Weather and Climate Extremes 11 (2016) 70–79

Contents lists available at ScienceDirect

Weather and Climate Extremes

journal homepage: www.elsevier.com/locate/wace

Trends in weather related disasters – Consequences for insurers and

society
Peter Hoeppe
Head of Geo Risks Research/Corporate Climate Centre, Munich Re, Munich, Germany

art ic l e i nf o a b s t r a c t

Article history: As extreme weather events affect the core business of insurance this industry has quite early addressed
Received 13 July 2015 potential effects of natural climate cycles and global warming on natural catastrophe hazards. Munich
Received in revised form Re’s experts have been researching loss events caused by natural hazards around the globe for 40 years.
16 October 2015
These losses are documented in the NatCatSERVICE database currently documenting more than 36,000
Accepted 19 October 2015
Available online 31 October 2015
single events. The analyses of the NatCatSERVICE data clearly show a high interannual variability, in some
regions decadal oscillations, and a long term trend to an increase in the number of natural catastrophes
Keywords: around the globe, with ever growing losses. The trend curve indicating the number of loss relevant
Disaster natural catastrophes worldwide reveals an increase by a factor of about three within the last 35 years.
Weather
As the rise in the number of natural catastrophes is predominantly attributable to weather-related
Losses
events like storms and floods, with no relevant increase in geophysical events such as earthquakes,
Climate change
Economic impact tsunamis, and volcanic eruptions, there is some justification in assuming that changes in the atmosphere,
and global warming in particular, play a relevant role.
However, the main contribution to the upward trend of the losses caused by natural catastrophes
comes from socio-economic/demographic factors such as population growth, ongoing urbanization and
increasing values being exposed. Prevention measures, especially flood protection programs, on the other
hand have a high potential to even reduce losses while the hazard has increased. Because of such factors
influencing the loss trends a clear attribution of at least part of the effects to global warming is very
difficult. There is, however, an increasing number of studies, which show significant increases in losses in
some regions and for some perils even after they have been normalized to the exposed values today.
Looking at trends of extreme weather events and their effects, natural climate variability has to be
considered. Short term oscillations such as ENSO as well as decadal oscillations in hurricane (Atlantic
Multidecadal Oscillation) or typhoon activity (Pacific Decadal Oscillation) still play a dominant role on
the variability of losses caused by weather extremes. As global warming will continue in the coming
decades, its contribution to increasing natural catastrophe losses will become more prominent, a pro-
jection also given by the 5th assessment report of the Intergovernmental Panel on Climate Change (IPCC
2014).
As long as the risks from weather related disasters stay calculable and the insureds can afford a risk
adequate premium there is no danger that such risks become uninsurable. Insurers, however, have to
invest more resources into analyses of trends and have to assure that premiums for the risks they cover
always reflect a dynamic hazard pattern.
There is no sensible way to interfere with natural climate oscillations influencing natural catastrophe
losses. Humankind, however, still has the chance to avoid catastrophic increases of losses caused by
global warming driven weather extremes by ambitious climate protection and adaptation measures.
& 2015 The Author. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND
license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction humans itself. Therefore an extreme event like an earthquake or

storm occurring in a region without human settlements never will
A disaster or catastrophe generally is defined as an event which be counted as a disaster. Some of the disasters cause primarily ma-
is causing significant damages and losses to human belongings or terial losses, especially in the richer countries with precautionary life
saving measures in operation, others in the poorer countries cause
primarily human losses. In the emerging countries both material
E-mail address: [email protected] and human losses can reach significant levels as the exposed values

http://dx.doi.org/10.1016/j.wace.2015.10.002
2212-0947/& 2015 The Author. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
 P. Hoeppe / Weather and Climate Extremes 11 (2016) 70–79 71

Table 1
Definitions of intensity classes of natural disasters as used in Munich Re NatCatSERVICE database in 2015. Monetary values are total economic losses in million US$. The
factors between classes of different income groups are chosen approximately proportionally to the income group ratios defined by Worldbank.

