration sizes and three storm magnitudes (Table 2-5).
In this chapter, we demonstrate how we derived Tier II and Tier III indicators from these
scenario results. First, we demonstrate how the scenario results can be used to produce Tier III
indicators, and illustrate their application to a comparison of wetland restoration in two
subbasins in the Woonasquatucket Watershed. We then present the development of Tier II
indicators using the scenario results combined with sensitivity analysis of the models presented
in Chapter 2.
The Indicators Framework
We compiled the ecosystem service benefit indicators using a framework developed by
Mazzotta and Wainger (in preparation). We summarize the framework here. It is based on four
questions that guide the process of indicator selection and measurement. It is not necessary to
fully answer all of the questions; they are ordered so that answering each question in turn adds
additional information to an ecosystem service benefits assessment. Each of these questions
may in turn be answered in lesser or greater detail. Thus, question 1 alone can provide some
information that adds value to a functional assessment in moving towards Benefit Relevant
Indicators (BRIs), and answering the first 3 questions provides basic BRIs. Adding information
that addresses question 4 will move the assessment closer to value assessment. The questions
are:
1. Is an ecosystem service supplied?
By definition (Munns et al. 2015), ecosystem services require use or appreciation by people and
thus are distinguished based on the interaction of supplied ecological outputs and demand for
those outputs by people. This step determines whether an ecosystem service exists by
assessing potential supply and demand of an ecosystem service by evaluating three things.
First, a service can only be supplied if wetland functioning meets thresholds required to provide
benefits to people. Second, people must care about, or demand, the service. For flood
regulation, this means that the wetland regulates flood waters at a level that will protect
structures and infrastructure that people care about. Third, in many cases other necessary
conditions, such as access points for recreational services, must be present for people to
benefit. These necessary conditions may include both physical supports to enjoying the
ecosystem service, such as required infrastructure and other conditions that allow physical
access, and institutional supports or constraints, such as regulatory limits to harvest (Olander
et al. 2015). In the case of flood regulation, this is typically not a relevant concern.
2. How likely is it that the service will continue to be provided over the long run?
This step assesses temporal reliability of ecosystem services by considering factors that affect
the probability that the wetland will continue to function at a sufficient level to provide services
over time. This is important to consider when comparing projects because two sites may
Chapter 3. Indicators
-------
provide identical benefits in the short run, but if one of the sites is threatened by stressors, it
may not continue to provide services into the future, resulting in a lower overall stream of
benefits for that site. The results of this step provide an indicator of whether and how long
a flow of benefits is likely to continue into the future.
3. How many people benefit?
This step assesses the number of people who benefit, which is the most basic measure of the
magnitude of benefits. If an ecosystem service exists (as determined through Question 1), the
number of people who benefit will be a strong indicator of its overall value. The total benefits
of an ecosystem change depend, to a large extent, on how many people stand to benefit. While
it is important to understand how much each individual values a change, the aggregate social
value of a change can be more sensitive to the size of the beneficiary pool than the magnitude
of change to an individual (Bateman et al. 2006), so as long as the average increase to an
individual is positive, the number of beneficiaries will be an important benefit metric.
Therefore, the number of beneficiaries can be used as a primary indicator in the most basic
of benefit indicator approaches.
4. By how much do people benefit?
This step assesses the magnitude of benefits to affected individuals or households, using
indicators of magnitude of the change in quality or quantity of the ecosystem service relative to
the pre-change baseline, availability and quality of substitutes, availability and quality of any
necessary complementary inputs, and strength of preferences for the ecosystem service. This
is a more difficult question to address using indicators than the first three questions.
This step may also include measures related to environmental justice regarding who is likely
to benefit in different locations. An example is the Social Vulnerability Index (Cutter et al. 2003).
Tier III Assessment
This section describes how we developed Tier III indicators based on the models presented in
Chapter 2, and presents an example of a hypothetical comparison of wetland restoration in two
subbasins (Figure 3-1). Because our model compares restorations at the subbasin scale, the
model cannot compare sites within a given subbasin (mean area=0.7 km2; 170 acres). Flows
from a subbasin depend on total wetland area within that subbasin, so a comparison of two
sites within a subbasin would result in the larger site ranked as superior. Therefore, a user
wanting to compare sites within a subbasin might make that decision either by simply selecting
the larger site, or by considering additional location-specific factors that might favor one site
over the other (e.g., slope), using best professional judgment or alternative functional
assessment tools. In addition, the decision maker might consider other ecosystem services
provided by each site.
Flood reduction results for the two restorations in the subbasins presented are from the
scenario results modeled (Chapter 2, Wetland Restoration Scenarios Section). Of the four
wetland restorations (1%, 5%, 10% and 25%) the largest, 25%, is shown in the hypothetical
34 Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
comparison to illustrate the largest predicted decrease in flooding. Of the three synthetic storm
conditions (1-year, 5-year, 25-year), we chose the 1-year storm because it produced a baseline
peak flow (1116 cfs, a 10-year flow) closest to those used to calibrate the hydrologic model
(1070 cfs). Although larger storms are used in flood mapping such as the DFIRM, regional
stormwater regulations primarily target small events. Stormwater regulations in Rhode Island
require stormwater conveyances to handle at least the 10-year, 24-hour precipitation storm
event (Rl DEM 2010), and Massachusetts requires that development peak discharge rates are
equal to or less than the predevelopment for the 2-year and 10-year 24-hour precipitation
events (MassDEP 2008). However, both states require onsite attenuation or no offsite impacts
for the 100-year 24-hour precipitation events. We would expect the same comparison based on
a larger storm to show more beneficiaries in the flood zone who would potentially benefit from
the restored wetlands.
Modeled Stream Readies
Subbasin 1540
Subbasin 2370
~ Subbasins
Figure 3-1. Subbasins selected for the example application
Chapter 3. Indicators
-------
The remainder of this section describes how we translated our model results into Tier III
indicators, and presents our example application of these indicators. Figure 3-2 illustrates the
process of applying the indicators. Throughout, we have made various assumptions that a user
of this approach might change, depending on the needs and context of the analysis. Wherever
possible we have specified our assumptions, and discuss how a user might modify the analysis
to address different concerns.
Initial Screening
Site Comparison
I. Assess Service Provision
Evaluate functional thresholds
Wetland volume
and percent
outflow diverted
both >0 ?
'
Evaluate demand
Valued assets in
the downstream
flood-prone
area?
II. Assess Temporal Reliability
x
Compile information to
indicate potential stressors:
• % imperviousness
• Projected development
III. Assess Who Benefits
x
Compile information on the
number of structures in the
flood zone:
• Number of E911 addresses
IV. Assess Benefit Magnitude
IT
Estimate reduction in flood
depth
Estimate effects of substitute
sources of flood protection
X
Estimate social vulnerability
Figure 3-2. Indicator application process
I, Assess existence of an ecosystem service
We have applied this step as an initial screening, using yes/no questions to determine whether
functioning meets the necessary threshold and potential beneficiaries exist.
Are functional thresholds required to supply the service met?
To provide flood regulation services the wetland needs to retain enough water to reduce
downstream peak flows. To perform this retention function, a wetland must provide adequate
retention volume and have a large enough source of water available for redirection into
retention. In our model, we had two metrics for the retention functioning of wetlands, wetland
volume and percent outflow diverted, both quantified using GIS calculations (see Chapter 2
Treatment of Wetlands in the Models Section and Appendix B). Based on observations of
modeled flows we have assumed that, provided both metrics are greater than 0, the service
threshold is met. This assumption can be adjusted within our approach if a stricter definition is
desired, but it was adequate for the smallest restoration scenarios we analyzed. Some of the
36 Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
existing functional assessments could also be used to answer this question. The Miller and
Golet (2001) functional assessment requires wetlands be basin wetlands to perform flood
regulation functions (see Chapter 2 Wetland Flood Protection Functional Assessment Section
for a more in depth comparison).
Is there evidence of demand for the service within the relevant provision area?
For people to benefit from flood regulation there must be structures, infrastructure, or other
valued assets that are vulnerable to a flood hazard within the downstream flood-prone area.
We evaluated this by determining, through GIS analysis, whether there are Rl E911 addresses
(RIGIS 2014) for buildings within the baseline scenario flood zone (Appendix H-l). The
delineation of the flood zone will vary with the size and type of storm evaluated. Therefore,
users of this approach in other locations may want to consider the most relevant storms to
model for their purposes. In addition, we have chosen to evaluate only the existence of
buildings, as indicated by E911 addresses, in the modeled area. The analysis could be expanded
to also include other valued assets such as important roads or other infrastructure, or could be
refined to only include those valued assets exposed to floods of a certain magnitude, such as
floods of 1 foot or more.
For our two comparison subbasins, we found that both conditions are met (Table 3-1). Percent
Flow Diverted and Wetlands Volume in both subbasins exceeded the threshold required to
perform flood regulation services (Table 3-2). There were also structures in the 10-year flow
baseline flood zone produced with the 1-year storm (Appendix H-l).
Table 3-1. Assessment of supply and demand
Indicator: Criteria Subbasin 1540 Subbasin 2370
Thresholds met? Wetland area >0 m2? Yes Yes
Wetland volume >0 m3? Yes Yes
Evidence of demand? Structures in flood zone? Yes Yes
Table 3-2: Subbasin and restoration characteristics
Subbasin 1540 Subbasin 2370
Total area
1.88 km2
1.76 km2
Original wetlands area
0.28 km2
0.23 km2
Restoration area
0.40 km2
0.38 km2
Percent diverted
36%
35%
Wetlands volume
272,702.8 m3
244,336.2 m3
Chapter 3. Indicators
-------
II. Assess temporal reliability of the service
This step addresses the question: What is the probability that the site will continue to provide a
service into the future? Our modeling did not estimate this probability. However, some of the
information used to parameterize the models might be used, combined with judgments by the
decision maker, to assess whether one site is likely to provide services more reliably over time.
CN (Appendix D) and imperviousness (Appendix D) both indicate level of urbanization. A
restoration site in an urbanized subbasin is likely more exposed to stressors and future
development, meaning it has lower reliability. Therefore, either lower CN or lower
imperviousness should indicate higher reliability. An additional indicator available for Rhode
Island is Rhode Island Statewide Planning's projected development for 2025 (Appendix H-2).
From the data on development projections for 2025, we used information on the percent of the
subbasin area which is protected or otherwise expected to have limited development in 2025
(Figure 3-3). A higher value for % limited development thus indicates higher reliability.
In the case of our simulated restorations, we summarized all three reliability indicators for each
subbasin. In cases where specific sites are being evaluated, a more detailed stressor assessment
based on the local conditions in the surrounding area may be possible and warranted. While it
is not practical to account for all possible future scenarios, modeling and expert judgment of
potential threats to persistence of benefits can strongly influence site prioritization and
therefore are important to consider. Based on the indicators % imperviousness, CN, and
projected 2025 land area protected from development for the subbasins we conclude that
Subbasin 2370 has a higher probability of reliably providing flood regulation services over time
(Table 3-3).
Table 3-3. Assessment of reliability
Indicator Subbasin 1540 Subbasin 2370
% Imperviousness 13.92% 6.83%
Curve number (CN)f 70.99 65.83
% Limited development 2025* 13.76 23.53
*Limited Development includes: "Conservation/Limited", "Major Parks & Open Space",
"Reserve" and "Water Bodies", and a higher number indicates higher reliability.
fA lower Curve Number or percent imperviousness indicates higher reliability.
38 Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
2025 Projected Landuse
| Sewered Urban Developed
Urban Development
Reserve
Conservation/Limited
i ; Prime Farmland
| Major Parks & Open Space
Water Bodies
N
A
0 0.125 0.25 0.5
r\
Oft
2025 Projected Landuse
| Sewered Urban Developed
Urban Development
Reserve
Conservation/Limited
Prime Farmland
| Major Parks & Open Space
Water Bodies
0 0.125 0.25
A
Figure 3-3. 2025 Projected Landuse for Subbasin 1540 (left) and Subbasin 2370 (right)
III. Assess who benefits
This step addresses the question: How many people benefit from the service? In our approach,
we quantify this using the number of addresses in the E911 database that fall within the
predicted flood zone, for the modeled floods. As discussed above, this analysis could be
extended to include critical roads and other infrastructure. To determine the number of
beneficiaries in each restoration scenario, we examined the addresses within the baseline
scenario flood zone (Appendix H-l) and used that as a count of the number of buildings that
were predicted by the model to experience a reduction of 0.04 ft or more in flood depth after
the restoration. We chose a depth of 0.04 ft because of the vertical accuracy of the Digital
Elevation Model (DEM) used to create and compare flood profiles. The DEM is provided with a
0.01 m vertical precision, and had a Root Mean Square Error of 0.067 m based on raw LiDAR
calibration control points. Although our hydraulic model showed changes in flooding of less
than 0.04 ft, we used this as a "detection threshold," since lesser changes could be attributed
to error.
For our example, the restoration in Subbasin 1540 reduced flooding for one building, whereas
the restoration in Subbasin 2370 protected five buildings (Table 3-4 and Figure 3-4). These
numbers are low mainly because of the limited number of downstream buildings vulnerable to
flooding during a 10-year peak flow flood. Based on these indicators, we concluded that
Subbasin 2370 has a greater number of beneficiaries.
Chapter 3. Indicators
-------
Table 3-4. Assessment of quality of service and beneficiaries' vulnerability
Social
Flood Depth Vulnerability
Subbasin 1540
Beneficiaries
Reduction
Substitutes
Index
DFIRM
Address #6
Subbasin 2370
Beneficiaries
0.13 ft
0
Medium Low
Minimal
Address #1
Address #2
Address #3
Address #4
Address #5
0.19 ft
0.11ft
0.11ft
0.06 ft
0.06 ft
0
0
0
0
0
Medium
Medium
Medium
Medium
Medium
Outside DFRIM
Outside DFRIM
Outside DFRIM
Outside DFRIM
Outside DFRIM
Subbasin 2370 (B)
Subbasin 1540 (C)
~ Model Boundary
9 Address With Decreased Flooding
Change in Flooding
Increased (>0.04 ft)
Non-Detectable Increase (0 to 0.04ft)
Non-Detectable Decrease (0 to -0.04ft)
Decreased (<-0.04ft)
Kilometers
A
Ki ometers
Figure 3-4. Building locations in the altered flood zone. Subbasin 2370 is shown in purple on map a, with
its flood zone magnified in map b. Subbasin 1540 is shown in orange on map a, with its flood zone
magnified in map c.
40 Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
IV. Assess the magnitude of benefits to individuals or households
This step addresses the question: How much do people benefit? It may incorporate a number
of measures, including the magnitude of change in the service relative to the baseline (i.e., the
quality of the service), the availability and quality of substitutes, any necessary complementary
inputs, and strength of preferences for the service.
We quantified the quality of flood regulation using an estimate of the reduction in flood depth
experienced at the protected buildings. For evaluation of quality of flood reduction to
infrastructure that is a two dimensional area, e.g., a road, park, or athletic field, the quality
of benefits could be quantified by flood extent (area), or the peak volume of water flooding
that area.
Other wetlands, dams or levees may provide substitute sources of flood regulation. However,
the way in which these substitutes provide flood regulation services differ and those
differences can determine the quality of services provided by substitutes. Assuming levee
design conditions are intended to completely block the flow of water to assets they protect,
upstream restoration will not benefit people protected by the levee unless water overtops the
levee. If water overtops the levee, additional flood reduction benefits from upstream
restoration could be provided. Similarly, if a dam detains water downstream of a restoration
site, the restoration may not add to flood regulation benefits for beneficiaries downstream
of that dam. Miller and Golet (2001) assumed this to be the case (Appendix A). However, the
effect of a dam on an upstream restoration's benefits depends on the flood regulation benefits
provided by the dam and the flood being considered. If there is flooding downstream of the
dam and the upstream restoration would reduce water flowing over the dam, the benefits
of the restoration may be decreased, but they are not completely substituted. In contrast to
levees and dams, existing wetlands already provide flood reduction benefits downstream but
do not prevent restored wetlands from providing additional benefits. Wetland flood reduction
benefits are cumulative and, once the maximum flood reduction benefits are received (i.e.,
there is no flooding), it is hard to discern which wetland provided the benefits.
Our hydrologic models and hydraulic models include dams, levees and existing wetlands, so
the effects of these in terms of substituting for restored wetlands is already incorporated in the
predictions of flood extent. It would be possible to remove dams and levees from the models,
to compare wetland benefits to benefits from substitutes. However, since the intent is to
compare restoration across locations, there is little to be gained from doing this. People may
also choose to adapt to flood risk by taking actions on their own property, for example by
raising the elevation of a vulnerable structure. We did not consider such potential adaptation
actions, but users of this approach could develop indicators of adaptation.
Strength of preferences and quality of complementary inputs are the last set of factors that
influence benefits. Flood benefits do not require complementary inputs, and strength of
preferences is difficult to quantify without directed data collection from beneficiaries. However,
there are a few ways to characterize beneficiaries based on their ability to recover after
Chapter 3. Indicators
-------
sustaining flood damage. A beneficiary with fewer resources, such as income, flood insurance,
or ability to find temporary shelter, would be expected to have a harder time recovering from
flooding. One existing indicator for this is the Social Vulnerability Index, which synthesizes
multiple census socioeconomic variables (Cutter et al. 2003). Another simple way to indicate
resources for recovery is to overlay the beneficiaries with FEMA's DFIRMs to analyze whether
the building is within the 100-year (1%) or 500-year (0.2%) flood zone, which is the basis for
National Flood Insurance Program's (NFIP) regulations and flood insurance requirements
(although homes without a mortgage are not required to have flood insurance, so one cannot
assume all homes within the DFIRM are insured) (Appendix H-5).
