Chapter 3 – Deconstructing disaster: patterns and trends in risk and poverty at the sub-national and local level 3.1 - Introduction Most disaster mortality and direct economic loss is concentrated in internationally reported large-scale disasters, which configure the global patterns and trends in disaster risk reviewed in Chapter 2.1 While such disasters capture international political and media attention, they are only the visible tip of a much larger risk iceberg. Largely invisible to global observers, this chapter opens a window on rapidly evolving and emerging trends in disaster risk at the sub-national and local levels and on their relationship with poverty. For example, we will explore evidence from Asia, Latin America2 and Africa that suggests a rapid increase in disaster risk associated with floods and heavy rains since 1990. These flood disasters are not associated with massive mortality and economic loss. Rather such emerging risks are highly dynamic, mirroring patterns of urban growth and economic, territorial and environmental change. We will also explore the evidence that different kinds of disaster have a disproportionate impact on the poor. In particular, we will highlight the hidden and sometimes long-term impact disasters have on poor households, and on particularly vulnerable groups such as women and children. The analysis builds on important recent advances in the compilation of national disaster data in both Asia as well as Latin America, which for the first time enables the exploration of disaster risk patterns and trends at a high level of resolution. As such, in this chapter global disaster risk is deconstructed and viewed from a national level of observation and a local, municipal level of resolution. Poverty data at a local or household resolution also exists in many of the same countries and has been compared with disaster data in a series of case studies in nine countries3 Together with a systematization of the findings of previous studies on disaster and poverty in Africa, this has enabled the identification of the different mechanisms through which disaster risk and poverty interact The country sample analyzed is characterized by a wide range of hazard types, development contexts and geographic conditions. Nevertheless, high resolution disaster data is still not widely available in Africa, Europe as well as in many small-island developing states and developed countries. As such, while the findings point to many broad patterns and trends across different contexts they may not be globally valid. 3.2 - A question of resolution Disaster risk reveals increasing complexity when observed at higher resolutions. Explored at the subnational and local levels, patterns and trends come into focus, which are essentially invisible when observed from a global perspective. International observers view large-scale disasters as a point or area on a map, which represents the statistics of mortality and direct economic loss associated with the disaster. A closer examination, however, reveals that losses are not evenly spread over the area affected: most of the losses tend to be concentrated intensively in specific locations. At the same time, diffuse patterns of damage and destruction spread extensively over wide areas become visible alongside these concentrations of intensive 1 Globally the vast majority of mortality and economic loss is concentrated in a few mega-disasters. An analysis of the EMDAT database, carried out for the ISDR, Disaster Risk Reduction: 2007 Global Review indicated that between 1975 and 2005, 82% of the disaster mortality registered in EM-DAT was associated with only 20 mega-events with over 10,000 deaths each. Economic loss is less concentrated but 38.5% of total economic losses were found to be concentrated in just 21 mega-events that caused more than US$10 billion of loss. 2 The Andean region (Bolivia, Colombia, Ecuador, Peru and Venezuela), Argentina, Mesoamerica (Costa Rica, and Mexico), South Asia (the Indian States of Orissa and Tamil Nadu, Nepal and Sri Lanka) and Iran. 3 Case studies on Bolivia, Ecuador, El Salvador, India, Iran, Mexico, Peru, Nepal and Sri Lanka were commissioned specially for this report by UNDP. Additional case-study material on Indonesia was contributed by the World Bank and on Fiji by SOPAC. 1 risk. For example, in the case of the Armenia earthquake, Colombia, 25 January 1999, eight municipalities intensively concentrated 98% of the deaths and 95% of the destroyed houses. However, the impacts of the earthquake were extended over a much wider area (Map 3.1). Map 0.1: Armenia, Columbia earthquake extensive and intensive impacts (1999) At a high resolution, other risk patterns also come into focus, associated with thousands of frequently occurring small-scale hazard events, such as floods, heavy rains, landslides, fires, and storms, which are invisible to the global observer and go largely unrecorded. In this chapter, the term intensive risk is used to describe those areas nesting within global disaster hotspots that concentrate most of the mortality and destruction. Intensive risk is generally associated with major hazards, such as large earthquakes, tsunamis, cyclones and volcanic eruptions and major concentrations of exposed and vulnerable population and assets. The term extensive risk is used to describe the diffuse, spatially extended risk patterns that only become visible at a higher resolution. Extensive risk is associated with specific local concentrations of exposed, vulnerable population and assets spread over wide areas and highly localized small-scale as well as major hazard events. Box 0.1: Intensive, extensive and everyday risk in Cali, Colombia 4 If the resolution is further increased, extensive risk can also be deconstructed. Risk patterns at the neighbourhood or household level come into focus, associated with everyday hazards such as house fires, domestic and occupational accidents. The number of disasters reported in four different disaster databases, illustrate in the case of Cali, Colombia, how different patterns of intensive, extensive and everyday risks become visible when risk is viewed at different resolutions. From a global perspective, the EM-DAT database reports 8 intensive disasters that affected Cali between 1906 and 2007. Increasing the resolution, a national DesInventar5 database reported 720 mainly extensive disasters between 1970 and 2007 and a city-wide database, 1,280 events. At an even more detailed resolution, between 1987 and 2007, the Cali Fire Corps reported 80,121 emergencies, reflecting everyday risk. Intensive and extensive disaster risk, therefore, are not mutually exclusive categories. Both have similar causes (hazards, physical exposure and vulnerability) and outcomes (mortality, physical destruction and damage, economic loss and livelihood interruption). The term intensive or extensive refers to the degree of concentration of the risk in space and time, at a given resolution. When risk is highly concentrated in space and time then it is more intensive in character. When it is spread out it is more extensive in character. 4 Jimenez, Nayibe, 2007, Urbanizacion, marginalizacion y prefiguracion de desastres en ciudades medianas de paises en desarrollo, Corporacion OSSO, Colombia. 5 See DesInventar at: http://gar-isdr.desinventar.net/DesInventar/main.jsp or http://online.desinventar.org/ 2 Whether a risk is described as intensive or extensive depends on the level of observation and resolution. As described in Box 0.1, the Armenia earthquake disaster appears intensive when observed globally but unfolds into a complex pattern of intensive and extensive impacts when viewed at a higher resolution. Some disasters, such as the landslide in Villa Tina, Colombia in 1987, which killed more than 500 people in a single neighbourhood are intensive, whether observed globally or locally. Except in the case of largescale disaster, however, it is not possible to fold extensive into intensive risk by decreasing the resolution. Extensive risk manifestations, associated with localized small-scale hazards become invisible if examined from a global perspective. 3.3 – Identifying extensive risk Extensive risk patterns in the sample were identified from a dataset specially updated, revised and compiled for this report from national DesInventar disaster inventories.6 The dataset consists of 126,620 disaster reports associated with both geological7 and hydrometeorological8 hazards occurring between 1970 and 2007 in the 12 countries:9 a statistically robust sample that allows a high degree of confidence in the trends and patterns identified. Disaster losses, reported in a given local area, by definition manifest the specific configurations of risk present in the area at the time. In DesInventar, a disaster report documents the losses that occurred at a given time in a given local administrative area, such as a municipality, district or block, together with information regarding the associated hazard type and cause. The impact of a large-scale earthquake or cyclone, therefore, is documented as multiple local disaster reports. The impact of a small-scale hazard affecting only one municipality appears as a single report. Relatively robust and comparable disaster loss data exists for attributes such as mortality, housing destruction and damage. Loss data on crop and livestock loss is far less robust. As a result, risks associated with droughts and rural agricultural livelihoods are not addressed in the first part of this chapter. This imbalance is to some extent redressed in the second part of the chapter, through the use of household poverty data from rural households. In order to identify extensive risk, a loss threshold of 50 deaths or 500 houses destroyed was established on the basis of a statistical analysis of the dataset (Technical Note 3.1). A disaster in a given local administrative area above the threshold is considered to represent intensive risk. A disaster with 49 deaths and 449 destroyed houses or less is considered to represent extensive risk. While any quantitative definition of extensive risk is arbitrary, the threshold of 50 deaths and 500 destroyed houses was considered both reasonable as well as statistically valid for the data sample at a local resolution. Applying the threshold to the sample, manifestations of intensive and extensive risk associated with both hydro-meteorological and geological hazards were identified. See Technical Note 3.1 for the distribution o 3.4 – The underlying risk patterns Extensive risk in time and space Of the 126,620 disasters in the sample, only 0.7% represent manifestations of intensive risk. Intensive risk disasters occur infrequently. Across the sample there is an annual average of only 27 disasters, 6 The dataset excludes hazard events without reported damage and any disaster reports without adequate source documentation, see DesInventar at: http://gar-isdr.desinventar.net/DesInventar/main.jsp or http://online.desinventar.org/ 7 Earthquakes, tsunamis and volcanic eruptions are considered as geological hazards in the analysis that follows. Landslides may be either geological or hydrometeorological and are often both. For the purposes of this report they have been classified as hydrometeorological, although recognising that many are related to earthquake occurrence. 8 Floods, flash floods, urban floods, rains, fires, forest fires, mudslides, avalanches, landslides, cyclones, storms, gales, strong winds, hailstorms, tornados, electric storms, hailstorms, lightning, thunderstorms, storms, droughts, heat waves, cold waves, frost, snowstorms. 9 The Peru database covers the time period 1997-2006; the Mexico database covers the time period 1980-2007 and the Tamil Nadu database 1976-2007. 3 approximately 1 disaster every two weeks. Intensive risk is also spatially concentrated. As Map 0.2 shows, only a handful of cantons manifested intensive risk in Ecuador over the reporting period. Map 0.2: Ecuador intensive disaster risk events (1970-2007) In contrast, 99.3% of the disasters represented manifestations of extensive risk. Extensive risk is continuous in time with an annual average of 3,395 disasters, equivalent to 9 disasters per day across the sample. Extensive risk is also widely spread in territorial terms. Map 0.3 shows, all except a handful of cantons in Ecuador manifested extensive risk in the reporting period. Map 0.3: Ecuador extensive disaster risk events (1970-2007) The spatial distribution of extensive risk is further highlighted by Table 0.1 which shows that across the 12 countries, 82% of local administrative areas were affected at least once during the reporting period and 48% have been affected 6 or more times. Table 0.1: Spatial distribution of extensive risk (1970-2007) Number of reports 0 Number of local administrative areas 982 4 % Cum % inv Cum % 17.90% 100.00% 17.90% 1 2 to 5 6 to 10 11 to 20 21 to 50 51 to 100 more than 100 Total 639 1218 717 729 647 291 262 5485 11.65% 22.21% 13.07% 13.29% 11.80% 5.31% 4.78% 100.00% 82.10% 70.45% 48.24% 35.17% 21.88% 10.08% 4.78% 29.55% 51.76% 64.83% 78.12% 89.92% 95.22% 100.00% Contemplating these maps and figures provides insight into a vision of disaster risk very different from that which most international observers are familiar. Rather than occasional catastrophes occurring along earthquake fault lines or on tsunami exposed shorelines, extensive disaster risk appears as a continuous ongoing manifestation of small-scale impacts affecting to a greater or lesser extent all of a country’s territory. The hazard profile of extensive risk Most intensive risk disasters in the sample were associated with the major hazard types reviewed in Chapter 2: earthquakes, volcanic eruptions, tsunamis, cyclones, floods and in several cases with landslides. Given the data limitations described in Section 3.2 intensive risk disasters could not be associated with drought. Across the sample there were four times more intensive risk disasters associated with hydro-meteorological than geological hazard. This reflects the hazard profile of the countries in the sample and the events that actually occurred in the reporting period, given that hazard events, such as large-scale earthquakes, eruptions or tsunamis, occur with long return periods in specific locations. In contrast, more than 92% of extensive risk disasters were associated with hydrometeorological hazard events. These include periodic cyclones and major floods but also large numbers of small-scale floods, landslides, storms, mudslides and other localized hydrometeorological events. As Figure 0.1 shows, floods, flash floods and heavy rains represent 40.9% of the extensive hydrometeorological risk in the sample. 