LETTER • OPEN ACCESS You may also like

- Vulnerabilities—bibliometric analysis and

A systems network approach for climate change literature review of evolving concepts
Carlo Giupponi and Claudio Biscaro
vulnerability assessment - An institutional approach to vulnerability:
evidence from natural hazard
management in Europe
To cite this article: Nathan S Debortoli et al 2018 Environ. Res. Lett. 13 104019 M Papathoma-Köhle, T Thaler and S
Fuchs

- Quantitative assessment of ecosystem

vulnerability to climate change:
methodology and application in China
View the article online for updates and enhancements. Jiangbo Gao, Kewei Jiao and Shaohong
Wu

This content was downloaded from IP address 182.76.70.66 on 25/08/2023 at 14:38

 Environ. Res. Lett. 13 (2018) 104019 https://doi.org/10.1088/1748-9326/aae24a

LETTER

A systems network approach for climate change vulnerability

assessment
OPEN ACCESS

RECEIVED
28 September 2017
Nathan S Debortoli1 , Jesse S Sayles1, Dylan G Clark and James D Ford2
REVISED
17 September 2018
McGill University, Department of Geography, 805 Sherbrooke Street West, Montreal, QC H3A 0B9 CA, Canada
1
These authors contributed equally to the project.
ACCEPTED FOR PUBLICATION 2
Current address: University of Leeds, Priestley International Centre for Climate, Leeds, United Kingdom.
18 September 2018
PUBLISHED
E-mail: [email protected]
15 October 2018
Keywords: climate change, multiplex network, Inuit, Canadian Arctic, vulnerability indices, vulnerability asessment, human-environment
system
Original content from this
work may be used under Supplementary material for this article is available online
the terms of the Creative
Commons Attribution 3.0
licence.
Any further distribution of Abstract
this work must maintain Vulnerability to climate change is a product of biophysical and social dynamics. Assessments of
attribution to the
author(s) and the title of community or regional vulnerability, however, often focus on quantitative infrastructure and
the work, journal citation
and DOI. environmental assessments, or qualitative assessments of a community’s social dynamics and
livelihood activities. A dearth of integrated quantitative assessments is a major barrier for decision
makers who require quantitative outputs and indicators, which can measure where vulnerability is
most severe and can be linked to climate projections. Our framework and analysis helps address such
gaps by identifying variables to build climate change vulnerability indices, which we pilot here
focusing on Inuit communities in the Canadian Arctic. We start with a systematic literature review of
community-based vulnerability studies and assess relationships among 58 social and biophysical
variables. We then use multiplex network analysis to determine how social and environmental
variables interact among and within the key component of vulnerability: exposure, sensitivity, and
adaptive capacity. We identify several structurally important variables that interact within and across
the three dimensions of vulnerability. This method is transferable as an integrative means of
understanding not only the direct causes of vulnerability, but also relations that are less tangible. The
approach of multiplex network analysis can be a building block to ongoing development of
vulnerability indices within the human dimensions of climate change field.

Introduction multiscale dynamics of vulnerability to climate change

[4–6].
Climate change is widely recognized as one of the major Community-level studies have been a central fea-
challenges of this century. Many regions and commu- ture of vulnerability research, and many focus on social
nities are already experiencing the effects of climate variables. There is, however, strong demand for inte-
change with shifts in precipitation patterns, loss of grated quantitative regional vulnerability studies that
coastline, and increases in frequency of severe weather consider both social and physical variables [7] and
events [1]. To help communities prepare for these demand by government officials for indicators to do so
changes, vulnerability research seeks to identify and as outlined by Canada’s Expert Panel on Climate
characterize the various interacting climatic and non- Change Adaptation and Resilience Results [8]. The
climatic factors which create susceptibility to harm potential to link downscaled climate model data with
[2, 3]. Community vulnerability case studies have been vulnerability assessments makes such assessments even
widely used for adaptation and disaster risk reduction more desirable [9–12]. With over 25 years of vulner-
planning, climate change cost analysis, and delineation ability case studies and literature summarizing com-
of relationships between social and physical environ- monalities and trends throughout the field, there are
ments, contributing to a growing knowledge of the enough data to begin moving beyond community-level

