Scholarly article on topic 'The data gap: An analysis of data availability on disaster losses in sub-Saharan African cities'

The data gap: An analysis of data availability on disaster losses in sub-Saharan African cities Academic research paper on "Social and economic geography"

CC BY
0
0
Share paper
Keywords
{}

Abstract of research paper on Social and economic geography, author of scientific article — Emmanuel Osuteye, Cassidy Johnson, Donald Brown

Abstract Urban centres in sub-Saharan Africa are increasingly affected by disasters as well as smaller, everyday hazards. Decision-makers in the region require better information about urban disaster impacts to plan how best to use their resources to reduce risks to the people most affected. This paper reviews the different kinds of publicly available data on human and economic losses from large and small disasters as well as on health impacts of everyday hazards to assess the quality and breadth of information available for urban areas. The findings reveal emergent information about disaster losses in urban areas generated by the DesInventar methodology, but the quantity of data and the coverage of disaster events is not enough to make robust conclusions for a particular city. Data about losses to health from everyday hazards are provided by demographic and health surveys, but their sample sizes are too small to provide accurate or detailed data on individual urban centres or on ‘slums’/informal settlements. The findings highlight the need for more robust data collection that would assist national and local decision-makers to make more informed and location specific choices about disaster risk management. Systematic collection and cataloguing is needed to make information robust enough for planning and policy-making – and to have relevant information for each ward and district within urban areas, including informal settlements.

Academic research paper on topic "The data gap: An analysis of data availability on disaster losses in sub-Saharan African cities"

Author's Accepted Manuscript

The Data Gap: An analysis of data availability on disaster losses in sub-Saharan African Cities

Emmanuel Osuteye, Cassidy Johnson, Donald Brown

www.elsevier.com/locate/iidr

PII: S2212-4209(17)30261 -3

DOI: http://dx.doi.org/10.1016/j.ijdrr.2017.09.026

Reference: IJDRR655

To appear in: International Journal of Disaster Risk Reduction

Cite this article as: Emmanuel Osuteye, Cassidy Johnson and Donald Brown, The Data Gap: An analysis of data availability on disaster losses in sub-Saharan African Cities, International Journal of Disaster Risk Reduction, http ://dx.doi.org/ 10.1016/j.ij drr.2017.09.026

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

The Data Gap: An analysis of data availability on disaster losses in sub-Saharan African Cities

Emmanuel Osuteye, Cassidy Johnson and Donald Brown, Development Planning Unit, University College London1

Corresponding author. Dr Emmanuel Osuteye Research Associate Development Planning Unit Faculty of the Built Environment 34 Tavistock Square London WC1H 9EZ e.osuteye@ucl.ac.uk

1 The authors are grateful for the funding received from the UK department for International Development (DFID) and the Economic and Social Research Council (ESRC) under the Urban Africa Risk Knowledge grant number ES/L008777/1

Abstract

Urban centres in sub-Saharan Africa are increasingly affected by disasters as well as smaller, everyday hazards. Decision-makers in the region require better information about urban disaster impacts to plan how best to use their resources to reduce risks to the people most affected. This paper reviews the different kinds of publicly available data on human and economic losses from large and small disasters as well as on health impacts of everyday hazards to assess the quality and breadth of information available for urban areas. The findings reveal emergent information about disaster losses in urban areas generated by the DesInventar methodology, but the quantity of data and the coverage of disaster events is not enough to make robust conclusions for a particular city. Data about losses to health from everyday hazards are provided by demographic and health surveys, but their sample sizes are too small to provide accurate or detailed data on individual urban centres or on 'slums'/informal settlements. The findings highlight the need for more robust data collection that would assist national and local decision-makers to make more informed and location specific choices about disaster risk management. Systematic collection and cataloguing is needed to make information robust enough for planning and policy-making - and to have relevant information for each ward and district within urban areas, including informal settlements.

1.0 Introduction

Sub-Saharan Africa (SSA) is one of the world's least urbanised, yet most rapidly urbanising regions (UNDESA, 2015). In the context of widespread poverty, climate change, and limited capacity to plan and manage rapid urban growth, towns and cities across SSA are becoming increasingly impacted by a wide range of hazards, encompassing everyday hazards (such as infectious and parasitic diseases, and road traffic injuries) and small disasters (such as localised landslides and floods) and large disasters (such as tropical storms, earthquakes, and floods) (Adelekan et al., 2015). Those living in informal settlements lacking basic infrastructure and services are often disproportionately affected by such impacts. Moreover, while most attention has traditionally been paid to large disasters, available evidence suggests that the cumulative impacts of everyday hazards and small disasters may be considerably greater (Pelling and Wisner, 2009).

This means that (a) integrated approaches to risk reduction, involving urban planning and environmental management, public health, disaster management, and climate change adaptation, are required if towns and cities in SSA are to be made more resilient; and (b) better, more detailed data for urban areas are required if planners and decision-makers in SSA are to effectively plan for and respond to these disasters (Adelekan et al, 2015). To support these objectives, we argue for a broader conceptualisation of risk that encompasses not just the potential impacts of large-scale events, but also those of smaller-scale events that often account for a larger share of overall losses in terms of health and income, particularly at the household scale (ibid; Bull-Kamanga et al., 2001; Pelling and Wisner, 2009; UNISDR, 2011, 2015. At present, only losses from larger events are recorded by the major disaster databases, but they do not consider the relative importance of smaller events (such as localised flooding, endemic diseases linked to poor quality food, water, hygiene, or lack of health facilities, accidents and fires), which are excluded. Accounting for the impacts of smaller events perhaps make the strongest argument for linking disaster risk management with development, especially in areas where the underlying social characteristics and

living conditions of people, in terms of their capacity to cope with, resist and recover from the impacts of disasters, make them more susceptible to harm (Bara, 2010).

This paper examines the availability of data on losses from disasters (small and large) and endemic health hazards in urban areas in SSA and what this data tells us about who is most affected, and why. It reviews the concept of disaster losses, and examines how they are used by the major databases. It also examines the extent to which the major disaster databases capture the effects of hazard events of different sizes, and presents an overview of the data sources from the health sector that could shed light on everyday risks, in particular those related to health. The paper concludes by examining the key limitations of the databases and data sources reviewed, and by identifying potential areas of interests for urban researchers.

2.0 Disaster Loss databases

The occurrence of disaster events have been increasingly documented and accounted for in international disaster databases. The United Nations Development Program's Global Risk Information Platform (GRIP) website has a comprehensive list of disaster databases classified as global, regional or national, and this indicates that there are four global disaster databases, EM-DAT, Global Disaster Identifier Number (GLIDE), University of Richmond Disaster Database Project and NatCatService. Additionally, the Dartmouth Flood Observatory has an archive of over close to 4000 large flood events and can be searched by country, and the DesInventar database is a collection of national databases, which currently includes 89 countries and is growing in scope.

