Introduction
Societal vulnerability to riverine flooding is a complex function of physical
hydrological processes, overlaid with our economic relationship to the land
(Wheater and Evans, 2009), community perceptions and responses to risk (e.g.,
Vinh Hung et al., 2007), and the fundamental ability of experts to clearly
communicate these risks to facilitate decisions by policy makers
(Pappenberger et al., 2013). Recent flood losses across North America,
Europe, and Asia have been exacerbated by uninformed and inaccurate prior
beliefs surrounding riverine flood hazards (Merz et al., 2015), fundamental
issues in how governmental organizations store and leverage data (Lane et
al., 2011; Harries and Penning-Rowsell, 2011), the cognitive biases of
individuals (Merz et al., 2015; Harries, 2008), socially organized apathy
(Norgaard, 2011), and governmental response to societal deviations from
anticipated rational behavior (Lupton, 2013).
Riverine flood risk analysis is inherently difficult due to the infrequency
of flooding events (Merz et al., 2015), a globally nonstationary climate
leading to more extreme precipitation (Trenberth et al., 2011), often
nonlinear hydrologic rainfall–runoff responses (e.g., Mathias et al., 2016),
and complex human–flood interactions (e.g., Collenteur et al., 2015), all of
which can act to limit the intuition of decision makers for understanding
flooding risks and selecting mitigation options (Merz et al., 2015). For any
given local flood risk, different data, models, and assumptions can be
combined in various ways to yield alternative, reasonable measures of the
“true” or “real” risk.
Flooding governance is typically discussed as being “top-down” or
“bottom-up.” Top-down typically refers to a technocratic hierarchy, often
in the form of national institutions acting as the sole decision makers (e.g., Serra-Llobet et al., 2016; Plate, 2002). These approaches have been
associated with a reliance on hazard-based assessments of risk. Bottom-up
approaches, in contrast, leverage the knowledge, experiences, preferences,
and vulnerabilities of end users explicitly in problem definition and
selection of mitigation actions. In bottom-up approaches, decision-making is
generally a collaborative process across institutions. Previous work has
shown that flood risk mitigation can benefit from a combination of top-down
and bottom-up approaches whereby decisions are collaboratively refined and
implemented across institutions (Serra-Llobet et al., 2016), often with
direct input from stakeholders (e.g., Edelenbos et al., 2017; J. Knighton et
al., 2017). Pahl-Wostl et al. (2013) demonstrate, through a case study of
three European nations, the relative benefits of different governance
schemes. Top-down mitigation, as found in Germany, allowed for clearer roles
in decision-making, centralized repositories of knowledge, and more rapid
action within limited windows of opportunity. The vertically integrated
approaches of the Netherlands and Hungary allowed for greater integration of
new information into policy decisions. In an effort to tackle some of the
“wicked problem” characteristics of water resources challenges, researchers
have been engaged in developing bottom-up frameworks for decision-making with
a focus on problems accompanied by deep uncertainty (e.g., Many-Objective
Robust Decision Making Kasprzyk et al., 2013; Scenario Neutral Planning
Prudhomme, 2010, Decision Scaling Brown et al., 2012) with applications
developed specifically to aid flood risk decision analysis (e.g., Evers et
al., 2018; J. Knighton et al., 2017).
Bottom-up decision analysis methodologies initially focus on understanding
system vulnerabilities (i.e., what are the negative consequences of a flood
that we wish to avoid?), mapping these vulnerabilities onto a wide range
of plausible hazard scenarios (e.g., if an n-year flood occurs, which of the
previously defined losses will we experience and possibly to what
severity?), and then evaluating which of these hazard scenarios are most
likely given our current understanding of atmospheric and hydrological
processes. Flood decision-making can be stalled by contentious discussions
about the reliability of hazard data (e.g., is climate change driving changes
to local storms? Should climate change be accounted for in mitigation
planning?). A bottom-up perspective can improve decision-making in that
uncertain data and potentially controversial methodologies can be evaluated
within the context of community risks. For example, climate-change-driven
changes to storm intensity may not increase frequency severe economic
losses, and therefore can possibly be disregarded. Such public debates over
the “accuracy” of hazard data and risk estimation, of the kind illustrated
by recurrent controversies surrounding flood insurance rate maps in the USA
(Elliott and Rush, 2017), reflect a technocratic faith that pegs
decision-making to the purported ability of risk analysis to arrive at
single true estimates of risk, which models typically do not and cannot
provide (Weinkle and Pielke Jr., 2017). Vulnerability-based assessments, by
contrast, map hazards directly to community vulnerabilities in order to
produce stakeholder-relevant predictions and outcomes, thereby enhancing the
broader legitimacy of any subsequent actions taken.
This is not, however, how risk mitigation planning and design has
historically been conceptually modeled within the USA. Instead, it has been
treated as a top-down process, with knowledge transfer between two distinct
groups comprised of lay people and experts (Wood et al., 2012; Birkholz et al.,
2014), the former typically understood as being ignorant or “overly emotional”
while the latter are presumed to be rational and “analytical” (Lupton,
2013). However, this fails to capture how both groups approach issues of
risk and natural hazards. As discussed in Norgaard (2011), a “knowledge
deficit” model that assumes lay people would take (rational) action if they
“only knew” is too limiting, as it leaves aside the institutional
structures (Harries and Penning-Rowsell, 2011) and cultural differences
(e.g., Masuda and Garvin, 2006) that shape orientations to risk,
institutional responses, and community vulnerabilities.
Furthermore, the simplistic distinction between professionals and lay people
is often blurred in practice. Though the momentum driving flood risk
mitigation originates at the federal level within the USA (Burby, 2001) and
many other nations, implementation of national policies and redistribution of
resources relies on social infrastructure at the local level (Few, 2003;
Rauken et al., 2015; Vogel and Henstra, 2015). Within the USA, Canada,
Australia, and Europe, flood governance is frequently the collective effort
of organizations operating across scales, including governmental
organizations, nongovernmental organizations (NGOs), privately owned firms,
citizen-led groups, and private research organizations (Plummer et al.,
2018). Within these organizations, the individuals who participate can be
considered experts in one subdiscipline of flood risk mitigation, with
shifting leadership roles throughout the process. For instance, governmental
organizations may take the lead on policy and legislation, while privately
owned firms contribute hydrologic modeling, and residents and citizen-led
groups share knowledge about local vulnerabilities to and effects of exposure
to flood risk. These coalitions benefit by leveraging the skills, knowledge,
and social position of the varied organizations to more effectively reduce
flooding risk.