24.6% of that risk is associated with fires and forest fires, 14% with landslides, mud-slides and avalanches, 12.3% with storm events, 4.6% with drought and heat waves and 3.5% with cold waves, frost and snowstorms. Figure 0.1: Hydrometeorological extensive risk disaster reports by hazard type across the sample While the underlying risk patterns in sample, therefore are principally linked to hydrometeorological hazard, particularly to floods and heavy rains, each country has its own extensive risk profile. In Orissa, India, for example, fires account for almost 59% of extensive risk disasters, given rural villages of tightly packed thatch houses that are extremely vulnerable to fire. The relationship between fires and weather in 5 Orissa is shown in Figure 0.2: seasonally the number of fire disasters peaks in the dry season between February and May and drops in the monsoon between June and November. In Iran, 42.9% of extensive risk disasters are associated with earthquakes and as Map 0.4 shows are widely distributed throughout the countries territory. Figure 0.2: Average seasonal occurrences of disaster reports associated with fire, Orissa, India (1970 – 2007) Map 0.4: Number of extensive geological disaster risk reports from the Islamic Republic of Iran (1970 – 2007) Loss and damage patterns Intensive risk concentrates a very large proportion of the mortality and direct economic loss10 that occurs in the sample. While, mortality statistics need to be approached with caution, as Box 0.2 indicates, the 0.7% of the disaster reports that manifest intensive risk account for 83.7% of the total deaths, as well as 74.6% of the houses destroyed. This kind of loss pattern is coherent with the effects of highly destructive hazards such as earthquakes, tsunamis, cyclones and volcanic eruptions. Mortality is particularly associated with geological hazard, which, while associated with only 19% of the intensive disaster reports, accounts for approximately 78% of the mortality. The distribution of these loss and damage patterns in time is shown in Figure 0.3. Mortality in intensive risk disasters is concentrated in a small number of major peaks, occurring every five years or so during the reporting period. In contrast, extensive risk mortality appears as a continuous underlying flow. Figure 0.3: Distribution of mortality associated with intensive and extensive risk across the data set (1980-2006) Box 0.2: When the dead go missing 10 Destruction in the housing sector is taken here to be a proxy for direct economic loss. 6 In the year 2000, the World Bank, describing the impact of natural catastrophes in 1999, stated that “the landslides in Venezuela alone caused 50,000 fatalities”.11 The EM-DAT database records 30,000 deaths due to the same set of floods, mudslides and landslides, which occurred in December 1999 and affected Vargas, Miranda, the Federal District of Caracas and to a lesser extent other states. Two recent papers published by anthropologist Rogelio Altez 12 of the Universidad Central de Venezuela paint a very different picture. After a lengthy forensic investigation in Vargas state, Altez documented a total of only 521 corpses found after the disaster, of which 290 were unidentified, and only 331 people reported missing. Given that number of missing was similar to the number of unidentified corpses, Altez concluded that “the total number of deaths does not exceed 700”. After flying over the affected area, the then Secretary General of the IFRC declared to the BBC that Venezuela's disaster was "certainly at least two or three times worse than Mitch as far as the death toll is concerned" and that “as many as 50,000 people may have been killed”.13 According to Altez, statements of this kind, disseminated in the media, later became widely accepted international statistics. The key message from Altez’s study is that there are still major deficiencies in the way corpses are dealt with after many largescale disasters. Documented cases exist of mass cremations and burials without an adequate process of identification or even quantification of the victims, often due to unjustified fear of epidemics. The enormous discrepancy in mortality in Venezuela, between the international statistics and the study carried out by Altez, may be unique. But it does highlight both serious deficiencies in the post-disaster treatment of corpses and the need for a critical approach when dealing with disaster mortality data. In general, the distribution of intensive risk mortality and housing destruction across the sample is coherent with the global patterns of risk outlined in Chapter 2. Higher risk countries, such as Iran have a higher percentage of intensive mortality and housing destruction. In lower risk countries, such as Costa Rica or Argentina, mortality and housing destruction is more extensive. While mortality and housing destruction is intensively concentrated, extensive risk disasters account for 51.3% of damaged housing across the sample. In most of the Latin American countries, over 75% of housing damage is extensive. In Asia the percentage is less given that no less than 58.5 % of the housing damage reported in Asia occurred in Orissa where 84.8% of the housing damage was intensive. This loss pattern is coherent with the effects of floods, rains and storms, which are more likely to damage housing rather than cause mortality or destruction. Research from other countries suggests that mortality and injury only increase significantly in very severe floods, with a large number of affected buildings14. Effectively, flood mortality in the sample is mainly concentrated in intensive events. Housing damage across the sample is particularly associated with floods, flash floods and heavy rains. Floods accounted for 34.7% of hydrometeorological extensive risk disasters but over 60% of corresponding housing damage. Heavy rains were associated with 6% of the disasters but 26.7% of the housing damage. Spatially, extensive housing damage is widely spread while intensive housing destruction is heavily concentrated as is illustrated in the case of Tamil Nadu by Map 0.5 and Map 0.6. 11 World Bank, 2000, Managing Disaster Risk in Emerging Economies, Volume 1. Altez, Rogelio, 2007, Muertes Bajo Sospecha: Investigación sobre el numero de fallecidos en el desastre del estado Vargas, Venezuela en 1999, Cuadernos de Medicina Forense, 13 (50); and Altez, Rogelio, Revet, Sandrine, 2005, Contar los muertos para contar la muerte: discusión en torno al numero de fallecidos en la tragedia de 1999 en el estado Vargas – Venezuela, Revista Geografica Venezolana, Numero especial 2005, 21 – 43. 13 http://news.bbc.co.uk/2/hi/americas/581579.stm 14 In Japan, the threshold above which flood mortality increases has been calculated at 1,000 inundated buildings by Zhai, Fukuzono and Ikeda (2006) reference…. Clearly this threshold will be different in other countries but suggest that in general extensive flood disasters are unlikely to cause major mortality. 12 7 Map 0.5: Extensive housing damage in Tamil Nadu, India (1976-2007) Map 0.6: Intensive housing destruction in Tamil Nadu, India (1976-2007) The cost of extensive risk Characterized by damage rather by destruction and by continuous small-scale manifestations rather than occasional catastrophic disasters, the economic losses associated with extensive risk are largely unaccounted for. Mechanisms which the international community uses to register and account for disaster loss include UN situation reports disseminated by OCHA, reports in an international database such as EM-DAT of which approximately one third include an estimate of economic loss; international appeals for assistance launched by the UN or by IFRC and post-disaster damage and loss assessments by the World Bank, other IFIs or the UN.15 Almost all intensive risk disasters are documented in EM-DAT, a few are described in situation reports and a small number lead to an international appeal and/or a damage and needs assessment. Figure 0.4 compares patterns of mortality and housing destruction and damage in both intensive and extensive risk disasters in Sri Lanka with the periodicity of disasters reported in EMDAT, OCHA situation reports, UN appeals and international damage and needs assessments and indicates. While most years with intensive disaster reports have a corresponding EM-DAT report, the production of OCHA situation reports is more sporadic. There were only three international appeals for assistance and only one international damage and needs assessment, following the 2004 tsunami. It has to be assumed that the losses associated with extensive risk between these international reports went unaccounted for, as well as losses in entire years such as 1997. [MESSY GRAPH: REDRAW] Figure 0.4: Deaths, houses destroyed and houses damaged in Sri Lanka and international awareness (1992-2005) 15 In general post-disaster damage and loss assessments use a methodology developed by the Economic Commission for Latin America and the Caribbean (ECLAC). Economic Commission for Latin America and the Caribbean, 2003, Handbook for Estimating the Socio-Economic and Environmental Effects of Disasters, Santiago, Chile: http://www.preventionweb.net/files/1099_eclachandbook.pdf 8 Despite its invisibility, it is likely that extensive risk represents a major economic cost for the countries in the sample and a significant percentage of global disaster loss that goes unaccounted for. In Mexico, for example, 316,928 houses were destroyed in intensive risk disasters between 1980 and 2006 and 471,708 houses were damaged. In contrast, 1,468,509 houses were damaged in extensive risk disasters and 29,510 houses destroyed. The cost of a destroyed house in Mexico has been estimated as US$16,80016 and the cost of a damaged house as 20% of that value. Based on these estimates, while the cost of destroyed and damaged housing in intensive risk disasters between 1980 and 2007 was US$5,477 million, the cost in extensive risk disasters represented another US$4,831 million. In Mexico, in the housing sector, therefore, extensive risk represented almost 47% of economic losses over the period. Applying the same methodology across the whole sample, extensive risk would represent 34% of the total cost of disaster loss in the housing sector. Additionally, in Mexico 55% of the costs in the housing sector in intensive risk disasters are concentrated in only two years, 2005 and 1988. In contrast, while 2005 also represented a peak, extensive risk housing costs are more evenly distributed over the whole 27 year period. Table 0.2 illustrates, across the sample, that losses in other sectors such as education, health and transport as well as people affected tend to be extensively distributed. 58% of schools, 65% of hospitals, 89% of roads damaged and destroyed as well as 89% of the people affected occurred in extensive risk disasters. Disaster loss data in other sectors is less robust than in housing, meaning that the absolute figures are not meaningful. The proportional distribution between extensive and intensive risk, however, is valid given the size of the sample. Table 0.2: Loss attributes by risk category across the sample (1970–2007) Risk category Loss attribute Schools Hospitals Kilometres of roads People Affected Total Extensive risk % Intensive risk % 32,157 1,037 64,917 182,989,857 18,488 677 57,695 144,627,235 57 65 89 79 13,669 360 7,221 38,362,622 43 35 11 21 This extensive distribution of a significant proportion of disaster costs has very significant implications. To the extent that international assistance following disasters is sensitive only to intensive risk disasters with major mortality and direct economic loss, the losses and costs of extensive risk are largely ignored as well as other less visible but more pervasive long-run impacts on poverty. If, as will be explored later, most extensive risk is associated with urban informal settlements and vulnerable rural livelihoods, international disaster assistance is failing to address a significant proportion of the impact of disasters on the poor. 3.5 – Underlying risk trends Intensive risk is relatively static in geographical terms. Located on seismic fault lines, around active volcanoes, on cyclone paths or on tsunami exposed coastlines, concentrations of intensive risk may change over time, according to the growth or decline of vulnerable population and economic assets exposed. But without a fundamental shift in global hazard patterns they do not move. While intensive risk patterns manifest in different hotspots in different periods, reflecting the return period of different hazards, they are strongly controlled by the global geography of hazard. In contrast, extensive risk is far more dynamic and geographically mobile, given that it is associated with myriad frequently occurring, spatially dispersed and small-scale hazards affecting highly localized concentrations of population and economic assets. Many small-scale hazards, such as localized flooding or fire, are socio-natural in character, given that the hazard itself is shaped by factors such as urbanization 16 Costs normalized using as a baseline the Indice de Precios de la Construccion, 2003 on the basis of an average sized social house of 42 m2 and an average construction cost per m2 of US$400. 