topics geo 2003 - sfu.ca re(2004).pdf · natural catastrophes in 2003 – pictures of the year...

11th year · 2004

Natural catastrophes in 2003 | Trends of great natural catastrophes since 1950 | Hot summer in Europe – The future has already begun | The California wildland fires | The Bam earthquake in Iran | Geographical underwriting – Applications in practice | The climate summit 2003 in Milan

M Münchener Rück

Munich Re Group

ANNUAL REVIEW: NATURAL CATASTROPHES 2003

TOPICSgeo

1

ContentsMunich Re TOPICS geo 2003

01 Natural catastrophes in 2003

Review – OutlookPictures of the yearStatistics

02 Major engineering and fire catastrophes in 2003

03 Great natural catastrophes

Long-term statistics 1950–2003

04 NatCatSERVICE-Information

Weather-related natural catastrophes in the insured and uninsured world

05 Documentation:

Hot summer in Europe – The future has already begun

06 Chronicle:

The California wildland fires of 2003

07 Earthquake report:

The Bam disaster in Iran

08 Rare events?

09 Geographical underwriting –

Applications in practice

10 The 2003 climate summit and the Kyoto Protocol –

Can we wait for Russia?

InsertsWorld Map of Natural Catastrophes 2003

MRNatCatPOSTER Natural catastrophes in 2003

Page 2

Seite 4Page 8

Page 10

Page 12

Page 16

Page 20

Page 27

2

Page 38

Page 44

Page 50

Cover: 01 In the summer of 2003, weeks and weeks of heat and drought led to

historically low levels in many important shipping routes throughoutEurope. The picture on the front cover shows the Rheinkniebrücke, abridge that crosses the Rhine River in Düsseldorf: there was so littlewater at many places that transportation was severely impaired.

Left:02 After the heat came the flood: after the drought, the autumn brought

strong thunderstorms to central Europe. Almost the entire Rhôneregion of southern France was under water at the beginning ofDecember after torrential rain.

2

The overall balance of catastrophes in 2003 was again clearly dominatedby atmospheric events and weather-related natural catastrophes. Thegreatest sadness and grief, however,was caused around the globe by thesevere earthquake in Iran at the endof the year, which claimed the lives ofsome 40,000 people and was thus thenatural catastrophe with the largestnumber of victims last year. As in theprevious year, the statistics were alsomarked by an extreme weather eventof historical dimensions: the greatheat wave that hit central and south-ern Europe in the summer of 2003,which killed at least 20,000 peopleand generated economic losses farexceeding US$ 10bn.

Loss figures

More than 75,000 died throughout theworld as a result of natural catastro-phes. The number of natural hazardevents came to around 700 and thusreached the same level as 2002. Eco-nomic losses rose to over US$ 65bn(2002: US$ 55bn). Droughts and for-est fires in Europe and the UnitedStates – but also floods in Europe andAsia – had a major impact. Insuredlosses totalled US$ 15.8bn (previousyear: US$ 11.5bn; 5-year averageUS$ 16bn).

Storms and floods

300 of the approx. 700 events record-ed were windstorms and severeweather events. They accounted fortwo-thirds of the insured losses.

– Two devastating series of severeweather events and tornadoes des-troyed buildings and infrastructurein the US Midwest. More than 400single tornadoes were counted in

just one tornado outbreak in May. Amassive hailstorm in Texas will godown in US insurance history aftergenerating insured losses of overUS$ 1bn.

– In mid-September, Maemi, a strongtyphoon, tore across South Korea(insured loss: around US$ 500m).

– At almost the same time, HurricaneIsabel was battering the US EastCoast, where it damaged more than360,000 homes (economic loss:US$ 5bn; insured loss: US$ 1.7bn).Although the hurricane season inthe Atlantic was of average severityon the whole, there were two furtherevents which deserve particular men-tion, Hurricane Fabian and Hurri-cane Juan. Fabian, which generatedlosses of US$ 400m at the begin-ning of September, was the severesthurricane in the history of the Ber-mudas. In October and thus relative-ly late in the season, Hurricane Juandeveloped off the US East Coast,moved an unusually long way north-wards (44.5°N), and hit large parts of eastern Canada (economic loss:US$ 100m; insured loss: US$ 60m).

Europe was largely spared severestorms and other extreme weatherevents this year, although there wasquite a blustery start to the year. Atthe beginning of January, WinterStorm Calvann swept over France,Switzerland, and Germany. It onlycaused moderate damage, however.

Altogether, there were about 200 size-able flood events around the world.

– Heat waves with temperatures of upto 50°C were followed by severefloods in many Asian countries inMay and June. As in previous years,

this was the result of an extremelywet monsoon. In China, the swollenwaters of the Huai and Yangtzeflooded hundreds of thousands ofhomes and caused an economicloss of almost US$ 8bn.

– Many places in the Rhône region of southern France were underwater at the beginning of Decemberwhen numerous rivers flooded theirbanks after extreme rainfalls of upto 300 mm in 24 hours in someplaces (insured losses: approx.US$ 1bn).

Hot summer in Europe –

A taste of things to come

The outstanding event of the pastyear in Europe was the extreme heatand drought in the summer. Initialclimatological analyses indicated that – in statistical terms – the eventas a whole was to be expected muchmore seldom than once in 450 years.(We refer you to our documentationbeginning on page 20.)

Nevertheless, the burden imposed on insurers by, for example, drought-related losses was relatively small be-cause reduced yields in the agricul-tural sector as a result of heat wavesand dry weather are – for the mostpart – not yet covered in the EuropeanUnion. On the other hand, economiclosses were extraordinarily high atapprox. US$ 13bn.

Devastating wildfires raged on nearlyall continents of the earth in 2003. Theyear had hardly begun when bushfires heading straight for Canberra(Australia) hit the headlines. Thousandof firefighters and volunteers wereengaged for weeks in an untiring bat-tle against huge fires in Italy, France,

Natural catastrophes in 2003Review – Outlook 01

TOPICS geo 2003 Munich Re

3

Portugal, and Spain, where more than5,000 km2 of forest burned down.Western Canada saw its worst firessince the beginning of the 20th cen-tury. In October and November, thou-sands of homes in California weredestroyed by fires, which presentedthe insurance industry with a bill ofsome US$ 2bn.

Serious earthquakes in Algeria

and Iran

There were 70 loss-producing earth-quakes around the globe last year,generating economic losses of someUS$ 6bn, about US$ 100m of whichwas insured. In February, the north-west of China was shaken by a quakewith a magnitude of 6.4, the worst inthe region for 50 years. 70,000 build-ings were destroyed or damaged. InMay, an earthquake with a magnitudeof 6.8 rocked Algeria near the capital,Algiers, and claimed the lives of atleast 2,200 people. A disturbinglysevere earthquake occurred off theJapanese island of Hokkaido on26 September, but in spite of its mag-nitude of 8.0, it only caused slightdamage owing to the position of itshypocentre.

Despite a magnitude of 6.5, the earth-quake that caused parts of Californiato tremble on 22 December did notcause any dramatic damage becauseit fortunately occurred in a thinly popu-lated region about half-way betweenLos Angeles and San Francisco. How-ever, its proximity to those two mega-cities underlines the extremely highrisk in the event of future strongquakes.

Finally, early in the morning on 26 December an earthquake with amagnitude of 6.5 devastated the re-

gion around the city of Bam in thesoutheast of Iran. The majority of the mud-brick houses in this city –situated on the legendary Silk Roadand with a population of approx.100,000 – collapsed, burying tens ofthousands beneath the debris. Theauthorities counted more than 40,000people killed and approx. 30,000 in-jured. We refer you to the article be-ginning on page 32.

Outlook

Along with 2002 and 1998, 2003 wasone of the warmest years ever record-ed. The hot summer in Europe was ademonstration of just how serious theeffects of climate change can be. Theresults of research performed by cli-mate scientists indicate that evenslight changes in temperature have atremendous impact on the correspond-ing extreme values. It is to be fearedthat extreme events which can betraced to climate change will have in-creasingly grave consequences in thefuture. This means that we must reckonwith new types of weather risks andgreater loss potentials. All this will bea challenge in every respect. Neitherhuman beings, buildings, and infra-structure nor the agricultural and live-stock sectors are prepared for suchextremes. In accordance with the pre-cautionary principle, we would bewell-advised to prepare ourselves fordramatic changes.

The 9th Conference of the Parties,which took place in Milan at the endof 2003, produced, to all intents andpurposes, no new results. What didemerge, however, was that climateprotection will shortly enter a newdimension, even if Russia does notratify the Kyoto Protocol immediately.Great expectations are being placed

on emissions trading, which willbegin in the European Union in 2005.This will create for the first time a real chance of influencing develop-ments with an eye to the future andthus alleviating the adverse effects ofclimate change (cf. our article on thissubject beginning on page 50).

Natural catastrophes in 2003 – Review – OutlookMunich Re TOPICS geo 2003

4

Natural catastrophes in 2003 – Pictures of the year TOPICS geo 2003 Munich Re

03 2003 was a year of extremes. Europe experienced its most extreme heat wave in centuries.But other continents suffered under the sweltering sun for weeks too. In India, water wellsand reservoirs dried up and the people had to walk miles just to fill a couple of vessels withthe precious liquid. In the wake of global climate change, water shortages will become acentral problem in many regions of the earth this century.

5

Munich Re TOPICS geo 2003

Pictures of the year

6

Natural catastrophes in 2003 – Pictures of the year TOPICS geo 2003 Munich Re

15–17 JulyHurricane Claudette, United States

In the United States last year, whole areaswere devastated on repeated occasions bytornadoes, and losses were also caused byhurricanes. The damage shown in this photowas caused by Hurricane Claudette, which,after sweeping through the Gulf of Mexico,had its main impact in Texas. The cyclone toreboats and yachts from their anchorage anddamaged oil rigs as well as houses and infra-structure. Altogether, 15,000 claims werelodged with insurers.

July–AugustForest fires, Portugal

2003 was a year of fires. There were extensiveforest fires in Australia, North America, andEurope. This inferno was photographed in theAçor Mountains, Central Portugal. The countryhad to cope with historic conflagrations, inwhich 10% of its forest area burned downcompletely.

12–13 SeptemberTyphoon Maemi, South Korea

With wind speeds of up to 215 km/h andrecord rainfall of 500 mm, Maemi was one of the strongest typhoons in the history ofSouth Korea. Thousands of buildings andbridges were destroyed, whole streets turnedinto raging torrents in next to no time andbecame impassable. Insured losses came toUS$ 0.5bn.

04

05

06

7

Natural catastrophes in 2003 – Pictures of the yearMunich Re TOPICS geo 2003

25 SeptemberEarthquake, Japan

Two submarine earthquakes in short succes-sion with magnitudes of 8.3 and 7.4 rockedthe eastern part of the North Japanese islandof Hokkaido. A major fire broke out in an oilrefinery, and there were power outages andsevere damage to the infrastructure. Thequake also triggered a sea wave that reacheda height of 4 m but fortunately did not causeany major flooding.