We present the results of our example analyzed in terms of marginal benefits, defined as the
decrease in flood depth, quantified for each beneficiary from Figure 3-4 (Table 3-4). Although
substitutes are already incorporated into the model, we counted any possible substitutes
between beneficiaries and the restoration subbasins (Appendix H-4; Table 3-4). We indicate
preferences based on the Social Vulnerability Index (Figure 3-5; Appendix H-3) and the DFIRM
(Figure 3-6; Appendix H-5).
Subbasin 1540
O Address With Decreased Flooding
| Model Boundary
Social Vulnerability Index (SoVi)
| High
Medium High
Medium
Medium Low
Subbasin 2370
O Address With Decreased Flooding
| Model Boundary
Social Vulnerability Index (SoVi)
| High
Medium High
0.125 0.25
0.5
0.75
—i Kilometers
0.1 0.2
0.4
0.6
—i Kilometers
Figure 3-5. Census tract Social Vulnerability Index for Subbasin 1540 (left) and Subbasin 2370 (right)
42 Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
iKilom eters
Kilometers
• Decreased Flood Address
Model Boundary
DFIRM Flood Zone
100 Year
500 Year
# •
Minimal Risk
Figure 3-6. Buildings where flooding is decreased by restorations in Subbasin 2370 and 1540 (black
points) compared to DFIRM flood zones (100-year and 500-year) in the FEMA modeled area (orange)
and outside the FEMA modeled area (white). Our model boundary (black outline) is provided as a
reference for building locations.
Based on these indicators, we concluded that Subbasin 2370 provides greater benefits to one
of its beneficiaries than Subbasin 1540 provides to its sole beneficiary. None of the beneficiaries
of either subbasin had dams (i.e., potential substitutes) between the restoration and their
location. The beneficiaries of wetland restoration in Subbasin 2370 had a higher vulnerability
index than those of restoration in Subbasin 1540. None of the buildings fall within the DFIRM
Chapter 3. Indicators 43
-------
flood zones, so would not be required to have flood insurance, despite being vulnerable to
flooding based on our model, which included smaller streams and smaller floods.
Tier II Indicators
Tier III indicators provide a rigorous, quantitative, and defensible evaluation of the benefits
of potential restorations. However, the need for extensive watershed-specific modeling makes
this type of assessment inaccessible to many decision makers. Although Tier II benefit
indicators are expected to be less rigorous than the Tier III indicators, they should be able to
inform decisions without intensive modeling, based instead on existing datasets. While Tier II
indicators are more easily applied, in many cases they still require some analysis, often
requiring knowledge of Geographic Information Systems (GIS).
Like the Tier III indicators, the Tier II benefit indicators address the four questions of the
indicators framework, although they do not answer all of the questions at the same level of
detail, and some of the optional sub-questions are not addressed. Rather than demonstrating
a simulated example application as we did forTier III, for the Tier II indicators we show how we
generalized model results and translated them into indicators that can be used by others.
The next sections describe this process.
Generalizing from Tier III Model Results
The primary consideration for generalizing to Tier II indicators is the need to determine the
relevant benefits area for flood regulation services. Flood regulation occurs at the site of a
wetland, but benefits people and structures downstream of the wetland site. Therefore, our
primary aim in generalizing from our detailed modeling was to determine the "benefits area"
and the factors that may influence how this varies.
Methods
Using the hydrologic and hydraulic models, we generated hydrologic and floodplain change
maps for 1,668 subbasin scenarios (139 subbasins*4 restoration scenarios*3 storm events)
(Table 2-5), plus a baseline, or no restoration scenario, for each synthetic storm (3 basin
scenarios). There were several options for how to evaluate this extensive data set to assess
where benefits are provided.
Option 1 was to perform the same comparison as Tier III for all wetland restoration simulations:
comparing the change in flood depth or flood extent from the baseline at all protected building
locations for all the basin scenarios. However, the actual location of the buildings or assets at
risk is unique to and characteristic of the development patterns and physical geography of the
study area, a limitation when developing indicators to be used beyond the study area.
Option 2 was to compare changes in flooded area from the baseline for all restoration scenarios
without consideration of where assets are located. Although this option would assess the effect
of restorations on the entire floodplain, there were drawbacks to this option as well. The
concern was that stream and floodplain morphology play an important role in determining
the extent of flooding. Two simulated wetland restorations may result in an equal reduction of
44 Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
flows or flood-water volume, but will have very different effects on the extent of flooding based
on the downstream morphology. A narrow-deep channel will show little change in flood extent
but a greater change in flood depth in comparison to a wide-shallow channel. This can be seen
in the Tier III indicator comparison maps (Figure 3-4).
Option 3 was to simply examine flood-water volume directly, as percent change from baseline
peak flows from the hydrologic model. These peak flows were available for all subbasin
simulations at all element locations along the stream network: basins, junctions, reaches,
reservoirs and the outflow point. Because the model is spatially lumped at the subbasin, the
location of a simulated wetland within the subbasins is unspecified, so the distance from the
simulated wetland to the outlet of the subbasin was assigned the longest flow path distance.
Results
Figure 3-7 shows plots of the change in flow from the baseline against the distance downstream
from the simulated restoration for each subbasin scenario, showing the distance decay in the
reductions in flow from the restorations (Figure 3-7; the same figure is provided with the two
subbasins from the Tier III application highlighted in Appendix I). Closer to the restoration, most
scenarios show the full change in flow from the restoration (-1%, -5%, -10%, or -25%), with a
more or less rapid decline in the percent change in flow with distance from the outlet of the
scenario subbasin (note that the y-axis scale changes with increased restoration size).
In order to understand how far downstream these changes in wetlands influence downstream
flow, we examined the changes in flow to estimate how far downstream the change in flow
became effectively zero. In many instances the percent change fluctuates and approaches but
never actually reaches zero. This suggests sustained low change in flow could be due to
rounding errors and not due to significant changes in flow as a result of the upstream
restoration.
We first defined a change in flow as effectively zero using a conservative cutoff—any change
less than -0.2%—by estimating where the subbasin scenario curves began to flatten. This
allowed us to estimate the distance downstream to the -0.2% threshold. We summarized these
results using box plots (Figure 3-8) by storm size (Figure 3-8a), by wetland change (Figure 3-8b),
and for all scenarios (Figure 3-8c).
Across storms (Figure 3-8a), the 1-year storm has the greatest variability and reaches no change
in flow slightly farther downstream, indicating that the simulated wetland restoration scenarios
had the strongest influence during small events. This makes sense given that each wetland
scenario holds a fixed volume of flood-water and that volume is a smaller proportion of a larger
event. Across wetland change scenarios (Figure 3-8b), the mean and variability in the distance
downstream to no change in flow increases as wetland scenario size increases indicating the
larger change in flow associated with a larger increase in wetland extent results in changes in
flow that are sustained farther downstream. Looking across both changes in wetland extent
Chapter 3. Indicators
-------
0 5000 10000 15000 20000 25000 0 5000 10000 15000 20000 25000 0 5000 10000 15000 20000 25000
Distance Downstream (m)
25-year
Storm Event
5-year
1-year
Figure 3-7. The change in flow (%) with distance downstream (m) from the simulated restoration for all
subbasins (n=139) for each restoration scenario (wetland change scenarios 1%, 5%, 10%, and 25%) and
synthetic storm event (1-year, 5-year, and 25-year). Note that the y-axis scale varies between sub-plots
for each of the wetland change scenarios.
and storm sizes (Figure 3-8c; Table 3-5), the longest downstream influences (mean of 7359.3 m)
were seen with the largest restoration scenarios (25%) in the smallest storm event (1-year),
again, because the largest wetland scenarios retain the largest volumes of flow which makes up
a larger proportion of a smaller event. The shortest downstream influences (mean of 3260.1 m)
were seen with the smallest restoration scenarios (1%) in the largest storm event (25-year).
The mean distance downstream where change in flow dropped below -0.2% for wetland change
scenarios, was 4311.7 m (2.7 miles), with individual simulations that ranged from 506.9 m (.31
mi) to 19464.2 m (12.1 mi) (Table 3-5).
Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
1 5 25
Storm Recurrence Interval (years)
1 5 10 25
Wetland Change Scenario (%)
01.1 05.1 25.1 01.5 05.5 25.5 01.10 05.10 25.10 01.25 05.25 25.25
Restoration Scenario
Figure 3-8. Boxplots of downstream distance (m) to no change in flow (<-0.2% change) by restoration
scenarios: (a) storm recurrence interval, (b) wetland change, and (c) both wetland and storm
recurrence interval.
Chapter 3. Indicators
-------
Table 3-5. Downstream distance (m) where change in flow became negligible (-0.2%).
Storm Event
Wetland
Restoration
1 Year
5 Year
25 Year
1%
Mean
3302.5
3255.8
3260.1
Max
6779.7
9234.6
9252.9
Min
506.9
506.9
506.9
SD
1632.6
1658.1
1701.7
5%
Mean
4696.3
3949.0
4137.5
Max
19464.2
12287.9
15411.5
Min
506.9
506.9
506.9
SD
3842.7
2438.6
2814.9
10%
Mean
6046.6
4093.4
4051.1
Max
19464.2
15099.1
13547.7
Min
506.9
506.9
506.9
SD
5110.4
2728.8
2593.4
25%
Mean
7359.3
3629.5
3959.9
Max
19464.2
9662.3
10997.6
Min
627.5
627.5
506.9
SD
5312.4
2182.3
2605.8
Next, we increased the threshold to a -1% change in flow as the cutoff to be considered
effectively no change in flow. We selected this new threshold based on inspection of flood
maps where changes in flow less than 1% frequently did not result in changes in flood depth
greater than the detection criterion of 0.04 ft. Figure 3-9 shows the box plots for this analysis.
Examination of results across storms (Figure 3-9a) showed that the mean and variability were
again greatest for the 1-year storm. Examination of results across wetland change scenarios
(Figure 3-9b) identified an issue with the 1% restoration scenarios when a 1% threshold is used
for defining change in flow as effectively no change in flow. Because the maximum change in
flow for the 1% restoration scenarios is already 1%, the threshold will be met as soon as
subbasin discharge joins other downstream flows. This results in distances to no change for the
1% scenarios equaling the distance of flow within the subbasin itself or the subbasin's longest
flow path distance. Excluding the 1% restoration scenario, the updated threshold showed the
same trend in variability and mean distance as with the smaller threshold, with mean and
variability in downstream influence increasing with restoration size.
48 Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
CD O
o o
£ o
~r
~r
1
1 5 25
Storm Recurrence Interval (years)
-------
Table 3-6. Downstream distance (m) where change in flow became negligible (-1%).
Storm Event
Wetland
Restoration
1 Year
5 Year
25 Year
5%
Mean
3302.5
3255.8
3299.4
Max
6779.7
9234.6
9252.9
Min
506.9
506.9
506.9
SD
1632.6
1658.1
1766.5
10%
Mean
3560.2
3634.2
3556.1
Max
8649.3
9662.3
11512.2
Min
506.9
506.9
506.9
SD
1804.1
2023.1
2029.8
25%
Mean
4545.7
3943.1
4101.8
Max
19464.2
12287.9
15411.5
Min
506.9
506.9
506.9
SD
3619.0
2438.0
2809.2
In general, for both cutoff points, the distance to zero change in flow stayed roughly the same
across storm recurrence intervals, but went up with the percent in wetland change. And as
described above, when combined, the longest distances were seen with the largest change in
wetland area and the smallest storm events.
Next, we examined how results vary with different restoration basin characteristics to see if
variability in downstream distance to the same thresholds (-0.2 and -1%) followed any general
trends (Figure 3-10; Appendix 1-2). These restoration basin characteristics included: the distance
to the gauged basin outflow, restoration basin area, initial percent wetlands, basin slope, basin
percent imperviousness, and the longest flowpath distance.
Scatter plot results based on the <-0.2 and <-1% thresholds were very similar, so we present the
-0.2% threshold here. Because some characteristics appeared to show slight trends, we
generated box plots based on splits at the mean value for each characteristic (Figure 3-11 for
<-0.2%; Appendix 1-3 for <-1% threshold).
50 Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
Dist USGS GAGE
Pet Wetlands
20000
0.0 0.2 0.4 0.6 0.3
Longest_Flowpath
Pctlmp
Figure 3-10. Downstream distance to no change in flow (<-0.2% change) plotted against subbasin
characteristics: distance (m) to the outlet (Dist_USGS_GAGE), subbasin area (Basin_Area_km2),
% wetlands (Pct_Wetlands), mean basin slope (Basin_Slope), % impervious cover (Pctlmp), and
the longest flowpath distance (m) (Longest_Flowpath).
o 1
Percent Wetland
0 1
Subbasin Area
-4-
0 1
Distance to Outlet
T"
~T~
T"
~T~
0 1 0 1
Basin Slope Percent Impervious
0 1
Longest Flowpath
Figure 3-11. Boxplots of downstream distance to no change in flow (<-0.2% change) plotted by subbasin
characteristics (distance to the outlet, subbasin area, % wetlands, mean basin slope, % impervious
cover, and longest flowpath) split into below (0) and above (1) the mean values for the basin
characteristics (Table 3-7).
Chapter 3. Indicators
-------
Though there were slight differences between the mean distance to zero change in flow above
and below the means for the subbasin characteristics, for the purposes of informing Level II
indicators these differences were very small (Table 3-7). The assumption that all wetland
restorations were the same distance from the subbasin outflow as the longest flow path for
that subbasin could have biased downstream distances, increasing the minimum distance to
the longest flowpath distance. However, the lack of correlation between longest flowpath
distance and mean distance to no change in flow (0.2%) suggests distance is not biased by the
longest flowpath assumption. Also, none of the means across subbasin characteristics varied
significantly, so this supports the selection of a threshold for downstream influence based on
the average downstream distances before changes in flow reach the no change threshold
(-0.2 or -1%). For the -0.2% threshold the mean distance was 4311.7 m and for the -1%
threshold this distance was 3688.7 m when the 1% restoration scenario was excluded. The
average of the two thresholds is 4000.2 m or 2.50 miles. There was considerable variation
about the mean for all subbasin characteristics, which indicates that there may be some
geographic settings in which downstream flow reduction extends over a larger area due to
variations in these characteristics. In an area where a particular characteristic has a strong
influence, restoration might optimize for that characteristic.
Interpretation of Results for Indicators
Based on our analysis, a reasonable distance for delineating the area where people could
benefit is 4 km (2.5 miles) downstream of the restoration, based on the mean distances to no
change in flow (both 0.2% and 1% thresholds) across restoration scenarios and storms. In some
cases there may be a larger distance where people could benefit—the mean distance to no
change in flow for the largest restoration (25%) and the smallest storm event (1-year, 10-year
flow) was higher, 7.4 km (4.6 miles). But based on our analysis, the 4 km (2.5 mile) distance
indicator was robust across most scenarios.
In our example results using the Tier III indicators, we found that the beneficiaries identified
were within 4 km (2.5 miles) of the subbasins with restoration (Appendix H-7). This is consistent
with our mean estimate for the Tier II indicators. Based on our modeling we found that, at least
for the watershed and conditions modeled, the downstream benefits area may not extend as
far as 5 miles (8 km), as assumed by Miller and Golet (2001).
Applying Tier II Indicators
I. Assess existence of an ecosystem service
In keeping with our goal of providing a relatively rapid assessment we have applied this step
using simple yes/no criteria. The indicators for service production are the same as those used
for the Tier III assessment. The user can either follow the GIS based process for calculating
wetland area and volume (see Chapter 2, Treatment of Wetlands in the Models), or can use a
functional assessment like Miller and Golet (2001) to define the thresholds for service
production. Flood reduction services do not require any complementary inputs so that, as long
as flood reduction service thresholds are met, a flood reduction service is produced.
52 Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
The quick check for demand is that there is evidence of demand for the service within the
relevant provision area being considered. This requires an initial determination of the relevant
benefits area. In the Miller and Golet (2001) functional assessment (Appendix A), the social
significance indicator determines beneficiaries based on their downstream distance from the
restored wetland: "Developedflood-prone areas downstream within 5 miles or to the nearest
dam (connection by stream or floodway required)." Through our modeling, presented above,
we found a 2.5 mile (4 km) downstream area to be the relevant area.
This initial question does not address how many people benefit but simply determines whether
there are potentially people who benefit. Therefore, it would simply answer the question: Are
there valued assets within the benefits area?
Tier III used the baseline map of modeled flooding as evidence of demand, meaning vulnerable
assets need to be identified some other way for Tier II. If available, FEMA DFIRMs (Appendix
H-5) can be used to represent the area expected to experience flooding. If a valued asset is
inside the 100-year or 500-year flood zone in the DFIRM map it is potentially vulnerable,
meaning there is potential demand for flood reduction services. In the Woonasquatucket
watershed, the DFIRM does not extend into the upper watershed, making it difficult to use as
an indicator if a proposed wetland restoration is in that part of the watershed. Also, in the Tier
III assessment we found that some locations outside the bounds of the DFIRM may experience
flooding.
