1Over the last 10 years, in the wake of the International Decade for Natural Disaster Reduction, a new science has emerged: cindynics, or risk analysis. This science, the purpose of which is to study and understand not only the natural hazards in their physical dimensions, but also vulnerability factors for property, territories and populations, so as to develop preventive action. While some work has been devoted to the understanding of a certain number of natural hazards, the way in which they can be forecast, and their prevention, the characterisation of the different types of vulnerability is as yet a delicate issue, on account of the numerous approaches to, and definitions of, the concept. This difficulty in assessing the vulnerability of a society or a territory, whether in terms of physical quantification or in terms of social causes and consequences, means that comparative studies between territories or between societies are not feasible. As vulnerability factors are frequently narrowly linked to the lives of the people in a given society, vulnerability studies have always concerned precise, well-outlined territories (Chardon, 1994). Comparisons between different territories and societies are rare. The lack of comparative elements means that it is difficult to position a natural event on a scale: thus when the river Aude and its tributaries in southern France flooded causing ten deaths, the media talked of a disaster, while at the same moment mud flows on the northern coast of Venezuela claimed nearly 50 000 lives.
2While the aim is not to denounce a self-centred streak in our societies, this study sets out to estimate the exposure of the different countries across the world to natural risk. As noted above, it is difficult to estimate the vulnerability of exposed property, material goods, territories and populations because the concept itself is fairly ambiguous. Nevertheless, if we refer to the French environment ministry definition, according to which “natural risk corresponds to the conjunction of a natural phenomenon said to be a ”hazard“ and the vulnerability of the exposed persons and property”, in a comparative study on levels of exposure to natural risks for each country across the world, we need to develop a global method for estimating vulnerability in all settings.
3An overview of the different definitions proposed for vulnerability shows that two approaches appear relevant: firstly an approach that aims to estimate vulnerability in terms of the financial loss that would be incurred following a natural disaster, and secondly an approach that integrates a wide range of social, demographic, cultural and historical factors. In our study we therefore sought to identify a series of indicators liable to enable the assessment of vulnerability according to these two approaches. These indicators needed to be readily available for all countries across the world, and to be reasonably reliable. In view of the great difficulty in obtaining detailed socio-economic and demographic data for all countries, we restricted our study to four synthetic indicators: the GDP per inhabitant, the Human Development Index, the urbanisation rate, and the demographic growth rate.
- 1 United Nations Development Programme
4The first task was therefore to compile the information required for the different calculations to determine vulnerability and to assess hazards. For vulnerability, we looked for indicators in different UN world databases (UNDP1 2003). The assessment of hazard was performed using the Centre for Research on the Epidemiology of Disasters (CRED) AMDAT database, which has catalogued 13000 natural disasters worldwide since the year 1900. Natural disasters are classified per country and per main type of phenomenon (earthquake, flood, storm, cyclone, tropical storm, avalanche, landslide, drought etc). The compilation was performed from heterogeneous sources, such as reports from NGOs, governments, and insurance companies, or articles in the press. Clearly, as in any compilation, the levels of reliability and exhaustiveness of the information differ from one country to another, and even more so from one period to another. We therefore chose to focus solely on the last 30 years, since we considered that the information available is more reliable and more exhaustive. From this data on natural disasters we grouped events according to four main classes: earthquakes, floods, droughts, and wind phenomena (cyclones and storms). We deliberately excluded phenomena such as avalanches or landslides, which account for only a small proportion of the hazards causing natural disasters.
5Once this data was compiled, it was entered into a geographical information system so as to calculate on the one hand the level of exposure to natural hazards for each country, and on the other the degree of vulnerability according to the two approaches selected. Thus in the first part of this paper we present the methods of calculation of hazard and vulnerability, and in the second part we discuss the results obtained.
6Considering the generally accepted definition according to which “a natural risk corresponds to the conjunction of a natural hazard and the vulnerability of the territory and the populations exposed”, we first need to set out the different methods of calculation of these parameters according to the approaches to vulnerability selected.
7Since a given hazard is independent from either of these approaches, the same method of calculation was used for both approaches to the estimation of natural risk.
