Professional Documents
Culture Documents
2003 Luers Meth Vuln
2003 Luers Meth Vuln
Abstract
We propose measuring vulnerability of selected outcome variables of concern (e.g. agricultural yield) to identified stressors (e.g.
climate change) as a function of the state of the variables of concern relative to a threshold of damage, the sensitivity of the variables
to the stressors, and the magnitude and frequency of the stressors to which the system is exposed. In addition, we provide a
framework for assessing the extent adaptive capacity can reduce vulnerable conditions. We illustrate the utility of this approach by
evaluating the vulnerability of wheat yields to climate change and market fluctuations in the Yaqui Valley, Mexico.
r 2003 Elsevier Ltd. All rights reserved.
0959-3780/$ - see front matter r 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/S0959-3780(03)00054-2
ARTICLE IN PRESS
256 A.L. Luers et al. / Global Environmental Change 13 (2003) 255–267
and change, and market fluctuations. Finally, we within the interactions of social and ecological systems
conclude with a discussion of the implications of our (Turner et al., 2003a, b).
analysis for future research and practice.
2.2. Quantifying vulnerability
and coping strategies (e.g. staple food production, access sets of stressors. This approach requires a set of generic
to infrastructure). The Pacific Northwest Laboratory metrics that can assess the relationships between a wide
(PNL) Vulnerability Assessment Program also uses a range of stressors and outcome variables of concern. To
composite approach to develop an index of vulnerability the extent that the selected variables of concern and the
to climate change for 38 countries (Moss et al., 2000, stressors characterize a given place, the vulnerability of
2002). The PNL vulnerability index represents a these variables may provide important characterization
composite of 16 variables selected from five sectors of of the vulnerability of that place. The selected variables
sensitivity (settlement, food security, human health, and stressors are likely to change over time and space,
ecosystem, and water) and three sectors for coping resulting in changes in the relative vulnerability; how-
capacity (economic, human resources and environmen- ever, the functional form of the generic vulnerability
tal). Examples of the variables include life expectancy, metric remains the same.
percent of unmanaged land, and GDP per capita. A few generic vulnerability metrics have been
Additional examples of the composite indicator ap- proposed. For example, the variability of selected
proach include the South Pacific Applied Geosciences variables of concern has been applied as a metric of
Commission (SOPAC) environmental vulnerability in- vulnerability especially in economic and agricultural
dex (EVI), which is a composite of 54 independent studies (Pritchett et al., 2000; Heitzmann et al., 2002).
variables categorized under the headings—degradation, Another generic metric is the probability that a variable
resilience, exposure (Briguglio, 1995; Kaly et al., 2002) of concern will cross a threshold (Schimmelpfennig and
and the Index of Human Security, which is a composite Yohe, 1999; Mansuri and Healy, 2002; Peterson, 2002).
of 16 indicators drawn from four major thematic areas While both of these measures are useful, neither is
(environment, economy, society, and institutions) sufficient to fully capture vulnerability. For example,
(Lonergan et al., 2000). contrast an elite business woman who might have an
While the indicator approach is valuable for monitor- extremely variable income but a mean income of well
ing trends and exploring conceptual frameworks, indices above a regional average with a poor laborer who may
are limited in their application by considerable sub- have a low base income but with much less variability.
jectivity in the selection of variables and their relative In this case, it is apparent that variability in income
weights, by the availability of data at various scales, and alone might not fully capture the relative vulnerabilities
by the difficulty of testing or validating the different accurately.
metrics. Perhaps most importantly, the indicator
approach often leads to a lack of correspondence
between the conceptual definition of vulnerability and 3. A new approach for quantifying vulnerability
the metrics. For example, researchers with the PNL
Vulnerability Assessment program define vulnerability In this section we derive a generic vulnerability metric
as ‘‘the extent to which climate may change or harm a by translating a general definition of vulnerability, the
system’’ and measure it, as described above, as a susceptibility to damage, into a mathematical expres-
composite of 16 variables that represent a system’s sion. To do this we first define a threshold of damage
sensitivity to climate and ability to adapt to change and then measure susceptibility in terms of the system’s
(Moss et al., 2002). Although these 16 variables may sensitivity to and exposure to stressors. We then propose
represent critical factors in a wide range of regions, they a framework for estimating a system’s ability to modify
are not likely to capture the ‘‘change or harm’’ that may its vulnerable conditions by adapting and responding to
result from climate change equally well in different changing circumstances. To illustrate the form of the
regions with distinct cultural systems, values and proposed vulnerability metric we introduce a simple
biophysical systems. By focusing on developing metrics idealized human–environment system where some mea-
that represent the vulnerability of a place, the indicator sure of human well-being (W ) is a parabolic function of
approach is forced to link the variables of concern for a an independent variable (X ) (Fig. 1a). For simplicity, we
region to the measure of vulnerability and as a result the present a one-dimensional example, however, this
measure becomes difficult to apply in diverse settings. approach could be applied to a system of multiple
Vulnerability measures can only accurately relate to stressors and multiple outcomes variables of concern.
specific variables, rather than the generality of a place,
because even the simplest system is so complex that it is 3.1. Vulnerability
difficult to fully account for all of the variables,
processes and disturbances that characterize it. There- 3.1.1. Sensitivity and threshold
fore, we argue that vulnerability assessments should Defining the vulnerability of a system first requires
shift away from attempting to quantify the vulnerability understanding the sensitivity of the system to different
of a place and focus instead on assessing the vulner- stressors and identifying a threshold of human well-
ability of selected variables of concern and to specific being at which the system is said to be ‘‘damaged.’’ The
ARTICLE IN PRESS
258 A.L. Luers et al. / Global Environmental Change 13 (2003) 255–267
Probability
Distribution A
Distribution B
Wo Wo
Wo
vulnerability of a system to small changes in forcings is a where PX refers to the probability of the occurrence of
function of the system’s sensitivity to a given perturba- stressor X :
tion and the relative proximity of the system to its Although it is impossible to determine the precise
damage threshold (Fig. 1a): functional relationships that include all of the stressors
and variables of concern, analysis based on simple
V ¼ f fsensitivity=state relative to a thresholdg; theoretical models and multivariate regressions from
jqW =qX j empirical data can provide valuable information about
¼f ; ð1Þ
W =W0 critical relationships that can be applied in this measure.