Bottom-up decision analysis frameworks place emphasis on adequately
understanding and conveying community vulnerabilities and historical risks
into the decision space. These frameworks rely heavily on strong social
networks among professionals and the public to bridge gaps among institutions
while articulating stakeholder interests (Morss et al., 2005). Collaboration
and stronger trust relationships among institutions and the public have
previously led to more effective means of disaster risk mitigation and
climate adaptation at the state level (Clarvis and Engle, 2015; Haer et al.,
2016). Opportunities to limit cognitive biases should be explored to reduce
institutional vulnerability to flooding (Merz et al., 2015), though few
research projects have considered how differences among flooding
practitioners may be understood or modified to further reduce vulnerability
to flooding hazards (e.g., Morss et al., 2005; Downton et al., 2005).
Given the intensity and complexity of the cooperation and coordination
required to plan and implement flood risk mitigation, establishing common
understandings – of community vulnerability to flooding, baseline flood loss
frequency, community willingness, and project-desired goals – is a challenging
and nontrivial task (Pahl-Wostl, 2009), particularly given diverse
backgrounds, education, work experience, and risk exposure of stakeholders
found in interdisciplinary working groups. de Brito and Evers (2016)
reviewed multi-criteria decision analysis efforts and found that
interdisciplinary decision analysis efforts across multiple stakeholder
groups (including both professionals and lay people) were rare. It is
therefore worth re-examining existing social networks and constructs at the
local level to determine where such common understandings can be enhanced in
the context of bottom-up methodologies for flood risk analysis. This need is
particularly relevant when flood risk mitigation planning incorporates
climate adaptation goals surrounded by deep uncertainties (Downton et al.,
2005; Merz et al., 2015).
We address this gap with a case study survey of 50 professionals working on
riverine flood risk mitigation within Tompkins County, New York, USA. We
define “professional” and “practitioner” here as a subset of
flooding risk stakeholders within Tompkins County who have more agency around
flood hazard mitigation than that of a community member stakeholder. Our
operational definition of professional includes professionals in private
practice and research, elected government officials, appointed government
officials, governmental employees, and volunteer members of advocacy groups
with a water resources focus. This focus allows us to understand how the
social connectedness among flooding practitioners and their community
influences the flood risk mitigation planning and design process. We
specifically focus this research on understanding how well positioned this
network of flooding professionals is to begin a bottom-up vulnerability-based
flood hazard mitigation plan. The results of this research show that
professionals are not a monolithic category, as they vary in their knowledge
of historical hydrologic events, perceptions of existing flooding
vulnerability and risk, and perceptions of the need to incorporate future
climate estimates into flood risk mitigation planning and design at the
outset of a flood mitigation planning process. We compare survey results to
available hydrologic data to determine how classic approaches focused on
hydrologic data can be supplemented with socio-hydrologic information, and
identify opportunities for strengthening interdisciplinary networks.
Methodology
Study region: Tompkins County
Tompkins County lies within central New York, USA. Areas of high population
density are clustered within 15 towns and villages, each developed adjacent
to a fourth-order or higher stream. The county population is approximately
100 000 people across 1300 km2, with a median household income of
approximately USD 48 000 (United States Census, 2017) (about USD 10 000
below the 2016 median household income for the USA). Tompkins County
legislature is presently composed of 10 registered Democrats, 4 registered
Republicans, and 1 Independent (TC, 2018), suggesting a Democratic partisan
lean.
Federal Emergency Management Adjacency (FEMA) flood insurance rate maps, last
updated in 1996, suggest that 3749 parcels lie within the 100-year special
flood hazard area, of which 1874 are located within the City of Ithaca. From
1978 through 2012, 229 flood loss claims (6.7 claims yr-1) were
submitted through the National Flood Insurance Program (NFIP), totaling
USD 1 593 201 (∼ 46 900 USD yr-1) (TC, 2013). The Tompkins
County Hazard Mitigation Plan, established in 2013 to review county flood
losses and propose corrective actions, is updated annually.
Tompkins County contains four fourth-order or greater streams that are
tributaries to Cayuga Lake. An active National Weather Service (NWS) flood
stage has been established for USGS gage 04234000 (USGS, 2018), which is
representative of flooding within the City of Ithaca, a low-lying densely
populated community within Tompkins County. Based on annual peak flow
records, exceedance of the 2 m levees or discharge in excess of
120 m3 s-1 within Fall Creek (adjacent to the City of Ithaca) is
estimated to be a 9-year event (J. Knighton et al., 2017), though we note
that this estimate may vary depending on the period of record considered.
Tompkins County receives an average of 1000 mm of precipitation annually,
with 15 % as snowfall (NCDC, 2018). The county is approximately 45 %
forested, 45 % agricultural land use (row crops), and 10 % urban (Fry
et al., 2011). A shallow confining layer (0.5 to 1.5 m) leads to a prevalence
of saturation excess runoff (Easton et al., 2007). Regionally, surface runoff
is primarily generated during the spring following extratropical rain-on-snow
coincident with frozen or saturated soils and during the fall period of
tropical-moisture-derived precipitation (Knighton and Walter, 2016;
J. Knighton et al., 2017; J. O. Knighton et al., 2017).
Recent trends in gaged streamflow across the Northeast USA suggest a more mild
increase in extreme discharge relative to the conterminous USA (Slater and
Villarini 2016). Downscaled CMIP5 projections of future precipitation
(projected years 2015–2100) suggest a slight increase in air temperatures
and an associated increase in annual maxima precipitation intensity (Schoof
and Robeson, 2016; Ning et al., 2015). Inter-seasonal predictions of future
precipitation for Tompkins County show high variability, and potentially
inaccurate estimates of seasonal extreme rainfall related to numerical and
physical limitations of current general circulation models (GCMs) (Wobus et
al., 2017; J. Knighton et al., 2017). The difficulty in predicting future
extreme precipitation and strong influence of the land surface on flood
runoff (Ivancic and Shaw, 2015; J. O. Knighton et al., 2017) has yielded
projections of mild to no increase in future riverine flooding hazard in the
Northeast USA, often accompanied by high uncertainty (e.g., Hirabayashi et al.,
2013; Wing et al., 2018), or an average increase in risk, but with high
spatial variability (e.g., Wobus et al., 2017). Broadly, this region exists
with relatively high uncertainty with respect to future climate trends and
riverine flood frequency. We anticipate that this lack of a clear signal from
state-of-the-art climate and flood projections on the anticipated future
flooding risk may create added difficulty and ambiguity for local decisions
concerning the need for local climate adaptation.
