Transcript Creation of a hazard index - EM-DAT

Creation of a hazard index:
Overview of the Hotspots
methodology
Piet Buys
[email protected]
19 August 2005
5Th EM-DAT
Technical Advisory Group Meeting
Project Objectives
• Identification of natural disaster risk
hotspots at sub-national scales
• Initial focus:
 Drought, floods, tropical cyclones,
earthquakes, volcanoes, landslides
• Where do they occur?
• Where might damage be most severe
(mortality and economic)
19 August 2005
5Th EM-DAT
Technical Advisory Group Meeting
Project Objectives
• Prioritization for local vulnerability
assessments and risk reduction in
highest-risk areas
• Support Bank efforts to engage clients
in hazard management activities
(Turkey Earthquake Insurance, CAS, ...)
19 August 2005
5Th EM-DAT
Technical Advisory Group Meeting
Ingredients for Disaster Hotspots
Identification
Hazard information / event probabilities
at a given location, including probable magnitude,
duration, timing
Elements at risk
people, infrastructure and economic
activities/assets that would be affected if the hazard
occurred
Vulnerability of the elements at risk
how damaged they would be, if they experienced a
hazard event of some level
19 August 2005
5Th EM-DAT
Technical Advisory Group Meeting
Global Hazard Data
Hazard
Hazardousness
Parameter
Period
Resolution
Source(s)
Storms
Frequency by wind
strength
19802000
30”
UNEP/GRID-Geneva PreView,
DECRG processing
Drought
Precipitation less than
75% of median for a 3+month period (WASP)
19802000
2.5°
IRI Climate Data Library
Floods
Counts of extreme flood
events
19852003*
1°
Dartmouth Flood Obs. World
Atlas of Large Flood Events
Earthquake
Expected PGA (10% prob.
of exceedance in 50 years)
n/a
sampled at 1’
Global Seismic Hazard
Program
Freq. of earthquakes > 4.5
on Richter Scale
19762002
sampled at
2.5’
Smithsonian Institution
Volcanoes
Counts of volcanic activity
79-2000
Sampled at
2.5’
UNEP/GRID-Geneva and
NGDC
Landslides
Estimated annual prob. of
landslide or avalanche
n/a
30”
Norwegian Geotechnical
Institute
19 August 2005
5Th EM-DAT
Technical Advisory Group Meeting
Global Data on
Elements at Risk
Exposure
Parameter
Period
Resolution
Land
Land area
2000
2.5”
GPW Version 3 (beta)
Population
Population counts / density
2000
2.5”
GPW Version 3 (beta)
Economic
Activity
National / subnational GDP
2000
2.5”
World Bank DECRG
Agricultural
Activity
National agricultural GDP
allocated to agricultural land
area
2000
2.5”
IFPRI
Road Density
Length of major roads and
railroads
c. 1993
2.5”
VMAP(0)
19 August 2005
5Th EM-DAT
Technical Advisory Group Meeting
Source(s)
Global Data on elements at risk
• Focused on two in this study
 Population / mortality (shown below)
 GDP per unit area / economic losses (not shown)
19 August 2005
5Th EM-DAT
Technical Advisory Group Meeting
Global Data on Vulnerability of
the Elements at Risk
• Vulnerability estimates guided by past events
• EM-DAT has records of mortality, persons
affected and direct economic damage
 http://www.em-dat.net/
• epidemiological approach based on mortality
rate (extension to economic loss is
straightforward)
19 August 2005
5Th EM-DAT
Technical Advisory Group Meeting
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 :
 240,000+ fatalities between 1981 and 2000
 1,312 million people in exposed area in 2000
 16.6 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
19 August 2005
5Th EM-DAT
Technical Advisory Group Meeting
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
• Country data in EM-DAT is noisy
19 August 2005
5Th EM-DAT
Technical Advisory Group Meeting
Geographic disaggregation
• => use regionally specific mortality rates
WB regions classified into four income groups
• geographically and hazard specific mortality
rates provide a better estimate of potential
vulnerability
19 August 2005
5Th EM-DAT
Technical Advisory Group Meeting
Geographic disaggregation