This metric of vulnerability provides a means for
where W0 represents a threshold value of well-being distinguishing between systems with different average
below which the system is said to be damaged. In this states of being, sensitivities and exposures (Fig. 1b). For
example, the sensitivity is represented as the absolute example, suppose there are two farming communities
value of the derivative of well-being with respect to the where total crop production per capita is a measure of
stressor, however, other measures of sensitivity could be well-being. The two communities may each have similar
used, for example the coefficient of variations. well-being functions. However, one community (System
W) may have more crop land per capita than another
3.1.2. Exposure community (System Y) resulting in a higher average
Different communities and ecosystems are exposed to well-being (Fig. 1b). Based on the proposed vulner-
varying magnitudes and frequencies of disturbing forces, ability measure, if these two systems were exposed to the
often resulting in differential vulnerabilities (IPCC, same distribution of stressors (say distribution A), then
2001; Turner et al., 2003a, b). We capture these System Y would be more vulnerable, even though they
differences in exposure by calculating the expected value both had the same exposure and sensitivity. On the
of the ratio of sensitivity to the state relative to a other hand, if there were two separate systems with
threshold based on the frequency distribution of the identical well-being functions (both represented by
stressors of concern: System Y) but dramatically different distributions of
stressors (distributions A and B) the vulnerabilities of
V ¼ Expected Value these two systems would be distinct from each other
ðsensitivity=state relative to a thresholdÞ; (Fig. 1b). For example, consider two farmers who have
Z identical wheat production functions (with crop produc-
jqW =qX j tion as a measure of well-being) but live in two
V ¼ PX dX ; ð2Þ
W =W0 drastically different climates. The first farmer might live
ARTICLE IN PRESS
A.L. Luers et al. / Global Environmental Change 13 (2003) 255–267 259
in a region that experiences a very steady moderate capacity including available technology, the structure of
temperature range while the second lives in an area that institutions, human capital such as education, and
is characterized by high climatic variability with access to risk spreading processes (IPCC, 2001). These
frequent extreme events. In this case, the second farmer are all examples of factors that could contribute to a
would have the higher vulnerability despite the fact that system’s ability to decrease its vulnerability.
they had identical well-being functions.
3.1.4. Units
3.1.3. Adaptive capacity The units of the above metrics for vulnerability and
We define adaptive capacity as the extent to which a adaptive capacity are the units of well-being divided by
system can modify its circumstances to move to a less the units of the stressor (W =X ). For example, in the case
vulnerable condition (Fig. 1c). We quantify adaptive study below our measure of well-being is agricultural
capacity (A) as the difference in the vulnerability under yield (measured as tons/ha) and the stressor we consider
existing conditions and under the less vulnerable is temperature ( C) and the resulting units of our
condition to which the system could potentially shift: vulnerability measure is ((tons/ha)/ C). However, for
A ¼ V ðexisting conditionsÞ V ðmodified conditionsÞ: practical purposes, in our analysis we represent both
ð3Þ vulnerability and adaptive capacity as relative unitless
measures by normalizing each by a reference state. In
A system could decrease its vulnerability by modifica- general, the reference state that is used will depend on
tions that would lead to one or more of the following: the focus of the particular study. In the case study
(1) a shift in the well-being function that decreases the below, the reference state we use is the vulnerability of
sensitivity to critical stressors; (2) a change in the the average farm in the study region.
position relative to a threshold of damage and (3) a
modification in the system’s exposure to stressors of
concern. 3.1.5. Combining vulnerability and adaptive capacity
The capacity to adapt is distinct from adaptations We argue that vulnerability studies that include
that a system has made in the past to accommodate adaptive capacity directly in a vulnerability character-
disturbing forces. Prior adaptations are captured in the ization are actually characterizing what we refer to as
characterization of the well-being function, specifically the minimum potential vulnerability, which we distin-
in the sensitivity and the position relative to the guish from the existing vulnerability. We measure the
threshold of damage. For example, a farmer may have existing vulnerability under current and future condi-
adapted to drought over the years by shifting manage- tions and the minimum potential vulnerability (Vmin ) as
ment practices, such as using drip irrigation and taking the existing vulnerability minus adaptive capacity:
measures to increase soil quality for water retention. Vmin ¼ V A: ð4Þ
This adaptation may lead the farmer to be less sensitive
to drought. This same farmer, however, may also have This distinction between the minimum potential
the potential to shift to more drought resistant crops or vulnerability and the existing vulnerability is important
dig groundwater wells to further decrease her sensitivity both conceptually and practically. For example, con-
to drought over the long-run. We refer to this potential sider two farmers who are faced with drought and whose
as ‘‘adaptive capacity’’. Once the potential to adapt has conditions are identical except that one has insurance
been fully realized it becomes part of the system’s and the other does not. One farmer may have a greater
normal functioning and is manifested as a decrease in adaptive capacity and thus a lower potential vulner-
sensitivity or an increase in the state relative to the ability because of his access to insurance. However, if
threshold of damage and a corresponding decrease in the insurance program does not respond as promised in
the vulnerability. a crisis then both farmers are just as vulnerable. The
Many factors may determine a system’s ability to insurance program only provides the potential for
modify its vulnerable conditions, including the rate of lowering the farmer’s vulnerability. On the other hand,
change of the disturbing forces, and the social and consider two farmers who are identical except they have
natural capital of the system. For example, if the different soil types and are both faced with drought. It
average temperature increases gradually, farmers may may be that a farmer with one soil type that has a
be able to adapt by changing crops or management greater water holding capacity—and therefore require
practices that result in a shift in the well-being function, less water—is less vulnerable to drought conditions than
which would result in lower vulnerability. However, if the other farmer with a different soil type. In these
the temperature change was rapid the same farmers examples, the soil influences both farmer’s existing
might be limited in their abilities to adapt the changing vulnerability, while the insurance program as part of a
conditions. The Intergovernmental Panel on Climate farmer’s adaptive capacity only influences the potential
Change (IPCC) identified eight determinants of adaptive of lowering the farmer’s vulnerable conditions.