Questionnaire design
Prior to designing the survey, we conducted semi-structured interviews
(methodology described by Hermanowicz, 2002) with 10 flooding professionals
within Tompkins County from January 2017 through August 2017. Each interview
was initiated with a series of general questions on the topic of flooding,
and shortly thereafter interviewees were encouraged to move the discussion
in their own direction of interest. The purpose of these interviews was to
develop an exhaustive inventory of themes concerning the challenges faced by
professionals engaging in group decision-making and to catalogue ideas about
flooding commonly held by flood risk mitigation practitioners. Common themes
included a professional's understanding of where flooding has occurred
frequently within the county, a professional's understanding of what
socioeconomic losses constituted a flood, a professional's perceived
community concern about shifting flooding risk under climate change, and
potential disagreement among professionals around the design goals of a
countywide flood mitigation project.
We distributed a questionnaire to community members who engage directly with
flooding through development of policy and legislation, science and
engineering, education, community outreach, and advocacy. Candidate
participants were identified by the Tompkins County Environmental Management
Council (EMC), the citizen advisory board to Tompkins County. The EMC's
varied experience, long-standing community connections, and formal liaison
role between the public, Tompkins County Planning Department, and Tompkins
County Legislature allowed it to make informed selections for this study. A
review of the final survey was performed by the Cornell University
Institutional Review Board and found to have no ethical implications related
to human subjects' participation. A draft of the survey is in the
Supplement.
Survey questions were a mixture of Likert-scale questions, multiple choice
selection, and open-ended response. The questionnaire was developed by the
authors and piloted with five members of the Tompkins County EMC. The
questionnaire was administered by email on 27 October 2017 via an online
platform. Four survey responses were delivered on paper due to limited
access to the internet.
The first goal of the questionnaire was to understand if historical
socio-hydrologic data are distributed broadly among flood professionals
within Tompkins County. The survey prompted recipients to enter anecdotal
information on historical flooding events, specifically, location, date of
event, magnitude (i.e., known high water elevations), and known economic
losses. The survey then focused on collecting information on participants'
perceptions of current regional flood hazard, risk, and community needs.
Individuals were asked to determine what forms of social or economic loss
constituted a flood. Individuals were then asked to report their perception
of current flooding frequency within their community, and what frequency of
flooding would be deemed acceptable. As recommended by Merz et al. (2014)
flood studies may benefit from a consideration of how unique weather types
and patterns impact flooding. In drawing this explicit link between global
climate and local weather we may better understand potential nonstationary
nature of flooding, and how and when climate adaptation should be considered
within flood hazard mitigation. The final section of the survey aimed to
determine an individual's understanding of how local and regional weather
drive flooding within Tompkins County. We asked several questions aimed at
understanding flood practitioners' perceptions of community knowledge and
desires for climate adaptation planning.
As will be discussed in the results, our survey population was comprised of a
relatively small working group of professionals within Tompkins County, New
York (n=50). Our focus on this specific population within Tompkins County
led to an inherently small sample size, though the surveyed population was
representative of a large proportion of the total population identified
(n=89). We therefore used qualitative interpretation of our survey results
in place of formal statistical tests. Our survey question and all results are
publicly available in Knighton (2018).
Hydrometeorological data analysis
Weather types and precipitation depth totals for historical events were
determined using the daily historical precipitation record (NCDC, 2018),
records of regional historical flooding (Johnson, 1936; Agel et al., 2015;
NCDC, 2018), a catalog of tropical storms (Roth and Weather Predictions
Center, 2012), and the personal account of Michael Thorne (City of Ithaca
Superintendent of Public Works, personal communication, 2018) to identify two
recent ice jam events. Return periods for extreme daily precipitation totals
were estimated with NOAA Atlas 14 (Percia et al., 2015).
We compare the spatial distribution of flooding as estimated from the FEMA
National Flood Hazard Layer (NFHL) 100-year floodplain (FEMA, 2018) and flood
practitioner reports.
Historical streamflow records were collected for Fall Creek (USGS, 2018) for
the period of 1925 to 2018. We use an annual block maxima approach to
identify the significant floods within the publically available long-term
hydrologic record. We compare reports of historical flooding from community
members to understand how we can best develop a complete record of county
flooding.
Results
Response rate
The survey was distributed to 89 professionals, of which 50 responded
(response rate of 56 %). Individuals were asked to self-sort into one of
six possible roles: community planning (n=8), education and outreach
(n=8), local government leadership (n=9), policy (n=5), advocacy
(n=9), and Engineering, Science and Research (ESR; n=11). We first
asked flooding professionals whether they believed they had a good
understanding of flood risk mitigation, to which 52 % indicated they had a
strong grasp of the subject, 38 % knew of a professional who could inform
them, and 10 % were not knowledgeable on the subject (two policy, one
education and outreach, one government, and one advocacy).
Spatial distribution of socioeconomic flood losses
Anecdotal reports of flooding were compiled to provide a spatial depiction
of commonly flooded locations within Tompkins County (Fig. 1), as recalled
by research participants. Anecdotal flood reports by community members
demonstrate that flooding is a countywide issue, with the most commonly
recollected flooding centered on the most densely populated areas. The
reported locations of flooding cover substantially more locations than those
falling within the official 100-year special flood hazard area, as depicted
on FEMA's flood insurance rate map (FEMA, 2018). This is typical of many
flood-prone areas; over 20 % of flood insurance claims come from
losses outside of currently mapped high-risk zones (where flood insurance is
available but not mandatory) (FEMA, 2015).
Spatial distribution of survey-reported flooding within Tompkins
County (filled circles) and FEMA 100-year flood plain (dark blue).
![]()
NWS flood stage on Fall Creek in Ithaca is estimated to be exceeded with a
9-year recurrence interval. However, reported dates of flooding events
(Table 1) suggest that for much of Tompkins County, professionals have
collected information on negative socioeconomic consequences from events
that are hydrologically more frequent than the 9-year Fall Creek baseline,
suggesting that primary sources of hydrologic data alone do not provide a
complete depiction of flooding hazards within the county.