are already high while precautionary measures have not been im- assures a high reliability of the data.
plemented yet. The loss data (both material and human) of natural This paper has the objective to share insurance industry data
disasters provide an independent source of information complem- on natural disaster losses with the scientific community and thus
enting the natural scientific data of meteorological or geophysical feed the scientific discussion on their trends and the processes
services. Especially for smaller and regional weather related events behind them. Such knowledge is the basis for risk mitigation and
like thunderstorms they can provide additional information, when prevention measures.
they have not been detected by the wide grid of routine meteor-
ological measuring stations. This is not valid for the geophysical events
as the seismic activity today is monitored in a worldwide network and 2. Methods
every event, even below the threshold of a damaging level, is detected.
In countries with a high insurance penetration like in the ones 2.1. Datamanagement at Munich Re NatCatSERVICE – methodology
in North America, Europe and parts of Asia losses caused by nat-
ural disasters have been documented for several decades. This is 2.1.1. Event data
an important information and basis for the risk assessment pro- All loss events caused by natural hazards resulting in property
cess of the insurers and reinsurers. The data of insured losses are damage and/or bodily injury are recorded in the NatCatSERVICE.
the most reliable natural catastrophe loss data available. The as- The objective of an entry in the database is to describe a cata-
sessment of direct total economic losses, i.e. the costs to re- strophe as detailed as possible. A full entry record consists of up to
construct, repair and replace the damaged infrastructure and pri- 200 attributes. The following are the most important: Date and
vate material losses, normally is based on assessment tools which duration, categorization of peril, geographical information, hu-
comprise quite some uncertainties. The uncertainties rise even manitarian and monetary impact.
more when secondary economic losses occur like business inter-
ruptions, people leave permanently the region of the catastrophe 2.1.2. Catastrophe classes
(loss of labor) or tourists avoid the affected areas. Depending on their monetary or humanitarian impact, the
Munich Re, founded in 1880, is a leading global reinsurance documented events are classified into four classes, ranging from a
company. The business model of a reinsurer is to cover parts of natural occurrence with small economic impact (1) to a major
major risks from primary insurers which they cannot cover totally natural catastrophe (4). The upper end of the scale includes the
themselves because of lack of enough risk capital or because being formerly used class of “great natural catastrophes”, which are part
confined to regional business activities they cannot diversify these of event class 4. In line with definitions used by the United Na-
with other risks in other regions. A relevant part of the risks tions, a “great natural catastrophe” clearly overstretches the af-
covered by global reinsurers are the risks caused by natural perils. fected region’s ability to help itself and interregional or interna-
They have the potential to cause so called accumulation risks, tional assistance is consequently required. As a rule, this will be
which means that they can affect different insurance business the case when thousands are killed and hundreds of thousands are
lines at a time and thus generate very large insured losses. Munich made homeless or if the overall loss reaches extreme dimensions,
Re has a long tradition to be a market leader in providing in- depending on the economic capacities and conditions of the
surance cover for natural catastrophe risks and thus has built up country concerned. These great natural catastrophes can be used
unique expertise in assessing and managing these risks. for long-term analyses starting in 1950 as such major disasters
For more than 40 years natural scientists at Munich Re have have always been reported in detail and the analysis is not dis-
been analyzing natural hazards and the losses they are causing. For torted by a reporting bias. In Table 1 the four classes of loss events
this purpose, Munich Re has set up the NatCatSERVICE, the most are defined, an additional class “0” for natural extreme events
comprehensive of the three existing global natural catastrophe without noteworthy losses exists but these events are not con-
loss databases (the other two are Sigma/Swiss Re and EmDat/ sidered in all further number statistics. The threshold values be-
CRED). The Munich Re NatCatSERVICE database currently com- tween classes 1 to 4 listed in this table are in million US$ and
prises about 36,000 data sets of loss events caused by any kind of somewhat arbitrarily chosen and prone to potential further
natural peril. It documents on a global level major events starting changes. More important are the ratios or factors between the
in 1950 and all loss relevant events from 1980 onwards, providing same classes in countries with different income groups. Only by
information on their effects on national economies, the insurance consequently applying these factors one can account for the fact
sector and the population. A stringent quality control system that for example a 100 million dollar loss event in a developing
72 P. Hoeppe / Weather and Climate Extremes 11 (2016) 70–79