2–4 DecemberFloods, France

Following the heat wave in the summer,December 2003 brought extreme torrentialrain to the South of France, affecting theentire southern part of the Rhône valley.300 mm of rain fell within a space of 24hours, an amount that normally takes several months to accumulate. This is apicture of an industrial estate in Arles.

22 DecemberEarthquake, United States

An earthquake between Los Angeles and SanFrancisco caused a stir shortly before the endof the year. Houses and a historic tower inPaso Robles collapsed. The epicentral areawas sparsely populated, seven people werekilled, 50 were injured. The quake with a mag-nitude of 6.5 highlighted once again the enor-mous earthquake hazard in California.

07

08

09

8


Statistics of natural catastrophes in 2003

Number of loss events: 699

Number of fatalities: 77,886

Loss events Fatalities

Breakdown by type of event

Ear

thq

uak

e

Volc

anic

eru

pti

on

Tro

pic

al

sto

rm

Win

ter

sto

rm,

bliz

zard

Torn

ado

Sev

ere

sto

rm

Hai

lsto

rm

Loca

l sto

rm

Flo

od

Flas

h fl

oo

d

Tsu

nam

i

Hea

t w

ave,

dro

ug

ht

Fore

st fi

re

Lan

dsl

ide

Ava

lan

che

Win

ter

dam

age,

fr

ost

300

250

200

150

100

50

0

3,000

2,500

2,000

1,500

1,000

500

0

47,748 fatalities 3,723 fatalities 22,036 fatalities

62%

2%

31%

5%

Percentage distributionworldwide

12%

43%

17%

28%

Loss events

Fatalities

Africa: 2,778

America: 946

Asia: 53,921

Australia/Oceania: 47

Europe: 20,194

Worldwide: 77,886


Africa: 57

America: 206

Asia: 245


Europe: 126

Worldwide: 699

Loss events and fatalities

Approx. 700 natural hazard events were registered in 2003. Compared with previous years, the distribution of the vari-ous types of event (earthquake, windstorm, flood, others) is nothing out of the ordinary. Unlike the fatalities: this figureis largely due to the earthquake in Bam (Iran) and the summer heat wave in Europe.

9

Natural catastrophes in 2003 – StatisticsMunich Re TOPICS geo 2003

US$ bn

Breakdown by type of event

10

9

8

7

6

5

4

3

2

1

0

US$ 11bn US$ 13bn US$ 14bn

Earthquake, volcanic eruption

Windstorm

Flood

Others

Economic losses (US$ bn)

Insured losses (US$ bn)

Ear

thq

uak

e

Volc

anic

eru

pti

on

Tro

pic

al

sto

rm

Win

ter

sto

rm,

bliz

zard

Torn

ado

Sev

ere

sto

rm

Hai

lsto

rm

Loca

l sto

rm

Flo

od

Flas

h fl

oo

d

Tsu

nam

i

Hea

t w

ave,

dro

ug

ht

Fore

st fi

re

Lan

dsl

ide

Ava

lan

che

Win

ter

dam

age,

fr

ost

Insured losses: US$ 15,810m

30%

10%

1%

21%

39%

Economic losses: US$ 64,604m

Africa: 5,158

America: 21,969

Asia: 18,230


Europe: 18,619

Worldwide: 64,604

Africa: 0

America: 13,274

Asia: 600


Europe: 1,690

Worldwide: 15,810


76%

8%


Economic and insured losses

In 2003, other events (heat waves, forest fires) were very prominent. The drought in Europe accounted for almost a thirdof the economic losses, whereas the wildland fires in California generated a large proportion of the insured losses.Otherwise, it was the picture that has become customary over the years: the statistics are dominated by windstorm losses.

15%

10

Major engineering and fire catastrophes in 2003 02


29 January, United StatesExplosion in a pharmaceutical factory

1 February, United StatesSpace shuttle Columbia explosion

18 February, South KoreaArson attack on underground train

20 February, United StatesInferno in a nightclub

3 June, SpainCollision between an express and a goods train

8 July, BangladeshFerry accident on the Meghna

14 August, United States and CanadaPower outage

19 September, outer spaceFailure of a Telstar 4 satellite

23 December, ChinaNatural gas accident during drilling operations

A large number of spectacular engineering catastrophes, explosions, and arson attacks occurred again in 2003. However,the extent of the losses caused by these events was much smaller than that due to natural catastrophes and theyclaimed fewer lives. There follows a selection of significant events:

10 11 12

13 14 15

16

19

17 18

11

Major engineering and fire catastrophes in 2003Munich Re TOPICS geo 2003

Date Region Loss event

29 January United States, Explosion in a pharmaceutical factory

North Carolina, An explosion in a pharmaceutical factory in North Carolina triggered a fire in which six people were killed Kinston and more than thirty injured. The explosion had probably been caused by fine plastic dust used in the

manufacturer of rubber.

1 February United States, Space shuttle Columbia explosion

sky over Texas The US space shuttle Columbia exploded during re-entry into the earth's atmosphere. This is thought to havebeen caused by a small piece of insulation material that had damaged the left wing during the launch phase.All seven members of the crew were killed in the accident.

18 February South Korea, Arson attack on underground train

Daegu An arsonist ignited a carton of flammable material in a packed underground train in the centre of Daegu. The flames spread in a matter of seconds and set fire to a second train coming in the opposite direction.138 people suffered severe burns, 120 died in the blaze.

20 February United States, Inferno in a nightclub

Rhode Island, A pyrotechnic device got out of control and set fire to the stage curtain and interior of a nightclub during a West Warwick rock concert. The club was not equipped with sprinklers, so that the fire was able to engulf the entire building

in just a few minutes, with the result that 100 people were killed.

3 June Spain, Collision between an express and a goods train

near Albacete The signals on a single-line track were green when an express train crashed head-on with a goods train at aspeed of 200 km/h. 27 people were killed in the accident. The catastrophe would not have happened if therehad been an automatic safety system installed on the line.

8 July Bangladesh Ferry accident on the Meghna

More than 650 people were drowned when an overloaded triple-deck ferry on the Meghna river hit turbulentwaters and sank. Accidents of such dimensions are almost regular occurrences because the number ofpassengers taken on board often greatly exceeds the permitted limit. The photo shows the situationencountered daily on passenger ferries in this region.

14 August United States Power outage in the northeast of the United States and Canada

and Canada, 50 million people in this densely populated region were without electricity for as long as 24 hours. The entire East Coast traffic network, communications, and large sections of economic life came to a standstill. The economic loss

is estimated to be US$ 6bn. The power outage in North America was just the first in a succession of incidentsthat affected various countries including the United Kingdom (28 August), Malaysia (1 September), Swedenand Denmark (23 September), and Italy (28 September).

19 September Space Failure of a Telstar 4 satellite

A short circuit in the power supply led to the total loss of a satellite that had been launched in 1995. As allcontact was cut off, it was of no further use. This Telstar 4 telecommunications satellite was insured forapprox. US$ 140m.

23 December China, Natural gas accident during drilling operations

Chongqing During drilling operations at a gas field in the southwest of China, there was a toxic gas blowout. 10,000people had to be treated for severe poisoning and chemical burns, more than 200 were killed. It took severaldays to close the burning well.

12

Great natural catastrophes TOPICS geo 2003 Munich Re

19 A strong earthquake with a magnitude of 6.8 on the Richter Scale wreaked enormous devas-tation near the capital of Algeria. The photo shows a typical damage pattern, which is knownas the soft-storey problem and is a concern to construction engineers throughout the world.The lack of adequate support leads to entire floors collapsing. 2,200 people died in thismajor natural catastrophe in May.

13


Great natural catastrophes 03

14


Great natural catastrophes 1950–2003

Hundreds of natural hazard eventsoccur around the world year in, yearout. For risk managers it is of decisiveimportance not only to look at the in-dividual events but also to consider inparticular future developments. Long-term trends can be derived frommajor natural catastrophes, becauseonly these can be researched far backinto the past. The last four years inour long-term statistics should notmake us forget that both the numberof events and the monetary losseshave increased perceptibly in recentdecades. This becomes clear particu-larly in the case of the decade com-parisons (cf. table on p. 15).

Definition of

great natural catastrophes:

Natural catastrophes are classedas great if the ability of the regionto help itself is distinctly over-taxed, making interregional or in-ternational assistance necessary.This is usually the case whenthousands of people are killed,

hundreds of thousands are madehomeless, or when a country suf-fers substantial economic losses,depending on the economic cir-cumstances generally prevailingin that country.

For risk managers in the insurance industry it is important not only to look at loss eventsin isolation but also and above all to keep track of identifiable trends.

02468

10121416

Earthquake, volcanic eruption

Windstorm

16

14

12

10

8

6

4

2

01950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000

Number of events

Number

A total of 700 loss events due to natural hazards were registered last year andfrom these we have selected the “great” natural catastrophes on the basis ofthe above definition.

– Tornadoes, severe storms, United States (May)– Earthquake, Algeria (May)– Heat wave, forest fires, Europe (July–August)– Heat wave, drought, forest fires, United States (October–November)– Earthquake, Iran (December)

The diagram shows for each year the number of great natural catastrophes, divided up by type of event.

Flood

Others

15

Great natural catastrophes – Long-term statistics 1950–2003Munich Re TOPICS geo 2003

Economic losses (2003 values)

of which insured losses (2003 values)

Average economic losses per decade

01020304050607080

> US$ 170bn

1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000

Economic and insured losses with trends

US$ bn

The table allows a comparison of the aggregate loss figures of recent decades. Comparing the last ten years with the1960s makes the increase in natural catastrophes particularly clear. This applies both to the number of events and to theextent of the losses incurred.

The chart presents the economic losses and insured losses – adjusted to present values. The trend curves verify theincrease in catastrophe losses since 1950.

Decade comparison 1950–2003

Last 10:60s

2.2

6.7

13.5

Decade 1950–1959 1960–1969 1970–1979 1980–1989 1990–1999 Last 10 years

Number of 20 27 47 63 91 60

events

Economic 42.7 76.7 140.6 217.3 670.4 514.5

losses

Insured losses – 6.2 13.1 27.4 126.0 83.6

Losses in US$ bn (2003 values)

The comparison

of the last ten

years with the

1960s shows

a dramatic

increase.

Trend of economic losses

Trend of insured losses

80

70

60

50

40

30

20

10

0

16

The year 2003 brought the UnitedStates devastating tornadoes andChina severe floods. In 2002 therewere hundred-year floods in centralEurope and at the same time floods inIndia with more than 1,200 fatalities.In 1992 Hurricane Andrew generatedbillion-dollar losses in Florida withinsured losses of US$ 17bn, in 1998Hurricane Mitch caused a humani-tarian disaster with some 10,000 fatal-ities. Mitch hit Honduras, where thereis little insurance coverage, and para-lysed the comparatively poor nationaleconomy for years to come. In the fol-lowing article we examine the insur-ance penetration in various marketsand the effect of weather-related nat-ural catastrophes.