EM-DAT and DesInventar are central to the analysis in this paper; they are two of the most popular international databases and are widely cited in policy documents and research analyses (Guha-Sapir and Hoyois, 2012),. Both EM-DAT and DesInventar attempt to aggregate and classify data to support analysis of both the types and effects of the disasters recorded. They work based on a common standardized classification and definition of types of perils and hazards (Integrated Research on Disaster Risk, 2014)., including sub-national data on 15 countries in SSA (Comoros, Djibouti, Ethiopia, Kenya, Madagascar, Mali, Mozambique, Morocco, Mauritius, Niger, Togo, Tunisia, Senegal, Sierra Leone, Seychelles and Uganda) and partial data on Tanzania (Zanzibar).

The databases have different thresholds about what they consider to be a disaster event. In EM-DAT, for a disaster event to be recorded it must fulfil one of the following criteria: have 10 or more people deaths, 100 or more people affected/injured/homeless, or declaration by the country of a state of emergency and/or an appeal for international assistance. Whereas the definition for DesInventar database is 1 or more human losses or $1 or more in economic losses. DesInventar uses national and local newspapers, police and public health reports as sources of information and will include a disaster event if there is any kind of human or economic loss. EM-DAT is compiled from various sources including UN, governmental and non-governmental agencies, insurance companies, research institutes and press agencies has a wide range of other sources. Thus, these two databases can portray quite different pictures of disaster losses in a country.

The DesInventar methodology gives a stronger indication of the spectrum of the type of disaster events and considerable analysis has been undertaken to understand the differences between losses from large (intensive) disaster events and smaller scale (extensive) events (UNISDR 2015, 2013, 2011) and is more conducive to understanding the breadth of everyday losses, small disasters and large disasters in urban areas of SSA. DesInventar defines extensive risk as "the risk layer of high-

frequency, low-severity losses [that] manifests as large numbers of recurrent, small-scale, low severity disasters which are mainly associated with flash floods, landslides, urban flooding, storms, fires and other localized events" (UNISDR 2015:90). The analysis of the available data showed that, since 1990, a majority of the loss and damages have been associated with extensive disasters and accounted for most of the disaster morbidity and displacement, which were a particular challenge for areas already vulnerable because of poor infrastructure and under-development (UNISDR, 2015:96). And where other socio-economic factors and access to resources, such as the reduced levels of education, instability of household income and reserves and the lack of social networks and external support during disaster events, affect the capacity of residents to cope and recover from disaster events (Kabisch et al, 2015).

3.0 National level disaster losses in SSA

EMDAT and DesInventar show that disasters in SSA are dominated by weather-related hazards such as floods, cyclones and storms, and drought. The major databases also show that in addition to the dominance of climatic-related hazards, the disaster profile of SSA highlights the occurrence of disease epidemics, fires, and accidents and earthquakes.

Based on an analysis of EM-DAT data over 80 million people in SSA were affected by large-scale natural disasters, resulting in 45,733 deaths between 2010 and 2015 (Table 1). Considering that there were only 354 disaster events recorded in all the SSA countries during the 5-year period, the fatality rate is relatively high at about 129 deaths per disaster incident. This was found to be significantly higher than natural disaster death rates in South East Asia and Latin America, within the same period.

Table 1: Natural disasters and extent of damage and loss in SSA 2010-2015, EM-DAT data2.

Region Country Occurrence Death Affected w Injured Homeless Total affected

West Liberia 2 4,500 25,714 0 0 25,714

East Somalia 12 20,211 7,649,380 0 20,200 7,669,580

West Sierra Leone 6 4,184 35,422 5 2,257 37,684

West Guinea 7 2,455 84,476 6 4,000 88,482

West Nigeria 17 4,073 8,739,594 1,229 500 8,741,323

West Burkina F 6 877 7,000,705 13 21,000 7,021,718

Central Cameroun 10 1307 327,436 95 34,980 362,511

Central Chad 10 928 2,383,360 0 0 2,383,360

Central Congo DR 18 1,569 134,917 1,057 12,490 148,464

East Uganda 14 709 1,008,838 1,255 3,368 1,013,461

West Ghana 12 529 178,917 262 0 179,179

East Madagascar 13 551 2,981,349 1,328 225,551 3,208,228

Central Congo 8 324 14,496 89 7,500 22,085

This table shows which SSA countries from the EM-DAT data available, suffered the biggest disaster losses over the 5 year period. From the authors' analysis, countries have been ranked by fatality rate (number of deaths divided by the frequency of disaster events).

West Niger 18 599 4,125,396 5,009 34,790 4,165,195

East Kenya 19 593 10,361,686 3,479 5,000 10,370,165

Central Angola 10 233 2,036,359 31 79,570 2,947,997

East Malawi 13 294 2,817,461 651 350 2,818,462

West Togo 4 89 111,954 0 0 111,954

South Namibia 6 133 956,150 518 0 956,668

East Mozambique 18 363 1,153,217 3,737 0 1,156,954

South Zimbabwe 13 213 3,883,536 3 475 3,884,014

West Cote D'Ivoire 5 80 6,425 0 0 6,425

East Burundi 9 139 3275 232 14,000 17,507

East Tanzania 9 137 1,109,000 312 6,776 1,116,088

South South Africa 9 135 347,011 540 3,500 351,051

East Ethiopia 10 145 5,979,334 0 0 5,979,334

East Mauritius 1 11 0 82 0 82

South Swaziland 1 11 400 0 0 400

West Mali 9 93 4,194,612 0 0 4,194,612

West Benin 9 82 775,098 1000 152,759 928,857

East Rwanda 5 42 18,173 43 5,920 24,136

South Lesotho 4 26 730,515 0 2,600 733,115

South Botswana 2 12 4210 0 0 4210

West Senegal 8 29 1,817,582 163 0 1,817,745

Central CAR 9 20 48,225 121 3,870 52,216

West Gambia 5 11 470,261 0 0 470,261

West Guinea Bissau 1 2 56,792 0 0 56,792

West Mauritania 4 8 1,550,975 0 2,305 1,553,280

South Zambia 6 11 4,528 0 20,150 24,678

East Comoros 3 4 84,498 150 0 84,684

Central Gabon 5 1 81,926 0 0 81,926

West Cape Verde 1 0 2,500 0 0 2,500

East Djibouti 1 0 200,258 0 0 200,258

East Seychelles 2 0 7,435 0 0 7,435

The disaster losses in the major databases, focus on loss of life, injury and displacement. Detailed information on economic (or monetised) losses are poorly documented in the EM-DAT database. Non-economic losses are poorly documented as well. For example, the categories for 'death' and 'injury' do not include morbidity as a secondary effect, even though disasters often create the conditions for disease transmission and the spread of epidemics (IFRC 2010). The data further suggest that the number of people injured may be under-reported considering the total number of people affected by event. The financial losses were only computed for flood events, despite the relative importance of other types of disasters from country to country.

The data available on DesInventar, for the same period provides a more detailed account of loss by itemizing the number of 'houses destroyed' and 'houses damaged', and the number of deaths, injuries and missing persons by disaster event. There is currently very limited information on the monetized losses from the disaster incidents, however unlike EM-DAT data that had financial losses computed for only floods, there was some detail on the financial losses from droughts, coastal erosion and fires, within the limited data available.

