Alineanummer51 (a) Copyright 1994, A. A. Olsthoorn and R. S. J. Tol. All rights reserved. For more information contact author. Socio-economic and policy aspects of changes in the incidence and intensity of extreme (weather) events A.A. Olsthoorn and R.S.J. Tol July 8 1994 PREFACE This report is the report of the workshop on the "Socio- economic and policy aspects of changes in the incidence and intensity of extreme (weather) events", held at the Institute for Environmental Studies, Vrije Universiteit, Amsterdam on 24 and 25 June, 1993. The workshop was jointly organised by the Environmental Change Unit, University of Oxford, W.J. Maunder Consulting Services Ltd, Tauranga, Ernst Lohman Consultants, Enschede, and the IES. The workshop was sponsored by the Dutch National Research Programme: Global Air Pollution and Climate Change. The authors hereby express their gratitude to all the participants of the workshop (cf. Annex 1) for cooperation and stimulating discussion, and especially those who contributed an annex to this report. In preparing the reflections of the talks, the texts and sheets of the speakers and the notes of mrs. U. van der Hul (only the second day), dr W.J. Maunder. dr A.A. Olsthoorn, drs R.S.J. Tol and prof dr ir P. Vellinga were used. We also thank dr W.J. Maunder and prof dr ir P. Vellinga for their comments to earlier versions of this report. 1. INTRODUCTION Both disaster statistics and recent press releases from the insurance industry show alarming increases in the cost and, in some cases, the occurrence of disastrous (extreme) weather events. It is obvious that these phenomena, having large economic, social and political impacts, will constitute an increasingly important subject in the current international debate on climate change. Consequently a number of questions need to be answered, including: - do these observed changes in severe weather events, particularly from a cost viewpoint, herald a sustained increase of climatic hazards? - what is the relationship between these changes and the increasing amounts of greenhouse gases in the atmosphere? - what might the future hold with respect to hazardous weather anomalies? - what are likely to be the direct impacts of an increase in weather-related natural disasters? - which strategies can be developed to cope with these possible effects of climate change? - how can institutions, most affected by weather disasters, contribute to greenhouse-gas policy responses? These and related questions constituted the framework of the IVM-ECU (Institute for Environmental Studies, Vrije Universiteit Amsterdam - Environmental Change Unit, University of Oxford) research project, of which the workshop 24-25 June 1993 was the beginning. The participants had different backgrounds: climatology, climate modelling, the insurance industry, development planning, disaster relief, environmental impact analysis and policy analysis. The aims of the workshop were to discuss the current knowledge and the relevance of the issues noted above and to identify specific research questions which are both relevant and manageable. IVM, ECU and other institutes have prepared research propo- sals based on the results of the workshop. The main interests of IVM and ECU are in impact and policy analysis; "weather generators", "insurance" and "natural disasters" are focal points. This research is undertaken in order to provide information relevant for the international debate on climate change policy. Results will be put forward within various research programmes, among them: - the World Climate Programme (WMO: World Meteorological Organisation, UNEP: United Nations Environmental Programme, ICSU: International Council of Scientific Unions, IOC: International Oceanographic Committee); - relevant working groups of the IPCC (Intergovernmental Panel on Climate Change); - the programme of the IDNDR (International Decade of Natural Disaster Reduction) of UNDRO-DHA (United Nations Disaster Reduction Organisation - Department of Human Affairs). The workshop was organised as a two-day event. The basic issues of the first day were: - what evidence exists, based on the historical climatological record, which indicates changes in the frequency, intensity, and geographical location of severe weather events? - do the results of current climate modelling provide evidence for changing patterns of (hazardous) weather (wind and rainfall)? - how can weather scenarios be derived from climate model results which can then be used in impact studies? The second day dealt with a variety of issues related to scenarios for the impacts of extreme weather events: - the impacts on the insurance industry - insurance as a mechanism for risk management of highly vulnerable small countries - international efforts to reduce the effects of disasters (IDNDR). - flood risk management (attitudes, preferences of individuals and the realities of disaster management schemes) This paper is the report of the workshop, its presentations, its discussions and its findings. Section 2 deals with the presentations of the first day; section 3 is on the discussion of the first day. The section 4 and 5 deal with the presentations and the discussion of the second day, respectively. Please note that the presentations are very briefly summarised. Section 6 summarises and concludes by presenting the present research agenda. The annexes (to be added) contain abstracts some of the presentations and the addresses of the participants. 