Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

Ecological Indicators 117 (2020) 106529

Contents lists available at ScienceDirect

Ecological Indicators
journal homepage: www.elsevier.com/locate/ecolind

A new composite climate change vulnerability index T



H.K. Edmonds, J.E. Lovell, C.A.K. Lovell
School of Economics, University of Queensland, Brisbane QLD 4072, Australia

A R T I C LE I N FO A B S T R A C T

Keywords: The earth’s climate is changing, with global warming attributable to anthropogenic greenhouse gas emissions
Climate change driven by economic and population growth. Human systems and ecosystems vary in their exposure, mitigation
Vulnerability and adaptive capacity, and vulnerability to various forms of climate change. Once mitigation and adaptation
Composite index efforts have been exhausted, vulnerability remains. Data compiled by the University of Notre Dame covering
Endogenous aggregation weights
over 100 nations in 2016 were used to construct a new composite climate change vulnerability index that
Data envelopment analysis
features endogenously generated weights to aggregate vulnerability indices across six vulnerable sectors. These
weights have the potential to inform policy aimed at allocating resources to reduce the cost of limiting vul-
nerability. The new composite vulnerability index, whose weights differ across sectors and across nations, is
compared with the Notre Dame vulnerability index, which uses weights equal across sectors and constant across
nations. Although the two indices agree on the identity of the most vulnerable nations, there is a statistically
significant difference between the two indices. In addition, a nonparametric statistical test failed to reject the
null hypothesis that one sectoral index could be deleted from the composite index without significant loss of
information. This also has potentially important policy implications.

1. Introduction IPCC (2001; 5,8,16) noted that nations exhibit different vulner-
ability to climate change depending on geographic location, and social,
Anthropogenic greenhouse gas emissions driven by economic and economic and environmental conditions. For example, small island
population growth are warming planet Earth and changing its climate. states and nations with low-lying coastal areas are particularly vul-
A succession of Intergovernmental Panel on Climate Change (IPCC) nerable to rising sea levels, a vulnerability not shared by high-altitude
Assessment Reports concluded that warming is “unequivocal”, and it is inland nations. Later IPCC (2014a; 6) wrote, with very high confidence,
“extremely likely” that human influence has been the dominant cause of that differences in vulnerability arise from both climatic and non-cli-
the observed warming since the mid-20th century. Climate change matic factors, the latter encompassing multidimensional inequalities
brings shrinking glaciers, rising sea levels, and increases in the fre- often produced by uneven development processes. Especially vulner-
quency of heat waves, warm temperature extremes, and heavy pre- able are socially, economically, culturally, politically, institutionally or
cipitation events, all of which adversely affect both natural and human otherwise marginalized groups. More recently IPCC (2018; 9) added,
systems. Continued growth in greenhouse gas emissions is predicted to with high confidence, communities dependent on agricultural or coastal
cause further global warming and climate change, increasing the like- livelihoods. In the academic literature, Yohe et al. (2006) have con-
lihood of severe and irreversible impacts for people and ecosystems. structed a vulnerability index using a methodology of Brenkert and
Human systems and ecosystems differ in their exposure, mitigation Malone (2005), and have emphasized the uneven global distribution of
and adaptive capacity, and vulnerability to various forms of climate vulnerability. Schleussner et al. (2018) relied on the Multidimensional
change. Some systems are exposed to droughts, others to cyclones, and Poverty Index (MPI) of Alkire et al. (2016) to identify key vulnerability
still others to rising sea levels. Once mitigation and adaptation efforts areas with their “1.5℃ Hotspots”. Both studies associated high vul-
have been exhausted, vulnerability remains. In its Third Assessment nerability regions with Sub-Saharan Africa, South Asia and the Small
Report the IPCC (2001; 6) defined vulnerability as “…the degree to Island Developing States (SIDS).
which a system is susceptible to, or unable to cope with, adverse effects Vulnerability is an important determinant of the allocation of the
of climate change, including climate variability and extremes.” In Least Developed Country Fund resources under the auspices of the
general, nations with the least resources are the most vulnerable. United Nations Framework Convention on Climate Change (UNFCCC)


Corresponding author.
E-mail address: k.lovell@uq.edu.au (C.A.K. Lovell).

https://doi.org/10.1016/j.ecolind.2020.106529
Received 21 November 2019; Received in revised form 5 April 2020; Accepted 12 May 2020
Available online 22 May 2020
1470-160X/ © 2020 Elsevier Ltd. All rights reserved.
H.K. Edmonds, et al. Ecological Indicators 117 (2020) 106529

