By Ursula Kaly, Lino Briguglio, Helena McLeod, Susanne Schmall, Craig Pratt and Reginald Pal

Small Island Developing States (SIDS) face disadvantages to their development associated with an interplay of factors such as remoteness, geographical dispersion, vulnerability to natural disasters, a high degree of economic openness, small internal markets, and limited natural resources. These issues have been recognised and increasingly highlighted in international fora during the last decade.

Vulnerability indices have been developed for the risks associated with economic conditions, climate change and sealevel rise, the El Nino-Southern Oscillation phenomenon, anthropogenic impacts and natural disasters. Most of these indices describe the vulnerability of human systems (economies, development, social systems) to these risks. Only a few studies have attempted to measure the vulnerability of the environment itself to anthropogenic and natural hazards.

Human systems and the environment are dependent on one another. Risks to the environment of a state will eventually translate into risks to humans because of their dependence on the natural environment for resources. In turn, the environment is susceptible to both natural events and appropriate management by humans. This means that overall vulnerability of a state should include measures of both human and natural systems and the risks which affect them. Such a 'Composite Vulnerability Index' (CVI) was proposed to the UN by the Maltese Ambassador in 1990. The development of an Environmental Vulnerability Index (EVI) will be an important step towards development of the CVI.

Definitions

Vulnerability is the potential for attributes of a system to respond adversely to the occurrence of hazardous events and Resilience is the potential for attributes of a system to minimise or absorb the impact of extreme events. The aim of vulnerability indices is to describe the relative vulnerability of states. Different forms of vulnerability of states are described. For example, economic vulnerability is concerned with external forces which act on the economy, while social vulnerability occurs when natural or other disasters force massive upheavals of residence, traditions and society.

The focus of this study is on vulnerability of the environment itself to both human and natural hazards. This includes effects on the physical and biological aspects of ecosystems, diversity, populations or organisms, communities and species. Unlike previously-developed vulnerability indices, the focus is not on risks to humans or their property.

Environmental vulnerability differs from vulnerability of human systems because the environment is complex with different levels of organisation from species to interdependent ecosystems and the complex linkages between them. Because data are often not available and indicators for health and vulnerability of the environment have to be physically measured, indicators may be heterogeneous in nature and not expressible in common units. This means that developing an index for the environment will need a new approach.

The need for an EVI

The need for an environmental vulnerability index was recognised at the Global Summit on Small Island States held in Barbados in 1994 where the United Nations formally expressed the desire in paragraphs 113 and 114 of the Barbados Plan of Action for the development of a such an index. The benefits of producing an EVI are that it can attract attention to certain states which are considered 'more vulnerable' and it summarises vulnerability based on meaningful criteria which can be considered by donors when allocating financial aid and projects.

The overall aim of this study was to begin the development of an environmental vulnerability index consistent with the Barbados Plan of Action and needs of the Alliance of Small Island States (AOSIS). The EVI developed here could then be combined with an economic vulnerability index to give a composite index which in a single figure format incorporates the environmental and economic concerns of a state. It is envisaged that the EVI and CVI would be recalculated every 5 years to examine changes through time as well as relative rankings of countries.

The specific aims were to:
- Review current work already completed or underway addressing environmental vulnerability of SIDS;
- Build on past work on environmental vulnerability, if appropriate, or approach the problem from a new perspective where other attempts have had limited success;
- Identify variables which may be used in the construction of an environmentally descriptive vulnerability index for Pacific SIDS;
- Develop a logical framework and methods of calculating and index for environmental vulnerability;
- Identify and collect data which would be used to calculate the environmental index;
- Identify gaps in the available data;
- Identify future directions for the further development of an internationally acceptable environmental vulnerability index;
- Compile a report for widespread circulation and consideration prior to the Donor Round Table leading up to CSD-7 and the UN General Assembly special session on SIDS in September 1999; and
- Further efforts towards the development of a composite vulnerability index as described in the Barbados Program of Action and meet the needs enunciated by AOSIS.

