Periods like heat waves (HWs) and long-time warm spells are common during the summer months. However, these events have become more frequent and severe in the last two decades with devastating consequences (Meehl and Tebaldi, 2004; Della-Marta et al., 2007a; Russo et al., 2015). Mega-Heatwaves (Barriopedro et al., 2011) like 2003 in Central Europe or 2010 in Eastern Europe and Russia led to ten thousands additional heat related deaths (Fouillet et al., 2006; Robine et al., 2007) and had an extensive impact on the economy and environment. Events like wildfires and droughts frequently accompany the actual HW increasing the death toll and economic losses foremost due to crop failures. Therefore, seasonal to decadal predictions as well as climate projections are of great importance for warning systems and adaptation planning. Long-term observations as well as large ensembles of model simulations are applied to analyse the mechanism of HWs and to assess their predictability. This study considers three research topics: Decadal predictability, dependence to soil conditions and future changes of HWs.
The impact of changes in soil moisture and soil temperature fields on a decadal simulation for Europe is explored with the land surface model (LSM) VEG3D with special focus on summer temperatures. LSMs as the lower boundary of climate models are sensitive to inaccuracies in the initialisation data since they disturb the LSM from its balanced state. The fluctuations of the drift back to the balanced state are transferred to a certain degree into the atmosphere and are part of the bias in climate simulations. Therefore, stand-alone (SA) simulations are conducted with VEG3D to investigate its sensibility to the atmospheric forcing, model resolution and external data (soil types and land use). The SA results are evaluated against measurements from stations as well as satellite data and simulations with other LSM. The comparison revealed that ERA-Interim (Dee et al., 2011) forced simulations offer the highest correlations to the observations. With this forcing, balanced soil initial fields are produced by a 0.22 VEG3D SA simulation for Europe. A decadal prediction with a coupled COSMO-CLM_VEG3D system was performed using the new initial fields of VEG3D SA. The soil initialization show an evident impact on the spatial and temporal variability of summer temperatures in Europe. In particular, transition zones are dominated by the dependence of the evapotranspiration to the soil moisture, the effect was detectable even in the second pentad of the simulation. An added value was found for the Iberian Peninsula in the summertime related to the dryer soil in the run with the balanced fields. This simulation produced a more realistic drought stress in summer due to the lower soil water and higher 2 m temperatures. On the other hand, the impact is detected for less than four years in Scandinavia. There, an overall higher soil moisture is available limiting the land atmosphere coupling, although traceable differences in the deeper soil layers still exist. It can be concluded, that accurate soil initializations can be crucial for the development of extreme summer temperatures depending on dryer and warmer soils in model simulations. This finding is highly important in regard of the climate change and a possible enlarging of the transition zones.
The predictability of HWs is investigated with the extensive ensemble of global and regional decadal hindcasts conducted in the Bundesministeriums für Bildung und Forschung (BMBF) funded Mittelfristige Klimaprognose project (MiKlip Marotzke et al., 2016). In the first step, two reanalyses driven CCLM simulations are employed to estimate the ability of the regional model to reproduce HW characteristics for Europe derived from gridded daily observations (EOBS, ECA&D Haylock et al., 2008). A warm bias in Southern Europe and a cold bias in Northern Europe is found for the HW temperatures, whereas the number of HWs and their averaged mean and maximum lengths are quite well simulated. Decadal variations of the HW temperatures are detected with a moderate correlation (r ~0.5) to the Atlantic meridional oscillation (AMO). In addition, mediocre to high correlations (positive as well as negative) are found between HWs and the large-scale circulation anomalies (LCA) mostly located in the area of their centre of actions. An increased number of HW days is found for Central Europe during the positive upper third NAO index, while the days are reduced in Turkey and Southeastern Europe. Similar correlation patterns are detected in the reanalysis driven CCLM simulations indicating the model is able to reproduce the teleconnection of HWs and LCA. However, the correlations between monthly HW day sums and the LCA indices are not symmetrically distributed, when the index is filtered into its positive and negative phase. Depending on the region, higher (anti-)correlations
are noticed for the different index phases. One phase seems to represent a dominant signal and mostly only smaller or no vice-versa signals are noticed for the contrary phase. For example, the correlation is 0 for the upper third of the EA pattern for the British Isles, whereas the negative lower third shows a moderate anticorrelation (r< -0.45). The yearly mean and maximum temperatures, as well as the lengths and HW per year numbers, are of similar quality in the regional hindcast ensemble as the ones from the reanalysis driven simulations. The global ensemble displays a larger cold bias for the temperatures and a positive bias for the HW lengths. A new approach to evaluate the ensemble on event basis was tested using the t-test to calculate fractional hit rates for each lead time and ensemble member. This method avoids a too strong smoothing of the events due to the established multi-year and ensemble-member averaging. In addition, HW days per year are introduced as a new user relevant variable. Furthermore, the evaluation showed, that they are predictable with a good skill overall in Europe.
A new ensemble of cloud-resolving climate projections was produced to bridge the gap between usual climate simulations with around 25 km resolution and impact models operating on several hundred meters and to fulfil the demand after very high resolved climate information. Forced by data from three global models, the ensemble is dynamically downscaled to 0.025 with three nesting steps (global to 0.44 to 0.0625 to 0.025). Besides the control (CTRL) period 1971-2000, two future period runs (2021-2050 and 2071-2100) using the RCP8.5 scenario were performed. Comparing the final nest, 0.025, to the coarser 0.0625revealed an added value of the
higher resolution due to a decreased temperature bias. In contrast to this, higher precipitation errors are produced by the cloud-resolving run. Daily HW temperatures are better represented in the 0.025 simulations in the mean but their variance is too large compared to observations. An increase in the HW temperatures of nearly 7C is detected for the future periods until 2100. In addition, an expansion of their occurrence time is noticed. HW conditions of the CTRL period are identified for nearly the whole summer half-year in the distant future. A reduction of precipitation was detected along with a decrease of relative humidity, soil moisture, latent heat flux and cloud cover and an increase of 2m and soil temperatures, sensible heat and incoming solar radiation. The lower soil moisture is one of the most key factors for HW development and amplification. More severe and longer lasting HWs can develop in the future periods due to the large reduction of the soil water content. Thus, conditions like 2003 will become more frequent rather than an exceptionally rare event.