8Using the French environment ministry definition, we propose a probabilistic approach to calculating hazard from the sum of the probabilities of occurrence of the four main types of natural phenomena retained for this study. This probability is calculated from observations made over the last 30 years that are available in the CRED EDM-DAT database. As natural phenomena do not all have the same intensity, each probability is weighted by the mean number of victims (deaths caused by these phenomena). A weight of 1 was chosen for floods, 2.2 for wind-related phenomena (cyclones, tornadoes, storms), 8.3 for earthquakes and 15.4 for drought.
9Thus the hazard is defined according to the following formula: [2.1]
10where Pi corresponds to the probability of occurrence of each type of phenomenon noted i, xi to the number of dead as a result of the event, and n to the total number of natural disasters according to type.
11In this approach to hazards, we could have focused solely on the probabilities of occurrence of the phenomena, but that would have amounted to placing phenomena with different consequences on the same scale of intensity (Dauphiné, 2001), and it therefore seemed appropriate to weight each natural phenomenon according to a mean intensity estimated from the number of dead.
12Following this definition of hazard, we can now go on to consider the quantification of vulnerability, using the two approaches, human or “social”, and “economic”.
13This first “social” approach to the concept of vulnerability focuses on social, demographic and cultural factors, and does not take account of economic factors derived from quantifying damage and losses incurred in a natural disaster in terms of financial cost. The choice of the parameters to include in the calculation of vulnerability was conditioned by two elements – the availability of information, and the scale. We chose a set of socio-economic and demographic parameters that were available for most of the countries in the world.
14The first parameter selected is the Human Development Index (HDI) proposed by UNDP. It is a “convenient synthetic index” according to Ercole & Pigeon (1999), which combines several socioeconomic factors (health, wealth, education) that are representative of the level of development in a country. In the calculation of vulnerability, the HDI is therefore entered as a inverse value, since it can be considered that the more developed is a country, the more likely it is to be able to protect itself against natural phenomena.
15The second parameter chosen is the gross domestic product (GDP) per inhabitant, estimated as an index and in inverse logarithmic value. Here again we consider that the wealth of a country is liable to enable it to develop means of preventing and forecasting natural disasters, and to respond fast in case of need. Indeed it is generally observed (André 2003) that the efficiency of emergency and health services, the number of hospitals available, and reactivity in case of epidemics are related to the wealth of the country. The use of the GDP in addition to the HDI might appear redundant. However, although GDP enters into the calculation of the HDI, the two indicators are not strongly correlated (correlation coefficient between GDP/inhabitant and HDI =0.30 over 174 observations). The conversion of the GDP/inhabitant value into a logarithmic value enables minimisation of size effects in the final calculation of vulnerability.
16By using the definition of vulnerability proposed by Ercole and Thouret (1994), that is to say the likelihood of sustaining damage in relation to the ability of a given society to respond to natural hazards, we consider that the wealthier a society, the less difficulty it will have in restoring balance following a natural disaster. Conversely, the poorer a society, the less ability it will have to respond to a disaster, and the longer it will take to return to a state of balance.
17The third parameter selected in the calculation of vulnerability is the share of the urban population in the total population of the country. Although this parameter cannot be directly correlated to increased risk, certain authors (Ledoux, 1995, Blaikie 1994) consider that urbanisation, which means spatial concentration of the population, is a factor that increases vulnerability. However, carefully planned urbanisation, avoiding settling populations in zones exposed to natural hazards (flood plains, unstable slope, areas at the foot of volcanoes etc) can also be factor that reduces vulnerability. Overall it can be said that the most vulnerable urbanised areas are those where uncontrolled exponential population growth has occurred as a result of migrations from impoverished rural areas. For Ledoux (1995) the percentage of urban population in developing countries is a good indicator of vulnerability, because the process of urbanisation is linked both to the migration of rural populations, and to vigorous demographic growth: this leads populations to settle massively in cities, without allowing time for the infrastructures required to accommodate them to be developed.
18To enable a distinction between countries where the large proportion of urban population is a historical characteristic from those where urbanisation is linked to recent processes mainly resulting from the impoverishment of rural areas and strong demographic growth, we decided to weight the percentage of urban population by the rate of demographic growth over the last ten years.