There are several alternatives for defining and identifying valued assets. Some options include
using address points to locate protected buildings, as was done for Tier III (Appendix H-l), using
town parcel data, using other critical infrastructure including roads or emergency response
assets, using population data, or using imagery like that available through Google Earth to
visually search for assets exposed in the DFIRMs (Appendix H-6). An alternative to using DFIRMs
would be to confirm that flooding occurs in the area being studied through other non-spatial
datasets, such as repetitive loss areas (available through FEMA as part of the Community Rating
System; FEMA 2014), other insurance claims data, or interviewing emergency response
personnel or others familiar with flooding in the area.
II. Assess temporal reliability of services
The Tier III indicator assessment of temporal reliability of services did not rely on any model
outputs and can be applied in the same way for Tier II.
III. Assess who benefits
This step addresses the question: How many people benefit from the service? This step would
further refine the yes/no assessment of demand to estimate the number of people who
benefit. The Tier III indicators quantified how many people benefit from the service using the
number of addresses to count protected buildings in the modeled flood zone. For Tier II
indicators, people can be quantified the same way, using number of addresses, but determining
the relevant flood zone requires further analysis. We list several methods for doing this under
question I above.
Chapter 3. Indicators
-------
IV. Assess the magnitude of benefits to individuals or households
This step addresses the question: How much do people benefit? It may incorporate a number of
measures, including the magnitude of change in the service relative to the baseline (i.e., the
quality of the service), the availability and quality of substitutes, any necessary complementary
inputs, and strength of preferences for the service. Since each measure included increases the
complexity of the assessment, it will be difficult to find Tier II indicators that assess the
magnitude of benefits, and the user may have to rely on the number of beneficiaries combined
with best professional judgment of the factors listed here that indicate magnitude (as in our
Tier I indicator approach).
Our modeling results provide some useful information that can assist in applying professional
judgment to this question. In terms of the magnitude of change in the flood regulation service
relative to the baseline, we did not find strong evidence for specific factors that lead to
variations in the benefits decay function for downstream distance. However, the mean and
range for the maximum downstream distance where benefits are delivered suggests that flood
regulation services decrease quickly beyond a certain distance. One potential way to approach
this with Tier II indicators would be to qualitatively rank beneficiaries that fall within multiple
downstream distance buffers. For example, beneficiaries 0 km to 4 km "likely receive benefits"
whereas beneficiaries 4 km to 7.4 km "may receive low benefits."
Tier II indicators for substitutes could be analyzed in the same way they were for Tier III,
quantifying the number of dams and levees within the downstream buffer being considered.
Although Miller and Golet (2001) identified downstream dams as playing a strong role in
eliminating downstream flood regulation benefits from wetlands, we were not able to observe
dams eliminating wetland flood regulation benefits in this way. Percent wetlands already in the
subbasin being restored did not show a strong negative influence on downstream benefits of
additional restored wetlands as might have been expected.
The Tier II benefit indicators successfully address the four question in the indicators although
they do not answer all of the questions at the same level of detail as Tier III indicators. While
Tier II indicators are more easily applied, in most cases they still require some analysis, often
requiring knowledge of Geographic Information Systems (GIS). We generalized model results
and translated them into indicators for the downstream distance within which flood regulation
benefits are able to be received. Instead of a quantitative decay function for the quantity of
benefits delivered, a qualitative downstream boundary can be used where beneficiaries 0 km to
4 km "likely receive benefits" whereas beneficiaries 4 km to 7.4 km "may receive low benefits."
The observed trends in this downstream boundary based on storm and restoration size also
provide useful information when adapting such boundaries to a more specific decision context.
54 Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
i: i'|i i » i l ¦ i» -l- -11 mi ni • i - 111 >
In this report, we have presented benefit indicators for the flood regulation service provided
by wetlands, and an approach for developing those indicators. The indicators presented in
Chapter 3 follow a framework that is grounded in economic theory (Mazzotta and Wainger
in preparation) and a quantitatively defensible modeling process (outlined in Chapter 2).
While the indicators themselves are useful for others evaluating flood regulation benefits,
equally important is the general approach to developing benefit indicators that we present,
and its compatibility with typical functional assessment tools and potential usefulness for
improving benefit transfer.
In this chapter, we summarize some caveats and present important considerations regarding
the broader applicability of our specific Tier II and Tier III indicators. Applicability beyond our
case study for the Woonasquatucket Watershed hinges on the applicability of the models used
to develop the indicators, as well as data availability for and characteristics of other
watersheds. For the Tier III indicators, we chose our models with future applications in mind,
and we provide suggestions for improving such applications to other geographic locations.
While some of the Tier II indicators developed are broadly applicable across locations, not all
are applicable to all decision contexts or all watersheds. We summarize limitations on Tier II
indicator applicability and what might be required to expand this applicability.
Modeling Summary
Although we chose the models we considered to be most appropriate, based on the selection
criteria detailed in Chapter 2, some aspects of modeling might be improved with additional
data and effort, or enhanced models. It is important to note that the primary reason for
developing benefit indicators is to inform decisions by facilitating comparisons across sites,
and not to provide the most precise predictions of flood impacts. Thus, we balanced precision
with the ability to develop useful benefit indicators that allow for comparison across wetland
restoration scenarios.
We point out some of the model complications and caveats here, to inform future modeling
efforts of this kind. These include:
• Past studies used HEC-HMS to effectively show the role of wetlands in the watershed,
but if trying to develop indicators for individual wetlands (rather than by subbasin) a
spatially distributed model has some clear advantages. HEC-HMS is a spatially lumped
model, and trying to make it function in a more distributed way by using more subbasins
increased computational requirements and strained time interval limitations, making
machine optimization arduous. A more distributed model might reduce optimization
time, at the expense of increased model complexity and data requirements. A spatially
distributed model would better account for where wetlands are located within the
subbasin and how much water is available to them. A spatially distributed model may not
Chapter 4. Discussion and Next Steps
-------
be the only solution; it may also be possible to model wetland catchments and then
parameterize that in HEC-HMS. Such a method still would not necessarily account for
water available to wetlands from adjacent streams however.
• With additional modeling it might be possible to improve upon the assumptions inherent
to how wetlands were integrated into these models, for example the size of their
catchment area or the effect of restorations on infiltration parameters. Our assumptions
about the runoff available to wetlands were purposely conservative. Sensitivity analysis
suggests that including even a 100m buffer catchment around wetlands would
significantly increase their impact on downstream flows. Better accounting for the actual
water available to wetlands would be a major improvement.
• Although model fit improved by using a spring storm, which is the time when flooding is
most likely, the variability in antecedent conditions for reservoir release and abstraction
made it very difficult to generate a single model that generalizes all the potentially
relevant conditions.
• Calibrating two models based on distinctly different storms showed how greatly
antecedent conditions can impact flood model results. A model that could better account
for these, and better data quality could improve model results.
• Sensitivity analysis of precipitation in the models showed small increases in total
precipitation cause large increases in flow relative to total flow, suggesting precipitation
data quality is very important to model accuracy. HEC-HMS is able to model gridded
precipitation data and this may be worth investigating in future modeling efforts,
especially when stationary precipitation gages may not be representative of rainfall in
the watershed
• In a watershed where more data were available on characteristics of dams and their
actual management, it would be possible to better model the actual role dams play as
substitutes for wetland flood reduction benefits.
• Baseflow was accounted for outside of the model and then removed from flow. Although
reservoirs were parameterized with an initial discharge, a model better equipped for
baseflow from small diffuse reservoirs might perform better. Better data on these
reservoirs, their actual storage-discharge relationships, and how they are managed,
such as drawdown before storms, could also improve models for watersheds like the
Woonasquatucket where there are numerous small dams.
• HEC-RAS performed well for hydraulic modeling but additional information, such as
better characterization of dams and more information about infrastructure in the
headwaters in general, might improve the flood model and better represent the role of
wetlands in flood retention. We used steady flow analysis, but non-steady flow analysis
may be able to better capture storage in the watershed. This becomes more important in
Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
a watershed with many diffuse reservoirs and could also help account for storage in
buffer wetlands.
Tier III Indicators
The Tier III application was based on model results and, although the modeling process is
transferable, the modeled results are specific to the modeled area within the Woonasquatucket
watershed. Where indicators did not require model results and instead used more widely
available datasets such as the Social Vulnerability Index, the indicators presented are applicable
to any location where the required data are available (NOAA 2015). Further, the specific models
used may not transfer well to another setting, but the process of summarizing model results to
inform our indicators framework is transferable.
The Tier III indicator development process also showed how the models could quantitatively
relate functional assessments, such as the Miller and Golet (2001) assessment, to benefit
indicators. OurTier III indicators might also replace or augment similar indicators in other
ecosystem service models. For example, ARIES uses FEMA 100- and 500-year flood zones.
Where developed, the modeled flood area from ourTier III modeling process could be used in
place of these FEMA flood maps. This would allow for a better quantification of the magnitude
of flooding that potential beneficiaries are projected to experience. Our overall indicator
framework can be applied to link the benefits indicators to values, and may allow for synthesis
with valuation methods. For example, the SolVES tool (Sherrouse and Semmens 2010) can be
used to transfer social-value models to physically and socially similar areas, and benefits
indicators such as those presented here could be one way to evaluate similar areas.
Tier II Indicators
Tier II indicators are designed to be more transferable than the Tier III indicators. Many of the
Tier II indicators were developed from existing datasets, but required some geospatial analysis
to characterize specific wetland restorations. In addition to these existing datasets, we used our
models to simulate restorations that could be generalized to a downstream distance for
transfer of flood regulation benefits. These are more transferable to watersheds with similar
hydrological and land use characteristics.
Though the trends in our estimated downstream distance for benefits are somewhat uncertain,
they give a valid range to consider in identifying potential beneficiaries. This range of
downstream distance also provides a quantitative model-based test of assessments that are
based on expert judgment, such as the 5 miles (8 km) suggested in Miller and Golet (2001). We
were unable to generalize the rate that benefits decay as they are transported downstream,
meaning the change in the level of benefits received with greater distance from a restoration
site cannot be quantified directly from our results, but will necessarily involve some expert
judgment regarding how benefits may decline with distance.
We were also unable to unambiguously quantify the role of dams in decreasing benefits from
wetlands. Although dams clearly play a role in the Woonasquatucket watershed, based on the
Chapter 4. Discussion and Next Steps
-------
data available and how dams and reservoirs were modeled, the presence of a dam downstream
from a wetland does not necessarily prevent the flood regulation benefits of the wetland from
reaching beneficiaries downstream of the dam. This may be explained by the fact that many of
the dams in the Woonasquatucket watershed are small and become run of the river during
larger storms. Larger dams that are managed more directly for flood abatement may prevent
additional downstream flood benefits from upstream wetlands, and our models did show that
some dams reduce wetlands benefits downstream. Therefore, our results regarding dams are
mixed, and could simply be because of the variations in conditions across dams, which may
need to be considered by decision makers on a case by case basis, based on local knowledge.
Our indicator framework guidebook will allow for this consideration.
Until similar modeling is conducted in other watersheds, it is difficult to say how transferable
Tier II indicators are. We expect that these indicators could be used in similar watersheds in the
Northeast. Downstream distances for benefits delivery did not appear to vary greatly within the
range of subbasin areas and imperviousness explored. Other factors such as stream
morphology and subbasin slope may play a role in this as well. Although we were unable to
develop generalized indicators quantifying the role of substitutes, such as dams, in decreasing
wetlands benefits, it is possible the number of dams in the Woonasquatucket influenced our
results for the downstream distance of benefits.
Future Directions
The overall process for developing tiered flood reduction benefits indicators from quantitative
modeling was successful. There are several areas for future research to further develop this
approach. Since wetlands provide many benefits, one area for future exploration is whether
the process we followed can be used for other wetlands benefits.
Although we suspect that Tier III indicators will provide more benefit relevant information than
Tier II indicators, it will be important to examine how the information provided by each relates
to results from valuation studies and the implications for decisions that are based on each level
of indicator. If there is little difference between decisions made based on these different levels
of information, then Tier II indicators could be used in most cases, because they require less
time and expertise. Tier II indicators will be more rapid than most of the methods previously
available and require much less analysis than the Tier III indicators. However, Tier II indicators
should be tested and compared to models for other watersheds before assuming
transferability.
Now that Tier II indicators have been developed, the resource requirements to implement
these indicators elsewhere using a similar approach could be reduced through tools we have
developed. For example, the geospatial tool used to calculate wetland volumes based on
characterizing their perimeters and depths from Digital Elevation Models will be available for
future use. Tools could also be created to characterize beneficiaries more easily or determine
which are within the downstream distance required to benefit from wetland sites. These tools
58 Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
could also be provided outside of proprietary geographic information systems through web
based services or as standalone programs.
Summary
What is most useful about the approach presented here is that it directly incorporates people
and the benefits they receive from ecosystem restoration. Further, it provides a framework that
can be used to compare potential wetland restoration scenarios based on these benefits
without the need for estimating dollar values. Using an approach to assessing non-dollar
benefits that is grounded in economic principles allows for more robust discussion of
alternatives through a disaggregated and transparent presentation of the various factors that
are likely to affect the level of benefits to people. This can inform many decision contexts
where a strict benefit-cost framework is either not appropriate or not necessary. This indicators
approach can also easily be extended to incorporate conceptions of value beyond the economic
definition of value, and can be transferred to other decision settings and benefit types. We
demonstrate how it can be used to complement an existing functional assessment approach,
and propose that it can also be used to inform benefit transfers to add more insight into
variations in restoration benefits across locations, and who might receive those benefits.
Chapter 4. Discussion and Next Steps
-------
-------
References
Ackerman, C.T. 2009. GIS Tools for Support of HEC-RAS using ArcGIS (HEC-GeoRAS) User's
Manual Version 4.2. US Army Corps of Engineers, H.E. Center, CPD-83, 246 p.
ACOE. 2000. Woonasquatucket River Dam Removal Flood Study. U.S. Army Corps of Engineers.
Providence, Rl, 26 p.
Anderson, B., Armsworth, P., Eigenbrod, F., Thomas, C., Gillings, S., Heinemeyer, A., Roy, D.,
Gaston, K. 2009. Spatial covariance between biodiversity and other ecosystem service
priorities. Journal of Applied Ecology 46(4), 888-896.
Bateman, I.J., Day, B.H., Georgiou, S., Lake, I. 2006. The aggregation of environmental benefit
values: Welfare measures, distance decay and total WTP. Ecological Economics 60(2), 450-
460.
Bedient, P.B., Huber, W.C., Vieux, B.E. 2007. Hydrology and Floodplain Analysis (4th ed.).
Upper Saddle River, NJ: Prentice Hall, 817 p.
Bengtson, M.L., Padmanabhan, G. 1999. A hydrologic model for assessing the influence of
wetlands on flood hydrographs in the Red River Basin: Development and application.
North Dakota Water Resources Research Institute, North Dakota State University, 75 p.
Bicknell, B.R., Imhoff, J.C., Kittle, J.L., Jr., Donigian, A.S., Jr., Johanson, R.C. 1997. Hydrological
Simulation Program—Fortran: User's manual for version 11:U.S. Environmental Protection
Agency, National Exposure Research Laboratory, Athens, Ga., EPA/600/R-97/080, 755 p.
Bousquin, J., Hychka, K., Mazzotta, M., and Berry, W. 2014. Flood Protection from Wetland
Restoration: Focus on Benefits and Beneficiaries. US EPA, ORD, Narragansett Rl, 66 p.
Brander, L.M., Florax, R.J.G.M., Vermaat, J.E. 2006. The Empirics of Wetland Valuation:
A Comprehensive Summary and a Meta-Analysis of the Literature. Environmental and
Resource Economics 33(2), 223-250.
Brinson, M.M. 1993. A Hydrogeomorphic Classification for Wetlands. US Army Corps of
Engineers, W. E. Station Vicksburg, MS, lOlp.
Brunner, G.W. 2010. HEC-RAS River Analysis System User Manual Version 4.1. US Army Corps of
Engineers, H.E. Center, CPD68, 790 p.
Collins, M.J. 2009. Evidence for Changing Flood Risk in New England Since the Late 20th
Centuryl." Journal of American Water Resources Association 45(2), 279-290.
Corcoran, M.K. 2007. Geomorphic Identification and Verification of Recent Sedimentation
Patterns in the Woonasquatucket River, North Providence, Rl. US Army Corps of Engineers,
Engineer Research and Development Center, Vicksburg, MS, 138 p.
Cutter, S.L., Boruff, B.J. Shirley, W., 2003. Social Vulnerability to Environmental Hazards. Social
Science Quarterly 84(2), 242-261.
Dahl, T.E., Allord, G.J. 1996. History of Wetlands in the Conterminous United States, in: Fretwell
J.D., J.S. Williams, J.S., Redman P.J. (eds.), National Water Summary on Wetland Resources,
US Geological Survey, Water-Supply Paper, 2425,19-26.
References 61
-------
DeGaetano, A., Zarrow, D, 2011. Extreme Precipitation in New York and New England. Technical
Documentation and User Manual. Northeast Regional Climate Center, Cornell University,
Ithaca, NY, 93 p.