Table 2: Summary of Natural disasters and extent of damage and loss in select SSA countries 20102015 (DesInventar)3

Country Event Events Counted Deaths Injured Houses Destroyed / Damaged

Kenya (2010 - 13) Drought 317 0 0 0

Epidemic 8 8 0 0

Extreme Rains 40 0 0 0

Flash Flood 3 0 0 75

Flood 341 478 28 13274

Landslide 14 37 7 96

Mudslide 3 4 0 0

Storm 1 5 0 0

Thunderstorm 18 25 14 2

Windstorm 7 0 0 16

Niger (2010 - 14) Drought 97 0 0 0

Epidemic 77 1288 0 0

Fire 66 26 0 217

Flood 567 730 174 65501

Forest Fire 110 0 0 0

Other 276 4957 4 0

Rains 1 0 0 0

Mali (2010 - 12) Drought 245 0 0

Flood 222 131 9892

Forest Fire 23 0 0

Thunderstorm 1 1 0

Senegal (2010 -14) Accident 376 728 2082 4

Mudslide (Dégradation Des Terres) 4 0 0 0

Building Collapse 104 74 338 203

Epidemic 7 3 0 0

Coastal Erosion (Erosion Côtière) 2 0 2 40

Bushfires 715 0 0 0

Lightning 53 68 52 7

Fires 205 77 55 864

Bird Investation (Infestation Aviaire) 1 0 0 0

Flood 93 54 15 12531

Shipwreck 22 174 68 0

3 The zero (0) entries from the DesInventar data represent fields with both a zero count and those that had no data reported.

(Naufrage)

Drowning 360 603 54 0

Rainstorm (Pauses Pluviometriques) 13 0 0 0

Drought 1 0 0 0

Extreme Temperature 2 13 0 0

Uganda (2010 -14) Accident 14 45 98 0

Animal Attack 19 1 16 0

Cyclone 1 0 0 75

Drought 189 12 0 0

Epidemic 30 259 0 0

Fire 79 19 30 518

Flash Flood 1 0 9 20

Flood 532 69 5 5595

Frost 1 0 0 10

Hailstorm 348 13 50 1786

Landslide 159 1461 22 1663

Lightning 31 57 214 0

Mudslide 22 19 0 298

Other 1 0 0 0

Rains 8 0 0 12

Rainstorm 40 8 0 512

Storm 89 0 72 871

Thunderstorm 3 0 0 0

Windstorm 8 0 0 66

Ethiopia (2010 - 12) Building Slide 1 0 0 0

Drought 1285 0 0 0

Fire 159 0 0 0

Flood 316 25 0 331

Hailstorm 2 0 0 0

Landslide 7 49 0 0

Plague 899 0 0 0

Rain 2 0 0 0

4.0 Data on disaster losses in urban areas of SSA

The 2015 UNISDR Global Assessment Report (UNISDR 2015) analysed sub-regional data across the 89 national datasets in DesInventar, including the 15 SSA countries4. For most of these countries, it is possible to infer that some of this analysis is of "urban data" because the events occurred in a sub -region that is almost entirely urbanised, which we can tell based on the cross referencing of the population densities of the sub-region to the known cities or urban centres from another data source (www.citypopulation.de), and on the assumption that the sub-region was a homogenous urban space. Thus we present below an analysis from a selection of large cities, Nairobi (Kenya), Niamey (Niger), Dakar (Senegal), plus Kampala (Uganda) and Freetown (Sierra Leone). It would be possible to extend this analysis to other major cities for the countries included in DesInventar, as there is data available to be analysed, although it requires cross-referencing with the urban population data, with Google maps and with local knowledge. This could yield some interesting results.

4 See http://www.desinventar.net/index_www.html

However, a note needs to be made about the quality of the "urban" datasets available in DesInventar. Overall the data for most of the urban areas is not of sufficient quality to reliably make conclusions about the totality of disaster losses in a particular urban area, but it does give an overview about the city and the kinds disasters that are prevalent. Often data is entered for one year and not others. The entries cover certain kinds of events (i.e. fires or traffic accidents or floods) but not for others. This depends on where the data comes from and who is responsible for upkeep of the dataset. One gets the impression that events are still vastly underreported, so the actual extent of losses is much greater than what is presented here. Nonetheless the analysis presents a range of the kinds of disaster events that we would expect to see in urban areas.

Kampala

In the District of Kampala, Uganda, the types of disasters between 1995 and 2013 were reported to be accidents, fire, flood and disease epidemics, lightning, rains, structural collapse and storm (Table 3). Out of the total number of deaths reported, 97% were caused by accidents5. Fire caused 63% of the housing damage/destroyed and flooding caused 36%. Flooding accounted for 95% of the people affected by disasters. However, the database as a whole accounts for only 67,529 (4.5% of the population of approximately 1.5 million) as affected by flooding over the period, which seems low considering that flooding is common is Kampala during the rainy seasons. There is no data on the economics losses from these events. The dominance of accidents is apparent here because there is a lack of data on other hazards.

Table 3: Losses by disaster type in Kampala District, 1995-2010 (DesInventar).

Event IS (A ■a (D u (O +J (D Q Deaths Injured Missing Houses Destroyed Houses Damaged Affected

Accident 14 4452 17013 0 0 0 0

Epidemic 7 36 0 0 0 0 787

Fire 15 7 27 0 212 0 2040

Flood 10 37 0 2 122 0 67529

Lightning 1 1 0 0 0 0 0

Rains 1 0 0 0 1 0 0

Storm 2 0 0 0 0 0 184

Structural Collapse 1 12 183 0 0 0 0

Nairobi

The database for Nairobi, Kenya, documents sporadic incidents of flooding, drought and disease epidemics (see Table 4) with only 52 entries across four years (2002, 2009, 2010 and 2012). It does not seem to record any entries for fires or accidents, which we would expect to see in a city of over 3

5These are largely police reports of traffic accidents amalgamated and entered as one data entry by year. From one entry of the database: "Total accidents reported in the year was 9558. There were 524 Government Vehicles, 14984 civilian vehicles, 1676 motor cycles, 849 ped/cycle, 2266 Pedestrian, 79 Forestry M/V, 100 Police M/V, 32 army Vehicles, 1 Prison Vehicles, 131 Diplomatic Vehicles and 16 U/Registered. 5058 of the accidents occurred during day and 4500, at night" 6 data cards refers to number of reports in the database

million people (Douglas et al, 2008)7. 46% of deaths were caused by flooding and 53% by epidemics, predominantly cholera outbreaks and swine flu. Flooding caused all of the reported housing damage/destruction. From the data, both flooding and disease epidemics are a problem in Nairobi. However, the database is very much lacking in detail and therefore does not provide an accurate picture of the risks in Nairobi; given its current level of detail, the database would not be useful for supporting policy-making.

Table 4: Losses by disaster type in Nairobi District, 2002-2013 (DesInventar).