Table 2
Categorization of perils in hazard families, main events and sub perils as used in Munich Re NatCatSERVICE database

country means a more severe impact than the same loss magni- casualties. Such data are often copied and further disseminated.
tude in an industrialized country. The validation process in the NatCatSERVICE database checks the
quality and number of the sources referred to as well as the
2.1.3. Categorization of perils plausibility of the loss figures and the description of the event. An
The data are classified according to the perils into four hazard evaluation system has been developed for this quality check,
families (geophysical, meteorological, hydrological, climatological) which assigns every data record to a quality level on a scale from 1
each of which is further subdivided into main events and sub (very good) to 6 (inadequate). Data records on quality level 4, 5 or
perils (see Table 2). This classification has been standardized with 6 do not meet with the quality standards of the database and are
the other global natural catastrophe databases in order to make not used for analyses.
the data comparable.
2.1.6. Statistical significance of trends
2.1.4. Economic and insured losses The quantification of statistical significances of the trends
The Munich Re NatCatSERVICE database is focusing on financial shown in the following figures is based on T-Tests of linear Re-
losses. They are subdivided into two categories: insured losses and gression calculations. P-values (probability of error)o0.01 are
economic losses. Figures for insured losses are most reliable be- regarded to represent highly significant trends. For economic and
cause they reflect claims actually paid by insurance companies. insured losses exponential trend calculation has been made on the
Economic losses, however, are more complex to assess. It is im- basis of the linear trends of the logarithms.
portant to differentiate between “direct losses”, “indirect losses”
and “secondary/consequential losses”. As a rule, the assessment of
overall losses in NatCatSERVICE consists of direct (tangible) losses 3. Results of analyses of natural disasters
plus, in the event of business interruption, the resulting indirect
losses. 3.1. Current status of natural disasters
Besides the material losses the NatCatSERVICE also documents
humanitarian figures of disasters like the number of fatalities, Fig. 1 shows exemplarily the width of coverage of disaster data
injured and homeless people. in the NatCatSERVICE for the year 2014. In this figure the different
colors represent the different groups of perils. In red (“geophysi-
2.1.5. Sources and data quality cal”) all disasters are shown which have their origin in the earth
The NatCatSERVICE employs around 200 sources that have crust like earthquakes, tsunamis and volcano eruptions. Green
been identified as first rate for a particular region and/or type of (“meteorological”) represents all kinds of storms like tropical cy-
event. The groups of main sources are: clones, winter storms or thunderstorm related events. The color
blue stands for all kinds of flood related events as river floods,

Insurance industry information. flash floods or landslides which have been triggered by intense

Meteorological and seismological services. precipitation. In orange (“climatological”) all other weather related

Reports and evaluations by aid organisations or NGOs, govern- events are shown like heat waves, droughts or wild fires. The
ments, the EU, the UN, the World Bank and other development larger dots represent the large loss events. They are also described
banks. in some more detail in the figure. For 2014 the number of docu-

Scientific analyses and studies. mented events has been 980.

News agencies. In Fig. 2 the mean distributions of the numbers of events,
fatalities, economic and insured losses to the different groups of
Despite first-class sources, the analysis process is sometimes perils are shown for the 35 years 1980–2014 in the form of pie
fraught with problems. Typical challenges include false reporting, charts. Storms (41%) have been the most frequent events followed
the use of incorrect conversion factors and double counting of by floods (36%). Most of the fatalities, however, have been caused
 P. Hoeppe / Weather and Climate Extremes 11 (2016) 70–79 73

NatCatSERVICE

Loss events worldwide 2014

Geographical overview
Floods Floods
Winter damage
United Kingdom, Bosnia and Herzegovina,
USA, Canada, 5–8 Jan Typhoon Rammasun
Dec 2013–Feb 2014 Serbia, Croatia, Romania, China, Philippines, Vietnam,
13–30 May 11–22 Jul
Severe storms
USA, 18–23 May

Severe storms
France, Belgium, Winter damage
Germany, Japan, 7–16 Feb
Drought
7–10 Jun
USA, 2014
Typhoon Kalmaegi
Flash floods China, Philippines, Vietnam,
USA,11–13 Aug 12–20 Sep
Hurricane Odile
Mexico, 11–17 Sep

Severe storms
USA, 2–4 Apr Cyclone Hudhud
India,
Severe storms Drought 11–13 Oct
980 USA, 27 Apr–1 May Brazil, 2014
Loss events
Severe storms Floods
USA, 3–5 Jun India, Pakistan,
Earthquake
3–15 Sep
China, 3 Aug
Source: Munich Re, NatCatSERVICE, 2015

Loss events Geophysical events Hydrological events

(Earthquake, tsunami, volcanic activity) (Flood, mass movement)

Selection of catastrophes Meteorological events Climatological events

Overall losses ≥ US$ 1,500m (Tropical storm, extratropical storm, (Extreme temperature, drought, wildfire)
convective storm, local storm)

