Constant climate simulations and continuous simulations to model extreme events
(1) Dorothée Kapsambelis - Modelling, R&D Department, CCR
(2) David Moncoulon - Modelling, R&D Department, CCR
(3) Martine Veysseire - Direction des Services Météorologiques, Météo-France
INTRODUCTION
Recent years have been marked by major climate-related events (flooding of the Seine and excess water in 2016, Hurricane Irma in 2017, droughts in 2018 and 2022) and France has recorded significant insured economic damage and agricultural losses. This series of events is a reminder of the French territory’s exposure and our society’s vulnerability to natural risks. To adapt public risk management policies, extreme weather events need to be quantified over the long term. In this context, to assess the risk exposure of our territories in the future, by integrating the impact of climate change, climate experts have developed numerous global and regional climate models. These models are built on the laws of thermodynamics and the concepts of fluid mechanics. They are complemented by complex parameterizations that allow certain climate phenomena to be represented. For a specific modelling on French mainland territory, the DRIAS database refers to the climate parameters simulating climate change from the EURO-CORDEX models. The DRIAS database has the advantage of having the data debiased on the SAFRAN re-analysis grid.
This raises the question of which model or combination of models to use, to make the results more robust. At CCR, the choice was made to use the ARPEGE-Climat model and to carry out constant climate modelling. The point of this methodology is to simulate a range of possible climate-related events for the same target year, which have not necessarily occurred. For the two target years 2000 and 2050, 400 simulations are produced and should be interpreted as possible outcomes of the year for which the probability of occurrence is calculated. This innovative methodology allows for a specific analysis of extreme events, which are rare by definition. However, the choice of using a single climate model is debatable. This article therefore proposes to compare the results obtained with the constant climate simulations of ARPEGE-Climat and those obtained with the continuous climate trajectory simulations of five EURO-CORDEX models. The results are presented on drought, and the climate indicator used to represent it is the agroclimatic index developed by CCR, which represents a cumulative ten-year water balance anomaly based on the difference between precipitation and potential evapotranspiration [1]. To represent future climate, the IPCC RCP 8.5 scenario is studied.
METHODOLOGY
Multi-model data used and index calculation
The data used are from the DRIAS database: regional EURO-CORDEX modelling and downscaling for France from SAFRAN data at 8 km. Five models were used, in addition to the CNRM- CM5 model. These include the following models:
- IPSL-CM5A developed by the Institut Pierre Simon Laplace (ISPL), couples the LDMZ atmosphere model, the OPA ocean model, the LIM ice model and the ORCHIDEE land surface model [2].
- CNRM-CM5, developed by the Centre national de recherches météorologiques (CNRM). It consists of several models. It thus integrates the ARPEGE-Climat model for the atmosphere, the NEMO ocean model, the GEMATO sea ice model, the SURFEX ocean-atmosphere surface and flow model and the TRIP model to simulate the transport of freshwater from rivers to oceans [3].
- NCC, developed by the Norwegian Climate Centre and the University Corporation for Atmospheric Research in Norway [4]. It integrates an ocean model, an atmospheric model, an ice model, a carbon cycle module and is used for climate forecasting.
- MPI, developed by Max-Planck-Institüt für Meteorologie. It couples an ocean, atmosphere and land surface model [5].
- MOHC, developed by the Met Office Hadley Centre, incorporates an ocean model, an atmosphere model and a land surface model [6].
- A final simulation integrating the values of all these models is also carried out and is called “multi-model”.
These models do not have several representations of a target year, they are simulations of forced climate trajectories according to different IPCC RCP scenarios. There is therefore only one representation of the year 2000 and one representation of the year 2050. Thus, to represent Climate 2000, the agroclimatic index is calculated over 1985-2005, which is the reference period for these models. To represent Climate 2050, the agroclimatic index is calculated over the 2040-2060 period (according to IPCC scenario 8.5). This gives a panel of 21 events over the two periods. The index is calculated at the scale of the 8km x 8km grid in Metropolitan France and its annual average value is then calculated for all of France, as well as the ten-year value of the index for all of France under both climates.