Event Data Cards Deaths Injured Missing s Ï 0 S 1 s dJ G Houses Damaged Affected

Drought 9 0 0 0 0 0 0

Epidemic 21 22 0 0 0 0 120

Flood 21 30 5 1 60 333 100

Forest Fire 1 0 0 0 0 0 0

Niamey

Niamey, Niger, has a population of 978,029, according to the census in 2012. The Niamey region, which is what is reported on in DesInventar, has a population 1,026,848, thus 95% of the regional population lives within the city boundaries, making it perhaps useful to analyse as an urban area. Data entries for every year between 2000 and 2013 show that epidemics and floods are the prevailing disasters (Table 5). Epidemics caused 99% of the 968 reported deaths. There is quite detailed reporting of epidemics, with 96 entries indicated and mostly measles and some meningitis outbreaks. For example, 5469 cases of measles, resulting in 161 deaths was reported on 21 April 2004. This is the kind of information and level of detail that can assist with policy-making and planning. Fifty-one floods were reported across years of 2000, 2006, 2010, 2013, including 34 incidents of flooding in 2013 ranging is size from 0 - 932 houses destroyed. There are no other types of events reported and likely this is a big gap in the data.

Table 5: Losses by disaster type in Niamey, 2000-2013 (DesInventar).

Event Data Cards Deaths Injured Missing Houses Destroyed Houses Damaged Affected

Epidemic 96 968 0 0 0 0 26302

Flood 51 4 6 0 13039 473 13810

The synthesized data on Dakar region includes 4281 data entries from four urban districts, including Dakar, Guediawaye, Pikine, and Rufisque (Table 6). While there are a few entries that go as far back as 1989, the database becomes more populated as of 2002. Road accidents are overwhelmingly the largest number of entries, totalling 82% of all events. Fires, drowning, structural collapse, industrial

7 Douglas et al (2008) report that there were 46 fatalities from floods and landslides in Nairobi in 2002, over a 2-week period of heavy rains alone. This detail on fatalities (and an additional hazard of landslides) illustrate the challenge of under-reporting in formal databases as DesInventar.

disasters and flooding are also prevalent. Drowning caused 441 deaths, road accidents caused 257 deaths and industrial disasters caused 147 deaths (in two incidents). Drowning is a major cause of death due to ferry accidents. Over 390,000 people (12% of population) were recorded as being affected by flooding and it accounted for 99% of all housing damages and destruction. This looks like quite a well-developed database compared to the others, in that it covers a lot of different kinds of events; this is more the kind of profile we would expect to see in a city. However, it is likely the numbers of events are still hugely under-reported.

Table 6: Losses by disaster type in Dakar, 1989-2015 (DesInventar)

Event Data Cards Deaths Injured Missing s S s 0 i 1 t c c Houses Damaged Affected

Coastal Erosion 1 0 2 0 0 0 0

Drowning 249 441 56 14 0 0 0

Epidemic 8 0 0 0 0 0 599

Fire 404 33 40 0 21 169 1800

Flood 47 19 7 0 3 63862 390239

Industrial Disaster 2 147 0 0 0 0 0

Locust 2 0 0 0 0 0 915

Road Accident 3890 257 5619 0 0 2 0

Sinking 4 10 4 0 0 0 0

Storm 2 0 0 1 0 0 0

Strong Wind 7 2 0 9 0 0 0

Structural Collapse 92 36 301 0 9 49 0

Swell 1 0 0 3 0 0 0

Wild Fire 5 0 0 0 0 0 0

Freetown

The Western Area of Sierra Leone includes the capital, Freetown as well as the Western Rural Area. The database has entries from 2006, and then 2009-2014 (Table 7). Deaths are attributed to all of kinds of events, although maritime accidents (235 deaths), road accidents (174 deaths) and epidemics (215 deaths) are the most prevalent, accounting for 79% of all deaths recorded in the database. Fire accounts for 83% of all the houses destroyed/damaged and flooding for 7%. The amount of losses from maritime accidents and fires is notable in this database. Again, like the Dakar database, there is a wide range of events, however the losses are likely under-reported.

Table 7: Losses by disaster type in Western Area, Sierra Leone, 2006, 2009-2014 (DesInventar).

Event Data Cards Deaths Injured Missing Houses Destroyed Houses Damaged Affected

Accident 9 9 4 0 0 0 0

Accident (Maritime) 119 235 52 0 0 0 3962

Accident (Road) 46 174 200 0 4 3 5

Conflict 10 1 47 0 10 4 1000

Drought 2 0 0 0 0 0 500

Drowning 2 7 0 0 0 0 0

Epidemic 46 215 39 0 0 0 1117

Fire 241 30 233 0 1338 457 10546

Flood 50 27 29 0 310 0 2308

Landslide 16 57 50 0 20 0 99

Lightning/ Electrical Storm 2 5 0 0 0 0 0

Storm/ Gale 40 14 69 0 944 654 5999

Structural collapse 5 4 10 0 0 1 0

Findings

This analysis of DesInventar for urban areas on SSA brings up some findings about the utility of this database for understanding disaster losses in urban areas:

• DesInventar data provides a general idea of the subnational distribution of disasters, but there is a lack of consistent data in the database, which makes accurate conclusions about any of the cities problematic. For an individual city, some kinds of events are reported in detail, whereas other kinds of events are not reported at all. The cities analysed all show quite different events and levels of losses. Some of this variability comes from the differences across the cities, for example, maritime accidents are a major loss of life in Freetown and measles is prevalent in Niamey (and likely to be a problem in other cities but not recorded). However, we assume that much of the variation in results across the cities is due to a lack of comprehensive data in the database.

• The methodology of DesInventar could be very useful for understanding a range of different kinds of disaster events that are typical in urban areas if the database was accurately populated. DesInventar picks up very small-scale events (for example a house fire or an accident). If a good level of data was reported in the database, it would provide a very telling picture about the extent of losses to a range of types of events within different areas of the city and could be a very useful tool for policy-making. Most of this data does actually exist in cities, held by police, hospitals, Red Cross and other organisations, but it is not being systematically reported in this database, and more effort to made to combine or cross reference data held at these local sources.

• DesInventar does not usually report on "everyday" human losses related to endemic conditions caused by environmental factors, such as deaths from diarrhoeal disease (unless it is classified as an epidemic, eg. cholera), malaria or deaths resulting from a lack of medical attention and morbidity. It is this everyday risks and health perspective that is needed to extend the range of analysis about human losses to include those related everyday hazards which would lead to a more comprehensive accounting of extensive risk losses.

Thus, what is needed is more comprehensive information collected at the city-level about the causes of pre-mature deaths and other losses. DesInventar is potentially a powerful for tool for analysing risks in an urban area, but the databases for SSA cities would need to be properly populated, using locally available data. Currently data is not being analysed and collated in this way in most urban jurisdictions in SSA. Secondly, DesInventar could be even more comprehensive if it included health

records from local hospitals that could account for the endemic health conditions caused by environmental factors.

