Abstract

Abstract This paper offers a practically motivated method for evaluating systems' abilities to handle external stress. The method is designed to assess the potential contributions of various adaptation options to improving systems' coping capacities by focusing attention directly on the underlying determinants of adaptive capacity. The method should be sufficiently flexible to accommodate diverse applications whose contexts are location specific and path dependent without imposing the straightjacket constraints of a "one size fits all" cookbook approach. Nonetheless, the method should produce unitless indicators that can be employed to judge the relative vulnerabilities of diverse systems to multiple stresses and to their potential interactions. An artificial application is employed to describe the development of the method and to illustrate how it might be applied. Some empirical evidence is offered to underscore the significance of the determinants of adaptive capacity in determining vulnerability; these are the determinants upon which the method is constructed. The method is, finally, applied directly to expert judgments of six different adaptations that could reduce vulnerability in the Netherlands to increased flooding along the Rhine River. Key Words : adaptive capacity, climate change, flood risk, coping capacity 2 Adaptive capacity has worked its way, as an organizing concept, into the research structures of those who contemplate the potential harm that might be attributed to global climate change and other sources of external stress. As such, it holds the potential of being a point of departure for the construction of practical indices of vulnerability that could sustain comparable analyses of the relative vulnerabilities of different systems located across the globe and subject to a diverse set of stresses that lie beyond their control. This paper offers a practically motivated method for evaluating adaptive capacity by assessing the potential contributions of various adaptation options to improving systems' coping capacities. We expect that this method will be sufficiently flexible to accommodate diverse applications whose contexts are location specific and path dependent without imposing the straightjacket constraints of a "one size fits all" cookbook approach. It is designed to produce, nonetheless, unitless indicators that can be employed to judge the relative vulnerabilities of diverse systems to multiple stresses and to their potential interactions. It is designed, as well, to help practitioners distinguish productively between macro and micro scale factors that work to define the underlying determinants of coping capacity -a distinction that can bring critical scale differences into clear focus. We begin in Section 1 with a brief review of the literature in which the notions of vulnerability, exposure, sensitivity and adaptive capacity have been developed. We focus particular attention on the determinants of adaptive capacity and their role in defining the boundaries of coping ranges -thresholds of relatively benign experience beyond which systems feel significant effects from change and/or variability in their environments. Section 2 uses a more formal representation of vulnerability to show how overall coping capacity might be judged by contemplating the feasibility and efficacy of alternative adaptation options; and Section 3 offers some empirical justification drawn from national data for this approach. Section 4 presents the adaptation-specific foundation of proposed methodology; it builds on the determinants of adaptive capacity for each adaptation option and shows how the complications of adaptation interaction and multiple stresses can be accommodated. The method is designed ultimately to produce unitless and comparable indicators of coping capacity. Section 5 illustrates how the method might be applied by constructing a hypothetical example around historical flows in the Nile River before Section 6 offers a more realistic application to the Rhine Delta in the Netherlands. Concluding remarks in Section 7 finally provide some discussion of context for subsequent application in response to the call from the United Nations Framework Convention on Climate Change to identify the most vulnerable nations and communities. A Brief Literature Review. The authors of Chapter 18 of the Report of Working Group II to the Third Assessment Report (TAR) of the Intergovernmental Panel on Climate Change (IPCC) focused considerable attention on the role of adaptation in judging the economic implications of climate change and climate variability [IPCC (2001)]. Even though they concentrated on climate-related stresses, their work can be applied by those who contemplate the potential for adaptation to diminish the costs (or to enlarge the benefits) of change in the mean or variability in any variable that defines a system's environment. Moreover, the care with which they assessed the broader adaptation literature in their approach to climate issues means that it is sufficient for present purposes simply to review their work with a careful eye toward converting their insights into workable and informative indicators of coping capacity. We begin the process by recalling four major conclusions that the lead authors of Chapter 18 highlighted in their Executive Summary (IPCC, 2001): 3 1. The vulnerability of any system to an external stress (or collection of stresses) is a function of exposure, sensitivity, and adaptive capacity. 2. Human and natural systems tend to adapt autonomously to gradual change and to change in variability. 3. Human systems can also plan and implement adaptation strategies in an effort to reduce potential vulnerability or exploit emerging opportunities even further. 