© 2015 Münchener Rückversicherungs-Gesellschaft, Geo Risks Research, NatCatSERVICE – As at January 2015

Fig. 1. Geographical overview of loss relevant events worldwide in 2014.

by geophysical events (51%) followed by storms (25%). The group attributed to the fact of a better reporting of smaller loss events
of perils with the largest contribution to the total economic losses today than in the past. Such a reporting bias dominated by the
has been the storms (40%), the second largest group (25%) being small events can be excluded when looking at the losses. The loss
the flood events. In the insured losses both the number of events figures are dominated by the larger loss events, for which re-
and the different insurance uptake for the different perils in the porting since 1980 certainly has not changed significantly.
richer countries play an important role. This leads to the fact that Fig. 4 shows the global annual direct economic and insured
storms have contributed 71% of all insured losses with a large gap losses of all weather related catastrophe events since 1980. All
to the earthquakes (11%) being the second most costly group and values have been adjusted for the country specific inflation based
the floods (10%) in the third rank. on the consumer price index. The figures show highly significant
The distribution of the losses clearly shows that they are (p o0.001) increases of economic and insured losses in the last 35
dominated by weather related events, 78% of the economic losses years with the peak in 2005 driven by the extremely active hur-
and even 89% of the insured losses. These results show clearly the ricane season, especially by the losses caused by hurricane Katrina
high potential of changes in the atmosphere, e.g. global warming, (original overall loss US$ 125bn, insured loss US$ 62bn).
to affect the economic and insured losses. This is the reason why As the losses shown in Fig. 4 are not normalized to the country
both the analyses of effects of natural climate cycles and global specific increases in wealth and thus to the total increases in ex-
warming on loss potentials have been in the focus of in house posed values the trend certainly is also driven by such socio-eco-
research of large insurers and reinsurers for a long time. nomic factors. As a complete normalization of losses on a global
level is very complex (due to lack of respective data and asym-
3.2. Global trends metric economic development in many countries especially in
China) there are hardly any studies on this. One of the few studies
Since the 1980s the frequency of loss relevant natural disasters showing significant trends of normalized insured losses is the
worldwide has increased highly significantly (po 0.001), has study by Barthel and Neumayer (2012). They found significant
about tripled since then (see Fig. 3). While in the early 1980s the normalized increases for all weather related disasters as well as for
annual number has been in the range of 300 this number has in- certain specific disaster types in the United States and West Ger-
creased to almost 900 in the last years. In the period of 1980–1989 many, two countries with high quality data. They normalised the
an annual average number of 320 events has been registered, it losses in respect to GDP growth and insurance penetration. In-
rose to 510 events in the 1990s, 660 events in the 2000s and to 830 terestingly they did not find such trends for the geophysical
in the last 5 years (2010–2014). Some part of this increase may be (nonclimate related) losses. They, as several other authors, could
74 P. Hoeppe / Weather and Climate Extremes 11 (2016) 70–79

NatCatSERVICE

Loss events worldwide 1980 – 2014

Percentage distribution

21,700 Loss events 1,740,000 Fatalities*

12% 51%
41% 25%
36% 13%
11% 10%

*Number of fatalities
without famine

Overall losses** US$ 4,200bn Insured losses** US$ 1,100bn

22% 11%
40% 71%
25% 10%
13% 8%
**Losses in 2014 values, **Losses in 2014 values,
adjusted to inflation adjusted to inflation
based on country CPI based on country CPI

Geophysical events Meteorological events Hydrological events Climatological events

(Earthquake, tsunami, (Tropical storm, extratropical (Flood, mass (Extreme temperature,
volcanic activity) storm, convective storm, movement) drought, forest fire)
local storm)
© 2015 Münchener Rückversicherungs-Gesellschaft, Geo Risks Research, NatCatSERVICE – As at January 2015

Fig. 2. Average global distributions of the different groups of natural perils to the number of events, number of fatalities, economic and insured losses in the time 1980–2014.