The next section of this paper turns to look at what kinds of information are available from the health sector that could shed more light on human losses in urban areas of SSA, particularly regarding everyday hazards. As mentioned earlier, it is the accumulation of these extensive every day human losses that accumulate and lead to bigger losses as compared to large disasters. Also, if premature death from diseases were included as everyday disasters it would change completely the picture of risk and vulnerability.

5.0 Data on health impacts in urban SSA

Health conditions and diseases (endemic and epidemic) in SSA have been documented by the World Health Organisation (WHO) through its Global Burden of Disease Project and by USAID through its Demographic and Health Survey (DHS) programme. The DHS programme provides data for urban and rural areas across most African countries. Data from the DHS permits analysis of indicators on health and its determinants at country level8. The WHO regional office for Africa also publishes the African Health Observatory, which provides comprehensive statistics on disease prevalence and burden across the region. The information is reported at country level, and some data are separated into rural and urban, but it is not possible to derive details of particular urban areas9. From the perspective of those who need to understand the most important risks facing their city, the lack of urban details is a major shortfall.

Available data compiled by the WHO shows that 43% of deaths in 2004 were caused by infectious and parasitic diseases (such as diarrhoea, respiratory diseases, malaria, etc.), many of which have large environmental contributions (Figure 1). This figure rises to 57% when respiratory infections are considered.

https://dhsprogram.com 9 See the latest 2014 Atlas of Health Statistics

http://www.aho.afro.who.int/sites/default/files/publications/921/AFRO-Statistical_Factsheet.pdf (last accessed 15/12/15).

Perinatal Conditions 8%

Respiratory infections 14%

Maternal conditions 5%

Disease prevalence for females

Nutrticna! denciences 1%

D aretes 2% Respiratory 2%

Circulatory 12%

Other NCD 6%

Respiratory infections 13%

Perinatal Conditions 9% \

Disease prevalence for males

Injuries 10%

Nutrticna! dencienc es 2%

Diabetes 1% Resp ratory 3%

Circulatory 9%

Other NCD 6%

Figure 1: Disease prevalence and percentage of mortality for females and males in Africa (redrawn from WHO, 2004)

There was similar distribution in individual SSA countries. For instance, the mortality distribution for Kenya (also in 2004) reported an average of 52% mortality from infectious and parasitic diseases. Looking into the prevalence of disease epidemics, the global distribution of recent cholera cases reported between 2010 and 2013, revealed that a majority come from SSA. A more detailed look at six countries (Kenya, Malawi, Niger, Nigeria, Senegal and Uganda) showed a total of 3,278 reported cases. The data on the reported number of deaths in these countries within the 3-year period was incomplete, although a comparison between the data that was available in both categories showed an average loss of 28 deaths per documented outbreak of cholera (WHO, 2004). But while much attention has been paid to disease epidemics, which have in some cases resulted in enough impacts to be classified as 'disasters', infectious and parasitic diseases (e.g. diarrhoea and malaria) tend to have larger cumulative health impacts over time.

Another example of major health losses in SSA is from the prevalence of malaria. Evidence suggests that those living in informal settlements face a heightened risk (Donnelley et al, 2005)10 and that there is a positive correlation between malaria prevalence and environmental conditions related to improper waste and sewerage channels at the household scale (Nkuo-Akenji et al, 2006). In Dar es Salaam, de Castro et al. (2004) found that a large number of breeding sites for the anopheles mosquito were concentrated in lower elevation parts of the city or near drains that require cleaning or rehabilitation, and that entomologic inoculation rates (i.e. the number of infective bites per person per year) were likely to be significantly higher in informal settlements and marginal localities near the periphery. From WHO records (2004), there were an estimated 627,000 malaria deaths worldwide in 2012 (uncertainty interval, 473 000-789 000), and out of this estimate, most occurred in sub-Saharan Africa (90%).

These losses from infectious and parasitic diseases may not be categorised as "events" in disaster databases because they occur regularly and are generally too small to meet the criteria to be recorded, nor do we have information about events in specific urban centres.

WHO introduces an interesting dimension of losses, which provides more depth into the conceptualisation of losses from disease prevalence. The measure is termed as the Disability-adjusted life years (DALYs), which is a summary of the disease-burden on a population accounting for the years of life lost as a result of premature death, disability or the years lost due to ill health (WHO, 2012). It gives a more detailed appreciation into how diseases that may not result in death have significant impact on the livelihoods and productivity of its sufferers.

The DALYs measure has recently been adapted and used to the impact of disasters and it promises to provide a view into a dimension of losses that is not accounted for by the focus on mortality, morbidity or damage to houses and infrastructure (Noy, 2015). The application of this novel DALYs to better understand the effects of extensive risk in SSA could be explored further.

Studies focusing on urban areas

One of the few detailed urban studies on health and population dynamics conducted in Nairobi, Kenya, by the Africa Population Health Research Centre (APHRC) showed that in 2012 residents sampled from a cross-section of slums, were generally disadvantaged compared to the rest of Nairobi and Kenya (APHRC, 2014). For instance, the lack of good drinking water and poor drainage were cited as the major needs for the slum residents, as access to good water was a problem for one-in-five slum residents in Nairobi (APHRC, 2014). The health risks and subsequent losses in these urban slums were on occasion even higher than those of rural Kenya. For instance child mortality and under-5 mortality rates in the urban slums were higher than the rates in rural Kenya, and Kenya as a whole (APHRC, 2014)11.

10 Data presented from studies in a number of sub-Saharan African cities (Brazzaville, Congo; Dakar, Senegal; Abidjan, Cote d'Ivoire; Cotonou, Benin; Ouagadougou, Burkina Faso; Dar es Salaam, Tanzania, and Accra and Kumasi, Ghana) showed clearly that malaria is a considerable urban health problem in Africa (Donnelley et al, 2005).

11 This trend was similar to surveys and studies conducted in 2000, where Under-5 and infant mortality rates in Nairobi's slums were about 20 and 35 %, respectively, higher in the slum communities of Nairobi compared to rural Kenya. AHPRC (2012) Urban Health in Kenya Key Findings: The 2000 Nairobi Cross-Sectional Slum Survey.

The usefulness of the above study is that they included a representative sample of informal settlement dwellers (which the DHS do not) and the disaggregated nature of the data shows other interesting demographics of the sampled populations (such as the subdivisions of the urban area, age, ethnic group, etc) that helps to understand the depth of urban health risk and disease burden. Again the methodology employed allows for an appreciation of other significant background information (on households and the population) that underlies these risks.

The lesson and challenge for disaster researchers working on SSA, from this example of the APHRC study would be to produce comprehensive data to cover the broad range of risks (including the health) and other extensive everyday risks that urban populations in the study sites face, with sufficient background data that enables for a broader knowledge of the risks and where possible the underlying factors that account for them.

6.0 Other databases and data sources on risks and everyday health losses

There are many other data sources that could be used to better understand the nature and scale of disaster and health related losses in urban areas in SSA. Table 10 summarises several databases that capture the losses associated with everyday hazards, and small and large disasters at different scales12.