Reported historical flooding events. Rainfall totals are the
maximum daily precipitation (NCDC, 2018). Return periods are determined from
NOAA Atlas 14 (Percia et al., 2015).
Date
Rainfall
Return period
Weather type
(m/d/yr)
(cm day-1)
(yr)
4/18/1905
1.2
< 1
–
6/3/1905
4.3
< 1
–
6/17/1905
4.6
< 1
–
7/3/1905
4.8
< 1
–
7/8/1935
20.0
> 1000
Tropical/local convective rain*
11/3/1954
4.0
< 1
Hurricane Hazel
6/23/1972
9.0
10
Hurricane Agnes
10/28/1981
12.9
25
Local convective rain
1/19/1996
4.7
< 1
Rain on snow
4/3/2005
5.7
2
Rain on snow
9/8/2011
11.3
25
Tropical Storm Lee
1/11/2014
0.0
< 1
Ice jam
6/14/2015
10.4
10
Local convective rain
7/1/2017
0.9
< 1
Local convective rain
1/12/2018
2.4
< 1
Ice jam
* Statewide flooding was reported to result from a mixture of a
tropical moisture export and local convective rainfall.
Weather types assigned to each reported historical flooding event indicate
that flooding has been induced by local extreme convective precipitation,
tropical-moisture-derived precipitation, extratropical rain-on-snow/snowmelt,
and ice jams. Weather types for events prior to 1930 were not
identified due to inconsistency among available sources.
Defining flooding by socioeconomic losses
Exploratory interviews with community leaders suggested that there were 13
socioeconomic losses that individuals commonly used to define a past
flooding event (Table 2). The survey presented these 13 possible flooding
losses and asked flooding practitioners to define which types of loss
constituted a flood. Surveyed individuals also had the option to write in
their own preferred definition.
Results of which socioeconomic losses were considered a flooding
event. Types 14 and 15 are write-in responses.
Type
Description of flood
Number of
Responses
1
Loss of life
29
2
Damage to private structures
32
3
Displacement of people
34
4
Damage to vehicles
30
5
Damage to public property
34
6
Inundation of public roads
34
7
Flow over private property
21
8
Backed up culverts
25
9
Loss of streamside vegetation
29
10
Stream flow out of channel banks
32
11
Substantial erosion in the stream channel
24
12
Minor erosion in the stream channel
4
13
Any flow greater than baseflow
11
14
Discharges above an n-year recurrence interval
4
15
Any negative impact to resources
1
No single type of reported flood was held common to all individuals surveyed
(Table 2). The belief that negative flood consequences related to minor
erosion in the stream channel and flow above baseflow constituted a flood
was only held by a few respondents. Individuals in planning, government, and
advocacy were more likely to hold a broad definition of flooding, whereas
individuals in outreach, policy, and ESR tended to hold a narrower definition
of flooding (Fig. 2). About 40 % of ESR responses opted to use a
write-in definition based on numeric description of flood frequency, for
example, “any flow exceeding a storm recurrence interval 100 years or greater”.
Socioeconomic losses that defined flooding events by discipline
(Table 2 subset by discipline). Values in parentheses indicate the number of
respondents who did not offer an answer.
![]()
Perceptions of current and desired flood frequency
Estimates of the current flooding return period for Tompkins County varied
slightly by discipline; however, most estimates were below the baseline
flood return period established for Fall Creek of the 9-year event (Fig. 3a).
The desired reduction in flooding return period varied considerably by
discipline. The median ESR, community planning, and outreach response
suggests that the expected flood frequency after mitigation efforts should
be slightly higher than current flooding hazard (Fig. 3b). The median
responses from governmental employees working on legislation and policy
desired flood frequency to be reduced to the 100-year event, suggesting a high
level of disagreement between disciplines on anticipated outcomes of flood
hazard mitigation. This difference could potentially be due to governmental
focus on well-established floodplain thresholds (FEMA, 2018) versus
perceptions of the physical limits of hydrologic alteration.
Estimated (a) current flood loss return period and (b) desired return
period resulting from flood hazard mitigation efforts. Reports of desired
frequency above 100 years are presented as 100-year for visualization. Red
lines indicate the median.
![]()
There was strong consistency in the perception of current flooding risks
(Fig. 3a), though the spatial distribution of affected locations was
highly individual (Fig. 1). This result suggests that individuals within
Tompkins County have a consistent understanding of the frequency of these
socioeconomic losses; however, there may not be a strong social network for
communication of risks as knowledge was spatially constrained by discipline.
The Tompkins County hazard mitigation plan contains county records of
historical events. This record is presently derived exclusively from federal
sources (TC, 2013; NCDC, 2018), with no formal mechanism to collect and
archive anecdotal accounts of flooding within the community. In the absence
of a centralized county database to collect and share personal experiences
among professionals, an individual's primary spatial knowledge of flooding
may be most derived from their own individual experiences, related to place
of residence or locations of previous work. Local flood hazard mitigation
plans in the USA typically suffer from low quality as they are primarily
developed as a requirement to maintain access to federal funding instead of
functional plans for risk mitigation (Lyles et al., 2014).
We compare aggregated reports of flooding (Fig. 1, Table 1) to the
long-term historical record of Fall Creek (Fig. 4a and b). Results
indicate that aggregated records from all community members successfully
identify the substantial flooding events for which Fall Creek overtopped the 2 m
levee or greatly exceeded the channel capacity of 120 m3 s-1. It is
worth noting that prior to 1970 annual peak flow frequently exceeded 120 m3 s-1,
yet did not exceed the current NWS flood stage of 2 m. This
could be due to a change stream discharge monitoring, or a physical change
in the stream rating curve. It was beyond the scope of this study to
investigate the cause of the shift in hydrologic response.
The continuous hydrologic data of Fall Creek discharge and stage do not
successfully identify all reported flooding events. Though Fall Creek is the
largest watershed within Tompkins County (contributing drainage area of 325 km2)
and has been monitored continuously for over a century, a purely
hydrologic-hazard-based assessment considering only primary hydrologic
measurements does not provide a complete picture of flooding across the
county. The large contributing drainage area (time of concentration
∼6 h) results in a hydrologic system sensitive only to
weather events on the order of 6 h or longer. Continuous hydrologic
monitoring records contain accurate information at the location of
measurement, which is useful for developing flood frequency curves and
hydrologic models. Survey reports of flooding, while less quantitative,
benefit from broader spatial coverage and often include anecdotal accounts
of socioeconomic losses (e.g., Marjerison et al., 2016). Together, primary
hydrologic measurements (e.g., continuous stream depth) and anecdotal survey
reports of flooding losses help to develop a more complete picture of the
flooding hazard and risk profile within Tompkins County.