NatCatSERVICE

Weather related loss events worldwide 1980 – 2014

Number of events

Number

1000

800

600

400

200

1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014

Meteorological events Hydrological events Climatological events

(Tropical storm, extratropical (Flood, mass (Extreme temperature,
storm, convective storm, movement) drought, forest fire)
local storm)
© 2015 Münchener Rückversicherungs-Gesellschaft, Geo Risks Research, NatCatSERVICE – As at January 2015

Fig. 3. Annual numbers of loss relevant natural events in the time 1980–2014 for the different kinds of perils.
 P. Hoeppe / Weather and Climate Extremes 11 (2016) 70–79 75

NatCatSERVICE

Weather-related loss events worldwide 1980 – 2014

Overall and insured losses

US$ bn

250

200

150

100

50

1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014

Overall losses (in 2014 values)* Insured losses (in 2014 values)* *Losses adjusted
to inflation based
on country CPI

© 2015 Münchener Rückversicherungs-Gesellschaft, Geo Risks Research, NatCatSERVICE – As at January 2015

Fig. 4. Annual aggregated global direct economic (green bars) and insured (blue bars) losses caused by weather related natural disasters. All values have been adjusted for
inflation and show the losses in 2014 values.

not find significant trends on a global level, though. In the last In Fig. 6 the annual number of convective loss events is shown
years there has been a controversial discussion in the literature for the U.S. While in the 1980s the number of events has been
on trends of normalized losses. Especially for developing and in the range of about 25 per year this has increased to a level
emerging countries the lack of the availability of high quality data of around 100 in recent years. The trend is highly significant
may be a major reason for this. (p o0.001).
The relative trend of the different groups of perils (Fig. 5) shows The inflation adjusted annually aggregated losses have in-
a distinct difference between all weather related perils and the creased significantly (p o0.001) and have peaked at a level of al-
group of geophysical perils. The number of loss relevant flood most US$ 50 bn in 2011 (see Fig. 7). The average loss level in the
events shows the most pronounced trend upwards (factor of larger last 5 years has been above US$ 20bn while in the first 5 years of
than 4) followed by the trends of storms and other weather related the 1980s it has been in the range of US$ 5bn. The figure also
events like heat waves, droughts and wild fires (both with a factor shows the high insurance penetration for all convective loss perils.
of about 3). All of these trends are highly significant (p o0.001). The ratio of the insured vs the overall losses in the last years has
The geophysical events only show a small but still significant been far above 50% in the last two years the insurance industry
(p ¼0,002) trend with an increase based on the slope of the re- had to pay even about 2/3 of the overall losses. This points to the
gression line of less than 50%. This slight increase most probably is high relevance of the convective events and their trends for the
driven by socio-demographic drivers like population growth and insurance industry.
more settlements in risk prone areas. Thus today a small earth- In a special study on those convective events in the U.S. con-
quake is classified as loss relevant, while some decades ago, be- ducted by a Ph.D. student and colleagues at Munich Re during the
cause of less people and/or lower values affected, such an earth- last years the losses have been also normalized to the growth of
quake has not been loss relevant. Such a socio-economic driver has wealth. This study has been published in the journal “Weather,
also to be expected in the weather related loss events but they still Climate and Society” (Sander et al., 2013). This study examines
show strong increases after the slope for the geophysical events convective events in the US with normalized losses exceeding US$
has been deducted and thus cannot explain the whole increase. 250 m in the period 1970–2009 (80% of all losses). The threshold
for the losses has been used to avoid a reporting effect, which
3.3. Convective events would have been relevant when considering smaller events. Past
losses have been normalized to currently exposed values. After
The weather related peril which has caused the largest in- normalization there are still increases of losses. In this study on
creases both in numbers of loss relevant events and also the losses losses for the first time also the meteorological conditions have
are convective events. All events with a loss potential developing been considered by using reanalysis data. The study found that the
out of thunderstorms, like hail, intense precipitation, tornadoes increases and the variability of the losses are correlated with
and strong straight line wind, are called convective events The changes in the meteorological potential for severe thunderstorms.
strongly increasing trends can be seen both in the U.S. as well as in The main driver of these changes have been changes in the hu-
Europe. midity of the troposphere. There is general evidence that the
76 P. Hoeppe / Weather and Climate Extremes 11 (2016) 70–79

NatCatSERVICE

Loss events worldwide 1980 – 2014

Number of events with relative trend

Number Geophysical events

(Earthquake, tsunami,
500%
volcanic eruption)
450%
Meteorological events
400% (Tropical storm,
extratropical storm,
350% convective storm,
local storm)
300%
Hydrological events
250% (Flood,
mass movement)
200%
Climatological events
150% (Extreme temperature,
drought, forest fire)
100%

50%

0%
1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014

© 2015 Münchener Rückversicherungs-Gesellschaft, Geo Risks Research – As at March 2015

Fig. 5. Relative (reference year 1980) trends of numbers of loss relevant natural events for the four different groups of perils. Lines represent the linear regressions of the
individual curves. Trends are significant for all peril families.