Table 10: Major data-bases for everyday hazards and small and large disasters at different scales

Scale Everyday hazards Small disasters Large disasters

International and regional WHO Global Burden of Disease indexes, including road traffic accidents by region and country EM-DAT NatCat Global Disaster Identifier Number (GLIDE) Disaster Database Project Asian Disaster Reduction Centre (ADRC)

National The Demographic and Health Survey (DHS); have data for rural and urban areas but not for specific cities let alone informal settlements DesInventar - African coverage: Comoros, Djibouti, Ethiopia, Kenya, Madagascar, Mali, Mozambique, Morocco, Mauritius, Niger, Togo, Tunisia, Senegal, Sierra Leone, Seychelles, Uganda, and Tanzania (Zanzibar) National databases (e.g. Australia, Canada, Nepal, Orissa, Philippines, St. Lucia, Sri Lanka, etc.)

Sub-National (urban and rural) Demographic and Health Surveillance Systems (DHSS) Desinventar, but urban districts not always differentiated from rural

Fact Sheet. http://aphrc.org/wp-content/uploads/2014/10/Urban-Health-in-Kenya_Key-

Findings_2000-Nairobi-Cross-sectional-Slum-Survey.pdf (last Accessed 16/12/15).

Urban ARK researchers have created a database of data sources relevant to understand risk in SSA, which can be accessed from here http://tinyurl.com/africa-datasets. These data sources provide information on all three of the ways of understanding risks (losses, vulnerability and natural/technological hazards).

districts

Individual Hospital episode data (mostly relevant to everyday hazards, but potentially useful for uncovering the health impacts of larger events) Health passports for some countries (e.g. Malawi)

Event specific Police records DesInventar records loss and damage from multiple hazards that can be disaggregated Earthquakes: The United States Geological Survey (USGS) Technological disasters: Awareness and Preparation for Emergencies on a Local Level (APELL) Floods: Dartmouth Flood Observatory (DFO and United States Weather Service (NWS) Tsunami: National and Geophysical Data Center (NGDC) Industrial Accidents: Major Accident Reporting System (MARS, Major Hazard Incident Data Service (MHIDAS)

National databases

Vital Registration Systems collect data on births and deaths at country-level; however, few are up-to-date in low- and middle-income countries (UN 2014). Nor are vital registration systems linked to population registers (which combine different sources of data, including from censuses) and Demographic and Health Surveys, etc.) for countries or urban and rural and areas. Consequently, comprehensive current information on the demographic characteristics and health outcomes for each resident of a given country, town or city, is often unavailable.

Sub-national databases

Demographic and disease surveillance systems collect data on disease burdens and their distribution for small geographic units (e.g. neighbourhood, town, city, district, etc.) (Nkuchia et al., 2015: 3). Such systems provide a critical foundation for disease prevention and control, but seldom exist in lower-income countries/towns or cities, and even less in informal settlements, with the notable exception of Nairobi (Emina et al., 2011).

Urban Health Observatories (UHOs) also monitor heath and its determinants at the intra-urban scale by collecting and compiling secondary data from censuses and Demographic and Health Surveys, among other data sources, and by analysing this data using remote sensing and GIS mapping. UHOs aim to provide urban policymakers with detailed data and maps showing the social and spatial distribution of disease burdens as a basis for informing targeted investment in deprived urban localities. Several pioneering cities, such as Belo Horizonte, Brazil (see Vlahov et al., 2011), have developed UHOs, but few exist (or have been documented) in SSA.

Household/individual databases

Hospital episode data provide details of the individual admitted (including age and gender), where they were admitted, and the diagnoses made. However, details about where the individual admitted lives or where their health was affected are often unrecorded. This makes it difficult to examine the potential links between social and environmental factors and health outcomes.

Hospital admission and mortality counts can be used to show temporal trends associated with exposure to different seasonal or climatic conditions in the short- or long-term, which could be of potential use for urban research and policy on climate change adaptation (Scovronick et al., 2015). Hospital data where residential details of patients are recorded, could also be cross-referenced with particular disaster events to examine the number of people whose health may have been affected.

Event-specific databases

Numerous data sources collect event-specific information on particular disaster hazards (e.g. floods, earthquakes, technological disasters, etc.) and everyday hazards. Data sources on the former are listed in Table 10 and are relatively straightforward. Data sources on the latter include police reports on crime specific events (e.g. homicide, assault) and accident specific events (e.g. road traffic accidents). Data on road traffic accidents at the national level make international comparisons possible, as demonstrated by the WHO global observatory13, although this data are not at a fine enough scale to permit sub-national analysis.

Newspaper records collect varied information on newsworthy events, including crime and violence, road traffic accidents, and disasters and their effects (e.g. number of people killed or displaced, number of buildings damaged or destroyed, etc.). But newspapers tend to cover events that warrant media attention, thus excluding smaller everyday events (e.g. localised disease burdens). Their archival value as a long-term secondary source may be limited by the priority to provide timely and accurate reporting of events as well (Yzaguirre et al., 2015).

7.0 Conclusions

Several key data limitations can be summarised from the databases and data sources reviewed above.

Lack of collated datasets on disasters that can show extent of losses in urban areas

This analysis has shown that there is some data available about disasters, everyday risks and health-related losses in sub-Saharan Africa, and that some of this data can shed some light on losses in urban areas. But there is still a major lack of collated data that can accurately show losses in urban areas, and even less in informal settlements. In DesInventar, many countries are yet to sign up to report national datasets, and the quality of data, in terms of being able to understand the losses in a particular urban area, is still weak. We contend that much of the data does exist (in accident records, police, fire and hospital records and news media and in some locations vital registration systems), but needs to be collated into databases in order to be useful for local analysis, planning and policymaking.

13 http://www.who.int/violence_injury_prevention/road_traffic/en/

Assimilating disasters losses and endemic health-related losses

Disasters are usually considered episodic, and thus there is a focus on disaster events as an occurrence of a particular phenomenon or losses over a period of time. As mentioned earlier, different databases assume different loss thresholds for considering an event as a 'disaster'; for example EM-DAT considers 10 or deaths or 100 or more people affected/injured/homeless. Desinventar considers much smaller events—1 or more human loss or $1 or more in economic loss, and therefore is more useful for characterising the overall human and economic impacts of risks.

However, none of the disasters methodologies go so far as to consider endemic kinds of health losses that occur everyday and are therefore not considered to be episodes or events; for example a death related to malaria or maternal health. We contend that, in order to have the full picture of losses for an urban area, we would need to understand the full range of disasters as well as everyday health impacts. The availability of this kind of information would be a powerful policy and decisionmaking tool.

Lack of disaggregated data for urban areas

The demographic and health surveys uses aggregate data to achieve national representation, and while they do report for urban and rural areas, the data does not allow for an analysis of losses in a particular urban area. Nor does it shed light on health inequalities across high- and low-income areas within an urban population). Desinventar does disaggregate data to potentially permit an analysis of urban areas, but the detail, coverage and consistency of data remains lacking.