World Bank regions by income group
19 August 2005
5Th EM-DAT
Technical Advisory Group Meeting
Incorporating hazard severity
• mortality rates will be higher in areas where
severity measures are larger
• some indication of how severely different areas
are affected within exposed area
• measures of severity: estimates of frequency or
probability, frequency by wind strength, expected
potential peak ground acceleration for
earthquakes
• use severity as a weight to adjust mortality rates
19 August 2005
5Th EM-DAT
Technical Advisory Group Meeting
In summary
1. mortality rate
2. weighted cell
mortality
3. adjustment
4. multi-hazard
where:
19 August 2005
h = hazard, i = grid cell, j = region_wealth
M = mortality (EM-DAT),
5
EM-DAT P = population (GPW3),
Technical Advisory Group Meeting
W = hazard severity weight
Th
Hurricane Severity and Intensity
19 August 2005
5Th EM-DAT
Technical Advisory Group Meeting
Uniform Global Mortality Rate
log of mortality
19 August 2005
5Th EM-DAT
Technical Advisory Group Meeting
Region Specific Mortality Rate
log of mortality
19 August 2005
5Th EM-DAT
Technical Advisory Group Meeting
Region Specific Mortality
Weighted by Hazard Severity
log of mortality
19 August 2005
5Th EM-DAT
Technical Advisory Group Meeting
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
• the same methodology can be applied to
economic losses (globally /proportion)
19 August 2005
5Th EM-DAT
Technical Advisory Group Meeting
Estimated Mortality rates
•highest mortality rates:
droughts:
AFR low income
earthquakes:
ECA low middle income
floods:
LAC upper middle income
storms:
SA low income
landslides:
EAP upper middle income
volcanoes:
LAC low middle income
•given the limited time period and quality of
input data => relative risk levels / deciles:
19 August 2005
5Th EM-DAT
Technical Advisory Group Meeting
Drought mortality risk hotspots
19 August 2005
5Th EM-DAT
Technical Advisory Group Meeting
Identification of areas affected by
multiple hazards
19 August 2005
5Th EM-DAT
Technical Advisory Group Meeting
All hazards mortality risk hotspots
note Africa vs. Europe
19 August 2005
5Th EM-DAT
Technical Advisory Group Meeting
All hazards total economic loss risk hotspots
note Africa vs. Europe
19 August 2005
5Th EM-DAT
Technical Advisory Group Meeting
All hazards Prop economic loss risk hotspots
note Africa vs. Europe
19 August 2005
5Th EM-DAT
Technical Advisory Group Meeting
Conclusion
• impact-weighted multi-hazard hotspots
index combines information on hazard
extent, exposed elements and vulnerability
(based on historic impacts)
• Scope for refinement
 Better weights / response function (feasible?)
 narrower definition of exposed area (hazards maps)
 better (more complete) damage estimates (EM-DAT)
 better definition of exposed economic assets
19 August 2005
5Th EM-DAT
Technical Advisory Group Meeting
Thank you
19 August 2005
5Th EM-DAT
Technical Advisory Group Meeting
19 August 2005
5Th EM-DAT
Technical Advisory Group Meeting
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.,
Mh = βo + β1 Hh + β2 Xh + ε
where
Mh = damage (mortality) from disaster event h
Hh = characteristics of the hazard leading to disaster
Xh =
exposure and vulnerability
characteristics of area affected
5
EM-DAT
19 August
2005
Technical Advisory Group Meeting
β1 = an estimate of severity
weight W
Th
Statistical determination of weights
hazard impact
• “dose-response function” could be any shape or
form
hazard severity
19 August 2005
5Th EM-DAT
Technical Advisory Group Meeting
Estimated
mortality rates
fatalities 1981-2000
per 100,000 inhabitants in 2000
19 August 2005
5Th EM-DAT
Technical Advisory Group Meeting
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?
19 August 2005
5Th EM-DAT
Technical Advisory Group Meeting
Mask areas of low pop, non-ag
55 % of area, 99 % of population remains
19 August 2005
5Th EM-DAT
Technical Advisory Group Meeting