Achtergrond

Climate aspects of the floods in Bosnia and Serbia May 2014

Over the last weeks, torrential rains fell in the valley of the Sava River that runs between Bosnia and Croatia and flows through Serbia into the Danube.

These rains caused large-scale flooding, with loss of life and extensive damage. A question that often arises is whether this event was related to climate change. In a first look at the event based on preliminary data we find no evidence that the warming trend has contributed significantly to the severity of the precipitation leading to the floods in this area.

MODIS false colour image of the flooding of the Sava River (source: NASA Earth Observatory)
MODIS false colour image of the flooding of the Sava River (source: NASA Earth Observatory)

The meteorological situation

After a wet April 2014, 34 mm of rain fell in the Sava basin on May 3 and 4, followed by 87 mm on May 13-16, which is the highest 4-day sum in the record since 1950 (based on preliminary data).

It was caused by a low-pressure system that was stationary over this area on these days. Locally the precipitation reached 150 mm in four days in this analysis, which is as much as usually falls in two months,

How rare was the event?

Based on the 34 years of data we estimated the return time of the maximum 4-day mean precipitation in Europe in the calendar months from July to June. Because of the relatively short time series return times above 100 years have very large uncertainty margins.

Figure 1: 4-day summed precipitation averaged over the basin of the Sava river, 1979-February 2014 ERA-interim, April/May 2014: ECMWF operational.
Figure 1: 4-day summed precipitation averaged over the basin of the Sava river, 1979-February 2014 ERA-interim, April/May 2014: ECMWF operational.
Figure 2: Precipitation in Europe 13-16 May 2014 [mm/4day], ECMWF analysis.
Figure 2: Precipitation in Europe 13-16 May 2014 [mm/4day], ECMWF analysis.
Figure 3: Extrapolation of a generalized Extreme Value distribution (GEV) fitted to the annual maximum of 4-day mean precipitation in the years July-1979 to June 2013 to the value observed in June 2013 to 20 May 2014. The 95% confidence interval of the re
Figure 3: Extrapolation of a generalized Extreme Value distribution (GEV) fitted to the annual maximum of 4-day mean precipitation in the years July-1979 to June 2013 to the value observed in June 2013 to 20 May 2014. The 95% confidence interval of the re

The return time of the observed 4-day precipitation in the Sava basin, 84 mm/4dy, is more than 60 years based on an extrapolation of the 34 years we have available at the moment.

On a map of the return times over all land points of Europe, 3% of the area is covered with return times larger than 50 year, mainly in the Sava / Danube basins. By chance an area of 2% is expected to have a return time of 50 years or more every year. This implies that an event like this occurs on average every year or two somewhere in Europe because of the natural variability of the weather.

Figure 4: Map of return times of the highest 4-day precipitation in July 2013 to 20 May 2014 in the distribution of the years July 1979 to June 2013. Values above 50 years have large uncertainty margins.
Figure 4: Map of return times of the highest 4-day precipitation in July 2013 to 20 May 2014 in the distribution of the years July 1979 to June 2013. Values above 50 years have large uncertainty margins.

Has the probability of an event like this increased due to global warming?

We consider here two ways to obtain a preliminary answer this question. The first is to look for trends in the observations. The time series of Figure 1 is compatible with no trend (there is a non-significant negative trend). Over the longer time series starting in 1950 there is a slight positive trend, but no significant difference in the return time of the heavy precipitation of mid-May 2014. The observations therefore did not give an indication that the probability of a heavy precipitation event like the one that occurred had increased before it happened.

The second way is to study the trend with climate models. This reduces the uncertainty due to natural variability at the expense of systematic errors in the representation of the weather in this region. This first look does not permit for a validation of the models in this respect. It is assumed they do a reasonable job in simulating the mechanisms responsible for changes in extreme multi-day precipitation. As shown in Figure 2, this extreme was caused by a large-scale weather system. These are in general are well-described by these models. However, it is known that in summer there are problems with the models underestimating precipitation trends in this area (see e.g., Bladé et al, Clim. Dyn., 2012, van Haren et al, Clim. Dyn., 2013), whereas regional models tend to simulate the observed trends better.

Figure 5 shows that a few models project a decrease in maximum 5-day precipitation in Bosnia-Herzegovina and Serbia, but the majority has a modest increase of around 15% at the end of the century in the highest CO2concentration scenario, RCP8.5. This translates to a few per cent increase up to now, much less than the natural variability.

Figure 5: Projected change in the annual maximum of 5-day precipitation in 23 CMIP5 models used for the IPCC Fifth Assessment Report for the RCP8.5 scenario that assumes that the highest rise in CO2 concentrations (source: CCCMA / KNMI Climate Change Atla
Figure 5: Projected change in the annual maximum of 5-day precipitation in 23 CMIP5 models used for the IPCC Fifth Assessment Report for the RCP8.5 scenario that assumes that the highest rise in CO2 concentrations (source: CCCMA / KNMI Climate Change Atla

On the map, the difference between the end of the century and the period 1981-2010 is about 10%, similar to the standard deviation of the natural variability (hatched) even in the highest scenario. Again, the trend up to now is much smaller than the trend to the end of the century in the highest scenario.

For a subset of these models the output has been further analysed with a hydrological model by Hirabayashi et al (Nature Climate Change, 2013). They find a decrease in flooding risk in this area, opposed to the very small increase in heavy precipitation events. This may well be due to the decrease in the mean precipitation, which gives rise to drier soils that absorb a larger fraction of the rain before it reaches the rivers. However, another study based on a small ensemble of regional climate models generated for the previous IPCC rapport found an increase in flood risks (Dankers and Feyen, GRL, 2009).

Conclusions

The precipitation event of 13-16 May 2014 on the Sava and Danube rivers was very rare, with a return time higher than 60 years. However, as it only covered a few per cent of the area of Europe an event with a return time like this is expected every year or two somewhere on the continent just due to the natural variability of the weather. There is no clear trend of 4-day precipitation events prior to this year in the observations and climate models only show a very small increase up to now. Results from hydrological models suggest that the probability of a flood on these rivers can decrease, rather than increase, due to climate change.

Literature

  • Bladé et al, 2012, Clim. Dyn. 39 709–727, doi:10.1007/s00382-011-1195-x
  • Dankers and Feyen, 2009, Geophys. Res. Lett. 114 D16108, doi:10.1029/2008JD011523
  • Hirabayashi et al, 2013, Nature Clim. Change 3 816-821, doi: 10.1038/nclimate1911
  • Van Haren et al, 2013, Clim. Dyn. 40 1–20, doi: 10.1007/s00382-012-1401-5
  • Van Oldenborgh et al, IPCC WG1 AR5 Annex I “Atlas” and KNMI Climate Change Atlas, climexp.knmi.nl/atlas. 


Many thanks to Rein Haarsma for a critical look at the manuscript. This work was supported by the NWO project `Trend or Coincidence' and the EU project EUCLEIA.

Niet gevonden wat u zocht? Zoek meer achtergrond artikelen