ARTICLE IN PRESS
260 A.L. Luers et al. / Global Environmental Change 13 (2003) 255–267
4. An example: assessing vulnerabilty in the Yaqui Valley that increasing temperatures resulting from global
warming may lead to decreased wheat yields (e.g. Lobell
4.1. Yaqui Valley et al., 2002). These concerns arose during a period when
a series of policy reforms were promulgated to increase
The Yaqui Valley, an intensively managed wheat- the efficiency of the agricultural sector by opening it to
based agricultural region, is located in Sonora, Mexico, the international market and by decreasing government
between the Sierra Madre Mountains and the Gulf of intervention in production and marketing decisions
California (Fig. 2). The Valley consists of approximately (Naylor et al., 2001). The effects of these various
225,000 ha of irrigated agricultural fields. Referred to as changes are not felt uniformly across the Valley but
the birthplace of the Green Revolution for wheat, it is depend in part on relative access to natural and social
one of the country’s most productive breadbaskets capital (Turner et al., 2003b; Lobell et al., 2002; Naylor
(Naylor et al., 2001). Using a combination of irrigation, et al., 2001). Here we apply the process-based approach
high fertilizer rates and modern cultivars (Matson et al., described above to begin to evaluate the relative
1998), Valley farmers produce some of the highest wheat vulnerability among Valley farmers faced with multiple
yields in the world (FAO, 1997). However, in a world of stresses. This case study is presented only as an example
globalized markets, reduced subsidies and price sup- application of the proposed methodology, and is not
ports, drought, and other forces, many Valley farmers intended to be a complete vulnerability analysis of the
and managers are concerned about sustaining yields and region.
maintaining household incomes.
The climate in the Yaqui Valley is semi-arid, with 4.2. Methods
variable precipitation rates averaging 317 mm yr1 and
an average daily temperature of 24 C. The natural 4.2.1. Defining the system
ecosystems of the Valley have co-evolved with, and are Our unit (or system) of analysis is the ‘‘farm unit’’—
thus likely to be resilient to, the natural climatic that is an agricultural field and the farmer or farmers
variability of the region. However, the agricultural responsible for the field. For practical purposes, we
system, which was developed by attempting to control define our agricultural field as a 30 m 30 m pixel as
the local environmental factors, such as water supply by described below. Of the many outcomes of concern to
building reservoirs, is likely to be vulnerable to climatic the Valley farmer, we focus on wheat yield as our
extremes. Prolonged droughts, such as the one that has measure of well-being. Wheat yield alone obviously does
persisted in the region since 1994, have lead to dramatic not fully capture the well-being of Valley farmers,
declines in total reservoir volume, increases in well however, we use it here to illustrate the proposed
pumping, and reduced water allocations to farmers. methodology. Wheat yield in any given farm unit is
Meanwhile, recent studies have pointed to the concerns affected by factors both internal and external to the
system. For example, regional temperature changes may hold management level constant and measure relative
affect plant growth, changes in national agricultural vulnerability. We, then hold soil type constant and
policies may influence farmer cropping decisions, and explore the relative effects of soil type in determining the
variation in farm unit technological resources may vulnerability within management levels. Each of these
influence planting and harvesting methods. We base methods is described in detail below.
our analysis on a previous study that suggests that over For each of the four years, we compute the distribu-
recent history wheat yields in the Valley are a function tion of yield within the entire Valley, and then rank
predominantly of temperature, soil type and manage- yields by percentile for each year. We then use a linear
ment (Lobell et al., 2002). We use the term management least-squares regression of yield with average night-time
loosely to refer to all factors other than temperature and temperature for January–April to define the average
soil that influence yield (Lobell et al., 2002). Lobell et al. yield and sensitivity for each percentile. To define the
(2002) showed that most of the variability in yield that vulnerability corresponding to each percentile, we run a
was not explained by soils or temperature was between- Monte Carlo simulation where temperature varies
field variability rather than within-field variability and according to a normal distribution with mean equal to
was therefore attributed primarily to management, 9.61 C and standard deviation equal to 0.99 C, as
where management included such factors as amount determined from 20 years of historical climate records.
and timing of fertilizer application, number and timing We then calculate the vulnerability according to Eq. (2)
of irrigations, tillage and cultivation practices and pest using a threshold value of 4 t/ha, which is the
control. approximate minimum yield required for farmer’s to
We focus our assessment on two external stressors— ‘‘break-even’’ (i.e. zero net profit) based on the average
climate (variability and change) and market fluctua- management practices (Matson et al. 1998). We normal-
tions—to calculate relative vulnerability of yields and to ize these vulnerability values by the average vulner-
begin to explore how biophysical and socio-economic ability calculated for the entire Valley. We generate a
conditions contribute to the variability of vulnerability map of vulnerability by matching the average yield
within the Valley. Specifically, we address the following percentile for each pixel to the associated vulnerability
three questions: (Fig. 3).