© 2018 The Author(s). Published by IOP Publishing Ltd

Environ. Res. Lett. 13 (2018) 104019

analysis in heavily studied regions, although few meth- dependent variance becomes increasingly complex.
ods have been proposed for upscaling integrated vul- Here, we apply network analytics as a means of char-
nerability case studies. Addressing this key acterizing how different socio-economic, environ-
methodology gap, we propose the use of multiplex net- mental, and climatic indicators interact to create
work analysis to draw together data from vulnerability vulnerability.
case studies and identify which variables are most influ- Informed by Ford and Smit [26], we contend that
ential for developing a regional vulnerability index. relationships of E, S, and AC can overlap and therefore
Vulnerability has been conceptualized in several are best represented by a multiplex network approach,
different ways among different discipline and study where objects in the network can have multiple types
perspectives [13–17]. Two distinct frameworks of vul- of relationships. Development of multiplex tools is
nerability appear throughout the scholarship—risk relatively new [32], and their application to human-
hazard or outcome-based approaches, and pressure environment research is even newer [33]. Therefore,
and release or contextual vulnerability [15, 18–21]. this study not only advances vulnerability scholarship,
Although both approaches fundamentally denote the but also network approaches to human-environment
susceptibility of a population to harm, the associated research. In this paper, we apply network analysis to
research questions and perspectives differ [16, 18, assess vulnerability in the Canadian Arctic, a region
20, 22]. Outcome-based vulnerability assessments undergoing some of the most rapid changes in climate
focus primarily on biophysical exposure and assess globally [34], with a focus on Inuit communities (see
vulnerabilities directly related to these exposures, map in SP-1, available online at stacks.iop.org/ERL/
while contextual vulnerability begins with human sys- 13/104019/mmedia).
tems and examines what makes populations vulner-
able to external conditions in the context of multiple
stresses [18, 23]. This framing of social vulnerability is Methods
also prevalent within the natural hazards scholarship
by which disasters are seen as products of social struc- Framework and research approach
tures, not solely biophysical magnitude [24]. This can Moving from qualitative community-based vulner-
be expressed as: ability studies to a regional system map of vulnerability
(a model that can be eventually linked to social and
V = f (E , S , AC ) ,
environmental projections), requires an iterative and
where E, S, and AC refer to exposure, sensitivity, and transferable approach. We begin by visualizing the
adaptive capacity of a community or region [25]. vulnerability framework in a way that promotes
Exposure is defined as a change in frequency, magni- quantitative operationalization of a systems analysis.
tude, or duration of biophysical interactions with Informed by existing vulnerability frameworks, we
human systems or other biophysical variables [26–28]. first subset the system into exposure (E), sensitivity (S),
Sensitivity is defined as conditions that alter the and adaptive capacity (AC). Since a given variable can
amount of impact an exposure may have on a influence multiple components of vulnerability (i.e.
biophysical or human system [26, 29, 30], and adaptive socio-economic status may influence both adaptive
capacity reflects the ability to address, plan for, or
capacity and sensitivity), we ensure that each vulner-
adapt to climate-related risks and take advantage of
ability component includes all variables; each comp-
new opportunities [26, 31]. These three components
onent can be represented as a two-dimensional matrix
of vulnerability are not mutually exclusive and the
relationships between variables can be scale dependent of system variables that together form a multi-layered
and may vary depending on the system being assessed system of interactions (figure 1).
[26]. We use contextual vulnerability in this study Next, we used a systematic literature review
because of the emphasis placed on social systems as [35, 36] of climate change vulnerability case studies
mechanisms for reduction (adaptation) or exacerba- conducted in the Canadian Arctic to select index vari-
tion of vulnerability and the ability to capture feedback ables and understand their relationships. Figure 2
loops and cross-sectoral interactions. illustrates the steps used starting with the systematic
While study approaches for contextual vulnerability literature review and ending with stakeholder engage-
vary throughout the literature, most community-based ment to validate and refine our results. Because the
case studies begin by examining the state of vulnerability current research builds on more than 16 years of work
(health burdens, food insecurity, unrest) and then dis- by our research group with local Inuit communities
sect the social and physical pathways and influences. and regional decision makers [5, 26, 37, 38], under-
When focusing on specific livelihood segments and pinned by principles of community-based participa-
locations, the classification and system mapping is often tory research [39, 40], we have the advantage of
straightforward. However, as we develop a model for implicitly integrating substantial local insight into our
operationalizing vulnerability assessments for regions project. We therefore, aim to re-engage stakeholders at
and transitioning from qualitative to quantitative repre- the end, to validate and refine our results. Community
sentations, selection of appropriate variables, identifica- work is a major investment for both researchers and
tion of system characteristics, and accounting for scale community members and some Inuit communities

2
Environ. Res. Lett. 13 (2018) 104019

Figure 1. Conceptualization of vulnerability as a multiplex network. On the left, two-dimensional matrices represent the coding
process (positive, negative or unknown/variable; blue, yellow and white respectively). Variables are also coded for membership in
four sub-networks from Canada’s Marine Coasts Assessment (top of figure) so that the total network can be subdivided to focus on
individual sectors/segments. Membership to sub-indices is not mutually exclusive. On the right, the matrices are translated into
networks. The gray circles represent vulnerability variables (called nodes) and the presence of E, S, and AC relationships are illustrated
by lines (called edges). Multiplex metrics are then calculated to understand the extent to which individual variables contribute to E, S,
and AC for an integrated understanding of vulnerability.

have already voiced concerns about research fatigue using repository searches (e.g. openDOAR), reference
[40]. To reduce fatigue, we propose coming to com- lists, and consultation with team members. For the
munities with a set of analyses and research findings in gray literature, the following syntaxes were used: (a)
hand that can catalyze conversations, which has pro- ‘Inuit climate change’ n=588 records; (b) ‘Inuit
ven fruitful in other places [41]. When generalizing vulnerability’ n=437; (c) ‘Inuit adaptation’ n=274
our approach to situations that lack such longstanding and finally ‘Inuit resilience’ n=10 records. See SP-1
relationships, initial community collaboration is likely for further details.
an important first step, as indicated in figure 2. The The 1758 articles, book, book chapters and con-
results of this paper focus on the literature review and ference proceedings were reviewed and organized
multiplex network analysis of vulnerability (i.e. steps 1 using a peer review tool for scrutinizing called Rayyan
through 5, figure 2). Further reflections about stake- Systemic Reviews Platform [42] (see systematic litera-
holder engagement are provided in the discussion. ture review SP-1). This online platform helps expedite
the initial screening of abstracts and titles using a pro-
Systematic literature review cess of semi-automation while incorporating a high
We began the review in Web of Knowledge using the level of usability [42]. Using this online tool, our team
following syntaxes during the 1970–2017 period: checked articles and agreed if they should be included
(a) ‘climat* chang*’ AND ‘Inuit*’ n=240 records; for a complete review. Articles and gray literature were
(b) ‘climat* chang*’ AND ‘vuln* Inuit*’ n=83 excluded from the full review process if they did not
records; (c) ‘climat* chang*’ AND ‘adapt* Inuit*’ relate to Inuit systems vulnerability or connect Inuit
n=87 records and finally (d) ‘res*’ AND ‘Inuit*’ lifestyle to human physical or environmental shifts.
n=39 records. We also searched gray literature not Papers that did not provide solid evidence linking cli-
computed by index tools, including institutional mate change, socio-economic, and environment fac-
reports, consultant reports, book chapters and con- tors were also excluded. The final review included 155
ference proceedings. These searches were conducted documents.