2. THE PRESENTATIONS OF THE FIRST DAY Pier Vellinga. The opening words were addressed by professor dr.ir. P. Vellinga, director of the Institute for Environmental Studies, the host of the meeting. The goal of the first day of the meeting was to provide the IVM and the ECU with a sound scientific and statistical basis for developing scenarios for possible futures of extreme weather events for the period up to 2030-2040. The issues so far identified as of possible interest were: - tropical cyclones; - coastal zones and storm surges; - agriculture and crop failure; - land subsidence and drought; - river transport and hydrology; - buildings and wind storms; - human health and heat stress; - economic development and flood hazard. Based on the state-of-the-art knowledge of the scientists, and the interests of the insurance industry, topics were to be selected and a choice needed to be made of both the areas to study and the methods to use. The second day would be taken up by the `customers', i.e. the insurers and the planners, and their views and needs. John Maunder. The next to speak was dr. W.J. Maunder, president of the WMO Commission for Climatology, the day's chairman. He posed the following questions. (i) What is an extreme weather event? Is it something above a certain threshold or is it a continuous process? Do extremes differ between certain locations? (ii) Has there been any change over the last 50 years in extreme events, in frequency, intensity or locations? (iii) If this is the case, why? (iv) What will happen in the future? Will it be different from the past or will it be the same? Do we know enough to answer this question? (v) What are the best methods to derive scenarios? Which combination of GCM outcomes, expert knowledge and statistics suits best? Who, including those not attending, can assist the ECU and the IVM in this? (vi) Which geographical areas and which extreme weather events should be studied? and (vii) What do the changes in extreme events mean to people, institutions, national governments, the international community and the international organisations? He further stressed the need to take not only the presumed climate change due to anthropogenic greenhouse gases into account, but also the natural climate variability, and to pay careful attention to the interactions between the physical and the socio-economic and political dimensions of extreme weather events. His final point was that if we consider climate change without extreme events, we might find that we have time to refine the scientific knowledge. However, weather extremes must be considered and, science has to move quickly to provide some answers because the decision makers, particularly in the insurance industry, are acting now! Peter Rowntree. Dr. P.R. Rowntree of the Hadley Centre of the U.K. Meteorological Office presented the results applicable to extreme weather events resulting from an analysis of the equilibrium and transient response experiments performed with the UKMO GCM0. According to these experiments, the gradient of the mean temperature changes1, mainly because land masses warm faster than the oceans2, and because of the more rapid rise in the temperatures at higher latitudes. Therefore, the circulation patterns change significantly which may lead to more northwesterlies over Europe, and an increase up to about one-third in the number of gales, es- pecially in the U.K.. Because of the presumed higher wind- speeds of storms, the tracks of such storms will be longer3, causing damage over a larger area. These conclusions are sensitive to the (weakly validated) details of the model; however, he indicated that the present meteorological observations confirm these details. The expected increase in humidity also contributes to increasing storminess. With respect to rainfall an increase Ñ in north west Europe Ñ of 20% to 40% in winter was derived from the model; however, an important feature of this increase is that the number of rainy days is likely to decrease, resulting in an increase of the maximum daily precipitation of up to 60%. In summer, Central European rainfall will probably equal the present amount or may be a little lower; Southern European summer rainfall might decrease by 20%. The UKMO models further predict a general soil moisture decrease in the summer, leading to longer dry spells (the maximum dry spell duration might increase by 50%) and longer spells of reduced transpiration. The above-mentioned wind circulation changes also have their repercussions on the temperature extremes. Minimum summer temperatures shift with the mean but maximum temperatures shift to a much larger extent (8ûC compared to 2ûC); in winter this is reversed: minimum temperatures rise faster than maximum temperatures. The frequency of tropical cyclones is likely to go up but the area affected stays about the same; the intensity increases slightly. The final remark of dr. Rowntree was on the sea level rise. The larger part of the rise will be due to the thermal expansion of the oceans. It should be noted that the ocean approaches the new equilibrium much slower than the atmosphere does, implying that, even if climate change stops, the sea level rise will continue to rise for centuries. The predicted rise of about 20 cm in the first 75 years is only the first part of the total rise. Francis Zwiers. Dr. F.W. Zwiers of the Canadian Climate Centre presented the temperature extremes derived from the statistical analysis of data from the runs with the CCC's second generation GCM. In the CCC model, 8 years are used for `the spin up' and then 20 further years are used to generate the scenarios with and without a doubling of CO2. After discussing more of the properties of the model and the similarities and differences between observations and control runs, dr. Zwiers elaborated on the threshold crossing frequency, the threshold crossing duration, and the number of above/below threshold days. The first threshold discussed was the 33ûC event. The crossing frequencies show a variety of changes; however, the crossing durations4 and the above threshold events show an increase5. For instance, the north east of the United States 25 more days of temperatures greater than 33ûC are likely under doubled CO2. For temperatures above 33ûC the `threshold crossing duration' is likely to be 2-4 days longer under 2xCO2. The second threshold discussed was the -1ûC event6. In this case, the threshold crossing frequency and duration showed mixed results, but in Eurasia and North America the duration of below -1ûC events decreases, and the number of below threshold days reveal an overall decrease. Next the changes in the 10 and 50 year return period maximum and minimum temperatures were presented. For the 10 year return period, the maximum temperatures showed a smaller increase (av. 3.14ûC) than the minimum temperatures (av. 5.0û). This might be due to the winter's temperatures rising faster than the summer's, the changes in soil moisture and the decrease in snow cover insolation. The 50 year return period maximum and