Uneven insurance penetration around

the globe

While industrial countries may be saidto be relatively well to very well in-sured, the people in developing coun-tries are more or less poorly insured.But what is the yardstick on which thisstatement is measured? There are vari-ous ways of calculating property in-surance premiums. The total premiumincome of individual countries cannotbe used as a criterion as this leads toconsiderable distortions in particularlyhighly populated and large countries.The best approximation for the per-centage of insured people or house-holds is obtained using the paid prop-erty insurance premiums per capita.

Given this extreme simplification,there is one important aspect that isobviously not to be overlooked: evenin countries where sufficient insur-ance premiums are paid on average,there will be a large number of peoplewho cannot afford any coverage at alland remain uninsured.

Our analysis is based on a breakdowninto seven country groups correspond-ing essentially to the following key:

– Uninsured (property insurance pre-miums US$ 0–5 per person andyear): material assets not or inad-equately insured

– Basically insured (property insur-ance premiums from US$ 5 to 50 ayear): most vital basic insurance justabout provided

– Adequately to well insured (propertyinsurance premiums exceed US$ 50,100, 500, or 1,000 a year): materialassets (households) adequately towell insured

– In Africa there are only seven coun-tries with an annual property insur-ance premium exceeding US$ 5 perperson. Of these, South Africa is theonly country with a well-developedinsurance market.

– The largest and most highly popu-lated countries of Eurasia, China,India, and Russia, are in differentgroups. Nevertheless, it must beborne in mind that the picture will

be distorted by the not inconsid-erable sums insured for industrialplants and international constructionprojects or businesses (e.g. interna-tional hotel chains). Also, the sizeand distribution of the populationlead to extreme regional differences,e.g. between urban and rural popu-lations. And the much-cited “emerg-ing markets”, especially India andChina, are developing so rapidlythat the insurance penetration maylook very different in just a fewyears.

– It is interesting to note that in SouthAmerica the average property insur-ance premium is without exceptionabove US$ 5 a year, i.e. in the basic-ally insured group. This means that – at least in purely statisticalterms – there is “basic protection”for the population's assets.

– In the wealthy industrial countries ofNorth America and western Europeand in Japan, South Korea, Aus-tralia, and New Zealand, a sum farexceeding US$ 100 per person peryear is spent on average for proper-ty insurance protection. As a rule,each household takes out severalinsurance policies; the people areconsidered to be well to very wellinsured.


04Weather-related natural catastrophes repeatedly affect both well-developed and less-developed insurance markets and countries. Against this backdrop the same questionsregularly arise: “Are weather-related natural catastrophes still insurable in the long term?Are all regions of the earth insurable in fact?” and “What is to be done about the millionsand millions of people who are not insured?”

Weather-related natural catastrophesin the insured and uninsured world

NatCatSERVICE-Information

17

Analysis of weather-related natural

catastrophes by country group

If we look at the global effects ofweather-related natural catastrophesthat have occurred since 1980 in thisworld classified in terms of insur-ance protection, we will notice thefollowing:

– Windstorms, severe weather, floods,drought, and other weather-relatednatural catastrophes occur through-out the world, regardless of whetherthe regions concerned are well orpoorly insured.

– Approx. 80% of all fatalities (morethan 700,000 people) were in theuninsured group in the period underexamination.

– As a result of the high level of in-dustrialisation and the comparative-ly high standard of living, a dispro-portionate majority of the totaleconomic loss (60%) was caused inthe rich industrial countries. Evenso, the economic loss incurred bythose without insurance far exceed-ed US$ 300bn (roughly a third).

– The overwhelming majority of in-sured losses (95%) was inevitablypaid out in the countries with good

insurance penetration; a sum ofmore than US$ 200bn. There are nomajor changes to be expected inthis distribution in the foreseeablefuture.

Will these losses continue to be

insurable in the future?

In the light of this development, whatare the limits of insurability of fre-quently affected markets or regions?There is no generally valid answer. Inthe medium to long term, it could bethat the insurance industry will haveto withdraw from individual regionsand zones that are regularly and al-most predictably affected by weather-

NatCatSERVICE-InformationMunich Re TOPICS geo 2003

Property insurance premium (non-life including health) per capita per year in US$Source: MR Economic Research/NatCatSERVICE®

US$ 51–100US$ 101–500US$ 501–1,000US$ 1,000+

US$ 0–5 US$ 6–25US$ 26–50

Uninsured group Basically insured group Well insured group

18

related natural catastrophes. This willapply in particular to areas along indi-vidual sections of river that are regu-larly flooded so that the basic premiseof insurance (sudden and unforeseenoccurrence of an event) is no longervalid. It is a different situation whenlarge regions or countries are con-cerned. Here the instruments avail-able to the insurance industry willsuffice for decades to come, insofaras at least some of the conditionsspecified below continue to be met.For example, a rigorous implemen-tation of just the first three of thefollowing measures will enable theinsurance industry to cope with evenhuge loss potentials :

– Risk-adequate premiums, prospec-tive underwriting

– Substantial deductibles geared tothe degree of exposure

– Accumulation control– Loss prevention and loss avoidance

(heightening of awareness, motiv-ation)

– Change in forms of cover from pro-portional to non-proportional excessof loss covers

– Limits of indemnity, going as far asthe exclusion of particularly exposed

areas (as already sketched out forflood) or of specific perils

Solutions for the poorly insured or

uninsured world

For years now international organisa-tions like the United Nations and theWorld Bank have been endeavouringto find solutions for uninsured or in-adequately insured people, communi-ties, and countries. An example of thiswas the International Decade forNatural Disaster Reduction (IDNDR),whose work spanned the world in the1990s and is being continued by itssuccessor organisation, the Inter-national Strategy for Disaster Reduc-tion (ISDR). The aim of these endeav-ours is, and was, to save human lives,reduce losses from natural catas-trophes, and to minimise their effects.We have selected a few examples toillustrate this:

– In conjunction with the insurance in-dustry, the World Bank has devisedconcepts for some countries (e.g.India) in order to ensure the provi-sion of reliable insurance protectionagainst natural hazards.

NatCatSERVICE-Information TOPICS geo 2003 Munich Re

Weather-related natural catastrophes 1980–2003: percentage distribution among insurance markets at various stages

of development

12%

20%

1% 0%3%

4%

22%

80%

8%2%

2%2%

2%4%

69%

1%

2%5%

11%14%

36%

Fatalities: 930,000Insured losses: US$ 225bnEconomic losses: US$ 1,000bn

US$ 0–5

US$ 6–25

US$ 26–50

US$ 51–100

US$ 101–500

US$ 501–1,000

US$ 1,000+

Property insurance premium per capital per year in US$Source: MR Economic Research/NatCatSERVICE®

19

– Financing catastrophe funds (CatBonds) is one way of making provi-sions before natural catastrophesoccur. In the event of a loss, speedypayments make it possible to re-build factories and restore damagedinfrastructure in a short time.

– Micro-finance and micro-insurancesolutions can even reach individualhouseholds in poor countries andpromote welfare and economicdevelopment even in the smallest of communities.

Of course, the state insurance con-cepts that have already been set up insome industrial countries to addressspecific problems (e.g. in France,Spain, Switzerland, and the UnitedStates) must continue to be observedclosely. Such state insurance conceptsmay be an effective solution in certaincountries.

Public-private partnerships as a

model

Roughly one-half of all the people onearth are de facto without adequateinsurance protection (in our analysis,the lowest class with US$ 0–5 perperson per year). At the same time,natural catastrophe losses will reachdimensions that are quite new – as aresult of global climate change, forexample. This makes it necessary tolook for practicable and adaptableconcepts. Also, a topic that is beingdiscussed more and more often at theConference of the Parties' meetingson climate protection is whether, andhow, the industrial countries, in otherwords the countries that are mainlyresponsible for man-made climatechange, can be held liable forweather-related and climate-relatedlosses in the uninsured world.

It will only be possible to rise to thechallenges we will be facing in ourever more complex world if politics,the insurance industry, and the peoplein the regions that are being affectedmore frequently all contribute towardsthat solution. The often called-forpublic-private partnerships (PPPs)which are based upon this conceptare a very promising start in the rightdirection.

NatCatSERVICE-InformationMunich Re TOPICS geo 2003

20 A heat wave held central Europe in its grip for weeks on end in the summer. It causedsevere damage particularly in the agricultural sector, with losses exceeding US$ 10bn, anddestroyed the entire harvest in many places.

20

Documentation TOPICS geo 2003 Munich Re

Hot summer in Europe – The future has already begun

Documentation:

05

21


22

Not only the floods in August 2002,but also the summer heat of 2003were real “hundred-year” events – asmeteorologists have now definitelyestablished. In 2002, Saxony hadrecorded local precipitation amountsof over 300 l/m2 within a period of24 hours, the likes of which had nevercome anywhere near to appearing inthe region’s climate statistics up untilthen. In 2003, the meteorologicalsummer (June to August) presentedGermany with mean temperatures noless than 3.4°C above the average forthe period 1961–1990. According to ananalysis performed by the FrankfurtUniversity Institute for Meteorologyand Climatology (Prof. Schönwiese),this corresponds to an occurrenceprobability of once in approx. 450years. But this does not even take into account the fact that the monthsof May and September were also un-seasonably warm or that the heatwave affected not only Germany butalso large parts of central, western,and southern Europe. Consequently,the event is an even more weightyand extreme “outlier” in the climatestatistics.

But are these two extreme weather-related events really as unusual todayas they would have been in pastdecades and centuries? Or do theyalready reflect the changes that globalwarming brings with it?

All in all, precipitation amounts in-crease if the earth gets warmer be-cause more water evaporates and the

atmosphere can hold more watervapour. The water cycle becomesmore intense as a result and the prob-ability of torrential rain rises substan-tially. Given an average increase intemperatures of 2°C, as is to be ex-pected in many regions by the middleof this century, the return periods ofextreme daily precipitation may bereduced to a third, and given a – notunimaginable – increase of 4°C toeven an eighth, e.g. from 100 to12 years.

The same applies to temperaturesand, generally speaking, to other cli-mate parameters as well. Even a rela-tively moderate shift in averages willlead to an inordinate increase in theexceedance probability of uncommonthreshold values. The Frankfurt meteor-ologists’ analyses show, for instance,that the probability of a hot summerlike the one in 2003 has increased by a factor of twenty in the past twodecades alone, which in terms of thetemperature were in themselves quiteexceptional. In accordance with therules of statistics, this means that, inonly a few decades, hundred-yearevents can turn into what are basic-ally quite normal occurrences thatmay be expected every couple ofyears. An additional danger is thatsuch rapid and enduring changes maycause unexpected feedback effectsand irreversible phase transitions inthe complex “atmosphere – earth –ocean – ice” system (e.g. changes inocean currents) and may thus makeall prognoses worthless.