Climate–flood linkages
Surveyed individuals were asked to report which type of weather mechanism (1
– extratropical system, 2 – local convective rainfall, 3 – tropical moisture
derived rainfall, 4 – snowmelt events, 5 – ice jam) contributed to flooding
within Tompkins County, with the option to write in flooding mechanisms.
Optional write-in mechanisms included “Sever [sic] thunderstorms,”
“dramatic increase in stream levels for any reason”, and “extreme rain”.
Responses and write-in results suggest that there is a general understanding
that rainfall and air temperatures relate to flooding events; however, there
was no strong agreement within any group that a given weather mechanism
contributed directly to flooding outside of county planners agreeing that
local convective rainfall contributed to flooding (Fig. 5). This result
suggests that practitioners may have a limited understanding of
weather–flood linkages.
Annual block maxima: (a) peak instantaneous discharge and (b) peak
stage for Fall Creek. Blue dots indicate that the event was identified as a
flood by at least one survey response. Red lines indicate the hazard
thresholds for (a) exceeding channel capacity, and (b) overtopping the 2 m
levee within the City of Ithaca.
![]()
Reported flooding mechanisms that contribute to local flooding: 1 –
extratropical system, 2 – local convective rainfall, 3 – tropical moisture
derived rainfall, 4 – snowmelt events, 5 – release of ice jam. Values in
parentheses indicate the number of respondents who did not offer an answer.
![]()
It is worth noting that on 11 January 2018, while the online survey was
active, a joint snowmelt–ice jam event caused nuisance flooding throughout
the City of Ithaca (Ithaca Times, 2018). We anticipated this specific event
would result in a strong agreement among professionals on the relationship
between ice jams, snowmelt, and flooding due to recency bias. However, this
was not reflected in the survey results, with only 48 % (n=24) and 38 %
(n=19) of all responses suggesting snowmelt and ice jam release, respectively, were important flooding mechanisms within Tompkins County.
Anticipation of the need to incorporate climate adaptation into flood risk
planning, as well as anxiety around “community perceptions” and “public
opposition to planning for climate change” were common themes that emerged
during the 2017 semi-structured interviews. Flooding practitioners were
asked which direction they anticipated future flooding risk within Tompkins
County would move. The majority of individuals, 30, believed that flooding
risk would increase, and 15 responded that they were not sure (Table 3).
Perceptions of future riverine flooding risk within Tompkins County
by flooding practitioners.
Not
Less
Same
More
sure
risk
risk
risk
Community planning
0
0
0
8
Education and outreach
1
0
1
5
Local government leadership
2
0
1
5
Policy development
3
0
0
2
Public advocacy
3
0
0
6
ESR
4
0
3
4
Total
15
0
5
30
Perceptions of community desire to implement climate adaptation
planning in flood risk mitigation.
Note
Definitely
Probably
Probably
Definitely
sure
not
not
yes
yes
Community planning
2
0
1
5
2
Education and outreach
1
0
0
4
0
Local government leadership
1
0
1
4
1
Policy development
5
0
0
0
1
Public advocacy
2
0
2
1
0
ESR
5
0
3
1
2
Total
18
0
7
15
6
Surveyed professionals were asked if they perceived a community desire to
implement climate adaptation practices in flood mitigation planning. The
result here was less clear, with 18 responding they were not sure, 7
probably not, 15 probably yes, and 6 definitely yes (Table 4). The Tompkins
County Planning Department acknowledges hazards posed by climate change and
the need for climate adaptation; however, the current county plan
only focuses on maintaining existing natural and built infrastructure. No
large-scale flood mitigation projects
currently incorporate climate adaptation in planning and design (TC, 2015). There was some disagreement
among disciplines on public preference for climate adaptation, with ESR and
public advocacy perceiving less interest, and outreach and government
perceiving more interest (Table 4).
We next asked practitioners to report their perceptions of the level of
climate science knowledge of residents of Tompkins County. Results were
divided, with 16 responding that they were not sure, 16 believing that
residents had basic knowledge, and 10 believe strong knowledge. Results were
not substantially different among the disciplines (Table 5).
Optional write-in responses
At the conclusion of the survey professionals were given the option to
provide any additional information or thoughts on the topic beyond the
survey responses provided. We summarize here the results of these
submissions. Though we do not aim to interpret these results, they can offer
important insights beyond what was captured in the survey questions.
Five professionals supplied optional comments in which they said that they
had little knowledge of community perceptions and expressed difficulty in
answering these particular questions, with one professional suggesting that
community perception was perhaps too broad to accurately define with one
single answer. Three responses suggested that they had a good understanding
of community perceptions through involvement with county government and
expressed that there was a willingness among the Tompkins County public to
involve climate adaptation practices in flood risk mitigation. Four
responses attributed recent flooding events to improper control of existing
flood mitigation infrastructure by local, state, and federal government. One
response listed the ecological benefits of flooding, and suggested that
rather than seek mitigation opportunities to control floods, we seek to
adapt human behavior.
Discussion
Definitions and extents of flooding
“How do we define a flood event?” and “What are the community
vulnerabilities?” appear to be critical questions for which establishing
consensus may prove difficult. Formal definitions of flooding used within
the ESR community, which focus on quantitative flood frequency, are more
hazard-oriented, whereas the definitions preferred by planning, government,
advocacy, and outreach utilized the socioeconomic losses to define
flooding. A focus on hazard may simplify engineering design and planning;
however, it can potentially be too limiting to properly address other
stakeholder needs. A conceptual disconnect on the definition of flooding
points to issues with the core problem statement that coalitions of flooding
professionals are self-organized to solve.
Perceptions of general community knowledge level of climate science
and adaptation.
