ALBERT

Description: The beginning of the 21st century was marked by a number of severe summer floods in Central Europe associated with extreme precipitation (e.g., Elbe 2002, Oder 2010 and Danube 2013). Extratropical storms, known as Vb-cyclones, cause summer extreme precipitation events over Central Europe and can thus lead to such floodings. Vb-cyclones develop over the Mediterranean Sea, which itself strongly warmed during recent decades. Here we investigate the influence of increased Mediterranean Sea surface temperature (SST) on extreme precipitation events in Central Europe. To this end, we carry out atmosphere model simulations forced by average Mediterranean SSTs during 1970–1999 and 2000–2012. Extreme precipitation events occurring on average every 20 summers in the warmer-SST-simulation (2000–2012) amplify along the Vb-cyclone track compared to those in the colder-SST-simulation (1970–1999), on average by 17% in Central Europe. The largest increase is located southeast of maximum precipitation for both simulated heavy events and historical Vb-events. The responsible physical mechanism is increased evaporation from and enhanced atmospheric moisture content over the Mediterranean Sea. The excess in precipitable water is transported from the Mediterranean Sea to Central Europe causing stronger precipitation extremes over that region. Our findings suggest that Mediterranean Sea surface warming amplifies Central European precipitation extremes.

Description: Much of our knowledge about future changes in precipitation relies on global (GCM) and/or regional climate models (RCM) that have resolutions which are much coarser than typical spatial scales of precipitation, particularly extremes. The major problems with these projections are both climate model biases and the gap between gridbox and point scale. Wong et al. developed a model to jointly bias correct and downscale precipitation at daily scales. This approach, however, relied on pairwise correspondence between predictor and predictand for calibration, and thus, on nudged simulations which are rarely available. Here we present an extension of this approach that separates the downscaling from the bias correction and in principle is applicable to free running GCMs/RCMs. In a first step, we bias correct RCM-simulated precipitation against gridded observations at the same scale using a parametric quantile mapping approach. To correct the whole distribution including extreme tails we apply a mixture distribution of a gamma distribution for the precipitation mass and a generalized Pareto distribution for the extreme tail. In a second step, we bridge the scale gap: we predict local variance employing a vector generalized linear gamma model (VGLM gamma) with the bias corrected time series as predictor. The VGLM gamma model is calibrated between gridded and point scale (station) observations. For evaluation we adopt the perfect predictor experimental setup of VALUE. Precipitation is in most cases improved by (parts of) our method across different European climates. The method generally performs better in summer than in winter and in winter best in the Mediterranean region with a mild winter climate and worst for continental winter climate in mid & eastern Europe or Scandinavia. A strength of this two-step method is that the best combination of bias correction and downscaling methods can be selected. This implies that the concept can be extended to a wide range of method combinations.

Description: To investigate the influence of atmospheric model resolution on the representation of daily precipitation extremes, ensemble simulations with the atmospheric general circulation model ECHAM5 at different horizontal (T213 to T31) and vertical (L31 to L19) resolutions and forced with observed sea surface temperatures and sea ice concentrations have been carried out for 01/1982 - 09/2010. All results have been compared with the highest resolution, which has been validated against observations. Resolution affects both the representation of physical processes and the averaging of precipitation across grid boxes. The latter, in particular, smoothes out localized extreme events. These effects have been disentangled by averaging precipitation simulated at the highest resolution to the corresponding coarser grid. Extremes are represented by seasonal maxima, modeled by the generalized extreme value distribution. Effects of averaging and representation of physical processes vary with region and season. In the tropical summer hemisphere, extreme precipitation is reduced by up to 30% due to the averaging effect, and a further 65% owing to a coarser representation of physical processes. Towards mid- to high latitudes, the latter effect reduces to 20%; in the winter hemisphere it vanishes towards the poles. A strong drop is found between T106 and T63 in the convection dominated tropics. At the lowest resolution, northern hemisphere winter precipitation extremes, mainly caused by large scale weather systems, are in general represented reasonably well. Coarser vertical resolution causes an equatorward shift of maximum extreme precipitation in the tropics. The impact of vertical resolution on mean precipitation is less pronounced; for horizontal resolution it is negligible.

Description: Extreme weather events can have serious impacts on human society as well as on ecosystems. Future and recent changes, as well as the underlying atmospheric mechanisms of the extremes, are usually estimated from climate model simulations that employ atmospheric general circulation models (AGCMs) of relatively coarse horizontal resolution (of the order of a few hundred kilometers). This coarse resolution has an impact on simulated extreme events, particularly precipitation extremes. Since heavy precipitation events are highly variable in space, precipitation extremes are expected to decrease with bigger grid size due to the effect of averaging individual small scale events across a bigger area. Coarser horizontal and vertical resolution may also degrade the representation of some physical processes. To study the impact of horizontal and vertical model resolution on the representation of extreme precipitation events, we analysed simulations with the ECHAM5 AGCM at different horizontal (T213, T159, T106, T63, T42, T31) and vertical resolutions (L31 for T213 to T42, and L19 for T63 to T31) using the same transient present day (1982-2009) boundary forcing (sea surface temperature and sea ice concentration). For each season and grid box, parameters of a stationary generalised extreme value (GEV) distribution were estimated and 20 season return values were derived as a measure of extreme precipitation. All results are compared to the simulation with the highest resolution, T213L31. To disentangle the effect of averaging from the scale dependent representation of physical processes, the high resolution T213 was averaged to the grids of the coarser resolutions for comparison on equal spatial scales. As expected, the return values decrease with coarser resolution. However, the scale dependency changes with region and season. Strongly decreasing return values were found between T106 and T63 covering an almost entire zonal band, which is particularly pronounced in regions of deep convection. It is worthwhile to note, however, that even at the most crude T31L19 resolution, the model was still capable of representing extreme precipitation over the northern hemisphere in winter reasonably well. These precipitation events are mainly caused by large frontal systems. We found high vertical resolution to also be very important for a better simulation of extreme precipitation, as a coarser vertical resolution yields an equatorwards shift of the intertropical convergence zone (ITCZ). The representation of mean precipitation was found to be practically independent of horizontal model resolution