9 and change in the use of ecosystem services. An observation of how patterns of extensive risk manifest and transform over time, therefore, provides a unique window to observe underlying trends in disaster risk accumulation and to gain insight into how risk is shaped by development. Extensive hydrometeorological risk Only 3% of the extensive risk disasters in the sample correspond to geological hazard, mainly the extensive impacts of large-scale earthquakes and tsunami. As such, we will concentrate our analysis on trends in extensive risk associated with hydrometeorological hazard. Figure 0.5 illustrates that the average annual occurrence of extensive hydrometeorological disasters has quintupled over the last 37 years across the sample. While the associated mortality has doubled, this implies that the average number of deaths per disaster is actually going down: possibly reflecting improved development conditions and local capacities in disaster preparedness and emergency management. As will be explored in chapter 5, most of the countries in the sample report substantial achievements in the development of institutional systems, policy and legislation, plans and mechanisms for disaster preparedness and response. In contrast, Figure 0.6 illustrates that the associated housing damage has increased at least ten fold. If the average number of damaged houses per disaster has approximately doubled then this can only be explained by an increase in the number of exposed and vulnerable houses or by an increase in the intensity of hydrometeorological hazard. The rapid growth of exposed and vulnerable settlements and the magnification of hazard levels through human intervention would thus seem to characterise urban development and territorial occupation in the countries of the sample. As will be explored in chapter 5, few of the countries in the sample report progress in mainstreaming disaster risk reduction considerations into economic and social development, land-use planning and building or into environmental management. Figure 0.5: Number of extensive hydrometeorological disaster reports and number of associated deaths for data set (1980-2006) Figure 0.6: Number of extensive hydrometeorological disaster reports and number of associated deaths for data set (1980-2006) Error! Reference source not found. provides further insight into these trends by comparing the annual average increase in mortality and housing damage with average annual population growth in each country. If population increase is accepted as a proxy for increased physical exposure,17 then increases or decreases in mortality and housing destruction and damage indicate changes in vulnerability. In most of Latin America, with the exception of Mexico and Ecuador, mortality is falling relative to population size, whereas in Asia it is increasing. This may indicate a greater achievement in disaster preparedness and response in Latin America or simply higher development levels. As was highlighted in chapter 2, risks associated with floods and cyclones heavily correlated with GDP per capita. While the Latin American countries in the sample are upper or lower middle income countries, all the Asian countries, with the exception of Iran are low-income. 17 Unlike the calculation of physical exposure presented in Chapter 2, it is impossible to compute the areas prone to small-scale hydrometeorological hazards, such as localised floods and fires, with a level of resolution to inform the analysis of extensive risk presented here. 10 In contrast, with the exception of Argentina and Orissa, housing destruction is increasing faster than population, indicating increased vulnerability in the housing sector. With the exception of Bolivia, housing damage is increasing faster still. Given the strong association of housing damage with flooding this confirms the trend of a combination of more intense flooding and increased vulnerability of human settlements. annual % change in mortality 1970-2007 from EXT HYDROMET events annual % change in houses destroyed 1970-2007 from EXT HYDROMET events annual % change in houses damaged 1970-2007 from EXT HYDROMET events annual % increase in population 1970-2007* Argentina 0.90 -1.88 2.92 1.33 Bolivia 1.48 6.52 0.03 2.17 Colombia -0.95 3.22 9.48 1.89 costa rica 0.93 3.71 8.18 2.35 Ecuador 3.93 9.41 26.12 2.09 Iran 13.11 / / 2.37 mexico (1980-2007) 6.48 3.48 17.94 1.88 nepal (1971-2007) 4.47 2.80 8.28 2.25 Orissa(1) 5.75 -3.55 7.80 1.66 peru (1970-2006) 1.31 2.80 3.03 1.96 sri lanka (1974-2007) 1.70 11.99 5.68 1.15 tamil nadu(1) (1976-2007) 11.67 4.66 12.23 1.25 Venezuela 0.51 2.56 5.96 2.49 country name *UNEP GEO Data: 1970-2010 5 year average (geodata.grid.unep.ch) (1) annual population increase over 1981-2001 period (Census if India, 2001) Error! Reference source not found., naturally, highlights important national variations. For example Figure 0.7 and Figure 0.8 illustrate that housing destruction in extensive risk disasters in Orissa, India experienced a drastic reduction over the reporting period. Fire is the cause of 92.5% of the houses destroyed in extensive risk disasters in Orissa, India. Over the reporting period thatch houses have been rebuilt with non-flammable materials, leading to a dramatic reduction in housing destruction. Figure 0.8: Housing destruction in Orissa, India (1970 – 2007) Figure 0.7: Causes for housing destruction in Orissa, India (1970 – 2007) An analysis of spatial trends indicates a centrifugal territorial expansion of extensive hydrometeorological risk. Figure 0.9 illustrates that the number of local administrative areas with manifestations of this kind of risk has increased consistently from decade to decade. It is interesting to note that the number of local areas affected by more than 50 disasters over the reporting period, or approximately 2 disasters per year has remained more or less constant. In other words there is a core of local areas, which are always affected. But the number of local areas affected by 1 to 9 disasters has doubled and those affected by 10 11 – 49 disasters has almost quintupled between 1980 and 2006, indicating that many new local areas, previously not affected by flood disasters now experience them while the number of disasters per municipality has also grown. Figure 0.9: Number of municipalities affected by extensive hydrometeorological risk (1980-2006) For example, in Colombia, 373 municipalities reported housing damage in extensive risk disasters in the 1970s. These figures rose to 661 in the 1990s and 692 in the first 7 years of the present decade. The number of municipalities reporting deaths and housing destruction rose from 147 and 388 respectively in the 1970s and 1990s to 507 in the first 7 years of the present decade. Simiarly, as Map 0.7 highlights extensive hydro-meteorological risk in Mexico was concentrated in only a small number of states in the 1980s. Over the last ten years almost all Mexican states now manifest this kind of risk. Map 0.7 Number of extensive risk events (1980-1989, 1990/1999, 2000/2007) Importantly, extensive risk patterns underpin the configuration of new patterns of intensive risk. When a large-scale hazard event, such as a major cyclone, flood or earthquake affects a region, it simultaneously activates a large number of extensive risks. For example, as can be seen in Figure 3.12 and 3.13 the 12 impacts of the super cyclone in Orissa, were concentrated in the same areas that manifest high levels of extensive risk. In other words the expansion of extensive risk constructs intensive risk over time. Map 0.8: Orissa, India extensive disaster risk reports (1970 – 2007) Map 0.9: Orissa, India intensive disaster risk reports (1970 – 2007) Floods and heavy rains Within this overall trend of emerging hydrometeorological risk, as Figure 0.10 illustrates, the number of disasters associated with floods and heavy rains, in the sample, is increasing at a far faster rate than other categories of hydrometeorological hazards, particularly since 1990. Figure 0.10: Number of disasters associated with floods and heavy rains (1980 – 2006) In Mexico, for example, the annual average number of extensive risk disasters associated with floods, rains and flash floods increased 8 fold increase since 1980 (Figure 0.11), accounting for 31% of all extensive hydrometeorological risk disasters in the 1980s but over 40% in the last decade. Similarly, in Colombia (Figure 0.12), floods, flash floods and heavy rains accounted for 43% of extensive hydrometeorological risk disasters in the 1970s but 53% in the last decade. 13 Figure 0.11: Extensive risk disasters in Mexico associated with floods, rains and flash floods (1980–2006) Figure 0.12: Floods and rains in Colombia as a proportion of all extensive risk disasters (1970-1979) and (1998 – 2007) Exceptions to this trend include Peru, where landslides and mudslides constitute the principal hydrometeorological hazard, given that the majority of the urban population is concentrated on a dry coastal desert strip in areas largely not prone to flooding. The city of Lima concentrates approximately one third of the countries population but has recorded an average of only 10 flood events per year, affecting regularly only two or three districts. As was emphasized in section 3.4 housing damage is a quintessential characteristic of disasters associated with flooding and heavy rains. A rapidly increasing number of flood and rain disasters affecting an increasing number of localities is therefore coherent with the rapid increase in housing damage noted above. A key emerging trend in extensive risk in Asia and Latin America, documented in this report, therefore, is a rapid and concurrent in increase in flood hazard, affecting exposed and vulnerable settlements in an ever expanding number of areas. 3.6 Interpreting the trend: urbanization, territorial occupation and extensive risk A number of factors could explain this trend of rapidly increasing and expanding extensive flood risk. Improved disaster reporting and an urban bias in the data ? There have been major improvements in the reporting of disaster events in recent years, due to improved communications, the development of decentralised institutional systems in many countries and the 14 introduction of the internet. In some disaster databases, such as Colombia, systematic government reporting contributed additional reports since the 1990s. We believe, however, that while improved reporting certainly will have lead to an increased number of reports, it is not sufficient to explain the trend, for a number of reasons. Improved reporting could be expected to lead to more reports across all hydrometeorological hazards and not just floods and rains. The contribution of new data sources would increase the average number of annual reports but not exponentially. At the same time, extensive flood risk is increasing not only in remote rural areas, where disasters may have gone unreported, but also in major metropolitan areas, such as Buenos Aires or Mexico, where reporting deficits are unlikely. In other words, while disaster reporting has improved since the 1970s, it, alone, does not convincingly explain the trend. At least in some countries, such as Peru and Colombia, many more disasters are reported from the major cities and from provincial capitals than from isolated rural areas18. This at least partly reflects a bias in the reporting, given that most databases are compiled from national sources. As Box 3.1 indicates local databases have generally registered many times more disasters than national databases, however, most of the additional reports are of very small disasters that are not visibly nationally. All the disasters reported from important areas certainly exist, while it is unlikely that nationally significant disasters from remote rural areas have gone unreported. While the bias certainly exists, again it is not sufficient to explain the trend. Increased precipitation ? Another possible explanation could be increases in precipitation levels or their concentration in more intense precipitation events. In some countries such as Colombia, Costa Rica, Ecuador and Venezuela the increase in flood and rain related disasters since the mid 1990s has coincided with a period of increased precipitation, as Figure 0.13 and Figure 0.14 highlights illustrates in the case of Costa Rica.19 Figure 0.13: Flood and rain disasters in Costa Rica (19902007) Figure 0.14: Precipitation Costa Rica 10 years averages (1957-2007) Again, however, while increases in precipitation certainly would increase flood hazard the data available does not validate the hypothesis. In Mexico and Nepal, for example, extensive flood risk is increasing, while average precipitation is decreasing. In the case of Nepal this may reflect increased glacier melt. In Peru, in contrast, average precipitation is increasing while extensive flood risk is decreasing. There is no consistent relationships between both variables, therefore, across the sample. Even in the countries where precipitation is increasing in the 1990s, this may be a short term trend within a longer term cycle. At the same time, a national average increase in precipitation may hide drastic local 18 Reference GRADE paper and Fernando Ramirez from Part II background papers Global Precipitation Analysis Products of the Global Precipitation Climatology Centre (GPCC), (2008). Full Data Reanalysis Product Version 4. 19 15 variations even in small countries like Costa Rica, implying that meaningful correlations between increased flooding and precipitation could only be explored with local level precipitation data, which had not been possible in the scope of this report. Similarly, average annual increases in precipitation do not reflect changes in seasonal distribution and intensity, which can have a dramatic influence on flooding. Finally, even if an increase in precipitation could be proved and a causal relationship with more intense flood disasters established, while this would explain the number of disasters and their impact it would not account for the drastic territorial expansion of risk highlighted in section 3.3. Urbanization, environment and territorial occupation ? Aggregated national statistics on urbanization and environment also provide little insight into the trend of increasing and expanding extensive flood risk. National level indicators on deforestation indicate that most countries have continued to lose forest cover at a rapid rate in the last decade and there is a relationship between declining regulatory services from ecosystems and increased flooding. However, as in the case of precipitation, it is impossible to infer from national averages whether this has contributed to increased extensive flood risk20 in specific watersheds. For example, in Costa Rica environmental protection policies and the application of a system of payment for ecosystem services have led to an increase in national forest cover at the same time as flood risk is increasing. An analysis of data in eacah watershed would be required to establish a relationship. Similarly as Table 0.321 indicates, all countries have been experiencing progressively slower urban growth rates over recent decades. As with precipitation, national averages of urban growth, therefore do not explain the trend. Table 0.3: Urban growth rates in % Accumulated Percentage increase in Urban Population Argentina Bolivia Colombia Costa Rica Ecuador India Orissa Tamil Nadu Iran Mexico Nepal Peru Sri Lanka Venezuela Average 78-87 22 54 38 49 56 41 36 18 69 35 88 38 7 44 40 88-97 19 44 29 55 37 34 32 34 39 26 91 25 2 33 30 98-07 14 32 21 34 25 27 22 16 75 15 -2 28 21 It is impossible, therefore, to easily explain the trend through an examination of national data. In contrast, case studies from different countries, point to a complex combination of interlocked factors related to changes in territorial occupation, declines in the regulatory services of ecosystems, models of urban growth and both rural and urban poverty as being factors that explain the trend, as well as its specific national and regional manifestations. Latin America In Latin America, more than 80% of the total population is now “urban” although definitions of what constitutes urban and rural vary from country to country. Extensive risk in Latin America, therefore, is essentially an urban phenomenon, with loss patterns in the region broadly reflecting the proportion of urban and rural population as shown in Table 0.4. 