Not
Little
Basic
Strong
sure
knowledge
knowledge
understanding
Community planning
1
0
3
5
Education and outreach
1
0
3
1
Local government leadership
1
2
2
2
Policy development
5
0
1
0
Public advocacy
1
1
2
1
ESR
5
0
5
1
Total
16
3
16
10
While formal guidelines exist at the US national level (Obama, 2015), these
definitions often conflate risk and hazard. Formal federal definitions in
the USA commonly focus on hydrologic hazard posed by a static water surface
elevation, as in FEMA flood insurance rate maps (FEMA, 2018), neglecting
hazard associated with discharge velocity, duration of inundation, and
suspended material. In addition, the focus on hydrologic hazard can obscure
from view the uneven socio-spatial distribution of exposure and
vulnerability, which often aligns with prevailing axes of inequality along
lines of race and class. The same flood event may be a nuisance for an
affluent community and complete devastation for a poor one. Differences in
reported flooding locations between the survey and established FEMA flood
zone maps (Fig. 1) could potentially be explained by the distinction
between hazard- and vulnerability-based definitions of flooding. For
instance, when floods hit particularly vulnerable communities, their impacts
may be dramatized in ways that allow them to retain salience in the minds
and memories of stakeholders.
Gober and Wheater (2015) propose a broad reconceptualization of flood risk
analysis that accounts for emergent and complex interactions between water
and society including the role of social memory in magnifying risk perception
(Di Baldassarre et al., 2015) or actual risks (Collenteur et al., 2015),
public perception, policy limitations, windows of opportunity, and an
imperfect flow of knowledge through society. Such conceptual models may
extend unrealistically beyond the capacity of local flood professionals,
though we can consider that expanding the definition of flooding beyond the
traditional flood frequency realm could allow practitioners to more easily
realize the benefits of bottom-up flood risk analysis frameworks. Flooding
practitioners increasingly face decisions about the appropriate level of
abstraction when defining socio-hydrology problem statements (Troy et al.,
2015; Blair and Buytaert, 2016). Preemptively limiting the complexity of a
problem in the planning stage possibly introduces new vulnerabilities in the
form of “surprise” (Merz et al., 2015). It is possible that surprise can be
avoided or reduced, even without consensus on community vulnerability through
encouraging interdisciplinary discourse (Di Baldassarre et al., 2016).
Perceptions of climate–weather–flood linkages and climate adaptation
planning
Researchers are moving towards reframing flooding risk from the perspective
of the distinct atmospheric mechanisms that induce floods (Merz et al., 2014)
in an effort to simplify the interpretation of how mesoscale global trends
(e.g., global climate change, decadal global oscillations) influence local
weather patterns and subsequently local flooding. Within the Northeast USA,
this problem is often expressed as nonstationary rainfall intensity and
warming air temperatures that may drive flooding hazards (e.g., DeGaetano,
2009).
Current climate projections of summer extreme precipitation for Tompkins
County predict increases in air temperatures and precipitation intensity
(DeGaetano, 2009), though regional estimates of future flooding hazard are
accompanied by high uncertainty (J. Knighton et al., 2017). While warming air
temperatures are likely to enhance the melt rate of the standing snowpack, it
could also limit total snowpack accumulation, tempering the effect of climate
change on winter flooding (J. O. Knighton et al., 2017). Similarly, during
the summer season warming air temperatures will likely result in reduced soil
moisture, tempering extreme runoff. Though an increase in the intensity of
precipitation is often expected to translate directly into increased
discharge (e.g., Trenberth, 2011), this outcome is not necessarily expected in
the Northeast USA (Ivancic and Shaw, 2015).
Emerging methods of flood analysis attempt to limit the need for coarse
interpretations of changes to extreme rainfall as projected by highly
uncertain GCMs, by drawing explicit connections from climate change to weather
types and then local flood risk (e.g., J. Knighton et al., 2017). While these
approaches can potentially better allow local professionals to address
climate adaptation within this high uncertainty decision space, they rely on
a baseline of knowledge around established climate–flood linkages. As
demonstrated in Table 1, individuals reported historically significant
flooding events. Through a simplistic weather-typing analysis, we determined
that these events encompassed the atmospheric mechanisms of extratropical
rain-on-snow events, snowmelt, local convective precipitation, tropical
storms, and release of ice jams. The reported knowledge of weather systems
that induce flooding suggested that professionals did not have a strong grasp
on climate–flood linkages, particularly among those engaged in policy
development (Fig. 5). Many write-in answers indicated that individuals often
did not consider the atmosphere–land surface complexity beyond that of a
simple input–output system.
Individuals were asked about their beliefs on future flooding risk and two
questions pertaining to perceptions of community knowledge and preferences
around climate adaptation planning. First, responses indicated that
approximately one-third of all professionals surveyed were unsure about
community perceptions of climate science and the importance of implementing
climate adaptation practices into flood hazard mitigation. Practitioners
agreed that communities had some understanding of climate science that was
either basic or good. Practitioners disagreed on the community desire to
incorporate climate adaptation practices into flood mitigation planning,
which may reflect the politically contentious nature of climate change in US
political discourse and policymaking more generally. Previous research has
demonstrated disconnections between public climate–flood risk perceptions
and expert opinion (e.g., Hamilton et al., 2016; Ogunbode et al., 2017;
Shepard et al., 2018). In this case study, professionals disagreed on both
community perceptions (Table 4) and climate education (Table 5). While there
may be conceptual differences in the beliefs of lay people and experts, we
possibly take for granted that experts effectively understand and represent
complex community needs and beliefs in the flood mitigation planning process.
Perceptions of flooding expertise and social networks of
professionals
As previously described, 52 % of flooding professionals within Tompkins
County reported that they had a good understanding of flood risk mitigation,
and 40 % reported that they did not, but had a resource who could inform
them, suggesting a well-connected network among professionals. Survey
results indicate strong disagreement among flooding professionals as to
which socioeconomic losses constitute a flood (Table 2), incomplete
knowledge of the spatial extent of flooding within Tompkins County (Fig. 1),
disagreement on anticipated community needs (Fig. 3b), weak
understanding of climate–weather–flood linkages (Fig. 5), and some
disagreement on community perceptions on climate adaptation (Tables 4 and
5).
The results of this research suggest a case in which practitioners may believe
that they are well informed and share commonly held beliefs, while in
reality the network of flooding professionals is less well established or
holds divergent perceptions or terminologies regarding flooding. Instances
in which decision makers believe their perspective is commonly held can open
problematic possibilities and new vulnerabilities. Flood risk is a
particularly difficult problem to address, owing to the infrequent nature of
hydrologic extremes, nonlinear relationship between rainfall, runoff, and
exposure, and potential socioeconomic feedbacks that develop between
society and flood hazards (e.g., the “levee effect”; Collenteur et al.,
2015). The large number of professionals who reported that they have a
strong grasp of the subject could potentially indicate a susceptibility to
cognitive biases influencing flood mitigation planning and design
(Merz et al., 2015). For example, practitioners may be particularly susceptible to
overconfidence and confirmation bias with respect to their currently held
understanding of existing flood risks.