minimum temperatures showed the same patterns but are noisier. The overall conclusions were (i) that a variety of inferences about the impact of climate change on the frequency, duration and size of extreme temperature events is possible, (ii) that changes in the location, scale and shape of the screen air temperature distributions are observed in the model and (iii) that inference about changes in extremes might be made relatively reliable with `short' climate simulations. Jerry Meehl. Dr. G.A. Meehl of the U.S. National Center for Atmospheric Research, based on the GCM of NCAR, addressed the question how the El Ni–o/Southern Oscillation phenomenon would be affected by a change in the climate. This is an important question because ENSO is the single largest contributor to climate variability over much of the Earth and is associated with extreme weather events in some areas of the tropics and the extratropics. Therefore, changes in ENSO effects could be of greater importance in some regions than changes of the mean climate. The NCAR model can only be used to identify physically plausible mechanisms, not to generate forecasts; this is due to systematic errors in the model. The conclusions are that, with increased atmospheric CO2, ENSO events in the tropical Pacific continue as at present, but that the associated effects are more intense, associated with higher mean sea surface temperature and strengthened anomalous Walker (eastwest) circulation. In addition, it was concluded that extratropical teleconnections are altered (they become more zonal), associated with major changes of mid-latitude atmospheric circulation. The NCAR model indicates that with a doubling of CO2, soil moisture is likely to decease by 7% in the present drought areas of southern Africa, 20% in the present drought areas of Australia, and 18% in the present drought areas of north east Brazil.The second topic discussed by the speaker was changes in the South-Asian monsoon. Because of global warming, the land-sea contrast may be increased, which will strengthen the mean monsoon: the average precipitation per day is likely to increase from 6.4 mm per day (1xCO2) to 6.8 mm per day (2xCO2). However, internal feedbacks (enhanced evaporation, increased soil moisture) increase interannual monsoon variability. Therefore, the model indicates that the number of floods seems to increase while the number of droughts remains the same or slightly increases as well. The current observations reveal the predicted increased variability but not the increased mean monsoon (the observation in fact indicate a decrease). This latter feature is due to the reduced land-sea temperature contrast (opposite to the model outcomes) which might be attributed to the increased pollution of the air over India with sulphate aerosols. GŸnther Kšnnen. Dr. G.P. Kšnnen of the Royal Dutch Meteorological Institute presented his views on climate prediction and climate predictability. Although there is no doubt that increasing atmospheric carbon dioxide enhances the greenhouse effect, the magnitude of the effect of a doubling of atmospheric carbon dioxide is difficult to estimate because it is both hard to measure and to predict among the effects of other greenhouse gases, such as water. There is a `jungle of feedbacks' both in climate and in climate models, and `everything depends on everything'. In particular, the influence of water (i.e. solid, liquid and gaseous) in the climate system is difficult to assess under changed circumstances. Therefore, the question arises as to which climate parameter is the most suited for climate change detection. Equally important is the question as to which parameter is the most suited for impact assessments. Differences between weather and climate prediction are that (i) the former requires prediction of the timing of specific weather events whereas the latter requires forecasting of the evolution of the climate characteristics, such as the means and standard deviations, (ii) the numerical models used for weather prediction only include the fast atmospheric processes whereas the models used for climate predictions also involve the slow processes, such as the ocean-atmosphere interactions and (iii) the predictability limit for weather is about 14 days whereas the climate limit is unknown, maybe infinite, depending on the requirements one imposes.Dr. Kšnnen next focused on the chaotic character of the climate and especially on the possibility of rapid modal shifts, the so-called flip-flop behaviour. With a simple two-variable model as an example, he argued that there may be more (than one) climate modes, i.e. states of equilibrium. Questions for present climatology are which states are possible, what is the probability of transition, and how to model this, if indeed this is possible?7 When dealing with climate change, questions arise such as how will equilibrium states and transition probabilities shift, and will new modes arise? The method suggested to derive a scenario for chaotically changed climate was to flip-flop one feature and to derive the others based on their present-day statistical rela- tionships. Also, we need to more completely understand the present climate. Nigel Arnell. Dr. N. Arnell of the U.K. Institute for Hydrology spoke on changes in extreme hydrological events. First he emphasised that due to the large interannual variability it is hard to separate trends from normal fluctuations. Second, it is important to separate the influence of climate variability and changes due to (other) human interference. Third, extreme hydrological events show a strong auto-correlation, i.e. they depend on their past, and have a tendency to cluster in successive years. Fourth, extreme hydrological events show world-wide spatial cor- relation, i.e. they reveal synchronicity, and have a tendency to occur everywhere at the same time. This reflects the teleconnections in the atmosphere of which the ENSO is the most important example. These last two features are especially important to the insurance industry. Changes in the hydrological extremes can be analysed by using the output of climate models as the input to the hydrological models. However, as the climate model outcomes with respect to extremes and especially precipitation