opinion & comment CORRESPONDENCE:
Adaptation to extreme heat in Stockholm County, Sweden To the Editor — Oudin Åström et al. argue1 that global climate change has doubled the incidence of heat-related mortality in Stockholm County, Sweden, because of an increase in the number of extreme-heat events between the periods 1900–1929 and 1980–2009. They link this to their finding that no adaptation to extreme heat occurred during the latter period. This conclusion is notable because many studies of trends in heat-related mortality across the United States2,3 and Europe4,5 have reported declines in both total mortality and the sensitivity of urban populations to extreme heat, despite an increasing frequency of extremeheat events. Such findings indicate that adaptation has more than kept pace with climate change. Oudin Åström et al. reach the opposite conclusion by failing to account for (1) influences on the Stockholm temperature record not related to global-scale warming, and (2) the large degree of adaptation to extreme-heat events that occurred over the entirety of the twentieth century. Oudin Åström et al. calculate heat-related mortality as the product of the number of extreme-heat events, the seasonal baseline mortality and the elevated relative risk of mortality during extreme-heat events. For the period 1980–2009, Oudin Åström et al. identify 378 extreme-heat events during which the risk of mortality was increased by 4.6%. This combination results in 689 heatrelated deaths. They identify 220 extremeheat days from 1900–1929 and conclude that large-scale climate change was responsible for the additional 158 occurrences of extreme-heat events observed during the period 1980–2009, which resulted in an additional 288 heat-related deaths with respect to the baseline climate. The temperature dataset used by Oudin Åström et al.6 includes the influence of natural regional variability, local processes associated with urbanization and globalscale climate change. The naturally occurring Atlantic Multidecadal Oscillation (AMO) is known to be a major driver of regional temperature variability across northern Europe7. The AMO was primarily in its cold phase during the 1900–1929 period 302
and primarily in its warm phase during the 1980–2009 period7 — a difference likely to be responsible for some portion of the increase in extreme-heat events identified by Oudin Åström et al. and inappropriately attributed to global climate change. Additionally, the Stockholm temperature record is not likely to be free of spurious warming from urbanization. The urban adjustment in their data, which reflects the influence of population growth during the mid-twentieth century, remains constant since the late 1960s, the result of a levelling of the urban population of Stockholm that extended from the 1960s through the early 1990s6. Stockholm has since become one of the fastest-growing cities in Europe with population growth of more than 25%8. It is possible that the increase in extreme-heat events reported by Oudin Åström et al. during the 1980–2009 period includes elements of this recent population expansion. Furthermore, there is no consideration of the effects of adaptation between the 1900–1929 and 1980–2009 periods. In another paper examining changes in heatrelated mortality in Stockholm County9, the enhanced relative risk from extreme-heat events during the base period 1900–1929 was derived using a similar methodology as in Oudin Åström et al. and found to be very close to 20%. This is substantially greater than the 4.6% identified during the period 1980–2009 and indicates that a sizeable adaptation to extreme heat has taken place between the two time periods, as noted in both papers1,9. To estimate the impact of this adaptation, we substitute the relative risk in the base period 1900–1929 for the relative risk in the 1980–2009 period, leaving the other parameters (number of extreme-heat events and seasonal baseline mortality) unchanged from the 1980–2009 values. This substitution indicates that in the absence of adaptation, 2,993 heat-related deaths would have occurred under the observed climate and population characteristics of the period 1980–2009. The difference between the unadapted (2,993) and the actual (689) heat-related mortality is 2,304. That number of averted deaths, presumably a result of