3
Environ. Res. Lett. 13 (2018) 104019

Figure 2. The figure illustrates the steps taken to implement the system’s network approach for vulnerability assessment in the
Canadian Arctic. Steps 1–5 are detailed throughout the article and step 6 is suggested as an exploratory and validation phase, which can
be an iterative process or an alternative starting point.

Reviewing these materials initially identified 137 affects ‘B’ which then affects ‘C’ would be coded as two
vulnerability variables. Many variables were con- direct effects, ‘A -> B’ and ‘B -> C’, and not ‘A -> C’.
sidered sub-sets of larger categories. We therefore The network is directed and feedback between two
merged variables so that the resulting dataset (n=58, variables ‘A’ and ‘B’ is permitted and coded as two
table 1) would contain variables of a similar specificity relationships (‘A -> B’ and ‘B’ -> ‘A’). Two research-
(see SP-1 for initial variables). For example, we con- ers independently conducted all coding and then
sidered walrus and ringed seal (animals important to codes were checked for agreement; discrepancies were
subsistence food systems) to be part of variable called reviewed and discussed by the research team and reco-
‘ice-flow animals’. ded following a consensus decision.
Among the 58 merged variables identified in the Lastly, we categorized variables into four sub-net-
literature review, we coded the effects of theoretically works as used in Canada’s Marine Coasts Assessment
changing the magnitude or quality of any given vari- [5] to conduct in depth analysis of specific dimensions
able on other variables. Coding used a nominal and of Inuit life. The four sub-networks are detailed in SP-
qualitative classification and was based on our detailed 3 and include the following sectors: (S1) infrastructure
reading of the literature and subsequent under- and transportation, (S2) business and economy, (S3)
standing of biophysical and socio-political dynamics health and well-being, and (S4) culture-education and
in the region. We coded effects as positive (an subsistence-harvesting.
increase/decrease in ‘A’ has the same effect on ‘B’),
negative (an increase/decrease in ‘A’ has the opposite Network analysis
effect on ‘B’), or variable (the relationships could be We use network analytics to analyze variables that
highly contextual, exhibit an inflection point, or influence exposure, sensitivity, and adaptive capacity.
demonstrate some other behavior). Positive, negative, A network approach is logical because it analyzes the
and variable effects were considered for each vulner- relationships among variables and vulnerability is
ability component (i.e. E, S, and AC) and each could fundamentally a relational concept; vulnerability is
have different interaction types (i.e. increasing ‘A’ produced by the interaction of various social and
could increase ‘B’ when considering sensitivity, but ecological factors [26]. A network consists of an
decrease ‘B’ when considering exposure). We only assemblage of units, called nodes in network science,
considered direct effects where, for example, changing that are connected by edges. Nodes in our network
‘A’ directly effects ‘B.’ Situations where changing ‘A’ represent the 58 variables identified in the literature

4
Environ. Res. Lett. 13 (2018) 104019

Table 1. Variables (n=58) included in the Inuit vulnerability system grouped by major category. Acronyms are provided and used in
figures 3 and 4.

Categories Variables Acronym Categories Variables Acronym

Weather Extreme weather event XWtr Biophysical environment Permafrost Prm

(Ecosystems)
Fog Fog Eustatic sea level rise SLev
Precipitation Prcp Native ice-flow animals WlfI
Temperature Tmp Native marine animals WlFM
Wind direction variability WDir Native plants Plts
Wind speed WSpd Native terrestrial animals WlfT
Hazards Coastal erosion Ersn Sea ice SIce
Floods Fld Health Mental health HthM
Slope failure Slp Physical health HthP
Storm surge SSrg Commercial Air transportation ArTr
transportation
Infrastructure Public buildings and roads Publ Overland transportation LnTr
Airport infrastructure Arp Shipping ShTr
Energy availability Eng Public Services Health care HthC
Ports Prt Education quality and quantity Edu
Telecommunications Telc Emergency response EmRs
Waste management Wst Natural resource management NRM
Housing Housing quality HsQt Risk management and future RMP
planning
Housing quantity HsQl Culture and Language Arts and traditional equip- Trd
ment/clothing production
Economy Cost of living Cost Environmental knowledge and EvKS
skills
Fishing (commercial) FshC Food sharing and social FdShr
networks
Informal income Inflc Indigenous language strength Lang
Nat. resource extraction NREx Water Security Water access WtrAc
Relative poverty Pov Water quality WtrQl
Tourism Trsm Water quantity WtrQt
Wage income WgIc Personal Land Travel Access to equipment Eqp
Food Security Country food quality CnQt Travel on the ice TrvI
Country food quantity CnQl Travel on the land TrvL
Food access FdAc Travel on water TrvW
Store-bought food quality StrQl
Store-bought food StrQt
quantity

review. Positive, negative, and variable effects were network. If the agency is inefficient, all communications
represented as edges, and each vulnerability comp- between parties will be slowed. If the agency dissolves,
onent (E, S, and AC) constitutes a network layer, or the network falls apart. The quality of actors (nodes)
one of multiple possible relationships in a multiplex and relationships (edges) are also important in a net-
network. We then analyzed a variable’s role based on work. When a central agency is trusted and all organiza-
network structures. For example, a highly connected tions agree on common objectives, a centralized
variable would be more central to understanding network can be very efficient [47, 48]. A structural per-
vulnerability than a variable with few connections. spective does not ignore node and edge function and
Such diagnostic approaches are commonly used for quality. It does however, recognize that real and poten-
guiding resource management programs (e.g. as tial function can be inferred by structural patterns that
reviewed by [43]) as well as understanding policy can then be further contextualized within specific cases
interactions [44]. Diagnostic application to economic as needed.
vulnerability also exist (e.g. vulnerably to peak A fundamental structural property of any node in a
oil [45]). network is connections to other nodes [46]. The num-
A network perspective often focuses on structural ber of connections is called a node’s degree centrality
relationships, or the number and arrangement of nodes (CD). The higher a node’s degree centrality, the more
and edges [46]. For example, a single state agency that important it is in the network from a structural per-
holds a governance network together, forming a net- spective. In a multiplex case, however, two nodes may
work with a central hub and few connections among have the same total number of edges but play very dif-
other nodes, will have significant influence over the ferent roles or function in the network. For example,

5
Environ. Res. Lett. 13 (2018) 104019

Table 2. Descriptive statistics of multiplex scores accords all nodes (n=58) in the entire network.