19From these four parameters we propose an estimate of the vulnerability of a country using the following formula [3.1.]:
20where HDI is the Human Development Index for 1999, the GDP/inhabitant is the figure for 1999; U is the share of the population that is urban in the total population of the country, and Cp is the rate of demographic growth from 1990 to 2000.
- 2 NR=hazard*asset
- 3 MunichRe
21The second “economic” approach to vulnerability is based on a global estimation, a priori, of the economic losses incurred in case of a natural disaster. It will therefore depend on the amount of interests at stake , generally expressed in monetary value. In this approach, the greater the value of the assets threatened, the greater the vulnerability. However, since the sole value of the items exposed is not adequate, (it would amount to returning to the first definition of natural risk proposed by Thouret2 in1996), we need to evaluate vulnerability from criteria other than that of value. Indeed, two items of the same value will not have the same degree of vulnerability towards a hazard, depending on how well protected they are, and the ability to foresee or prevent the natural phenomenon. If the degree of protection can be estimated on local scale, it is virtually impossible to assess the degree of protection of a country or a continent - the level of protection is closely linked to prevention policies and the level of development of the country. For instance, we can consider the example of the Munich reinsurance company3 which estimates vulnerability, for some thirty megapolises across the world, according to the type of building, policies for prevention and protection, and the density and the urbanisation level of the population. This is then combined with a value for properties, GDP per inhabitant, and the weight of each city in the global economic network (MunichRe, 2002). This approach, which has the advantage of combining physical and economic factors, cannot be used for research on a global scale, since the value of buildings and physical vulnerability cannot be estimated satisfactorily. It is however worth considering for work on the scale of a single city.
22The method of calculation of vulnerability that we propose here according to an economic approach is based on the combined use of economic indicators such as the GDP per inhabitant, socio-economic and demographic indicators such as the HDI, the share of urban population in the total population, and the population growth rate over the last 10 years. The formula for the calculation of economic vulnerability is as follows [3.2]:
23where HDI in the Human Development Index, GDP/inhabitant is the figure for 1999, U is the share of urban population in the total population of the country, and Cp is the rate of demographic growth from 1990 to 2000.
24In this second approach, GDP per inhabitant (expressed as a logarithmic value) enables estimation of the global wealth of a country, and hence the value of the exposed items, assets, property etc. Thus GDP is a factor that increases vulnerability. So as to provide a partial assessment of the ability of a country to foresee and protect itself from natural disasters, we used the inverse HDI. Here we consider that the more developed is a country, the more likely it is to implement policies for forecasting and preventing natural disasters, and the more likely it is to make resources available to protect the assets that are exposed. Thus the HDI is used as a factor reflecting reduced vulnerability. For instance, concerning protective measures against earthquakes, it can be noted that anti-seismic building regulations are overall well implemented and complied with in “wealthy” countries, while in developing countries such as Turkey, Algeria or countries in Latin America, these rules are virtually non-existent, or not complied with. One telling example is the Izmit earthquake in Turkey (1999) where it can be noted that a large proportion of the buildings destroyed dated from the 1990s, a period in which anti-seismic building regulations were already in force in the country (André, 2003). The same phenomenon was observed for the Algiers earthquake in May 2003 (André et al, 2003). Likewise, risk prevention and urban planning established in developed countries is often lacking in the developing countries. On the basis of these observations, it appeared necessary to use a development indicator to weight the calculation of “economic” vulnerability. Finally, as noted by Ledoux (1995) the urbanisation rate is also a reflection of the vulnerability of a society. It can be thought that highly populated, very urbanised zones are the most vulnerable because of the concentrations of assets and wealth. However, long-standing urban settlements are for the most part provided with urban planning systems. They have already experienced natural disasters, and have drawn lessons from them. They therefore appear less vulnerable than recent urban settlements resulting from fast, uncontrolled demographic expansion, and migration from impoverished rural areas. These new trends in urbanisation are not accompanied by control over immigrant flows, and populations often settle in the peripheral areas outside the city centres, in sectors that are more exposed to natural phenomena. Therefore, so as to include a historical component characterising the urban structure of a country, two last factors are integrated into the calculation of vulnerability - the share of the urban population in the overall population of the country, weighted by the inverse demographic growth. This differentiates two types of urban structure.
