What did we learn in this project, what are the limitations, and what should be further
For hourly precipitation intensity on a local scale, we found evidence for a 10-14
% increase per degree warming. In recent years, the intensity of summer showers appears
to be 15 % higher than before 1990, which is likely attributed a one-degree warming
of the Netherlands. Extrapolating this relation to the future, increases of up to
100 % in intensity appear possible by the end of this century (with a warming of
5 degrees, which is approximately the upper range in the Dutch KNMI’06 climate scenarios).
Extrapolating observed relations to such high temperatures is obviously dangerous
– although results from Hong Kong suggest that it may work reasonably well – and
therefore we also looked for evidence in regional climate models. Unfortunately,
models appear to be very uncertain at predicting changes in hourly extremes with
changes between close to zero and up to 60-80%. What is even worse is that the models
fail to reproduce the observed relations for high temperatures, casting serious doubt
on their ability to predict future changes. A major cause of these model deficiencies
is related to the fact that these model do not resolve the physics of convective
clouds that give rise to extreme precipitation intensities; instead they use simplified
prescriptions, so-called parameterizations.
For large scale precipitation extremes connected to synoptic scale (>500 km) low
pressure systems which occur mostly in the winter season, it was shown that the role
of natural variability is large even on a 30-year time scale. When the natural variability
was averaged out, it turned out that extremes increased at a rate similar to the
mean precipitation change in winter. Typical increases are 3-7 % per degree warming,
which are consistent with the KNMI’06 scenarios.
For the present-day climate, It was shown in that the probability of a NNW storm
surge is larger after a period of 5-20 days of extreme precipitation in the Rhine
catchment than climatology. The probability of a simultaneous occurrence of a high
river discharge and NNW storm surge could be a factor 4 higher than in the case that
these events are independent (as commonly assumed). This could have considerable
implications for the safety norms. Yet, we would also like to note two limitations
of this study. First, these results are obtained in a model with relatively coarse
resolution. Second, in order to have sufficient statistics we could only look at
moderate extremes occurring approximately once a year.
We studied precipitation extremes on two totally different scales: local intensity
on an hourly time scale and multi-day precipitation on a scale of hundreds of km.
In between there is a whole range of time and spatial scales, which are relevant
for different users such as, for instance, the water boards. The event on the 26-27
August 2006, as discussed in here, underlined the importance of these intermediate
scales. We do not have sufficient insight into how these intermediate-scale (mesoscale)
precipitation extremes could change. They are affected by processes acting on different
scales. Convective processes could give rise to increases in the order of 10-14 %
per degree warming, yet large scale processes give rise to changes that are 2-3 times
smaller. In addition, the influence of non-local processes, such as the occurrence
of atmospheric rivers channeling moist air from the sub-tropics to our regions, are
poorly understood and could give rise to unforeseen changes.
Much of the research will continue in the KfC research program “Theme 6: High quality
climate projections” which runs from 2011 to 2015.
Atmospheric mesoscale atmospheric dynamics (on a scale of 100 m to 100 km) are responsible
for many high impact weather events (heavy precipitation, lightning, wind gusts,
hail). There is only marginal understanding of how these mesoscale systems could
change in the future. Climate models do not resolve (or only partly resolve) these
type of events (even regional models run at a high resolution of 10-20 km) and analysis
of observations generally suggest (much) stronger dependencies of the dynamics of
such events on temperature (humidity) than is obtained in climate models.