(2019) and financed through the Global Environment Facility. The the allocation of international support to less developed nations.
United Nations Department of Economic and Social Affairs and The remainder of the paper is structured as follows. In Section 2
Committee for Development Policy (2018) currently contains 47 na- composite indices and their structure, with emphasis on the literature
tions in its Least Developed Country (LDC) category, which aggregates concerned with the construction of composite environmental indices,
an Economic Vulnerability Index (EcoVI), a Human Assets Index, and are summarized. A description of the specific Notre Dame data used in
Gross National Income per capita; Guillaumont (2009) provides details the analysis is also included. In Section 3 the methodology adopted to
on its construction. While the EcoVI differs from the IPCC concept of construct the composite CCVI is outlined. In Section 4 empirical find-
vulnerability to climate change, it shares crucial components (e.g., the ings are presented and discussed with suggestions for further research.
percent of population in coastal zones of low elevation). One finding highlights the superiority of an endogenously-weighted
Numerous factors constrain adaptation, and therefore vulnerability. composite CCVI over a conventional fixed weight index, and another
The key constraints identified by IPCC (2001; 8 and IPCC (2014a; demonstrates the feasibility of exploiting the endogenously-generated
26,27) include limited financial and human resources, limited in- weights to construct an equally informative CCVI having fewer sub-
tegration or coordination of governance, absence of key adaptation indices.
leaders and advocates, and poor planning. IPCC (2014b) augmented
this list of constraints to adaptation and provided a detailed survey of 2. Literature review
the literature.
The existence of adaptation constraints leads naturally to the con- The IPCC’s definition of vulnerability has evolved little through
cept of adaptation thresholds, and hard and soft limits to adaptation, successive Assessment Reports, and has guided the subsequent aca-
and then to adaptation deficits, or adaptation gaps, the difference be- demic literature. Adger (2006), Eakin and Luers (2006), and Gallopín
tween (or ratio of) actual and maximum feasible adaptation attributable (2006) have developed a holistic social-ecological system framework
to constraints and inefficiency, as discussed by United Nations for addressing the human and environmental sources of vulnerability
Environment Programme (UNEP) (UNEP, 2014, UNEP, 2016, UNEP, and establishing suitable metrics for it. Beroya-Eitner (2016) summar-
2017, UNEP, 2018). UNEP (2018; 2) defined adaptation gaps as “…the ized the issues with a focus on the relatively neglected ecological di-
difference between the actual level of adaptation and the level required mension, and Weißhuhn et al. (2018) provided a recent survey of the
to achieve a societal goal”. Maximum feasible adaptation can be re- literature. The University of Notre Dame (2015a, 2015b) vulnerability
presented by an adaptation frontier introduced by Preston et al. (2013). index fits into this holistic framework, containing both social and
Sectors generally identified with food security, water resources, ecological components within its sectoral indices.
human health, ecosystem services, human habitat, and infrastructure The transition from conceptualization to quantification has gener-
inclusive of energy resources experience differences in vulnerability. ated a search for ways to aggregate multiple vulnerability indices into a
Data from the University of Notre Dame containing vulnerability in- composite index. The OECD (nd) provided guidance, defining a com-
dices across these six sectors for over 100 nations were used to con- posite index as a single index comprised of multiple indicators based on
struct a composite climate change vulnerability index (CCVI) for each “…an underlying model of the multi-dimensional concept that is being
nation, and to contrast it with the Notre Dame vulnerability index. The measured.” Key features of the underlying model include the mathe-
CCVI was constructed in a data envelopment analysis (DEA), a linear matical structure of the composite index and its appropriateness, re-
programming technique introduced by Charnes et al. (1978) to analyze gardless of structure, as a basis for policy-making.
the performance of not-for-profit entities. DEA is a particularly appro- The encyclopedic OECD Handbook edited by Nardo et al. (2008)
priate methodology for analyzing environmental outcomes that are not stressed the importance of the underlying theoretical framework for
priced on markets. In the present analysis decision-making units are generating a “meaningful” composite index. It also notes that weighting
nations, variables are the six sectoral vulnerability indices, and the and aggregation of component indices should be consistent with the
relative performance of nations is measured with the CCVI. underlying theoretical framework, a desideratum satisfied by few com-
DEA also provides a flexible, endogenously-generated weighting posite indices, and fewer composite environmental indices.
system to aggregate environmental outcomes and to guide decision- The list of extant composite indices is long; Böhringer and Jochem
making and resource allocation aimed at enhancing the efficacy of (2007) surveyed 11 composite indices, including the Human Develop-
policies designed to reduce vulnerability to climate change. DEA-gen- ment Index (HDI) (United Nations Development Programme, nd), and
erated endogenous weights can play a critical role in addition to their found each to have structural shortcomings. The HDI is the most pro-
role as aggregators. Mendelsohn (2006) and Anderson et al. (2019) minent composite index, and has received the most criticism; see for
have stressed that market prices would provide signals eliciting public example, Despotis (2005), Cherchye et al. (2008), and Klugman et al.
policy responses that differ across nations. These price signals would (2011). Much of the criticism has revolved around the mathematical
encourage efficient adaptation, thereby lowering the overall cost of structure of the model underlying the composite index. The mathema-
limiting vulnerability to climate change. In the absence of market prices tical structure of a composite index can be characterized in terms of
of sectoral vulnerability indices, DEA weights can serve the same pur- normalization of variables, weighting of variables in aggregation, and
pose, by providing proxies for market prices that can elicit nation- the functional form of the aggregator function (Gan et al., 2017; Greco
specific policy responses that direct resource allocation to reduce the et al., 2019). It is widely recognized that most composite environmental
cost of limiting vulnerability to climate change. IPCC (2014a; 26) indices have restrictive mathematical structures. The EcoVI provides a
identified these signals with economic instruments such as payments prominent example; for a detailed account of its construction illus-
for environmental services, improved resource pricing, and charges and trating its restrictive structure see Feinduono and Goujon (2016). An-
subsidies. The ability to quantify unpriced effects addresses the call of other is provided by the Physical Vulnerability to Climate Change Index
Weaver and Miller (2019) for cross-disciplinary participation that (PVCCI) (Guillaumont, 2015; Closset et al., 2017; Feindouno and
would bring improvements in our ability to value impacts of climate Guillaumont, 2019; Guillaumont and Simonet, 2011a, 2011b). A third
change as an input to the design of policy mechanisms. is the Yale Environmental Performance Index (EPI) (Wendling et al.,
Generalizing from adaptation to vulnerability, the CCVI introduces 2018). A fourth is the Environmental Vulnerability Index (EnvVI) of the
vulnerability gaps relative to a vulnerability frontier, and is capable of South Pacific Applied Geoscience Commission and the United Nations
identifying the most vulnerable nations. Observed variation in vulner- Environmental Programme (South Pacific Applied Geoscience
ability gaps across sectors within nations makes it important to treat Commission and United Nations Environmental Programme, 2005).
nations differently when constructing a composite vulnerability index. Eakin and Luers (2006), Eriksen and Kelly (2007), Barnett et al.
This is particularly important given the link between vulnerability and (2008), Füssel (2009) and Hinkel (2011) have offered a variety of