4. The economic cost of vulnerability to an external stress is the sum of the incremental cost of adaptation plus any residual damages that cannot be avoided. Moreover, the authors of Chapter 18 emphasized that adaptive capacity varies significantly from system to system, sector to sector and region to region. Indeed, the determinants of adaptive capacity include a variety of system, sector, and location specific characteristics: 1. The range of available technological options for adaptation, 2. The availability of resources and their distribution across the population, 3. The structure of critical institutions, the derivative allocation of decision-making authority, and the decision criteria that would be employed, 4. The stock of human capital including education and personal security, 5. The stock of social capital including the definition of property rights, 6. The system's access to risk spreading processes, 7. The ability of decision-makers to manage information, the processes by which these decision-makers determine which information is credible, and the credibility of the decision-makers, themselves, and 8. The public's perceived attribution of the source of stress and the significance of exposure to its local manifestations. Finally, it is essential to note that exposure to variability and to extreme events is an important source of vulnerability. In fact, systems typically respond to variability and extreme events before they respond to gradual changes in the mean. In summary, the vulnerability cum adaptation literature recognizes explicitly that systems' environments are inherently variable from day to day, month to month, year to year, decade to decade, and so on [see For most systems, though, change and variability over short periods of time fall within a "coping range" -a range of circumstances within which, by virtue of the underlying resilience of the system, significant consequences are not observed [see Downing, et al (1997) or Pittock and Jones The broad relationship between vulnerability, sensitivity, and adaptive capacity illustrated in the river flow example can be expressed in its most general form with a little notation. Let V denote vulnerability, E denote exposure, S denote sensitivity, and A denote adaptive capacity, and let D i index each of the eight determinants of adaptive capacity identified in Section 1. 1 Each of these variables, except adaptive capacity, is recorded in bold because each can be a vector. Vulnerability can be measured simultaneously along many different dimensions. A system can be exposed to many different stresses simultaneously. Many different sectors and many different people can feel sensitivity to any particular exposure at the same time. All of the determinants of adaptive capacity can, in principle, be depicted in multiple ways. Adaptive capacity is uniquely represented as a scalar, because we think that a scalar can be defined to serve as an aggregate measure of the potential to cope. Formally, then, Both F{--} and AC{--} are multivariate and complex functions that are location specific and path dependent. It is reasonable to expect that vulnerability would, at least eventually, increase monotonically with exposure and sensitivity at increasing rates. It is also likely that exposure and sensitivity would fall at decreasing rates with higher adaptive capacity. The relationship between adaptive capacity and its determinants reflected in equation Functional representations of this sort are, of course, only useful if they offer ways of organizing thoughts and sifting through the complexity. Many of the variables cannot be quantified, and many of the component functions can only be qualitatively described. Still, working through their content from the bottom up can be useful in uncovering practical insights that can support the creation of indexes of coping capacity. Some of the determinants of adaptive capacity should, for example, operate on macro-scales in which national or regional factors play the most significant role, but other determinants should function on more micro scales that are precisely location specific. This will turn out to be a useful distinction. The set of available, applicable, and appropriate technological options (Determinant 1) for a given exposure at a particular location should, for example, be defined on a micro-scale, even though the complete set of possible remedies might have macro roots. Flood control options would, for example, be determined by the local conditions of the river bed and available engineering knowledge; and this knowledge may be restricted to indigenous knowledge on the one hand or informed by worldwide consultants on the other. Determinants 2 through 6 should all have large macro components to them, but their micro-scale manifestations could vary from location to location or even from adaptation option to adaptation option. Resources (Determinant 2) could be distributed differently across specific locations, but adaptive capacity may be more sensitive to larger scale distributional issues across different locations. The essential questions here focus on whether sufficient funds are available to pay for adaptation and whether the people who control those funds are prepared to spend them on adaptation. Empirical results reported in the next section are evidence that these questions can be critical at the most fundamental level. We will show that poorer people are more likely to fall victim to natural catastrophes than are richer people. We will also report that more densely populated areas are more vulnerable. This is as expected, because the same disaster affects more people. Moreover, we find a positive relationship between income inequality and vulnerability; i.e., people in more egalitarian societies seem to be less likely to fall victim of natural 5 disasters than are people in a society with a highly skewed income distribution. This is expected, as well. It is consistent with the negative correlation between income and vulnerability, and it suggests that measures designed to highlight a skewed distribution would confirm the notion that the poorest communities within a country would face similar resource deficiencies when it comes to protecting themselves. Some other explanatory variables are insignificant, but it is important to note that health care and education have strong positive correlations with per capita income. Macro-scale and even international institutions (Determinant 3) could certainly matter even at a micro level, especially in