As Mitlin and Satterthwaite (2013: 104) point out, the lack of disaggregated data can be partly overcome by examining census data on vulnerability factors and the determinants of health linked to housing conditions, access to basic services, space per person, cooking fuels used, protection against extreme weather, etc. This data can shed light on neighbourhoods or areas where social and vulnerability factors that make them more susceptible to disasters and everyday hazards are high, and where investments to address these underlying factors need to be made. The use of census data has also been promoted to assess vulnerability to climate change at the intra-urban scale (Schensul and Dodman 2013), but many census authorities do not provide census data in a form that allows for a proper identification of informal urban settlements or the detailed analysis of each urban centre and by small area units within each centre - that would show the scale and nature of intra-urban inequality.

Limited spatial coverage

While Desinventar has expanded to include 15 countries in SSA, its geographic coverage remains limited in the region. Spatial coverage within SSA countries is also limited by the different ways in which 'urban' and 'rural' settlements are classified14. In countries where many small urban centres

14 Satterthwaite (2015b) discusses the problem of how 'small urban centres' are often not accounted for in major databases because the criteria that national statistical offices use to classify 'urban' and 'rural' populations - for example, settlement size, administrative importance, economic structure - often mean that small settlements are classified as 'rural' or as 'large villages' rather than as 'urban' or 'small urban centres'. This has significant implications for SSA where a large proportion of the urban population live in towns of less than 20,000 inhabitants that could be classified as either urban or rural.

are classified as 'rural' or as 'large villages', the proportion of damage and losses from disasters occurring in smaller urban centres is likely to be under-estimated. This is particularly problematic considering that disaster risk may be increasing fastest in small and medium-sized urban centres since their capacities to plan and manage urban growth are relatively weak (UNISDR 2011).

Tying loss data to the underlying drivers of risk

Acting on loss data also requires making links into the underlying drivers of those losses; for example, collecting data on social factors (e.g. age, gender, income, ability, migrant status, etc.), environmental factors (e.g. access to quality housing, basic services, etc.), and political and institutional factors related to planning and decision-making processes at multiple levels. External factors linked to climate change must also be better understood in relation to internal factors linked to urban growth and change, and poverty. But information on climate change has its own limitations, and is seldom used directly for urban decision-making, as observed in Accra and Maputo (Steynor et al., 2015).

If sub-national databases become more precise and comprehensive in capturing urban loss and damage from everyday hazards and disasters (small and large), it is highly likely that the observed trends would reinforce the view that urban risks are increasing fastest in urban areas, in particular those where local governments are unable to effectively plan and manage urban growth, and unwilling to address the needs of their low-income populations (IFRC 2010; Satterthwaite et al., 2007).

Areas of interest for research

There are several areas of interest for researchers that are identifiable from the analysis above. Firstly, we think more needs to be done to bring the health and disasters research (and sectors) together to be able to better address risks. There is the need for further research on losses related to everyday hazards (which would be picked up by health data), and small and large disasters, and more research to assess their relative importance for urban losses. This could lead to the inclusion of health measurements and metrics in disasters databases, which we think should be seriously considered as it would give a much more nuanced picture of the spectrum of losses people are facing. Secondly, and related to the point above, for individual cities, more research needs to be done that combines different data sources (e.g. DesInventar, newspaper archives, hospital data, police reports, Demographic and Health Surveys, etc.) to create a more detailed and comprehensive picture of urban risk. Where detailed demographic and health data is lacking this research could consider using census data to examine underlying social and environmental risk factors within and between urban populations. There are also lessons to be learnt from the experience of African cities that have implemented DesInventar and disease surveillance systems, and identify the data sources and lessons they present for developing innovative research methodologies. Thirdly, there is a potential of using DALYs and other health metrics to uncover the significant repercussions that disease burdens often have for low-income urban populations (see Kovats et al., 2014). Fourthly, on the concerns of data quality raised, it will be useful for more disaggregated data collection to uncover the social and spatial distribution of urban risk within and between populations of small and intermediate urban centres and large cities.

Policy Implications:

Our analysis in this paper is useful for policy and decision-makers in SSA cities, particularly planners and those working on disaster risk reduction to acknowledge this critical issue of the lack of reliable disaggregated data and the need for innovative approaches to policy formulation and implementation in the interim, as datasets are improved. It draws attention to the need for a shift from disaster recovery oriented policies to focus more on mitigation, preparedness and basic services that address everyday risks, alongside the paradigm shift that the cumulative losses from smaller everyday risks were higher than large-scale intensive risks.

Also, in the absence of detailed disaggregated data in SSA cities, there is the need for policy makers to adopt more grassroots participatory approaches to verify and triangulate information and involve different interest groups in the process of policy formulation and data collection to create more inclusive and socially just outcomes. Georeferenced granular data is important for city planners in producing risk sensitive policy and interventions in sectors that interact with the production of everyday risks such as healthcare provision, transport policy, and provision of water and sanitation infrastructure.

References

Adelekan, I., Johnson, C., Manda, M., Matyas, D., Mberu, B.U., Parnell, S., Pelling M., Satterthwaite, D. and Vivekananda, J. (2015). Disaster risk and its reduction: an agenda for urban Africa. International Development Planning Review. Vol. 37, No.1, pp. 33-43. doi:10.3828/idpr.2015.4

African Population and Health Research Center (APHRC). 2014. Population and

Health Dynamics in Nairobi's Informal Settlements: Report of the Nairobi Cross-sectional Slums

Survey (NCSS) 2012. Nairobi: APHRC

AHPRC (2012) Urban Health in Kenya Key Findings: The 2000 Nairobi Cross-Sectional Slum Survey. Fact Sheet. http://aphrc.org/wp-content/uploads/2014/10/Urban-Health-in-Kenya_Key-Findings_2000-Nairobi-Cross-sectional-Slum-Survey.pdf (last Accessed 16/12/15).

AU (African Union), UNISDR (United Nations International Strategy for Disaster Reduction), and World Bank. 2008. Status of Disaster Risk Reduction in the Sub-Saharan Africa Region, by Rakhi Bhavnani, Seth Vordzorgbe, Martin Owor, and Franck Bousquet. Washington, DC: World Bank.

Bara, C. (2010) 'Social Vulnerability to Disasters: Factsheet', Crisis and Risk Network (CRN), Center for Security Studies (CSS), ETH Z urich.

City Population www.citypopulation.de (last accessed 20/06/17).

de Castro, M.C., Yamagata, Y., Mtasiwa, D., Tanner, M., Utzinger, J., Keiser, J. and Singer, B.H. (2004) 'Integrated Urban Malaria Control: A Case Study in Dar es Salaam, Tanzania', Am. J. Trop. Med. Hyg., 71(2), pp. 103-117.

Donnelly, M. J., McCall, P. J., Lengeler, C., Bates, I., D'Alessandro, U., Barnish, G., Konradsen, F., Klinkenberg, E., Townson, H., Trape, J., Hastings, I. M. and Mutero, C. (2005) 'Malaria and urbanization in sub-Saharan Africa', Malaria Journal 4:12

Douglas, I., Alam, K., Maghenda, M., Mcdonnell, Y., Mclean, L., and Campbell, J. (2008) 'Unjust waters: climate change, flooding and the urban poor in Africa', Environment & Urbanization Vol 20(1): 187-205.