Mini review of previous work on vulnerability indices

Fifteen studies were reviewed which examine the relative vulnerabilities of states in terms of risks to human and natural systems. Most of the studies were concerned with risks to human economic and social systems (13), while only 5 attempted to describe effects on the natural environment. The risks of concern also varied among studies. Anthropogenic risks were considered in 11 studies; 6 studies examined climate change and sealevel rise; and 6 studies considered natural disasters as part of their risk. Only 1 study specifically examined the effects of both humans and natural hazards on the environment. It is the object of this study to fill this gap by developing an environmental index based on a wide array of environmental indicators which includes both natural and anthropogenic risks.

Features of past vulnerability indices

1. There has been some confusion with terminology. In one study, an Ecological Vulnerability Index was developed which actually looked at the vulnerability of human systems to natural disasters and inherent geographic characteristics, rather than vulnerability of the environment. We propose that vulnerability indices should be named by their responders, not risks. Human vulnerability indices include Economic, Quality of Life and Human Development Indices. This is the first study to attempt a true Environmental Index.

2. The logical framework for past indices has tended to be lacking. The successful development of an EVI requires a logical framework to ensure that the index is not just driven by data availability, terms are fully defined, appropriate indicators are found and the model can be tested.

3. The number of indicators varied among studies. Studies with only a few indicators (3-6) tended to focus on human systems. When the number of indicators used was moderate (12-15) more emphasis was placed on natural disasters and ecological variables. Only one study used a large number of indicators (60) and it was the only one which produced a list of indicators for anthropogenic pressures on the environment. The lesson here is that more indicators are required when complex ecological systems become the focus of the index being constructed.

4. Five different methods of evaluating or scoring the indicators were identified from the studies reviewed. Some of these were considered for the present study.

5. Five different methods were identified for aggregating the value of the indicators into an index or sub-indices. None of these was considered appropriate for the EVI. A modification of two methods identified from past indices was used.

Theoretical framework for the EVI

The maintenance of ecosystem or ecological integrity is at the heart of the development of a vulnerability index for the environment, because it is ecosystem integrity that is threatened by natural and anthropogenic hazards. The notion of ecosystem integrity is so complex that it cannot be expressed through a single indicator, but rather requires a set of indicators at different spatial, temporal and hierarchical levels of the ecosystem. Ecosystem integrity depends on biodiversity, ecosystem functioning and resilience, all of which are such interrelated variables, that factors which affect just one of these can have far-reaching ecosystem-wide consequences.

The risks to the environment are any events or processes that can cause damage to ecosystem integrity. These include natural and human events and processes such as 'the weather' and 'pollution'. Some researchers have identified natural hazards as those in which natural environmental conditions depart from 'normal' to such an extent that systems of interest (human, environmental) may be adversely affected. The problem with this definition is that unless we identify certain natural events as being anthropogenically altered (eg. anthropogenically-accelerated sea-level rise), all events are 'normal'. The implication from this line of reasoning is that the changes we see to the natural world as a result of natural hazards are deemed 'unacceptable' from a human perspective. This means that except in the case of anthropogenic risks, in an assessment of environmental vulnerability, what we really are examining is unacceptable departures from our (human) view of how the environment should change. For the purposes of this study, we will accept that risk events should include those which cause sudden and seemingly-negative impacts on natural systems as a way to evaluate vulnerability.

Although most identifiable risk events are capable of causing damage, it is only the larger and more intense events that are likely to cause wholesale changes in the environment, at least in the short to mid term. Some of the more important risks which can impact on the environment include meteorological events (e.g. cyclones, droughts, heatwaves), geological events (earthquakes, tsunamis, volcanos), anthropogenic impacts (mining, habitat destruction, pollution), climate change and sealevel rise.

The entities at risk, termed the 'responders' include ecosystems, habitats, populations and communities of organisms, physical and biological processes (e.g. beach building, reproduction), energy flows, diversity, ecological resilience and ecological redundancy.