2
H.K. Edmonds, et al. Ecological Indicators 117 (2020) 106529

somewhat skeptical assessments of the mathematical structure, and governance. This abbreviation of a conventional DEA program is the
hence the practical value, of these and other composite environmental contribution of Adolphson et al. (1991) and Lovell and Pastor (1999).
indices. Barnett et al. (2008; 102) even argued that indices such as Unlike most models of business or economic behavior that contain
EnvVI “should not be used as the basis for disbursing funds, comparing variables to be minimized, such as expenses or resource use, and vari-
countries, or for measuring the performance of countries in environ- ables to be maximized, such as revenues or educational outcomes, this
mental management”. climate change vulnerability model is restricted to variables to be
The Notre Dame vulnerability index is structured similarly to the minimized, namely sectoral vulnerability indices. The envelopment
composite environmental indices mentioned above, being calculated as program in Table 1 calculates the potential of nation “o” to shrink its
the arithmetic mean of its six normalized sectoral vulnerability indices. vector of sectoral vulnerability indices xo as much as possible, subject to
This has two implications: (i) sectoral weights are equal to 1/6, forcing N constraints, one for each sectoral index, that bound the shrunken
all nations to value the six sectors equally, regardless of their circum- vector xo below by a nonnegative convex combination of the least
stances; and (ii) aggregation is linear (rather than log-linear, for ex- vulnerable nations in the sample. The optimal value of θ ∈ (0,1], with
ample), implying perfect substitutability among all sectoral indices for smaller values of θ indicating greater vulnerability. This lower bound
all nations. This form of weighting and aggregation is common among defines a composite vulnerability frontier, analogous to the adaptation
composite indices, and it has generated much criticism in the literature, frontier introduced by Preston et al. (2013). The optimal value of θ
including Cherchye et al. (2007, Cherchye et al., 2008), Bowen and forms the basis for a measure of a nation’s composite vulnerability gap,
Moesen (2011), Zhou et al. (2006, 2007, 2017), Gan et al. (2017), Seth the difference between (or ratio of) its actual vulnerability xo and its
and McGillivray (Seth and McGillivary, 2018), Zhou and Zhang (2018), potential vulnerability θ xo. The series of adaptation gaps examined by
and Greco et al. (2019). UNEP can be viewed as conceptual components of a vulnerability gap.
Surprisingly few attempts have been made to introduce flexibility The reciprocal θ -1 ∈ [1, +∞) is the composite climate change vul-
into the mathematical structure of composite environmental indices. nerability index CCVI, with larger values of θ -1 indicating higher levels of
One such attempt, by Brooks et al. (2005), used statistical methods to vulnerability. θ -1 provides a ranking of nations on the basis of their
identify key indicators and to introduce a degree of flexibility into their overall vulnerability to climate change, independently of any other
relative importance in constructing an aggregate vulnerability index for national characteristics, which are ignored in the present analysis.
ranking nations. The CCVI constructed in the present analysis in- The multiplier program in Table 1 calculates for nation “o” a vector
troduces flexibility by adopting a mathematical programming approach of endogenous weights νn ∈ (0,+∞) with which to aggregate its sec-
proposed by Melyn and Moesen (1991), and extended by Cherchye toral vulnerability indices into its CCVI. For a nation with excessive
et al. (2007), that generates a “benefit-of-the-doubt” composite index amounts of sectoral vulnerability index xn the program will implicitly
whose aggregation weights are endogenously generated, and thus free attach a small weight to this sectoral index in an effort to minimize its
to differ across sectors and nations, thereby respecting their varying overall vulnerability score. A very small weight for a specific sector
circumstances. Recent applications of endogenous weight aggregation suggests that the marginal improvements from mitigation efforts in that
in an environmental context include Zhou, Delmas and Kohli (2017), sector are likely to be infinitesimal. By reflecting different vulnerability
and Edmonds et al. (2017). The set of sectoral indices to be aggregated across sectors that in turn reflect different national circumstances, these
are identified by adopting a nonparametric statistical approach devel- weights have the potential to assist in resource allocation and the de-
oped by Pastor et al. (2002) that identifies superfluous sectoral indices sign of policies intended to reduce vulnerability, as noted in Section 1
that may be discarded. with reference to Mendelsohn (2006), Anderson et al. (2019), and
Weaver and Miller (2019). The IPCC refers to these policies as adap-
3. Methodology and data tation and mitigation pathways.
By the duality theorem of linear programming, at optimum θ = μ,
The use of DEA and its variants has spread to the construction of a and can therefore be expressed as an endogenously-weighted sum of its
N
host of composite indices, most notably alternatives to the HDI sectoral vulnerability indices, θ = ∑n = 1 νn xni for any nation i = 1,…,o,
(Despotis, 2005; Cherchye et al., 2008) and recently to the construction …,I. These endogenous weights provide a considerable improvement
of a variety of composite environmental indices, at micro (Zhou et al., over the fixed weights used in most composite indices. Fixed weights
2006; Bellenger and Herlihy, 2009; Edmonds et al., 2015; Zhou et al., impose perfect substitutability among sectoral indices, with rates of
2017; Liu et al., 2019) and macro (Färe et al., 2004; Zhou et al., 2007; substitution constant across nations. The endogenous weights gener-
Lo, 2010; Edmonds et al., 2017; Huang et al., 2018) levels. The CCVI ated by DEA also impose perfect substitutability among sectoral indices,
was also constructed using a variant of DEA. but with the important advantage that weights, and rates of substitution
Let nations be indexed by i = 1,…,I, and let a nation’s vulnerability among sectoral indices, are allowed to differ across nations according to
be tracked across N sectors with sectoral vulnerability indices labelled a nation’s circumstances. Weight flexibility is particularly important in
xn and indexed by n = 1,…,N. The DEA program that evaluates the the construction of composite vulnerability indices, since nations differ
aggregate vulnerability to climate change of nation “o” is given by the in their exposure, sensitivity and adaptive capacity across sectors.
dual pair of linear programs in Table 1. As an auxiliary benefit of generating endogenous weights of sectoral
The envelopment and multiplier programs contain sectoral vulner- vulnerability indices, DEA provides a methodology for implementing
ability indices xn but no additional variables that might influence vul- Occam’s Razor, which states that when faced with competing models
nerability such as geographic location, national income or form of that make the same predictions, one should select the model with the
fewest assumptions. In the context of this study DEA provides a fra-
Table 1 mework within which to conduct a statistical test of the hypothesis that
DEA CCVI programs. a candidate sectoral vulnerability index xn is superfluous in the con-
struction of a CCVI, and can be deleted from the initial list of N sectoral
Envelopment Program Multiplier Program
indices, because deleting it has an insignificant impact on the overall
minθ,λ θ maxμ,ν μ
subject to subject to CCVI used to rank nations.
I
θxno - ∑i = 1 λi xni ≥ 0 n = 1,…,N
N
∑n = 1 νn xno = 1 The procedure is to run DEA using all N sectoral vulnerability in-
λi ≥ 0 i = 1,…,I N dices, and to compare the resulting CCVI(N) ranking with another CCVI
μ − ∑n = 1 νn xni ≤ 0 i = 1,…,I
(N-1) ranking obtained by deleting a candidate sectoral vulnerability
μ,νn ≧ 0 n = 1,…,N
Calculating CCVI index. The candidate sectoral vulnerability index is that index having