Emina J, Beguy D, Zulu EM, et al. (2011) Monitoring of Health and Demographic Outcomes in Poor Urban Settlements: Evidence from the Nairobi Urban Health a2nd Demographic Surveillance System. Journal of Urban Health 88: 200-218.

Global Risk Information Platform (GRIP) www.gripweb.org (last accessed 15/12/15). Gupa-Sapir, D. and Hoyois, P. (2012) 'Measuring the Human and

E conomic Impact of Disasters', Report produced for the Government Office of Science, Foresight project 'Reducing Risks of Future Disasters: Priorities for Decision Makers'.

Freire, M. E., Laa, S. and Leipziger, D. (2014) 'Africa's Urbanisation: Challenges and opportunities'

International Federation of the Red Cross (2010)

http://www.ifrc.org/docs/appeals/10/MDRZM007fr.pdf (accessed 08/12/15).

IFRC. (2010) World Disasters Report 2010: Focus on urban risk. Geneva: IFRC.

Integrated Research on Disaster Risk. (2015). Guidelines on Measuring Losses from Disasters: Human and Economic Impact Indicators (IRDR DATA Publication No. 2). Beijing: Integrated Research on Disaster Risk.

Integrated Research on Disaster Risk. (2014). Peril Classification and Hazard Glossary (IRDR DATA Publication No. 1). Beijing: Integrated Research on Disaster Risk.

Kabisch, S., Jean-Baptiste, N., John, R. and Kombe, W. J. (2015) 'Assessing Social Vulnerability of Households and Communities in Flood Prone Urban Areas' in Pauleit, S., Coly, A., Fohlmeister, S., Gasparini, P., J0rgensen, G., Kabisch, S., Kombe, W. J., Lindley, S., Simonis, I., and Yeshitela, K. (Eds) Urban Vulnerability and Climate Change in Africa: A Multidisciplinary Approach. London:Springer.

Kovats S, Lloyd S and Scovronick N. (2014) Climate and health in informal urban settlements. Working Paper, November 2014. London: IIED.

Mitlin, D and Satterthwaite D (2013) Reducing Urban Poverty in the Global South, Oxford:Routledge.

Nkuchia M, M'ikanatha M and Iskander JK. (2015) Concepts and Methods in Infectious Disease Surveillance, Chichester, West Sussex: John Wiley & Sons.

Nkuo-Akenji, T., Ntonifor, T. N., Ndukum , M. B., Kimbi, H. K., Abongwa, E. L., Nkwescheu, A., Anong, D. A., Songmbe, M., Boyo, M. G., Ndamukong, K. N. and Titanji, V. P. K. (2006) ' Environmental factors affecting malaria parasite prevalence in rural Bolifamba, South- West Cameroon', African Journal of Health Sciences, Volume 13, Number 1-2.

Noy, I. (2015) 'A DALY Measure of the Direct Impact of Natural Disasters', VOX C EPR Policy Portal, http://www.voxeu.org/article/daly-measure-direct-impact-natural-disasters (accessed 15/12/15).

Osuteye, E., Johnson, C., and Brown, D. (2016) The data gap: An analysis of data availability on disaster losses in sub-Saharan African Cities, Urban Africa Risk Knowledge Working Paper No. 11.

Pelling M and Wisner B. (2009) Disaster risk reduction: cases from urban Africa, London: Earthscan.

Satthertwaite, D. (2005) 'Urbanisation in sub-Saharan Africa: trends and implications for development and urban risk', Urban Transformations,

http://www.urbantransformations.ox.ac.uk/blog/2015/urbanization-in-sub-saharan-africa-trends-and-implications-for-development-and-urban-risk/ (accessed 15/12/2015).F

Satterthwaite D, Huq S, Pelling M, Reid, H and Lankao, P R (2007) Adapting to Climate Change in Urban Areas: The possibilities and constraints in low- and middle-income nations. Human Settlements Discussion Paper Series. Theme: Climate Change and Cities - 1. London: International Institute for Environment and Development.

Schensul D and Dodman D. (2013) Populating Adaptation: incorporating Population Dynamics in Climate Change Adaptation Policy and Practice. In: Martine G and Schensul D (eds) The Demography of Adaptation to Climate Change. New York, London and Mexico City: UNFPA, IIED and El Colegio de México, 1-23.

Scovronick N, Lloyd SJ and Kovats RS. (2015) Climate and health in informal urban settlements. Environment and Urbanization 27: 657-678.

Steynor, A., Jack, C., Padgham, J. and Bharwani, S. (2015) Using climate information to achieve long-term development objectives in coastal Ghana and Mozambique. CDKN Policy Breif. January 2015.

Sverdlik A. (2011) Ill-health and poverty: a literature review on health in informal settlements. Environment & Urbanization 23: 123-155.

UNDESA. (2015) World Urbanization Prospects, 2014 Revision. New York: United Nations. UN-Habitat (2010) The State of African Cities. Nairobi: UN-Habitat.

UN-Habitat. (2006) Meeting Development Goals in Small Urban Centres: Water and Sanitation in the World's Cities 2006. Nairobi: UN-Habitat.

UNISDR (2011) 2011 Global Assessment Report on Disaster Risk Reduction: Revealing Risk, Redefining Development. Geneva: UNISDR.

UNISDR (2013) From shared risk to shared value - The business case for disaster risk reduction. Global Assessment Report on Disaster Risk Reduction. Geneva, Switzerland: United Nations Office for Disaster Risk Reduction (UNISDR).

UNISDR (2015) 'Making Development Sustainable: The future of Disaster Risk Management. Global Assessment Report on Disaster Risk Reduction. Geneva, Switzerland: United Nations Office for Disaster Risk Reduction (UNISDR).

Vlahov D, Agarwal S, Buckley R, et al. (2011) Roundtable on Urban Living Environment Research (RULER). Journal of Urban Health 88: 793-857.

World Bank, 2009: Climate Resilient Cities: A Primer on Reducing Vulnerabilities to Disasters. By Neeraj Prasad, Federica Ranghieri, Fatima Shah, Zoe Trohanis, Earl Kessler, Ravi Sinha. Washington.

World Bank (2010) Report on the status of Disaster Risk Reduction in Sub-Saharan Africa http://www.gfdrr.org/sites/gfdrr/files/publication/AFR.pdf

World Health organisation (2004) 'World infobases'

https://apps.who.int/infobase/Mortality.aspx?l=&Group1=RBTCntyByRg&DDLCntyByRg=AFR&DDLC ntyName=1001&DDLYear=2004&TextBoxImgName=go (accessed 14/12/15).

Yzaguirre, A., Warren, R. Smit, M. (2015) Detecting environmental disasters in digital news archives. Paper presented at Big Data (Big Data), 2015 IEEE International Conference.