Doehring, D.O., Smith, M.E. 1978. Modelling the Dynamic Response of Floodplains to
Urbanization in Eastern New England. Environmental Resource Center, Colorado State
University, Fort Collins, CO. 95 p.
Duelli, P., Obrist, M.K. 2003. Biodiversity indicators: the choice of values and
measures. Agriculture, Ecosystems and Environment, 98(1), 87-98.
ESRI, 2012. ArcGIS 10.1. Environmental Systems Research Institute, Redlands, California.
Faber-Langendoen, D., Rocchio, J., Thomas, S., Kost, M., Hedge, C., Nichols, B., Walz, K.,
Kittel, G., Menard, S., Drake, J., Muldavin, E. 2012. Assessment of wetland ecosystem
condition across landscape regions: A multi-metric approach. Part B. Ecological Integrity
Assessment protocols for rapid field methods (L2). EPA/600/R-12/021b. U.S. Environmental
Protection Agency, Office of Research and Development, Washington, DC, lllp.
FEMA. 2014. National Flood Insurance Program Community Rating System, Coordinator's
Manual. FIA-15/2013. Federal Emergency Management Agency, 615 p. URL:
https://www.fema.gov/media-library/assets/documents/8768.
Flemming, M.J. 2013. Hydrologic Modeling System HEC-HMS Quick Start Guide Version 4.0. US
Army Corps of Engineers, H.E. Center. CPD-74D, 50 p.
Flemming, M.J., Doan, J.H. 2013. Geospatial Hydrologic Modeling Extension (HEC-GeoHMS)
User's Manual Version 10.1. US Army Corps of Engineers, H.E. Center. CPD-77, 193 p.
Flynn, K.M., Kirby, W.H., Hummel, P.R. 2006. User's manual for program PeakFQ, annual flood-
frequency analysis using Bulletin 17B guidelines (No. 4-B4), 52 p.
Ghermandi, A., van den Bergh, J.C.J.M., Brander, L.M., de Groot, H.L.F., Nunes, P.A.L.D. 2010.
Values of natural and human-made wetlands: A meta-analysis. Water Resources
Research, 46(12).
Golet, F.C., Myshrall, D., Miller, N., Bradley, M. 2003. Wetland restoration plan for the
Woonasquatucket river watershed, Rhode Island. Kingston, Rl: Department of Natural
Resources Science, University of Rhode Island. Final research report prepared for the
Rl Department of Environmental Management and US EPA, 161 p.
Hall, F.R. 1968. Base-Flow Recessions—A Review. Water Resources Research, 4(5), 973-983.
Hardmeyer, K., Spencer, M. 2007. Using Risk-Based Analysis and Geographic Information
Systems to Assess Flooding Problems in an Urban Watershed in Rhode Island.
Environmental Management, 39(4), 563-574.
Hey, D.L., Philippi, N.S. 1995. Flood reduction through wetland restoration: the Upper
Mississippi River Basin as a case history. Restoration Ecology, 3(1), 4-17.
Hirsch, R., Ryberg, K. 2012. Has the magnitude of floods across the USA changed with global
C02 levels? Hydrological Sciences Journal, 57(1), 1-9.
62 Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
Hodgkins, G.A. 2010. Historical Changes in Annual Peak Flows in Maine and Implications for
Flood-Frequency Analyses. U.S. Geological Survey Scientific Investigations Report
2010-5094, 38 p., at http://pubs.usgs.gov/sir/2010/5094.
Hubbard, D.E., Linder, R.L. 1986. Spring runoff retention in prairie pothole wetlands.
Journal of Soil and Water Conservation, 41(2), 122-125.
King, D.M., Price, E.W. 2004. Developing Defensible Wetland Mitigation Ratios. A Companion to
"The Five-Step Wetland Mitigation Ratio Calculator", Report prepared for NOAA Office
of Habitat Conservation, Habitat Protection Division by King and Associates, Solomons
Island, MD, 43 p.
King, D.M., Wainger, L.A., Bartoldus, C.C., Wakeley, J.S. 2000. Expanding Wetland Assessment
Procedures: Linking Indices of Wetland Function with Services and Values, Report prepared
for Wetlands Research Program. US Army Corps of Engineers, Document Number ERDC/EL
TR-00-17, 63 p.
Lane, C., D'Amico, E. 2010. Calculating the Ecosystem Service of Water Storage in Isolated
Wetlands using LiDAR in North Central Florida, USA. Wetlands, 30(5), 967-977.
Legates, D.R., McCabe, G.J. 1999. Evaluating the use of "goodness-of-fit" measures in
hydrologic and hydroclimatic model validation. Water Resources Research, 35(1), 233-241.
Mazzotta, M., Ojo, C., Bousquin, J., Hychka, K.C., Druschke, C.G., McKinney, R., Berry, W.,
in preparation. Assessing the Benefits of Restoration: A Step-By-Step Guide for
Decision Makers.
Mazzotta, M., Wainger, L., in preparation. Non-Monetary Benefit Indicators without Apology:
The Economic Theory and Practice of Ecosystem Service Benefit Indicators.
McCarthy, G.T. 1938. The Unit Hydrograph and Flood Routing. Proceedings of the Conference of
North Atlantic Division, US Army Corps of Engineers, Providence, Rhode Island, 24 p.
McCuen, R.H., 1982. A guide to hydrologic analysis using SCS methods. Prentice-Hall, Inc.
Miller, N.A., Golet, F.C. 2001. Development of a statewide freshwater wetland restoration
strategy: Site identification and prioritization methods. Kingston, Rl: Department of Natural
Resources Science, University of Rhode Island. Final research report prepared for the Rl
Department of Environmental Management and US EPA, 161 p.
Moriasi, D., Arnold, J., Van Liew, M., Bingner, R., Harmel, R., Veith, T. 2007. Model evaluation
guidelines for systematic quantification of accuracy in watershed simulations. Transactions
of the American Society of Agricultural and Biological Engineers, 50(3), 885-900.
Munns, W.R., Rea, A.W., Mazzotta, M.J., Wainger, L.A., Saterson, K. 2015. Toward a standard
lexicon for ecosystem services. Integrated Environmental Assessment and Management.
Integrated Environmental Assessment and Management. Preprint online.
Nash, J., Sutcliffe, J, 1970. River flow forecasting through conceptual models part I—A
discussion of principles. Journal of Hydrology, 10(3), 282-290.
NOAA. 2014. COOP: 376698. Url: http://www.ncdc.noaa.gov/about-ncdc (Accessed: 7-30-2014)
NOAA. 2015. Coastal Social Vulnerability Index (SoVI) dataset
http://coast.noaa.gov/dataregistrv/search/dataset/info/sovi (Accessed: 7-30-2015)
References
-------
NRCS. 1984. Computer Program for Project Formulation-Hydrology. Technical Release No. 20.
US Department of Agriculture, Natural Resources Conservation Service, Washington, DC,
330 p.
Ogawa, H., Male, J.W. 1986. Simulating the flood mitigation role of wetlands. Journal of Water
Resources Planning and Management, 112(1), 114-128.
Olander, L., Johnston, R.J., Tallis, H., Kagan, J., Maguire, L., Polasky, S., Urban, D., Boyd, J.,
Wainger, L., Palmer, M. 2015. Best Practices for Integrating Ecosystem Services into Federal
Decision Making. National Ecosystem Services Partnership, Duke University, Durham, NC,
54 p.
Oliver, F., Maidment, D.R. 2000. GIS tools for HMS modeling support. Hydrologic and hydraulic
modeling support with geographic information systems. Environmental Systems Research
Institute Press, USA, 85-112.
Peterson, T.C., Karl, T.R., Kossin, J.P., Kunkel, K.E., Lawrimore, J.H., McMahon, J.R., Vose, R.S.,
Yin, X. 2014. Changes in weather and climate extremes: State of knowledge relevant to air
and water quality in the United States. Journal of the Air and Waste Management
Association, 64(2), 184-197.
Potschin, Haines-Young. 2011. Ecosystem services: Exploring a geographical perspective.
Progress in Physical Geography, 35(5), 575-594.
Qaiser, K., Yuan, Y., Lopez, R. 2012. Urbanization Impacts on Flooding in the Kansas River Basin
and Evaluation of Wetlands as a Mitigation Measure. Transactions of the American Society
of Agricultural and Biological Engineers, 55(3), 849-859.
Richardson, L., Loomis, J., Kroeger, T., Casey, F. 2015. The role of benefit transfer in ecosystem
service valuation. Ecological Economics, 115, 51-58.
Rhode Island Statewide Planning Program. 2006. Land Use 2025: Rhode Island State Land Use
Policies and Plan. Providence, Rl: Rhode Island Department of Administration. 154 p.
RIGIS. 1993. Wetlands (s44wwt93). Rhode Island Geographic Information System (RIGIS) Data
Distribution System, URL:http://www.edc.uri.edu/rigis, Environmental Data Center,
University of Rhode Island, Kingston, Rhode Island (Accessed: 7-20-2011).
RIGIS. 2010. Land Use 2025; Ianduse2025. Rhode Island Geographic Information System
(RIGIS) Data Distribution System, URL:http://www.edc.uri.edu/rigis, Environmental Data
Center, University of Rhode Island, Kingston, Rhode Island (Accessed: 3-16-2010).
RIGIS, 2013. Digital Elevation Model; DEM11. Rhode Island Geographic Information System
(RIGIS) Data Distribution System, URL:http://www.edc.uri.edu/rigis, Environmental Data
Center, University of Rhode Island, Kingston, Rhode Island (Accessed: 7-25-2013).
RIGIS. 2014. Rl E-911 Sites (e911Sitesl4r3). Rhode Island Geographic Information System
(RIGIS) Data Distribution System, URL:http://www.edc.uri.edu/rigis, Environmental Data
Center, University of Rhode Island, Kingston, Rhode Island. Accessed: 2-23 -2015.
Soil Conservation Service. 1985. National Engineering Handbook. Section 4: Hydrology. US
Department of Agriculture: Washington DC, 30 p.
Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
Sherrouse, B.C., Semmens, D.J. 2010. Sot jes for Ecosystem Services (SolVES)-Us to
include social values information in ecosystem services assessments. US Geological Survey
Fact Sheet 2010-3118, 2 p. (also see http://solves.cr.usgs.eov/publications.shtml for other
publications).
Smith, J.A., Baeck, M.L., Villarini, G., Krajewski, W.F. 2010. The hydrology and hydro-
meteorology of flooding in the Delaware River Basin. Journal of Hydrometeorology, 11(4),
841-859.
Turner, R.K., van den Bergh, J.C.J.M., Soderqvist, T., Barendregt, A., van der Straaten, J., Maltby,
E., van lerland, E.C. 2000. Ecological-economic analysis of wetlands: scientific integration for
management and policy. Ecological Economics 35(1), 7-23.
USGS. 2014. Instantaneous Data Archive (IDA). US Geological Survey.
http://ida.water.usgs.gov/ida/available_records.cfm?sn=01114500 (Accessed: 7-30-2014)
Vallee, D., Giuliano L., 2014. Overview of a Changing Climate in Rhode Island. Developed for the
storm guidance working group of the Rl executive climate change coordinating council. 7 p.
Villa. F,, Bagstad, KJ \ B.. Johnson. G.W.. Portela. R., Honzak, M.. Batker, D. 2014.
A methodology for adaptable and robust ecosystem services assessment. PLoS ONE
9(3):e91001.
Villarini, G., Smith, J.A. 2010. Flood peak distributions for the eastern United States. Water
Resources Research, 46(6).
Ward, A.D., Trimble, S.W., 2003. Environmental hydrology. CRC Press.
Williams, J.R., LaSeur, W.V. 1976. Water yield model using SCS curve numbers. Journal of the
Hydraulics Division, 102(9), 1241-1253.
Winter, T.C. 1999. Relation of Streams, Lakes, and Wetlands to Groundwater Flow Systems.
Hydrogeology Journal 7.1 (1999): 28-45.
Woodward, D., Hawkins, R., Jiang, R., Hjelmfelt, Jr., A. Van mullem, J., Quan Q. 2003. Runoff
Curve Number Method: Examination of the Initial Abstraction Ratio. World Water and
Environmental Resources Congress 2003. June 2003, 1-10.
Yuan, Y., Qaiser, K., 2011. Floodplain Modeling in the Kansas River Basin Using Hydrologic
Engineering Center Models: Impacts of Urbanization and Wetlands for Mitigation. US EPA,
Las Vegas, NV. EPA/600/R-11/116, 42 p.
Zarriello, P., Ahearn, E., Levin, S. 2010. Magnitude of Flood Flows for Selected Annual
Exceedance probabilities in Rhode Island Through 2010. US Geological Survey. Scientific
Investigations Report 2012-5109, 94 p.
Zarriello, P., Straub, D., Westenbroek, S. 2013. Simulated and Observed 2010 Flood-Water
Elevations in Selected River Reaches in the Moshassuck and Woonasquatucket River basins,
Rhode Island. US Geological Survey. Scientific Investigations Report 2013-5191, 35 p.
References
-------
-------
APPENDICES
Appendix A-Wetland Flood Protection Functional Assessment
Functional indicators from Miller and Golet (2001) were incorporated as model inputs, to
demonstrate how Tier III benefits indicators derived from the model results can be used as an
extension to existing functional assessments.
Function
Criteria -r
CO
UJ
o
Soiree
Hores
Flood
Impervious surfaces cover > 20% of land within 500 feet
0
L
Abatement
Slopes within 500 feet of wetland are > 15%
0
L
Point-source inflow
0
L,F
Bordering or containing a lower perennial stream
0
L,F
E
F
E
L,F
1,2
Constricted outlet
E
L,F
Developed flood-prone areas downstream within 5 miles or to
nearest dam (connection by stream or floodway required)
S
L
*Mark each box as Y, N, D, or NA (i.e., yes, no, don't know, or not applicable)
+0 = opportunity; E = effectiveness; S = social significance
tL = lab data; F = field data
^ot applicable if entire wetland unit has been destroyed.
2Not applicable if the wetland types of the existing unit and the destroyed portion are different.
Appendices
-------
Appendix IB - Wetland Parameters
Wetland short-term surface water retention was represented in HEC-HMS as diversions—or
subbasin flow that is diverted from entering the main channel. Model flow diverted from a
given subbasin is a factor of (1) the rate at which subbasin outflow is diverted and (2) the
maximum cumulative volume of water that can be diverted. This appendix details the process
used to quantify these two diversion parameters from spatial wetland data (Figure B-l).
Parameters for wetlands in the HEC-HMS model included the maximum volume that could be
diverted to the wetland (Table B-l, Avg. Vol.) and percent of subbasin flow diverted (Table B-l,
% Wetland). Percent of subbasin flow diverted was based on the surface area of wetlands
(Table B-l, Area of wetlands) as a percent of total basin area (Table B-l, % Wetland). This
percentage informed the diversion's Paired data table (Table B-2 demonstrates the diversion
D_W1570 for subbasin W1570), which HEC-HMS used to generate a curve for the Inflow-
Diversion function. When total volume diverted to a wetland reaches the maximum volume,
flow is no longer diverted from subbasin outflow (Figure B-2).
A Python toolbox in ArcMap 10.2 was used to estimate potential maximum volume retention.
The tool calculates volume based on surface elevations as bottom contour, the minimum or
average elevation from perimeter vertices as the water surface elevation for height, and areal
extent based on wetland polygons. Surface elevations were taken from the LiDAR-derived
Digital Elevation Model (DEM). LiDAR returns bounce off standing water, meaning the DEM
identifies the bottom of the wetland as the elevation of surface water in the wetland at the
time of survey (April 22-May 6, 2011). Any water detected by LiDAR was considered the
residual storage in wetlands before flood events, meaning wetland storage was not adjusted in
any of the scenarios for more or less residual water.
The water surface elevation for each wetland was determined using the Lane and D'Amico
(2010) method where water surface elevation in the wetland is estimated based on the average
elevation at vertices around the wetland polygon perimeter. Adjacent wetland polygons were
combined and treated as one continuous wetland. This eliminated doughnut or stepped water
surfaces formed where interior or upstream wetlands would otherwise have a lower water
surface elevation than exterior or downstream wetlands. Although the toolbox allows for
volumes to be separated based on wetland type, we did not use this information. Each basin
may still contain multiple wetlands (Table B-l, No. provides the original count of wetlands in
that basin). Surface elevations were attributed to each wetland and then wetlands were divided
into the subbasins before volume was calculated. This ensured water surface elevations were
consistent for each wetland but volumes were calculated only for the area of the wetland inside
the subbasin where that volume was assigned. Wetland volumes (m3) were calculated for each
subbasin based on both the minimum (Table B-l, Min. Vol.) and average (Table B-l, Avg. Vol.)
water surface elevation.
Sensitivity analysis of wetland parameters included the comparison of methods used to
calculate wetland maximum diversion volume and percent subbasin flow diverted. To compare
68 Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
methods, the model was run using volumes calculated based on both minimum water surface
elevations and average water surface elevations. Sensitivity analysis of percent flow diverted to
increased wetland catchment area was modeled by shifting non-wetland area to wetland area
in each subbasin by 10%, 25%, 50%, and 75%.
