Réf. Schär & al. 2004 - A

Référence bibliographique complète

SCHÄR C., VIDALE P. L., LÜTHI D., FREI C., HÄBERLI C., LINIGER M. A., APPENZELLER C. The role of increasing temperature variability in European summer heatwaves. Nature, 2004, Vol. 427, 332-336.

Abstract: Instrumental observations and reconstructions of global and hemispheric temperature evolution reveal a pronounced warming during the past ~150 years. One expression of this warming is the observed increase in the occurrence of heatwaves. Conceptually this increase is understood as a shift of the statistical distribution towards warmer temperatures, while changes in the width of the distribution are often considered small. Here the authors show that this framework fails to explain the record-breaking central European summer temperatures in 2003, although it is consistent with observations from previous years. They find that an event like that of summer 2003 is statistically extremely unlikely, even when the observed warming is taken into account. They propose that a regime with an increased variability of temperatures (in addition to increases in mean temperature) may be able to account for summer 2003. To test this proposal, they simulate possible future European climate with a regional climate model in a scenario with increased atmospheric greenhouse-gas concentrations, and find that temperature variability increases by up to 100%, with maximum changes in central and eastern Europe.

Climate change, heatwaves, summer 2003, simulations, temperaure variability, return period, Europe.

Organismes / Contact
Atmospheric and Climate Science, ETH Zürich, Winterthurerstrasse 190, 8057 Zürich, Switzerland. schaer@env.ethz.ch
MeteoSwiss, Krähbühlstrasse 58, 8044 Zürich, Switzerland

(1) - Paramètre(s) atmosphérique(s) modifié(s)
(2) - Elément(s) du milieu impacté(s)
(3) - Type(s) d'aléa impacté(s)
(3) - Sous-type(s) d'aléa

Pays / Zone
Massif / Secteur
Site(s) d'étude
Période(s) d'observation
Northern Switzerland

(1) - Modifications des paramètres atmosphériques
Large-scale analysis of summer 2003
A record-breaking heatwave affected the European continent in summer 2003. In a large area, mean summer (June, July and August, referred to as JJA below) temperatures have exceeded the 1961–90 mean by ~3°C, corresponding to an excess of up to 5 standard deviations. Even away from the centre of action, many long-standing temperature records have tumbled.

In the Northwestern foothills of the Alps, the year 2003 is far off the distribution in three of the four panels (Basel-Binningen, Geneva, Bern-Liebefeld, and Zürich). For instance, the previous record holder for JJA was 1947 with a temperature anomaly of T’ = 2.7°C (with respect to the 1864–2000 mean). The corresponding value for 2003 is as high as T’ = 5.1°C and this amounts to an offset of 5.4 standard deviations from the mean.

Estimation of return period

The two distributions [1864–1923 & 1941–2000] show similar characteristics in general, but the 1941–2000 distribution is shifted by the mean warming (DT = 0.8°C) between the two periods. This shift also implies a change in the frequency of extremes. For instance, the frequency of a month with an anomaly of T’ = 3°C has increased by ~100%. Hence, a month in the 1941–2000 period with an excess temperature of T’ = 3°C can be tied with a probability of 50% to the warming between the two periods, in a probabilistic sense. This illustrates how comparatively small shifts in climate mean may imply pronounced changes at the tails of the statistical distribution and in the frequency of extremes.

With respect to the reference period 1864–2000, the summer 2003 return period is several million years, but such an excessive estimate based on a short series is dubious. To account for the warming in the last decades, a more recent reference period 1990–2002 (with DT = 1.25°C warmer mean temperature) has been used. With respect to this climatology, the resulting return period for summer 2003 still amounts to t = 46,000 yr. The uncertainty of this estimate is considerable, however, and the lower bound of the 90% confidence interval is t = 9,000 yr.

This large return period should not be overstated, and is here merely used to express the rareness of such an extreme summer with respect to the long-term instrumental series available. In particular, the analysis does not exclude the possibility that such warm summers might have occurred in the more distant historical past, for instance in the Medieval Warm Period, in 1540 or in 1757. It suggests, however, that an event like summer 2003 does not fit into the gaussian statistics spanned by the observations of the reference period, but might rather be associated with a transient change of the statistical distribution. This interpretation is consistent with the idea that small changes of the statistical distribution can yield pronounced changes in the incidence of extremes.
Northern Switzerland:
In the SCEN simulation [2071–2100], the distribution of JJA temperatures is shifted by ~4.6°C towards warmer temperatures for the grid point in northern Switzerland. More important, SCEN also exhibits a pronounced widening of its statistical distribution, with the standard deviation increasing by 102%. This widening is statistically highly significant and only slightly affected by the transient warming within the two periods (a revised estimate using detrended temperature series yields a somewhat smaller variability increase of 86%).