extremes are highly unreliable, this approach is not very promising. A second option is to design a scenario for extreme climate inputs and use this as an input to the hydrological model. A third option, dr. Arnell mentioned and demonstrated, is to perturb the parameters of the probability distributions describing the present hydrological extremes. Tom Downing, Paula Harrison and George Blumberg. Dr. T.E. Downing of the Environmental Change Unit of the University of Oxford asked the participants for their comments and advice on the methodology ECU were using in a pilot study on the first order impacts of changes in extreme events. Using these comments, the ECU will continue their study of which the first results will be published in the report of the Oxford Roundtable on Climate Change and Extreme Events on 11 and 12 November, 1993. P.A. Harrison presented the future risk of subsidence in Europe using five scenarios. This risk was defined by a shortfall of accumulated December-August rainfall below 500 mm.8 The first two scenarios demonstrated the use of a sensitivity analysis. The present yearly average rainfall was perturbed by -20% and +20%. In the first scenario the area at risk largely expands, in the second it largely shrinks. The last three scenarios used GCM output based on equilibrium 2xCO2 experiments. The GCMs differ to a large extent with respect to the future area at risk but all show a some improvement of the baseline situation. G.M.C. Blumberg presented a statistical approach to the analysis of extreme high temperature events in a changing climate. To this end, the three extreme value distributions (Gumbel, Frechet and Weibull) were fitted to the observed daily maximum temperatures for 200 sites in Europe, and compared. The sensitivity of the probability on the temperature crossing a certain threshold to perturbations in the mean and scale parameters of the Gumbel distribution was analysed. Richard Tol. R.S.J. Tol of the Institute for Environmental Studies of the Vrije Universiteit of Amsterdam was the final speaker of the first day. His first remark was that, based on a series of statistical models of the global mean temperature (well-validated: the global mean temperature could be forecast over a period of 50 years), he concluded with a 95% statistical confidence that the atmospheric concentration of carbon dioxide Ñ or a feature very much alike Ñ has influenced the climate during the last 100 years9.The actual topic of this talk was how to improve stochastic weather generators. Although SWGs, used to obtain time series of future weather, do a reasonable job in representing the mean weather, he said, their performance with respect to extreme weather is poor. This may be due to changes in the stability of the weather10, inducing changes in the variability and predictability. This feature was shown for the De Bilt daily mean temperatures. Similar patterns are observed in economic time series. The models developed there, called GARCH, were used for the analysis of the daily temperature variability and shown to perform reasonably. Therefore, future SWGs should include a GARCH-like variance structure. 3. THE DISCUSSION OF THE FIRST DAY After a short break, the discussion was initiated by dr. T.E. Downing to answer the question how to develop scenarios for the above mentioned phenomena? Is this possible? If so, how good are they? What data are needed? What time-scales should be used and which areas should be studied? What are the relevant climatic parameters? What are the plausible scenarios? Several extreme weather events were considered as being worthy of study in the initial phase of the project, as follows:¡With respect to the risk of subsidence, it was concluded that, although U.K. precipitation and subsidence show a good correlation, more sophisticated climate parameters, such as the potential evaporation, would do better. The water-balance and the type of soil need to be incorporated in the analyses. The scenario should be based on the GCM of which the control run resembles the present climate best11. It was suggested that one parameter should be picked from the GCM output, preferably temperature, while the other parameters should be derived from the present-day statistical relationships between the parameters of interest and the steering parameter. It was further decided to study the U.K. on a monthly base and to perform perhaps an addi- tional case, for instance in France. More sophisticated studies can only be performed at a much smaller scale whereas the project aims at a broad overview of possible risks. Part of the data are already available to the ECU, more data could be made available from the (re)insurance industry and perhaps from the Hadley Centre. ¡With respect to the transport problems due to low river run- off it was concluded that the Rhine would be the most suitable object of study. The hydrological modelling as well as the first order impact assessments are (being) done, the project could contribute by assessing the socio-economic impacts and possible policy responses. The IVM will contact the relevant institutes. ¡Flood hazards can best be studied using perturbations of probability distributions for a small number of well- documented sites, for instance the track near Rheinland Pfalz near Ludwigshafen (Germany). Cooperation with other institutes, projects or persons, having done hydrological modelling is another option, for instance Delft Hydraulics, the U.K. National River Authority/Commission of Hydrological Modelling, EuroFlood (in which inter alia Middle-sex Polytechnic University is involved; it is funded by the EC- EPOCH programme) or Russell and Smith. If the second option is chosen, the project will again focus on the second order impacts and the policy responses. The IVM will contact the relevant institutes. ¡Storms are to be studied over the whole of Europe; small, localised damage has too little impact on the insurance industry. For the same reason, not only the highly intensive core of the storms is of interest but the whole of the storm track. Based on expert knowledge (partly based on the GCM outcomes from UKMO, NCAR and MPI) and sensitivity analyses scenarios for changed frequencies, intensities and widths of the stormtracks will be devised. An important problem in this is that the knowledge of the present situation is still very limited. This will also be tackled. The (re)insurance industry will provide the data on storm vulner- ability. ¡Heat stress and human health will be studied in some Mediterranean cities. The CCC and USEPA have already addressed some aspects of this problem for the cities in the North-East U.S.A., and the USEPA, in association with prof. L. Kalkstein (University of Delaware), and John Maunder will continue to as further study this area during April/May 1994. ¡Tropical cyclones will be studied in the South-West Pacific, partly because of the political sensitivity of this region. John Maunder will sample the data on the present tracks, frequency and intensity, E.J.A. Lohman will assist in assessing the first and second order impacts. 