more effective adaptation, is eight times the number identified to have occurred as a result of global climate change (288), itself a likely overestimate. The observed decline in the relative risk from extreme-heat events is described by Oudin Åström et al. as “probably due to, among other factors, improvements in health care and dwellings and behavioural changes during temperature extremes.” Some portion of this response probably reflects the temporal increase in the frequency of extreme-heat events, an increase that elevates public consciousness and spurs adaptive response. In this manner, climate change itself leads to adaptation10,11. It is insufficient to ignore this effect when compiling and discussing the impacts of climate change2. If an increasing frequency of heat events raises public awareness and gives rise to an adaptive response that lowers the population’s relative risk due to extreme heat, this must be properly weighed against any increases in mortality that result from a greater number of mortality-inducing heat events. In the case of Stockholm County, raised awareness from climate change need only be responsible for 288 out of 2,304 (~13%) deaths saved through adaptation to have completely offset the climate-related increase in heat-related mortality identified by Oudin Åström et al. For any greater contribution, climate change would have resulted in an overall decline in heat-related mortality in Stockholm County despite an increase in the frequency of extreme-heat events. Such a result would be consistent with findings for other major cities in Europe and the US2–5 and would stand in opposition to the conclusions drawn by Oudin Åström et al. Our analysis highlights one of the many often overlooked intricacies of the human response to climate change. ❐ References 1. Oudin Åström, D., Forsberg, B., Ebi, K. L. & Rocklöv, J. Nature Clim. Change 3, 1050–1054 (2013). 2. Davis, R. E., Knappenberger, P. C., Michaels, P. J. & Novicoff, W. M. Environ. Health Perspect. 111, 1712–1718 (2003). 3. Kalkstein, L. S., Greene, S., Mills, D. M. & Samenow, J. Nat. Hazards 56, 113–129 (2010). 4. Matzarakis, A., S. Muthers, S. & Koch, E. Theor. Appl. Climatol. 105, 1–10 (2011). 5. Kyselý, J. & Plavocá, E. Climatic Change 113, 437–453 (2012).
NATURE CLIMATE CHANGE | VOL 4 | MAY 2014 | www.nature.com/natureclimatechange
© 2014 Macmillan Publishers Limited. All rights reserved
opinion & comment 6. Moberg, A., Bergström, H., Ruiz Krisman, J. & Svanerud, O. Climatic Change 53, 171–212 (2002). 7. Sutton, R. T. & Dong, B. Nature Geosci. 5, 288–292 (2012). 8. Statistics Sweden (accessed 28 October 2013); http://www.scb.se/ 9. Oudin Åström, D., Forsberg, B., Edvinsson, S. & Rocklöv, J. Epidemiology 24, 820–829 (2013).
10. Fouillet, A. et al. Int. J. Epidemiol. 37, 309–317 (2008). 11. Palecki, M. A., Changnon, S. A. & Kunkel, K. E. Bull. Am. Meteorol. Soc. 82, 1353–1367 (2001).
Paul Knappenberger1*, Patrick Michaels1 and Anthony Watts2
Cato Institute, 1000 Massachusetts Ave, NW, Washington DC 20001, USA, 2 IntelliWeather, 3008 Cohasset Rd Chico, California 95973, USA. 1
*e-mail:
[email protected]
Reply to ‘Adaptation to extreme heat in Stockholm County, Sweden’ Oudin Åström et al. reply — We thank Knappenberger and colleagues for their interest in our research1. Their correspondence expresses two concerns: a possible bias in the temperature data2 and appropriate consideration of adaptation to extreme-heat events over the century. To clarify, we estimated the impacts of observed climate change over the century on temperature-related mortality; our purpose was not to determine what caused the climatic changes. Our study aimed to examine the health impacts of temperature extremes on the population during the period 1980–2009, given the societal and infrastructure changes that occurred over the twentieth century, if this population had experienced the climate of the period 1900–1929. We did not adjust for actual adaptation responses because the low public awareness of the health hazards of high ambient temperature suggests that there would have been limited autonomous adaptation, and because data were not available to adjust for any actual adaptation responses. We did not compare the relative risk of mortality during an extreme day between 1900–1929 and 1980–2009, as this would be misleading. With respect to the temperature data, we compared the station data recorded during the two study periods (1900–1929 versus 1980–2009). To limit the influence of regional and decadal variability, we used the standard
