globalchange  > 气候变化事实与影响
DOI: doi:10.1038/nclimate2124
论文题名:
Increasing stress on disaster-risk finance due to large floods
作者: Brenden Jongman
刊名: Nature Climate Change
ISSN: 1758-1388X
EISSN: 1758-7508
出版年: 2014-03-02
卷: Volume:4, 页码:Pages:264;268 (2014)
语种: 英语
英文关键词: Climate-change impacts ; Environmental sciences ; Hydrology ; Economics
英文摘要:

Recent major flood disasters have shown that single extreme events can affect multiple countries simultaneously1, 2, 3, which puts high pressure on trans-national risk reduction and risk transfer mechanisms4, 5, 6. So far, little is known about such flood hazard interdependencies across regions7, 8 and the corresponding joint risks at regional to continental scales1, 9. Reliable information on correlated loss probabilities is crucial for developing robust insurance schemes5 and public adaptation funds10, and for enhancing our understanding of climate change impacts9, 11, 12. Here we show that extreme discharges are strongly correlated across European river basins. We present probabilistic trends in continental flood risk, and demonstrate that observed extreme flood losses could more than double in frequency by 2050 under future climate change and socio-economic development. We suggest that risk management for these increasing losses is largely feasible, and we demonstrate that risk can be shared by expanding risk transfer financing, reduced by investing in flood protection, or absorbed by enhanced solidarity between countries. We conclude that these measures have vastly different efficiency, equity and acceptability implications, which need to be taken into account in broader consultation, for which our analysis provides a basis.

Major river floods are typically driven by large-scale atmospheric circulations8, 13, 14. As a result, single flood episodes can affect vast areas in a short period of time, irrespective of economic and political boundaries1, 3. This was demonstrated in June 2013 by the blocking of the planetary waves of the atmospheric flow regime in the Northern Hemisphere2, which led to extensive flooding and €12 billion losses15 in nine different countries across central and eastern Europe. Understanding the risk posed by large-scale floods is of growing importance, as their impacts are rising owing to socioeconomic development6, 16, and their frequency and intensity may increase under a changing climate1, 9, 12, 17.

Well-devised risk management of climate-related extremes, including floods, is therefore considered to be an important pillar of climate adaptation18. Rising flood losses already force insurance companies to increase their capital base and may lead to more years of below-zero profitability5. Uninsured risks are a growing concern, as a lack of financial means for relief, recovery and reconstruction negatively affects the wellbeing of people, the economy and a country’s budget6, 19. Accurate information on the joint probability of flood losses that takes into account spatial correlations between river basins across different countries is essential for developing insurance mechanisms5 and public compensation schemes10 robust to present and future extreme losses. This information is especially required and informative in the European Union (EU), where international disaster financing is increasingly connected through insurance regulations21, climate change adaptation strategies20, and a joint compensation mechanism between member states10.

So far, methods for producing large-scale flood risk estimates have either been based on specific hazard event scenarios, or are upscaled from lower to higher spatial levels by summation of basin-level risk16, 22, 23, 24. In both cases, natural correlation between events is neglected (that is, full spatial independence across river basins is assumed) and reliable estimates of extreme losses cannot be made. Hence, flood risk projections available to the disaster risk reduction community do not accurately represent geographical risk patterns and are not probabilistic in nature. We demonstrate here that natural dependencies among risks in different regions can be accounted for (Methods), and we present probabilistic projections of flood risk in the EU.

We find monthly peak river discharges in the 1,007 sub-basins to be a good proxy for the occurrence of reported damaging flood events17 on a European scale, as shown in Supplementary Fig. 1. The results show high positive cross-correlations in observed peak discharges between the river sub-basins in Europe, indicating a large degree of spatial interdependence in extreme river flows. Spearman’s correlations are significant (α=0.05) in 63% of all sub-basins, and in 98% of the sub-basins showing strong correlations (that is, r >0.7; Supplementary Table 1).

Strong positive cross-correlations in peak discharge occur between basins in central and eastern Europe, following the patterns exhibited during the 2002 and 2013 floods across multiple countries in this region (Fig. 1a). Peak discharges in this area are often linked to the atmospheric circulation pattern Vb, or Genoa Low; that is, a low-pressure system travelling from the Atlantic southeast across the Mediterranean towards central Europe1. High-to-strong cross-correlations amongst southern European basins (Fig. 1b) are known to be caused by the occurrence of regular Mediterranean depressions25, whereas regional negative cross-correlations are also observed under the influence of Atlantic depressions26. We also find high-to-strong correlations in peak discharges amongst basins in western European countries, which have been linked to the occurrence of atmospheric rivers and extra-tropical cyclones13 (Fig. 1c). On the basis of the peak discharge correlations, we assigned countries to 5 main regions, which are used for computing country-specific losses and the required compensation payments (Fig. 1d; Methods). In this study, the correlations were computed over the entire time series for which discharge data were available (1990–2011; Methods). The results may vary depending on the selected time periods, because some of the atmospheric circulation patterns and resulting peak discharges show seasonal variation (Supplementary Fig. 2); and the circulation patterns, and hence rainfall distributions and intensities, may be influenced by climate change1, 9, 13, 17. Uncertainty in these changes, however, remains high12.

Figure 1: Correlations of monthly peak discharges between basins in Europe.
Correlations of monthly peak discharges between basins in Europe.

Spearman’s rank correlation coefficient of extreme monthly discharges among European river sub-basins, calculated on the basis of LISFLOOD30 simulations forced by observed daily climate data for the period 1990–2011 (Supplementary Fig. 1; Supplementary Table 2). ac, The correlations of all river basins with the basin containing the cities of Vienna, Austria (a); Rome, Italy (b); London, United Kingdom (c). d, We derived best estimates of aggregated natural discharge correlations between countries and identified 5 main regions of correlated extreme discharges. These regions are used to convolute probability distributions on an EU level and calculate country losses (Methods).