To better understand our conservative assumption of wetland catchment size (only the areal
extent of wetlands), the areal extent of wetlands plus a 100-m buffer was also calculated
(Figure B-l). Following this assumption, a catchment defined by areal extent plus a 100-m
buffer would increase the average percentage diverted by 51% (Table B-l, % Wetland or Buffer).
Similarly, restoration scenarios converted a percentage of non-wetland area in each subbasin to
wetland. An example of the percent subbasin flow diverted from the 25% scenario is given in
Table B-l, 25 % Restoration Scenario % Wetland, along with the surface area of wetlands that
would be added (Table B-l, Increase in Wetlands Area).
Table B-l: Subbasin wetland parameters and characteristics related to their sensitivity analysis
(subbasin names; number of wetlands in each subbasin; minimum volume (m3); average volume (m3);
surface area of wetlands (m2); percent area of subbasin in wetlands; increase in wetland area in 25%
Restoration Scenario (m2); percent area of subbasin in wetlands in 25% Restoration Scenario; and
percent of subbasin area in wetlands or 100-m buffer around wetlands).
25% Restoration Scenario
Subbasin
No. of
Min.Vol. Avg. Vol.
Wetland Area
%
Increase in
%
W1260
21
235
57644
193663
12%
342450
34%
70%
W1270
8
914
70937
97075
14%
144361
36%
69%
W1280
11
4
19809
38937
11%
82532
33%
76%
W1290
13
1154
126771
294886
24%
227590
43%
77%
W1300
11
3289
381545
225733
20%
229077
40%
70%
W1320
12
513
84013
128685
17%
162337
37%
75%
W1330
7
0
49754
53526
10%
125513
32%
59%
W1340
3
51
43557
47238
52%
11004
64%
100%
W1350
3
11
3170
2041
61%
332
70%
100%
W1360
3
96
17508
17023
50%
4220
63%
100%
W1370
9
119
211908
246137
24%
197529
43%
72%
W1380
10
1404
289029
156578
19%
167333
39%
87%
W1390
13
2966
711624
191622
17%
229178
38%
80%
W1400
9
683
148042
127818
17%
161110
37%
80%
W1410
1
0
154
130
11%
258
33%
100%
W1420
10
1150
140516
365574
25%
274747
44%
76%
W1430
18
3170
158881
314003
15%
457938
36%
60%
W1440
16
4158
353654
402552
25%
306405
44%
80%
W1450
7
1842
152580
121413.82
15%
168712
36%
86%
W1460
4
31
18098
44694.76
13%
73768
35%
66%
Appendices
-------
25% Restoration Scenario
Subbasin
No. of
Min.Vol. Avg. Vol.
Wetland Area
%
Increase in
%
W1470
11
573
747595
243987.75
19%
252717
40%
88%
W1480
11
3902
642571
334141.92
18%
376178
39%
75%
W1490
8
67
25634
72026.12
9%
176692
32%
56%
W1500
16
4282
305844
98652.97
7%
318029
30%
61%
W1510
2
6
22354
5240.58
3%
41856
27%
49%
W1520
11
774
28637
67077.94
7%
211334
31%
70%
W1530
7
33
7533
50521.56
10%
114884
32%
66%
W1540
15
1239
272703
281186.94
15%
400508
36%
65%
W1550
6
0
654556
195225.45
24%
155950
43%
66%
W1560
14
1843
155684
230047.00
19%
242986
39%
86%
W1570
6
0
179516
123019.13
21%
113302
41%
84%
W1580
1
0
23399
10990.35
56%
2172
67%
100%
W1590
3
0
89714
95334.85
17%
113859
38%
71%
W1600
4
0
145882
199681.28
27%
136066
45%
66%
W1610
12
396
552908
332883.06
25%
253966
44%
81%
W1620
1
0
91312
74880.77
37%
31441
53%
81%
W1640
2
0
92534
11997.66
15%
16898
36%
70%
W1650
3
4
54294
72873.22
34%
34758
51%
84%
W1660
2
0
2917
28589.77
6%
122138
29%
41%
W1670
14
4381
629244
182537.18
15%
268474
36%
74%
W1680
6
113
292082
143675.38
18%
160362
39%
72%
W1690
8
10132
554300
309140.92
24%
251226
43%
89%
W1700
5
0
225748
274304.30
37%
114993
53%
87%
W1710
7
38
48143
63010.56
8%
183345
31%
53%
W1720
5
0
304838
55524.59
26%
40157
44%
95%
W1750
6
0
149209
249967.08
36%
108944
52%
78%
W1770
14
3458
351952
375347.15
20%
376688
40%
55%
W1780
2
0
66699
200640.02
27%
135027
45%
61%
W1790
12
1380
781163
349811.83
19%
364828
40%
67%
W1800
17
19167
320454
228021.39
15%
320136
36%
80%
W1810
8
1060
554893
245656.43
15%
343687
36%
55%
W1820
3
0
169474
66012.48
9%
171945
32%
38%
W1830
3
0
468
4181.93
1%
198018
25%
20%
W1840
13
728
157789
201449.01
25%
152306
44%
94%
W1850
1
0
19853
1474.22
40%
551
55%
100%
W1860
3
0
130707
147106.76
27%
97173
46%
87%
W1870
10
260
90006
112974.04
16%
148549
37%
80%
W1880
7
108
640710
79077.85
19%
86093
39%
90%
W1890
11
175
239144
147579.03
17%
182883
38%
75%
W1900
1
0
14850
24095.52
23%
20296
42%
80%
Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
25% Restoration Scenario
Subbasin
No. of
Min.Vol. Avg. Vol.
Wetland Area
%
Increase in
%
W1910
9
2288
17060
52259.91
15%
75369
36%
88%
W1920
1
0
3043
2913.63
66%
377
74%
100%
W1930
9
20
50590
41680.26
5%
218165
28%
51%
W1940
1
0
41420
64608.43
21%
62376
40%
54%
W1950
1
0
12398
9604.70
49%
2458
62%
100%
W1960
4
41
200861
158143.22
20%
160850
40%
78%
W1970
1
0
124231
123065.50
51%
29227
63%
85%
W2000
11
773
38571
62813.93
6%
239331
30%
53%
W2010
7
9860
432051
251801.56
13%
410642
35%
57%
W2020
1
0
26544
24396.66
47%
6929
60%
87%
W2030
6
29
37534
181523.45
14%
279230
35%
43%
W2040
4
614
59862
51956.44
10%
120197
32%
54%
W2050
9
2548
470700
124138.52
18%
144305
38%
81%
W2060
10
3493
94479
107295.19
10%
232401
33%
63%
W2070
6
69
83776
122890.47
19%
130319
39%
80%
W2090
12
3932
235553
157307.53
17%
187053
38%
79%
W2100
1
0
0
1054.41
1%
17795
26%
41%
W2110
2
0
17375
16093.27
29%
9750
47%
93%
W2130
12
5022
128521
188634.27
13%
315597
35%
49%
W2140
6
489
105906
155634.15
29%
93281
47%
79%
W2150
7
2755
98471
102372.58
15%
143018
36%
77%
W2160
11
479
32317
80443.34
11%
171264
33%
71%
W2170
3
2
18826
79644.14
10%
173033
33%
46%
W2180
1
0
171989
189351.28
69%
21255
77%
97%
W2190
1
0
142670
145424.19
87%
5633
90%
100%
W2200
3
387
339594
215897.11
20%
210022
40%
53%
W2210
4
0
15209
60515.13
11%
117141
34%
55%
W2220
7
141
71316
117015.17
12%
207448
34%
59%
W2230
1
0
195137
274283.28
70%
28831
78%
99%
W2240
2
9
12486
26674.51
25%
19524
44%
100%
W2250
7
134
157678
265799.66
26%
191836
44%
78%
W2260
10
6
204971
238941.99
26%
174472
44%
89%
W2290
5
428
11695
9762.75
2%
137027
26%
38%
W2300
1
0
20962
12497.82
7%
43861
30%
67%
W2310
12
314
62571
104856.52
9%
273492
32%
60%
W2320
8
85
133593
66027.31
10%
149844
32%
75%
W2330
10
12
36466
76387.15
7%
257761
30%
61%
W2340
5
13
85093
69026.59
7%
216849
31%
47%
W2350
1
0
97976
46119.13
16%
60070
37%
62%
W2360
2
0
5022
17454.86
4%
118811
28%
35%
Appendices
-------
25% Restoration Scenario
Subbasin
No. of
Min.Vol. Avg. Vol.
Wetland Area
%
Increase in
%
W2370
11
39
244336
225732.24
13%
382694
35%
58%
W2380
3
1295
70701
76911.96
13%
130673
35%
60%
W2390
1
11
24842
9700.12
47%
2768
60%
100%
W2400
1
0
65559
100393.48
82%
5496
87%
100%
W2410
9
428
227016
391316.74
27%
264586
45%
75%
W2430
4
60
95257
24699.08
4%
131646
28%
43%
W2440
1
0
9257
4500.30
59%
794
69%
100%
W2450
11
1535
109888
88025.92
11%
183230
33%
68%
W2470
3
11
26671
144754.12
26%
101132
45%
69%
W2480
1
0
9174
23838.05
37%
10130
53%
79%
W2490
15
7241
406071
394654.75
19%
431940
39%
63%
W2500
3
2
260337
105841.56
19%
110475
39%
78%
W2520
4
111
8053
11134.11
8%
30910
31%
83%
W2530
1
10
68243
96882.10
19%
101589
39%
75%
W2570
15
882
19439
56768.31
7%
183290
30%
78%
W2580
19
1416
178303
248633.32
19%
273789
39%
84%
W2630
1
51
38474
11000.56
5%
52017
29%
42%
W2670
4
185
15421
20729.86
15%
29980
36%
71%
W2680
1
0
8909
14271.67
21%
13266
41%
77%
W2720
5
97
93044
40297.99
10%
93781
32%
67%
W2770
2
0
279262
68439.70
16%
91035
37%
58%
W2780
6
7
52577
111963.67
14%
170537
36%
66%
W2830
11
386
363106
349482.82
27%
238317
45%
76%
W2870
5
14
273007
114541.26
20%
117037
40%
76%
W2880
2
0
14944
20832.01
10%
46984
32%
44%
W2920
4
712
27457
69361.67
10%
163515
32%
37%
W2930
2
0
10591
61538.51
21%
59622
40%
70%
W2970
2
39
34263
34170.66
27%
22554
46%
100%
W2980
2
0
124
7684.74
6%
29006
30%
59%
W3030
1
0
8414
18645.33
41%
6822
55%
97%
W3070
4
126
14948
24423.18
39%
9512
54%
96%
W3080
5
0
8064
47338.45
12%
88460
34%
70%
W3120
7
1
79033
48285.28
3%
390897
27%
25%
W3130
7
21
58498
74560.22
15%
109924
36%
69%
W3180
17
918
162826
303998.12
18%
351670
38%
70%
W3220
9
21
111732
64923.78
13%
104917
35%
78%
W3230
9
46
56079
50818.03
12%
92895
34%
77%
W3240
8
1558
95405
90992.84
9%
224448
32%
60%
W3250
3
52
84289
44922.39
16%
60700
37%
53%
Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
Table B-2: Diversion D_W2980 paired data table
Inflow
Diversion
(m3/s)
(m3/s)
0
0
1
0.062
100
6.212
100000
6212
] Subbasins
Existing Wetlands
100-m Buffer
Non-wetland
0 0.75 1.5 3 4.5 6
I Kilometers
Figure B-l: Modeled area showing wetlands area (blue) a 100m buffer around wetlands (green) and
the remaining non-wetlands area (yellow). Black lines show the borders of individual subbasins.
Appendices
-------
W2980 Flow Diverted to Wetland
— W2980 Flow After Diversion
W2980 Inflow
0.10-
0.03"
GI 0.04-
0.02-
0.00
Figure B-2: Inflow, flow diverted to wetlands, and flow after diversion (m3/s) for subbasin W2980. When
the total diverted to Diversion D_W2980 reaches the maximum, flow diversion (blue) goes to zero and
subbasin Inflow (green) equals subbasin outflow (red).
Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
Appendix C - PeakFQ Analysis
The PeakFQ program (Flynn et al. 2006) was used to analyze observed peak flows for the
Woonasquatucket watershed (Figure C-l systematic peaks), estimate recurrence intervals for
storms modeled (Table C-l), and relate flows from synthetic storms to flows of standard
probabilities (Table C-2). All available annual peak stream flows for the Woonasquatucket USGS
gage (USGS #01114500) from 1936 to 2014 were used for a total of 73 data points (Figure C-l).
Results from PeakFQ include a frequency fitted to the observed peaks and upper and lower
uncertainty bounds for those peaks (Figure C-2). A linear regression of the observed peak flows
(Figure C-l) showed an average increase of 46 cfs each decade, though the trend is weak due to
high annual variability (R2= 0.0958).
; c
,,,, i,,,,,,,,,
Systematic Peaks
a
a
....
o
¦n a
....
* o
1 1 1 1
o
....
c
o
o
....
1 1 1 1
- n a a oo
o °o°
oo ^
° o° ° o 0
Oo ° °
° o
' 9?
}<& o o
°o
~ O O
°o
o
o
=1
o
o
....
. . . ,
....
....
....
....
....
....
....
1330 154C 1350 1360 1370 1380 1S60 2Q0C 2010 2020 2D30
Wat* Yasr
Station -01114500
Figure C-l: All annual peak discharge values (cfs) for water years at USGS station #011145000.
The general trend in annual peak flows over time increased an average of 46 cfs each decade
(y = 4.6588X -8538.9, R2 = 0.0958).
Appendices
-------
10,000
Fitted frequency
o Systematic Peaks
Confidence limits
cr:
2
—
u
l/i
5
JL
ra
c
0.
1,000
2
3
c
Peakfq v 7.1 run 3/19/2015 1:50:19 PM
B17B using Weighted Skew option
0.0545 = Skew (G)
0 Zeroes not displayed
0 Peaks below PILF(LO) Threshold
Single Grubbs-Beck
100
99.5
98
95
90
80
70
50
30
20
10
5
2
0.5
0.2
Annual Exceedance Probability, Percent
Station -01114500 WOONASQUATUCKET RIVER AT CENTERDALE, Rl
Figure C-2: Annual exceedance probabilities for all observed (systematic) peaks, a curve fit to those
probabilities (red), and its upper and lower confidence intervals (blue).
Table C-l: Probability and recurrence intervals for the storm of record and the four storms considered
for use in hydrologic modeling
Discharge
Probability
Recurrence
Rank
Date
Use
(cfs)
(B17B)
(Years)
1
Mar. 30, 2010
Not Used: Missing data
1810
0.0135
74.1
2
Oct. 15, 2005
Calibration Storm 1
1530
0.0270
37.0
3
Mar. 22, 2001
Calibration Storm 2
1070
0.1351
7.4
4
Apr. 03, 2005
Validation Storm 1
943
0.1892
5.3
6
Apr. 16, 2007
Validation Storm 2
851
0.2973
3.4
Table C-2: Expected discharge (cfs) required for peak flows of standard probabilities
Probability
Recurrence
B17B
Record
Lower
Upper
0.002
500
2621.0
2371.0
2151.0
3377.0
0.005
200
2239.0
2072.0
1868.0
2821.0
0.010
100
1968.0
1851.0
1664.0
2434.0
0.020
50
1710.0
1635.0
1466.0
2074.0
0.040
25
1463.0
1422.0
1273.0
1739.0
0.100
10
1152.0
1142.0
1023.0
1330.0
0.200
5
922.4
926.0
830.8
1041.0
0.500
2
605.8
614.2
550.0
667.0
0.995
1.005
174.0
161.0
139.1
207.4
Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
Appendix ID HUMS Parameters
Table D-l: Spatial data used to produce variables and parameterize HMS model
Name
Type
Creator
Date
Resolution
Availability
Dams
Dams
Inventory
Rl DEM Division of
Compliance & Inspection's
Dam Safety Program
2000
NA
Statewide
LiDAR DEM (lm):
Statewide Spring 2011
Digital
Elevation
Model
URI Environmental Data
Center
2011
1 m
Statewide
Impervious Surfaces
2003-2004
Impervious
Surfaces
Sanborn
2003-
2004
2ft
Statewide
Land Use- 2003/2004
Land
Use/Land
Cover
Sanborn Map Company,
State of Rl
2007
minimum
mapping unit
of 0.5 acre
Statewide
National Hydrography
Dataset Plus (Version 2)
Flowlines
Streams
USGS and Horizon Systems
2012
1:100,000
Most of the
continental
U.S.
5K Streams (1:5000)
Streams
RIDOT
1997
1:5,000
Statewide
Wetlands
Wetlands
IEP inc
1993
1km2
(0.25 acre)
Statewide
Curve Number Loss Method Parameters
At the subbasin scale, we used the loss method in HEC-HMS to generate runoff volume based
on the Soil Conservation Service Curve (SCS) Curve Number (CN) approach and the percent
impervious surface in the subbasins (Table D-2). The SCS method requires a grid of CN values,
based on soil type and land cover (Table D-l), where land cover types are assigned CN values
(USDA 1986). The average basin CN value was 67 and the average subbasin percent
imperviousness was 12% (averages not weighted by basin area). The transform method uses
the SCS Unit Hydrograph with the lag time calculated using TR-55 method (McCuen 1982). We
estimated parameters for the TR-55 method, including impervious surface area and Manning's
N for sheet flow. Parameters left as default values were: cross-sectional flow area (2 m2),
wetted perimeter (6 m), and Manning's N for channel flow (0.03). Basin slopes for the TR-55
method were determined using the original DEM rather than the filled DEM, meaning
occasionally channel slope (28/139 instances) and watercourse slope (4/139 instances) were
<0. For many of these subbasins slope was minimal because the channel passed through a
reservoir. The slopes for these channels and watercourses were adjusted to 0.0001 to reflect a
flat slope while still allowing for TR-55 calculations.