approach of comparing patterns over 30-year time periods. The observed changes are the result of natural processes, including regional climate variability, and anthropogenic influences, including urbanization3. Our method of comparing the climate during two 30-year periods is valid for any two periods. Sensitivity analyses using different reference periods when calculating the cut-off temperatures (1910–1939, 1920–1949, 1930–1959, 1940–1969 and 1950–1979) limit the influence of the Atlantic Multidecadal Oscillation (AMO)4. For all periods that were investigated, the increase in the number of excess heat extremes ranges from 77 for the reference period 1930–1959 to 158 for the reference period 1950–1979. The AMO during the 1990s was similar to the warm state of 1931–1960 during which there was an increase in the number of heat extremes, albeit not to the extent of the original reference period. We appreciate the opportunity to correct any misperceptions about adaptation to heat extremes in Stockholm. Our data indicate that there is no adaptation to heat extremes on a decadal basis or to the number of heat extremes occurring each year. Although another study observed a reduction in the population health impact of hot and cold extremes over the twentieth century5, this decrease should not be confused with adaptation to climatic change. As in the
studies cited by Knappenberger et al., socio-economic development, epidemiological transitions and health system changes were and continue to be the main drivers of changes in population sensitivity — not explicit, planned actions to prepare for climate change impacts. These changes also apparently increased population resilience to climate change. Whether future development pathways will continue to increase resilience will also depend on many factors other than climate change. Importantly, it is not appropriate to assume that historic trends will continue, with or without climate change. ❐ References 1. Oudin Åström, D., Forsberg, B., Ebi, K. L. & Rocklöv, J. Nature Clim. Change 3, 1050–1054 (2013). 2. Moberg, A., Bergström, H., Ruiz Krisman, J. & Svanerud, O. Climatic Change 53, 171–212 (2002). 3. IPCC Climate Change 2013: The Physical Science Basis (eds Stocker, T. F. et al.) (Cambridge Univ. Press, 2013). 4. Sutton, R.T. & Dong, B. Nature Geosci. 5, 788–792 (2012). 5. Oudin Åström, D., Forsberg, B., Edvinsson, S. & Rocklöv, J. Epidemiology 24, 820–829 (2013).
Daniel Oudin Åström1*, Bertil Forsberg1, Kristie L. Ebi1 and Joacim Rocklöv2 1 Department of Public Health and Clinical Medicine, Division of Occupational and Environmental Medicine, Umeå University, 901 87 Umeå, Sweden, 2Department of Public Health and Clinical Medicine, Division of Epidemiology and Global Health, Umeå University, 901 87 Umeå, Sweden. *e-mail:
[email protected]
COMMENTARY:
Costing natural hazards Heidi Kreibich, Jeroen C. J. M. van den Bergh, Laurens M. Bouwer, Philip Bubeck, Paolo Ciavola, Colin Green, Stephane Hallegatte, Ivana Logar, Volker Meyer, Reimund Schwarze and Annegret H. Thieken The proposed ‘cost assessment cycle’ is a framework for the integrated cost assessment of natural hazards.
R
eported costs of natural hazards are at historically high levels, and are rising due to the ever increasing cost of events with large-scale effects. The Thailand flood in 2011, for example, shut down scores of factories, damaging
global car manufacturing and electronics industries. In 2013, Typhoon Haiyan in the Philippines caused many casualties and displaced thousands of people. Globally in 2013, natural hazards caused damage estimated at US$125 billion1. Property
NATURE CLIMATE CHANGE | VOL 4 | MAY 2014 | www.nature.com/natureclimatechange
© 2014 Macmillan Publishers Limited. All rights reserved
damage has doubled about every seven years over the past four decades2. But such assessments generally do not reflect the complete set of costs of natural hazards, which comprise direct, business interruption, indirect, intangible and risk 303