1. On which farm units are wheat yields most vulner- 4.2.3. Measuring adaptive capacity
able? As described above, the critical factors that may
2. To which stressors are wheat yields most vulnerable? influence a farm unit’s yield function include soil type
3. What factors explain differences in vulnerability of and management level. Management is the only one of
wheat yield between farm units in the Valley? these factors that farmers can potentially manipulate to
move to a less vulnerable condition. Therefore, in our
4.2.2. Measuring vulnerability analysis we estimate adaptive capacity from our time
To illustrate an application of the proposed metric, series of yields as the extent to which a farm unit has
we utilize remotely sensed estimates of yields in the exceeded its average management percentile over the
Yaqui Valley for four years: 1994, 2000, 2001, and 2002. study period. We assumed that the highest relative yield,
Yield estimates are derived from Landsat TM and as represented by the yield percentile, could be achieved
ETM+ data, as described in detail by Lobell et al. every year with the appropriate management. We
(2003). The four years span a range of climatic estimate the adaptive capacity as the difference between
conditions useful for assessing sensitivity of yields to the vulnerability calculated as above and the vulner-
climate. Yields in the Yaqui Valley are strongly ability calculated for a yield temperature function where
determined by average night-time minimum tempera- we assume the expected yield is equal to the maximum
tures during the growing season, which govern rates of yield percentiles observed over the four years. To create
plant respiration and development, where lower tem- a unitless measure we normalize this difference by the
peratures correlate with higher yields. Previous work has average value of the difference calculated for all pixels
shown that the sensitivity to temperature is greater for over the Valley:
fields with lower yields, suggesting a significant interac- ðVR mean VR max Þpixeli
tion between climate and management (Lobell et al., A¼ ; ð5Þ
ðVR mean VR max Þvalley ave
2002). In addition, the Lobell et al. (2002) analysis
suggests that while soil type and temperature variation where R refers to the relative yield percentile.
all represented significant sources of yield variability,
farm management, as estimated by yield variation 4.2.4. Explaining vulnerability and adaptive capacity
within soil type, was most important for determining To explore the factors contributing to the different
wheat yields. Based on these results, we pursue a two- vulnerabilities among farm units we compare our
tiered approach to our vulnerability analysis. First, we vulnerability map with other spatially explicit data,
ARTICLE IN PRESS
262 A.L. Luers et al. / Global Environmental Change 13 (2003) 255–267
Fig. 3. Vulnerability, Yaqui Valley irrigation district. The most vulnerable are shown in red and the least vulnerable in dark blue.
including soils type and management. To assess the vulnerable yields are those on poorly managed farm
relative effect of soils we calculate the vulnerability for units with stony-clay and compacted-clay soil.
what we refer to as different ‘‘management categories’’ The frequency distribution of vulnerability for the
within each soil type. We define three management Valley as a whole is skewed towards vulnerabilities
categories within each soil type as the top, middle and lower than the average (Fig. 5), indicating that while the
bottom thirds of the distribution of average yields over majority of farm units are relatively resilient in the face
the four years of data. We calculate the distribution of of variable climate a few farm units remain highly
vulnerability and adaptive capacity for the farm units vulnerable. However, estimates of relative vulnerability
within each of these three management categories. are not fixed. As changes occur inside and outside of the
system, so will the vulnerability. For example, if the
average minimum temperature increases by 1 C the
4.3. Results and discussion vulnerability of the average farmer would increase by
roughly 10%, resulting in an effective shift to the right in
4.3.1. Vulnerability the whole Valley distribution (Fig. 5). Meanwhile, a
Fig. 3 shows the spatial variability and patterns of 10% decrease in the effective price of wheat (either as a
vulnerability throughout the Valley. Overlaying these result of shifts in market prices or shift in subsidy
relative vulnerability estimates with spatially explicit policies) would result in a rise in the identified threshold
biophysical and institutional factors allows us to explore of damage in this model and result in approximately a
links between these factors and the spatial patterns of 30% increase in the vulnerability of the average farmer.
vulnerability. Our analysis indicates that both soil type These results illustrate how the proposed methodol-
and management practices appear to contribute to the ogy can provide a framework for assessing the relative
spatial variation in vulnerability. Not surprisingly, the importance of market fluctuations compared to tem-
‘‘best managed’’ soils are the least vulnerable, irrespec- perature changes in determining vulnerability. The
tive of the soil type (Fig. 4). However, the calculated differential effects of fluctuations in international
vulnerabilities suggest that soil type becomes more markets is particularly relevant to the region, which
important in the poorly managed lands. The most has begun a transition to an open economy after the
ARTICLE IN PRESS
A.L. Luers et al. / Global Environmental Change 13 (2003) 255–267 263
2
Deep Clay (V) Silt Loam (V) by shifting to higher value crops, or to a change in
Stony Clay (A) Comp Clay (A) temperature by shifting the timing and amount of
Deep Clay (A) Silt Loam (A) irrigation. The extent to which farmers are able to
respond to changing environmental and economic
conditions will depend on a range of social, political
1 and biophysical factors. Our analysis here does not set
out to explain these factors for the Yaqui Valley but
rather presents an example of a framework in which
these affects can be explored.
