Periodic Reporting for period 1 - ANDANTE (AttributioN of DynAmic and thermodyNamic components in exTreme weather and climate Events)
Période du rapport: 2020-03-01 au 2022-02-28
Extreme weather and climate events, such as heat waves, droughts, and their combinations, are intrinsic to the climate system, and they can have substantial environmental and socio-economic impacts. Every extreme event is the result of a combination of external drivers, natural and anthropogenic, and internal variability. Risk-based or probabilistic extreme event attribution assesses to what extent anthropogenic drivers modify the probability and magnitude, and hence the risk of an extreme event or a class of events to understand regional impacts of climate change. Surface conditions depend on the patterns of atmospheric circulation. Thus, in a specific region human-induced thermodynamic influence can be amplified or counteracted by human-induced change in the atmospheric circulation. The main goal of project ANDANTE is to explore both human-induced dynamic (i.e. circulation/flow) and thermodynamic contributions to the risk of selected extreme events in Europe and Africa. The project makes important contributions to the field of extreme event attribution. Anthropogenically driven thermodynamic aspects in analysed extreme events can be only partially offset with changes in dynamic/circulation elements due such anthropogenic forcings. The multi-model climate simulations and multi-member reanalysis products are used to make robust multi-method estimates of the univariate and multivariate (i.e. multi-variable) risk indicators. The risk assessment of selected extreme events can be useful to a wide spectrum of stakeholders interested in climate change impacts.
The primary research focus was on European summer heatwaves and droughts in 2018, 2021 and 2022. Also, particular attention was given to 2015-2016 and 2019-2021 droughts in Africa. The work performed for ANDANTE followed the planned project structure consisting of three work packages (WP):
WP1 Production of very large ensembles of model simulations
We used HadAM3P/RM3P and HadAM4 simulations for large ensembles produced through the climateprediction.net based in Oxford, and combined them with large ensembles of CMIP5 and CMIP6 simulations for multi-model analysis and attribution of selected extreme events. The large ensembles of climate simulations helped resolving tails of distributions, i.e. extreme events and the associated circulation as we performed extreme event attribution using flow-conditional probabilities.
WP2 Process-based evaluation of models and bias correction
This WP involved multi-model evaluation of not only mean state, trend, variability, and extremes, but it also addressed the question of how well key physical processes are represented in historical period simulations to assist with model selection. Besides several observational datasets of temperature and precipitation (GHCN, CRU-TS, GPCC, E-OBS, etc.) we also used state-of-the-art reanalyses (e.g. ECMWF ERA5) as observation-based references. We explored the structure of biases (deviations of model state and processes from observations) to investigate sources of biases that were subsequently corrected/adjusted (as much as possible) with two novel bias correction methods. To study dynamics and representation of weather regimes and their relation to extremes in Europe we were examining their patterns and time evolution using unsupervised machine learning (ML) methods (firstly k-means clustering). In this process, we analysed various variables that capture key aspects of atmospheric dynamics (e.g. sea level pressure, 500 hPa geopotential heights, etc.). In Africa we examined the representation of continent-scale climate modes of variability (in surface temperature, precipitation, etc.) and their teleconnection to the tropical Pacific Ocean by again relying on unsupervised ML techniques. Additionally, we examined the influence of ENSO on mean climate and extremes in Africa versus influence of the Indian and Atlantic basins through regression and composite analysis.
WP3 Multi-method risk-based event attribution of dynamics and thermodynamic components
We performed event attribution of extreme events of interest in Africa and Europe. To answer the question of what is the role of anthropogenic drivers in the characteristics of an extreme event or a class of events, we estimated the return times and key risk indicators with the uncertainties. More specifically, we used multi-method approach to perform unconditional and flow-conditional event attribution. We improved our understanding of human-induced dynamics, thermodynamics and cross-term, dynamics-thermodynamic interaction, contributions in the selected extremes. Distinction of event attribution of human-induced dynamic and thermodynamic components was enabled by the flow-conditional probabilities (determined by counting the occurrences among similar circulation/flow structures).