Appendices
-------
Table D-2: Subbasin parameters; % impervious, Curve Number (CN) values directly from HEC-GeoHMS
based on AMC II, after machine calibration to the 2005 Fall storm, after machine calibration to the
2001 Spring storm, based on the Dry AMC I method, and based on the Wet AMC III method
Subbasin
Impervious
Initial CN
'05 CN
'01 CN
Dry CN
Wet CN
Name
(%)
(AMC II)
Calibration
Calibration
(AMC 1)
(AMC III)
W1260
6.1%
59.4
40.0
69.0
39.8
77.3
W1270
7.6%
64.1
43.1
74.4
42.9
83.3
W1280
4.6%
60.6
40.7
70.3
40.6
78.7
W1290
4.8%
60.4
40.6
70.1
40.5
78.5
W1300
3.7%
57.2
38.4
66.3
38.3
74.3
W1320
4.8%
62.3
41.9
72.3
41.7
81.0
W1330
9.0%
64.2
43.2
74.5
43.0
83.4
W1340
7.4%
68.7
46.2
79.7
50.1
83.1
W1350
1.2%
71.1
47.8
82.6
51.9
86.1
W1360
0.4%
69.3
46.6
80.4
50.6
83.8
W1370
4.8%
70.3
47.3
81.6
51.3
85.0
W1380
5.6%
61.3
41.2
71.2
41.1
79.7
W1390
3.1%
58.8
39.5
68.2
39.4
76.4
W1400
6.7%
67.9
45.7
78.9
49.6
82.2
W1410
32.9%
82.2
55.3
95.4
64.9
93.7
W1420
0.8%
56.9
38.3
66.1
38.1
74.0
W1430
8.5%
62.9
42.3
73.0
42.1
81.8
W1440
2.9%
67.9
45.6
78.8
49.5
82.1
W1450
1.6%
60.4
40.6
70.2
40.5
78.6
W1460
6.8%
63.0
42.4
73.1
42.2
81.9
W1470
3.0%
63.9
43.0
74.2
42.8
83.0
W1480
2.2%
63.5
42.7
73.7
42.6
82.6
W1490
4.2%
60.7
40.8
70.5
40.7
78.9
W1500
6.4%
61.0
41.0
70.9
40.9
79.3
W1510
4.3%
56.8
38.2
66.0
38.1
73.9
W1520
16.0%
68.7
46.2
79.8
50.2
83.2
W1530
1.1%
59.6
40.1
69.1
39.9
77.4
W1540
13.9%
71.0
47.7
82.4
51.8
85.9
W1550
13.7%
71.3
48.0
82.8
52.1
86.3
W1560
8.4%
66.1
44.5
76.7
48.3
80.0
W1570
21.2%
76.0
51.1
88.2
60.0
86.6
W1580
0.0%
55.9
37.6
64.9
37.5
72.7
W1590
0.0%
65.0
43.7
75.5
47.5
78.7
W1600
14.9%
76.9
51.7
89.3
60.7
87.7
W1610
4.1%
62.6
42.1
72.6
41.9
81.3
W1620
11.9%
73.8
49.6
85.7
53.9
89.3
W1640
6.2%
56.3
37.9
65.4
37.7
73.2
W1650
22.7%
77.4
52.1
89.9
61.2
88.2
W1660
16.4%
64.3
43.2
74.6
43.1
83.5
W1670
5.2%
65.8
44.3
76.4
48.0
79.6
W1680
8.2%
75.6
50.8
87.8
59.7
86.2
W1690
2.3%
63.3
42.6
73.5
42.4
82.3
W1700
7.5%
71.4
48.0
82.9
52.1
86.4
W1710
16.7%
67.0
45.0
77.7
48.9
81.0
W1720
8.1%
66.7
44.9
77.5
48.7
80.7
Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
Subbasin
Impervious
Initial CN
'05 CN
'01 CN
Dry CN
Wet CN
Name
(%)
(AMC II)
Calibration
Calibration
(AMC 1)
(AMC III)
W1750
7.3%
72.9
49.0
84.6
53.2
88.2
W1770
11.0%
67.1
45.1
77.9
49.0
81.1
W1780
14.0%
67.6
45.4
78.4
49.3
81.7
W1790
9.9%
66.9
45.0
77.7
48.9
81.0
W1800
4.8%
67.3
45.2
78.1
49.1
81.4
W1810
1.2%
58.5
39.3
67.9
39.2
76.1
W1820
13.0%
64.6
43.5
75.0
43.3
84.0
W1830
15.4%
70.3
47.3
81.7
51.3
85.1
W1840
2.9%
64.0
43.1
74.4
42.9
83.3
W1850
0.0%
70.9
47.7
82.3
51.8
85.8
W1860
5.2%
63.7
42.8
73.9
42.7
82.8
W1870
4.9%
67.1
45.1
77.9
49.0
81.2
W1880
0.2%
60.6
40.8
70.4
40.6
78.8
W1890
13.1%
68.5
46.1
79.5
50.0
82.9
W1900
9.8%
63.0
42.4
73.1
42.2
81.9
Muskingum Routing Method Parameters
At the reach-scale, the hydrologic routing of flows followed the Muskingum approach. In this
approach, the travel time in the reach (K) is estimated using lag time. Reach K (hrs) values were
calculated using K = L/3600V, where L = reach length (m) and V = reach velocity (m/s) (Olivera
and Maidment 2000). Reach length was already available, and channel average velocity from
the TR-55 method was used for reach velocity. The average Reach K value of reaches included
in the HEC-HMS model was 0.446 hours. Attenuation within the reach is accounted for through
a constant (X) which ranges from 0 to 0.5, with 0 meaning storage is controlled by outflow (full
attenuation), and 0.5 giving equal weighting to inflow and outflow (no attenuation). Reach X
values were set to a default of 0.2, except where reaches overlapped large bodies of open
water, where Reach X was set to 0.4 to represent increased storage. We estimated the
minimum and maximum number of reaches based on: the length of the river, the reach velocity
(from the TR-55 method), and the one minute time step of the model (Olivera and Maidment
2000). This method allows for the time it takes for water to move through a reach to be less
than the time step of the model. We also assumed no baseflow from each subbasin, instead
adjusting our gage data within HEC-HMS to reflect baseflow.
Table D-3: Reach parameters; Initial Muskingum K and X values, Muskingum values after machine
calibration to the 2005 storm, Muskingum values after machine calibration to the 2001 storm
Reach
Initial
'05 Calibration
'01 Calibration
K
X
K
X
K
X
R100
0.13
0.2
0.16
0.23
0.20
0.25
R1010
0.27
0.2
0.30
0.23
0.38
0.25
R1040
0.21
0.2
0.24
0.23
0.32
0.25
R1050
0.30
0.2
0.33
0.23
0.37
0.25
R1060
0.14
0.2
0.17
0.23
0.22
0.25
Appendices 79
-------
Initial
'05 Calibration
'01 Calibration
Reach
K
X
K
X
K
X
R1070
0.11
0.2
0.14
0.23
0.20
0.25
R1090
0.02
0.4
0.05
0.43
0.12
0.46
R110
0.21
0.2
0.24
0.23
0.32
0.25
R1100
0.97
0.2
1.00
0.23
1.06
0.25
R1130
1.09
0.2
1.12
0.23
1.14
0.26
R1150
0.03
0.2
0.06
0.23
0.12
0.25
R1160
0.44
0.4
0.47
0.43
0.53
0.46
R1200
0.72
0.2
0.75
0.23
0.81
0.25
R1220
0.37
0.2
0.40
0.23
0.43
0.25
R1240
0.33
0.2
0.36
0.23
0.43
0.25
R130
0.65
0.2
0.68
0.23
0.73
0.25
R160
0.02
0.2
0.05
0.23
0.12
0.25
R190
0.08
0.2
0.11
0.23
0.18
0.25
R20
0.23
0.2
0.26
0.23
0.34
0.25
R200
0.05
0.2
0.08
0.23
0.15
0.25
R2550
1.97
0.4
2.00
0.43
2.06
0.46
R260
0.19
0.2
0.22
0.23
0.29
0.25
R2640
0.30
0.4
0.33
0.43
0.39
0.46
R2690
0.11
0.4
0.14
0.43
0.21
0.46
R2750
0.17
0.2
0.20
0.23
0.28
0.25
R2800
0.17
0.2
0.20
0.23
0.28
0.25
R2890
0.22
0.2
0.25
0.23
0.34
0.25
R2940
1.08
0.2
1.11
0.23
1.16
0.25
R2990
0.11
0.2
0.14
0.23
0.20
0.25
R300
0.13
0.2
0.16
0.23
0.24
0.25
R3040
0.37
0.2
0.40
0.23
0.44
0.25
R3090
0.35
0.2
0.38
0.23
0.42
0.25
R3140
0.04
0.2
0.08
0.23
0.13
0.25
R3200
1.45
0.4
1.48
0.43
1.52
0.45
R330
0.76
0.4
0.79
0.43
0.84
0.46
R340
0.23
0.2
0.26
0.23
0.34
0.25
R350
0.06
0.4
0.09
0.43
0.15
0.46
R380
0.52
0.4
0.55
0.43
0.59
0.46
R400
0.10
0.2
0.13
0.23
0.21
0.25
R410
0.15
0.2
0.18
0.23
0.27
0.25
R420
1.51
0.4
1.57
0.43
1.61
0.46
R430
1.53
0.4
0.20
0.43
0.29
0.46
R460
0.30
0.4
0.33
0.43
0.39
0.46
R470
0.11
0.4
0.14
0.43
0.19
0.46
R480
0.57
0.2
0.60
0.23
0.67
0.25
Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
Initial
'05 Calibration
'01 Calibration
Reach
K
X
K
X
K
X
R510
0.02
0.2
0.05
0.23
0.12
0.25
R530
0.55
0.2
0.58
0.23
0.64
0.25
R560
0.09
0.4
0.12
0.43
0.21
0.46
R580
0.02
0.3
0.05
0.30
0.14
0.36
R610
0.12
0.2
0.15
0.23
0.21
0.25
R620
0.18
0.2
0.21
0.23
0.29
0.25
R630
0.04
0.2
0.07
0.23
0.14
0.25
R640
0.93
0.4
0.96
0.43
1.05
0.46
R650
0.28
0.2
0.31
0.23
0.40
0.25
R660
2.90
0.28
2.93
0.28
2.98
0.33
R690
0.82
0.4
0.86
0.43
0.96
0.46
R70
0.42
0.2
0.45
0.23
0.49
0.25
R700
0.09
0.2
0.13
0.23
0.21
0.25
R710
0.02
0.2
0.05
0.23
0.15
0.25
R720
0.38
0.4
0.41
0.43
0.45
0.46
R770
0.21
0.2
0.24
0.23
0.35
0.25
R790
0.20
0.4
0.23
0.43
0.31
0.46
R80
0.02
0.2
0.05
0.23
0.22
0.25
R810
0.77
0.4
0.80
0.43
0.86
0.46
R820
0.28
0.2
0.31
0.23
0.39
0.25
R840
0.11
0.2
0.14
0.23
0.20
0.25
R850
0.21
0.2
0.24
0.23
0.34
0.25
R870
1.66
0.4
1.69
0.43
1.74
0.46
R880
0.55
0.4
0.58
0.43
0.63
0.46
R890
0.57
0.4
0.60
0.43
0.64
0.46
R910
1.39
0.4
1.42
0.43
1.47
0.46
R920
0.44
0.4
0.47
0.43
0.53
0.46
R940
0.92
0.2
0.95
0.23
1.01
0.25
R960
1.23
0.2
1.26
0.23
1.30
0.25
R980
0.62
0.2
0.65
0.23
0.70
0.25
R990
0.34
0.2
0.37
0.23
0.44
0.25
Parameter Calibration
Every reach in the HEC-HMS model was optimized for both its Muskingum X and K values. In
confirmed water storage areas where the initial Muskingum X value was 0.4, the X value was
restricted between 0.2 and 0.5 so the values for X would not be less than the overall default or
greater than the theoretical maximum. For all other reaches where Muskingum X values were
initially the default (0.2), the value was restricted to between 0.1 and 0.4, since these reaches
should have an X value greater than the default and less than that of open water. Reach
Muskingum K values were restricted to an order of magnitude above (10K) or below (K/10) the
Appendices
-------
original value (Table D4). Since less emphasis was placed on accurately assessing initial values
for other parameters, this restriction helped to ensure that the calibration process didn't skew
K values in place of other parameters, at least for the initial calibrations. There were two
exceptions where reach Muskingum X values were not optimized because increased X would
have resulted in instability that sub-reaches could not correct. These reaches were R580 (X=0.3)
and R660 (X=0.28).
Table D-4: Example of restrictions on reach component optimization for parameters Muskingum
X and K
Reach
Initial X
MinX
Max X
Initial K
Min K
Max K
R100
0.2
0.1
0.4
0.13
0.013
1.3
R920
0.4
0.2
0.5
0.44
0.044
4.4
Basin CNs were uniformly calibrated by a scale factor. A scale factor works the same way as our
restrictions on Muskingum K, where all values are increased by that factor. CNs were held to a
factor of 0.01 to 100 (HEC-HMS default).
As an alternative to the machine calibrations, CNs were adjusted for both the Fall 2005 model
using AMC 1 methods (D-l AMC I Dry) and the Spring 2001 model using AMC III methods (D-l
AMC III Wet). Without further calibration these methods yielded similar or improved fits for
both models and were chosen over the machine calibrated basin CNs.
Both the Spring 2001 and Fall 2005 models were used to compare the three sets of routing
parameter values (those from the '05 calibration, the '01 calibration, and an average of the two
"blended"). Both model peaks performed better with the wet parameters; whereas Root Mean
Square (RMS) Error and Nash-Sutcliff were slightly better for both models using the blended
parameters (Table D-5). Model peak fit is more important for our application than general
model fit, so the routing parameters from the Spring '01 calibration were used for both the
spring and fall models.
Table D-5: Comparison of both model results using three sets of routing parameters
Routing Parameters
Storm
Peak Error (m3/s)
Peak Error (Hours)
RMS Error
Nash-Sutcliff
Dry Parameters
Oct '05
7.5
~2 early
4.6
0.885
Blended Parameters
Oct '05
6
~1 early
3.8
0.921
Wet Parameters
Oct '05
2.8
~1.5 early
4.2
0.904
Dry Parameters
Mar '01
-3.7
~2 late
5.1
0.547
Blended Parameters
Mar '01
-3.6
~3 late
4.9
0.580
Wet Parameters
Mar '01
-2.8
~2.5 late
5.0
0.564
82 Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
Appendix E - Reservoirs
Reservoirs were parameterized in the models by identifying reservoirs associated with dams
with a maximum storage of 12 acre-feet or more from a statewide spatial data layer of dams
(Table D-2). Each reservoir was assigned an outflow curve (e.g. Figure E-l for Georgiaville
Reservoir) based on storage-discharge (e.g. Table E-l for Georgiaville Reservoir) with initial
conditions as inflow equal to outflow meaning any baseflow actually exiting at the time was
ignored. Once the reservoir volume exceeded storage the discharge increased with no
corresponding increase in storage.
Table E-l: Georgiaville Reservoir storage-discharge table
Storage Flow
Normal 1603.5 0
Maximum 2405.3 106.84
300-
300
250-
250"
W 200-
1T 200-
g, 150-
.« 1 Do-
ra 100-
50-
1,700
1,900 2.000 2,100 2,200 3,300 2.100
THOU M3 STORAGE
1,700 1,600 1.900 2,000 2,100 2,200 2,300 2,400
THOU M3 STORAGE
Figure E-l: Georgiaville Reservoir Storage-Discharge Function: original (from Table E-l, left) and with
an exponential function (from Table E-2, right).
After initial calibration model reservoirs were updated in two ways. First, four data points were
added to the storage-discharge table based on an exponential relationship between the two
original data points. This change allowed reservoirs to fill up faster and drain more slowly. The
impact of changing individual reservoirs in this way was variable, but overall peak fit for the
Spring '01 model increased by 0.5 m3/s. The second update was to increase the initial reservoir
discharge to better represent conditions before the Spring '01 storm. Modeled flows were
calculated using initial discharges for all reservoirs of 0, 1, 2, 5, and 10 m3/s (Figure E-3). The
results suggested the impact of initial discharge varies over the duration of the model. Initially
modeled flows were very sensitive to the discharge, but modeled flows converged again after
about 3 days. However, modeled flow where initial reservoir discharge was equal to inflow did
not converge with the other modeled flows. An initial reservoir discharge of 0.5 m3/s was
Appendices
-------
chosen because this value was large enough to increase peak flow but small enough not to
exceed baseflow.