20 21 For national level data on deforestation rates see: http://faostat.fao.org/site/377/DesktopDefault.aspx?PageID=377#ancor Department of Economic and Social Affairs Population Division, 2008, World Urbanisation Prospects: The 2007 Revision. 16 Table 0.4: Urban and rural distributionof extensive risk in Peru and Colombia [CHANGE TABLE HEADINGS] Table 0.5: I am asking Cristina since the 15 October in 5 spererate emails to send me the talbe, but she refuses to send an excel version. Only what I have here, and another one with Mexico in Power Point! Rangos de población < 100 100-500 500-1000 1000-5000 > 5000 Totales Población 113 606 846 50 735 724 20 658 731 37 989 536 46 414 503 269 405 340 % 42 19 8 14 17 100 Registros 42149 9516 5978 4260 6330 68 233 % Muertos 1397 62 14 9 6 9 100 4142 1302 3137 12 490 22 468 % 6 18 6 14 56 100 Viviendas destruidas 39 151 29 772 5390 14 330 61 282 149 925 % 26 20 4 10 41 100 Viviendas afectadas 1 510 855 368 756 130 724 73 245 224 492 2 308 072 % 65 16 6 3 10 100 As Table 0.5 compares the number of extensive risk disasters and their attributes in eight countries in Latin America with the population size of the affected administrative areas. Areas with a population inferior to 500,000 contain 61% of the total population. These administrative areas include most of the rural population as well as many small and medium urban centres. They account for 76% of the extensive risk disasters and 81% of the housing damage but only 24% of the deaths and 46% of the housing destruction. At the other extreme, administrative areas with populations that exceed 1 million, including all the major cities in the region, account for 31% of the population,15% of the extensive risk disasters and 13% of the housing damage but 70% of the deaths and 51% of the housing destruction. This distribution of extensive disaster risk suggests two different but complementary processes of urbanization and territorial occupation operating simultaneously: The large number of extensive risk disasters and housing damage occurring in rural areas and small urban centres is coherent with the geographical expansion of extensive risk described above. As can be clearly visualized in the case studies of Ecuador and Peru presented below, extensive risk follows processes of territorial occupation, characterised by the opening or improvement of roads, the extension of the agricultural frontier and the growth of small and medium urban centres. Increasing extensive risk is usually due to a concatenation of factors including a decline in the regulatory services provided by ecosystems, inadequate water management, land-use changes, rural-urban migration, unplanned urban growth which increases run-off, the expansion of informal settlements in low-lying areas and an under-investment in urban drainage. The reduced number of disasters and housing damage but the relatively high proportion of mortality and housing destruction, occurring in large and mega-cities, indicates a process of intensification of risk over time. As the case studies of Buenos Aires, Mexico and Costa Rica show extensive flood risk is closely linked to the increased run-off caused by new urban 17 development, a chronic underinvestment in city-wide pluvial drainage, the location of informal settlements and social housing projects in low-lying flood prone areas and inadequate water management in the surrounding watersheds. As cities develop and consolidate, however, many localized extensive risks are gradually reduced, as drains are installed, houses are improved and flood protection works are carried out. However, risk becomes more intensive over time as the underlying city-wide causes of risk are not addressed. When disasters do happen, therefore, they tend to be more destructive with high rates of mortality and housing destruction. The cases of Tabasco, Mexico and the Federal District of Mexico itself illustrate this process in the contexts of a mega-city and of a large city in a complex hydrographic basin Box 3.x Flooding in Tabasco, Mexico, 2007: In 2007, 62% of the area of the state of Tabasco in Mexico and 40% of the city of Villahermosa were affected by the most severe floods in the history of the state. At least 1.2 million people were affected, 168,000 houses damaged, 570,000 hectares of crops lost as well as major damages to infrastructure and the economy. The exceptionally high rainfall that contributed to the flooding meant that climate change and variability were invoked as the principal culprit of the disaster. In reality, the causes were a combination of inadequate water management in a very complex watershed that covers 89,000 km2 and a process of urbanisation of low-lying areas without a corresponding investment in drainage. The canalisation of rivers in the upper basin, through a series of isolated and uncoordinated civil engineering works, has tended to increase flooding in urban areas built on the floodplains of the Grijalva and Carrizal rivers further downstream. Four of the six most devastating floods that have occurred in Villahermosa in the last 95 years have occurred since 1980. [IF WE HAVE SPACE INLCUDE TWO MAPS OF POP. DENSITY AND FLOODED AREAS FROM UNOSAT http://unosat.web.cern.ch/unosat/freeproducts/mexico/Tabasco_Inundations_Nov2007/UNOSAT_pop_dens_Tabasco_Nov200 7_A3_Spanish_highres.jpeg http://unosat.web.cern.ch/unosat/freeproducts/mexico/Tabasco_Inundations_Nov2007/UNOSAT_Tabasco_Nov2007_A3_Spa nish_highres.jpeg Floods in Mexico: The area now occupied by Mexico’s mega-capital was originally a series of lakes in a large basin surrounded by high mountains. Tenochtitlan, capital of the Aztecs, was founded on an island in these lakes and after the Spanish conquest became the capital of Nueva Espana and later Mexico city. Given the geography and hydrology of the basin floods occurred both in the pre-hispanic period and with increasing intensity in the colonial period. One flood between 1629 and 1635 may have caused 30,000 deaths. The problem was addressed by the government of Porfirio Diaz which drained the basin through the construction of a 50km long canal and a tunnel through the mountains of Tequixquiac allowing the waters of the basin to drain into the Gulf of Mexico. While this monumental example of civil engineering initially greatly reduced flooding, the exponential growth of the city and it’s subsidence due to the drainage and urbanisation of wetlands have progressively increased hazard levels and increased the frequency of urban flooding. Further engineering works, such as the construction of a deep drainage system between 1967 and 1975, have partially addressed the problem. However, the subsidence has reduced the gradient of the canal of Porfirio Diaz and even threatens to invert it, which would mean even more destructive floods in the future. As will be explored in more depth in the next chapter, cities with poor urban governance and which have only been able to absorb growth through the expansion of informal settlements often tend to have the highest levels of risk. In Mexico, for example, almost 20% of the disasters occurred between 1980 and 2007 were concentrated in only 22 municipalities. 7 of these municipalities have more than 1 million inhabitants, 8 with between 500,000 and 1 million inhabitants and 7 with between 200,000 and 500,000 inhabitants. Most of the most affected municipalities, had high percentages of their population with high or very high levels of marginality, for example: Acapulco (54.4%), Coatzocoalcos (54.1%) (Tijuana (31.3%), Juarez (45%), Veracruz (31%) or Tapachula (54.1%)22. In the metropolitan areas of San Salvador, both disaster incidence and housing damage and destruction were found to be closely correlated with urban growth. The municipalities with recurrent extensive disasters were those with the most rapid urban growth23, in some cases up to 16% per year. Argentina In Argentina, floods and rains account for approximately 45% of extensive hydrometeorological disasters. At the national level, unlike in other countries in the sample, the annual average number of events is actually declining, along with mortality, housing destruction and housing damage. However, no less than 32% of the disasters occur in the Province of Buenos Aires or in the Federal Capital. In contrast to the country as a whole, extensive flood and rain risk in the urban area has increased between 1970 and 2007 as illustrated in Figure 0.15. 22 23 Mansilla, Mexico document #2 Mansilla, El Salvador document. 18 Figure 0.15: Extensive flood and heavy rain disasters in the province of Buenos Aires and the Federal Capital (1970-2007) This increase in extensive flood risk is closely related to the increased run-off caused by the paving of the land area and the reduction in woodland and parks, as well as by underinvestment in drainage. Whereas in the 1970’s the average annual investment in urban infrastructure in Buenos Aires was US$350 million per year, by 1989 this figure had fallen to US$69.9 million. The city’s drainage system was completed in 1953 for a calculated population of 800,000. The city proper of Buenos Aires had a population of 2,776,138 in 2001, while the Greater Buenos Aires area had a population of 11.4 million people. The greatest impact of the floods is in neighbourhoods such as La Boca and Barracas, which were traditionally inhabited by lowincome groups.24 Costa Rica In Costa Rica, floods, heavy rains and flash floods accounted for 63% of the extensive risk disasters reported since 1996 and 83% of the housing damage. Since 1996, the number of disasters has increased 10 fold, while the number of houses damaged has increased more than 3 fold. In contrast, the number of deaths and destroyed houses have actually decreased, reflecting relatively efficient emergency management and an overall improvement in housing conditions, except in the case of informal precario settlements. According to the Municipality of San Jose, more than 80% of the floods occurring in the countries capital are caused by either inadequate drainage to cope with the increased run-off caused by urban growth or by the accumulation of garbage and waste in drainage channels. Most housing damage is concentrated in precarios occupying marginal land adjacent to the streams and torrents that drain the city. Some cities such as Turrialba have historically prohibited construction within 15 metres of a river bank. This kind of urban zoning paradoxically guides formal urban development to safe areas, while leaving hazard prone areas for illegal or informal occupation by poor households. According to the World Bank25, Costa Rica has one of the lowest poverty headcounts in Latin America. Using a US$2 per day poverty line, Costa Rica has a headcount of only 9% compared to a regional average of 25%. Key socio-economic indicators are considerably better than the average in upper-middle income countries. However, since the mid 1990’s, while the country experienced relatively consistent economic growth, income inequality rose, per capita household income of the poor fell and the decline in poverty experienced in previous decades stagnated. In the same period, the number of extremely poor people living in precarios has almost doubled to 8% of the population. This period coincides with the increase in flood events and damage to housing. The rapid increase in the number of extensive flood risk in Costa Rica is linked therefore to urban growth, inadequate water management, a lack of investment in drainage and an increase in precario housing. The latter is a symptom of the lack of progress made in addressing poverty and inequality over the last decade. Cali, Colombia Box. 3.x Floods and informal settlements in Cali, Colombia. The city experienced 179 extensive flood disasters between 1970 and 2007, representing 53% of the extensive hydrometeorological disasters and accounting for 95.6% of the housing damage. Between 1950 and 2000, risk patterns arose in different areas of the city as informal settlements were created through 24 Herzer, Hilda and Clichevsky, Nora, 2000, Floods in Buenos Aires, Learning from the Past, in World Bank, 2000, Managing Disaster Risk in Emerging Economies, Washington DC. 25 World Bank, 2007, Costa Rica Poverty Assessment: Recapturing Momentum for Poverty Reduction. http://wwwwds.worldbank.org/external/default/WDSContentServer/WDSP/IB/2007/05/23/000020953_20070523093115/Rendered/PDF/3 59100CR.pdf 19 land invasions by poor families and with inadequate drainage. In the 1950s and 1960s flood disasters were concentrated in informal settlements along the flood plains of the Cali and Cauca rivers, with inadequate or non-existent storm drains. In the 1970s, the focus of flood disasters shifted to the right bank of the river Cali and the flood plain of the river Canaveralejo. Again, the incapacity of the drainage system to deal with the increased run-off posed by urban growth was responsible for the flooding. In the 1980s and 1990s, the focus of flood disasters shifted to Aguablanca, an area which had been subject to major investments in hydraulic infrastructure to habilitate 640 hectares of land for intensive agriculture. In the 1980s this area was occupied by informal settlements and by social interest housing promoted by the city government. 