Participation coefficient (Pi) Total overlapping degree (oi) Normalized total overlapping degree (z)

Min 0.19 14.00 −1.24

Max 0.98 85.00 3.49
Mean 0.77 32.62 0.00
Median 0.75 30.50 −0.14
Standard deviation 0.13 14.90 0.99

consider a node ‘A’ with nine exposure edges and a considered for these metrics. Our objective was to pro-
node ‘B’ with nine edges spread evenly across the E, S, vide an initial coarse-grained diagnostic of vulner-
and AC categories. Despite having the same number of ability. We then analyzed the weighted data to
edges, the two nodes clearly play different roles in the understand system feedbacks using network diagrams
network, the exact nature of which is contextual to the and edge frequency counts. Further details and exam-
particular case. Node ‘B’ in the previous example is ple code for calculating multiplex Pi, Z-scores, and
said to be more multiplex [49]. Analyzing the degree total overlapping degree are provided in SP-2.
centrality and multiplex nature of vulnerability com-
ponents helps untangle the complex web of interac-
tions that create the coupled human-environment
Results
phenomena of vulnerability [16].
Multiplex participation analysis
To understand how multiplex a node is, we calcu-
Nodes in the network vary greatly in terms of how
late its participation coefficient (Pi) which is a measure
multiplex they are (table 2, min and max Pi). No node,
of how evenly a node’s edges are distributed among
however, is perfectly multiplex or participating only
different categories. Following Battiston et al [49] who
within a single layer (0<Pi <1.00, table 1). The most
expanded the concept of single layer network partici-
multiplex variables in the whole vulnerability system
pation [50] to multiplex networks, we consider Pi to be
(figure 3) includes travel on ice (Pi =0.98), ports (Pi
zero when a node has edges of only one kind (e.g. only
=0.98), water quality Pi=0.97), public buildings and
exposure edges). Pi equals one when a node has an
roads (Pi =0.96), shipping (Pi =0.95) and travel on
equal number of edges among all the categories in
land (Pi =0.95), implying that these are cross cutting
question (e.g. 3 E, 3 S, and 3 AC). Formally, Pi is
variables that play an integral role in the whole
defined as follows [49]:
vulnerability system.
M ⎡ a=1 ⎛
k i [a] ⎞ ⎤
2 A different set of nodes has the highest total over-
Pi = ⎢1 -
M - 1 ⎢⎣
å ⎝ o ⎠ ⎥⎥,
⎜ ⎟ lapping degrees: cost of living (oi =85), poverty (oi
M i ⎦
=77), wage income (oi =71), natural resource man-
where M=the number of layers in the network (i.e. agement-planning (oi =64) and extreme weather
E, S, and AC), ki =a node’s degree centrality for each events (oi =62). These variables concomitantly have
of a layers, and oi =a node’s total degree centrality medium-high participation scores (Pi) in the range of
across all network layers, also known as its total 0.64–0.78 indicating that they too are important cross
overlapping degree. cutting issues affecting Inuit vulnerability to climate
Following Battiston et al [49], we also consider change, though not the most multiplex. Within the
Z-scores, which normalize the total overlapping network, variable-undetermined relationships are the
degree to allow comparisons of networks of different most frequent relationships (n positive edges=500, n
sizes. negative edges=460, n variable edges=932). Inuit
cost of living, for example has a total overlapping
oi - á oñ
z (oi ) = , degree (oi) for variable-undetermined relationships of
so 76 and only nine positive and negative relationships
where 〈o〉 =the average overlapping degree of all combined. Cost of living is relatively multiplex (Pi
nodes in the system and σo =the standard deviation. =0.75) and structurally important (oi =85) for
Unlike Pi, which is bound by 0 and 1, the Z-score is not understanding Inuit vulnerability to climate change,
bounded and its magnitude and range illustrate the but its specific interactions with other variables are
variability of total degree overlap in the system. uncertain illustrating a challenge when trying to para-
Our vulnerability network is directed, meaning meterize a total vulnerability index. Any index of Inuit
that edges did not have to be reciprocal; there could be vulnerability should include a sensitivity analysis to a
an edge from i to j in layer a, but not from j to i. Degree variable with multiplex scores such as those found in
centrality scores include edges to and from a node (i.e. cost of living. The index might be calibrated under sev-
total or freeman’s degree centrality). When calculating eral scenarios or assumptions about a variable such as
centrality, Z-scores and participation we unweighted cost of living to present a range of results to aid deci-
the data and only edge presence or absence was sion makers.

6
Environ. Res. Lett. 13 (2018) 104019

Figure 3. Multiplex vulnerability network plot considering the whole vulnerability system. The y-axis Z represents the normalized
total overlapping degree centrality. The x-axis Pi measures the distribution of edges among the different vulnerability dimensions E, S
and AC. For clarity, abbreviated variables names are listed. Full names are available in the list in the right side of the figure.