Our analysis indicates that average estimated adaptive
capacity varies only slightly between the three manage-
ment categories, yet distinct patterns appear between
0
soil types in the poor and best managed fields (Fig. 4). In
Poor Average Good
poor management areas, those farm units on the best
Mangement Category soil types (silt loam and deep clay) with the lowest
Fig. 4. Vulnerability (V) and adaptive capacity (A) by soil type and vulnerabilities exhibit above average-adaptive capacities
management type. The values shown are means of the distributions for while the farm units on the worst soil types (compacted
each category (soil type and management category). Po0:0001 from clay and stony clay) with highest vulnerabilities exhibit
Two-way ANOVAs with soil type and management category as fixed below-average adaptive capacities. These results suggest
effects and V and A as response variables.
that soil type may limit the ability of some poorly
managed farm units to adapt to stressors such as
0.5 increases in temperature. However, the best managed
existing conditions farm units on the best soils (e.g. silt loam) exhibit the
1 degree increase in mean temp
lowest adaptive capacities and those on the poorer soils
exhibit slightly above-average adaptive capacities. These
10% decrease in wheat prices
results suggest that the best managed farm units with
silt-loam soils have reached a maximum yield for the
Frequency
Temperature (C)
Temperature (C)
Temperature (C)
10 10 10
10 10 10
Yield (T/ha)
Yield (T/ha)
Yield (T/ha)
8 8 8
5 5 5
6 6 6
0 0 0
4 4 4
-5 -5 -5
2 2 2
0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
Year Year Year
Y Stony Yo (Threshold) Y Stony (Future) Yo (Threshold) Y Stony (Future+AC) Yo (Threshold)
Y Silt Temperature Y Silt (Future) Temperature Y Silt (Future+AC) Temperature
Y Stony-existing (V=1.6) 0.012 Y Stony (NoAC; V=1.9) 0.012 Y Stony (NoAC; V=1.9) 0.012
10 10 10
Y Silt-existing (V=0.56) 0.010 0.010 Y Silt (NoAC; V=0.6) 0.010
Probability
Probability
Probability
Yield (T/ha)
Yield (T/ha)
Y Silt (NoAC; V=0.6)
Yield (T/ha)
vulnerability and yield pattern in the silt-loam units. for vulnerability analysis. In this case study, we selected
When potential adaptations are included in the analysis, wheat yield as our outcome variable of concern. Our
however, the vulnerability of the stony-clay units analysis suggests, however, that yield alone might not be
decreases to 1.2 times that of the average farm unit sufficient to capture the vulnerability of Valley farmers.
under existing conditions, and the variability of yields For example, important coping strategies for farmers,
shifts to primarily above the threshold. However, the such as shifting production patterns in response to price
relative vulnerability of the silt-loam farm units remains shifts, are not easily captured with wheat yield. An
relatively unchanged in the future conditions even when alternative outcome variable to consider would be
the adaptive capacity has been realized. household income. Focusing on the vulnerability of
Overall, this study suggests that farm units with the household incomes to a set of multiple stressors
lowest minimum potential vulnerability (Vmin ) are those including temperature variability and change and
that are well managed on silt-loam soils and the farm market fluctuations would allow us to capture the
unit with the highest minimum potential vulnerability differential abilities of households to diversify, not only
are those that are poorly managed on stony-clay and within the agricultural sector but also across industries.
compacted-clay soils. However, this analysis suggests We will explore these issues in future work.
that management can overcome soil type constraints. As Examining the management and coping strategies of
a result, the factors that determine management level are farmers in the Yaqui Valley highlights the importance of
likely to have the most effect on vulnerabilities in this spatial scale in vulnerability assessments. In this case
region in the future. Our management variable incorpo- study, we chose the farm unit as our system of analysis
rates many different factors that need to be examined to for simplicity. The farm unit however is not isolated, it
identify the causes of vulnerability and constraints on affects and is affected by its surroundings. For example,
adaptive capacity. For example, can different manage- high fertilizer application rates for wheat production,
ment practices be attributed to differential access to which have increased from 80 to 250 kg N ha1 between
credit, information or other institutional factors? We are 1968 and 1995 and contributed to more than a doubling
currently conducting a survey that will address some of of wheat yields during the same period, have led to large
these questions. nitrogen losses in the ground and surface waters
(Matson et al., 1998; Panek et al., 2000; Harrison,
4.3.3. Defining the system 2003). Although the ecological consequences of the
Accurately defining the system, including identifying transfers of these and other wastes to downstream
the appropriate outcome variables of concern and estuaries have not been evaluated, they may pose a
setting the relevant spatial and temporal scale, is critical threat to important ecosystems on which coastal
ARTICLE IN PRESS
A.L. Luers et al. / Global Environmental Change 13 (2003) 255–267 265
communities depend for their livelihoods (Valdez et al., model the differential vulnerability of wheat yields to
1994). These sorts of externalities must be considered climate variability and change, and to market fluctua-
when defining the system of analysis, because, as tions. Our analysis suggests a skewed distribution of
suggested in this example, factors that may contribute vulnerability exists within the study region, with most
to a farmer’s resilience, such as fertilizer application, farmers exhibiting low vulnerabilities and a few farmers
may lead to a fisherman’s increased vulnerability. with high vulnerability. In addition, our method reveals
The dynamic nature of vulnerability illustrated in this that Valley farmers, without adaptations, are on average
case study highlights the importance of the temporal more vulnerable to a 10% decrease in wheat prices than
scale of analysis. Ultimately, for policy makers, the a 1 C increase in average minimum temperature. Soils
vulnerability implications of annual fluctuations in the and management both contribute to relative vulnerabil-
world wheat prices may not be as important as other ities in the region, however, it appears that the
factors such as resource scarcity. For example, in the constraints imposed by poor soil types can be overcome
semi-arid region of the Yaqui Valley, after 7 years of by improved management practices.