In accordance with the precautionaryprinciple, we would be well-advisedto be prepared for dramatic changes.It is essential to consider trends thatare already recognisable today whendefining design values for buildings(e.g. for thermal or windstorm loads),or when designating high-risk zonesin which land use is prohibited, orwhen calculating the risk premiumsfor natural hazard covers on aprospective basis.

Regrettably, the experience gainedfrom last year’s hot summer has con-firmed in full what Munich Re’s GeoRisks Researchers have been predict-ing for many years: 1. The risk of pro-nounced and prolonged periods ofheat and drought is increasing dra-matically in large parts of Europe.2.The potential effects on the econ-omy and the insurance industry arecompletely underestimated or evenignored. 3. There has been a ruinousneglect of precautions for adaptationto such situations.

Indeed, the heat and drought in thesummer of 2003 had some remark-able features, as the following keydata impressively demonstrate:

– Size of the affected area: severalmillion square kilometres

– Countries most severely affected:France, Spain, Portugal, Italy, Ger-many, Switzerland, Austria; England,Benelux, Poland, Slovakia, Romania,and Hungary.


Summer 2003: Weeks and weeks of extreme temperatures above the 30°C mark made thepeople in Europe sweat. Are the extreme weather-related events of the recent past stillatypical occurrences or do they already reflect the normality of global warming?

23

– Duration of the heat period withoutextended interruptions: approx.3 months (mid-May–mid-August)

– Number of fatalities: over 20,000(see table)

– Property damage, especially in agri-culture and as a result of forest fires:US$ 13bn

Other grave effects on the economy:

– Inland shipping (low water: cancella-tions and losses in cargo transporta-tion and tourism)

– Industry, power plants (river watertoo warm, problems with cooling:production bottlenecks)

– Stark reduction in worker efficiency(resulting in an economic loss thatis substantial but still very difficultto quantify in monetary terms)

A particularly critical aspect for manysectors of the economy was the ex-tremely low water level of numerousEuropean rivers, in which the lowestvalues of all time were often meas-ured. This meant that normal shippingwas impossible for a long time. Aboveall, however, numerous electricity-gen-erating plants had to cut back on theiroutput because there was either notenough “water power” or not enoughcooling water to stay within the dis-charge temperature limits. Electricalpower was therefore in short supplyand the price rose; occasionally thepower supply failed altogether.

Even in sectors that usually profitfrom nice summers like open-air en-tertainment and tourist attractions,there was a shortfall in daytime re-ceipts because it was just too hot tomake the effort and go. The owners ofcafés, ice parlours, garden restau-rants, beer gardens, and swimmingbaths did a roaring trade though.

Documentation Munich Re TOPICS geo 2003

At the peak of the heat wave, water had to be sprayed onto the outside surfaces of numerouspower plants above ground (as here in Fessenheim, France) as the cooling systems were nolonger working efficiently due to the heated rivers and the shortage of water. Some powerplants had to reduce their output, some had to be shut down altogether.

21

24


The effects of the exceptional heat on the people

The thermal climatic conditions are among the environ-mental factors that have the strongest impact on personalwell-being and health.

The human organism reacts in many ways to changingtemperature conditions, in order to keep up its internalregulating system, in which maintaining a constant bodycore temperature has top priority. In hot weather the bloodpressure falls, the heart has to beat faster, and the bodyloses large amounts of liquid through perspiration. Thesestresses and strains may be too much for the cardiovascu-lar system especially in the infirm and may cause death.Many studies prove that there are definite links betweenthe number of fatalities and the thermal conditions on anyone day. In New York and Shanghai, for example, threetimes as many people die on extremely hot days as onnormal warm days – and these are two cities in two com-pletely different regions of the earth (cf. the graphs below).

As the strain exerted on the body is influenced not only byair temperature but also by humidity, wind speed, and ir-radiation conditions, all these complex influences aretaken into account in the more recent biometeorologicalanalyses. Of paramount importance therefore is perceivedtemperature, which many meteorological services havebeen using for some years now.

The analysis of nine different heat waves shows that per-ceived temperature and the number of fatalities run moreor less parallel in time. The number of fatalities dropsbelow the average expectancy value once a heat wave hasended. This phenomenon is called the “harvesting effect”in the field of epidemiology. It means that the elevatedmortality during heat waves is to some extent due to thetime of death being advanced by no more than a few days.In the majority of heat-mortality studies, however, modernepidemiological analyses point not only to this “harvestingeffect” but also to fatalities that would not have been ex-pected in the subsequent few weeks without the effects ofthe heat, in other words to a real increase in mortality overa relatively long period of time.

Temperature and mortality in New York and Shanghai.

Maximum temperature ( ºC)

The graphs show the number of fatalities as a function of the maximum temperature. At elevated temperatures, the numbers of deaths increaseenormously. Source: McMichael, A.J., et al., 1996. Climate Change and Human Health, WHO/WMO/UNEP.

Daily mortality

10 15 20 25 30 35 40

300

250

200

150

100

50

0

Maximum temperature (ºC)

Daily mortality

10 15 20 25 30 35 40

300

250

200

150

100

50

0

25

Documentation Munich Re TOPICS geo 2003

The heat in the summer of 2003 claimed the lives of morethan 20,000 people in Europe (see table). And the distribu-tion of the heat victims correlates strikingly with the analy-ses of thermal stress performed by the German WeatherService (DWD) (see map of Europe below).

The death rate in France in August 2003 was more thantwice as high (25 cases per 100,000 inhabitants) as the tem-perature-related fatality rate in Chicago during the 1995heat wave (12 deaths per 100,000 inhabitants, totalling514); in Ile-de-France and Centre, the regions most severe-ly affected, the rate was no less than four times as high.

The effects of the summer heat wave on morbidity due tothe consumption of spoilt food (e.g. salmonella poisoning)cannot be quantified. The WHO estimates, however, thatthroughout the world about 2.4% of all diarrhoea cases arethe result of higher temperatures due to climate change inrecent decades.

For the experts, it is not only the maximum daytimetemperatures that are important but also the night-timetemperatures. They influence the quality of sleep andhence a person’s heat tolerance on the following day.Particularly in the concrete jungles of the megacities, night cooling is reduced considerably by the heat islandeffect. The negative effects of heat waves may thus beconsiderably aggravated.

Correlations: weather, health, accidents

A recent study performed at the University of Munichlooked into the connections between the number of roadaccidents and various meteorological parameters. Therewere significant correlations not only with primary wea-ther influences like storms, rain-soaked roads, and blackice, but also with thermal conditions. The scientific analysisof daily accident figures and weather factors revealed thatthere were 18% more road accidents on the hottest daysthan on cool days. Even in countries where heat load iscommon like Saudi Arabia (Riyadh), there is a positive cor-relation between the frequency of road accidents and airtemperatures.

Map of perceived temperatures in Europe

on 8 August 2003

Victims of the hot summer of 2003 in Europe

Snapshot of perceived temperatures in Europe on 8 August at thepeak load time (1 p.m.). Paris was in the extreme zone for days andrecorded by far the largest number of victims. The figures publishedon numbers of victims varied widely from country to country andwere not always easy to verify. The figures quoted in the table re-flect the best estimates of ministries and state statistical offices atthe end of the year.

Fatalities

France 14,800

Spain 2,000

Portugal 1,300

Italy 4,000

Germany 3,500

United Kingdom 900

Netherlands 500

Service: German Weather Service

extremestrong

moderateslight

comfortableslight

moderatestrong

extreme

heat load

cold stress

ºC383226200

–13–26–39

26


Occupational health studies reveal that heat load increasesthe number of occupational accidents and reduces produc-tivity. Air temperatures above 27°C reduce cognitive per-formance (e.g. perception, reaction times, concentration).

There is also evidence that heat is capable of raising thelevel of aggression and the incidence of violent acts. Acurrent study on the effects of the weather shows that therescue services are called out to deal with acts of violenceup 75% more often on hot days than on cool days.

The direct negative effects of climate change on humansare not so much the result of slow and continual increasesin the temperature as the increasing frequency of extremeevents like the summer of 2003 in Europe.

The hot summer in 2003 was also indirectly responsiblefor the frequent exceedance of ozone threshold values inwestern and southern Europe. This may also affect thehealth of the population and have a negative impact onagriculture and forestry. The increase in forest damage incentral Europe in the summer of 2003 was attributed to the elevated ozone values as well as to moisture stress.

Summary

To sum up: heat waves not only cause a distinct increasein the mortality of the elderly but also leave their marks onnational economies: more accidents occur, agriculture andforestry suffer, and labour productivity declines.

During the summer months,record temperatures wererecorded in many regions ofEurope. This thermometer in Alicante shows an unpreced-ented temperature of 64°C –which represents a new dimen-sion even for southern Spain.

22

27


The California wildland fires of 2003 06Chronicle:

23 Natural forests, like most forests, are not usually insured for fire, but large losses are oftento be expected if buildings or entire settlements are caught in the blaze. In California, wheremore than 4,000 buildings went up in flames, the insurance industry paid more thanUS$ 2bn to the homeowners.

28

Several additional fires started duringthe following week, fuelled by dryvegetation and fanned by strongSanta Ana winds (warm, dry, offshoreblowing winds), before a combinationof vigorous fire-fighting activities andmore cooperative weather conditions(cool, damp, onshore blowing winds)started to whittle them down. The late-October fires in southern Californiadestroyed more than 3,600 homesand damaged a few hundred more.The fires also destroyed or damagedseveral hundred automobiles, trailers,

and boats. The counties of SanBernardino and San Diego were theworst hit, but the fires did also affectVentura, Los Angeles, and Riversidecounties. A total of 22 deaths havebeen attributed to these fires. Esti-mates of total insured losses from allfires combined are around US$ 2.2bn,and economic losses have been pro-jected on the order of US$ 3.5bn.

The fires of October/November 2003

The Cedar fire in San Diego County isthe largest wildland fire in California’shistory. It burned more than 270,000acres (110,000 hectares) and des-troyed over 2,200 homes. It is be-lieved that the Cedar fire was ignitedby a distressed hunter sending up asignal flare. The Old and the GrandPrix fires have both been blamed onarson. The fires merged on 26 Octo-ber, separated only by Lytle Creek. Theso-called Padua fire was merely the

Chronicle TOPICS geo 2003 Munich Re

Location and extent of the October 2003 wildland fires in southern California.

County Line

Interstate

Highway

Burned Area

Urbanized Area

In late October 2003, southern California experienced an unusually active period ofwildland fires. The first fires in this episode were ignited on 21 October.

29

western flank of the Grand Prix fire,the portion of it that was on the LosAngeles side of the county line. Thewhole conflagration (Old/GrandPrix/Padua) in San Bernardino Countyburned over 160,000 acres (65,000hectares) and destroyed more than1,100 homes. As of December 2003,the causes of all the Californian fireshave not been determined, but theVerdale fire has been blamed onarson, the Paradise fire has been at-tributed to human activity (either anaccident or arson), and the Simi firehas been attributed to spot fires fromthe Verdale fire. The causes of theRoblar, Piru, Mountain, and Otay firesare still under investigation.