4. THE PRESENTATIONS OF THE SECOND DAY Lennart Bengtsson. Prof. dr. L. Bengtsson of the Max Planck Institut fŸr Meteorologie (MPI) in Hamburg presented the first talk of the second day Ñ tropical cyclones according to the MPI GCM Ñ which was an extension of the first day. The textbook determinant of tropical cyclones is that the sea surface temperature exceed 27ûC. However, William Gray identified a number of empirical hurricane genesis parameters, i.e. the low-level vorticity, the inverse of the vertical shear of the horizontal wind between the lower and the upper troposphere, the ocean thermal energy, and the middle troposphere relative humidity. Gray's theory does a good job in predicting hurricane activity but his results cannot be reproduced by the MPI.Bengtsson explained that a simple but clear way to view hurricane genesis is by looking at it as a Carnot heat engine (Emmanuel). According to this model the driving `force' is the difference between the tem- peratures at the sea-surface and the tropopause. Because this difference decreases according to most climate models, cyc- lones should intensify, but the incidence would fall. This might, however, be a model artefact, since the observations of the temperature differences are not clear. The MPI GCM can as yet not simulate tropical cyclones, but from the out- comes it is possible to identify hurricane type vortices based on some criteria for the relative vorticity, wind- speed, temperature anomalies and the duration. While the MPI GCM generates less tropical cyclones in a 2xCO2 climate, the UKMO model finds more cyclones. In this regard, the modelling of the behaviour of clouds is of crucial importance, but this is still not sophisticatedly tackled. Another weakness is the neglect of other greenhouse gases beside carbon dioxide, since these change the vertical temperature distribution. The GCM's indicate that the geographical areas in which tropical cyclones are found roughly stay the same12.Finally prof. Bengtsson presented a video of the genesis and life or a hurricane type of vortices as they `occur' in GCM experiments. Pier Vellinga. Prof.dr.ir. P. Vellinga, the day's chairman, reported on the conclusions the IVM/ECU project team had drawn from the first day. Three scenarios will be designed. The first is to based on the more optimistic of the GCM outcomes in order to investigate whether the insurance industry can deal with (minor changes in) the present climate. The second scenario is the middle scenario; it is to be based on the more pessimistic GCM outcomes. The third scenario is a surprise scenario, a worst-case analysis; it is to be based on (possibly perturbed) historical analogues. To match the maximum planning horizon of the insurance industry, the investigations will deal with the period up to 2010. Gerhard Berz. Dr. G. Berz of the Munich Reinsurance Company presented the views of the insurance industry on the changing incidence and intensity of extreme weather events. Since 1987, the insurance industry faced a dramatic increase in claims, due to especially high winds. The main reasons for this are the rise in population, the higher standard of living, the concentration of people and goods, the settlement in and industrialisation of extremely exposed areas, the increased susceptibility of modern societies and technologies to natural hazards, the increasing insurance density and the change in environmental conditions. The total capacity of the insurers and reinsurers is at present about a hundred billion U.S. dollars13. This implies that if the 1992 hurricane Andrew (economic losses about $30 bln, insured losses about $16 bln) would have hit 50 miles north (with expected losses 3 to 5 times as high), even the big insurance companies would have become bankrupt. Because of these trends, and the demand for cover outreaching the supply, the position of the reinsurance companies has improved, resulting in higher rates, stricter conditions and increased deductibles14. The goal of this change is to shift the burden from the reinsurers to the insurers and the insured with the aim to enforce better building codes and maintenance. Other problems possibly related to human-induced changes are the increasing hailstorms (observed in some areas), more rainfall, larger agricultural losses due to rising tem- peratures, and possibly ozone pollution, and the expected increase in skin cancer. The losses due to wind depend on the third power of the windspeed15. During warm winters with little snow cover, European storm tracks are longer. Obser- vations suggest that the number of strong extratropical storms increases but levels off. A major question for the response of the insurance industry is which part of the observations they should view as due to natural variability, and which part they should view as due to climatic change; at present it has not yet been possible to separate these two. Important features of climate change from the insurer's viewpoint are the increased weather variability, new extreme weather events in certain regions, new hazard exposures, more frequent and larger natural disasters, greater claims potential, poorer claim experience in such conditions, lagging premium adjustment, and rising demand for natural hazard cover. Policies are (potentially) adjusted every year, but the planning horizon of the most insurance companies is 10, maybe 20 years. An important improvement would be