The multiplex analysis condenses several dimen- Inuit well-being. Sea ice extent could serve as a key
sions of information into a single metric providing a indicator variable within a vulnerability index.
powerful overview of the entire system. This analytical
power, however, should not overshadow the merit of Inuit sensitivity to climate change
simultaneously learning through the lens of a more Within the sensitivity layer, there are numerous
detailed analysis of each layer. For this reason, we fur- variable-undetermined relationships, especially con-
ther explore and dissect the whole vulnerability system cerning the socioeconomic, cultural and well-being
into the three vulnerability layers of exposure, sensitiv- aspects of Inuit. It is logical that these variables have
ity and adaptive capacity (figure 4). many variable-undetermined relationships; they are
driven by people’s actions, governance and the idio-
Inuit exposure, sensitivity and adaptive capacity to syncrasies of specific regions and not necessarily
climate change reducible to simple direct relationships. Relative
Inuit exposure to climate change poverty has the highest degree centrality (CD =48,
In the exposure layer, extreme weather has the largest figure 4, table 3). Other nodes with high degree
degree centrality (CD =27, figure 4, table 3) and would centrality include cost of living (CD =43), extreme
therefore affect many variables within the Inuit weather events (CD =35), and natural resource
vulnerability system. Temperature is the second most management (CD =24). The large number of vari-
central variable (CD =24) and is expected to increase able-undetermined edges indicate that it is not always
with climate change, leaving the region more prone to possible to associate a given driver of change to
disasters such as floods (CD =24) or storm surges (CD perceived or documented consequences; there will be
=20) [51, 52]. Combining quantitative metrics with complex cumulative impacts of different drivers and
qualitative systems understanding can help parame- trade-offs. This uncertainty highlights the need to
terize a vulnerability model. For example, sea ice, calibrate vulnerability models at a local scale. Course
another highly central variable (CD =17), has many grained assessments, as presented here, can help tease
outgoing positive relationships (blue arrows, figure 4) out assumptions and identify areas needing further
to other variables that influence traditional food research.
systems including food security, access, and animal
resource abundance. Traditional activities have been Inuit adaptive capacity to climate change
shown to improve Inuit physical and mental health The adaptive capacity layer has several variable-undeter-
[53, 54] and therefore a variable like sea ice is not only mined and positive relationships. Wage income, with the
structurally important in the network, but essential for highest degree centrality (CD =49, figure 4, table 3), has

7
Environ. Res. Lett. 13 (2018) 104019

Figure 4. The network figure illustrates (a) exposure, (b) sensitivity and (c) adaptive capacity layers for the whole vulnerability system.
Node size is proportional to the total number of network connections (Freeman’s degree centrality, CD). Node colors represent the 13
category groups (see legend at bottom). Blue lines are positive relationships (an increase / decrease in ‘a’ causes the same change in ‘b’).
Red lines are negative relationships (an increase / decrease in ‘a’ causes the opposite change in ‘b’). Black dashed lines are variable/
undetermined relationships (outcome is contextual and/or a non-linear relationship). For clarity, only variable names discussed in
the text are shown. See SP-3 for graphics with all variables labeled.

high structural importance followed by cost of living (CD which are highly central, are affected by a wide range of
=42), risk management and future planning (CD =40) local, regional, and national programs and capacities [55].
and natural resource management (CD =34). Natural The vulnerability index will need to be calibrated for local
resource management and risk management, both of variations in these important variables.

8
Environ. Res. Lett. 13 (2018) 104019

Table 3. Names and Freeman’s degree centrality (CD) scores of the top 10 rank ordered variables within each vulnerability layer (E, S, AC).
The same nodes (n=58) are present in each layer, but the number of relationship varies. In the table, management is abbreviated mgnt.

Rank
order Exposure Sensitivity Adaptive capacity

1 Extreme weather events (CD =27) Relative poverty (CD =48) Wage income (CD =49)
2 Temperature (CD =24), Cost of living (CD =43) Cost of living (CD =42)
Floods (CD =24)
3 Storm surge (CD =20) Extreme weather events (CD =35) Risk mngt. and future planning (CD =40)
4 Wind speed (CD =19) Natural resource mngt. (CD =24) Relative poverty (CD =29)
5 Sea ice (CD =17), County food quantity (CD =23) Physical health (CD =25)
Coastal erosion (CD =17)
6 Slope failure (CD =16) Precipitation (CD =22), Tourism (CD =23),
Wage income (CD =22) Mental health (CD =23),
Informal income (CD =23)
7 Precipitation (CD =13), Temperature (CD =21) Food sharing and social networks (CD =21)
Eustatic sea level rise (CD =13)
8 Permafrost (CD =9), Coastal erosion (CD =20), Arts and traditional equipment/clothing pro-
duction (CD =20),
Travel on ice (CD =9) Fishing (commercial) (CD =20) Environmental knowledge and skills (CD
=20)
9 Fog (CD =8), Floods (CD =19), Natural resource extraction (CD =19)
Overland transportation (CD =8), Country food quality (CD =19)
Public buildings and roads (CD
=8),
Ports (CD =8)
10 Travel on land (CD =7) Tourism (CD =18), Emergency response (CD =17),
Storm surge (CD =18), Indigenous language strength (CD =17)
Risk mngt. and future palming (CD
=18)