Figure 1 - Distribution of average annual values of the agroclimatic index for Metropolitan France on the ARPEGE-Climat model (box plot) and values of the agroclimatic index for the 5 EURO-CORDEX models (coloured dots) between Climate 2000 and Climate 2050.
Calculation of the agroclimatic index on constant climate simulations
To compare the results obtained from the continuous and constant climate simulations, it was necessary to adapt the methodology to compute the agroclimatic index so that it would be computed over 21 years with constant climate. The agroclimatic index values were therefore calculated over 21 successive years of constant climate simulations with the ARPEGE-Climat model. Thus, 100 draws of nationwide index values are obtained. The distribution of the average and tenyear values of the agroclimatic index of the 100 constant climate runs is studied and compared with the results of the index values on the continuous path models.
RESULTS
Figure 1 shows the comparison of the average annual values of the agroclimatic index over Metropolitan France for Climate 2000 and Climate 2050. This distribution of 100 constant-climate draws contains the set of average annual values for each of the continuous simulations. All models show a generalised drying out with a decrease in the ten-year water balance anomaly. However, the extreme values of the index are better represented with constant-climate simulations due to the large number of simulations over the target year. Constant climate simulations are therefore more suitable for studies on extreme events. The approach for this analysis is a national average of index values. It may conceal greater variations on a finer geographical scale. This is the subject of ongoing work.
# extreme climate events
# constant climate simulations
# continuous simulations
Figure 2 shows the evolution of the average annual values of the index over Metropolitan France between current and future climates. Simulation average in constant climate is higher than the evolution obtained for each of the EURO-CORDEX models.
Figure 3 shows the comparison of the ten-year values of the agroclimatic index over Metropolitan France for the current and future climates. The distribution of the index values calculated with a constant climate on the ARPEGE-Climat model contains all the ten-year values of the other models, which validates the robustness of the constant climate simulation for the consideration of extreme events.
In continuous modelling, only 21 representations of the target year are possible. Thus, the ten-year value of the agroclimatic index is only calculated over 2 values, which leads to many uncertainties. Moreover, one of the models (MOHC model) shows an opposite trend to the other models, implying that ten-year droughts would become less intense by 2050. These results seem unlikely in view of the numerous studies carried out and published, and the trend observed by the other models [7].
The changes in ten-year droughts in Metropolitan France on EURO-CORDEX simulations and on 100 ARPEGE-Climat runs with constant climate are presented in Figure 4. It appears that, depending on the model, variations are very different and not homogeneous. Nevertheless, the distribution of the changes in the agroclimatic index values of the 100 ARPEGE-Climat simulation runs contains all the developments of the models, which shows the robustness of constant-climate simulations for the simulation of extreme events in the future.
Finally, because of the significant variations in the index values between the models, and particularly on the ten-year values, by coupling the different models to make a multi-model, and by calculating extreme values in a statistical way (10-90 quantiles), a 90th percentile is calculated which corresponds to the value generated by one of the models used and a 10th percentile which corresponds to the value generated by another EUROCORDEX model. Thus, a bias is generated which is added to the biases of each of the models and which results in the simulation of a non-homogeneous distribution of climatic events which is not necessarily realistic.
Figure 2 - Change (%) in the agroclimatic index values between the future climate (RCP 8.5) and the current climate calculated on the ARPEGE-Climat model simulations of Météo-France (100 runs - box plot) and on the EURO-CORDEX models (one dot per model).
Figure 3 - Distribution of ten-year values of the agroclimatic index for Metropolitan France on the ARPEGE-Climat model (box plot) and values of the agroclimatic index for the 5 EURO-CORDEX models (coloured dots) between Climate 2000 and Climate 2050
Figure 4 - Change (%) in agroclimatic index values between future climate (RCP 8.5) and the current climate calculated on the simulations of the ARPEGE-Climat model of Météo-France (100 runs - box plot) and on the EURO-CORDEX models (one dot per model).