Potential flood losses for the period 2000–2050 for each of the 1,007 river basins were computed at a high (100×100 m) resolution using simulated daily discharge data, extreme value analysis, spatial inundation modelling, and an economic damage model, following the method described in earlier work24, 27. The projections of flood hazard up to 2050 are based on LISFLOOD (ref.30) simulations driven by an ensemble of 12 climate experiments derived from a combination of 4 general circulation models and 7 regional climate models24 (Supplementary Methods). Both the climate change and economic development components of the flood and damage models were forced by the SRES-A1B scenario.

Flood protection standards, defined as the minimum statistical probability discharge that leads to flooding, were modelled for each basin in three steps. First, minimum and maximum flood protection standards in the EU were estimated from literature study (Supplementary Table 2) at 1 per 10 years and 1 per 500 years, respectively. A 1 per 1,000 years standard was manually assigned to The Netherlands, following the national flood defence levels in place there. Second, the EU-average flood protection was estimated by running the flood damage model with all hypothetical protection levels and analysing the intersection with reported losses15. A flood protection standard ranging between the minimum and maximum was assigned to each basin as a function of the average potential damage per square metre relative to the EU average. The resulting protection levels map shows a range between 1 per 10 (basin with the lowest potential damage per square metre) and 1 per 500 (basin with the highest potential damage per square metre), and a separately assigned 1 per 1,000 level in The Netherlands (Supplementary Fig. 3). Modelled protection standards are thus higher in areas where the potential damages are high, owing to a relative concentration of people and assets, than in areas where potential damages are low. This is in line with policies in major European river watersheds such as the Rhine and the Danube, where higher levels of protection are maintained in densely populated areas than in rural areas. The cubic interpolation over basins was calibrated using points of known flood protection (Supplementary Table 2). Flood protection standards were assumed constant over time for the projections presented in Fig. 2 (that is, the protection measures are assumed to be only upgraded to maintain the same failure probability under climate change, without further adaptation), and were increased to the new potential minimum standards in Fig. 3c (corresponding to the values on the y axis in this panel).

We combined the estimated potential damage with the modelled basin-level protection standards to derive probability loss curves for each year in which all modelled losses below the protection level were set to zero. We then used the peak river discharge time series to estimate natural dependencies between basins. To account for nonlinearities in the dependency structure, we use the (flipped) Clayton copula Cθ(u, v), a specific copula from the Archimedean family:

The Clayton copula provided sufficient flexibility in modelling dependencies given the data at hand (Supplementary Methods, Supplementary Figs. 5 and 6). This model was used to aggregate basin loss curves to the country level in a stepwise manner using the estimated copula parameters as the ordering criteria (Supplementary Methods). In more detail, the selection of the next basin to be aggregated is based on maximizing the smallest tail dependency between the already selected basins and the potential candidates. This procedure avoids underestimation as well as overestimation of the risk. A stepwise conditional copula approach was adopted to estimate dependencies at the country level within the derived country groups (Fig. 1d): in the order of descending estimated pairwise Clayton copula parameters θ, the conditional copulas were used as stepwise extensions of the joint loss distributions. Between country groups independence was assumed. The conditional (flipped) Clayton copula with parameter θ is given by:

Finally, we computed 1 million random samples from the multivariate flood loss model to calculate statistical loss probabilities on a country basis, for each year in the time series. Expected average losses are correspondingly defined as the mean of all samples in the year of analysis. We factored total modelled direct losses for each country by empirically estimated insurance coverage rates (Supplementary Fig. 4) to approximate expected average insurance payouts. Expected claims to the EUSF were calculated from total estimated losses following the payout regulations governing this fund, which are based on the size of the damage relative to the national GDP and an arbitrary payout threshold10. Finally, we computed residual losses for national public and private sectors as the total flood losses minus insurance and Solidarity Fund payouts.

  1. Becker, A. & Grünewald, U. Flood risk in central Europe. Science 300, 10991099 (2003).
  2. Wake, B. Flooding costs. Nature Clim. Change 3, 778 (2013).
  3. Kundzewicz, Z. W., Pińskwar, I. & Brakenridge, G. R. Large floods in Europe, 1985–2009. Hydrol. Sci. J. 58, 17 (2013).
  4. Michel-Kerjan, E. & Kunreuther, H. Redesigning flood insurance. Science 333, 408409 (2011). URL:
http://www.nature.com/nclimate/journal/v4/n4/full/nclimate2124.html
Citation statistics:
资源类型: 期刊论文
标识符: http://119.78.100.158/handle/2HF3EXSE/5209
Appears in Collections:气候变化事实与影响
科学计划与规划
气候变化与战略

Files in This Item:
File Name/ File Size Content Type Version Access License
nclimate2124.pdf(2666KB)期刊论文作者接受稿开放获取View Download

Recommended Citation:
Brenden Jongman. Increasing stress on disaster-risk finance due to large floods[J]. Nature Climate Change,2014-03-02,Volume:4:Pages:264;268 (2014).
Service
Recommend this item
Sava as my favorate item
Show this item's statistics
Export Endnote File
Google Scholar
Similar articles in Google Scholar
[Brenden Jongman]'s Articles
百度学术
Similar articles in Baidu Scholar
[Brenden Jongman]'s Articles
CSDL cross search
Similar articles in CSDL Cross Search
[Brenden Jongman]‘s Articles
Related Copyright Policies
Null
收藏/分享
文件名: nclimate2124.pdf
格式: Adobe PDF
此文件暂不支持浏览
所有评论 (0)
暂无评论
 

Items in IR are protected by copyright, with all rights reserved, unless otherwise indicated.