25The two approaches to vulnerability being thus defined, natural risk can be defined by the two equations below ([3.3, 3.4]:
26Natural risk according to a social approach: [3.3]
27Natural risk according to an economic approach: [3.4]
28Where Pi corresponds to the likelihood of occurrence of each type of phenomenon noted i, xi to the number of dead according to the phenomena, and n to the total number of disasters per type of phenomenon; HDI is the Human Development Index, GDP/inhab is the gross domestic product per inhabitant in 1999, U is the share of the urban population in the overall population of the country, and Cp is the demographic growth rate from 1990 to 2000.
29From these two definitions of natural risk, for each country in the world we calculated levels of risk exposure according to the two approaches. The following chapter presents the results in the form of maps of risk exposure
30Before presenting the results of these two approaches to natural risks in the form of maps, it seems essential, for proper interpretation of the results, to consider the levels of exposure to natural hazards leading to disasters in the different countries. As the natural risk is a combination of a hazard and a level of vulnerability, if the different levels of exposure of the different countries are not taken into account, the final results concerning risks on the global scale and the comparison of the two approaches used in the calculation of vulnerability could be misinterpreted.
31To define the level of exposure to hazards for each country we first of all developed a map estimating exposure on the basis of probabilities for occurrence of each natural hazard retained in the study, calculated over a period of 30 years. As this first method does not take account of the intensity or the dangerousness of the hazards, this led to two thirds of the world being classified as strongly exposed to damaging natural hazards.
32However, as remarked above, the different natural hazards do not all have the same destructive power, so that the probability of occurrence of each hazard needs to be weighted by a global dangerousness or intensity factor. We therefore returned to the formula [2.1] proposed above to calculate a global level of exposure to hazard according to the country (see figure 4.1), weighted by the intensity of the events. This second method of calculation of exposure to a hazard shows that the countries that are the most exposed to destructive natural phenomena are the USA, Mexico, the northern part of South America, Asia Minor, Central and Eastern Asia, Oceania and Sub-Saharan Africa and Sudan and East Africa. Regarding Europe, high values are observed on the Atlantic coastline, and moderate values in the eastern areas.
Figure 4.1. Map estimating levels of exposure of the different countries, developed from the sum of probabilities for occurrence of each type of hazard, and weighted according to the dangerousness of the phenomenon.
33This second map thus enables spatial differentiation of the levels of exposure to hazards leading to natural disasters. Regarding the physical component of natural risk, it can therefore be seen that the different countries and regions across the globe are not on an equal footing. Given this observation, we will now consider the estimation of the “non-pshysical” and non-random component of risk: vulnerability, estimated here according to the two approaches presented above.
34This section presents the final results concerning the characterisation of natural risks, based firstly on data relating to exposure to hazards presented above, and secondly on the two approaches to vulnerability described above. We will start by the estimation of natural risk via a social approach, and then go on to the economic approach. The results are presented in the form of maps so as to confront the two methods applied. For each, four levels of risk were defined, ranging from low to very high, based on means and standard deviation.
35The social representation of risk (see figure 4.2) confirms our assumptions concerning inequalities between North and South for natural risks. Indeed, it can be seen that there is a dichotomy across the planet, with “wealthy” countries in the northern hemisphere, Oceania and South America, for which the level of natural risk is fairly low, contrasting with “poor” or developing countries in Central America, Africa and Asia, where levels of natural risk are relatively high.
Figure 4.2. Map of estimated levels of risk for the different countries according to a social approach to vulnerability
36Careful analysis shows that there is a first group of countries with a high level of risk, comprising Sub-Saharan Africa and Sudan, the Great Lakes region of Africa, the Indian sub-continent, China, the Indonesian archipelago, and Mexico, Peru and Bolivia. In this first group, it can be noted that there are marked disparities in levels of development, whether in social, technological or economic terms. Certain countries like China, India or Mexico are fairly well developed economically and technologically. Their allocation here to a group of countries with a high level of risk can therefore be explained in part by the unequal distribution of wealth in the population, or else by high levels of exposure to destructive natural phenomena. The Africa group cannot however explain its high levels of risk by high levels of exposure to natural phenomena. Although they cause large numbers of deaths, the droughts that affect this area of the globe are nevertheless infrequent, and cannot on their own explain the high risk values. Therefore the answer needs to be sought in the social vulnerability of these populations. The countries in this part of Africa are the least developed, with socio-demographic, health and economic indicators at very low levels.