29 neighbourhoods and 24,000 houses sprung up on an area intended for agriculture and where the construction of drainage infrastructure was costly and difficult. The draining of lakes, intended to regulate flood waters, the leveling of dykes for urbanization and the obstruction of natural drainage channels with garbage were additional causes of flooding in newly urbanized areas in the east of the city since the 1990’s. Ecuador In Ecuador, 96.1% of disaster reports are extensive, accounting for 67.9% of the mortality, 75.6% of the destroyed housing and 90.3% of the damaged housing. Even the impacts of major climate events such as the 1997-1998 ENSO were extensive in Ecuador, given that they were spread over large areas of the country. The spatial accumulation of extensive hydrometeorological risk is closely associated with the evolution of territorial occupation and urbanization in the country. Map 0.10 shows the spatial evolution of extensive hydrometeorological disasters since 1970. In the 1970’s, disasters were concentrated in the metropolitan areas of Quito and Guayaquil. In successive decades, extensive hydrometeorological risk spread first along the country’s Andean backbone, then into the tropical lowlands of the Pacific Coast and finally into the Amazon basin, closely reflecting the dynamics of territorial occupation in the country. Map 0.10: Spatial evolution of extensive hydrometeorological disasters in Ecuador from 1970-2007 20 Peru Extensive hydrometeorological risk in Peru is characterized by landslides and mudslides. Traditionally these hazards occurred in the coastal valleys on the western slopes of the Andes. Over the last two decades there has been a drastic redistribution of hazard patterns as the opening of roads down the forested eastern slopes of the Andes and into the Amazon basin led to the clearing of forest for agricultural purposes, including coca cultivation, and rapid urban growth. Given the topography and land cover of the area, this process of territorial occupation was accompanied by a rapid growth in the number of landslides and mudslides. Error! Reference source not found. shows the redistribution of extensive risk associated with these hazards between 19701985 and 1986-2006. In the first period, risk was concentrated in the coastal valleys, in the Andes and in areas such as the province of La Convention in Cuzco, an early area of territorial occupation in the Amazon. In the second period, concentrations of risk appear in the central jungle, along the upper Huallaga Valley and in the northern Amazon, closely mirroring the process of territorial occupation. Asia The interpretation of the trend in Asia poses a very different set of challenges. Firstly, the five databases from India (Orissa and Tamil Nadu), Iran, Nepal and Sri Lanka are not a representative sample of Asia as a region. They provide a north-south cross section of south Asia, and a glimpse of west Asia. East, southeast and central Asia are unrepresented. Secondly, the compilation of the disaster databases is far more recent. The Latin America databases were first compiled in 1995 and subsequently have been subject to numerous updates, revisions and analysis. The Asian databases in contrast were initiated between 2003 and 2006. 2003. Only Sri Lanka has published a summary of the disaster risk patterns and trends26, although a similar publication is underway in Tamil Nadu. Given that the database themselves are so 26 Reference Historical disaster database in Sri Lanka publication. 21 recent it is not surprising that current analysis of disaster data in Asian countries focuses principally on describing the main spatial patterns and temporal trends of the different risks and of hazards. Very little research would seem to have taken place into the role that underlying risk factors, such as urbanisation and rural livelihoods, play in explaining the risk trends and patterns manifest in the disaster databases. Given these limitations, our interpretation of the trends in extensive risk in Asia is less comprehensive and more incipient than in Latin America. The rural-urban contrast In contrast to Latin America, all of the Asian countries, except for Iran, are still markedly rural. Iran had 66.9% urban population in 2005. In contrast, Nepal had 15.8%, Sri Lanka 15.1% and India 28.7%. At the same time, however, there is evidence that cities in Asia are experiencing the problems of increasing damage due to urban floods, for the same reasons as in Latin America. For example, Colombo experienced 240 flood disasters since 1974. Almost half the reported disasters and about 80% of the reported housing damage has occurred since 2005. Chennai, similarly experienced major flooding in 1990, 1994 and 1996 and Kathmandu in 2000 and 2002. Flooding in these, as in other cities in south Asia, would tend to suggest that in urban areas, many of the same factors that are driving risk in Latin America, such as rapid urban growth, the expansion of informal settlements, inadequate water management and an underinvestment in drainage. In many cities in the region, given higher levels of poverty and a greater proportion of the urban population living in unserviced informal settlements and slums than in Latin America, it is possible that the problem may be even more severe. Elsewhere, however, flooding has very different causality and manifestations. In Iran, flooding would appear to be mainly associated with torrential flash flooding as the snows melt in spring. In Nepal, Orissa and Tamil Nadu, flood disasters are also associated with major concentrations of rural population living on the floodplains and deltas of major rivers. Environmental degradation, in particular declines in the regulatory services provided by forest ecosystems, may be contributing to increased flooding in some watersheds. In Nepal, increased flooding in the south-eastern Terai region may be related to increased glacier melt in the Himalaya as well as to environmental changes in upland watersheds. In both Indian states as well as Sri Lanka and Nepal, housing damage in rural areas would seem to be closely associated with the high density of rural settlement in flood prone areas and the vulnerability of rural housing. Box 3.x Iran: The number of people killed and houses destroyed and damaged in extensive flood disasters in Iran is minimum compared to the losses in major earthquakes. Extensive risk floods are responsible for less than 2% of the total deaths over the reporting period and thus have tended not to attract the attention of policy-makers in the country. Nevertheless, there were a total of 2,481 extensive flood disasters over the reporting period. While as Figure 3.x shows the number of disasters is trending up as in other countries and regions, most of the reports are concentrated in a few peak years, for example 1982, 1987, 1993, 1995 and 1998, 2001 and 2007. Most flood events in Iran occur from March to May and are torrential floods related to winter snow melt. Unlike other countries, the number of damaged houses is trending down which may indicate either decreasing exposure due to migration to cities or else improvements in housing. As in other countries, the number of areas affected by extensive flood disasters has increased over the reporting period. In general geographic distribution of flooding, is controlled by the countries topography and associated with the mountain ranges and fertile valleys in the east and south-east, north and south-west of the country. STEPHANE: INSERT FIGURE OF TEMPORAL EVENT DISTRIBUTION WITH REGRESSION OF EXT. FLOOD/ FLASH FLOOD/HEAVY RAIN EVENTS + MPA OF SPATIAL DISTRIBUTION Box 3.x Orissa: Orissa is one of the most least developed states in India with a per capita Net State Domestic Product about two thirds of the national average, as low HDi plcing it 11 th out of 15 Indian states, and a rural poverty head count approaching 50%. The economy is primarily rural with 64% of the population working in the agricultural sector. A large part of the mortality and housing damage reported in Orissa occurred in the 1999 super-cyclone and a large percentage of the housing destruction, as mentioned is due to rural fires. Extensive riverine and coastal flood disaster reports are concentrated along the eastern flood-plain, both in the northern districts of Bhadrak, Kendrapara and Jajapur, which have low-incomes, low urbanisation and a high proportion of scheduled tribes as well as in more urbanised Districts to the south, in the floodplain and delta of the Mahanadi river. Factors, which could be contributing to both increased extensive flood disaster risk as well as drought in Orissa are the lack of access to and control over land and common property resources like water, forest and 22 common land, land degradation, the decline in regulatory ecosystem services provided by forests, and ineffective water management. The distribution of mortality and affected people, however, highlights the extreme vulnerability of southern Orissa. The Districts of Kalahandi, Bolangir and Koraput are characterised by repeated droughts, floods, extreme poverty, food insecurity and chronic income poverty and localised near famine conditions. While the distribution of extensive risk disasters is skewed towards the central eastern coastal region, the distribution of mortality is skewed towards the extremely poor rural Districts mentioned above. In other words, while flood incidence would seem to be associated with higher levels of urbanisation and with relatively affluent agricultural areas on floodplains and deltas, mortality is far higher in the poorest most food insecure inland areas. INSERT MAP OF DISTRIBUTION OF EXT. RISK REPORTS + EXT. RISK DEATHS AT THE BLOCK LEVEL Box 3.x Sri Lanka. In Sri Lanka the number of houses damaged in extensive flood disasters has approximately doubled since the early 1990s. while the number of disasters reported has also increased. As Map 3.x shows, there are two major concentrations of extensive flood risk in Sri Lanka. There is a major concentration of housing damage due to flooding in south-western Sri Lanka, which is relatively highly urbanised, with associated issues of inadequate drainage. For example, flooding reported in early November 2006 in Colombo destroyed 221 houses, damaged 1,674 houses and affected 80,128 people. Map 3.x (http://unosat.web.cern.ch/unosat/freeproducts/srilanka/UNOSAT_Sri_Lanka_radarsat_Colombo_v1.0_highres.jpeg) shows the distribution of flooded areas while Map 3.x shows http://unosat.web.cern.ch/unosat/freeproducts/srilanka/UNOSAT_Sri_Lanka_popden_v1.0_highres.jpeg the population density. Floods were concentrated in urban areas indicating typical problems of settlement of low-lying areas and in adequate drainage. [ASK FRANCESCO IF THEY CAAN OVERLAY BOTH IMAGES FOR US] In eastern Sri Lanka, there is another concentration of extensive flood risk primarily associated with monsoon rains in the lowlying flood plains in north-central and eastern Sri Lanka, for example around Polonnaruwa. Tamil Nadu: Tamil Nadu is India’s most urbanized large state, with 44% of its population living in urban areas and a diversified economy. On the Human Development Index it is third among India’s 15 largest states and has been successful in reducing poverty. The rural poverty head count ration declined from 32 to 23 % between 1993 and 2004., while the urban head count ratio fell even faster, from 40 to 22%. However, the state has high differential vulnerability, with Scheduled Castes having twice the mean poverty head count in urban and 50% higher in rural areas. Extensive risk is spread throughout the state, including droughts, floods, fires and epidemics. While most of the mortality was concentrated in the 2004 tsunami, housing damage and destruction is much more distributed in extensive risk disasters. Extensive flood risk, as elsewhere in south Asia would seem to present two different patterns. As Map 3.x shows [[INSERT MAP SHOWING DISTRIBUTION OF EXT. RISK FLOOD /RAIN HOUSING DAMAGE ] there is a concentration of extensive risk housing damage around urban centres in the north-east of the state. While poorly studied, flooding would seem to be associated with the high level of urbanization, and associated problems of settlement of low-lying areas, increased run-off and inadequate drainage. In Chennai, for example, 17.7 % of the urban population was living in slums in 2001 27 The other concentration of housing damage is along the floodplain of the Kaveri river, in the watershed of the Ponnaiyar river and in the Kambam valley. Tamil Nadu has a long history of chain tank irrigation , which has fallen into disuse. Much flood damage in rural areas is associated with chain-tank failure and the silting of irrigation channels. In upland areas, increasing extensive flood risk may be associated with issues of environmental degradation. Box 3.x Nepal: In Nepal, extensive risk flood disaster and associated housing damage is concentrated in the densely populated alluvial plains of the Terai in south-east Nepal See Map3.x. Floods in this case are not a consequence of urbanisation or population density, but a cause. The richness of the soils is a consequence of the frequent replenishment of nutrients through flooding and the reason why the region can support such a large dense population. Other extensive risks, associated with landslides for example are spread across the mountainous centre, north and west of the country and are responsible for more mortality than floods. [INSERT MAP OF EXT FLOOD HOUSING DAMAGE BY DISTRICT] Africa Systematic historical disaster data is not readily available in African countries. In South Africa, a detailed database has been developed and maintained in Capetown28. A national database in now being compiled in Mozambique. However, these are still exceptions in the region. Given this lack of data, it is not possible to explore extensive risk patterns and trends in the same way as in Latin America and Asia. Nevertheless, a series of case studies from African cities29 highlight that most 27 Maiti, Sutapa, Agrawal, Praween, 2005, Environmental Degradation in the context of growing urbanisation: a focus on the metropolitan cities of India, Journal of Human Ecology, 17(4): 277-287 28 Reference MANDISA from DiMP 23 of the underlying factors of extensive flood risk in Latin American and Asian cities are also manifest in Africa and that, if anything, these manifestations are more extreme. Four types of flood have been identified in African cities: (1) localised flooding due to inadequate drainage (2) flooding from streams whose catchment is entirely within the urban areas (3) flooding from major rivers on whose banks the cities and towns are located and 94) coastal flooding from the sea or by a combination of high tides and river flows. According to Action Aid, the first two kinds of flood are most prevalent. The underlying causes are very rapid urban growth is fuelled by migration from rural areas, which themselves suffer from the combined impacts of flood, drought, conflict, HIV/AIDS and other hazards on the vulnerable, stressed livelihoods of poor rural households. Due to poor urban governance most cities absorb growth through the expansion of informal settlements, which often occupy low-lying flood prone areas. 