Wildfire history

The continuously expanding wildland-urban interface in the south-westernUnited States is susceptible to disas-trous fire losses. Recent examples

include the Rodeo-Chediski (Arizona)conflagration in June 2002, whichcaused an estimated US$ 125m ininsured property losses (all historicallosses are reported in 2003 dollars,adjusted for inflation using the Con-sumer Price Index), and the CerroGrande (New Mexico) fire in May2000, which caused an estimatedUS$ 150m of insured property losses.Historically, however, the state ofCalifornia has been hit the hardest.The most costly fire in the state’s his-tory, in terms of insured propertylosses, is the Oakland Hills firestormin October 1991, which caused anestimated US$ 2.3bn in insured prop-erty losses. In 1993, which is the thirdcostliest wildland fire year in Califor-nia’s history (after 1991 and 2003),two fires in southern California (Lagu-na Canyon in October and Malibu-Topanga in November) caused a totalof US$ 930m in insured propertylosses.

Return period approaches

Based on data on insured losses dur-ing the past 34 years (1970 to 2003),annual aggregate property losses inCalifornia from wildland fires haveaveraged approximately US$ 180m,adjusted for inflation. It is interestingto note that in this 34-year period, theannual average has only been exceed-ed four times. Applying a simple,least-squares logarithmic curve fittingto the inflation-adjusted Californiawildfire loss data (annual aggregates)yields an estimated return period offive years (the reciprocal of the an-nual frequency of exceedance) forUS$ 100m, ten years for US$ 650m,and a return period of 25 years for anannual aggregate industry loss ofUS$ 2bn. Extrapolating this curve,which is not a recommended practice,generates an estimate of US$ 2.8bn

ChronicleMunich Re TOPICS geo 2003

Summary of the October 2003 wildland fires in southern California

(data as of December 2003)

No. Fire Cause of ignition Date Date Area Deaths Destroyed structures

started contained km2 Commercial Residential Outbuildings

1 Roblar Under investigation 21 Oct. 28 Oct. 35 0 0 0 0

2 Grand Prix Arson 21 Oct. 8 Nov. 241 0 1 135 60

3 Piru Under investigation 23 Oct. 10 Nov. 259 0 1 1 6

4 Verdale Arson 24 Oct. 28 Oct. 35 0 0 0 1

5 Old Arson 25 Oct. 5 Nov. 370 6 10 993 0

6 Cedar Accident 25 Oct. 4 Nov. 1,107 14 22 2,232 566

7 Simi Spot fire from (4) 25 Oct. 1 Nov. 438 0 0 22 278

8 Paradise Accident/arson 26 Oct. 6 Nov. 230 2 2 221 192

9 Otay Under investigation 26 Oct. 28 Oct. 186 0 0 1 5

10 Mountain Under investigation 26 Oct. 29 Oct. 42 0 0 21 40

11 Padua West end of (2) 27 Oct. 2 Nov. 42 0 0 0 0

Total: 2,985 22 36 3,626 1,148

The number of the fire (No.) in the table refers to the fire’s number on the map.Outbuildings refer to detached garages, sheds, pools, fences, and the like.

30

Chronicle TOPICS geo 2003 Munich Re

The local fire-fighting teams were completely at the mercy of natural forces. Firemen are powerless in the face of a gigantic ball of fire in SimiValley, northwest of Los Angeles, where an area of more than 400 km2 burned down.

24

31

for the 50-year industry gross loss onan annual aggregate basis from Cali-fornia wildland fires.

The estimated losses and returnperiods mentioned above are onlygiven here for rough reference; theyare not based on a probabilistic riskmodel. While the insured wildland firelosses incurred over the past onethird of a century have been adjustedfor inflation, they have been correctedneither for increased exposure nor foroverall growth in wealth and propertyvalues during the period. Since theincrease in exposures has dispropor-tionately taken place at the wildland-urban interface, it is very likely thatthe empirical loss estimates men-tioned above are lower than the actu-al, present-day probabilistic loss esti-mates would be.

The wildland fire risk in the UnitedStates in general, and in California inparticular, is on the rise, due to in-creasing urbanisation and an ongoingpopulation shift towards the southand the west. Homeowners andowners of other property located at or near the wildland-urban interfacewould be well advised to take all rea-sonable measures to reduce the firehazard. The recommended actions in-clude, but are not limited to, creatingand maintaining a clear fire break sur-rounding the property, using non-combustible roofing and wall claddingmaterial, and keeping the roof, guttersand eaves clear of debris.

Insurance and reinsurance aspects –

Fire event issues

Since a fire known as the Padua firemerged with a fire known as theGrand Prix fire and subsequentlymerged with the Old fire (PCS Cat 98),

the above-mentioned list of elevenfires has to be reduced. Munich ReGroup therefore consider that the Cal-ifornia fires of October 2003 consistedof several fires or events. This consid-eration also maintains a consistencywith the 1993 wildfires, where each ofthe 27 different fires had a separateorigin and each was considered to bea separate event. We are aware thatthe intention of this paper is to de-scribe the events physically. But thecommitment of Munich Re finally de-pends of the definition of “event” inboth the direct policy wording and thereinsurance wording.

Therefore, Munich Re recommend thefollowing:

a) Review the respective wordings b) Define the number of events based

on the respective wordingsc) If necessary, revise the claims esti-

mations/amounts based on the findings of the number of events

ChronicleMunich Re TOPICS geo 2003

33

Scientific aspects, features

of the quake

The strong earthquake with a magni-tude of 6.5 occurred at 5.26 a.m. localtime. The hypocentre was only 8 kmbelow the city of Bam, some 180 kmsouth of Keman, the provincial capi-tal, and just under 1,000 km south of Teheran.

Scientists were surprised neither bythe strength of the quake nor by itslocation. Iran is located in the pres-sure zone of the Arabian and Eurasianplates, which are moving towardseach other at a speed of some 3 cm ayear. This plate movement has causedstrike slip and thrust faults throughoutIran, a region of high seismic activity.The Bam Fault with a north-south hori-zontal displacement runs almost dir-ectly through Bam. It was on this faultthat the quake on 26 December 2003occurred (see Fig. 2). As the seismichazard map for Iran shows, there is an elevated earthquake hazardthroughout the country. The city ofBam is in intensity zone 3, correspond-ing to an expected earthquake inten-sity of at least VIII (MM scale) on aver-age once in just under 500 years (see Fig. 1).

Historical records give no indicationthat Bam or its immediate environshad ever been hit by a quake of suchmagnitude before. This is why thecity’s 2,000-year-old citadel had neverfallen victim to an earthquake. In

1981, however, two strong earth-quakes with magnitudes of 6.9 and 7.3 shook the region some 100 kmnorth of Bam, claiming the lives ofseveral thousand people. The list of devastating historical earthquakecatastrophes in Iran is particularlylong. In the last century alone, sevenquakes with over 5,000 victims wereregistered, the last being in 1990,when 40,000 people were killed in the Rasht region on the Caspian Sea.

One distinct feature of the Bam earth-quake was the relatively small size ofthe heavily affected area around theBam Fault activated in the earthquakeand in particular around the epicentre.Severe damage was only reported inBam itself and a few villages in theimmediate vicinity. This is due to whatwas – by global standards – a moder-ate magnitude of 6.5 and to the lowfocal depth of 8 km. The maximumintensity recorded in Bam, which layin the epicentral area, was IX on theEMS scale (see Fig. 3).

Losses

The people of Bam were still sleepingwhen the earthquake struck, claimingthe lives of at least 40,000, injuringmore than 30,000, and making100,000 homeless. The city now pre-sents a scene of total destruction, with70% of the houses having collapsedat least partially. Besides the unfortu-nate time of the quake and its posi-tion directly beneath the town, the

main reason for the large number offatalities was the generally poor con-struction quality. Traditional buildingsof mud-brick construction and heavyroofs are particularly unsafe whenearthquakes strike. The binding mater-ial is poor to non-existent so that theunreinforced masonry possesseshardly any resistance against theground motion in strong earthquakes.With little tensile and shear strength,the walls are unable to resist dynamicearthquake loads, which is why themajority of the buildings in the epi-central area collapsed. The compara-tively loose construction of the mud-brick walls prevented the creation ofcavities which would have given theinhabitants some protection from thefalling masses of stone and giventhem the chance of being rescuedalive. Consequently, the 44 inter-national relief teams with 1,500 staff and some 20,000 Iranian volunteerhelpers were only able to rescueabout 3,000 people. Three days afterthe earthquake, only 20 survivorswere pulled from the ruins.

Situated on the legendary Silk Road,Bam is famous for its 2,000-year-oldcitadel Arg-e-Bam. This fortified settle-ment is one of the country’s most im-portant cultural monuments and wasbadly damaged in the earthquake. The destruction of Bam and its citadelwill cripple tourism in the region foryears, if not decades. The area hasthus been robbed of an importanteconomic pillar.


The Bam disaster in Iran

25The collapsing mud-brick buildings form compact heaps of rubble which create hardly anycavities and thus make it almost impossible to survive under the debris for any length oftime.

07On 26 December, shortly before the end of the year, an earthquake below the city of Bam in the southeast of Iran highlighted once again the lethal connection between poorbuilding quality and large numbers of victims: 40,000 people died.

Earthquake report:

34

Insurance aspects

On average, only 10 to 15% of thehouses in Iran are insured for fire.Earthquake cover is not automaticallyincluded but has to be purchased sep-arately.

– The insurance penetration in theregion around Bam is likely to bemuch lower – on account of the lowstandard of living there.

– Although many of the country’slarge oil or gas industrial plants arecovered by insurance, none of themare located in Bam. Insured lossesare therefore expected to be small.The economic effects will have to beborne mainly by the impoverishedpopulation itself. Besides having tocope with the immeasurable humantragedy, the region thus also facescomplete economic collapse.

– By far the highest concentration ofliabilities anywhere in the country is to be found today in the capital,Teheran. In the event of a strongearthquake, the insurance industrywould have to reckon with lossesamounting to several hundred mil-lion dollars.

Accumulation control and possiblythe introduction of limits of indemnityare techniques that can make the riskacceptable even in such extremely ex-posed regions.

Earthquake report TOPICS geo 2003 Munich Re

60

40

50

30

AzerbaijanArmenia

Iraq

Saudi ArabiaBahrain

Qatar

United Arab Emirates

Oman

Pakistan

Afghanistan

Turkmenistan

KermanKhorramshahr

Yazd

Esfahan

Arak

RashtMashhad

Birjand

Persian Gulf

Gulf of Oman

Caspian Sea

1

2

3

4

1

Bandar-e 'Abbas-

Zahedan--

-

-

Bojnurd-

Gorgan-

Theran(Teheran)

-

Ardabil-Tabriz-

Orumiyeh- -

Kermanshah- -

Ilam--

Hamadan-

-Kashan- -

-Dezful-

Ahvaz-

Bushehr-

Shiraz- -

0 200 400 km

Fig. 1: Seismic hazard map of Iran and the position of the epicentre.