NatCatSERVICE

Convective loss events in the U.S. 1980 – 2014

Number of events

Number

Loss events caused by straight-line winds, tornadoes, hail, heavy precipitation, flash floods, lightning

© 2015 Münchener Rückversicherungs-Gesellschaft, Geo Risks Research, NatCatSERVICE – As at July 2015

Fig. 6. Annual numbers of loss relevant convective events in the U.S. in the time 1980–2014.
 P. Hoeppe / Weather and Climate Extremes 11 (2016) 70–79 77

NatCatSERVICE

Convective loss events in the U.S. 1980 – 2014

Overall and insured losses

bn US$

Loss events caused by straight-line winds, tornadoes, hail, heavy precipitation, flash floods, lightning
Overall losses (in 2014 values)* Insured losses (in 2014 values)* *Losses adjusted
to inflation based
on country CPI
© 2015 Münchener Rückversicherungs-Gesellschaft, Geo Risks Research, NatCatSERVICE – As at July 2015

Fig. 7. Annual aggregated direct economic (green bars) and insured (blue bars) losses caused by convective events in the U.S. All values have been adjusted for inflation and
show the losses in 2014 values.

humidity of the air has increased significantly in the last decades, storms in the summer season (predominantly convective storms)
driven by the warming oceans and the increased evaporation from will increase by 25% on average and for the time between 2041
their surfaces. and 2070 by 61% (see Fig. 10).
In Europe a similar pattern of trends of convective loss events is The study also analysed flood events. The results showed that
visible, though at a lower absolute level. Here the largest annual e.g. flood events causing losses of € 750 m, which currently are
aggregated losses have reached about € 7.5bn. In terms of the in- expected every 50 years will turn to become about 20 year events
creases of the number of events the trends are comparable and within the next 30 years.
highly significant (po0.001). Also in Europe they have risen around The GDV study focussed on Germany, but as weather extremes
a factor of 4 from about 30 events per year to about 120 (see Fig. 8). are not confined to political boarders the assumption that similar
The largest convective losses have occurred in 2013 when a effects will occur at least in the neighbouring countries, if not in
large hail storm hit Germany. This hail event consisting of two hail whole Europe certainly is valid.
storm lines, one in Lower Saxony and North-Rhine-Westphalia, the As the results of the GDV study clearly show the causal relation
other in Baden-Wuerttemberg, caused overall losses of € 3.6bn between loss trends and climate change and with the assumption
and insured losses of € 2.8bn. This makes it to the ever most ex- that such changes will be driven by a rather continuous process
pensive global loss event caused by hail. The trends of the eco- not starting suddenly in the next years, there is a suggestion that
nomic and insured losses are highly significant (p o0.001) Fig. 9). part of the already detected loss trends in Germany but also most
probably in many parts of Europe may be attributed to climate
change already.
4. Perspectives of future developments In another study based on climate models Diffenbaugh et al.
(2013) found a robust signal for further increases in the occurrence
According to the 5th assessment report of IPCC (IPCC, 2014) of severe thunderstorms in the U.S. driven by global warming.
global warming will result in many regions in increases of the
frequency and intensity of weather extremes. The highest con-
fidences for this is for heat waves and intense precipitation events. 5. Conclusion
In an unique study commissioned by the German Association of
Insurers (GDV, 2011) the future insured losses caused by storms The Munich Re NatCatSERVICE data clearly show that the
and floods have been modelled on the basis of the current prop- number of loss relevant weather extremes has increased sig-
erty insurance portfolio in Germany with a wide range of global nificantly. There is increasing evidence that at least part of these
and regional climate models for Germany. The results are very increases are driven by global warming. The increases in losses are
robust, i.e. all model runs have shown future increases in insured driven predominantly by higher exposed values due to increasing
losses caused by climate change. wealth and population in many regions. The task to quantify the
So e.g. already within the next 30 years the losses caused by significantly smaller signal of climate change is very difficult as
78 P. Hoeppe / Weather and Climate Extremes 11 (2016) 70–79

NatCatSERVICE

Convective storms in Europe 1980 – 2014

Number of events

Number

160

140

120

100

80

60

40

20

1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014

Flash Flood Tornado Hailstorm Severe storm

© 2015 Münchener Rückversicherungs-Gesellschaft, Geo Risks Research, NatCatSERVICE – As at April 2015

Fig. 8. Annual numbers of loss relevant convective events in Europe in the time 1980–2014.