72% of Africa’s urban population lives in such settlements, meaning that floods predominantly affect the urban poor. Investment in drainage infrastructure is often non-existent while little or no maintenance of any existing drainage infrastructure is carried out. While climate change and sea level rise is already magnifying flood hazard levels, the principal causes are related to the urbanisation process itself. [INSERT CITY EXAMPLES FROM ACTION AID PUBLICATION ON SPECIFIC CITIES + CONSIDER MOVING AURAN ST. LOUIS BOX FROM CHAPTER 4] 3.7 Poverty and disaster risk: the empirical evidence [NEED TO REFRENCE ALL THIS SECTION] The evidence presented in this chapter so far tends to validate the hypothesis that both the impact and incidence of disasters at the national and local level tends to be disproportionately concentrated amongst the poor. Across Latin America, in parts of Asia and in African cities, urban informal settlements are certainly experiencing rapidly increasing flood disasters, with significant patterns of housing damage. There is also an expansion of extensive risk in small towns and rural areas, associated with processes of territorial occupation and a decline in the regulatory services provided by ecosystems. While it is less clear which social groups are affected by such disasters, it is unlikely that the housing damage reported corresponds to high-income groups. Given that intensive risk is overlaid on patterns of extensive risk, the impacts of major disasters will also disproportionately affect the same social groups affected by extensive risk disasters. While a clear relationship between disaster risk and poverty can be inferred from the evidence presented above it is far less easy to demonstrate empirically for a number of reasons: Empirical studies tend to be opportunistic rather than systematic, given the limited availability of both disaster and poverty data at a suitable scale. Spatially aggregated data may smooth important local differences in disaster risk and poverty making it difficult or impossible to identify relationships Disaster incidence and loss may be highly correlated with hazard intensity as well as with poverty. Information on the distribution and intensity of localized hazards is often not available, making it difficult to control for what may be a key variable. As mentioned in 3.6 national disaster databases are biased towards major cities and towards local political capitals and may underestimate disaster risk in poor rural areas. When disaster risk is compared correlated nationally with data on the poverty headcount, poverty gap or unsatisfied basic needs, it may produce counter-intuitive results in countries with predominantly urban populations, given relative differences between urban and rural poverty. The case studies referred to in this chapter are drawn from publications by Action Aid (unjust waters – reference) and by AURAN. 29 24 Despite these challenges, the case studies from Latin America and Asia commissioned for this report, together with the results from other research in Africa do provide empirical insight into the relationship between disaster risk and poverty. Many of these empirical studies address the issue of drought and it’s impact on rural livelihoods, thus balancing the analysis presented in the first part of this chapter, which focused particularly on the impact of floods on the housing sector. 3.7.1 The poor are more at risk A number of the case studies provide empirical evidence that the incidence and impact of disasters is greater in areas inhabited by poor households, although the studies do not show whether this was due to greater exposure to hazard, due to physical vulnerability or due to the magnification of hazard levels. Additionally, the studies do not take into account other less quantifiable variables such as social capital, which often plays a significant role in reducing risk30. In Mexico, the number of destroyed and damaged houses between 1980 and 2007 was compared with the total housing stock in each municipality and with an Index of Municipal Marginality developed by the national population council. It was found that in absolute terms most houses were affected in urban municipalities with a relatively low level of marginality. 69% of the houses lost occurred in municipalities with very low levels of marginality and a further 12.7% in municipalities with a low level of marginality, in general in large cities. However, when compared with the total housing stock, disaster losses affected an average of only 8% of the housing stock across these municipalities. In contrast, while in absolute terms, only 18% of the destroyed and damaged houses occurred in municipalities with high or very levels of marginality, these destroyed a far higher percentage of the housing stock. In a third of these municipalities less than 10% of the housing stock was affected over the 27 year period. In contrast, in another third, between 10 and 25% of the housing stock was lost, while in another third more than 25%. Over 20% of the municipalities with high and very high levels of marginality had more than 50% of their housing stock affected. Most of these municipalities contained a mixture of rural areas and small towns. This indicates that while disaster risk in absolute terms may be principally urban, it’s impact may be far greater in marginal rural areas and small towns. In Peru it was found that rural households that reported a disaster impact in 2002 had on average less access to public services, were less well integrated into the market and had a higher proportion of agricultural income. It was also shown that households that experienced disaster were between 2.3 and 4.8 times more likely to be always poor than never poor. Those rural families with more livestock holdings had slightly less probability of being always poor, which indicates that livestock is important as a buffer against lost income. In Bolivia, it was found that at the municipal level, factors such as improved education, immigration, declining child mortality (associated with improving living conditions) and a decrease in unsatisfied basic needs all contributed to a reduction in monetary poverty. However, in municipalities exposed to natural hazards, the effectiveness of these variables in reducing poverty declines by about 50%. In Bolivia, poverty was least 3% higher in municipalities exposed to hazard. Similarly, in Ecuador it was found that drought exposure led to an increase in poverty incidence by 2% approximately. In Nepal areas affected by floods tended to have lower poverty rates and higher per capita expenditures while those affected by epidemics and landslides tended to have higher poverty rates and lower per capita expenditure rates. As described in 3.6, flooding incidence and impacts are concentrated in the highlyproductive lowland agricultural plains of the Terai in south-eastern Nepal. As flooding contributes to the fertility of the soil of the region, it contributes to the wealth of the area. In contrast, epidemics are heavily concentrated in districts in western Nepal and in mountainous western Nepal, which concentrate the 30 Yamamura, Eiji ……… see list of papers consulted for second part. 25 countries rural poverty. Poverty rates are significantly higher in the mid-western (44.8%) and far western (41%) of the country than in the central (27.1%) and eastern regions (29.3%), where the Terai is located. In urban areas, the municipalities with most disaster affected people had a significantly higher poverty rate and poverty gap. In Orissa, India the absence of disaggregated data on household expenditure, assets and income meanst that it was not possible empirically to explore correlations between disaster incidence and loss and poverty. However, as in Nepal the areas most subject to floods, along the coast and floodplains and deltas of major rivers are relatively wealthy compared to extremely poor southern Orissa. It is unlikely, therefore, that there would be a clear relationship between poverty and flooding. The poorest areas of Orissa in the south are drought prone, but as already stressed, the disaster database does not provide adequate data on drought impacts on livelihoods. A statistically significant relationship, however, was found between households living in houses with earth walls and thatch roofs and those most affected by cyclone, flood, fire and lightening. In the case of Sri Lanka, a very strong correlation was found between population living below the poverty line and the number of houses damaged due to floods and a less strong but significant correlation between population living below the poverty line and houses damaged due to landslides. As in the case of Mexico, mentioned above this does not necessarily imply that there is a high incidence of floods in the areas where the poor live but it does confirm that unsafe, vulnerable housing is a poverty factor that conditions the likelihood of suffering greater loss in a disaster. In Tamil Nadu, as in Orissa, the absence of disaggregated data on household expenditure, assets and income made it difficult to explore disaster risk poverty interactions. However, the use of poverty proxies demonstrated a strong relationship between the vulnerability of housing and disaster loss. As in Sri Lanka, this confirms that unsafe, vulnerable housing is a poverty factor that conditions the likelihood of suffering greater loss in a disaster. For example, mortality in extensive flood disasters was highly correlated with vulnerable housing. Similarly, cyclone housing damage was inversely related to the literacy rate. If literacy is taken to be a proxy for poverty again this indicates that the poor were more likely to suffer housing damage. Mortality amongst scheduled castes was also higher in Blocks with a high proportion of vulnerable housing. 3.6.2 Disaster losses increase poverty The empirical evidence also confirms that disaster impacts, associated with both intensive and extensive risks, have a direct and negative impact on the welfare at both the household as well as local and regional levels. Impacts may include an immediate increase in monetary poverty, both in terms of its depth and breadth, as well as a deterioration in other welfare indicators. If as examined in 3.6.1, the poor are more likely to experience disaster incidence and loss, they are therefore at the same time less able to absorb the impact of the loss and recover. In hazard prone areas, therefore, the recurrent losses due to disasters is a central component of the experience of poverty. In Mexico, municipalities that experienced disasters between 2000 and 2005 experienced a 0.8% drop in the municipal-level Human Development Index (HDI), a 3.6% increase in food poverty a 3% increase in capacities poverty and a 1.5% increase in assets poverty. Drought increased these different poverty indicators more than floods. Droughts led to a decrease of 1.2% in the municipal HDI, increased food poverty by 4.1%, capacities poverty in 3.7% and assets poverty in 2.5%. In Iran, only a minor percentage of the population lives below the World Bank’s US $1 per day poverty line. This percentage fell from 3.5% in 1991 to 0.3% in 2004. Other human development indicators also steadily improved over the period, including life expectancy, education and health. In Iran, unlike any of the other countries surveyed in this chapter, intensive earthquake disasters, affecting entire provinces account for the vast majority of mortality (95%) and houses destroyed (73%), while 69% of the population is urban. Given this pattern of major earthquakes affected entire provinces and their urban 26 centres, it was therefore possible to identify correlations between high disaster risk and decreased household expenditure a high level of spatial aggregation, at the provincial level. Those provinces (Ardebil, Chaharmahal-o-Bakhtiari, Lorestan, Mazandaran, Khorasan, Hormozgan and Yazd )which had experiences the largest number of deaths and houses destroyed per 10,000 inhabitants had also experienced the greatest negative effect on the expenditures of urban households. However, in some provinces with major housing destruction, such as Mazandaran, Lorestan and Khorasan, household expenditure actually increased, probably due to the positive impact of financial assistance from government or charities. In El Salvador, the 2001 earthquakes killed more than 1,200 people, affected approximately 300,000 houses (or 32% of the stock) and caused US $1.6 billion in direct and indirect damage (12% of GDP in 2000). Between 2000 and 2002, average household income per capita was actually increasing in El Salvador (from 5449 to 6957 colones) while extreme poverty rates had fallen from 33.8% to 26.6%. In poor rural households, affected by the earthquakes, average household income per capita was reduced by approximately one third. Those most affected by the earthquakes suffered higher loss of housing, land, livestock, farm machinery and other physical capital, which reduces the future earning capacity. In the city of Trinidad, following a major flood in 2006, the level of poverty increased more than five times the national increment while the poverty gap and the severity of poverty also increased. The impact of education on reducing poverty also fell from 8.7% in 2006 to 6.6% in 2007 3.6.3 Disaster loss disproportionately affects the most poor and vulnerable The ability of households to access, use and transform both tangible and intangible assets, such as reserves and savings, an extended family, social safety nets provided by governments, land and property and skills and knowledge influences their income-generating potential and welfare in normal times. In disaster situations, asset holdings offer a crucial means to buffer impacts and to recover and therefore, condition their level of resilience or vulnerability to hazards. Households susceptible to experience hazard impacts and disaster losses, typically suffer a reduction in return to assets, leading to a collapse in income and a loss in assets themselves. The most vulnerable, who are less able to access, use and transform assets in normal times are the most affected by disaster. Disasters therefore not only increase poverty, they increase inequality. In Peru, it was found that the occurrence of disasters between 2002 and 2006 had a drastic effect on households monthly per capita consumption after 2006. This impact was significantly greater for poor families, whose consumption was reduced by 2.6 % than for wealthier families, whose consumption was reduced by only 1.2%. In Trinidad, Bolivia, the impact of floods in 2006 had a negative impact on inequality. The Gini coefficient rose from 0.37 to 0.43 in the city between 2006 and 2007. The disproportionate affect of disasters in the poorest households is further substantiated by evidence from Africa. In Burkino Faso, the 1984-1985 drought affected the poorest third of the population 10% more than the wealthier third. In Madagascar, cyclone impacts on the volume of agricultural production of the poorest quintile were 11% compared to only 6% on the riches quintile. In Ethiopia, low rainfall mainly affected households with low initial asset holdings. Other studies have shown that in countries where the socio-economic status of women is low disasters have a significant effect on the gender gap in life-expectancy31 given that disasters exacerbate previously existing patterns of discrimination that make women more vulnerable. In countries, where the socioeconomic status of women improves the effect vanishes. For example, following the 2006 floods in Trinidad, Bolivia, women’s income was negatively affected more than that of men. 31 Neumayer and Piper 2007, see reference papers. 