Discussion Several nodes were less multiplex than expected, such

as access to education, knowledge of traditional lan-
Quantitative vulnerability indices are an important guages and health care. Low multiplex scores by no
tool for adaptation planning [56]. Indicator variables means challenge the established importance of these
should be easily applicable, measurable, accessible, variables in the literature. Simply, these variables,
transferable, and non-redundant [57]. While uncer- which primary interact within a single vulnerability
tainty in climate models and social and economic layer, may be more straightforward to parameterize in
projections makes predicting future vulnerability an index.
challenging, vulnerability indices, nevertheless, pro- Our results provide important assessment of vari-
vide a foundation for integrating future climatic and ables contributing to Inuit vulnerability. A few limita-
socio-economic drivers of vulnerability. Our systema- tions and potential future expansions are worth noting
tic literature review and multiplex analysis advances however. First, several highly multiplex and central
vulnerability indicator creation by integrating social variables are categorically related such as poverty,
and biophysical variables derived from both commu- costs of living, and wage income. These variables,
nity-based studies and large-scale regional reports. We however, cannot be treated as surrogates. Raising
also provide a methodology for identifying key indi- wages will have little effect on a person’s life if the cost
cator variables based on their centrality within a of living rises. Both variables must be considered,
systems framework of vulnerability and ability to though perhaps could be calibrated into a living wage
crosscut the three vulnerability components of expo- index. Our perspective is that the network analysis
sure, sensitivity, and adaptive capacity. should always be paired with expert opinion when
In our analysis, cross cutting, multiplex variables, crafting vulnerability indices to account for such
such as cost of living, poverty and wage income will nuanced relationships and interdependences among
play a vital role in future vulnerability index creation. nodes.
Context, however, is important to consider when Second, our analysis uses several very powerful
parameterizing. For example, tourism was highly cen- network metrics including multiplex participation,
tral in S and AC dimensions. While potently a source Z-scores, and degree centrality. We unweighted the
of income (and possible adaptive capacity), studies network when calculating most metrics but included
have shown that tourism can be maladaptive when not weights when considering edge counts and visual
properly integrated into local economies [58], so the interpretation. Future expansions of our work might
effects of this variable must be evaluated carefully. take advantage of other network metrics (see [59] for a

9
Environ. Res. Lett. 13 (2018) 104019

review). For example, average node strength accounts network analysis aim to support indicator develop-
for the value or strength of all edges and inverse parti- ment for the entire Canadian Arctic, an area com-
cipation ratio (not to be confused with multiplex prised of 53 communities spread out over a vast area
participation, Pi, which we analyzed) evaluates how >2 million km2 and four regional governments, each
evenly weights are distributed among the edges [60]. with rules and norms for conducting research (SP-1).
These metrics could help highlight variables known to Time, logistics, and finances may inhibit consulting
be important to Inuit well-being and resilience, such every community. Researchers will have to work hard
as education and traditional skills, that did not score to achieve a legitimate engagement plan that yields the
high in our multiplex participation analysis. right level of stakeholder input as there are likely no
Third, we do not take geography into account. Our panacea approaches.
review pulls together numerous case studies and regio-
nal reports to synthesize a set of vulnerability variables.
Conclusion
It is possible that some variables are more pressing in
specific regions of the Canadian Arctic due to local
Our multiplex network approach provides a flexible,
variability in resources, environment, and socio-poli-
quantitative framework to evaluate exposure, sensitiv-
tical organization. Expansion of our work might
ity, and adaptive capacity variables while recognizing
attempt to integrate a geographic weighting or regio-
that these dimensions of vulnerability are not mutually
nal membership into the network analysis [43].
exclusive. Our results can help select indicators from
Finally, since our analysis started with the litera-
available census and other government data to create
ture review, our results only reflect variables captured
vulnerability indices. For instance, accessing fine-
in the literature. Fortunately, there is a strong tradition
resolution, up-to-date data for highly central and
of vulnerability and adaptation research in the Cana-
multiplex variables would be a top priority when
dian Arctic, providing a robust dataset to pull on.
compiling vulnerability indices. We can further cali-
While no minimum number of local case studies are
brate indicator models based on their positive and
needed for the network analysis, case study availability
negative relationships. Undetermined effects will have
could be a barrier in other locations; but our approach
to be scrutinized depending on how they behave in
can be applied to any population or setting provided
context specific situations. Our approach, however, is
there is data. Data availability and reliability are not
not without limitations. We recommend using the
unique concerns to the network approach. They are a
network approach as one of several methods to
challenge for any synthesis method. Network analysis
identify indicator variables and that any index be
is actually rather robust to small datasets. The multi-
developed with adequate and necessary involvement
plex analysis could be conducted with as few as three
and feedback from local communities, decision
nodes, the bare minimum that constitutes a network.
makers, and subject experts.
(Recall, nodes in our analysis are vulnerability vari-
ables, not the case studies from which variables are dis-
tilled.) Where quality case studies are lacking, other Acknowledgments
data sources such as media reports, or stakeholder
interviews could also be used to generate nodes for the Several anonymous reviewers provided important
network analysis. comments that improved this manuscript. This pro-
Even with a rich case study literature, stakeholder ject was supported by funding from the Social Sciences
engagement, as outlined in figure 2, is a fundamental and Humanities Research Council of Canada
step to validate findings and elicit additional impor- (SSHRC), Canadian Institute of Health Research
tant variables. While we have chosen to engage stake- (CIHR), Transport Canada’s Northern Transporta-
holders with a set of results in hand to catalyze tion Adaptation Initiative, and ArcticNet.
discussions, reflecting extensive previous research and
consultations conducted by the team in the region, ORCID iDs
other situations may warrant starting with stake-
holders. The network analysis might also be treated as Nathan S Debortoli https://orcid.org/0000-0002-
one of several methods, such as focus groups or expert 2467-243X
opinions, to identify variables.
Stakeholder involvement should not be taken
lightly or done superficially. Beyond essential issues References
related to community-based research, such as cultural [1] IPCC Climate Change 2014 Impacts, Adaptation, and
understanding, trust, integrating disparate world- Vulnerability. Part B: Regional Aspects. Contribution of Working
views, and considering what is realistic, feasible, and Group II to the Fifth Assessment Report of the Intergovernmental
desired involvement by communities [40, 61], there Panel on Climate Change ed V R Barros vol 2014 (Cambridge:
Cambridge University Press)
are inherent challenges of depth and breadth when [2] Jurgilevich A, Räsänen A, Groundstroem F and Juhola S 2017
engaging stakeholders for a locally informed and A systematic review of dynamics in climate risk and
regionally focused project. Our literature review and vulnerability assessments Environ. Res. Lett. 12 013002