3
H.K. Edmonds, et al. Ecological Indicators 117 (2020) 106529

the fewest positive weights, and the smallest average weight, among the Table 2
N indices in CCVI(N). If the CCVI(N-1) ranking is sufficiently similar to Summary statistics for sector vulnerability indicators.
the CCVI(N) ranking, the candidate index is deemed superfluous and Ecosystem Food Human Health Infrastructure Water
deleted. Somewhat more formally, if a sufficiently large number of Services Habitat
nations experiences a sufficiently small reduction in their CCVI when a
Mean 0.4451 0.5266 0.4673 0.4176 0.3478 0.3258
candidate sectoral vulnerability index is deleted, that index is deemed
Std dev 0.0885 0.1377 0.1114 0.1698 0.1081 0.0938
superfluous and removed. The test is a nonparametric statistical test Max 0.6588 0.8015 0.7454 0.8104 0.7688 0.6921
based on the distribution of the ratio γ = CCVI(N-1)/CCVI(N) ≦ 1. Min 0.2251 0.2027 0.2213 0.1778 0.0845 0.0424
− −
From γ calculate Tn = {1 if γn < γ and 0 otherwise}, where γ is the
largest acceptable reduction in CCVI. Finally define the test statistic
N
T = ∑n = 1 Tn , which counts the number of nations whose CCVI declines vulnerability is particularly severe in the Infrastructure and Water
− −
by more than γ . Under the null hypothesis that γ ≦ γ this statistic is sectors. This dispersion highlights the need for a weighting and ag-
distributed as binomial (N,po), with 0 < po < 1 a probability level. gregation methodology that respects and incorporates nations’ differing
Pastor et al. (2002) provide details of the statistical test of similarity. distributions of vulnerability to climate change.
Summarizing, the methodology used in this analysis makes three The analysis of these data proceeded as follows: (i) a composite
contributions to the construction of a composite vulnerability index: (i) climate change vulnerability index CCVI(6) was created with the DEA
the ability to generate endogenous weights with which to aggregate program in Table 1 to aggregate six sectoral vulnerability indices; (ii)
sectoral indices; (ii) the ability of these weights to guide policy intended after testing the null hypothesis that a sector is redundant and can be
to lower the cost of reducing vulnerability in an efficient, i.e., resource- deleted from CCVI(6) without statistically significant loss of informa-
saving, manner; and (iii) the ability to identify superfluous sectoral tion, one sector was deemed redundant, and deleted; and (iii) DEA was
indices. again used to aggregate vulnerability indices across five key sectors to
National vulnerability data are available from The University of create another composite climate change vulnerability index CCVI(5).
Notre Dame (2015a). The data are normalized on [0,1], with higher
values indicating greater vulnerability, and arranged in matrix format, 4. Results and discussion
with three components (exposure, sensitivity and adaptive capacity) that
influence vulnerability, and six sectors (ecosystem services, food, human The six-sector endogenously weighted vulnerability index CCVI(6)
habitat, health, infrastructure and water) across which vulnerability is differs significantly from the Notre Dame vulnerability index ND that
tracked. Each sectoral vulnerability index is itself an arithmetic mean of uses fixed weights to aggregate vulnerability across six sectors, with the
its six component indicators. To illustrate, the Health index is the ar- probability of the two indices being equal less than 1%. The dis-
ithmetic mean of two exposure indicators (projected change of deaths crepancy between the two indices is due to their different weighting
from climate change induced diseases and projected change of length of schemes; all 141 nations are forced to assign equal weights of 1/6 to
transmission season of vector-borne diseases), two sensitivity indicators each sector in the ND index, despite their widely differing vulner-
(slum population and dependency on external resources for health abilities identified in Table 2, but they exploit the freedom to assign
services) and two adaptive capacity indicators (medical staffs and ac- different weights that reflect their different vulnerabilities across sec-
cess to improved sanitation facilities). Details are available from The tors in the CCVI(6) index, a distinct information advantage over the
University of Notre Dame (2015b). The three components are those fixed weights in ND. The two indices are depicted in Fig. 1, with linear
identified as vulnerability drivers in the IPCC Assessment Reports, and trend line in red.
the six sectors are generally consistent with those identified in IPCC Despite the significant difference between the two vulnerability
Assessment Reports and in the academic literature (e.g., Brooks et al., indices, there is strong agreement between them on the identity of the
2005; Eriksen and Kelly, 2007; Hallegatte, 2009; and Füssel, most vulnerable nations, with CCVI(6) agreeing on 11 of the 15 most
2010). Data obtained from the vulnerability component have supported vulnerable nations in ND, contradicting an assertion of Eriksen and
research, particularly as it has related to its determinants (Sarkodie and Kelly that commonly used vulnerability indices fail to agree on the
Strezov, 2019) and its impact on human migration from less developed identity of the most vulnerable countries. In addition, 14 of the 15 most
to more developed nations (Grecequet et al., 2017). vulnerable nations in CCVI(6), and 12 of the 15 most vulnerable na-
The analysis was conducted using the six sectoral vulnerability in- tions in ND, are among the 47 nations in the Least Developed Country
dices, which comprised an unbalanced panel covering up to 192 nations (LDC) category of the United Nations Department of Economic and
over varying time periods beginning in 1995. The 2016 data were Social Affairs and Committee for Development Policy (2018).
chosen in an effort to include as many nations as possible. Eliminating The summary statistics for the endogenous weights for all six sectors
nations failing to report a value for one or more sectors reduced the
sample size in the present study to 141 nations. In contrast, the Notre
Dame Vulnerability Index is available for all 192 nations because it
aggregates as few as 4 of 6 sectors if one or two sectors are missing data,
with exogenous weights of 1/4 or 1/5 replacing 1/6 for these 51 na-
tions. This practice implicitly assigns zero weights to missing sectors in
51 nations and weights of 1/6 to these sectors in 141 nations.
Unfortunately, many of the 51 nations eliminated from the CCVI
data are among the most vulnerable developing nations, SIDS in par-
ticular. Of 38 UN member SIDS, the sample contains just 18, including
four in the Pacific and nine in the Caribbean. However not all elimi-
nated nations are among the most vulnerable; two of the wealthiest
nations, Liechtenstein and Monaco, also were eliminated.
Summary statistics for six sectoral vulnerability indices and 141
nations appear in Table 2. Even after normalization, considerable dis-
persion remains within each sectoral vulnerability index, as indicated
by the Max – Min range within each sectoral index. Dispersion in
Fig. 1. CCVI(6) and ND Vulnerability Indices.