27 3.6.4 Disaster has long term effects on the poor Disaster losses are often measured only in terms of mortality or direct economic impact. There is evidence, however, that disasters can have longer term impacts in areas such as health, education and nutrition that can seriously affect the possibilities of recovery and further development. While strategic economic sectors and wealthy households can recover quite quickly after a disaster, the setbacks to poor households may last for years and in some cases decades. If households have few assets to buffer the impact of a disaster, the transitory impacts of a hazard can turn into longer-term impacts, particularly on highly vulnerable groups such as children. Nutrition shortfalls in children can affect their human development later in life, while households may adopt short-term strategies with high costs in the future, such as taking children out of school or depleting their few remaining assets. In other words, disaster loss not only has an immediate impact on welfare but can erode medium and long term welfare through an erosion of human and physical capital Within poor households, children and women may be particularly affected. Temporary poor health and malnutrition during drought can produce stunting, lower school achievement and attainment as well as lower health and lower wages and productivity as an adult. In the case of children, there are quite dramatic negative impacts if nutritional deficiencies occur from the womb to about 2 years old. In Zimbabwe, for example, 16 years after the 1982-84 drought in Zimbabwe affected children had 7% loss of adult earnings. Drought shocks in Ethiopia, for example, had a major detrimental effect on child health. Children aged 624 months at the time of the 1994-1995 drought experienced about a 0.9 cm growth loss in height over a six month period in those households where half the crop was damaged. In Bangladesh, the growth of children in landless and credit constrained households was severely affected by the 1988 floods32. During the 1994-95 drought in Zimbabwe, women’s body mass fell by about 3% while no impact was found on men’s health. Impacts on adult health were also observed in Ethiopia, where during the 19941995 drought the Body Mass Index in communities with poor rainfall and low landholdings dropped by 0.9%. Impacts on adult health can impair productivity and increase the risk of premature mortality In Nepal, people living in areas which had been affected by floods in the past, were more likely to suffer from wasting and low weights. Similarly the population in areas affected by landslides, were associated with higher percentages of stunting. In El Salvador, following the 2001 earthquakes, the probability of school enrollment for children in the most affected households decreased by 5.3%. Similarly, there was a worsening in school retention and progression in some areas of Nicaragua affected by Hurricane Mitch33. In Cote d’Ivoire, enrollment declined by about 20% in drought prone areas. 3.6.5 Coping with risk and loss can further erode development In rural areas, options to recover from major asset depletion are far more constrained than in urban areas, which can lead to permanently decreased welfare. Informal coping mechanisms such as selling livestock are only partially effective as a buffer. During the 1999 drought in Ethiopia, livestock herds declined by almost 40% and it was estimated that 25% of livestock reductions were distress sales where the seller received less than 50% of the normal price of the animal sold. Similarly, in Burkino Faso, livestock sales only covered 20 -30 % of crop income short-falls due to rainfall deficiencies. Only a quarter of the income variability was smoothed via stored grains. The limited success of such strategies is evident from the experience of the 2004 drought in Ethiopia, where despite limited mortality, consumption levels of 32 33 Foster, 2005 (from Baez background paper) Ureta, 2005 (from Baez background paper0 28 those affected by the drought fell by 16%. Similarly in Burkino Faso, the 1984 -85 drought apparently produced no mortality but lead to increases in poverty of 2 – 19% in the Sahelian region and 12 – 15% in the Sudanian region. In El Salvador, households had to sell productive assets such as animals or land, use savings or borrow and stop or cancel planned investments in physical capital as a result of the earthquake. Alternative sources of income and consumption played only a limited role in absorbing losses associated with the disaster. 26.8% of the households reported less food consumption because of the earthquake. Given that the vast majority of households in rural areas in Africa, Asia and Latin America do not have access to insurance and credit, it is very difficult and long to recover assets lost in a disaster. This means that many years after a drought or famine many rural households are still facing difficulties in recovery. 10 years after the devastating famines in Ethiopia in the 1980s cattle holdings in asset-poor households were still two-thirds of what they were before. In some cases, recovery may be so long that welfare is permanently affected creating what are known as poverty traps. Given the risk of poverty traps, evidence from Burkino Faso, for example, shows that some households prefer to drastically reduce consumption in order to protect livestock assets. Consumption growth loss would seem to be common 10 years or more after drought disasters. Recovery would also seem to be wealth differentiated meaning that the poorest recover slower than the wealthy. In the case of successive disasters, or where extensive risk disasters have eroded assets, the use of stores of assets to smooth consumption and buffer against loss is less likely to be successful. The impact of disaster loss on poverty is likely to be particularly drastic in contexts that experience frequent successive droughts and floods. The poorest households do not have the ability to recover assets from one disaster before being affected by the next, which leads into accelerated decline in welfare. Livelihoods in hazard prone areas are affected not only by asset loss itself but by as yet unrealized risks. For example, farmers are less likely to invest in fertilizer in drought prone areas, given that it a harvest fails due to drought the investment is lost. However, without fertilizer productivity is very low constraining the capacity of farmers to accumulate sufficient assets to absorb loss when a drought occurs. For example, poor households in drought prone areas of Tanzania tend to grow sweet potatoes more than any other crop because it is drought resistant, despite the fact that it’s returns per hectare are about 25% less than sorghum, maize or cotton. A simulation exercise in Zimbabwe, for example, found that exposure to hazards would reduce the capital stock across households by 46% after 50 years, due to such ex-ante strategies to minimize risk.. Similalry, in urban areas, households may prefer a central but hazard prone location if it improves access to work and income, even though asset loss due to flooding, for example impairs future welfare. 3.7 Conclusions For most international observers, it is no exaggeration to say that the disaster problematic per se is largely associated with manifestations of intensive risk. The national and international political impact of intensive loss events, the continued predominance of disaster response and international humanitarian assistance as approaches to deal with disaster risk, the focus of scientific research on major geological and climatic processes, including climate change and the compilation of disaster loss statistics centred on mortality and direct economic loss have all contributed to this perception. As stressed in Chapter 2 of this report, major concentrations of intensive risk, associated with both mortality and economic asset loss risk do exist and represent an enormous challenge to global stability and sustainability. However, in the shadow of the intensive risk disasters that capture media and political attention, patterns of extensive risk are emerging, manifested through thousands of smaller disasters each year, expanding geographically over wide areas and associated mainly with hydrometeorological hazards, such as floods, landslides and fires. Extensive risk has been largely ignored by the humanitarian 29 community, the media and the private sector. In a recent IFRC report, extensive risk was described as uncounted risk34 within an overall typology of neglect. Since about 1990, the number of extensive risk disasters has increased rapidly and has expanded territorially: in particular disasters associated with floods and heavy rains. This expansive trend is driven by different factors in each country, including: the decline in the regulatory services provided by ecosystems, land use changes in rural areas, the increase in the urban population living in informal settlements, increases in run-off due to urban growth and in appropriate water management and a chronic underinvestment in drainage in urban areas. While extensive risk manifests more in terms of recurrent damage than in terms of massive death and destruction, it is now both perennial and pervasive in many countries and importantly prefigures new patterns of intensive risk. Put simply the extensive risks of today are the intensive risks of tomorrow. While it is the way countries are managing their economic, social and urban development and their environments that is driving the expansion of extensive risk, climate related variables such as more intense precipitation also increase risk. Given that most extensive risk is related to hydrometeorological hazard, it is extremely sensitive to an intensification of the watercycle, as well as to other climate change outcomes such as glacier melt. Extensive risk, therefore, also represents a major challenge for climate change adaptation. There is evidence that poor urban and rural households are disproportionately affected by manifestations of extensive risk. The erosion of assets and loss of income caused by housing damage and crop loss, as well as by the destruction of local infrastructure generally goes uncounted and represents the invisible side of disaster costs. According to our calculations its total cost may be very significant, however. When extensive risk is overlaid with intensive risk disasters causing covariate impacts in a large number of households, those effects are multiplied and can lead to increased poverty in the short-run and a series of very damaging effects on health, welfare and productivity in the long-run. These disaster impacts particularly affect the poorest households and their most vulnerable members such as young children and women. The impact on children is particularly critical given that it compromises the development potential of future generations. The ex-post mechanisms designed by the international community to respond to disasters, such as humanitarian assistance and reconstruction loans, are designed to respond to events with major mortality or direct economic loss. Recurrent asset loss in poor households due to extensive risk as not addressed by international appeals for assistance, damage and needs assessments and similar mechanisms. Neither is much attention paid to the longer term effects of disasters on the most poor and vulnerable. This implies that poor rural and urban households are not disproportionately affected by disaster, they also receive the least assistance. The underlying risk factors that are configuring extensive risk and future intensive risk are explored in the next chapter. Given the relationship between disaster risk and poverty highlighted in this chapter, addressing those factors is a fundamental challenge, for both disaster risk and poverty reduction. 34 International Federation of Red Cross and Red Crescent Societies (IFRC), 2006, World Disaster Report 2006 Neglected Crisis. 30 Technical Note 3.1 – Establishment of the threshold between intensive and extensive risk Previous studies undertaken in 2002 by the ISDR Working Group 3 on Risk, Vulnerability and Disaster Impacts35 and in 2005 at the National University of Colombia36 applied different methodologies to identify the number and impact of small disasters in DesInventar. These methods were reviewed in order to see if they provided a viable approach for the identification of extensive risk. The Working Group 3 study compared the DesInventar with EM-DAT in four countries: Chile, Colombia, Jamaica and Panama. A large number of reports in DesInventar with losses below the EMDAT threshold of 10 deaths or 100 people affected were identified, most of which would reflect extensive risk. However, large disasters above the EM-DAT threshold would generally have intensive as well as extensive impacts in different local areas. At the same time, the EM-DAT threshold was not established as a measure of intensive risk. Identifying DesInventar events which were not reported in EM-DAT was therefore not a valid approach to identify extensive risk. The second study, which examined the impact of small disasters in Colombia, used a statistical procedure to calculate outlier events which exceeded 3.5 standard distributions from the mean distribution of DesInventar reports in Colombia. While this approach has a transparent statistical procedure, the definition of the number of standard deviations from the mean is by definition arbitrary and relative. If the number of standard deviations is increased or decreased then the threshold between intensive and extensive risk would change and the number of outliers would increase or decrease. After ruling out the above methods, the approach chosen was to compare the distribution of DesInventar reports across the dataset of 126,620 disaster reports in the 13 countries and states between 1970-2007 with the numbers of deaths and of houses destroyed. As Figure 0.16 and Figure 0.17 show a statistical analysis was carried out to identify the point where the maximum percentage of losses is concentrated in the minimum number of reports for both the variables, in this case where the fatality slope fell below a gradient of 0.1%. A similar exercise was carried out for destroyed houses. 