37If we consider the countries presenting a moderate to low level of risk, the analysis of the levels of exposure to natural hazards shows that most of these countries are strongly exposed to natural phenomena. What in fact reduces their overall level of risk and their social vulnerability is the high level of economic and technological development, and a more even distribution of wealth and knowledge in their populations.
38While the characterisation of natural risk according to a social approach to vulnerability has highlighted inequalities between societies in the North and South, what can be said of levels of risk in an economic approach to vulnerability?
39As for the social approach, the level of risk was calculated for each country according to an economic approach to vulnerability using the formula presented in 2.5. We then developed a map showing four levels of risk (figure 4.3.) using the same breakdown as for the social approach (from means and standard deviation).
Figure 4.3: Map of estimated levels of risk for the different counties according to an economic approach to vulnerability.
40The first observation to be dawn from this map (Figure 4.3) is that the inequalities between North ad South observed in the first approach are less marked. Of course, here it is the wealthy countries that present the highest risk, but a very large proportion of South America, and India and China too, are also classified as presenting high risk levels. This observation can be explained by the fact that these countries, considered as “poor”, have vigorous economic activity, and produce as much wealth if not more than a number of “wealthy” countries. Only Africa, Indonesia and Scandinavia present low levels of risk. Regarding Scandinavia, it is probable that the low level of risk cannot be attributed to the low level of economic activity, but rather to low levels of exposure to natural hazards. For Indonesia and Africa, the low level of risk can be explained more by low levels of production of economic wealth than by low levels of exposure to natural phenomena.
41This second approach to vulnerability thus shows reverse results compared with those derived from the social approach. Here the most exposed countries are those that in the first approach presented the lowest levels of risk.
42Without claiming to offer an ideal solution, this work on vulnerability towards natural disasters provides a basis for reflection on the approaches to be envisaged in estimating natural risks on the scale of a continent or worldwide. Although the two methods of calculation contrast one with the other, they are to degree complementary, and their results are not actually contradictory. Combining them to assess risk highlights three groups of countries: one characterised by low risk (Scandinavia, southern Africa, the Indochina peninsula), a second that has high exposure to risk (Mexico, the north of South America, India, China, Indonesia, the Philippines), and a third group with moderate risk exposure made up of the other counties in the world.
43These two approaches to risk are rarely combined. The focus is either on financial cost (the viewpoint of insurance companies) and it is the economic approach that is adopted, or the focus is on the social and cultural aspects of the risk. In the first instance the geography of risk will present the wealthier counties as being more exposed, along with a few developing countries in Asia and South America; in the second, marked inequalities between North and South will be observed, with the poorest nations being the most exposed to natural risks.
44Finally, in studies on natural risks, a dichotomy is frequently observed between the proponents of risk analysis according to natural hazards, and the proponents of an approach based on levels of vulnerability, yielding an incomplete analysis of natural risks. By combining the study of hazards in a probabilistic perspective with the study of vulnerability according to two empirical approaches (social and economic) we propose a more exhaustive method of analysis of natural risks across the globe, enabling comparisons among different regions in the world, and removing the focus from the local setting, which is the most often studied. There are however some limitations to this new approach to risk, still experimental in nature. Since we aimed to propose methods enabling comparisons between countries, we were obliged to resort to the synthetic socio-economic and demographic indicators that were available for all countries. These indicators may present a degree of correlation, but for the world as a whole, we did not note any significant correlations between GDP and HDI, or between urbanisation rates and urban growth. Finally, to allow for size effects linked to GDP per inhabitant, we opted for a logarithmic conversion of this value, which enables a reduction of the weight this indicator in the two methods for calculating vulnerability levels.