Zone 1: MMVIZone 2: MMVIIZone 3: MMVIIIZone 4: MMIX and

above

35

Cities in danger

Millions of people in Iran live in build-ings that would collapse in a strongearthquake; throughout the world it isseveral hundred million. It is to befeared that the future will bring moretragedies like the quakes of Gujarat(India, 2001) and Izmit (Turkey, 1999),with death tolls of some 20,000. Infact, it is only a matter of time beforea strong quake occurs again directlybeneath a city with over a million in-habitants and claims the lives of sev-eral hundred thousand people, as inTangshan (China) in 1979. Owing tothe rapid growth of many Third Worldmetropolises in highly-exposedregions, such a scenario has becomedistinctly more probable in recentdecades in spite of the possibilitiesprovided by modern earthquake engi-neering. Experts fear that if a strongquake were to be triggered beneathTeheran, which has a population oftwelve million and has already beendestroyed on several occasions in thecourse of its history, as many as onemillion people could die.

Munich Re has long been involved ina series of global initiatives set up bythe United Nations aimed at disasterreduction; in this way, it makes a sub-stantial contribution in its own field ofcompetence to the sustainable eco-nomic and social development ofcountries in the Third World.

Earthquake reportMunich Re TOPICS geo 2003

Fig. 3: Earthquake intensity map (EMS scale) of the earthquake on26 December 2003 (Source: Inter-national Institute of EarthquakeEngineering and Seismology, Iran).

56ºE 57ºE 58ºE 59ºE 60ºE

31ºN

30ºN

29ºN

28ºN

Fig. 2: Historical survey of earthquake activity in the region (Source: International Institute ofEarthquake Engineering and Seismology, Iran).

58.00E 58.50E

30.00N

29.50N

29.00N

VI

VII

VIII

V

Bam

Epicentre

Bam

Fau

lt

IXArg-e BamBam airport

New Arg of Bam(Arg-e Jadid)

ParavatAbaragh

Source: International Institute of Earthquake Engineering and Seismology, Iran.

Main Shock Ms6.5

36

Summary

The quake on 26 December 2003 dev-astated 70% of the city of Bam; morethan a third of its inhabitants werekilled. The reason for this horrificscene was the poor quality of thebuildings. What makes it all the moretragic is that the region had been clas-sified as a highly exposed zone evenbefore the earthquake, so that thebuildings should have been of saferconstruction.

The catastrophe demonstrates onceagain the connection that existsaround the world between a lowstandard of development and a highnumber of fatal victims. Even if thehazard is known, the people in suchregions simply do not have themoney to build earthquake-resistanthouses as a means of cushioning oravoiding catastrophes.

Earthquake report TOPICS geo 2003 Munich Re

26, 27The 2,000-year-old citadel in Bam – in 1975 (left) and after the earthquake.

26

37

Earthquake reportMunich Re TOPICS geo 2003

Date Region Magnitude Fatalities

23.1.1909 Borujerd 7.3 6,000

1.5.1929 Kopet Dagh 7.4 3,330

2.7.1951 Demavend 3,000

1.9.1962 Buyin-Zara, Quazvin 7.3 12,225

31.8.1968 Khorasan, Dasht 7.3 12,000

10.4.1972 Fars, Firuzabad, Ghir 7.0 5,044

24.11.1976 Northwest, area on border to Russia 7.3 5,000

16.9.1978 Tabas 7.7 20,000

20/21.6.1990 Gilan Province, Manjil-Rasht 6.2 40,000

26.12.2003 Bam 6.5 40,000+

The ten most devastating earthquakes in Iran in the last one hundred years

27

38


Rare events? 08

Wind and storm surge damage at a hotel complex in the Bermudas. The damage caused by Hurricane Fabian reached from the roof of thiscomplex to its foundations.

28

Tropical cyclones in 2003 caused exceptional losses in the Bermudas, Korea, and Canada. The tropical cyclone season presented a mixed picture in the meteorologists’ and insurers’ statistics.

39

Hurricane Fabian:

Insured losses in the Bermudas:US$ 300m. Largest hurricane insured loss onthe Bermudas to date: HurricaneEmily in 1987 with US$ 35m

Fabian hit the Bermudas on 5 Sep-tember 2003 at wind speeds of195 km/h (10-minute mean). With anintensity of 3 on the Saffir-SimpsonScale, this was the strongest hurri-cane since 1926 on these islands,which in the last decade have devel-oped into a significant offshore finan-cial centre. In addition to the lossesdirectly caused by the wind through-out the islands, there were also lossesin the coastal regions due to stormsurge and wave impact. The sea levelrose by as much as 7 m, especially in

As in the years 1998–2002, windstormactivity in the Atlantic (US/Caribbean)was, with a total of 16 tropical cyc-lones, well above the long-termaverage of just under 10 (referenceperiod: 1950–2002). On the otherhand, the seven hurricanes (tropicalcyclones in the Atlantic with windspeeds above 118 km/h, i.e. at leaststage 1 on the 5-stage Saffir-SimpsonHurricane Scale [SS]) were more orless average.

In the Northwest Pacific (SoutheastAsia), the year ended with 22 ob-served tropical storms (17 of typhoonforce, corresponding to at least SS1).Windstorm activity in this region wasthus below average in 2003; by wayof comparison: the averages in theperiod 1950–2002 were about 26 fortropical storms and 20 for typhoons.

Tropical cyclones did not cause anyoutstanding losses for the insuranceindustry in 2003. With insured lossesof US$ 1.7bn, the costliest event ofthe year was Hurricane Isabel, whichattracted intense media coverage be-cause of the very high wind speedsover the Atlantic (at times SS5) andthe associated threat to the US EastCoast. Fortunately, the hurricaneweakened considerably just beforelandfall in North Carolina and hit theregion around Cape Hatteras, wherethe population density and concentra-tion of values are both comparativelylow.

Nevertheless, the insurance industrywas hit by losses in 2003 that meantnew record highs for the areasaffected:

the south of the main island. Thereare no dependable estimates on theextent to which water damage con-tributed to the overall loss. But theanalysis of insured losses reveals thathotels and coastal resorts were affect-ed more severely on average thanresidential buildings and other com-mercial risks.

The average loss ratio in the countryas a whole (the ratio between theoverall insured loss and the overallsum insured) was in the lower single-digit percentage range. There weresome risks – especially hotels withbusiness interruption cover – wherethe individual loss ratio was muchless favourable, however, with valuesexceeding 20%. How does this vulner-ability compare with the experiencederived from hurricanes in the United

Rare events?Munich Re TOPICS geo 2003

The storm surge whipped up by Hurricane Fabian washed sand and rubble into numerousbuildings near the shore. Even if high wind-load standards are incorporated in the buildingcode, they will not reduce such losses to any substantial degree.

29

40

States and windstorms in Europe? Theanswer is indisputable: given a certainwind speed, the vulnerability of build-ing and contents values in the Bermu-das is much lower than in the coastalareas of the United States but some-what higher than in Europe. This ispartially due to the difference be-tween tropical cyclones and extratrop-ical storms in terms of the secondaryhazards of torrential rain and stormsurge. However, the main reason forthe overall low vulnerability in theBermudas is the generally meticulousattention to constructing buildingsand roofs with solid materials.

Hurricane Fabian was another cleardemonstration that if constructionmethods take the hazards into account,they can reduce the claims burdenfrom catastrophe events considerably.This has a positive effect on the riskpremiums. There is one thing thatshould not be forgotten, however.Fabian was only an SS3 hurricane inthe Bermudas and thus far from beinga potential accumulation scenario. Ifthe islands had been hit by HurricaneIsabel (2003) with its intensities of up to SS5, it would have potentiallybeen a probable maximum loss (PML)scenario.

Typhoon Maemi:

Insured loss in South Korea:US$ 500m.Largest insured typhoon loss inSouth Korea to date: Typhoon Rusain 2002 with US$ 170m

Maemi, the tropical cyclone that wascalled a super-typhoon in many re-ports, crossed parts of South Koreaon 12 September 2003 and presentedthe insurance industry with a bill thatwas more than three times as high asthe record established the previousyear.

In meteorological terms, Maemi wasanything but an exceptional event,though. Only one weather station on the island of Cheju off the Koreanmainland actually recorded a recordwind speed of 216 km/h (in peakgusts). All the other 70 wind speedrecordings (two of which were also made on Cheju) were in the110–180 km/h range. In Pusan, the top reading was 154 km/h, a valuethat had already been reached orexceeded on two occasions since1950. The position in Ulsan was quitesimilar: in the previous 50 years therehad been three events in whichMaemi’s top speed of 120 km/h wasexceeded.

Unlike Typhoon Rusa in 2002, Maemiwas a relatively dry storm. In the ex-posure centres of Pusan and Ulsan,the 24-hour precipitation depths were64 and 99 mm respectively (corres-ponding to a return period of lessthan one year). Only the southerncoastal regions registered values of200–300 mm/24 h and locally of up to410 mm/24 h.

Rare events? TOPICS geo 2003 Munich Re

The figures show the highest peak gusts (in km/h) recorded during Typhoon Maemi. The 10-minute average wind speeds are approx. 20% lower in each case (source: KoreaMeteorological Administration; cartographic processing: Geo Risks Research Dept., Munich Re).

41

If the geographical distribution of in-sured values in Korea – essentiallycommercial and industrial risks – istaken into account, the followingoverall assessment of the Maemi lossemerges:

– The vulnerability to (windstorm)damage of the risks currentlycovered by insurance is reduced bythe existence and monitoring ofbuilding codes. The local buildingcodes for Pusan and Ulsan aregeared to a wind speed of 144 km/h.On Cheju the minimum require-ments are 162 km/h.

– Compared with the loss experiencefrom tropical cyclones in the UnitedStates and extratropical storms inEurope, Maemi is to be considered –in terms of its intensity over theKorean mainland – as “medium”.The average loss ratio (cf. HurricaneFabian in item 1) was below theexperience values for hurricaneswith landfall in the United States,but clearly above the empiricalvalues of gales in Europe in recentyears (e.g. Lothar, 1999).

Hurricane Juan:

Insured loss in Canada: US$ 60m.Largest hurricane loss to date:Hurricane Hazel in 1954 withUS$ 100m (total economic loss,original value at that time); since1950, no insured loss from a topical cyclone has been higherthan US$ 5m.

Although the loss burden was gener-ally moderate, Juan was an excep-tional hurricane. What made it excep-tional was its meteorological data:when Juan reached the Canadian


A loading crane toppled by the wind in the port of Pusan.