NatCatSERVICE

Convective storms in Europe 1980 – 2014

Overall and insured losses

EUR bn

1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014

Loss events caused by straight-line winds, tornadoes, hail, heavy precipitation, flash floods, lightning
Overall losses (in 2014 values)* Insured losses (in 2014 values)* *Losses adjusted
to inflation based
on country CPI
© 2015 Münchener Rückversicherungs-Gesellschaft, Geo Risks Research, NatCatSERVICE – As at April 2015

Fig. 9. Annual aggregated global direct economic (green bars) and insured (blue bars) losses caused by convective events in Europe. All values have been adjusted for
inflation and show the losses in 2014 values.
 P. Hoeppe / Weather and Climate Extremes 11 (2016) 70–79 79

Fig. 10. Forecast of changes of climate change driven insured losses caused by storms in summertime in Germany in the next 30 and next 60 years.

some more confounding parameters have to be considered for caused by weather related events will make them uninsurable in
which data availability if confined. One important aspect in this the next decades as long as so called non linear “tipping points” of
sense are loss prevention measures taken in many countries. Thus climate change are not reached. The GDV study also showed at the
an increasing hazard does not necessarily always mean increasing concrete example of Germany that the premiums, though rising
losses. Some of the hazard driven increases of loss events may for some perils according to the change in hazard, will stay in an
even have been overcompensated by such loss prevention mea- affordable range. This could be different for countries located in
sures. Such a confounder most probably plays an important role in more extreme climatic zones, being exposed also to tropical cy-
the flood loss data.
clones or droughts. The higher volatility of weather extremes in
Flood losses can be influenced much more than e.g. wind storm
many regions as expected by IPCC, will increase the need for more
losses by preventive measures. So e.g. after the large flood in
risk capital of the insurers and a better – regional and peril re-
Hamburg in 1962 with economic losses in the range of € 1.6 bn (in
2010 Euros) and 347 fatalities more than € 2 bn have been in- lated-diversification of their risks.
vested in flood protection measures. Since then nine times the
level of the Elbe river has reached even higher levels without
causing significant losses anymore. So in this case the loss trend References
even points downwards while the weather driven hazard has
increased.
Bartel, F., Neumayer, E., 2012. Trend analysis of normalized insured damage from
We have to expect that the already visible increases in weather natural disasters. Clim. Change 113 (2), 215–237.
related loss events will continue in the future as climate change Diffenbaugh, N.S., Scherer, M., Trapp, R.J., 2013. Robust increases in severe thun-
will proceed. Societies will have to invest ever more money in derstorm environments in response to greenhouse. PNAS, 16361–16366.
GDV (German Association of Insurers), 2011. The Climate Change Challenge, GDV,
prevention measures and reconstruction after an extreme event. Berlin. 〈http://www.en.gdv.de/wp-content/uploads/2014/01/study_Climate_
Many studies have shown that pre disaster investments (loss Change_Challenge_Answers_Demands_German-Insurers_2011.pdf〉.
prevention) pay back manifold in the long term as they decrease IPCC, 2014. Climate Change 2014: Impacts, Adaptation, and Vulnerability. Con-
tribution of Working Group II to the Fifth Assessment Report of the Inter-
losses and thus the costs after a disaster more than they cost.
governmental Panel on Climate Change. In: Field, C.B., Barros, V.R., Dokken, D.J.,
After having applied prevention measures there still will be Mach, K.J., Mastrandrea, M.D., Bilir, T.E., Chatterjee, M., Ebi, K.L., Estrada, Y.O.,
risks which cannot be prevented in an economical sensible way. Genova, R.C., Girma, B., Kissel, E.S., Levy, A.N., MacCracken, S., Mastrandrea, P.R.,
For these risks the insurance industry can provide cover and make White, L.L. (Eds.), Cambridge University Press, Cambridge, United Kingdom and
New York, NY, USA, p. 1132.
quick money available for reconstruction after a large event. This
Sander, J., Eichner, J., Faust, E., Steuer, M., 2013. Rising variability in thunderstorm-
risk transfer mechanism increases the resilience of societies effi- related U.S. losses as a reflection of changes in large-scale thunderstorm for-
ciently. There is no indication that the expected changes of losses cing. Weather, clim. Soc. 5, 317–331.