4
H.K. Edmonds, et al. Ecological Indicators 117 (2020) 106529

Table 3
Summary statistics for sector endogenous weights.
Ecosystem Food Human Health Infrastructure Water
Services Habitat

Mean 0.1910 0.1016 0.2340 0.0377 0.1686 0.2742


Std dev 0.2038 0.1328 0.2013 0.1423 0.1484 0.2542
Max 0.8228 0.6794 0.9434 1.1593 0.4770 0.9344
Min 0 0 0 0 0 0
# > 0 97 81 108 16 112 108

for all 141 nations in Table 3 illustrate the wide variation in weights
assigned by nations, both within and across sectors. These en-
dogenously selected weights differ dramatically from the fixed ND
weights of 1/6 (0.1667) for all sectors and all nations.
Table 3 also indicates that one sector, Health, has by far the smallest
average weight, and the least number of positive weights, in the sample. Fig. 2. CCVI(5) and ND Vulnerability Indices.
This is because many nations are relatively vulnerable in the Health
sector, and so DEA assigns a low weight to Health in an effort to the highest weight in the sample to the Health sector, and has γ = 0.93.
minimize their overall vulnerability CCVI(6). Since 125 of 141 nations The next two most affected nations are France (γ = 0.95) and Cuba
assign a zero weight to the Health sector, this in turn suggests that the (γ = 0.96). All three have excellent health care systems, and would
Health sector contributes little to CCVI(6), and is a candidate for de- benefit most from inclusion of the Health sector. Nonetheless the null
letion. Table 4 provides supporting evidence. It shows that Health hypothesis that the Health sector can be deleted could not be rejected,
sector vulnerability is highly correlated with vulnerability in the Food since the probability that the distributions of the two vulnerability in-
sector and, reading across the Health row, Health is the sector that is dices differ is less than 1%. The two vulnerability indices are essentially
the most highly correlated with each of the remaining sectors. It follows the same, a conclusion illustrated in Fig. 3. The endogenous weights in
that high Health vulnerability goes hand in hand with, and may be CCVI(5) are almost identical to the same five weights in CCVI(6) re-
driven by, high vulnerability in the remaining sectors, particularly ported in Table 3. Following the same testing procedure, the null hy-
Food. The World Health Organization (2018, nd, nd, nd) has agreed, pothesis that any sector other than Health can be deleted from CCVI(6)
and has asserted causality. It has found that adverse impacts of climate without statistically significant loss of information could be rejected.
change on food security, safe water and sanitation, ecosystem services, The same testing procedure was followed to reject the hypothesis that
human habitat, and energy infrastructure all exert a negative influence any one of the five remaining sectors can be deleted from the composite
on human health. This does not mean that human health has no impact climate change vulnerability index CCVI(5).
on a nation’s vulnerability; it does influence vulnerability, but much of The global distribution of CCVI(5) is mapped in Fig. 4. The most
its impact is attributable to vulnerability in other sectors. The re- vulnerable nations are located in Sub-Saharan Africa, South Asia and
mainder may be due to a lag effect; throughout its Assessment Reports among the few SIDS in the data. Most of these nations are included in
the IPCC has asserted, with very high confidence, that climate change the global 1.5 °C hotspot regions identified by Schleussner et al. (2018),
affects human health primarily by exacerbating existing health pro- the extremely and significantly vulnerable regions identified by Yohe
blems. et al. (2006), and the most vulnerable developing nations identified by
The reduced index CCVI(5) also differs significantly from the ND Closset et al. (2017). As IPCC has claimed throughout its Assessment
index, although CCVI(5) also agrees on 11 of the 15 most vulnerable Reports, vulnerability varies with non-climatic variables, with location
nations in ND. The two indices are depicted in Fig. 2, with linear trend being closely correlated with the components of the three sub-indices of
line in red. The strong similarity between Fig. 1 and Fig. 2 provides an the Economic Vulnerability Index.
initial indication that the distribution of CCVI(5) may not differ sig-
nificantly from that of CCVI(6), and that the Health sector can be de-
leted without significant loss of information about the distribution of 5. Conclusions
vulnerability to climate change across nations.
Using the ratio γ = CCVI(5)/CCVI(6) ≦ 1, the results show a mean The results illustrate the analytical and practical contributions of
value of γ = 0.998, with γ = 1 for the 125 nations assigning a zero the methodology to the study of national vulnerability to climate
weight to the Health sector, and for which deleting the Health sector change.
has absolutely no impact on its CCVI, and 0.99 < 1 < γ for six na-
tions for which deleting the Health sector has a barely perceptible
impact on its CCVI. The most affected nation, The Netherlands, assigns

Table 4
Correlation coefficients among sector vulnerability indexes.
Health Ecosystem Food Human Infrastructure Water
Services Habitat

Health 1 0.600 0.807 0.630 0.297 0.308


Ecosystem 1 0.583 0.409 0.018 0.232
Services
Food 1 0.492 0.157 0.295
Habitat 1 0.123 0.064
Infrastructure 1 0.111
Water 1
Fig. 3. CCVI(6) and CCVI(5) Vulnerability Indices.

5
H.K. Edmonds, et al. Ecological Indicators 117 (2020) 106529

Fig. 4. Global Distribution of CCVI(5).