In developing the current estimate of flood frequency from aggregated reports
of historical events, we reached a relatively accurate appraisal of flooding
hazard (Figs. 1 and 4). This result demonstrates that flood mitigation can
benefit strongly if the knowledge of independent institutions is properly
leveraged (e.g., Serra-Llobet et al., 2016). In other cases, this research
identifies gaps in the social networks of flooding professionals.
Inflexibility among professionals to consider variations in project goals or
risk tolerance can lead to undesirable flood mitigation outcomes (Downton et
al., 2005). Disagreements on the definitions of flooding and community
preferences for flood mitigation are perhaps expected results at the
inception of a flood mitigation project and present opportunities for
flooding professionals to engage more directly across disciplines.
Implementation of emerging robust decision-making frameworks (e.g., Prudhomme
2010; Brown et al., 2012) commonly dictate that a bottom-up approach be taken
with regard to defining the problem and weighting desirable outcomes. These
approaches place significant emphasis on understanding community
vulnerability (Pielke et al., 2012) as the initial objective. Our results
suggest that professionals engaging in flood mitigation within Tompkins
County are in agreement about exposure (Fig. 3a), but have some
disagreement around flood vulnerabilities (Fig. 1) and anticipated needs
(Fig. 3b). Substantial differences emerged in how the six groups of
professionals responded to certain questions. ESR preferred to define
flooding from a classic perspective, focusing on hydrologically relevant
metrics (e.g., the n-year discharge and exceedance of bankfull discharge),
whereas individuals in planning, government, and advocacy defined flooding
very broadly, from a bottom-up perspective of the associated socioeconomic
losses. This particular result is not surprising as there is substantial
variation in professional norms among the participating flood mitigation
roles owing to education, experience, and their respective
audiences/constituencies.
Distinct institutions working on flood mitigation may engage differently
with primary and secondary data sources. For example, engineering, science,
and biophysical research groups are typically focused on the collection and
interpretation of meteorological and hydrologic data, while those engaged in
public outreach may be more responsible for the collection and
interpretation of socio-hydrologic data (i.e., historical flooding economic
losses, migration), and governmental organizations may best understand
policy implications of flood investments. The organization of social
networks of professionals working in the realm of flood risk reduction
remains a fairly unstudied subject, though it is particularly relevant as
communication of highly uncertain information among professional disciplines
remains a challenging task (e.g., Pappenberger et al., 2013; Morss et al.,
2005; Downton et al., 2005). Social organizations may lead to the development
of risk perception networks, whereby clusters of individuals share the same
risk perception (Scherer and Cho, 2003). Conversely, dissimilar groups may
tend to develop less connected relationships, where information quality is
only high within a subset of organizations and shared only when necessary.
Finally, disagreement surrounding the role of primary data and expression of
hydrologic uncertainty can lead to suboptimal solutions or inaction (e.g., Downton et al., 2005).
Independently operating institutions at the local level commonly hold
individual goals (Butler and Pidgeon, 2011; Measham et al., 2011), which may
be distinct from the collective flood risk mitigation goal. Given their
orientation and obligations toward voters and taxpayers, governing
organizations are commonly driven by a desire for continued local economic
(re)development (Molotch, 1999) as well as preoccupied with concerns around
“blame” and “credit” for social outcomes (Leong and Howlett, 2017). This
sometimes results in a narrow temporal focus and decisions that are made
with limited consideration for decadal or longer processes (Gober and Wheater,
2015). Private firms may be concerned with maintaining profitable contracts,
job security, ethics, and liability. NGOs and advocacy groups may be
concerned with developing and maintaining public interest (Lorenzoni and
Pidgeon, 2006). Research organizations are often concerned with developing
new science and engineering techniques, with a tendency to avoid advocacy
and maintenance of impartial stances on controversial subjects (Singh et al.,
2014).
Disconnections among the network of professionals could also be related to
compartmentalization of urban and nonurban problems. Tompkins County
contains 1 urban center (City of Ithaca) and 14 less densely populated
towns and villages. Observed differences in flooding perception among
professionals could stem from the differences in community structures and
land use within Tompkins County. Urban fill substantially alters the
hydrologic response of urban areas, potentially confounding the relationship
between rainfall and runoff (Knighton et al., 2014) that is not experienced
in less developed areas with native soils. We note that reports of flooding
and flood loss claims are greatest within the City of Ithaca (TC, 2013), possibly
suggesting that a combination of regional hydrology and population density
is driving the perception of flooding risk and the need for mitigation
within the City of Ithaca; however, Marjerison et al. (2016), through a
broader spatial study, suggest that local population density may not be a
sufficient regional factor to determine flooding frequency perception.
Beyond density, urban vulnerabilities and problems are often considered to
be distinct from nonurban areas due to greater sociopolitical
institutional complexity and less deliberate community planning (Zevenbergen
et al., 2008).
Strong networks of professionals have been demonstrated to simplify
decision-making, even in challenging situations (e.g., Bracken et al., 2016).
Efforts to refine the flood hazard and risk profile of Tompkins County that
make the effort to involve members of all disciplines explicit may build
trust and communication among practitioners (Morss et al., 2005) and the
public. Work that strives to leverage the knowledge of independently
operating institutions and the public to define the problem statement and
guide mitigation practices ultimately improves outcomes (e.g.,
Serra-Llobet et al., 2016; Edelenbos et al., 2017). Public willingness to
take risk reduction measures has previously been attributed myriad variables
including actual risk and societal norms (Lo, 2013), income (Lo, 2014),
exposure to or protection from past events (Di Baldassarre et al., 2013), and
trust in expert opinion (Wachinger et al., 2013). Though disagreement between
flooding practitioners and the public may occur, establishing these social
connections could be an important step towards establishing trust, building
public support for mitigation projects, and opening opportunities to collect
socio-hydrologic data which could improve flood mitigation planning and
design.
