Impact-weighted multi-hazard disaster hotspots index. Piet Buys and Uwe Deichmann Development Research Group Infrastructure & Environment World Bank

Transcription

1 Impact-weighted multi-hazard disaster hotspots index Piet Buys and Uwe Deichmann Development Research Group Infrastructure & Environment World Bank

2 Hotspots indicators rather than one single indicator, build up a series of indexes of increasing complexity maps of areas with high probability of a natural disaster occurring measurement is difficult (how to define hazardousness based on frequency and magnitude?) combine this with areas where elements at risk are concentrated coincidence of high disaster probability and high predicted impacts (on people, economically and socially productive assets)

3 Incorporating vulnerability propensity to be damaged when subjected to a hazard event use observational data: information on past impacts EM-DAT has records of mortality, persons affected and direct economic damage epidemiological approach based on mortality rate (extension to economic loss is straightforward)

4 Mortality rates compute mortality rates using EM-DAT cumulative number of persons killed by a given hazard and divide by the total population in the area exposed to that hazard e.g. globally, for storms (prelim. estimates): 243,551 fatalities between 1981 and ,312 million people in exposed area in fatalities per 100,000 population (note time periods) we can apply this rate to the population grid in areas exposed to the hazard to produce an estimate of expected fatalities over a 20 year period

5 Geographic variations in mortality but: mortality is not distributed uniformly e.g., earthquake of a given magnitude does more damage in India than in Japan social, economic and physical factors that reduce vulnerability: building codes, emergency response, education, topography, geology many of these are related to the wealth of a country use regionally specific rates

6 draft figures based on EM-DAT and GPW3, likely to be revised significantly! Estimated mortality rates fatalities per 100,000 inhabitants in 2000 Region Drought Earthquakes Floods Storms Landslides Volcanoes Total impact zone Non-OECD OECD AFR EAP ECA LAC MENA NAM SAS

7 Geographic disaggregation geographically and hazard specific mortality rates provide a better estimate of potential vulnerability country data too noisy, so we use WB regions classified into four income groups highest mortality rates: droughts: AFR low income earthquakes: ECA low middle income floods: LAC upper middle income storms: SAS low income (ignoring MENA outlier) landslides: LAC low middle income volcanoes: LAC low middle income

8 Geographic disaggregation World Bank regions by income group

9 Incorporating hazard severity some indication of how severely different areas are affected within exposed area measures of severity: estimates of frequency or probability, storm speed and duration, potential peak ground acceleration for earthquakes mortality rates will be higher in areas where severity measures are larger use severity as a weight to adjust mortality rates

10 In summary 1. mortality rate 2. weighted cell mortality 3. adjustment 4. multi-hazard where: h = hazard, i = grid cell, j = region M = mortality (EM-DAT), P = population (GPW3), W = hazard severity weight

11

12 Uniform global mortality rate log of mortality

13 Region specific mortality rate log of mortality

14 Region specific mortality rate weighted by hazard severity log of mortality

15 Global results although the model output presents an estimate of predicted cumulative mortality from all hazards over a twenty year period, we interpret it as a notional index (low high) hazard specific mortality-weighted indexes combined, multi-hazard hotspots index number of hazards contributing

16 Mask areas of low pop, non-ag 55 % of area, 99 % of population remains

17 Drought disaster mortality risk hotspots

18 Earthquake disaster mortality risk hotspots

19 Flood disaster mortality risk hotspots

20 Landslide disaster mortality risk hotspots

21 Storm disaster mortality risk hotspots

22 Volcano disaster mortality risk hotspots

23 Multi-hazard mortality risk hotspots (sum of predicted mortalities in each grid cell)

24 How many hazards contribute to the index? (number of high impact hazards affecting grid cell)

25 Drought disaster economic loss risk hotspots

26 Earthquake disaster economic loss risk hotspots

27 Flood disaster economic loss risk hotspots

28 Landslide disaster economic loss risk hotspots

29 Storm disaster economic loss risk hotspots

30 Volcano disaster economic loss risk hotspots

31 Caveats this is an intuitive approach and relatively easy to implement (but: it builds on many years of diligent data development!) main problem: weighting is ad hoc and deterministic need to know: what should be the cutoff for exposed area? at what level of severity does damage occur? how does damage vary with changes in severity?

32 Statistical determination of weights consider hazard severity as the dose and hazard impacts as the response requires ability to link specific hazard events (e.g., hurricanes) to their impacts (fatalities, economic damage) statistical estimation also yields measures of accuracy e.g., M h = β o + β 1 H h + β 2 X h + ε where M h = damage (mortality) from disaster event h H h = characteristics of the hazard leading to disaster X h = exposure and vulnerability characteristics of area affected = an estimate of severity weight W β 1

33 Conclusion impact-weighted multi-hazard hotspots index combines information on hazard extent, exposed elements and vulnerability (based on historic impacts) plenty of scope for refinement response function (feasible?) narrower definition of exposed area (hazards maps) better (more complete) damage estimates (EM-DAT) better definition of exposed economic assets

34 Interpretation

35

36

37 Global hazard data Hazard Storms Drought Floods Hazardousness Parameter Frequency by wind strength Precipitation less than 75% of median for a month period (WASP) Counts of extreme flood events Period Resolution Source(s) * 30 UNEP/GRID-Geneva PreView, DECRG processing 2.5 IRI Climate Data Library 1 Dartmouth Flood Obs. World Atlas of Large Flood Events Earthquake Exp. PGA (10% prob. of exceedance in 50 yrs) n/a sampled at 1 Global Seismic Hazard Program Freq. of earthquakes > 4.5 on Richter Scale sampled at 2.5 Smithsonian Institution Volcanoes Counts of volcanic activity Sampled at 2.5 Landslides Estimated annual prob. of landslide or avalanche UNEP/GRID-Geneva and NGDC n/a 30 Norwegian Geotechnical Institute

38 Number of people Gridded Population of the World, GPW II(I)

39 Characteristics of affected population Cities above 100k population new: Global Rural Urban Mapping Project (GRUMP) outputs Characteristics of people (and buildings) from surveys (LSMS)

40 Economic activity

41 Statistical determination of weights dose-response function could be any shape or form hazard impact hazard severity

42 Ingredients for disaster hotspots identification hazards drought, earthquakes, floods, landslides, storms, volcanoes probability, magnitude, duration elements at risk population, economic activity and assets (GDP) vulnerability how much elements at risk would be affected if hazard event would occur mortality, direct monetary damage