drought, water scarcity has become a serious threat to No single measure will be able to capture completely
the region (Osuna, 2003). Over the short-term, water the multiple dimensions of vulnerability. Ultimately,
limitation does not appear to be a major factor in the vulnerability research will require a set of metrics that
vulnerability of yields, in part because farmers have can help analyze and explain vulnerability character-
adapted to drought by extracting groundwater. Over the istics within and between systems. The most effective
long-term, however, if groundwater is extracted at a rate metrics will be those that are generic enough that they
above the regional recharge rate, this may lead to can be applied to a wide range of settings. There are
increased vulnerabilities as pumping costs increase. several reasons why we believe the proposed metric is an
Pumping groundwater may be increasing farmer’s example of the type of measures that are needed. Firstly,
resilience to drought today while contributing to the the form is sufficiently general to apply to simple one-
vulnerability of farmers in the future. Evaluating these dimensional systems or complex multi-dimensional
sorts of resilience and vulnerability tradeoffs between systems modeled in matrix form. Second, because the
different temporal and spatial scales must be a focus of measure is unitless it is easily comparable between and
vulnerability assessments in the future. within systems. Third, the metric can be used in a
modeling framework where the vulnerability implica-
tions of future environmental or political scenarios can
5. Conclusions be evaluated and the uncertainties can be incorporated.
Fourth, the proposed measure allows researchers to
This paper presents a methodology for quantifying analyze four essential aspects of vulnerability indepen-
vulnerability as the expected value of the sensitivity of dently—the state relative to a threshold of damage,
selected variables of concern to identified stressors sensitivity, exposure and adaptive capacity. Finally, the
divided by the state of the variables of concern relative proposed metric is not confined to a particular
to a threshold of damage. In addition, we present a conceptual framework but rather could be applied to
method for estimating the minimum potential vulner- test and compare the appropriateness of multiple
ability by accounting for a system’s ability to adapt and frameworks in different systems.
respond to changing circumstances.
The application of the proposed metric in a vulner-
ability assessment requires: (1) the identification of Acknowledgements
outcome variables of concern (e.g. income, yield, health,
ecosystem function); (2) the identification of stressors of We thank the members of the Research and Assess-
concerns (e.g. climate change, drought, market fluctua- ment Systems for Sustainability (SUST) Program for
tions); (3) a model of the relationship of the outcome many thoughtful discussions. We are particularly grate-
variables of concern to the stressors; and (4) base-line ful to the following from the SUST group for valuable
data from which the stressors-outcome model can be insights on the challenges and opportunities of measur-
calibrated. The final assessment will depend on the ing and modeling vulnerability: Colin Polsky, Billie Lee
strength of the data and model on which they are based. Turner II, Hans Joachim Schellnhuber, Lindsey Chris-
However, the proposed approach does not require tensen, William Clark, and Roger Kasperson. We also
detailed multi-variable predictive models of human– thank Rosamond Naylor, Karen Seto and Karen
environmental systems, and can be applied using simple Carney for helpful comments on the manuscript. This
regression models that incorporate the critical factors. work was supported in part by a grant from the
We illustrate the proposed methodology in a vulner- National Science Foundation (BCS-0004236) with con-
ability assessment of the Yaqui Valley, Mexico. Using a tributions from the National Oceanic and Atmospheric
combination of remote sensing and GIS techniques we Administration’s Office of Global Programs for the
ARTICLE IN PRESS
266 A.L. Luers et al. / Global Environmental Change 13 (2003) 255–267
Research and Assessment Systems for Sustainability Gunderson, L., 2000. Ecological resilience—in theory and application.
Program. Additional support was provided by John D. Annual Review of Ecology and Systematics 31, 425–439.
and Catherine T. Macarthur Foundation, and The Harrison, J.A., 2003. Nitrogen dynamics and greenhouse gas produc-
tion in Yaqui Valley surface drainage waters. Ph.D. Thesis,
David and Lucile Packard Foundation, the UPS Stanford University, Stanford.
Foundation, by an NSF Graduate research Fellowship Heitzmann, K., Canagarajah, R.S., Siegel, P.B., 2002. Guidelines for
to D. Lobell, and by graduate fellowships to A. Luers assessing the sources of risk and vulnerability. World Bank, Social
from the Switzer Family Foundation and The Teresa Protection Discussion Paper Series 0218.
Heinz Environmental Scholars Program. Holling, C.S., 1973. Resilience and stability of ecological systems.
Annual Review of Ecology and Systematics 4, 1–23.
IHDP (International Human Dimensions Program) Update,
2001. Special issue on vulnerability 2, 1–16. Available on line at
/http://www.ihdp.orgS.
References IPCC (Intergovernmental Panel on Climate Change), 2001.
Impacts, adaptation, and vulnerability climate change 2001,
Alwang, J., Siegel, P.B., J^rgensen, S.L., 2001. Vulnerability: a view Third Assessment Report of the IPCC, University Press,
from different disciplines. Social Protection Discussion Paper Cambridge, UK.