to have available better information of the risks, i.e. the meteorological features (including the influence of the teleconnections), as well as the exposure, and to have better control over the latter, through land use zonation and improved design16. In response to the question whether the insurance industry can be a `burden sharing mechanism' for the costs of climate change, Dr. Berz responded that up to now it could by spreading and atomising risks but in the long run the aforementioned $100 billion is not enough to perform this task adequately. A proper response of the insurance industry to climatic changes could/would include withdrawal from certain markets (for instance, Bangladesh is uninsurable), sound technical pricing, increasing deductibles, limiting liabilities, increasing risk and exposure transparency, disaster prevention and mitigation, raising awareness and improving education, and uniform loss adjustments. These responses constitute a shift from cover selling/premium gathering to integrated risk management. The role governments play in this is important, not only in prevention and mitigation but also in enabling the insurers to build large reserves, which is at present troublesome because of the tax-structures. The final remark considered the desirability to couple atmospheric models to socio-econ- omic and especially insurance models17. Micheal Wilford. M. Wilford of the Foundation for International Environmental Law and Development in London presented the AOSIS18 insurance scheme. One of the more troublesome impact of climate change and sea level rise is the effect they have on the small island states. The insurance industry is not able to cope with small markets with a low probability of a large damage claim. The gov- ernments of these often rather poor countries can also not deal with this matter. Therefore, the thought has emerged that the industrialised countries should develop a `financing pool' to support the poor countries in case of excessive damage due to sea level rise. As the majority of the AOSIS members are at present more concerned with the danger of tropical cyclones, ways are sought to extend the existing scheme for sea level rise to extreme weather events. Two historical analogues exist for this scheme. The first is the international convention on oil pollution. If an area is hit by oil pollution and the responsible shipowner cannot pay (or cannot be traced?), an international fund, gathered from the premiums paid by all oil importers, pays. The second analogue is the OECD nuclear damage convention to cover especially the top-layer of the liability. Premiums are gathered based on the Gross National Product and the number of nuclear installations. These analogues and the AOSIS scheme differ from common insurance in that the claimers do not match the premium payers. The present scheme was submitted to the Rio Climate Convention but has not been included. The scheme contains the following features. The developed countries will contribute money to a fund if a certain sea level threshold (absolute plus relative rise) has been crossed; the first instance to evaluate the sea level threshold is 10 years after the agreement is signed; the threshold will be reviewed every 5 years. Contribution to the fund will be made once the threshold is crossed. The claims will not occur before a certain threshold (relative sea level rise) is crossed; this is expected after 30 to 50 years. Present problems related to this proposal are which countries are to be included in the scheme? How to include tropical cyclones and sea surges? What would be the trigger events for this? How to value areas in small island nations? A real problem is the loss of livelihood instead of damage to infrastructure. Therefore, health insurance might be a better analogue than property insurance, i.e. paying for recovery instead of damage. After a question in the audience, it was remarked that the total claims would be relatively small Ñ the problem is rather political than economic. James Bruce. J.P. Bruce represented the International Decade of National Disaster Reduction (IDNDR), initiated by UNDRO- DHA. He started by restating the dramatic catastrophe record of the last years Ñ between 1966 and 1987 no $100 million loss catastrophes took place, after 1987 15 of these have occurred. Mr. Bruce's feeling suggests that the climate is changing, either due to the natural variability or due to the enhanced greenhouse effect. A lesson drawn from hurricane Andrew, which hit Florida in 1992, is that the development of infrastructure and buildings needs to be restricted by enforcing building codes and landuse zonation. Another lesson from the success of an Indian programme stresses the important of early warning systems in the timing of appropriate measures to mitigate effects. The targets of the IDNDR for the year 2000 are (i) to have assessed the risks of natural disasters, (ii) to have prevention planned in all disaster-prone areas, and (iii) to have installed early warning systems. Jšrg von Seggern. Dr. J. von Seggern of the Frankona Reinsurance Company presented a simple method to map the flooding risk potential. This is important because flooding constitutes a large part of the economic losses due to natural perils, and because floods will be insured in Germany in the near future. The method needs to be simple to keep the amount of information needed, and so the costs, low. The flooding risk in a certain region (defined by postal codes) is determined based on the amount or intensity of losses/damage, the rainfall recurrence intervals, land use, the hydrological order of the rivers crossing the region, the type of landscape in conjunction with the flood history, the upstream catchment, and the rainfall recurrence intervals in these catchments. Criteria for the design of a flood severity index were also proposed as following: the spatial extent, the duration and the cause of the flood, the possibilities of rescue in case of a flood, the number of people injured, the number of people dead, the physical impacts, the percent of buildings affected, and the average damage per risk due to the flood. This results in four scales of severity which combined with the probability of occurrence defines the