Three aspects of environmental vulnerability were identified which would need to be incorporated into an EVI. These are:

1. The level of risks (or pressures) which act on the environment within the state, forming the Risk Exposure sub-Index (REI) which examines the frequency and where possible, the intensity of risk events which may affect the environment. These are based on levels observed over the past 5-10 years for most risks, but may include data for much longer periods for geological events. The REI is a measure of potential risk only: There is no logical expectation that patterns of risk expression during the immediate history of a state will necessarily result in similar risk levels today or in the future;

2. Intrinsic vulnerability or resilience of the environment to risks, forming the Intrinsic Resilience sub-Index (IRI) which refers to characteristics of a country which would tend to make it less/more able to cope with natural and anthropogenic hazards; and

3. Extrinsic vulnerability or resilience as a result of external forces acting on the environment, forming the Environmental Degradation sub-Index (EDI) which describes the ecological integrity or level of degradation of ecosystems. The more degraded the ecosystems of a country (as a result of past natural and anthropogenic hazards), the more vulnerable it is likely to be to future risks.

Features of the EVI

In developing the EVI, we set criteria on certain features of the index to ensure that it would be able to perform the tasks for which it was developed. The criteria were that the EVI should be intuitively understandable (set within a range from which highly vulnerable states have immediate recognition as such), impartial, suitable for international comparisons and able to differentiate among countries, applicable at different spatial scales (regional, country, island), refinable, presented in breakdown and single figure formats and easy to calculate using a user-friendly computer interface.

Methodology

Because the risks are many and ecosystem resilience and integrity are complex in character, it was necessary to use indicators to characterise them. This means that not all aspects are covered, but that a subset of variables is selected which describes frequency and intensity of risks, intrinsic vulnerability, effects on ecosystems, groups of organisms, physical features of the environment and mitigators of effects.

For the purposes of the EVI the following definitions relating to indicators and indices were used:
- An indicator was defined as any variable which characterises the level of risk, resilience or environmental degradation in a state;
- A sub-index (the REI, IRI and EDI) was defined as an aggregated average of the scores for indicators which related separately to risk, resilience or degradation; and
- An index (the EVI) was defined as an aggregated average of each of the three sub-indices (REI, IRI, EDI) to give an overall measure of the environmental vulnerability of a state. The EVI is then, a composite of each of the three sub-indices.

The criteria for the selection of indicators was that they should be applicable over the entire scale of interest (countries, regions), spread over different geographic, habitat and climatic types, relatively easy to understand, well defined, have data available now or with assistance in the future and be as uncorrelated as possible (to limit redundancy).

A total of 57 indicators of environmental vulnerability were finally selected for inclusion in the index. This included 39 indicators of risk (REI), 5 indicators of resilience (IRI) and 13 indicators of environmental integrity or degradation (EDI) (Annex 2). Many of the indicators were expressed as a fraction of area of land or coast rather than simply absolute numbers because it is risk density or proportion of area degraded that is of interest from an environmental perspective.

Although a larger number of indicators would have been preferred to obtain a wider picture of risks, resilience and ecological integrity, many of the indicators initially selected were discarded because they either did not have data available and data were unlikely to be procured in the near future, they were ambiguous or bimodal in their responses; or were redundant and the information they intended to capture was present in another indicator.

Data for calculating the EVI (and initially setting the response levels) were collected for three countries: Fiji, Tuvalu and Australia to provide some initial testing of the model. These data were obtained from country reports, UN, WHO, SOPAC, SPREP, FAO, and other publications from international agencies, centres for risk assessment and management (e.g. Tsunami Centre, NOAA), local experts and government officers. Without being able to go to the countries to train and focus attempts of local authorities to the task of collecting or collating the required data, some indicators were unavailable for this initial testing.

Quantifying vulnerability

Our overriding principle in constructing the EVI was not to introduce complexities into the model unless there was a justifiable reason to do so.

Environmental indicators are of a heterogeneous nature, that is they include variables for which the responses are numerical, qualitative and on different scales (linear, non-linear, or with different ranges). To deal with the heterogeneity, it was necessary to map the possible responses to the variables onto a 0-7 scale where 1-7 was used for the spread of values and 0 or N was used for 'non-applicable' and 'no-data' responses.