No. 0115, World Bank. Kaly, U., Pratt, C., Howorth, R., 2002. A framework for managing
Blaikie, P., Cannon, T., Wisner, B., 1994. At Risk, Natural Hazards, environmental vulnerability in Small Island Developing States.
People’s Vulnerability and Disasters. Routledge, London. Development Bulletin 58, 33–38.
Bohle, H.G., Downing, T.E., Watts, M., 1994. Climate change and Kasperson, J.X., Kasperson, R.E., 2001. International Workshop on
social vulnerability: toward a sociology and geography of food Vulnerability and Global Environmental Change. 17–19 May 2001,
insecurity. Global Environmental Change 4 (1), 37–48. Stockholm Environment Institute (SEI), Stockholm, Sweden. A
Briguglio, L., 1995. Source of this description: small island developing Workshop Summary prepared on behalf of workshop participants.
states and their economic vulnerabilities. World Development 23 SEI Risk and Vulnerability Programme Report 2001-01.
(9), 1615–1632. Stockholm.
Carpenter, S., Walker, B., Anderies, J.M., Abel, N., 2001. From Kasperson, J.X., Kasperson, R.E., Turner II, B.L., Schiller, A., Hsieh,
metaphor to measurement: resilience of what to what? Ecosystems W.-H., 2003. Vulnerability to global environmental change. In:
4, 765–781. Diekmann, A., Dietz, T., Jaeger, C., Rosa, E.S. (Eds.), The Human
Chambers, R., 1989. Vulnerability, coping and policy. Institute of
Dimensions of Global Environmental Change. MIT, Cambridge,
Developmental Studies Bulletin 20, 1–7.
forthcoming.
Clark, W.C., J.ager, J., Corell, R., Kasperson, R., McCarthy, J.J.,
Kates, R.W., Clark, W.C., Corell, R., Hall, J.M., Jaeger, C.C., Lowe,
Cash, D., Cohen, S.J., Desanker, P., Dickson, N.M., Epstein, P.,
I., McCarthy, J.J., Schellnhuber, H.J., Bolin, B., Dickson, N.M.,
Guston, D.H., Hall, J.M., Jaeger, C., Janetos, A., Leary, N., Levy,
Faucheux, S., Gallop!ın, G.C., Gruebler, A., Huntley, B., J.ager, J.,
M.A., Luers, A., MacCracken, M., Melillo, J., Moss, R., Nigg,
Jodha, N.S., Kasperson, R.E., Matson, P., Mooney, H., Moore III,
J.M., Parry, M.L., Parson, E.A., Ribot, J.C., Schellnhuber, H.-J.,
B., O’Riordan, T., Svedin, U., 2001. Sustainability science. Science
Seielstad, G.A., Shea, E., Vogel, C., Wilbanks, T.J., 2000.
292, 641–642.
Assessing vulnerability to global environmental risks. Report of
Kelly, P.M., Adger, W.N., 2000. Theory and practice in assessing
the Workshop on Vulnerability to Global Environmental Change:
vulnerability to climate change and facilitating adaptation.
Challenges for Research, Assessment and Decision Making,
Warrenton, VA. Research and Assessment Systems for Sustain- Climatic Change 47, 325–352.
ability Program Discussion Paper 2000-12. Environment and Lobell, D.B., Ortiz-Monasterio, J.I., Addams, C.L., Asner, G.P., 2002.
Natural Resources Program, Belfer Center for Science and Soil, climate and management impacts on regional wheat
International Affairs (BCSIA), Kennedy School of Government, productivity in Mexico from remote sensing. Agricultural and
Harvard University, Cambridge, MA. Forest Meteorology 114, 31–43.
Cutter, S.L., 1996. Vulnerability to environmental hazards. Progress in Lobell, D.B., Asner, G.P., Ortiz-Monasterio, J.I., Benning, T., 2003.
Human Geography 20, 529–539. Remote sensing of regional crop production in the Yaqui Valley,
Downing, T.E., Butterfield, R., Cohen, S., Huq, S., Moss, R., Mexico: estimates and uncertainties. Agriculture, Ecosystems, and
Rahman, A., Sokona, Y., Stephen, L., 2001. Climate Change Environment 94, 205–220.
Vulnerability: Linking Impacts and Adaptation. University of Lonergan, S., Gustavson, K., Carter, B., 2000. The index of human
Oxford, Oxford. insecurity. Aviso 6, 1–11. Available on line: /http://www.gechs.
Folke, C., Carpenter, S., Elmqvist, T., Gunderson, L., Holling, C., org/avisoS.
Walker, B., Bengtsson, J., Berkes, F., Colding, J., Danell, K., et al., Malkin, E., 2002. Mexican leader aims to offset tariff cuts. New York
2002. Resilience and sustainable development: building adaptive Times, November 19. Available online: http://query.nytimes.com/
capacity in a world of transformations. Scientific Background gst/abstract.html?res=F70914F63A540C7A8DDDA80994-
Paper Prepared for the Environmental Advisory Council to the DA404482.
Swedish Government. ICSU Series for Sustainable Develop- Mansuri, G., Healy, A., 2002. Vulnerability prediction in
ment, No. 3. Available online: http://www.resalliance.org/reports/ rural Pakistan. Paper Presented at IFPRI-World Bank
resilience and sustainable development.pdf. Conference on Risk and Vulnerability: Estimation and
Food and Agricultural Organization. 1997. FAOSTAT Rome. Policy Implications, Washington, DC, September 23–24. Avail-
Glewwe, P., Hall, G., 1998. Are some groups more vulnerable to able online: http://www.ifpri.org/events/conferences/2002/092302/
macroeconomic shocks than others? Hypothesis tests based on mansuri.pdf.
panel data from Peru. Journal of Development Economics 56, Matson, P.A., Naylor, R.L., Ortiz-Monasterio, I., 1998. Integration of
181–206. environmental, agronomic, and economic aspects of fertilizer
Grimm, V., Wissel, C., 1997. Babel, or the ecological stability management environment. Science 280, 112–115.
discussions: an inventory and analysis of terminology and a guide Mitchell, J., Devine, N., Jagger, K., 1989. A contextual model of
for avoiding confusion. Oecologia (Berlin) 109, 323–334. natural hazards. Geographical Review 79, 391–409.