Table E-2: Updated Georgiaville Reservoir storage-discharge table based on an exponential
relationship between normal and maximum discharge
Storage
Flow
Normal
1603.5
0
10%
1683.7
0.003
25%
1804.0
0.017
75%
2044.5
0.55
90%
2325.1
31.27
Maximum
2405.3
106.84
— Observed Flow
0 CMS Initial Discharge
1 CMS Initial Discharge
— 2 CMS Initial Discharge
5 CMS Initial Discharge
— 10 CMS Initial Discharge
Mar2001
26 '27'28' 29
Figure E-3: Results of altering initial reservoir discharge compared to observed flow (black). Initial
reservoir discharges tested ranged from 0 CMS (green) to 10 CMS (red).
Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
Appendix F - HMS Calibration and Validation Data
The storm of record occurred in March of 2010, however much of the data around the peak of
the event were not available. Therefore we chose to calibrate to the second largest event
(October 15, 2005; Figure F-l). The model was started at 00:00 on October 12, 2005 and
stopped at 00:00 on October 17th, 2005, for a total duration of 121 hours. Baseflow, the
minimum flow during this time period, was 0.878 m3/s.
50
45
40
_ 35
CO
^ in
0
0.5
1
1.5
i
N
¦ i II
5 7S
z
9 R
o Zb
LL.
g 20
ro
* 15
1 n
Z.J
3
' 3.5
A
±VJ
*+
A R
*+. D
0
5
LD
O
o
r\i
LD
o
O
r\i
LD
O
O
on
LD
o
O
on
LD
o
O
LD
O
o
LD
O
o
LD
LD
O
O
LD
LD
O
O
LD
O
o
LD
O
o
E
u
c
o
'+-•
f0
a.
'u
oj
i—
o.
Figure F-l: Hydrograph (USGS Station 01114500; black line in m3/s (CMS) on left axis) and hourly
precipitation (NOAA Coop 376698; gray line in cm on right axis) for the calibration storm
(October 15, 2005). Raw flows are shown without baseflow removed.
The third largest flow (March 22, 2001) was used to calibrate the second flow model
(Figure 13). Calibration data started at 03:00 the morning of March 13, 2001 and ran until 23:00
March 29, 2001, for a total of 404 hours. Baseflow, the minimum flow observed during the run
time, was 2.5 m3/s.
Appendices
-------
50
45
40
— 35
to
u 30
5 25
o
Ll_
g 20
(0
* 15
10
5
0
0.5
o
r\j
r\j
r\j
LO
f\J
r\j r\j
00
f\J
CD
r\j
Q.
<
O *~H
m ro
Figure F-2: Hydrograph (USGS Station 01114500; black line in m3/s (CMS) on left axis) and hourly
precipitation (NOAA Coop 376698; gray line in cm on right axis) for the calibration storm (March 22,
2001). Raw flows are shown without baseflow removed.
The fourth largest flow (April 03, 2005) actually occurred in the spring of the same year as the
storm used to calibrate the model. Precipitation and flow data surrounding this flow were used
as the second validation storm (Figure F-3). Validation started at 4:00 on April 2, 2005 and ran
until 23:00 April 7, 2005. Daylight savings was observed during this time period, at 02:00 on
April 3rd, resulting in a duration of 138 hours. Baseflow, the minimum flow observed during the
run time, was 5.15 m3/s.
86
Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
LO
lO
lO
lO
LO
LO
LO
LO
LO
LO
LO
LO
LO
LO
LO
LO
o
o
o
o
o
o
o
O
o
o
o
o
o
o
o
o
1—
1—
1—
1—
1—
i_
1—
1—
HJ
HJ
HJ
nj
nj
nj
nj
HJ
nj
Q.
Q.
Q.
Q.
Q.
Q.
Q.
<
<
<
<
<
<
<
i
m
i
i
LO
i
tO
i
r*>
i
00
i
CD
i
o
i
r—1
T—\
r\J
ro
LO
ub
r\]
r\j
r\j
r\J
r\J
r\J
r\J
m
m
Figure F-3: Hydrograph (USGS Station 01114500; black line in m3/s (CMS) on left axis) and hourly
precipitation (NOAA Coop 376698; gray line in cm on right axis) for the validation storm (April 03, 2005).
Raw flows are shown without baseflow removed.
The fifth largest flow occurred in December, when average temperatures drop below freezing
(Figure 7). Although HEC-HMS has parameters to be able to accommodate sub-freezing
temperatures, we had not collected the necessary data or calibrated the model for these
conditions. Skipping the fifth largest flow, the sixth largest flow (Table 5; April 16, 2007) became
the third and last validation storm. The third validation started at 00:00 on April 13, 2007 and
ran until 15:00 April 23, 2007 for a total duration of 255 hours (Figure F-4). Baseflow, the
minimum flow observed during the run time, was 4.13 m3/s.
Appendices
-------
:Y
u 30
Figure F-4: Hydrograph (USGS Station 01114500; black line in m3/s (CMS) on left axis) and hourly
precipitation (NOAA Coop 376698; gray line in cm on right axis) for the calibration storm (April 16,
2007). Raw flows are shown without baseflow removed.
Table F-l: Summary of storm baseflows.
Estimated
Storm
Baseflow (m3/s)
October '05
0.88
March '01
2.52
April '05
5.15
April '07
4.13
88 Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
Appendix G - Synthetic Storms
Synthetic storms of known probability were used as precipitation data to generate simulated
flows with the calibrated model. The duration of the synthetic storm was 24 hours based on
observed storm events used to calibrate and validate the model. Peak precipitation intensity
was positioned in the center of that 24-hour duration. Duration intervals are used in HEC-HMS
to create precipitation curves for the synthetic storm, where higher intensity precipitation is
received during shorter intervals, and longer intervals are move evenly distributed across time.
Duration intervals of as little as 5 minutes were available. Based on comparisons of 5-minute
(blue), 1-hour (red), and 2-hour (green) duration intervals (Figure G-l) peak precipitation during
the 1-hour interval still exceeded peak precipitation observed in calibration and validation
storms, but 5-minute duration intervals were off by degrees of magnitude. Ultimately, 1-hour
duration interval was used for the synthetic storms. This interval also corresponded to the
observed hourly precipitation used to calibrate and validate the models.
0.Q9
u.Jb
007
0/J0
0.05
a.
£ 0.04
a
0.03
00?
0.0' 1
0.00
Q3:0v 06:00 0900 2 SO 15 00 19:0; 2100 00 CO 03 CO
'SVatfCOl | MMsr2001
Figure G-l: Peak precipitation values were compared across three duration intervals: 5-minute (blue),
1-hour (red), and 2-hour (green).
— s
Mfi
Hou
2
k.«J
J
1
1
1
1
xu
Appendices
-------
Appendix H - Maps
~ Addresses in Flood Zone
o Addresses in Modeled Area
~ Model Boundary
| | Subbasins
1 Year Flood
High : 5.5 m
Low : 0 m
¦ Kilometers
8
Figure H-l: Addresses (orange) included in the area where flood modeling was performed (black border)
and the subset of those addresses (red) that experienced flooding in the 1-year Baseline Scenario (blue).
Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
Non-urban Developed
Prime Farmland
~ Modeled Subbasins
2025 Projected Landuse
I I Sewered Urban Developed
Urban Development
Figure H-2: 2025 Projected lariduse
across modeled subbasins.
I Modeled Subbasins
Social Vulnerabillity Index (SoVi)
| High
Medium High
Medium
Medium Low
Low
Figure H-3: Social Vulnerability Index
across modeled subbasins and flood
modeled area.
• Dams
Levee
~ Model Boundary
Subbasins
Kilometers
A
Figure H-4: Dams (brown) and levees (purple) in the flood modeled area (black border).
Appendices
-------
J I,
I f
N
A
Legend
Model Boundary
DFIRM Flood Zone
100 Year
500 Year
Minimal Risk
fx tr-A
o 1
i Kilometers
3
Figure H-5: Flood model boundary (red boundary) compared to FEMA DFIRM mapped flood zones.
Minimal risk areas are outside of the 500 year flood zone.
Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
DFIRM Flood Zone
Minimal Risk
irnentsi tJi3C& £ei?er
Figure H-6: RIGIS imagery with DFIRM to identify assets in the 100- and 500-year flood zones as defined
by FEMA.
Appendices 93
-------
Address With Decreased Flooding
Subbasin 2.5-mile (4-km) buffer
Subbasin 1540 (Left)
Subbasin 2370 (Right)
Modeled Change in Flooding
Increased (>0.04 ft)
Non-Detectable Increase (0 to 0.04ft)
Non-Detectable Decrease (0 to -0.04ft)
Decreased (<-0.04ft)
, 0
_ 0
( (i )
f )
\ • /
V
N
A
0 1 25 2 5 5 7 5
Figure H-7: A 2-mile downstream buffer identifies the same beneficiaries that were counted in the
Tier III example assessment.
Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
Appendix I - Tier I! Indicator Development
Storm Event
1-year 5-year 25-year
0 5000 10000 15000 20000 250000 5000 10000 15000 20000 25000 0 5000 10000 15000 20000 25000
Distance Downstream (m)
Figure 1-1: The change in flow (%) as the distance downstream (m) from the simulated restoration for all
subbasins (n=135) for each restoration scenario (wetland change scenarios {1%, 5%, 10%, and 25%) and
synthetic storm event (1-year, 5-year, and 25-year)). Two subbasins are shown in red (W2370) and blue
(W1540). Note that y-axis vary between sub-plots with the changes in wetland change scenarios.
Appendices
-------
Dist USGS GAGE
O
o
o
-
(N
O
o
-
o
°o
OO
o
o
w
kn
oo,
o
5000 10000 20000
g Basin_Slope
5 10 15
o Basin Area km2
o — —
0.0 0.5 1.0 1.5 2.0
g Pctlmp
o Pct_Wetlands
0.0 0.2 0.4 0.6 O.S
Longest_Flowpath
O
o
•-
o
(N
o
o
o
o
o
-
o
LTJ
O
0
1000 2000 3000 4000
Figure I-2: Downstream distance (m) to no change in flow (<1% change) plotted against subbasin
characteristics: (distance to the outlet (m), subbasin area (km2), % wetlands, mean basin slope (degrees),
% impervious cover, and longest flowpath (m)).
i
Q
4-
0 1
Percent Wetland
0 1
Subbasin Area
4
0 1
Distance to Outlet
0 1
Basin Slope
0 1
Percent Impervious
Longest Flowpath
Figure I-3: Boxplots of downstream distance (m) to no change in flow (<0.2% change) plotted by
subbasin characteristics (% wetlands, subbasin area (km2), distance to the outlet (m), mean basin slope
(degrees), % impervious cover, and longest flowpath (m)) split into below (0) and above (1) the mean
values for the basin characteristics.
Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
Appendix J - IR Script to read in IDSS Peak Flows and save as csv
#DESCRIPTION: This script pulls in all dss files in a folder, finds the peak flows for specified points and creates
#a table of these peak values. That table can be used to find differences between scenarios.
########################PACKAGES########################
install.packages("rJava")
library(rJava)
install.packages("devtools")
dss-rip package
devtools::install_github("eheisman/DSS-Rip",args="--no-multiarch")
library(dssrip)
#######################FUNCTIONS#######################
reachTOrun <- function(element, run, type){ #ASSEMBLES THE PATH NAME
data= pasteO("//", element,"/", type, 7/1MIN/", run,"/")
return(data)
}
maxFROMxts <- function(path, file){ #GETS THE PEAK FLOW FROM THE XTS FILE
XTS = getFullTSC(file, path)
return (max(XTS))
}
getPeak <- function(elements, run, file, type){ #THIS ONE DOES ALL THE WORK
file<-opendss(file)
elements.paths <- sapply(elements, reachTOrun, run, type)
elements.peaks <- sapply(elements.paths, maxFROMxts, file)
file$close()
return(elements.peaks.dt <-data.frame(elements.peaks, row.names = elements))
#return(names(elements.peaks.dt)<-run)
}
singleMaxFlow <- function(HMS_location, file, run, flows, type){ #BASICALLY JUST CALLS getPeak
myFile = pasteO(HMS_location, file)
table_ALL=getPeak(flows, run, myFile, type)
names(table_ALL)<-run
return(tableALL)
}
IstMaxFlow <- function(tableName, myRunLst, myFileLst, flows, type){ #BASICALLY JUST CALLS getPeak
across a list
tableName <- data.frame(row.names = flows)
i=0
if (length(myRunLst)==length(myFileLst)){
for (file in myFileLst){
i=i+l
#find peaks for all HMS Elements in flows
table = getPeak(flows, myRunLst[i], file, type)
names(table) <- myRunLst[i]
tableName <- (cbind(tableName, table))
}
}
return(tableName)
}
##########ELEMENT LISTS##########
#all HMS elements in hydrologic order
###COPIED FROM HMS, DSS REQURIES ALL CAPS###
FlowsALL hydro <-c('W1480', 'DW1480', 'W1590', 'DW1590', 'J597', 'R260', 'W1490', 'DW1490',
'W1580', 'DW1580', 'J600', 'R530', 'W1790', 'DW1790', 'W1610', 'DW1610', 'W1870', 'DW1870', 'J539',
'R610', 'W1840', 'DW1840', 'W1900', 'DW1900', 'J531', 'R630', 'W1810', 'DW1810', 'W1710', 'DW1710',
'W1860', 'D W1860', 'J542', 'R510', 'W1690', 'DW1690', 'W1850', 'DW1850', 'J545', 'R620', 'W1880',
Appendices
-------
'DW1880', 'W1950', 'DW1950', 'J518', 'R880', 'W1800', 'DW1800', 'W2000', 'DW2000', 'J506', 'R790',
'W2090', 'DW2090', 'W2070', 'DW2070', 'J483', 'R820', 'W2050', 'DW2050', 'W2110', 'DW2110', 'J478',
'R870', 'W1960', 'DW1960', 'W2140', 'DW2140', 'J470', 'R890', 'W2160', 'DW2160', 'W2180',
'DW2180', 'J460', 'R920', 'W2220', 'DW2220', 'W2260', 'DW2260', 'W2190', 'DW2190', 'J440', 'R960',
'W2230', 'D_W2230',''RES_WATERMAN'','J448', 'R940', 'W2250', 'DW2250', 'W2240', 'DW2240',
"RES GREENVILLEMILLPOND", 'J443', 'R2800', 'W2780', DW2780', "RESKNIGHTMILLPOND",
"RESKNIGHTMILLPOND", 'R2750', 'W2770', 'D W2770', "RESSTILLWATERMILLPOND",
"RESSTILLWATERMILLPND", 'R1010', 'W2490', 'DW2490', 'J380', 'R1200', 'W2470', 'DW2470',
'W2500', 'DW2500', 'W2480', 'DW2480', 'J385', 'R1160', 'W2410', 'DW2410', 'W2400',
'D_W2400',"RES_SLACK",'J402', 'R1090', 'W2380', 'DW2380', 'W2390', 'DW2390', 'J407', 'R2640',
'W2630', 'DW2630', "RES_HOPKINSPOND", "RESHOPKINSPOND",'R2690', 'W2680',
'D_W2680',"RES_MOWRYPOND", "RESMOWRYPOND",'R1070', 'W2720', 'D W2720', 'W2670',
'DW2670', 'J435', 'R980', 'W2580', 'D_W2580',"RES_UPPERSPRAGUE", "RESUPPERSPRAGUE",'R720',
'W2060', 'DW2060', 'W2570', 'D W2570', 'J491', 'R2550', 'W2530', 'D_W2530',"RES_LOWSPRAGUE",
"RESLOWSPRAGUE",'R840', 'W2130', 'DW2130', 'W2520', 'DW2520', 'J473', 'R2990',
"RESGRANITEMILLPOND", 'W2980', 'D_W29807'RESGRANITEMILLPOND'','R810', 'W2830',
'D_W2830',"RES_HAWKINSPOND", "RESHAWKINSPOND",'R2890', 'W2880', 'DW2880',
"RESREAPERPOND", "RESREAPERPOND",'R1240', 'W2870', 'DW2870', 'W2430', 'DW2430', 'J397',
'R1100', 'W2200', 'DW2200', 'W2210', 'DW2210', 'J453', 'R2940', 'W2930', 'DW2930',
"RESMOUNTAINDALE", "RESMOUNTAINDALE", 'R910', 'W2920', 'DW2920', 'W2970', 'DW2970',
'J486', 'R660', 'W1770', 'DW1770', 'W1780', 'DW1780', 'J561', 'R460', 'W1290', 'DW1290', 'W1270',
'DW1270', 'J672', 'R20', 'W1300', 'DW1300', 'W1280', 'DW1280', 'J669', 'R3040', 'W3030',
'D_W3030',"RES_CESARIOPOND", "RESCESARIOPOND", 'R3090', 'W3080', 'DW3080',
"RESPRIMROSEPONDLOWER", "RESPRIMROSEPONDLOWER",'R70', 'W3070', 'D_W3070','J664', 'R80',
'W1260', 'DW1260', 'W1330', 'DW1330', 'W1340', 'DW1340', 'W1350', 'DW1350', 'J657', 'R100',
'W1380', 'DW1380', 'J649', 'R110', 'W1360', 'DW1360', 'J654', 'R130', 'W1370', 'DW1370', 'W1320',
'DW1320', 'W1400', 'DW1400', 'J644', 'R160', 'W1390', 'DW1390', 'W1410', 'DW1410', 'J641', 'R330',
'W1420', 'DW1420', 'W1470', 'DW1470', 'J628', 'R190', 'W1530', 'DW1530', 'W1460', 'DW1460', 'J614',
'R200', 'W1450', 'DW1450', 'W1510', 'DW1510', 'J619', 'R340', 'W1500', 'DW1500', 'J580', 'R350',
'W1430', 'DW1430', 'W1640', 'DW1640', 'J588', 'R380', 'W1660', 'DW1660', 'W1650', 'DW1650', 'J583',
'R420', 'W1700', 'DW1700', 'W1750', 'D_W1750',"RESSTUMPPOND", 'R3200', 'W3180', 'DW3180',
"RESSTILLWATERPOND", "STILL WATERPOND",'R470', 'W1540', 'DW1540', 'W1440', 'DW1440',