Response levels (maximum, minimum and intermediate divisions) for each of the indicators were set wherever possible using the technical literature or by consultation with generalists and specialists in each field. Some levels were set using the data from Tuvalu, Fiji and Australia where available, or as estimates when these sources were unavailable.

Six of the 57 indicators were assigned an intrinsic weighting factor of 5, while the remaining indicators were given the default weighting of 1. This is the equivalent of giving the six weighted indicators the equivalent value of 5 indicators. This weighting was applied to indicators considered to be of central importance to the question of vulnerability in the model. To ensure that the final EVI, REI, IRI and EDI scores remained between the values of 0 and 7, it was necessary to adjust the weighting factors by dividing the intrinsic weighting value (1 or 5) by the average of all weighting values within each sub-index. The 0-7 scores were then multiplied by the adjusted weighting factor prior to accumulation in the sub-index to which they belonged.

The EVI and sub-indices were calculated using an EXCEL workbook. The workbook (Version 7-EVI-calculator.xls) is comprised of seven linked worksheets, each dealing with a different aspect of calculation and reporting. Report Level 1 was the highest level and gives the value of the EVI and sub-indices for each country and measures of confidence in the data. Report Level 2 gives a breakdown of the REI and EDI sub-indices showing relative contribution of meteorological, geological, anthropogenic risks and mitigating factors and ecosystem and biodiversity indicators. Report Level 3 gives the adjusted and raw scores for each individual indicator. A separate copy of the calculator is required to evaluate the vulnerability indices for each country.

After adjustment for intrinsic weighting, the scores for each indicator within a sub-index were averaged to produce a sub-index value of between 0 and 7. Where data were unavailable for an indicator, that indicator was omitted from the average, not given a 0 score, so that it made no contribution to the mean. Because there were also indicators for which the response was 'not applicable' (such as volcanic eruptions in Tuvalu), we calculated two types of index for the EVI and sub-indices. These we termed the Nett and Gross vulnerabilities. Nett vulnerability omits those indicators considered not applicable in a country and describes vulnerability to risks that actually apply in a country. Gross vulnerability assigns a zero value to non-applicable indicators and describes vulnerability in relation to all risk indicators used in the model and therefore vulnerability in a total sense.

In parallel with scoring each indicator against the 1-7 scale, we built into the EVI a way of assessing the reliability of data. These reliability values are reported alongside each index and should be read with them. The data reliability scores give the number indicators which were not applicable, the number with no data; the number of responses which were based on real data; and the number of responses based on 'best guess' or estimated by the operator and/or authorities.

Preliminary results for Tuvalu, Fiji and Australia

The preliminary EVI value for Tuvalu was the highest of the three countries indicating that its environment is the most vulnerable. The score obtained for Fiji was intermediate in value, and that for Australia was the lowest, though the difference between Fiji and Australia was relatively smaller than that between Fiji and Tuvalu.
 

	Australia	Fiji	Tuvalu
EVI (Nett)	3.04	3.79	5.04

When the EVI was decomposed into the REI, IRI and EDI sub-indices and categories of risk, a more complex pattern emerged. The risk exposure index (REI) was similar for the three countries, varying only between 3.13 nett for Australia and 3.49 nett for Tuvalu. The intrinsic resilience index (IRI) varied the most among the countries with Australia having the best resilience score (1.00 nett) and Tuvalu having the worst (7.00 nett). An almost reverse pattern was seen for the EDI. In this sub-index, the highest score (worst conditions) was obtained by Australia and the lowest by Fiji. The values of the sub-indices suggests that the different aspects which form vulnerability can operate independently of one another. Tuvalu was characterised by poor intrinsic resilience, and moderate risks and degradation. In contrast, Australia was characterised by very good intrinsic resilience, but high degradation. Fiji had the greatest vulnerability to meteorological and geological events, and Australia the greatest risk to pollution.

The values obtained for nett and gross indices differed little. Although this suggests that there may be no need to separately calculate the two index types, the present study did not provide a good test because there were very few indicators which were 'not applicable' across all three countries. It is expected that the nett and gross index values will be required when the EVI is extended globally and a greater range of climatic and geographic areas are built into the model.