ARTICLE IN PRESS
A.L. Luers et al. / Global Environmental Change 13 (2003) 255–267 267
Moss, R., Brenkert, A. Malone, E.L., 2000. Measuring vulnerability: a Magalhaes, A.R., Panagides, S. (Eds.), Climate Variability,
trial indicator set. Pacific Norwest Laboratories. Available online: Climate Change and Social Vulnerability in the Semi-Arid Tropics.
http://www.pnl.gov/globalchange/projects/vul/indicators.pdf. Cambridge University Press, Cambridge.
Moss, R.H., Malone, E.L., Brenkert, A.L., 2002. Vulnerability to Schellnhuber, H.J., Black, A., Cassel-Gintz, M., Kropp, J., Lammel,
climate change: a quantitative approach. Prepared for the US G., Lass, W., Lienenkamp, R., Loose, C., Ludeke, . M.K.B.,
Department of Energy. Available online: http://www.globalchange. Moldenhauer, O., et al., 1997. GAIA 6, 19–34.
umd.edu/cgi-bin/Details.pl?sref=PNNL-13765. Schimmelpfennig, D., Yohe, G., 1999. Vulnernability of crops to
Murdoch, J., 1994. Poverty and vulnerability. The American climate change: a practical method of indexing. In: Frisvold, G.,
Economic Review 84, 221–225. Kuhn, B. (Eds.), Global Environmental Change and Agriculture.
Naylor, R.L., Falcon, W.P., Puente-Gonzalez, A., 2001. Policy Edward Elgar Publishing, UK.
reforms and Mexican agriculture: views from the Yaqui Stephen, L., Downing, T.E., 2001. Getting the scale right: a
Valley. CIMMYT, International Maize and Wheat Improvement comparison of analytical methods for vulnerability assessment
Center, Economics Program Paper No. 01-01, CIMMYT, and household-level targeting. Disasters 25, 113–135.
Mexico, D.F. Turner, B.L., II, Kasperson, R.E., Matson, P.A., McCarthy, J.J.,
Osuna, P.M., 2003. Año seco ser!a el 2003, Tribuna, Mexico, April 14. Corell, R.W., Christensen L., Eckley, N., Kasperson, J.X., Luers,
Available online: www.tribuna.com.mx. A., Martello, M.L., Polsky, C., Pulsipher, A., Schiller, A., 2003a. A
Panek, J., Matson, P.A., Ortiz-Monasterio, I., Brooks, P., 2000. framework for vulnerability analysis in sustainability science.
Distinguishing nitrification and denitrification sources of N2O in a Proceedings of the National Academy of Sciences of the United
Mexican wheat system using 15N as a tracer. Ecological Applica- States of America 100 (14) (8 July), 8074–8079.
tions 10 (2), 506–514. Turner, B.L., II, Matson, P.A., McCarthy, J.J., Corell, R.W.,
Peterson, G.D., 2002. Estimating resilience across landscapes. Con- Christensen, L., Eckley, N., Hovelsrud-Broda, G., Kasperson,
servation Ecology 6(1), 17. Available online: http://www.consecol. J.X., Kasperson, R.E., Luers, A., Martello, M.L., Mathiesen, S.,
org/vol6/iss1/art17/. Naylor, R., Polsky, C., Pulsipher, A., Schiller, A., Selin, H., Tyler,
Petschel-Held, G., Block, A., Cassel-Gintz, M., Kropp, J., Ludeke, . N., 2003b. Illustrating the coupled human–environment system for
M.K.B., Moldenhauer, O., Reusswig, F., Schellnhuber, H.J., 1999. vulnerability analysis: three case studies. Proceedings of the
Syndromes of global change: a qualitative modelling approach to National Academy of Sciences of the United States of America
assist global environmental management. Environmental Modeling 100 (14) (8 July), 8080–8085.
and Assessment 4, 295–17314. USAID Famine Early Warning System Various publications. Avail-
Pritchett, L., Suryahadi, A., Sumarto, S., 2000. Quantifying vulner- able online: http://www.fews.org/fewspub.html.
ability to poverty: a proposed measure with application to Valdez, C., Tordesillas, M., Bechtel, B., Chavarria, E., Oriza, A.,
Indonesia. Social Monitoring and Early Response Unit Research Bravo, E., Pease, J., 1994. Evaluation and management require-
Institute (SMERU) Working Paper, May 2000. Available online: ments for the coastal wetlands in Southern Sonora, Mexico.
www.smeru.or.id. Instituto Techynologico y de Estudios Superiors de Monterrey,
Ribot, J.C., 1995. The causal structure of vulnerability: its application Guaymas, Mexico.
to climate impact analysis. GeoJournal 35, 119–122. Watts, M., Bohle, H.-G., 1993. The space of vulnerability. The causal
Ribot, J.C., 1996. Climate variability, climate change and vulner- structure of hunger and famine. Progress in Human Geography 17
ability: moving forward by looking back. In: Ribot, J.C., (1), 43–67.