Both data sets (observations and climate change simulations) exhibit a similar (statistically significant) relationship between temperature and precipitation anomalies. The regression analysis yields slopes of - 11% / °C and - 8.2% / °C for the observations and the simulations, respectively. Thus, although there is some underestimation of summer precipitation in CTRL (at the grid point under consideration, by 21%), the simulations credibly represent the observed precipitation sensitivity. Despite a general trend towards drier conditions with increasing temperatures, there is also an increase in the incidence of heavy precipitation events.

Scatter diagrams showing summer mean temperature and precipitation anomalies for northern Switzerland demonstrate that in terms of temperature and precipitation the climatic conditions in JJA 2003 were not unlike those simulated by SCEN for the period 2071–2100. For northern Switzerland, the 2003 observation is located approximately in the middle of the SCEN data points. Thus, the RCM simulations suggest that towards the end of the century about every second summer could be as warm or warmer (and as dry or dryer) than 2003.

The spatial distribution of the relative increase in variability shows a pronounced signal throughout central and eastern Europe that is not directly linked to the simulated mean temperature change. The warm summers of SCEN show signs of drought, with the semi-arid Mediterranean climate progressing towards central Europe. In SCEN, central Europe is more often (but not always) affected by summer droughts than in CTRL [1961–90], and this implies an increase in variability. The drought conditions develop in response to large-scale anticyclonic forcing, and they nonlinearly amplify local temperature anomalies. During droughts the net balance of solar and infrared radiation is almost entirely balanced by local heating, while evapotranspiration is suppressed owing to the lack of soil moisture. This process may be further amplified by a positive feedback between soil moisture and precipitation.

All of GCM and RCM scenarios analysed exhibit a substantially increased level of variability over large parts of Europe. The simulated increase in variability also implies an increase in extremes relative to mean climatic conditions. For illustration, a 50% increase in the standard deviation of the long-term JJA temperature series (standard deviation = 0.94°C) would raise the probability of a 2003-like event (T’ = 3.85°C with respect to 1990–2002) by a factor of ~150. For an event with T’ = 5°C, it would increase by a factor of ~5100. This tremendous sensitivity of extremes to the width of the statistical distribution has led to the statement “variability is more important than averages”. A recent increase in variability is thus a plausible hypothesis to explain extreme JJA 2003 conditions.

This results demonstrate that the European summer climate might experience a pronounced increase in year-to-year variability in response to greenhouse-gas forcing. Such an increase in variability might be able to explain the unusual European summer 2003, and would strongly affect the incidence of heatwaves and droughts in the future.

Informations complémentaires (données utilisées, méthode, scénarios, etc.)

Large-scale analysis of summer 2003
The continental-scale temperature anomaly for JJA 2003 is based on ERA-40 reanalysis data (for 1961–90) and operational meteorological analysis data (for 2003) of the European Centre for Medium-Range Weather Forecasts (ECMWF). Monthly temperatures are computed as means of daily Tmin and Tmax. Small height differences between the ERA-40 and ECMWF topographies are accounted for by the use of an adiabatic lapse rate (0.6 °C per 100 m).
For further analysis, the authors consider long-term temperature series from Switzerland, located close to the centre of the anomaly. Twelve carefully homogenized series are available with daily resolution since 1864. To minimize contamination by local meteorological and instrumental conditions, they amalgamate four independent and particularly reliable stations (Basel-Binningen, Geneva, Bern-Liebefeld, and Zürich) into one single series with monthly temporal resolution. This series is representative for the northwestern foothills of the Alps.

Estimation of return period
In a first step, they restrict attention to the time period 1864–2000 and compile compound statistics for all monthly temperature anomalies (January–December). The purpose of this is to identify changes near the tails of the statistical distribution that result from the warming trend in the series. To this end, they consider two 60-yr periods, one covering the beginning of the series (1864–1923), and one the end (1941–2000). The resulting statistical distributions are given both in terms of cumulative probability and probability density functions.
To quantitatively assess the 2003 situation, they have estimated its return period. The return period t is an estimate of the frequency of a particular event (or its exceedance) based on a stochastic concept. Here they employ a gaussian distribution fitted to JJA temperatures to estimate t with respect to a selected reference period.