10
Environ. Res. Lett. 13 (2018) 104019

[3] Räsänen A et al 2016 Climate change, multiple stressors and [23] Ford J D, Smit B and Wandel J 2006 Vulnerability to climate
human vulnerability: a systematic review Reg. Environ. Change change in the Arctic: a case study from Arctic Bay, Canada
16 2291–302 Glob. Environ. Change 16 145–60
[4] McDowell G, Ford J and Jones J 2016 Community-level [24] Hewitt K 1983 Interpretations of Calamity from the Viewpoint of
climate change vulnerability research: trends, progress, and Human Ecology (London: Taylor and Francis)
future directions Environ. Res. Lett. 11 033001 [25] Smit J B and Pilifosova O 2003 From adaptation to adaptive
[5] Ford J D, Bell T and Couture N J 2016 Perspectives of capacity and vulnerability reduction Climate Change, Adaptive
Canada’s North Coast region Climate Change Impacts and Capacity, and Development. ed J B Smith et al (London:
Adaptation Assessment of Canada’s Marine Coasts. 1 ed Imperial College Press) pp 9–28
D S Lemmen et al (Ottawa: Government of Canada) [26] Ford J D and Smit B 2004 A framework for assessing the
pp 153–206 vulnerability of communities in the Canadian Arctic to risks
[6] Ford J D, Champalle C, Tudge P, Riedlsperger R, Bell T and associated with climate change Arctic 57 389–400
Sparling E 2015 Evaluating climate change vulnerability [27] Turner B L et al 2003 A framework for vulnerability analysis in
assessments: a case study of research focusing on the built sustainability science Proc. Natl Acad. Sci. 100 8074–9
environment in northern Canada Mitigation Adaptation [28] Ford J D, Knight M and Pearce T 2013 Assessing the ‘usability’
Strateg. Glob. Change 20 1267 of climate change research for decision-making: a case study of
[7] Bisaro A and Hinkel J 2016 Governance of social dilemmas in the Canadian International Polar Year Glob. Environ. Change-
climate change adaptation Nat. Clim. Change 6 354–9 Hum. Policy Dimens. 23 1317–26
[8] Douglas A et al 2018 Measuring Progress on Adaptation and [29] Hewitt K 1983 Interpretation of Calamity: From the Viewpoint
Climate Resilience: Recommendations to the Governement of of Human Ecology (Boston, MA: Allen and Unwinn)
Canada (Gatineau QC: Environment and Climate Change [30] Watts M J and Bohle H G 1993 The space of vulnerability: the
Canada) causal structure of hunger and famine Prog. Hum. Geog. 17 43–67
[9] Debortoli N S, Camarinha P I M, Marengo J A and [31] Ford J D et al 2013 The dynamic multiscale nature of climate
Rodrigues R R 2017 An index of Brazil’s vulnerability to change vulnerability: an Inuit harvesting example Ann. Assoc.
expected increases in natural flash flooding and landslide Am. Geogr. 103 1193–211
disasters in the context of climate change Nat. Hazards 86 [32] De Domenico M, Sole-Ribalta A, Gomez S and Arenas A 2014
557–82 Navigability of interconnected networks under random
[10] 2017 Índice de vulnerabilidade aos desastres naturais failures Proc. Natl Acad. Sci. USA 111 8351–6
relacionados às secas no contexto da mudança do clima / [33] Baggio J A, BurnSilver S B, Arenasc A, Magdanzd J S,
Ministério do Meio Ambiente, Ministério da Integração Kofinas G P and Domenico M D 2016 Multiplex social
Naciona (Brasil: World Wildlife Fund) (https://www.wwf.org. ecological network analysis reveals how social changes affect
br/natureza_brasileira/reducao_de_impactos2/clima/copy_ community robustness more than resource depletion Proc.
of_mudancas_climaticas2_20062017_1938/) Natl Acad. Sci. 113 13708–13
[11] Debortoli N S, Camarinha P I M and Rodrigues R R 2016 [34] Ford J, Bell T, Couture N and Clark. D 2018 Climate change
Natural disasters caused by water Third National and Canada’s northern coast: current knowledge, future
Communication of Brazil to the United Nations Framework research needs Environ. Rev. 26 82–92
Convention on Climate Change Volume II 1st edn (Brasília: [35] Mahood Q, Van Eerd D and Irvin E 2014 Searching for grey
Brazilian Ministry of Science, Technology and Innovation) literature for systematic reviews: challenges and benefits Res.
pp 145–58 Synth. Methods 5 221–34
[12] Nguyen T T X, Bonetti J, Rogers K and Woodroffe C D 2016 [36] Ford J D, Berrang-Ford L and Paterson J 2011 A systematic
Indicator-based assessment of climate-change impacts on review of observed climate change adaptation in developed
coasts: a review of concepts, methodological approaches and nations Clim. Change 106 327–36
vulnerability indices Ocean Coast. Manage. 123 18–43 [37] Ford J D, McDowell G and Jones J 2014 The state of climate
[13] Ford J D and Smit B 2004 A framework for assessing the change adaptation in the Arctic Environ. Res. Lett. 9 104005
vulnerability of communities in the Canadian Arctic to risks [38] Ford J D, Pearce T, Duerden F, Furgal C and Smit B 2010
associated with climate change Arctic 57 389–400 Climate change policy responses for Canada’s Inuit
[14] Pearce T, Smit B, Duerden F, Ford J D, Goose A and Kataoyak F population: the importance of and opportunities for
2010 Inuit vulnerability and adaptive capacity to climate adaptation Glob. Environ. Change 20 177–91
change in Ulukhaktok, Northwest Territories, Canada Polar [39] Pearce T D et al 2009 Community collaboration and climate
Rec. 46 157–77 change research in the Canadian Arctic Polar Res. 28 10–27
[15] Smit B and Wandel J 2006 Adaptation, adaptive capacity and [40] Ford J D et al 2016 Community-based adaptation research in the
vulnerability Glob. Environ. Change 16 282–92 Canadian Arctic Wiley Interdiscip. Rev. Clim. Change 7 175–91
[16] Turner B L et al 2003 A framework for vulnerability analysis in [41] Vance-Borland K and Holley J 2011 Conservation stakeholder
sustainability science Proc. Natl Acad. Sci. 100 8074–9 network mapping, analysis, and weaving Conservation Lett. 4
[17] Williamson S, Sharp M, Dowdeswell J and Benham T 2008 278–88
Iceberg calving rates from northern Ellesmere Island ice caps, [42] Ouzzani M, Hammady H, Fedorowicz Z and Elmagarmid A
Canadian Arctic, 1999–2003 J. Glaciol. 54 391–400 2016 Rayyan-a web and mobile app for systematic reviews Syst.
[18] O’Brien K, Eriksen S, Nygaard L P and Schjolden A 2007 Why Rev. 5 210
different interpretations of vulnerability matter in climate [43] Sayles J S and Baggio J A 2017 Who collaborates and why:
change discourses Clim. Policy 7 73–88 assessment and diagnostic of governance network integration
[19] Adger W N et al 2007 Assessment of adaptation practices, for salmon restoration in Puget Sound, USA J. Environ.
options, constraints and capacity Climate Change 2007: Manage. 186 64–78
Impacts, Adaptation and Vulnerability. Contribution of Working [44] Champalle C, Ford J and Sherman M 2015 Prioritizing climate
Group II to the Fourth Assessment Report of the change adaptations in Canadian Arctic Communities
Intergovernmental Panel on Climate Change (Cambridge: Sustainability 7 9268–92
Cambridge University Press) 719–43 [45] Kerschner C, Prell C, Feng K and Hubacek K 2013 Economic
[20] Ribot J 2014 Cause and response: vulnerability and climate in vulnerability to peak oil Glob. Environ. Change 23 1424–33
the Anthropocene J. Peasant Stud. 41 667–705 [46] Borgatti S P, Mehra A, Brass D J and Labianca G 2009 Network
[21] Ford J D et al 2010 Case study and analogue methodologies in analysis in the social sciences Science 323 892–5
climate change vulnerability research Wiley Interdiscip. Rev.- [47] Bodin Ö and Crona B I 2009 The role of social networks in
Clim. Change 1 374–92 natural resource governance: what relational patterns make a
[22] Adger W N 2006 Vulnerability Glob. Environ. Change 16 268–81 difference? Glob. Environ. Change 19 366–74