'J611'R300', 'W1560', 'DW1560', 'W1550', 'DW1550', 'J606', 'R430', 'W3240', 'D W3240', 'W3230',
'D W3230', 'J345', 'R480', 'W3220', 'D W3220', 'W3250', 'D_W3250',"RES_CAPRON",'J342', 'R560',
'W1830', 'D W1830', 'W1820', 'D W1820', 'J550', 'R580', 'W1930', 'D W1930', 'W1920', 'D W1920', 'J524',
'R640', 'W1680', 'D W1680', 'W1600', 'D W1600', 'J577', 'R410', 'W1520', 'D W1520', 'W1570',
'D W1570', 'J603', 'R400', 'W1720', 'D W1720', 'W1620', 'D W1620', 'J568', 'R650', 'W1670', 'D W1670',
'W1940', 'D W1940', 'J521', 'R690', 'W1890', 'D W1890', 'J534', 'R700', 'W1910', 'D W1910', 'W1970',
'D W1970', 'J513', 'R710', 'W2030', 'D W2030', 'W2020', 'D_W2020V'RES_GEORGIAVILLE", 'J499',
'R770', 'W2010', 'D W2010', 'W2040', 'D W2040', 'J496', 'R850', 'W2150', 'D W2150', 'W2100',
'D W2100', 'J467', 'R990', 'W2170', 'D W2170', 'W2290', 'D W2290', 'J432', 'R1040', 'W2330', 'D W2330',
'W2320', 'D W2320', 'J422', 'R1060', 'W2370', 'D W2370', 'W2360', 'D W2360', 'J412', 'R1050', 'W2310',
'D W2310', 'W2300', 'D W2300', 'J427', 'R1130', 'W2340', 'D W2340', 'W2350', 'D W2350', 'J417',
'R1150', 'W2450', 'D W2450', 'W2440', 'D W2440', 'J392', 'R3140', 'W3130', 'D W3130',
"RES GREYSTONEDAM", "RESGREYSTONEDAM", 'R1220', 'W3120', 'D W3120', 'USGS GAGE')
#List with the correct order to import directly into RAS both FLOW and COMBINE data
Full RAS <- c('J577', 'R2890', 'R1240', 'J397', 'J453', 'R910', 'R720', 'J491', 'R840', 'J380', 'J385', 'R1090', 'J407',
'R2690', 'R1070', 'J440', 'J506', 'J483', 'J478', 'J542', 'J545', 'J412', 'J534', 'J603', 'J568', 'J611', 'J606', 'J597', 'J600',
'J539', 'J531', 'J518', 'J470', 'J460', 'R940', 'R2800', 'R2750', 'R1010', 'J435', 'J473', 'R810', 'J486', 'J628', 'J614', 'J619',
'J580', 'J649', 'J672', 'J669', 'R3090', 'R1220', 'J657', 'J654', 'J644', 'J641', 'J588', 'J583', 'R3200', 'R470', 'J345', 'R560',
'J524', 'J521', 'J513', 'R770', 'J496', 'J467', 'J432', 'J427', 'J392', 'R70', 'USGS GAGE')
#list of Flows to subset
RAS FLOW <-c('J577', 'J397', 'J453', 'J491', 'J380', 'J385', 'J407', 'J440', 'J506', 'J483', 'J478', 'J542', 'J545',
'J412', 'J534', 'J603', 'J568', 'J611', 'J606', 'J597', 'J600', 'J539', 'J531', 'J518', 'J470', 'J460', 'J435', 'J473',
Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
'J486', 'J628', 'J614', 'J619', 'J580', 'J649', 'J672', 'J669', 'J657', 'J654', 'J644', 'J641', 'J588', 'J583', 'J345',
'J524', 'J521', 'J513', 'J496', 'J467', 'J432', 'J427', 'J392', 'USGSGAGE')
#list of reach combines to subset
RASCombine <-c('R2890', 'R1240', 'R910', 'R720', 'R840', 'R1090', 'R2690', 'R1070', 'R940', 'R2800',
'R2750', 'R1010', 'R810', 'R3090', 'R1220', 'R3200', 'R470', 'R560', 'R770', 'R70')
####PARAMETERS AND IMPLEMENTATION#####
HMSlocation = "C:\\HMS\\Outputs\\FinalScenario\\"
####single file####
run = "RUN:MARCH_2001_HEAD_MW_MIDCALIB_01WET"
file = "ResultsAll.dss"
flows = FlowsALLhydro
#Run baseline analysis
baselineTable_ALL<-singleMaxFlow(HMS_location, file, run, flows, "FLOW")
baselineTable_Combine<-singleMaxFlow(HMS_location, file, run, RAS Combine, "FLOW-COMBINE")
####file list####
runConvention = "RUN:MARCH_2001_HEAD_MIDCALIB_01_SCENARIO_"
runNumber= 1:139
#1% scenario
dssFileLstl = 1:139
myFileLstl <- pasteO(HMS_location, "Opt Results ", dssFileLstl, ".dss")
runConventionl = pasteO(runConvention, "1","_")
#10% scenario
dssFileLstlO = 279:417
myFileLstlO <- pasteO(HMS_location, "Opt Results ", dssFileLstlO, ".dss")
runConventionl0 = pasteO(runConvention, "10","_")
#25% scenario
dssFileLst25 = 418:556
myFileLst25 <- pasteO(HMS_location, "Opt Results ", dssFileLst25, ".dss")
runConvention25 = pasteO(runConvention, "25","_")
#5% scenario
dssFileLst5 = 140:278
myFileLst5 <- pasteO(HMS_location, "Opt Results ", dssFileLst5, ".dss")
runConvention5 = pasteO(runConvention, "5","_")
myRunLstlO <- pasteO(runConventionlO, runNumber)
myRunLst25 <- pasteO(runConvention25, runNumber)
myRunLst5 <- paste0(runConvention5, runNumber)
myRunLstl <-pasteO(runConventionl, runNumber)
#Run analysis
scenario 10_table_All <- lstMaxFlow(scenariolO_table_All, myRunLstlO, myFileLstlO, FlowsALL hydro,
"FLOW")
scenario25_table_All <- lstMaxFlow(scenario25_table_All, myRunLst25, myFileLst25, FlowsALL hydro,
"FLOW")
scenario5_table_All <- lstMaxFlow(scenario5_table_All, myRunLst5, myFileLst5, FlowsALL hydro, "FLOW")
scenarioI table All <- lstMaxFlow(scenarioI table All, myRunLstl, myFileLstl, FlowsALL hydro, "FLOW")
#Run FLOW-Combine
scenariolO table combined <- lstMaxFlow(scenario 10_table_All, myRunLstlO, myFileLstlO, RAS Combine,
"FLOW-COMBINE")
scenario25_table_combined <- lstMaxFlow(scenario25_table_All, myRunLst25, myFileLst25, RAS Combine,
"FLOW-COMBINE")
scenario5_table_combined <- lstMaxFlow(scenario5_table_All, myRunLst5, myFileLst5, RAS Combine, "FLOW-
COMBINE")
scenariol table combined <- lstMaxFlow(scenariol_table_All, myRunLstl, myFileLstl, RAS Combine, "FLOW-
COMBINE")
#subset RAS flows
scenario 10RASFLOW <-scenario 10_table_All[RAS_FLOW,]
Appendices
-------
scenario25_RAS_FLOW <-scenario25_table_All[RAS_FLOW,]
scenario5_RAS_FLOW <-scenario5_table_All[RAS_FLOW,]
scenario 1RASFLOW <-scenario l_table_All[RAS_FLOW,]
#Join the two preppared for RAS
scenario 10RAS <- rbind(scenariolO_table_combined, scenario 10RASFLOW)
scenario 10_RAS<- scenario 10_RAS[match(Full_RAS, row.names(scenario 10RAS)),]
scenario25_RAS <- rbind(scenario25_table_combined, scenario25_RAS_FLOW)
scenario25_RAS<- scenario25_RAS[match(Full_RAS, row.names(scenario25_RAS)),]
scenario5_RAS <- rbind(scenario5_table_combined, scenario5_RAS_FLOW)
scenario5_RAS<- scenario5_RAS[match(Full_RAS, row.names(scenario5_RAS)),]
scenario I RAS <- rbind(scenariol_table_combined, scenario 1RAS FLOW)
scenario 1_RAS<- scenario I RAS [match(Full_RAS, row.names(scenario I RAS)),]
####Write to csv####
stormRun-' springO 1"
#raw flows
write.csv(scenariolO_table_All, file=pasteO(HMS_location, "Flows scenariolO All ", stormRun,".csv"))
write.csv(scenario25_table_All, file=pasteO(HMS_location, "Flows_scenario25_All_", stormRun,".csv"))
write.csv(scenario5_table_All, file=pasteO(HMS_location, "Flows_scenario5_All_", stormRun,".csv"))
write.csv(scenariol_table_All, file=pasteO(HMS_location, "Flows scenariol All ", stormRun,".csv"))
######baseline to CSV#####
baselineFLOW <- data.frame(baselineTable_ALL[RAS_FLOW,], row.names = RAS FLOW)
colnames(baselineFLOW)<-"RUN:MARCH_2001_HEAD_MW_MIDCALIB_01WET"
scenarioBaseline <- rbind(baselineTable_Combine, baselineFLOW)
write.csv(scenarioBaseline[Full_RAS,], file=pasteO(HMS_location, "Baseline ", stormRun,".csv"))
#pct change from baseline
baseline = baselineTable_ALL$"RUN:MARCH_2001_HEAD_MW_MIDCALIB_01WET"
pctChange scenario lO all <- ((scenario 10_table_All-baseline)/baseline)* 100
write.csv(pctChange_scenariolO all, file=pasteO
(HMS location, "pctChange scenariolO All ", stormRun,".csv"))
pctChange_scenario25_all <- ((scenario25_table_All-baseline)/baseline)* 100
write.csv(pctChange_scenario25_all, file=pasteO
(HMS location, "pctChange_scenario25_All_", stormRun,".csv"))
pctChange_scenario5_all <- ((scenario5_table_All-baseline)/baseline)*100
write.csv(pctChange_scenario5_all, file=pasteO
(HMS location, "pctChange_scenario5_All_", stormRun,".csv"))
pctChange scenario l all <- ((scenario l_table_All-baseline)/baseline) * 100
write.csv(pctChange_scenariol all, file=paste0
(HMS location, "pctChange scenariol All ", stormRun,".csv"))
#raw RAS tables
write.csv(scenariolO_RAS, file=pasteO(HMS_location, "Flows scenariolO RAS ", stormRun,".csv"))
write.csv(scenario25_RAS, file=pasteO(HMS_location, "Flows_scenario25_RAS_", stormRun,".csv"))
write.csv(scenario5_RAS, file=pasteO(HMS_location, "Flows_scenario5_RAS_", stormRun,".csv"))
write.csv(scenariol_RAS, file=pasteO(HMS_location, "Flows scenario 1_RAS_", stormRun,".csv"))
100 Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
Appendix IK - Data Quality and Limitations
Information on data quality, as well as limitations on the use of model and indicator results,
have been incorporated into the main text where appropriate. Here, we compiled all the
relevant data quality and limitation information in one place.
Spatial datasets were selected to optimize spatial resolution, temporal relevance, and accuracy
(Bousquin et al. 2014). Metadata for all spatial datasets are available from RIG IS. Of particular
significance to results are accuracy of the Digital Elevation Model (DEM) and E911 address
points. The DEM had a 0.01-m vertical precision and an RMSE of 0.067 m based on raw LiDAR
calibration control points. Based on this vertical precision, 0.012 m was used as a "detection
threshold" for our hydraulic model, since lesser changes in flooding could be attributed to
error. E911 addresses have a horizontal accuracy of 5-10 m and may be as old as 2001. This
dataset is updated frequently as structures change and better data become available. The most
recent update (December 31, 2014) available at the time of analysis was used to demonstrate
indicators. This analysis is intended to demonstrate the indicator development approach and to
develop general indicators (Tier II) of flood protection by freshwater wetlands. This analysis is
not a formal flood zone delineation or an actual assessment of structures currently at risk of
flooding.
All hydrologic and meteorologic time series data were obtained from USGS (station 01114500)
and NOAA (COOP station 376698) and were subject to their quality assurance standards. Any
provisional or incomplete flow data, such as those for the March 2010 flood, were rejected for
use in calibration or validation of the hydrologic model. Flow data during prolonged cold
periods that could be erroneous due to ice effects, such as those for the December 2008 flood,
were rejected for use in calibration or validation of the hydrologic model. We did not model the
entire watershed, because the area below the USGS gage used for model calibration is highly
urbanized and lacks information on stormwater infrastructure, which is likely to lead to errors
in predicting flood extent.
We evaluated hydrologic model fit using comparison to gauged storm peak flows, Nash-
Sutcliffe (N-S) efficiency, and Root Mean Square Error (RMSE) statistics. Two models were
calibrated, one for dry conditions (October 2005) and one for wet conditions (March 2001).
These two models were not expected to cross validate well, representing opposing antecedent
moisture conditions. The March 2001 calibrated model had an N-S of >0.5 for the storm
calibrated and the April 2005 validation storm, an N-S of approximately 0 for the 2007
Validation storm, and an N-S <0 for the October 2005 storm. This suggests the calibrated model
performs well with wetter antecedent moisture conditions and larger storms, is less satisfactory
for smaller storms (April 2007 had a flow recurrence interval of 3.4 years), and unfit for dry
antecedent moisture conditions (October 2005).
Sensitivity Analysis showed the hydrologic model was particularly responsive to precipitation
and assumptions of divergence of runoff to wetlands based on wetland area. The precipitation
Appendices 101
-------
station (NOAA COOP #376698) is outside the modeled basin and 15.9 km south of the USGS
gage station. The assumptions behind treatment of runoff diverted to wetlands received
particular attention in our methods and analysis. These assumptions should be considered
when using the model orTier II indicators developed here.
Simulated water levels from the hydraulic model obtained from USGS were previously
compared to 2010 flood high-water marks in Zarriello et al. 2013. Based on those comparisons,
the model was considered valid, provided structures and stream channels remained clear of
debris. However, that hydraulic model was adapted for use in this study. The adapted hydraulic
model included additional cross section and stream reach geometry and flows from
hydrographs from the hydrologic model. The hydraulic model and derived flood maps were
used for Tier III indicator demonstration purposes only and should not be used as an actual
assessment of actual flood risk. Given this use of the adapted hydraulic model, no further
validation was performed.
Tier II indicators of downstream distances for flow of flood benefits were based on changes in
flow in simulated restoration scenarios. The number of simulations necessitated automation,
and results were checked for quality. Of the 139 subbasins, flows from four subbasins did not
correspond to the restoration scenario in one or more simulations (W2980, W1830, W1660,
and W2360). To be consistent, those four subbasins were removed from the dataset, reducing
the number of restoration simulations to 1,620.
There were two further quality restrictions on individual data points within simulations. First,
no percent change in flow should show an increased flow and, second, no percent change in
flow should exceed the percent restoration, based on the way that restorations were
implemented in the model (i.e. 5% scenario results in a 5% reduction of flow). The number of
instances where these errors (1-lncreased or 2- Exceed) were greater than 0.1% are shown in
Table K-l. These errors are extremely low, considering there are 2,484 data points in each of
the 12 basin scenarios. Closer inspection of these errors showed they typically originated from
flows leaving dams. These errors were assumed to be non-systemic given their rarity (e.g. in the
scenarios with the most errors, 25% restoration and 25-year storm scenarios, only <0.006% of
data points had increased flow), and only individual data points were removed from
consideration.
Table K-l: Single reach data points where percent change was outside of the expected bounds.
1-Year
5-Year
25-Year
Restoration Increased
Exceed
Increased
Exceed
Increased
Exceed
1%
5%
10%
25%
0
0
0
0
3
4
5
3
0
0
0
0
0
2
2
1
8
6
9
11
5
5
4
5
102 Benefit Indicators for Flood Regulation Services of Wetlands: A Modeling Approach
-------
We expect that Tier II indicators developed in this study could be used in similar watersheds in
the Northeast, but until similar modeling is conducted in other watersheds, it is difficult to say
how transferable the indicators are. Downstream distances for benefits delivery did not appear
to vary greatly within the range of subbasin areas and imperviousness explored. Other factors
such as stream morphology and subbasin slope may play a role in this as well. Although we
were unable to develop generalized indicators quantifying the role of substitutes, such as dams,
in decreasing wetland benefits, it is possible the number of dams in the Woonasquatucket
influenced our results for the downstream distance of benefits.
Appendices
------- |