THE PARTNERS
Météo-France, as the French national meteorological and climatological service, carries out work, studies and research on climate and its future development. As such, Météo-France is an essential source of information and expertise for the proper performance of CCR missions. Météo-France and CCR have been in partnership since 2013.
CONCLUSION
This paper highlights a methodology to compare the constant climate simulations of the ARPEGE-Climat model and the climate trajectories simulated by the EURO-CORDEX models on the 2050 climate using the IPCC RCP 8.5 scenario. This study highlights the need to have a large panel of climatic events to carry out a specific study on extreme events. To combine the advantages of having several models and the need, demonstrated here, for a large number of simulations of the same target year, it will be essential in the future to combine these two approaches. However, to date, the computation time required, and the volume of output data limit the operational capacity to perform this type of simulation. But, in view of the permanent increase in computing capacity, this option is still possible in the future./
REFERENCES
1. Kapsambelis, Dorothée, David Moncoulon, and Jean Cordier. 2019. “An Innovative Damage Model for Crop Insurance, Combining Two Hazards into a Single Climatic Index” Climate (7) (11): 125. https://doi.org/10.3390/ cli7110125.
2. Marti, Olivier, Pascale Braconnot, Jacques Bellier, Rachid Benshila, Sandrine Bony, Patrick Brockmann, Patricia Cadule, et al. 2006. “The new IPSL climate system model: IPSL-CM4”. Other Notes scientifiques du Pôle de modélisation du climat. 26. IPSL. https:// hal. archivesouvertes.fr/hal03319443.
3. Voldoire, Aurore, Emilia Sanchez-Gomez, David Salas y Mélia, B. Decharme, Christophe Cassou, Stéphane Sénési, Sophie Valcke, et al. 2013. “The CNRM-CM5.1 Global Climate Model: Description and Basic Evaluation”. Climate Dynamics 40 (May): 2091. https://doi. org/10.1007/S00382-011- 1259-Y.
4. Bentsen, M., I. Bethke, J. B. Debernard, T. Iversen, A. Kirkevåg, Ø Seland, H. Drange, et al. 2013. “The Norwegian Earth System Model, NorESM1-M – Part 1: Description and Basic Evaluation of the Physical Climate”. Geoscientific Model Development 6 (3): 687-720. https://doi.org/10.5194/gmd6-687-2013
5. Gutjahr, Oliver, Dian Putrasahan, Katja Lohmann, Johann H. Jungclaus, Jin-Song von Storch, Nils Brüggemann, Helmuth Haak, and Achim Stössel. 2018. “Max Planck Institute Earth System Model (MPI-ESM1.2) for High-Resolution Model Intercomparison Project (HighResMIP)”. Preprint. Climate and Earth System Modeling. https://doi. org/10.5194/gmd-2018-286.
6. Bellouin, N., W. J. Collins, I. D. Culverwell, P. R. Halloran, S. C. Hardiman, T. J. Hinton, C. D. Jones, R. E. McDonald, A. J. McLaren, and F. M. O’Connor. 2011. “The HadGEM2 family of met office unified model climate configurations”. Geoscientific Model Development 4 (3): 723-57.
7. Soubeyroux, Jean-Michel, Sébastien Bernus, Lola Corre, Agathe Drouin, Brigitte Dubuisson, Pierre Etchevers, Viviane Gouget, et al. 2020. “Les nouvelles projections climatiques de référence DRIAS 2020 pour la métropole”, Météo-France. 98.
CITATION
Kapsambelis D. et al, Comparison of constant climate simulations and continuous simulations to model extreme drought events. In CCR 2022 Scientific Report; CCR, Paris, France, 2022, pp. 34-37