The stochastic concept adopted in the estimation of return periods assumes independent, identically distributed JJA temperatures with the underlying distribution being gaussian. The distribution parameters are estimated from the data of the reference period, using the method of moments (which in the case of a gaussian distribution is identical to maximum-likelihood estimation). The return period of the event (expected frequency of threshold exceedance) is then calculated from the fitted distribution. Confidence bounds of the return period were calculated by parametric resampling. These take into account the uncertainty of the parameter estimates given the finite sample size (that is, the number of summers in the reference period), but not the uncertainty in the underlying stochastic concept. We have also tested whether the data are reasonably gaussian distributed, checking quantile-quantile plots.

Climate change simulations
As a shift of the statistical distribution by the observed mean warming is unable to explain the record-breaking summer 2003, they hypothesize that the heatwave might be due to a change of the distribution’s width, representing an increase in year-to-year variability. Support for this hypothesis comes from a regional climate model (RCM) driven by a greenhouse-gas scenario representing 2071–2100 conditions (SCEN). The scenario integration is compared against a control integration covering the period 1961–90 (CTRL). At the lateral boundaries, the RCM is driven by a model chain consisting of two general circulation models.
The climate change scenario is based on the A2 scenario of IPCC. The scenario computations involve three numerical models:
the low-resolution HadCM3 global coupled atmosphere–ocean GCM;
the intermediate-resolution HadAM3H atmospheric GCM;
the CHRM limited-area high-resolution RCM, used with a horizontal resolution of 56 km and 20 levels in the vertical, and driven at its lateral boundaries by HadAM3H.

Summer temperature and precipitation anomalies are displayed against each other, both for the observations and for the climate change simulations. Both panels include a data point representing observed JJA 2003 conditions, and the results apply to northern Switzerland. The observed data are based on averages of conventional temperature and precipitation (rain-gauge) observations at the four stations referred to above, while the simulated data are shown for a single grid point roughly corresponding to the location of our long-term series.

(2) - Effets du changement climatique sur le milieu naturel

Sensibilité du milieu à des paramètres climatiques
Informations complémentaires (données utilisées, méthode, scénarios, etc.)

(3) - Effets du changement climatique sur l'aléa

Paramètre de l'aléa
Sensibilité du paramètres de l'aléa à des paramètres climatiques
Informations complémentaires (données utilisées, méthode, scénarios, etc.)

(4) - Remarques générales


(5) - Syntèses et préconisations


Référence citées :

Folland, C. K. et al. Global temperature change and its uncertainties since 1861. Geophys. Res. Lett. 28, 2621–2624 (2001).

Easterling, D. R. et al. Observed variability and trends in extreme climate events: A brief review. Bull. Am. Meteorol. Soc. 81, 417–425 (2000).

Aschwanden, A. et al. Bereinigte Zeitreihen: Die Ergebnisse des Projekts KLIMA90 (MeteoSwiss, Zürich, 1996).

Bergert, M. et al. Homogenisierung von Klimamessreihen und Berechnung der Normwerte 1961–1990 Veröffentlichungen der MeteoSchweiz, 67, MeteoSwiss, Zürich, (2003).

Allen, M. R. Liability for climate change. Nature 42, 891–892 (2003).

Katz, R. W. & Brown, B. G. Extreme events in a changing climate: Variability is more important than averages. Clim. Change 21, 289–302 (1992).

Schär, C., Lüthi, D., Beyerle, U. & Heise, E. The soil-precipitation feedback: A process study with a regional climate model. J. Clim. 12, 722–741 (1999).

Wild, M., Dümenil,L.& Schulz ,J.P. Regional climate simulation with a high resolution GCM: surface hydrology. Clim. Dyn. 12, 755–774 (1996).

Christensen, J. H. & Christensen, O. B. Severe summertime flooding in Europe. Nature 421, 805–806 (2002). [Fiche biblio]

Simmons, A. J. & Gibson, J. K. The ERA-40 Project Plan ERA-40 (Project Report Series, No. 1, ECMWF, Reading, 2000).

Johns, T. C. et al. Anthropogenic climate change for 1860 to 2100 simulated with the HadCM3 model under updated emissions scenarios. Clim. Dyn. 20, 583–612 (2003).

Jones, R. G., Murphy, J. M., Hassell, D. C. & Taylor, R. Ensemble Mean Changes in a Simulation of the European Climate of 2071–2100 Using the New Hadley Centre Regional Modelling system HadAM3H/ HadRM3H (Hadley Centre Report, Hadley Centre, Exeter, UK, 2001).