The ClimEx Project: Investigating Climate Variability to Study Extreme Events in a Warming World
Ludwig-Maximilians-Universität München (Germany)
Local Project ID:
HPC Platform used:
SuperMUC of LRZ
Hydrometeorological extremes, such as droughts and floods are one of the grand challenges of our future and pose great interest and concern for water management and public safety. Hence, the ClimEx project disaggregates the response of the climate system into changing anthropogenic forcing and natural variability by analyzing a novel large-ensemble of climate simulations, operated using High-Performance Computing. The comprehensive new dataset (CRCM5-LE) generated 50 transient independent and evenly likely realizations of the past and the future climate (1950-2099) over two large domains (Europe, Eastern North America) in high spatial (12km) and temporal (1h-1d) resolution. The resulting 7500 model years allow for a thorough analysis of extreme statistics and derivation of robust estimates of return values for hydrometeorological extreme events (e.g. floods) under current and future climate conditions.
The ClimEx project evolved from the long-term collaboration of research institutions and authorities from Bavaria and Québec on topics related to climate change and its effects on water resources . The ongoing project focuses on hydro-meteorological extreme events (e.g. extreme rainfall events and major river floods) as well as their links to natural climate variability and human induced climate change.
Since extreme events are rare by definition and observations are hampered by short time series, they are of limited suitability for the analysis of climate variability. Therefore, the ClimEx project created a climate model ensemble of 50 equally likely simulations over Québec (Canada) and Europe for the past and the future to better asses the role of climate change, climate variability and their effects on hydro-meteorological extreme events. Furthermore, this unique dataset will provide an ideal basis to determine if and when a clear climate change signal emerges from natural climate variability.
To accomplish these tasks, a complex model chain involving three modelling steps was introduced to connect global climate conditions with local hydrological impacts. The first step of this chain was performed by the Canadian Centre for Climate Modelling and Analysis of Environment and Climate Change Canada (ECCC) (further described in ) by providing 50 realizations of the driving Global Climate Model (GCM) CanESM2 (Canadian Earth System Model, version 2). It simulates the entire Earth’s surface from 1950 to 2100. From 1950 to 2005 observed emissions (in CO2 and other greenhouse gases, aerosols), volcanic eruptions and solar cycle forcing are applied, whereas for the following years (until 2100) the high concentration scenario RCP 8.5 (Representative Concentration Pathway ) is used. The differences between each of the 50 realizations (members of the GCM) are based on slight perturbations in their initial conditions (i.e. in one aspect of the model’s cloud properties), which produce an inter-member variability that is considered to represent natural variability in the climate system. Furthermore, the extensive number of model years (50x 150 years = 7500 years) permits to create an appreciable amount of extreme events which otherwise would rarely occur. This dataset was made available to the ClimEx project for the subsequent spatial refinement of the rather coarse resolution of the GCM for further applications.
In a second modelling step, the Canadian Regional Climate Model version 5 (CRCM5; developed by Université du Québec à Montréal (UQAM) in collaboration with ECCC ) was used to increase the spatial resolution of the CanESM2 (310 km) data to 12 km by dynamical downscaling (Figure 1) resulting in the CRCM5 large-ensemble (CRCM5-LE). This was the first computational step of the ClimEx project performed for the European and the northeastern North American domain. Since this step required a vast amount of computational resources (computation time and storage), the project relied on the provision of high performance computing (HPC) capacities. Started in 2016, 88 million core-hours on the SuperMUC supercomputer at the Leibniz Supercomputing Centre (LRZ) were employed over a one year calculation period for the dynamical downscaling. The resulting dataset presents a valuable data source for research communities considering climate change and its impact on the environment. Furthermore, it provides a profound basis for extreme value analysis.
As a third step, the meteorological variables of the CRCM5-LE are used to drive the hydrological model WaSiM  for 98 catchments in Bavaria in a high spatial (500 m) and temporal resolution (3 h) setup. A novel global calibration strategy (i.e. finding model parameters which are valid for the entire Bavaria to fit the model to observations) using an automatized approach (Dynamically Dimensioned Search (DDS)  in combination with Simulated Annealing (SA)) was applied. A set of various objective functions were employed to find the optimal parameterization of the hydrological model. Since this iterative process was computationally demanding, large parts were performed on LRZ’s CoolMUC2 as well as SuperMUC.
The analysis of the CRCM5-LE reveal strong climate change signals of temperature and precipitation for future periods (Figure 2, ). Over the European domain, the results indicate an increase in winter precipitation as well as extreme precipitation, whereas during the summer period droughts will occur more frequently.
For Bavaria, the analysis on extreme precipitation (e.g. 5-day maximum precipitation) indicates an increase in the frequency of occurrence and in volume. Furthermore, the data shows a clear climate change signal by 2040.
Since the computation of the hydrological simulations is still in progress, statements about the development of runoff and related extreme events in a changing climate for the entire Bavarian domain are not available yet. However, the application of the CRCM5-LE for the hydrological model will provide a profound database for a robust estimation of rare hydrological extreme events (e.g. HF100 – high flow of a recurrence period of 100 years, Figure 3).
 Fyfe, J. C. et al., 2017: Large near-term projected snowpack loss over the western United States. Nature Communications, 8, 14996, doi:10.1038/ncomms14996
 Meinshausen, M. et al., 2011: The RCP greenhouse gas concentrations and their extensions from 1765 to 2300. Climatic Change, 109 (1-2), 213-241, doi: 10.1007/s10584-011-0156-z
 Šeparović, L., et al., 2013: Present climate and climate change over North America as simulated by the fifth-generation Canadian Regional Climate Model. Clim Dyn, 41, 3167–3201, doi:10.1007/s00382-013-1737-5.
 Schulla, J., 2017: Model Description WaSiM. Technical Report. pp. 347.
 Lespinas, F. et al., 2017: Performance of the dynamically dimensioned search algorithm: influence of parameter initialization strategy when calibrating a physically based model. Hyd Res, 49/1, doi:10.2166/nh.2017.139
 Leduc et al., 2018. ClimEx project: a 50-member ensemble of climate change projections at 12-km resolution over Europe and northeastern North America with the Canadian Regional Climate Model (CRCM5). Submitted to Journal of Applied Meteorology and Climatology (submitted Feb. 2018; major revisions Apr. 2018; major revisions Oct. 2018).
Prof. Dr. Ralf Ludwig
Fakultät für Geowissenschaften
Lehrstuhl für Geographie und geographische Fernerkundung
e-mail: r.ludwig [@] lmu.de