11
Environ. Res. Lett. 13 (2018) 104019

[48] Berardo R and Scholz J T 2010 Self-organizing policy networks: (https://www.amap.no/documents/doc/Adaptation-

risk, partner selection, and cooperation in estuaries Am. J. Actions-for-a-Changing-Arctic-AACA-Barents-Area-
Political Sci. 5 632–49 Overview-report/1529)
[49] Battiston F, Nicosia V and Latora V 2014 Structural measures [56] Hinkel J 2011 Indicators of vulnerability and adaptive capacity:
for multiplex networks Phys. Rev. E 89 032804 towards a clarification of the science–policy interface Glob.
[50] Guimerà R and Amaral L A N 2005 Functional cartography of Environ. Change 21 198–208
complex metabolic networks Nature 433 895–900 [57] Birkmann J (ed) 2006 Indicators and criteria for measuring
[51] Diaconescu E P, Mailhot A, Brown R and Chaumont D 2018 vulnerability: theoretical bases and requirements Measuring
Evaluation of CORDEX-Arctic daily precipitation and Vulnerability to Natural Hazards: towards Disaster
temperature-based climate indices over Canadian Arctic land Resilient Societies (New Delhi: United Nations University
areas Clim. Dyn. 50 2061 Press)
[52] Bintanja R and Andry O 2017 Towards a rain-dominated [58] Carson M and Peterson G 2016 Arctic Council Arctic Resilience
Arctic Nat. Clim. Change 7 263–7 Report Stockholm Environment Institute and Stockholm
[53] Cunsolo Willox A et al 2013 Climate change and mental health: Resilience Centre (https://www.sei.org/mediamanager/
an exploratory case study from Rigolet, Nunatsiavut, Canada documents/Publications/ArcticResilienceReport-2016.pdf)
Clim. Change 121 255–70 [59] Boccaletti S et al 2014 The structure and dynamics of
[54] MacDonald J P, Harper S L, Cunsolo Willox A, Edge V L and multilayer networks Phys. Rep. 544 1–122
Government R I C 2013 A necessary voice: climate change and [60] Menichetti G, Remondini D, Panzarasa P, Mondrago R L J and
lived experiences of youth in Rigolet, Nunatsiavut, Canada Bianconi G 2014 Weighted multiplex networks PLoS One 9
Glob. Environ. Change-Hum. Policy Dimens. 23 360–71 [61] Kouril D, Furgal C and Whillans T 2016 Trends and key
[55] AMAP 2017 Adaptation Actions for a Changing Arctic elements in community-based monitoring: a systematic review
(AACA) Barents Area Overview Report Arctic Monitoring of the literature with an emphasis on Arctic and Subarctic
and Assessment Programme (AMAP) (Oslo, Norway) regions Environ. Rev. 24 151–63

12