determining how decisions among various adaptation options might be made and who has access to the decision-making process. For example, the World Bank follows certain procedures in its investment decisions, and adaptation projects in countries seeking World Bank support must satisfy Bank criteria before even being considered. The European Union also has a framework (on procedures as well as consequences) into which all water management projects must fit, so macro-scale influences can be felt even in developed countries. On the other side of the coin, though, adaptation projects in other places can be decided and implemented completely according to local custom alone. The stock of human capital (Determinant 4) could be a local characteristic, as well, but its local manifestation would likely be driven in large measure by macro-scale forces such as national support of local education. The stock of social capital (Determinant 5) and efficacy of risk-spreading processes (Determinant 6) should be largely functions of macro-scale structures and rules; but they could again take different forms from location to location and option to option. Property rights may be well defined through national institutions, and they may be the basis of private insurance markets; but issues of moral hazard and adverse selection may or may not be particularly severe in one location or another. Risk can be spread through national markets for commercial insurance and the international reinsurance markets, but some companies may refuse to sell flood insurance. Risk can also be spread through mutual obligations in the extended family, the strength of which varies between cultures and city and countryside. By way of contrast, Determinants 7 (managing information) and 8 (attributing signals of change to their sources) may have some general macro-scale foundations, but their primary import would be felt on a micro-scale. Indeed, decision-rules and public perceptions could take on forms that would be quite particular to the set of available options. Taken in its most general form, the vulnerability model reflected in equations (1) and (2) can lead to the conclusion that everything is connected to everything else, but this is not a productive insight. A more practical approach would, instead, read carefully through the determinants of adaptive capacity to recognize that the local manifestations of macro-scale determinants of adaptive capacity are their most critical characteristics. Taking this more modest observation to heart, we will show that it is possible to build coping indicators directly from systematic evaluations of the feasibility of available adaptations, taken one at a time, and their relative efficacy in reducing either sensitivity or exposure. Some Empirical Justification for the Roles of Selected Factors in Determining Vulnerability. The determinants of adaptive capacity are based on a synthesis of many case studies in the literature on natural hazards and other risks. The determinants are hypothesized to be important, with a lot of anecdotes to back that claim, but the relative strength of the various determinants is unclear. In this section, we look at national data of vulnerability to natural hazards and seek to explain the observed lack of adaptive capacity with selected proxies for adaptive capacity. We argue above that adaptive capacity is a local characteristic; unfortunately, data availability does not allow us to look at a higher resolution than country level. We do not measure vulnerability to natural hazards directly, but rather look at natural disasters and their impacts. The determinants of adaptive capacity can also only be measured by proxy, particularly notions such as social and human capital. These caveats need to be kept in mind when interpreting the results below. 6 The Centre for Research on the Epidemiology of Disasters (CRED), Catholic University of Louvain, collects information on the consequences of natural and man-made disasters. Their databases are made available through the US Office of Foreign Disaster Assistance (OFDA) at http://www.cred.be/emdat/. The database contains information of about 12,000 disasters, covering the entire world, a range of disasters and the period of 1990-2000. Indicators included are the time and place of disaster, its type, its strength, and the damage done, measured in the number of people killed, injured, made homeless or otherwise affected and the economic damage done. Sources include (re)insurance companies, development and disaster aid agencies (non-governmental, national or multilateral), and the press. Based on these data, we calculated three vulnerability indices. First, we computed the number of people killed by natural disasters in the period 1990-2000 in each country, and normalized this with the size of the population in 1995, the middle of the decade. Second, we calculated the number of people affected (but not killed), normalized with population. Third, we computed material damage, measured in US dollars, normalized with Gross Domestic Product (GDP). Population data were obtained from the World Bank Group Economic Growth Research (http://www.worldbank.org/research/growth). From the same source, we obtained information on GDP, income per capita, enrollment in education, life expectancy, and land area. We obtained indicators of political rights and civil liberties from Freedom House (http://www.freedomhouse.org). The Gini coefficient, measuring the distribution of income in a country, was obtained from the UNDP World Income Inequality Database from the United Nations University World Institute for Development Economics Research (http://wider.unu.edu/wiid/wiid.htm). All data are for 1995. Setting aside data problems, we obtain the following results. where R is the risk, I is an indicator of adaptive capacity (indexed by i), c is the index for countries, α and β are parameters and ε is noise. The multiplicative form of the superindicator of adaptive capacity corresponds to the notion in the main text that the weaker links in the adaptive chain matter most. Equation The noise term η is assumed to be Normally distributed. An additional