(i) Exogenously-generated fixed-weight aggregation has been widely, composite climate change vulnerability index without statistically
and properly, criticized as a methodology for constructing com- significant loss of information, and with potentially considerable
posite indices. The Notre Dame vulnerability index ND is one of cost savings, was conducted. This test helps to identify a minimal
several environmental indices based on a fixed-weight aggregation set of key sectors whose inclusion is essential to the construction of
methodology. An alternative aggregation methodology has been an informative vulnerability index. The World Health Organization
proposed and implemented here, one that is popular in business has provided support for the deletion of the Health sector from the
and economics, and that generates endogenous-weight aggrega- composite index, on the basis that human health is itself a function
tion. National data from the University of Notre Dame Global of other sectoral indices such as food security. From a methodo-
Adaptation Initiative have been used to compare climate change logical perspective, explanatory power can be improved by re-
vulnerability indices generated by the ND and CCVI methodolo- moving the Health index or replacing it with another index that is
gies, and the two indices are found to differ in a statistically sig- less highly correlated with other sectoral indices.
nificant way. (v) The most vulnerable nations were identified. They are located
(ii) The endogenous weights in the CCVI index may be interpreted as primarily in Sub-Saharan Africa and South Asia, and among the
shadow prices that respect differences in nations’ exposure, sensi- SIDS, in accordance with previous findings.
tivity and adaptive capacity to climate change, and it is argued that
these weights have the potential to serve as proxies for missing The methodology and empirical findings have generated two re-
market prices with policy implications for the development of commendations for future research.
adaptation and mitigation pathways designed to reduce the cost of
limiting vulnerability to climate change. (i) By studying what sectoral source(s) of national vulnerability can be
(iii) The orientation of a DEA program is specified by the analyst, deleted without statistically significant loss of information, this
ideally guided by policy considerations. The orientation of the DEA analysis responds to the challenge of quantifying the magnitude of
program in the present study is guided by the belief that vulner- variation in vulnerability across nations in a parsimonious way. An
ability to climate change is an undesirable characteristic of a na- altogether different challenge might also be addressed by asking
tion, and is therefore to be minimized, presumably through policy what variable(s) or indices could be added to the analysis that
actions designed to increase adaptation and mitigation efforts. It is would provide a statistically significant gain in information. This
possible to reverse the orientation of the DEA program from one of alternative challenge would seek to identify the sources of variation
minimizing vulnerability to one of maximizing vulnerability, al- in vulnerability across nations. In its Assessment Reports the IPCC
though the policy considerations behind the reversal are unclear. has identified numerous non-climatic sources. Geographic variables
However in this case the DEA program would assign relatively (Nordhaus, 2006) could be added to these, such as distance to the
large aggregation weights to the Health sector for the majority of equator, a popular explanatory variable in the economic develop-
nations. ment literature (Hall and Jones, 1999). An alternative candidate is
(iv) A nonparametric statistical test of the null hypothesis that a vul- an index of social economic progress, as conceived by Stiglitz et al.
nerability sector is superfluous and can be deleted from a (2009) and currently under development at the OECD (2018). The