30

42

coast near Halifax at a latitude ofabout 44.5°N, it was a full-blown trop-ical cyclone with an intensity of SS2and wind speeds of up to 158 km/h (ingusts). The region had not been hit bya hurricane of comparable strength in40 years. Even Hazel – the most dam-aging tropical storm in Canada up tothat time – had reached peak gusts ofonly 115 km/h over Canada. Extremeflooding on account of copious rain-fall with values of up to 225 mm/24 h

caused 83 deaths and an (uninsured)loss which – in current terms – wouldprobably have come to more thanUS$ 1bn.

Rare events? TOPICS geo 2003 Munich Re

31

Satellite image of Hurricane Juan shortly before it reached the Canadian mainland near Halifax on 29 September 2003. In spite of the surfacetemperature of the North Atlantic already being quite cool at this high geographical latitude, the typical spiral structure of the hurricane vortexremained completely intact until landfall.

Halifax

43


Statistics on tropical cyclones in the North Atlantic and theNorthwest Pacific reveal an increase in windstorm activityin the period 1950–2003.

It is noticeable that in both regions the increase onlyrelates to less extreme events. There is no change in theaverage number of hurricanes and typhoons. The reasonsfor the increase in the frequency of less extreme events

are still unclear. For the time prior to 1970 at least – whencomplete monitoring of tropical cyclones became possiblewith the aid of satellites – changes in the observationmethods could influence the quality of the data. But inrecent decades there has been an increase in the expanseof sea areas with water temperatures exceeding 26°C – aprecondition for the growth of tropical cyclones – and thiscould also raise the frequency of this type of storm.

Windstorm activity in the North Atlantic and the Northwest Pacific 1950–2003

The annual number of tropical cyclones and hurricanes in

the period 1950–2003 in the North Atlantic (including theCaribbean). The 54-year average is 10 for tropical storms(including hurricanes) and 6 for hurricanes.

The annual number of tropical cyclones and typhoons in

the period 1950–2003 in the Northwest Pacific.The 54-yearaverage is 26 for tropical storms (including typhoons) and20 for typhoons.

201816141210864201950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000

Tropical cyclonesNumberLinear trend

Number

40

35

30

25

20

15

10

5

01950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000

Number

Hurricanes NumberLinear trend

Tropical cyclones NumberLinear trend

Typhoons NumberLinear trend

44

Natural hazard business in propertyinsurance calls for increasingly pre-cise estimates of threat potentials(e.g. terrorism) and their interrela-tions. The only way to ensure successin the long term is to project appropri-ate prices for current and future risks.The hallmark of quality in the deal-ings of insurers and reinsurers is first-class risk management, because itguarantees the sustainability of busi-ness relations.

In order to meet the requirements ofsustainable risk management, MunichRe’s Geo Risks Research Dept.(GEO)has developed a new method de-signed to ensure in the long term thatthe complex business of insurance re-mains calculable and profitable: thismethod is called geographical under-writing.

In our last issue of topics – NaturalCatastrophes, we presented the theor-etical approach behind this method; in this issue, we will describe the ini-tial experience we have gainedthrough applying it in practice. Wewill also sketch out the steps thatneed to be taken prior to implement-

ing the new processes in the manage-ment of natural hazards and man-made risks and in the course of lossanalysis.

The process of geocoding

The starting point for future-orientedcontrol and optimisation of portfolioswith regard to natural hazards andman-made risks is to know as precise-ly as possible the geographical loca-tion of the risks concerned (georefer-encing). Anonymised client portfolioand loss data provide the basis forthis. Many companies in the insur-ance industry are already in a positionto supply this important information.

The geocoding of portfolios of liabil-ities and losses (georeferencing)makes it possible to produce analysesand computer models in accordancewith the geographical underwritingmethod.

Geocoding may be performed usingvarious levels of detail (addresses,municipalities, postcodes) (Fig. 1). Inview of the precision and quality re-quired in the future, coarse geocod-

ing, e.g. at country or state level, is no longer sufficient. The spatial defin-ition (risk allocation) on the basis ofCRESTA zones (cf. www.cresta.org)often used today in property insur-ance must be refined and made moretransparent for important core mar-kets and risk types.

The level of detail required in thegeocoding process depends on thehazard for which the risk exposure isbeing examined. Given the modellingand simulation programs availablenowadays, a distinction can be madebetween three levels of spatial detail(resolutions):

Coarse: Data with a coarse resolutionusually only relate to regions, coun-tries, federal states, or very largepostcode units. Such a resolution maybe used for large-scale hazards orscenarios (e.g. the effects of climatechange, environmental influences) butit is of limited value as far as manyother (natural) hazards are concerned.

Medium: If data of medium qualityare available – precise postcode units,municipalities, and local authority

Geographical underwriting –Applications in practice 09


AnalysesComputer models Scenarios

Risk-adequateaction

GeocodingPortfolio and/or loss data

Creating transparency in the underwriting process is prerequisite to sustainability andlong-term efficiency in risk management. The key to success is provided by a new methodand technique: geographical underwriting.

The process of geocoding

The insurance companies' portfolio and/or loss data are geocoded on the basis of different levels of spatial detail. This is crucial for analyses of specific portfolios, computer models, and scenario calculations. These ultimately provide an essential basis for underwriting and riskmanagement decisions.

45

Geographical UnderwritingMunich Re TOPICS geo 2003

2-digit postcodes (yesterday) 5-digit postcodes (today) Address level (today/tomorrow)

Fig. 2: Levels of detail and their suitability for computer modelling and simulation

Ideally, as much liability information as possible should be available with a high geographical resolution. It is then possible to aggregate to themost suitable level of detail, depending on the issues and objectives involved. This only functions from a lower to a higher level, however, i.e.from addresses (individual coordinates) to the next aggregation level upwards (postcodes, countries).

highmediumlowNote: Suitability of analyses

Fig. 1: Risk allocation with varying levels of detail

Using a simulated client portfolio as an example, the input datasets(e.g. liabilities) are visualised in various resolutions. The differentcolour intensities reflect the different liability concentrations. InGermany referencing is performed almost exclusively on the basis

of 5-digit postcodes (in line with CRESTA). The address level result – inthis case a property treaty with about one million addresses – gives anidea of how much better analyses can be if this level of resolution canbe attained. Switching from 2-digit postcodes to data based on specificaddresses increases a portfolio's transparency quite considerably.

Countries/federal states/2-digit postcodes

CRESTA (by country)/local authorities/5-digit postcodes

Streets/addresses/GPS georeferencingto longitude/latitude

Windstorm Earthquake Flood Man-made risks

46

boundaries – it is possible to producequite realistic analyses for certain nat-ural hazards like windstorm or earth-quake. The majority of CRESTA zonesare based on this classification.

Fine: If the focus of the hazards to beanalysed is on small areas (as is thecase with terrorist attacks and indus-trial accidents, for example) or if thequestion as to whether damage oc-curs or not depends on a difference of only a few metres (floods, hail), itwill be necessary to work with moredetailed geodata. This means the kindof information provided by individualaddresses and GPS (Global Position-ing System), which yield exact resultsdown to a few metres.

The current position of geocoding

All geocoding services at addresslevel are currently based on datacollected for vehicle navigation androute planning purposes. At presentthere are two large firms whose offer-

ings are used by all downstream ser-vice providers. These services arecalled location based services (LBS) in the trade and may be combinedwith the spatial analysis tools of Geo-graphical Information Systems (GIS).

The data have a high level of cover-age particularly in the important in-dustrial countries. These countries are mainly the core markets in theinsurance sector; the US market andEurope are very well covered. Theywill soon be joined by the Asianmarkets.

Geocoding is technically quite maturenowadays and has been tested suffi-ciently for it to be used in practice.What is more, external services canbe used via the internet or integratedin a company’s own IT landscape.

From the insurance point of view, notonly risk analyses but also clients’master file data can be examined toensure their postcodes are correct

and can thus be used for efficientgeo-marketing activities.

From the underwriting point of view,geocoding can be used in both treatyand facultative business when it hasbeen enhanced to address level. Tothis end, a preliminary series of feasi-bility studies was performed in whichreal portfolios were analysed andtheir technical and operative viabilityverified.

In the following sections we presentapplications (computer models,scenarios) used in risk managementthat are possible on the basis ofgeocoding.

Case study: Holistic accumulation

study

For the purposes of “transparency”,visualising the spatial exposure distri-bution or the spread of risks is be-

Geographical Underwriting TOPICS geo 2003 Munich Re

Fig. 3: Detailed accumulation and spreadeffects from several individual portfolios (red, blue: insured risks from two differentportfolios).

Fig. 4 a: Postcode areas potentially exposedto flood (yellow).

Fig. 4 b: Individual risks exposed to the flood hazard (yellow) on the basis of datawith exact addresses (red: all risks).

47

coming increasingly important. As theexample of Manhattan demonstrates,centres of accumulation can be dis-played precisely on the basis ofaddresses. In this example, the blueand red dots represent two portfolios,thus revealing spatially cumulativeeffects and the spread of risks. Theability to spatially analyse and visu-alise a complete accumulation, whichmay comprise several separate port-folios or client accounts, is of particu-lar importance in this connection (Fig. 3).

Case study: Levels of detail and

their effect in connection with

accumulation

In order to calculate and analyse theproperty PML (probable maximumloss) in, for instance, areas prone toflooding, it is a great advantage tohave exact knowledge of the risk (risksituation). The example (Fig. 4) showshow various levels of spatial detail in-

fluence the PML calculations on thebasis of coarse and fine observations.The violet area in the diagram on theleft is the area prone to flooding; theyellow areas are postcode areas thatcould be affected by a flood. As entirepostcode areas are only flooded inexceptional cases, determining theareas that really will be affected isvery difficult and prone to error. Thereason for this is that the risks are notdistributed evenly but have additionalpoints of concentration (e.g. tradingestates, built-up areas). Exact ad-dresses were used for the data in theright-hand map. All the yellow dotscoincide with the flood-prone area.

How strong an impact the level of de-tail may have on the PML calculationwill depend on the composition andspread of the portfolio. In this specificexample (Fig. 5) the PML derivedfrom the postcode calculation (greenline) is almost one-third higher thanthe address-based PML (blue line).

There are likely to be similar correla-tions for many portfolios, although di-vergent shifts are also conceivable.

Case study:The multi-location

problem

Many portfolios contain multi-locationpolicies, which are difficult to identifyand analyse exactly – especially in the case of treaty business. Theseoften involve housing associations orchains of companies, where the ad-dress quoted only refers to the head-quarters while the policy actually in-cludes many other risks at differentlocations as well. This can lead to asituation in which existing exposuresare not identified or are incorrectlyevaluated. This may prove to be adrawback both for insurers and forreinsurers.

In this example of a multi-location risk(Fig. 6), the known situation (red dot)would not lead one to expect an ex-


Fig. 5: Levels of detail and their possibleeffects on PML calculations.

Fig. 6: The multi-location problem (overviewand detail; red: address of the multi-locationpolicy; green: actual location of the risks).