These results are only very preliminary. The EVI and sub-indices will only provide a reasonable measure of vulnerability if most of the indicators can be filled-in for a country. We suggest here that at least 80% of the indicators (46 of the 57 questions) should be filled out by any one country for a reasonable estimate of vulnerability. In this report, we were unable to reach this threshold for any country. The data are available for the remaining indicators, but are buried, or need to be requested from authorities and will take some time and effort to procure. There were also problems with quality of the data, as we often found conflicting estimates in the literature. These temporary deficiencies in the data means that all the EVI values obtained here are indicative: they are by no means completed estimates and should be read with caution.

In conclusion, the EVI model gives single-figure measures of environmental vulnerability that appear to be able to distinguish countries and identify sources of vulnerability within a country. For the moment, it appears that of three countries tested, Australia is the least vulnerable, and Tuvalu the most. There were similar levels of risk in each of the countries, the most degradation in Australia and the least intrinsic resilience in Tuvalu. These results are only preliminary because there was insufficient time to collect all of the data required for these three countries and there were some problems with reliability of the data. We expect that each of these problems can be overcome and the results suggest that the EVI will be a useful tool for characterising vulnerabilities of states.

Strengths and weaknesses of the EVI

As for all methods of summarising and modelling data, the EVI developed here is associated with a number of strengths and weaknesses which must be understood for its proper application and use.

Strengths: The EVI is based on a theoretical framework that prompted us to find indicators for all identified aspects of vulnerability. It is able to incorporate quantitative and qualitative data on different response scales and identifies two types of vulnerability (Nett and Gross) simultaneously allowing for a world-wide comparison of states and as assessment of the real risks likely to affect a state. It also identifies areas in which local environmental agencies could improve data collection. Although at present it focuses on Oceania, it is extendable world-wide by the incorporation of new indicators.

Weaknesses: The index does not exclusively rely on published data resulting in omissions and a high cost of data collection. There is some subjectivity in assigning weights to indicators and non-linearities to the scores. The mapping of data on a 7 point scale may result in a loss of detail compared with directly using numerical data. In common with all indices, the EVI is affected by the indicators chosen and the results obtained may differ if different variables were chosen. Local variations, short and long term effects and other details could not be incorporated into the model without making it too complex. The index is also subject to problems with differences in interpretation of in-country users, though this could be minimised with training.

Future directions and conclusions

The environmental vulnerability index developed here will require further refinement before it becomes fully operational. This will include review of the indicators selected and the levels selected in a world-wide context, adding indicators for parts of the model which are underrepresented, and identifying indicators which should be incorporated and for which data should be collected. Refinement of the EVI will require peer review and inputs from highly-specialised experts. Mechanisms for this include running a 'think tank' and by publishing in an international reviewed journal. It will also be necessary to carry out consultations with the SIDS. A second mechanism for refining the index will be to go to, say, 10 SIDS to build their capacity to work with the EVI and to collect data which may at present be buried rather than lacking. Data should also be procured from developed countries by request. When sufficient data have been collected and the indicators refined, it will be necessary to test the performance of the model to identify biases, remove redundant variables, test its ability to differentiate countries, the method of accumulating scores and assess the availability and confidence in the data. It will also be necessary to develop a user-friendly computer calculator for the EVI. The final stage in the development of the EVI will be to combine it with other indices, such as the Economic Vulnerability Index to give an indication of overall vulnerability of states.

The results show that it is possible to obtain a single figure measure of vulnerability which incorporates the risks, intrinsic resilience and health or integrity of the environment. This study also shows that data which were previously thought to be difficult to obtain can be obtained. The methodology selected in the computation of the index can produce results which could have operational usefulness for ranking countries according to their environmental vulnerabilities. It is envisaged that the EVI would be recalculated every 5 years to provide updates on the vulnerability status of countries. This index highlights the need for governments to upgrade their collection and collation of environmental statistics. In addition, the breakdown of results into meteorological, geological, anthropogenic, mitigating and other categories of risk highlights areas of concern for environmental action.