6
H.K. Edmonds, et al. Ecological Indicators 117 (2020) 106529

two challenges are complementary. The first challenge has been Change: The PVCCI, an Index to be used for International Development Policies,”
addressed in the present study, and research into the second can be Policy Brief 190, FERDI. http://www.ferdi.fr/en/publication/b190-measuring-phy-
sical-vulnerability-climate-change-pvcci-index-be-used-international (accessed 26/
accomplished simply by adding the desired additional variables to 07/2019).
the DEA program in Table 1 and following the same testing pro- Füssel, H.-M. (2009), “Review and Quantitative Analysis of Indices of Climate Change
cedure. Exposure, Adaptive Capacity, Sensitivity, and Impacts; Background Note to the World
Development Report 2010 ,” http://documents.worldbank.org/curated/en/
(ii) The Notre Dame Vulnerability Index ND contains 192 nations in 885101468164982209/pdf/520590WP0Adapt1round0note101PUBLIC1.pdf.
2016, while the index CCVI generated by DEA contains just 141 Füssel, H.-M., 2010. How inequitable is the global distribution of responsibility, cap-
nations. DEA requires an equal number of sectors for each nation, ability and vulnerability to climate change: A comprehensive indicator-based as-
sessment. Global Environ. Change 20, 597–611.
so 51 nations had to be removed from the data, many of which are Gallopín, G.C., 2006. Linkages between vulnerability, resilience and adaptive capacity.
LDCs and SIDS. Kuosmanen (2009) has proposed a DEA model that Global Environ. Change 16, 293–303.
can accommodate missing observations, essentially replacing Gan, X., Fernandez, I.C., Guo, J., Wilson, M., Zhao, Y., Zhou, B., Wu, J., 2017. When to
use what: methods for weighting and aggregating sustainability indicators. Ecol. Ind.
missing observations with a “big M” from linear programming. This
81, 491–502.
approach is a promising way of increasing degrees of freedom in Grecequet, M., DeWaard, J., Hellman, J.J., Abel, G.J., 2017. Climate vulnerability and
any DEA analysis, the present study of the vulnerability of nations human migration in global perspective. Sustainability 9, 720–730.
to climate change in particular. Greco, S., Ishizaka, A., Tasiou, M., Torrisi, G., 2019. On the methodological framework of
composite indices: a review of the issues of weighting, aggregation, and robustness.
Soc. Indic. Res. 141, 61–94.
Declaration of Competing Interest Guillaumont, P., 2009. An Economic vulnerability index: its design and use for interna-
tional development policy. Oxford Development Studies 37, 193–228.
Guillaumont, P., 2015. Measuring vulnerability to climate change for allocating funds to
The authors declare that they have no known competing financial adaptation. Chapter 35. In: Barrett, S., Carraro, C., de Melo, J. (Eds.), Towards a
interests or personal relationships that could have appeared to influ- Workable and Effective Climate Regime. CEPR Press, London.
Guillaumont, P., and C. Simonet (2011a), “Designing an Index of Physical Vulnerability to
ence the work reported in this paper. Climate Change,” Policy Brief 18, FERDI. http://www.ferdi.fr/sites/www.ferdi.fr/
files/publication/fichiers/B18.pdf (accessed 26/07/2019).
References Guillaumont, P., and C. Simonet (2011b), “To What Extent are African Countries
Vulnerable to Climate Change? Lessons from a New Indicator of Physical
Vulnerability to Climate Change,” Working Paper 8, FERDI. http://www.ferdi.fr/
Adger, W.N., 2006. Vulnerability. Global Environ. Change 16, 268–281. sites/www.ferdi.fr/files/publication/fichiers/I08_WEB.pdf (accessed 26/07/2019).
Adolphson, D.L., Cornia, G.C., Walters, L.C., 1991. A unified framework for classifying Hall, R.E., Jones, C.I., 1999. Why do some countries produce so much more output per
DEA models. Operational Research ’90. Pergammon Press, New York. worker than others? Quart. J. Econ. 114, 83–116.
Alkire, S., C. Jindra, G. Robles and A. Vaz (2016), “Multidimensional Poverty Index – Hallegatte, S., 2009. Strategies to adapt to an uncertain climate change. Global Environ.
Summer 2016: Brief Methodological Note and Results,” OPHI Briefing 42. https:// Change 19, 240–247.
ophi.org.uk/multidimensional-poverty-index-summer-2016-brief-methodological- Hinkel, J., 2011. Indicators of vulnerability and adaptive capacity: Towards a clarification
note-and-results/. of the science-policy interface. Global Environ. Change 21, 198–208.
Anderson, S.E., Anderson, T.L., Hill, A.C., Kahn, M.E., Kunreuther, H., Libecap, G.D., Huang, J., Xia, J., Yu, Y., Zhang, N., 2018. Composite eco-efficiency indicators for china
Mantripragada, H., Mérel, P., Plantinga, A.J., Smith, V.K., 2019. The critical role of based on data envelopment analysis. Ecol. Ind. 85, 674–697.
markets in climate change adaptation. Climate Change Econ. 10. https://doi.org/10. IPCC (2001), “Summary for Policymakers” in Climate Change 2001: Impacts, Adaptation
1142/S2010007819500039. and Vulnerability. Contribution of Working Group II to the Third Assessment Report
Barnett, J., Lambert, S., Fry, I., 2008. The hazards of indicators: insights from the en- of the Intergovernmental Panel on Climate Change. Cambridge UK: Cambridge
vironmental vulnerability index. Ann. Assoc. Am. Geogr. 98, 102–119. University Press.
Bellenger, M.J., Herlihy, A.T., 2009. An economic approach to environmental indices. IPCC (2014a), “Summary for Policymakers,” in Climate Change 2014: Impacts,
Ecol. Econ. 68, 2216–2223. Adaptation, and Vulnerability. Part A: Global and Sectoral Aspects. Contribution of
Beroya-Eitner, M.A., 2016. Ecological Vulnerability Indicators. Ecol. Ind. 60, 329–334. Working Group II to the Fifth Assessment Report of the Intergovernmental Panel on
Böhringer, C., Jochem, P.E.P., 2007. Measuring the Immeasurable: A Survey of Climate Change. Cambridge UK: Cambridge University Press.
Sustainability Indices. Ecol. Econ. 63, 1–8. IPCC (2014b), “Adaptations, Opportunities, Constraints, and Limits,” in Climate Change
Bowen, H.P., Moesen, W., 2011. Composite competitiveness indicators with endogenous 2014: Impacts, Adaptation, and Vulnerability. Part A: Global and Sectoral Aspects.
versus predermined weights: an application to the world economic forum’s global Contribution of Working Group II to the Fifth Assessment Report of the
competitiveness index. Competitiveness Rev. 21, 129–151. Intergovernmental Panel on Climate Change. Cambridge UK: Cambridge University
Brenkert, A.L., Malone, E.L., 2005. Modeling vulnerability and resilience to climate Press.
change: A case study of India and Indian States. Clim. Change 72, 57–102. IPCC (2018), “Summary for Policymakers,” in Global Warming of 1.5C. An IPCC Special
Brooks, N., Adger, W.N., Kelly, P.M., 2005. The determinants of vulnerability and Report on the Impacts of Global Warming of 1.5C above Pre-Industrial Levels and
adaptive capacity at the national level and the implications for adaptation. Global Related Global Greenhouse Gas Emission Pathways, in the Context of Strengthening
Environ. Change 15, 151–163. the Global Response to the Threat of Climate Change, Sustainable Development, and
Charnes, A., Cooper, W.W., Rhodes, E., 1978. Measuring efficiency of decision making Efforts to Eradicate Poverty. Cambridge UK: Cambridge University Press.
units. Eur. J. Oper. Res. 2, 429–444. Klugman, J., Rodríguez, F., Choi, H.-J., 2011. The HDI 2010: New Controversies, Old
Cherchye, L., Moesen, W., Rogge, N., Van Puyenbroeck, T., 2007. An introduction to Critiques. J. Econ. Inequality 9, 249–288.
‘Benefit of the Doubt’ composite indicators. Soc. Indic. Res. 82, 111–145. Kuosmanen, T., 2009. Data envelopment analysis with missing data. J. Operat. Res. Soc.
Cherchye, L., Ooghe, E., Van Puyenbroeck, T., 2008. Robust human development rank- 60, 1767–1774.
ings. J. Econ. Inequality 6, 287–321. Liu, X., Xiang, J., Douqing, Z., Yang, J., Wang, Y., 2019. How public environmental
Closset, M., S. Feindouno, P. Guillaumont and C. Simonet (2017), “A Physical concern affects the sustainable development of Chinese cities: an empirical study
Vulnerability to Climate Change Index: Which are the Most Vulnerable Developing using extended DEA Models. J. Environ. Manage. 251. https://doi.org/10.1016/j.
Countries?” Policy Brief 213, FERDI. http://www.ferdi.fr/sites/www.ferdi.fr/files/ jenvman.2019.109619.
publication/fichiers/p213-ferdi-closset-feindouno-guillaumont-simonet_-octobre_ Lo, S.-F., 2010. The differing capabilities to respond to the challenge of climate change
2018_0.pdf (accessed 26/07/2019). across annex parties under the Kyoto protocol. Environ. Sci. Policy 13, 42–54.
Despotis, D.K., 2005. A reassessment of the human development index via data envel- Lovell, C.A.K., Pastor, J.T., 1999. Radial DEA models without inputs or without outputs.
opment analysis. J. Operat. Res. Soc. 56, 969–980. Eur. J. Oper. Res. 118, 46–51.
Eakin, H., Luers, A.L., 2006. Assessing the vulnerability of social-environmental systems. Melyn, W., and W. Moesen (1991), “Towards a Synthetic Indicator of Macroeconomic
Annu. Rev. Environ. Resour. 31, 365–394. Performance: Unequal Weighting when Limited Information is Available,” Public
Edmonds, H.K., Lovell, J.E., Lovell, C.A.K., 2015. A framework for guiding the manage- Economics Research Paper 17, CES, KU Leuven, Belgium.
ment of urban stream health. Ecol. Econ. 109, 222–233. Mendelsohn, R., 2006. The role of markets and governments in helping society adapt to a
Edmonds, H.K., Lovell, J.E., Lovell, C.A.K., 2017. A new composite index for greenhouse changing climate. Clim. Change 78, 203–215.
gases: climate science meets social science. Resources 6, 62. https://doi.org/10. Nardo, M., Saisana, M., Saltelli, A., Tarantola, S., Hoffmann, A., Giovannini, E., 2008.
3390/resources6040062. Handbook on Constructing Composite Indicators: Methodology and User Guide.
Eriksen, S.H., Kelly, P.M., 2007. Developing credible vulnerability indicators for climate OECD, Paris https://www.oecd.org/sdd/42495745.pdf.
adaptation policy assessment. Mitig. Adapt. Strat. Glob. Change 12, 495–524. Nordhaus, W. D. (2006), “Geography and Macroeconomics: New Data and New Findings,”
Färe, R., Grosskopf, S., Hernandez-Sancho, F., 2004. Environmental Performance: An Proceedings of the National Academy of Sciences 103, 3510-3517.
Index Number Approach. Resour. Energy Econ. 26, 343–352. OECD (nd), Glossary of Statistical Terms. https://stats.oecd.org/glossary/detail.
Feinduono, S., and M. Goujon (2016), “The Retrospective Economic Vulnerability Index, asp?ID=6278.
2015 Update,” Working Paper 147, FERDI. https://ferdi.fr/en/publications/the-ret- OECD (2018), Beyond GDP: Measuring What Counts for Economic and Social
rospective-economic-vulnerability-index-2015-update. Performance. https://www.oecd.org/social/beyond-gdp-9789264307292-en.htm.
Feindouno, S., and P. Guillaumont (2019), “Measuring Physical Vulnerability to Climate Pastor, J.T., Ruiz, J.L., Sirvent, I., 2002. A statistical test for nested radial DEA models.