50 deaths and 500 destroyed houses were considered to represent a reasonable threshold that was both easy to understand and met the above criteria. Figure 0.16: Maximum percentage of losses GAR Asian & LAC DesInventar Fatality profile for All & Fatal Events (percentile) F atality s hare (perc entile) 100.00% 95.00% 90.00% 85.00% 80.00% 75.00% 70.00% 65.00% 60.00% - 20 40 60 80 100 Number of Deaths All As ia events All As ian F atal events All L AC events All L AC F atal events 35 OSSO, LA RED, 2002, Comparative Analysis of Disaster Databases, report produced for ISDR Working Group 3.. http://www.unisdr.org/eng/task%20force/tf-working-groups3-eng.htm 36 Mabel C. Marulanda, Omar D. Cardona, and Alex H. Barbat, 2007, Revealing the Impact of Small Disasters to the Economic and Social Development. The need of a proposal to cover the losses of low-income people and a framework to measure and reduce the vulnerabilityby: http://www.ehs.unu.edu/file.php?id=295 31 Figure 0.17: Asia and LAC DesInventarfatality profile (by percentile of all reported & fatal events) Asia Fatality class intervals All Asia events All Asian fatal events 1 2 3 5 7 10 15 25 50 75 100 250 500 1,000 2,500 5,000 7,500 10,000 15,000 25,000 0.00% 83.00% 91.50% 94.70% 96.90% 97.80% 98.40% 98.90% 99.30% 99.60% 99.70% 99.70% 99.80% 99.90% 99.90% 99.90% 99.90% 99.90% 99.90% 99.90% 99.90% 0.00% 50.30% 68.70% 81.90% 87.30% 90.70% 93.70% 95.90% 97.60% 98.30% 98.60% 99.30% 99.50% 99.70% 99.80% 99.80% 99.90% 99.90% 99.90% 99.90% 50,000 100% 100% n' value 45,217 7,652 All fatality slope LAC Cumulative deaths % Cumulative death 0.000 0.503 0.184 0.066 0.027 0.011 0.006 0.002 0.001 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 3,851 6,663 10,084 12,294 14,346 16,959 20,139 24,495 27,637 29,365 37,525 43,633 53,646 63,486 75,220 112,265 112,265 112,265 112,265 203,536 1.89% 3.27% 4.95% 6.04% 7.05% 8.33% 9.89% 12.03% 13.58% 14.43% 18.44% 21.44% 26.36% 31.19% 36.96% 55.16% 55.16% 55.16% 55.16% 100.00% 0.00% 89.60% 94.20% 96.30% 98.00% 98.70% 99.10% 99.40% 99.70% 99.80% 99.90% 99.90% 99.90% 99.90% 99.90% 99.90% 99.90% 99.90% 99.90% 99.90% 100.00% 0.00% 44.20% 64.70% 81.50% 87.70% 91.60% 95.00% 97.20% 98.60% 99.20% 99.40% 99.80% 99.80% 99.90% 99.90% 99.90% 99.90% 99.90% 99.90% 100.00% 0.442 0.205 0.084 0.031 0.013 0.007 0.002 0.001 0.000 0.000 0.000 0.000 0.000 86,403 79.02% 109,345 100.00% 0.000 203,536 100.00% 100.00% 100.00% - 109,345 100.00% 81,399 8,431 All LAC events All LAC fatal events All fatality slope Cumulative deaths % Cumulati ve death 3,728 3.41% 7,182 6.57% 11,945 10.92% 14,796 13.53% 17,290 15.81% 20,588 18.83% 23,945 21.90% 28,114 25.71% 30,920 28.28% 32,175 29.43% 36,317 33.21% 37,407 34.21% 40,903 37.41% 45,403 41.52% 48,403 44.27% 48,403 44.27% 48,403 44.27% 68,403 62.56% No events Fatal events with asymptotic fatality < 0.1% slope Figure 0.18: Intensive and extensive risk in Asia and Latin America (1970 – 2007) Country Risk Type Hazard type Disaster reports % Deaths % Argentina Extensive Hydrometeorological Extensive Intensive 14,525 98.7 2,554 84.8 Geological 165 1.1 68 2.3 1,041 2.0 3,031 2.2 Hydrometeorological 25 0.2 360 12.0 24,775 46.8 28,600 20.5 Intensive Geological 4 0.0 30 1.0 8,800 16.6 6,643 4.8 14,719 100.0 3,012 100.0 52,959 100.0 139,369 100.0 Extensive Hydrometeorological 1,363 98.5 435 55.1 2,369 56.7 2,066 100.0 Extensive Geological 16 1.2 0 0.0 8 0.2 0 0.0 Intensive Hydrometeorological 4 0.3 249 31.6 0 0.0 0 0.0 Intensive Geological 1 0.1 105 13.3 1,800 43.1 0 0.0 1,384 100.0 789 100.0 4,177 100.0 2,066 100.0 SUB TOTAL Bolivia SUB TOTAL Colombia 18,343 % 34.6 Houses Damaged 101,095 % 72.5 Extensive Hydrometeorological 21,316 97.2 6,645 18.5 86,655 50.5 397,968 83.8 Extensive Geological 548 2.5 221 0.6 6,489 3.8 35,386 7.4 Intensive Hydrometeorological 58 0.3 2,387 6.7 27,224 15.9 2,991 0.6 Intensive Geological 14 0.1 26,603 74.2 51,282 29.9 38,754 8.2 21,936 100.0 35,856 100.0 171,650 100.0 475,099 100.0 Extensive Hydrometeorological 8,957 97.7 414 86.1 2,981 39.0 36,497 80.9 Extensive Geological 207 2.3 21 4.4 835 10.9 2,538 5.6 Intensive Hydrometeorological 0 0.0 0 0.0 0 0.0 0 0.0 Intensive Geological 3 0.0 46 9.6 3,830 50.1 6,068 13.5 SUB TOTAL Costa Rica Houses Destroyed 32 SUB TOTAL Ecuador 100.0 7,646 100.0 45,103 100.0 3,237 96.1 1,944 67.9 8,223 75.6 36,088 90.3 124 3.7 43 1.5 406 3.7 3,687 9.2 Hydrometeorological Geological Intensive Hydrometeorological 4 0.1 550 19.2 663 6.1 178 0.4 Intensive Geological 2 0.1 327 11.4 1,590 14.6 0 0.0 3,367 100.0 2,864 100.0 10,882 100.0 39,953 100.0 Extensive Hydrometeorological 2,065 55.3 3,389 2.5 3,379 2.4 119,705 37.1 Extensive Geological 1,601 42.9 375 0.3 3,729 2.7 46,569 14.4 Intensive Hydrometeorological 29 0.8 2,099 1.5 29,420 21.3 2,800 0.9 Intensive Geological 36 1.0 131,430 95.7 101,485 73.5 153,606 47.6 3,731 100.0 137,293 100.0 138,013 100.0 322,680 100.0 71.5 Extensive Hydrometeorological 11,978 96.8 7,692 32.5 28,198 8.1 1,387,289 Extensive Geological 317 2.6 242 1.0 1,312 0.4 81,220 4.2 Intensive Hydrometeorological 65 0.5 3,549 15.0 236,801 68.4 370,011 19.1 Intensive Geological 14 0.1 12,171 51.5 80,127 23.1 101,697 5.2 12,374 100.0 23,654 100.0 346,438 100.0 1,940,217 100.0 40.8 SUB TOTAL Nepal 481 Extensive SUB TOTAL Mexico 100.0 Extensive SUB TOTAL Iran 9,167 Extensive Hydrometeorological 11,295 98.8 8,513 80.6 113,580 58.1 59,936 Extensive Geological 72 0.6 116 1.1 2,001 1.0 5,300 3.6 Intensive Hydrometeorological 49 0.4 1,180 11.2 48,062 24.6 31,811 21.6 Intensive Geological 19 0.2 757 7.2 31,709 16.2 50,023 34.0 11,435 100.0 10,566 100.0 195,352 100.0 147,070 100.0 Orissa, Extensive Hydrometeorological 7,298 94.8 6,572 22.0 281,154 23.4 398,913 15.2 India Extensive Geological 6 0.1 0 0.0 3 0.0 11 0.0 Intensive Hydrometeorological 395 5.1 23,296 78.0 917,797 76.5 2,227,441 84.8 Intensive Geological 0 0.0 0 0.0 0 0.0 0 0.0 7,699 100.0 29,868 100.0 1,198,954 100.0 2,626,365 100.0 SUB TOTAL SUB TOTAL Peru Extensive Hydrometeorological 13,134 92.1 5,191 13.0 39,860 9.2 244,275 67.1 Extensive Geological 1,062 7.4 876 2.2 19,847 4.6 43,865 12.0 Intensive Hydrometeorological 36 0.3 20,409 51.0 61,402 14.1 51,312 14.1 Intensive Geological 29 0.2 13,550 33.9 312,894 72.1 24,732 6.8 14,261 100.0 40,026 100.0 434,003 100.0 364,184 100.0 Extensive Hydrometeorological 9,745 98.8 1,801 8.7 30,416 13.0 242,128 55.3 Extensive Geological 35 0.4 221 1.1 1,836 0.8 3,000 0.7 Intensive Hydrometeorological 40 0.4 2,383 11.5 147,072 62.6 144,923 33.1 Intensive Geological 42 0.4 16,360 78.8 55,472 23.6 47,515 10.9 9,862 100.0 20,765 100.0 234,796 100.0 437,566 100.0 12,325 98.7 1,847 36.6 109,869 48.5 550,346 60.9 SUB TOTAL Sri Lanka SUB TOTAL Tamil Nadu, Extensive Hydrometeorological India Extensive Geological 62 0.5 213 4.2 3,171 1.4 1,703 0.2 Intensive Hydrometeorological 81 0.6 839 16.6 90,511 40.0 326,994 36.2 Intensive Geological 23 0.2 2,145 42.5 22,991 10.1 24,056 2.7 12,491 100.0 5,044 100.0 226,542 100.0 903,099 100.0 61.8 SUB TOTAL Venezuela Extensive Hydrometeorological 4,135 98.6 1,395 52.4 13,975 28.6 78,290 Extensive Geological 44 1.0 10 0.4 6 0.0 235 0.2 Intensive Hydrometeorological 15 0.4 1,258 47.2 34,955 71.4 48,115 38.0 Intensive Geological 0 0.0 0 0.0 0 0.0 0 0.0 4,194 100.0 2,663 100.0 48,936 100.0 126,640 100.0 121,373 95.9 48,392 15.5 739,002 24.1 3,654,596 48.3 4,259 3.4 2,406 0.8 40,684 1.3 226,545 3.0 SUB TOTAL All TOTAL Extensive Hydrometeorological Extensive Geological Intensive Hydrometeorological 801 0.6 58,559 18.7 1,618,682 52.7 3,235,176 42.7 Intensive Geological 187 0.1 203,524 65.0 671,980 21.9 453,094 6.0 126,620 100.0 312,881 100.0 3,070,348 100.0 7,569,411 100.0 33 List of Figures Figure 3.1: Number of extensive risk disaster reports seperated in several hydrometeorological categories (1970-2007) ...........................................................................................................................................5 Figure 3.2: Average seasonal occurrences of disaster reports associated with fire, Orissa, India (1970 – 2007) ......................................................................................................................................................6 Figure 3.3: Distribution of mortality associated with intensive and extensive risk across the data set (1980-2006) ...........................................................................................................................................6 Figure 3.4: Deaths, houses destroyed and houses damaged in Sri Lanka and international awareness (1992-2005) ...........................................................................................................................................8 Figure 3.5: Number of extensive hydrometeorological disaster reports and number of associated deaths for data set (1980-2006).......................................................................................................................10 Figure 3.6: Number of extensive hydrometeorological disaster reports and number of associated deaths for data set (1980-2006).......................................................................................................................10 Figure 3.7: Causes for housing destruction in Orissa, India (1970 – 2007) ................................................11 Figure 3.8: Housing destruction in Orissa, India (1970 – 2007) .................................................................11 Figure 3.9: Number of municipalities affected by extensive hydrometeorological risk (1980-2006) ........12 Figure 3.10: Number of disasters associated with floods and heavy rains (1980 – 2006) ...................13 Figure 3.11: Extensive risk disasters in Mexico associated with floods, rains and flash floods (1980–2006) .............................................................................................................................................................14 Figure 3.12: Floods and rains in Colombia as a proportion of all extensive risk disasters (1970-1979) and (1998 – 2007) .......................................................................................................................................14 Figure 3.13: Flood and rain disasters in Costa Rica (1990-2007) ...............................................................15 Figure 3.14: Precipitation Costa Rica 10 years averages (1957-2007) .......................................................15 Figure 3.15: Extensive flood and heavy rain disasters in the province of Buenos Aires and the Federal Capital (1970-2007) .............................................................................................................................19 Figure 3.16: Extensive flood disaster and housing damage is concentrated in urban areas (1996 – 2007) ............................................................................................................. Error! Bookmark not defined. Figure 3.17: Maximum percentage of losses ...............................................................................................31 Figure 3.18: Asia and LAC DesInventarfatality profile (by percentile of all reported & fatal events) .......32 Figure 3.19: Intensive and extensive risk in Asia and Latin America (1970 – 2007) .................................32 List of Maps Map 3.1: Armenia, Columbia earthquake extensive and intensive impacts (1999) ......................................2 Map 3.2: Ecuador intensive disaster risk events (1970-2007) .......................................................................4 Map 3.3: Ecuador extensive disaster risk events (1970-2007) ......................................................................4 Map 3.4: Number of extensive geological disaster risk reports from the Islamic Republic of Iran (1970 – 2007) ......................................................................................................................................................6 Map 3.5: Extensive housing damage in Tamil Nadu, India (1976-2007) .....................................................8 Map 3.6: Intensive housing destruction in Tamil Nadu, India (1976-2007) .................................................8 Map 3.7 Number of extensive risk events (1980-1989, 1990/1999, 2000/2007) ........................................12 Map 3.8: Spatial evolution of extensive hydrometeorological disasters in Ecuador from 1970-2007........20 Map 3.9: Redistribution of extensive risk associated with these hazards between 1970-1985 and 19862006. .................................................................................................... Error! Bookmark not defined. Map 3.10: Houses damaged in extensive disasters in North-Central Venezuela (1980 – 2007)......... Error! Bookmark not defined. Map 3.11: Distribution of flood events (1950-1959; 1960-1969) ............ Error! Bookmark not defined. Map 3.12: Orissa, India extensive disaster risk reports (1970 – 2007) .......................................................13 Map 3.13: Orissa, India intensive disaster risk reports (1970 – 2007) ........................................................13 List of Tables Table 3.1: Spatial distribution of extensive risk (1970-2007) .......................................................................4 34 Table 3.2: Loss attributes by risk category across the sample (1970–2007) .................................................9 Table 3.3: Annual increase in mortality and housing damage compared with annual population growth (1980-2006) ......................................................................................... Error! Bookmark not defined. Table 3.4: Urban growth rates in % .............................................................................................................16 Table 3.5: .....................................................................................................................................................17 Table 3.6: I am asking Cristina since the 15 October in 5 spererate emails to send me the talbe, but she refuses to send an excel version. Only what I have here, and another one with Mexico in Power Point!....................................................................................................................................................17 ALSO INCORPORATE BOX INTO SECTION 4.2 9 (or eventually into chapter 3 from Action Aid (Climate change, urban flooding and the rights of the urban poor in Africa: key findings from six African cities) 35