Loss as a percentageof sum insured

Return period in years

4

3

2

1

0

0 10 100 1000 10000

48

posure to flood. If all the individualrisks subsumed under this policy areobserved, however, it turns out that anot inconsiderable number of theseindividual risks are highly exposed.This frequently encountered problemcan be solved if all the individual riskscollected together in one multi-locationpolicy are available with their addressand the corresponding indemnitylimit.

Case study: Improved loss

management

A small section of the August 2002floods in Germany reveals the enor-mous potential for analyses if data-sets with exact addresses are used(Fig. 7). The area that has been flood-ed is marked on the basis of satellitedata (cross-hatched in red). The

yellow borders and captions relate to postcode areas and clearly showhow coarse this “grid” is for this kindof analysis. The individual risks areshown as red dots. A clear distinctionbetween risks that are affected andthose that are not is both possibleand important in order to be able toestimate loss potentials after anevent. With the help of loss advices(green triangles) certain loss patternsmay be derived which can be used for such purposes as claims handling.Claims relating to risks located out-side the derived loss zones can beidentified accurately so that any casesof doubt can be clarified.

Case study: Proactive underwriting

The expected risks of change (e.g. dueto climate change) will very probablylead to an increase in natural catas-trophes and extreme events.

Experience gained in the past is there-fore of limited value as an indicator ofthe future. For this reason, proactiveunderwriting is required in which the,in all probability, higher losses of thefuture are appropriately considered.

The same applies to man-made haz-ards, the intensity and frequency ofwhich do not follow scientific pat-terns. If these altered conditions inrisk management are to be visualised,all possible scenarios must be takeninto account and underwriting toolsmust be developed to deal with them.

Geocoding may also be used for thepurposes of proactive risk control (riskclearance). This is the case, for ex-ample, with a Munich Re rating toolfor facultative risks (Fig. 8): the risks

Geographical Underwriting TOPICS geo 2003 Munich Re

Fig. 7: Loss management can be improved considerably by linking up portfolio data withcurrent weather or flooding information.

49

in question are geocoded interactivelyon the basis of their addresses (reddots) and compared online with risksthat have already been written andgeocoded (triangles). On the basis ofthese data, the maximum loss amountin the direct vicinity of the new risk(s)is modelled (violet circle, all riskswithin this circle). In this way, an un-desirable accumulation situation canbe identified and calculated in ad-vance, and if necessary avoided.

Next steps

We will only be able to manage therisks we assume to best advantage ifwe can assess them as accurately aspossible. The advantage of the methodoutlined above is that accumulativeeffects are identified in good time,allowing underwriting options andrisk control to be activated quicklyand effectively for the purposes ofsustainable portfolio management.

The following conditions must be met:

– The flow of data between insurerand reinsurer must be further inten-sified and optimised. Examples from the US market show that thisis possible.

– In order to evaluate portfolios better,the data must be available at least inmedium and preferably in fine de-tail. Only the latter makes it possibleto conduct an analysis using anygeographical spatial pattern desired.Until this step can be performed byall market players autonomously,service providers or reinsurers canfill the gap.

– Models already in place for naturalhazards need to be further refined.Models for other hazards (e.g. ter-rorism) still have to be developed.

These approaches have already beenrealised successfully in a number ofareas at Munich Re. The establishmentof a geo-data platform is one of thecentral components; many applica-tions will be able to access this analy-sis potential. It is essential that fulladvantage be taken of these en-hanced geo-information services inthe future.


49

Fig. 8: Timely identification of accumulationsituations by means of proactive risk control.The circle denotes the area of the maximumloss amount.

50

From 1 to 12 December 2003, some5,000 experts from almost 200 coun-tries – including politicians from 180countries – gathered in Milan for thisyear’s climate summit. The Confer-ence of the Parties (COP) has the ob-jective of establishing globally bind-ing, effective climate protectionmeasures.

The central issue of the discussions isthe much-cited 1997 Kyoto Protocol,which sets out emission limits for alarge number of countries. The Proto-col is not only supported by all themajor emerging and developingcountries (e.g. Brazil, South Africa,India, China) but has also been rati-fied by 120 countries, including largeindustrial nations like the EU memberstates, Japan, and Canada. Australia,Russia, and the United States are still missing from the list – but thearrangements are that the KyotoProtocol can become globally bindinginternational law as soon as it hasbeen ratified by either the UnitedStates or Russia.

The US government accepts climatechange as a fact – but is still lookingfor alternative ways of dealing with it.It is calling for even more evidencethat environmental change is attribut-able to human activity and is concen-trating its efforts on research into newclimate protection technologies. Thisis the policy increasingly being fol-lowed by Australia too. Against thisbackdrop it was clear that the whole

world and particularly the mediawould concentrate on the stanceadopted by Russia and so reportingfocused on the question: Will Russiaratify – yes or no? In the months priorto the meeting, the signals fromMoscow – even from individual min-istries – were sometimes positive,sometimes negative. The ratificationissue will probably continue to cause

heated debate; but experts and in-siders still expect the Kyoto Protocolto be signed and sealed in the fore-seeable future.

What subjects did the negotiations

in Milan address?

– The climate conference in Milanclosed further important gaps in the provisions that shape the KyotoProtocol. This included defining themodalities for the recognition of af-forestation and reforestation proj-ects in developing countries. Suchprojects must not only be designedto absorb CO2 permanently, theymust also be scrutinised and per-formed in accordance with ecologi-cal and socio-economic criteria.

– It was also important to settle whatis to happen with reforestation pro-jects and plantations that have beendestroyed after being recognised as CO2 sinks. If a forest has beencleared or has burned down, itmakes sense to take this fully intoaccount and deduct the certifiedemission reductions that have beengranted.

– Another topic addressed was thefuture of the climate convention“after Kyoto”. There was a discussionof the issues and obligations thecommunity of states should be aim-ing for after the first commitmentperiod (2008–2012).

The 2003 climate summit and the Kyoto Protocol – Can we wait for Russia? 10


“Climate change remains the most important global challenge to humanity, and its ad-verse effects are already a reality in all parts of the world” – this was a key message in thefinal agreement of the climate summit in Milan. Even so, it would seem the 9th Confer-ence of the Parties did not achieve very much. Or did it?

On time for the climate summit in Milan, thebanks and insurers that are taking part in theUNEP Finance Initiatives published a paperexplaining their position with regard to emis-sions trading. It presents the most importantaspects in clear terms.

32

51

– One particularly pleasing outcomeof COP 9 was that a number ofcountries committed themselves todeveloping their renewable energysources. These included numerousArabian and African countries(Egypt, Djibouti, Morocco, Syria, and others).

Is Russia alone the key to climate

protection?

Reporting directly from Milan, theBBC said, “Russia holds climate key”.Although the Kyoto Protocol and itsratification represent the first andhence one of the most importantglobal steps towards climate protec-tion, there is no sense in simply wait-ing for ratification to happen. Thereare other important developmentsrunning parallel that also promoteclimate protection.

– The banks and insurers taking partin the UNEP Finance Initiatives pro-gramme have stepped up their dia-logue with the politicians – for thefollowing reason: a good basis forcooperation is necessary becausethe Kyoto Mechanisms specified inthe Protocol will generate many en-vironmental projects throughout theworld and these will depend on sup-port from financial services providers.

– At the beginning of June 2004,Germany will host a major worldconference on renewable energies(Renewables 2004). It is being or-ganised by the federal government,which gained further support for thisinitiative at the Milan summit. “Re-newables 2004” will undoubtedlyhave a stimulating effect – also onthe economy. The significance ofrenewable energies will be given afurther boost.

– 2005 sees the start of internationalemissions trading in the EU memberstates. Approx. 12,000 industrialfacilities in such sectors as energy,steel, and minerals will be givenclearly delineated emission alloca-tions and limits. Energy efficiencyought then to be seen relativelyquickly as an economic opportunity,which will promote the implementa-tion of climate-friendly and ecologic-al technologies. In this context,Munich Re is investigating possiblenew fields of business for the insur-ance industry.

A further question is whether the nofewer than 120 countries that havealready ratified the Protocol shouldnot forge ahead with their climateprotection commitments parallel tothe Kyoto process. After all, thesecountries have more than underlinedthat they intended to implement anearnest climate protection policy. Then why wait for Russia?

“Climate change remains the mostimportant global challenge to human-ity, and its adverse effects are alreadya reality in all parts of the world” wasone of the final statements made byministers and representatives from188 nations around the world at COP9 in Milan. As the effects of global cli-mate change become more and morenoticeable, every possible effort mustbe made to save the climate – andwithout delay.

The 2003 climate summit and the Kyoto ProtocolMunich Re TOPICS geo 2003

52

Picture credits

01 Associated Press, Frankfurt am Main, Germany02, 03 Reuters, Berlin, Germany04 Associated Press, Frankfurt am Main, Germany05 Reuters, Berlin, Germany06 Associated Press, Frankfurt am Main, Germany07, 08 Reuters, Berlin, Germany09 Associated Press, Frankfurt am Main, Germany10 Reuters, Berlin, Germany11 Associated Press, Frankfurt am Main, Germany12, 13, 14, 15, 16 Reuters, Berlin, Germany17 Loral Skynet, Bedminster, USA18 Reuters, Berlin, Germany19 Associated Press, Frankfurt am Main, Germany20 Reuters, Berlin, Germany21 Associated Press, Frankfurt am Main, Germany22 fotocommunity, Bonn, Germany23, 24, 25 Reuters, Berlin, Germany26, 27 Associated Press, Frankfurt am Main, Germany28 Reuters, Berlin, Germany29 Mark Bove, Princeton, USA30 Reuters, Berlin, Germany31 NOAA, Washington, USA32 Munich Re archives33 Reuters, Berlin, Germany

MRNatCatPOSTER1, 2, 3, 4 Reuters, Berlin, Germany

Graphics, maps, tables (unless stated otherwise): © Geo Risks Research, Munich Re


33 A waterspout near Limassol on the southcoast of Cyprus on 27 January.

© 2004Münchener Rückversicherungs-GesellschaftKöniginstrasse 10780802 München GermanyTel.: +49(0)89/3891-0Fax: +49(0)89/399056http://www.munichre.com

Responsible for content

Corporate Underwriting/Global Clients, Geo Risks Research Dept.

Persons to contact

Thomas LosterTel.: +49(0)89/3891-5287Fax: +49(0)89/3891-75287E-mail: [email protected] WirtzTel.: +49(0)89/3891-3453Fax: +49(0)89/3891-73453E-mail: [email protected]

Order number

German 302-03970English 302-03971French 302-03972Spanish 302-03973Italian 302-03974

Printed by

F-Media Druck GmbH, Weissenfelder Strasse 4a, 85551 Kirchheim/Heimstetten, Germany

topics geo 2003 - sfu.ca re(2004).pdf · natural catastrophes in 2003 – pictures of the year...

Documents