final_program_abstracts[1]

11 IMSC Session Program
Model evaluation of the hydrological cycle in the present and
future climate
Thursday - Parallel Session 2
Nathalie Schaller, Jan Cermak and Reto Knutti
Institute for Atmospheric and Climate Science, ETH Zurich, Switzerland
The intensification of the hydrological cycle is expected to have major impacts on
human societies and ecosystems, which is why reliable statements about future
changes in the precipitation patterns are urgently needed. Unfortunately, the model
spread for precipitation is particularly broad. Another source of concern is the
disagreement in trends between observations and model simulations of precipitation.
While both climate models and observations predict an increase of the total amount of
water in the atmosphere of 7% per Kelvin of surface warming, all climate models
indicate a precipitation increase between only 1 to 3%/K, which is lower by a factor 2
or more than the observed increase in precipitation during the last decades.
Here we explore new ways of evaluating and understanding the intensification of the
hydrological cycle in the climate models along with, as background motivation, the
identification of potential reasons for the discrepancy between models and
observations.
A common way to evaluate the model simulations is to use statistical measures to
quantify their biases with respect to observations on the global scale. However, since
precipitation is highly variable on both the spatial and temporal scales, metrics
representing regional features of the modeled precipitation response to climate change
might be more suitable to identify the good models. Three different ways of ranking
the climate models are therefore compared, considering: a) biases in a broad range of
climate variables, b) only biases in global precipitation and c) regional features of
modeled precipitation in areas where future changes are expected to be pronounced. A
surprising result is that the multimodel mean performs only average for the featurebased
ranking, while it outperforms all single models in the two biasbased rankings. It
is further shown that many models have similar biases and that the observation
datasets are often located at one end of the model range. This outcome indicates that
instead of using the “one model, one vote” approach, weighting the models according
to their ability to simulate the present climate on a regional scale might lead to more
reliable projections for precipitation.
Abstracts 238