7
H.K. Edmonds, et al. Ecological Indicators 117 (2020) 106529

Oper. Res. 50, 728–735. https://unfccc.int/Adaptation-Fund.


Preston, B.L., Dow, K., Berkhout, F., 2013. The climate adaptation frontier. Sustainability University of Notre Dame (2015a) University of Notre Dame Global Adaptation Initiative.
5, 1011–1035. https://gain.nd.edu/our-work/country-index/.
Sarkodie, S.A., Strezov, V., 2019. Economic, social and governance adaptation readiness University of Notre Dame (2015b), University of Notre Dame Global Adaptation Index:
for mitigation of climate change vulnerability: evidence from 192 countries. Sci. Country Index Technical Report. https://gain.nd.edu/assets/254377/nd_gain_tech-
Total Environ. 656, 150–164. nical_document_2015.pdf.
Schleussner, C.-F., et al., 2018. 1.5℃ hotspots: climate hazards, vulnerabilities, and im- Weaver, C.P., Miller, C.A., 2019. A framework for climate change-related research to
pacts. Annu. Rev. Environ. Resour. 43, 135–163. inform environmental protection. Environ. Manage. 64, 245–257.
Seth, S., McGillivary, M., 2018. Composite indices, alternative weights, and comparison Weißhuhn, P., Müller, F., Wiggering, H., 2018. Ecosystem vulnerability review: proposal
robustness. Soc. Choice Welfare 51, 657–679. of an interdisciplinary ecosystem assessment approach. Environ. Manage. 61,
South Pacific Applied Geoscience Commission and United Nations Environmental 904–915.
Programme (2005), Environmental Vulnerability Index: Measuring the Resilience of Wendling, Z.A., Emerson, J.W., Esty, D.C., Levy, M.A., de Sherbinin, A., et al., 2018. 2018
Our Planetary Life-Support Systems http://www.vulnerabilityindex.net/. Environmental Performance Index. Yale Center for Environmental Law & Policy, New
Stiglitz, J. E., Sen, A., and Fitoussi, J-P (2009), “The Measurement of Economic Haven CT https://epi.yale.edu.
Performance and Social Progress Revisited,” OFCE Centre de recherché en économie World Health Organization (2018), The State of Food Security and Nutrition in the World
de sciences Po. Working Paper 2009-33 (December). www.ofce.sciences-po.fr/pdf/ 2018. https://www.who.int/nutrition/publications/foodsecurity/state-food-se-
dtravail/WP2009-33.pdf. curity-nutrition-2018-flyer-en.pdf?ua=1.
United Nations Department of Economic and Social Affairs and Committee for World Health Organization (nd), “Water, Health and Ecosystems”. https://www.who.int/
Development Policy, 2018. Handbook on the Least Developed Country Category: heli/risks/water/water/en/.
Inclusion, third ed. Graduation and Special Support Measures https://www.un.org/ World Health Organization (nd), “Climate Change and Human Health: Ecosystem Goods
development/desa/dpad/publication/handbook-on-the-least-developed-country-ca- and Services for Health”. https://www.who.int/globalchange/ecosystems/en/.
tegory-inclusion-graduation-and-special-support-measures-third-edition/. World Health Organization (nd),” Health and Sustainable Development: Energy Access
United Nations Development Programme (nd), Human Development Index (HDI) http:// and Resilience”, https://www.who.int/sustainable-development/health-sector/
hdr.undp.org/en/content/human-development-index-hdi. health-risks/energy-access/en/.
United Nations Environment Programme (UNEP) (2014), The Adaptation Gap Report: A Yohe, G., Malone, E., Brenkert, A., Schlesinger, M., Meij, H., Xing, X., 2006. Global dis-
Preliminary Assessment. https://www.unenvironment.org/resources/adaptation- tributions of vulnerability to climate change. Integ. Assess. Bridg. Sci. Policy 10,
gap-report-2014. 35–44.
United Nations Environment Programme (UNEP) (2016), The Adaptation Gap Report: Zhou, P., Ang, P.W., Poh, K.L., 2006. Comparing aggregating methods for constructing
Finance. http://www.unepdtu.org/newsbase/2016/05/uneps-adaptation-finance- the composite environmental index: an objective measure. Ecol. Econ. 59, 305–311.
gap-report-released?id=377aa3d4-32c1-4100-8bee-ae65390b60ba. Zhou, P., Ang, P.W., Poh, K.L., 2007. A Mathematical programming approach to con-
United Nations Environment Programme (UNEP) (2017), The Adaptation Gap Report: structing composite indicators. Ecol. Econ. 62, 291–297.
Towards Global Assessment. https://www.unenvironment.org/resources/adaptation- Zhou, P., Delmas, M.A., Kohli, A., 2017. Constructing meaningful environmental indices:
gap-report-2017. a nonparametric frontier approach. J. Environ. Econ. Manage. 85, 21–34.
United Nations Environment Programme (UNEP) (2018), The Adaptation Gap Report: Zhou, P., Zhang, L.P., 2018. Composite indicators for sustainability assessment: metho-
Health. https://www.unenvironment.org/resources/adaptation-gap-report. dological developments, Chapter 2. In: Pang, R., Bai, X., Lovell, K. (Eds.), Energy,
United Nations Framework Convention on Climate Change (2019), “Adaptation Fund” Environment and Transitional Green Growth in China. Springer, Singapore.

You might also like