map legend. Maureen Fordham. Dr. M. Fordham of the Flood Hazard Research Unit of the Middlesex Polytechnic presented her views on the public attitude to environmental impacts. With respect to flood risk she adopted the IPCC CZMS response options, i.e. protection, accommodation, retreat and do nothing, which should be judged in their context according to the economic viability, the environmental sustainability, the public acceptability and the institutional flexibility. The results from a study of the attitudes of people living along two stretches of the Thames19 reveal that the perceived benefits of living along the river outweigh the perceived risks of flooding (although some have no choice but to live there). Next, dr. Fordham presented the results of a case study of showing how well-designed flood protection schemes (seawalls, procedures for taking appropriate measures when risks for flooding is high, landuse zonation) might be subject to failure due to human behaviour. 5. THE DISCUSSIONS OF THE SECOND DAY During the discussion of the second day, the research programme was further refined. Three climate change scen- arios are to be designed for the period up to 2010. The first will describe, based on GCM outcomes, only minor changes in the incidence and intensities of extreme weather events. The second scenario will describe, based on other GCM outcomes, larger changes in extreme events. The third scenario will reflect, based on (possibly perturbed) historical analogues, the maximum credible loss. The GCM outcomes will, insofar as possible, be taken from the transient experiments with coupled models, and where possible use regional models nested in global ones. Sensitivity analyses of the impacts to the GCM outcomes should be performed. The need to couple the climate models to the socio-economic models was recognised, although it was also recognised hat initially this would be difficult. Several studies will be undertaken in the first phase of the research: ¡The impact of windstorms in Europe is to be studied at a geographical scale on at least the size of the southern parts of the UK or the Netherlands but preferably broader. The Dutch greenhouses might provide a suitable case-study. Data are needed on the insurance policies and on the windspeed-loss ratio. ¡River flooding will be studied in the river Rhine in detail, for a certain, yet to be identified stretch, and more generally for a larger region; the flood scenarios will be derived by perturbing present frequency distributions or, where possible, by available hydrological models. ¡The impact of tropical cyclones, taking sea level rise and ENSO into account, is to be studied in the South-West Pacific; other options mentioned, Bangladesh and the South- West Indian Ocean, including Madagascar, are postponed for the time being because of the accessibility of data in the SW Pacific and the political importance of this region in the climate negotiations. ¡Subsidence is to be studied in the UK, with case studies elsewhere. The impact on agriculture and crop loss will be studied European-wide, particularly with respect to drought. Health and heat stress will be studied, particularly in southern Europe and north east U.S.A. as funds permit. 6. THE RESEARCH AGENDA This final section contains the research agenda as it is presently perceived by the ECU and the IVM. It is based on the results of the workshop as well as on the followup discussions between the institutes and the consultants. This agenda constitutes a guideline for the drafting of research proposals. The overall methodology comprises the following research topics: - development of appropriate weather scenarios - studies of the vulnerability of areas, in both the geographic and economic sense, and their developments (baseline scenarios) - combining weather scenarios with vulnerability studies, resulting in impact scenarios - elaboration of socio-economic effects of impacts - policy analysis These elements are briefly elucidated below. Finally some remarks are made with respect to organization and distribution of the eventual research results. Weather scenarios Statistical analysis of existing weather records in order to find the relationship between synoptic weather data, and to design stochastic weather generators. What might an historical review of extreme weather events contribute? Methodologies to derive appropriate weather scenarios from GCM-results Statistical analysis of GCM-results Perturbations of GCM-parameters How to deal with the possibility of shifts in climate modes? Selection of type of weather events and of geographical areas ENSO related Tropical cyclones Extratropical depressions (gales) Rainfall over Europe - flooding, subsidence, droughts Hot spells in urban areas What scenarios to develop? Character, time horizon, how many? Requirements from socio-economic backgrounds (see below) Vulnerable areas Assessments of vulnerabilities of: Small island states North-West Europe (as exposed to gales) Geographic information system of stock-at-risk/exposed areas Cities (subject to hot spells) Public health Tourism Transport Hot spells and security Relationship between hot spells and air pollution River flood plains Coastal zones Agriculture Crop failure due to drought Crop failure due to heat stress, in combination with ozone pollution To what extent are greenhouses (vulnerable to wind and hailstorms, important in Dutch horticulture) insurable Property Housing, building codes and vulnerability Infrastructure Exposure scenario How will the stock-at-risk (exposure) in the different areas develop. Are scenarios available describing developments in economy, landuse, demography and other areas, from which appropriate risk scenarios can be derived? How to do this? What does the event-effect relationship look like? Increased risk scenarios How to combine the exposure scenario with weather scenario to produce risk scenarios? How to describe risks? Through analysis of possible impacts? Insurance industry Economic importance of the insurance industry Strategic behaviour of the various actors on the (re)insurance market Effects of sudden large claims. Effects on financial markets. Policy scenarios What policies are possible in what areas? What can be concluded from an historical