complication is that, for a number of countries in the sample, we observe a zero risk. We interpret equation For people killed, we find a negative, almost significant relationship between per capita income and risk. That is, poorer people are more likely to die of natural violence than are richer people. This is as we would expect. The fact that deaths are probably underreported in less developed countries can only make this relationship stronger. Other explanatory variables are insignificant, and the explanatory power of the regression is low. This may be partly due to the noisy link between a disaster occurring and people dying as a result. We obtain more interesting results for the fraction of people affected by natural disasters. The data are more reliable, and the link between natural disaster, vulnerability, and observed effect is more straightforward. We again find that poorer people are more likely to fall victim of natural violence than are richer people. The relationship is highly significant. For every percent economic growth, vulnerability falls by a percent. We also find that more densely populated areas are more vulnerable. This is as expected, because the same disaster affects more people. We also find a positive relationship between income inequality measured by a Gini coefficient and vulnerability. As expected, people in more egalitarian societies are less likely to fall victim of natural violence than are people in a society with a highly skewed income distribution. This result is consistent with the reported correlation with per capita income on a national level. 2 Alternative income inequality measures like top to bottom ratios or percentages of populations below certain thresholds could easily underscore this observation and bring into focus the notion that the poorest citizens of any country would find it difficult to devote sufficient resources to protect themselves from natural hazards. Other explanatory variables are insignificant, but we should be aware that health care and education have a strong positive correlation with per capita income. The explanatory power of this regression is reasonable. Building an Indicator for Coping Capacity from the Determinants of Adaptive Capacity. The construction of an index of the potential contribution of any adaptation option (to be denoted by j) to an indicator of overall coping capacity (denoted by PCC j ) will begin with a step by step evaluation of feasibility factors -index numbers that are judged to reflect its strength or weakness vis a vis the last seven determinants of adaptive capacity. These factors will be subjective values assigned from a range bounded on the low side by 0 and on the high side by 5 according to systematic consideration of the degree to which each determinant would help or impede its adoption. Let these factors be denoted by ff j (k) for determinants k = 2, …, 8. We will argue that an overall feasibility factor for adaptation (j) should be reflected by the minimum feasibility factor assigned to any of these determinants; i.e., FF j ≡ min{ff j (2), … ff j (8)}. Each factor inserted into equation The ability of adaptation option (j) to, in fact, influence a system's exposure or sensitivity to an external stress can meanwhile be reflected in an efficacy factor EF j -a subjective index number assigned from a range running from 0 to 1. Efficacy factors will thereby reflect the likelihood that adaptation (j) will perform as expected to influence exposure and/or sensitivity compounded by the likelihood that actual experience would exceed critical thresholds if it were adopted. The potential contribution of any adaptation to a system's social and economic coping capacity can then, finally, be defined as the simple product of its overall feasibility factor and its efficacy factor; i.e., PCC j ≡ {EF j }{FF j }. Any method that focuses attention on the feasibility and efficacy of adaptation options through equations To be operational, though, a feasibility-based evaluation of an adaptation's potential contribution to overall coping capacity must be informed by a compounding evaluation of feasibility and efficacy for each possible adaptation. The method described in equations Suppose, for the sake of argument, that m potential adaptation options have been identified for a specific vulnerability and that they have been assigned potential coping capacity indicators {PCC 1 , …, PCC m }. The spirit of equations A high R ratio would indicate that several options of relatively comparable potential were available, but a low ratio would indicate a lack of diversity that could prove harmful if the assessment of adaptation were flawed. An Illustrative Application to a Hypothetical Example. The various panels of We can now begin to apply the indicator methodology of Section 4 to this illustrative example by exploring how three different adaptation options might alter either the flow of the Nile or the indicated thresholds of significant impact: Option A: Construction of a series of protection levies. Option B: Building a dam upstream. Option C: Periodically dredging the river. The point of this section is to show how the method described in Section 4 might be applied by demonstrating how feasibility factors and coping capacity indicators might be assigned in this river flow example. Assume, to that end but only for the time being, that any of the options listed above would work as displayed in Differences in magnitude and distribution of cost across the population and over time would surely be critical in contemplating the potentially limiting role of resource availability (Determinant #2) for options A through C. If the threats appeared when modest resources were available to the host economic/political/social systems, for example, assessors might find levies or dredging more attractive than building an upstream dam, and we could assign feasibility factors of 3 or 4 to those options and 0 or 1 to the dam. If resources were limited but relatively certain over the foreseeable future, however, we might lower the factor assigned to levies to 0 or 1 but hold the dredging option at 3 or 4. And if the systems had