Broader impacts
European nations (e.g., Næss et al., 2005), the UK (e.g., Brown and
Damery, 2002), African and Asian nations (e.g., Huntjens et al., 2011), and
the USA face the prospect of enacting water governance and developing policy
within a changing climate. Recent research has focused heavily on the
shortcomings of top-down approaches to flood hazard mitigation as enacted by
a variety of governmental organizations. For example, Brown and Damery (2002)
explore the structural issues present in a top-down governance scheme
within the UK and conclude that a focus on hazard leads to improper problem
definitions and “undersocialized” solutions. Subsequent research has
proposed that governance leveraging both top-down and bottom-up schemes could
improve the efficiency with which a nation incorporates societal
vulnerability information into policy (e.g., Pahl-Wostl et al., 2010),
thereby lowering societal risk.
Recent applications of decision analysis frameworks for flood mitigation
within Europe (e.g., Evers et al., 2018) and the USA (e.g., J. Knighton et al.,
2017) highlight the technical potential of these approaches, yet as a
research community, we have not fully explored these frameworks outside of a
handful of carefully controlled case studies. As more attention is being
given to bottom-up approaches as a potential panacea for flood hazard
mitigation, a critical assessment of governmental organizations,
institutions, and practitioners becomes more necessary to explore
possibilities for new unforeseen vulnerabilities that may emerge.
It is worth discussing that several aspects of our study catchment may have
implications for how these results can be interpreted and applied to other
locations. Our research focuses exclusively on a network of professionals
within Tompkins County, New York (USA), with a distinctly bottom-up structure
for flood governance. The county is moderately sized (population of 100 000)
and experiences moderate flooding (∼9-year recurrence
interval for socioeconomic riverine flood losses).
With respect to county population and flood loss frequency, we can possibly
anticipate that the social connectedness of professionals would increase with
increasing community size and flood frequency. Both larger populations and
increased frequency of hazards could lead to more complete records of
historical floods and increased connectivity among professionals. It is possible that increased
exposure to flood frequency would reduce cognitive biases (e.g., Merz et al.,
2015), leading to an “adaptation effect” (Di Baldassarre et al., 2015).
Conversely, less exposure, as would be expected with a smaller population and
less flooding risks, would be expected to decrease social understanding of
flooding risks (Collenteur et al., 2015) and less established networks of
professionals.
Broadly, there are several aspects of this research which may allow our
results to be more globally applicable. First, the local governmental and
institutional organization of this case study mirrors that of other US and
European cities, which suggests the possibility of similar institutional
vulnerabilities associated with local governmental, private, and
community organizations. For example, Bracken et al. (2016), studying flood
management within the UK, describe a similarly loose coalition of experts
from governmental and nongovernmental organizations to those observed
within our research. Second, the reliance of bottom-up decision analysis
frameworks on networks of people exists independent of local governmental
structure, and would likely contribute similar vulnerabilities as we have
observed. Merz et al. (2015) review a series of historical floods across
Europe that resulted in increased devastation as a result of surprise.
Surprise is then attributed by Merz et al. (2015) to cognitive biases
“hardwired in the human brain.” It is possible that the divergent
perceptions and definitions that we observe among Tompkins County
professionals are indicative of universal human traits, rather than simply a
local phenomenon. Finally, our methodology is easily adapted, and could be
applied to uncover new vulnerabilities in parallel governmental structures
in other nations.
Conclusion
Complex decisions involving highly uncertain inputs can potentially be
reduced to more manageable problems, given the recent improvements to, and
wider application of, bottom-up vulnerability-based decision analysis
frameworks. This family of methodologies relies heavily on the precise
definition of system vulnerabilities so that uncertain inputs (e.g., climate
projections of extreme rainfall) can be readily mapped to a
stakeholder-relevant metric of concern. In flood risk analysis these
frameworks are proving particularly useful as they may help to avoid debates
around the reliability of future climate projections by placing this
information in the context of sensitivity to stakeholder-relevant outcomes.
In cases in which risk sensitivity to climate variability is low, discussions
on the reliability of climate projections can be readily avoided. Further,
there is an emerging consensus that flood risk analysis should directly
consider the unique characteristics of the atmospheric mechanisms that
induce flooding. Research advances have demonstrated that explicitly
separating storms by weather types allows for stronger inferences on future
flooding hazards.
Flood hazard and risk mitigation in the USA is often carried out at the local
level by an informal collection of governmental, nongovernmental, private,
and academic institutions. Previous research has observed variations in how
institutions approach flood hazard, owing to differences in educational and
professional backgrounds, variations in their relationship to
socio-hydrologic data, and which goals or outcomes are deemed desirable.
Given these discrepancies, there is a strong need to review differences
among disciplines to ensure that emerging bottom-up vulnerability-based
frameworks can be readily incorporated into local planning efforts.
Our research demonstrates that there are broad differences in belief among
practicing professionals within Tompkins County, New York, USA, as to what
socioeconomic losses constitute a flooding event and spatially
disaggregated knowledge of historically flooded locations. There was strong
agreement on the frequency of flooding experienced by residents, but
disagreement around the desired level of protection from a flood mitigation
effort. These results suggest that there is some variance among flooding
professionals in the definition of community vulnerability to flooding.
Undisclosed or unknown disparities in perceptions among flooding
practitioners could serve as barriers to successfully implementing
vulnerability-based frameworks for decision analysis. Developing strong
definitions of flooding vulnerability may not require explicit agreement
among all practitioners, but rather venues that allow for productive
processes of deliberation. These venues necessarily involve multiple,
diverse stakeholders, whose input shapes an outcome that all parties can
agree is procedurally fair and acceptable.
Explicit consideration of climate–flood linkages showed similar barriers
based on practitioners' knowledge. Professionals identified historical
events induced by five unique mechanisms, but failed to identify these types
of weather events as important causes of local floods. Among the survey
results there is an intuitive sense that intense rain causes flooding,
though it is possible this limited understanding prevents conceptual
connections of local events to regional and global climate patterns. These
discrepancies could serve as a barrier to implementing important advances in
flood risk engineering that aim to use relevant climate projections to
inform local planning.
In aggregate, the knowledge base of the Tompkins County flood practitioners
served to provide a well-defined picture of community vulnerability and
perceptions, though the beliefs of individuals varied. Previous research
suggests that collaborative efforts can work to improve connections between
social networks of experts and lay people. This research demonstrates the
need for interdisciplinary research, planning, and design throughout flood
risk mitigation and climate adaptation planning to maintain strong social
connections, not just between lay people and experts, but among experts.