The first multi-model ensemble of regional climate simulations at kilometer-scale resolution part 2: historical and future simulations of precipitation

Emanuela Pichelli, Erika Coppola, Stefan Sobolowski, Nikolina Ban, Filippo Giorgi, Paolo Stocchi, Antoinette Alias, Danijel Belušić, Segolene Berthou, Cecile Caillaud, Rita M. Cardoso, Steven Chan, Ole Bøssing Christensen, Andreas Dobler, Hylke de Vries, Klaus Goergen, Elizabeth J. Kendon, Klaus Keuler, Geert Lenderink, Torge Lorenz, Aditya N. Mishra, Hans-Juergen Panitz, Christoph Schär, Pedro M. M. Soares, Heimo Truhetz and Jesus Vergara-Temprado

Climate Dynamics
Work Package 3
Link: https://doi.org/10.1007/s00382-021-05657-4

Highlights

Limiting the effects of climate change will require accurate predictions of its effects. Advances in computer technology are now allowing us to run high-resolution climate simulations at kilometre-scale, which can more accurately model the evolution of the atmosphere and its fine scale processes than lower resolution models. This paper presents the first multi-model ensemble of 10-year regional climate simulations at km-scale for the present climate and for the end of the century. This new generation ensemble is able to simulate present day precipitation more reliably than the driving model ensemble at lower resolution, improving some aspects badly represented at lower scale such as the simulation of too frequent and weak rainfall events, the early onset of convection and the underestimation of severe event intensity. The high-resolution simulations exacerbate changes in the future climate, predicting, among others, an increase in the number of extreme rainfall events in the Alpine region studied. Improvements to climate projections introduced by such an approach might be crucial to define more effective climate change adaptation and mitigation strategies, helping policymakers make the best-informed decisions.

Background
Regional Climate Models (RCMs) are area-limited models used to simulate local climate at a relatively high resolution, improving the representation of the atmosphere and, therefore, extreme weather events. At their highest resolutions they are capable of explicitly simulating atmospheric deep convection, a key process in many extreme weather events, rather than using approximated physical schemes to represent it, which are known to be one of the biggest sources of uncertainty in these simulations. RCMs at high resolution are widely used in weather forecasting, but they require large supercomputing resources, limiting their use in longer-term climate modelling. Improved computer power, however, has now made their use in climate prediction more viable. This paper presents the first multi-model ensemble of 10-year, kilometre-scale RCM simulations for the present day and the end of the 21^st century. The ensemble of ‘convection-permitting’ models is compared with high-resolution observations and with a lower resolution ensemble of their driving models to assess the benefits of simulating regional precipitation and the most severe events at such fine scale. Improving these predictions helps people and policymakers formulate climate adaptation and mitigation measures with the best available information.

Results
This study is performed on the greater Alpine domain, with a special focus over sub-regions where high-resolution observations are available (Switzerland, France and Italy), simulating the current climate and projecting the future one in a high-emissions scenario. The team found that the ensemble of high-resolution simulations brought clear improvements over its lower resolution counterpart when looking at the recent climate, when compared with observations. In particular, it improved the so called ‘drizzle problem’, the tendency of low-resolution models to predict too frequent, weak rainfall events. The higher resolution models better simulated hourly and daily rainfall frequency and intensity, and showed a lower spread in their results over some regions, leading to increased confidence in the ensemble. The two sets of future projections were mostly in agreement in terms of climate change signal, but with larger changes over today’s climate in the high-resolution ensemble. Summer mean rainfall is predicted to decrease over the Alpine region by the end of the century mostly because of a decrease of the frequency of rain events. In autumn, on the other hand, the mean precipitation is shown to increase overall in the north of the region and decrease in the south, in association with less frequent but more intense events. The high-resolution ensemble reduces the uncertainty of precipitation statistics over most of the analysed areas, and especially for heavy precipitation events in autumn, which are projected to increase in number and intensity. These results could be very useful in developing resilience to climate change.

Methods
This paper utilised five different RCMs, run with different physical configurations and/or combinations with driving global models, for a total of 12 high-resolution simulations over the Alpine domain. Their ensemble was compared with one of lower resolution simulations that covered the whole Europe and served to create their boundary conditions. The high-resolution simulations were run with a 2–3 km grid size, high enough to allow the models to simulate deep convection without any use of dedicated approximations, which are needed for the low-resolution simulations (12–25 km). These simulations covered three time periods. 1996-2005 and 2090-2099 were used in this study, while 2041-2050 simulations are still underway. The future projections used a high-emission scenario of future climate change (RCP8.5). Datasets of observed rainfall were used to assess the performance of the present-climate simulations.

Policy relevance
These results demonstrate the improvements that the use of RCMs at convection-permitting scale in a model-ensemble approach can bring to our predictions of precipitation in areas with complex topography, which also allow for a better representation of the most extreme rainfall events. Improving our predictions with such an approach will allow communities and policymakers to make better-informed decisions around adapting to and mitigating the impact of this important aspect of climate change.