access to significant resources from home or abroad, then we might give values of 4 or 5 to all three options. In any case, these feasibility factors would not represent the likelihood that any adaptation option might be implemented; they would simply reflect an assessment about whether or not requisite financing could be found. Resource distribution would also play a role, here. The general population might be unwilling divert limited resources from efforts designed to address other sources of stress to dredging the river, but one of its superrich or a political leader could decide that an upstream dam would a suitable monument to his or her stature in the community. It might even secure his or her place in history. Less prestigious adaptation options such as levies and dredging might then be assigned feasibility factors equal to 0 or 1 while an upstream dam could score 2 or 3 or perhaps higher. The key here would be to decide if there were other potential projects around that would be equally noticeable so that the investment decision could actually turn on personal stature and not necessarily social usefulness. Determinant #3 speaks to institutional requirements for each adaptation option and the decisionmaking frameworks with which they will be evaluated. Both need to be considered, and both can be evaluated independently of resource availability. Human capital and social capital are the broad categories identified in Determinants #4 and #5. Both could play a role in this example in many ways by working through the definition of property rights and effective access to the decision-making process. We have already noted that constructing levies or the dam would require the taking of land that would either be used to support the levies or be inundated by the upstream lake created by the dam. The definition of property rights would, therefore, implicitly identify who would be hurt if either option were implemented. Levies could, though, benefit riverside propertyowners. If that were perceived to be the case, then a high feasibility factor of 4 or 5 could be assigned to indicate the high likelihood that the people who would be harmed by their construction could be convinced that they would see a compensatory benefit. The people harmed by the construction of the dam would, by way of contrast, not necessarily benefit from downstream flood management; nor would they necessarily be compensated fully by associated recreational or energy benefits. If their access to the decision-making process gave them sufficient power to block the construction of the dam and if institutions did not exist that could compensate them fully for their losses, then this access could be reason to assign a low feasibility factor for the dam. Determinant #6 speaks to systems' sensitivities to various exposures because it focuses attention on their ability to spread or reduce the risk. Suppose, in our example, that the broader community were "plugged into" global markets so that it could compensate easily for agricultural shortfalls caused by interruptions of irrigation practices. The macro-scale system would then be able to spread the risk of these shortfalls, and the feasibility factor assigned to any adaptation whose implementation would reduce vulnerability to low river flow, the dam in this case, should the decline irrespective of the decision rule applied in Determinant #2. By the same token, if sensitivity to flooding were fundamentally financial, then private or social insurance programs could spread risk, reduce the damage, and thereby diminish the feasibility of any of the specific options listed above. Notice, however, that both of these stories rely on structures and institutions that have not yet been introduced; and so the ability of these structures and institutions to accommodate additional stress needs to be evaluated. Indeed, the best approach in any case where access to risk spreading mechanisms might lead to assigning low feasibility factors to any specific adaptation would subject exploiting or enhancing these mechanisms to the very same feasibility assessment methodology. Determinants #7 and #8 highlight informational needs, perceptions, and decision-making credibility. The significance of each depends upon decision-making structures drawn from macro sources that can have parallel effects on all three options. Perceptions about the sources of the vulnerability to flooding could have a significant effect on the derivative perception that any or all of the options would work. Perceptions about the significance of the vulnerability, independent of its source, could also have comparable and consistent effects on the likelihood that any option would be contemplated. Low confidence in attribution or low opinion of significance would make all of the options relatively less feasible because none of them would be subjected to serious evaluation; low feasibility factors should then be assigned. High confidence in attribution and widespread recognition of significant exposure would, of course, have the opposite effect. Some options, like the dredging option in this example, could add micro-scale dimensions to feasibility considerations in this category, as well. Since dredging would be repeated as needed as the future unfolds, its feasibility would additionally depend in part on the ability of river managers to collect information and to process it properly so that renewed dredging could be implemented in a timely fashion. Absent this ability, a low factor would be assigned to dredging. Determinant #6 identified access to risk-spreading mechanisms, and so it highlighted the need to describe risk appropriately. The previous discussion highlighted the possibility that spreading risk might diminish a system's sensitivity to a given exposure, but there is another side of the risk equation that also 12 needs to be evaluated. Will the adaptation options evaluated for their feasibility in terms of Determinants #2 through #8 actually work to reduce exposure or diminish sensitivity? This question leads directly to assessments of the second factor in equation (5) -assigning efficacy factors for each option.