analysis? Conclusions from the area of flood risks (perception of risks, attitudes of individuals, quality of risk management schemes) What level of policies (from local to global, from government regulation to free "risk markets") What policies are viable? Which institutions? What institutional changes are needed? Are new institutions necessary? How to implement appropriate policies? Calculated risks versus perceived risks? Organisational aspects Identification of relevant related research done elsewhere. Climate change and flood risks EuroFlood study - EPOCH-funded Studies in the US? Other Climate change and crop failure (ECU involvement?) IDNDR studies (disaster reduction) Identification of available results from research done in the insurance industry Building a research network Expanding the "workshop network". Dissemination of the research results Workshops Publications in journals distributed in various disciplinary areas Publication of an edited volume ANNEX 1. LIST OF ADDRESSES PARTICIPANTS: Mr. James P. Bruce Canadian Climate Programme Board (IDNDR) 1975 Juno Avenue, Ottawa, Canada K1H 656 Telephone: 1 613 7315929 Fax: 1 613 7313509 Dr GŸnther P. Kšnnen KNMI PO Box 201 3730 AE De Bilt The Netherlands Telephone: 31 30 206451 Fax: 31 30 210407 Dr Gerald A. Meehl National Centre for Atmospheric Research P.O. Box 3000, Boulder, Colorado 80307-3000 United States of America Telephone: 1 303 4971000 Fax: 1 303 4971137 Dr Francis W. Zwiers Canadian Climate Centre/University of Victoria, PO Box 1700 MS 3339, Victoria, B.C Canada V8W 2Y2 Telephone: 1 604 3638229 Fax: 1 604 3638247 Dr Jšrg von Seggern Frankona RŸckversicherung Maria Theresiastrasse 35, 8000 MŸnchen Germany Telephone: 49 89 92280 Fax: 49 89 9228395 Dr W. John Maunder WMO, Chairman Climatology Commission, World Climate Programme Department Case Postale 2300 CH-1211 Geneva 2 Switzerland Fax: 41 22 7342326 Ernst J.A. Lohman, M.Sc. Ernst Lohman Consultants Bolhaarslaan 83, 7522 CV Enschede The Netherlands Telephone: 31 53 337003 Fax: 31 53 329618 Dr Peter R. Rowntree Hadley Centre, Meteorological Office London road, Bracknell Berkshire RG12 2SY United Kingdom Fax: 44 344 854898 Mr. Michael Wilford Clyde and Company 51 Eastcheap London EC3M 1JP United Kingdom Telephone: 44 71 6231244 Fax: 44 71 6235427 Mrs. Paula A. Harrison, Dr Thomas E. Downing, Mr. George M.C. Blumberg ECU, University of Oxford 1a Mansfield Road, Oxford OX1 3TB United Kingdom Telephone: 44 865 281180/6 Fax: 44 865 281181 Dr Maureen Fordham Flood Hazard Research Unit, Middlesex Polytechnic Queensway Enfield, Middlesex EN3 4SF United Kingdom Telephone: 44 81 3625359 Fax: 44 81 3625403 Dr Gerhard Berz MŸnchener RŸckversicherungs Gesellschaft Koeninginstrasse 197, 8000 MŸnchen, 40 Germany Telephone: 49 89 38910 Fax: 49 89 399056/7 Drs Hans van Zwol Nationale Nederlanden Warmondstraat 79 I Telephone: 31 20 6177889 Dr Nigel Arnell UK Institute for Hydrology Crownmarsh Gifford, Wallingford, Oxfordshire OX10 8BB United Kingdom Telephone: 44 491 838800 Fax: 44 491 832256 Prof. dr Lennart Bengtsson Max-Planck-Institut fŸr Meteorologie Bundesstrasse 55 D-2000 Hamburg 13 Deutschland Telephone: 49 40 411730 Prof. dr ir Pier Vellinga, Dr Xander A. Olsthoorn, Drs Richard S.J. Tol, Drs Richard J.T. Klein, Mrs. Use van der Hul Institute for Environmental Studies, Vrije Universiteit De Boelelaan 1115 1081 HV Amsterdam The Netherlands Telephone: 31 20 5483827 Fax: 31 20 6445056 ANNEX 2. CONTRIBUTIONS OF THE PARTICIPANTS 0 This model has 20 vertical levels but the grid is still quite crude. For example, only two grids cover New Zealand. The transient experiment uses a `deep ocean' of up to 5 kilometers, the equilibrium model has a `slab ocean'. 1 Thermal contrasts over the North Atlantic have increased over the last 30 years. 2 Therefore, much greater warming is likely in the Northern Hemisphere that in the Southern Hemisphere. 3 In the audience it was remarked that the reduced snow- cover, due to higher temperatures, further enhances this effect. 4 except for Northern India; this may be due to the enhanced monsoons 5 The crossing frequency might decrease due to a crossing duration increase. 6 In the audience it was mentioned that this event is more sensitive to the structure of the model than the 33ûC event. 7 According to dr. Kšnnen, there is some evidence from atmospheric-ocean models that the North-Atlantic has in fact two equilibrium states. 8 In the audience it was remarked that rainfall is not the sole determinant of the risk of subsidence, e.g. the type of soil and the temperature interfere as well Ñ as a demonstration of a methodology, though, this definition suffices. 9 In the audience it was remarked that this study in fact only shows that there is non-linearity in the observed temperature rise. 10 not between the reference case and the future scenario, but within the reference period 11 Note that this conclusion was altered later on; all scenarios are to be based on the most sophisticated models available, i.e. the transient experiments of GCMs with a realistic representation of the oceans. 12 In the audience it was remarked that during ENSO events the distribution of sea surface temperatures changes, possibly extending the area in which cyclones occur. 13 American billion, i.e. 1,000,000,000. 14 A large part of the losses due to windstorms consists of a large number of small claims. 15 In the audience it was remarked that it is more likely to be a step function; with increasing windspeeds the relation might shift to even higher powers. 16 A striking example of unknown exposure is the insurance of the factories of a large company Ñ the company is insured as a whole, the locations of the individual factories is unknown; a practical problem in tackling this is that, for instance, hotels have a global policy for building design whereas the exposure differs from place to place; a problem with land use zonation is that it cannot influence the present stock-at-risk. 17 In the audience it was remarked that Lloyd's have a model of the insurance market and that others are developing models. 18 Association Of Small Island States 19 Maidenhead (1990) and Datchet, Whatsbury, Staines and Chertsey (1989 and 1990); between the two DWSC studies, a flood actually occured.