We found that human-induced thermodynamic aspects can be at the most only partially offset with changes in dynamic/circulation aspects due to anthropogenic forcing factors, while cross-term, dynamics-thermodynamic interaction plays more important role on weather timescales (i.e. days and weeks) than on monthly and longer timescales in Europe and Africa. Furthermore, human-induced thermodynamics changes of extreme events in a changing climate will only amplify later this century in most of the emission scenarios.
Dissemination and networking activities were tampered by global health conditions, but I was fortunate to be able to present results at EGU general assemblies from April 2021 to May 2023. I also gave presentations at RBI in Zagreb, Croatia (October 2022), IIASA in Laxenburg, Austria (April 2023), and ICTP in Triste, Italy (May 2023). Furthermore, I engaged with various stakeholders interested in health and socio-economic impacts of extreme events in a changing climate, so in collaboration with the UK Health Security Agency I presented my results at the Public Health Research and Science Conference in May 2021. Additionally, I started collaborating with the World Bank Group in Washington DC, USA, were I presented my results on extreme events in Africa in October 2022. No website has been developed for the project yet.
WP1 Production of very large ensembles of model simulations
We used HadAM3P/RM3P and HadAM4 simulations for large ensembles produced through the climateprediction.net based in Oxford, and combined them with large ensembles of CMIP5 and CMIP6 simulations for multi-model analysis and attribution of selected extreme events. The large ensembles of climate simulations helped resolving tails of distributions, i.e. extreme events and the associated circulation as we performed extreme event attribution using flow-conditional probabilities.
WP2 Process-based evaluation of models and bias correction
This WP involved multi-model evaluation of not only mean state, trend, variability, and extremes, but it also addressed the question of how well key physical processes are represented in historical period simulations to assist with model selection. Besides several observational datasets of temperature and precipitation (GHCN, CRU-TS, GPCC, E-OBS, etc.) we also used state-of-the-art reanalyses (e.g. ECMWF ERA5) as observation-based references. We explored the structure of biases (deviations of model state and processes from observations) to investigate sources of biases that were subsequently corrected/adjusted (as much as possible) with two novel bias correction methods. To study dynamics and representation of weather regimes and their relation to extremes in Europe we were examining their patterns and time evolution using unsupervised machine learning (ML) methods (firstly k-means clustering). In this process, we analysed various variables that capture key aspects of atmospheric dynamics (e.g. sea level pressure, 500 hPa geopotential heights, etc.). In Africa we examined the representation of continent-scale climate modes of variability (in surface temperature, precipitation, etc.) and their teleconnection to the tropical Pacific Ocean by again relying on unsupervised ML techniques. Additionally, we examined the influence of ENSO on mean climate and extremes in Africa versus influence of the Indian and Atlantic basins through regression and composite analysis.
WP3 Multi-method risk-based event attribution of dynamics and thermodynamic components
We performed event attribution of extreme events of interest in Africa and Europe. To answer the question of what is the role of anthropogenic drivers in the characteristics of an extreme event or a class of events, we estimated the return times and key risk indicators with the uncertainties. More specifically, we used multi-method approach to perform unconditional and flow-conditional event attribution. We improved our understanding of human-induced dynamics, thermodynamics and cross-term, dynamics-thermodynamic interaction, contributions in the selected extremes. Distinction of event attribution of human-induced dynamic and thermodynamic components was enabled by the flow-conditional probabilities (determined by counting the occurrences among similar circulation/flow structures).
We found that human-induced thermodynamic aspects can be at the most only partially offset with changes in dynamic/circulation aspects due to anthropogenic forcing factors, while cross-term, dynamics-thermodynamic interaction plays more important role on weather timescales (i.e. days and weeks) than on monthly and longer timescales in Europe and Africa. Furthermore, human-induced thermodynamics changes of extreme events in a changing climate will only amplify later this century in most of the emission scenarios.
Dissemination and networking activities were tampered by global health conditions, but I was fortunate to be able to present results at EGU general assemblies from April 2021 to May 2023. I also gave presentations at RBI in Zagreb, Croatia (October 2022), IIASA in Laxenburg, Austria (April 2023), and ICTP in Triste, Italy (May 2023). Furthermore, I engaged with various stakeholders interested in health and socio-economic impacts of extreme events in a changing climate, so in collaboration with the UK Health Security Agency I presented my results at the Public Health Research and Science Conference in May 2021. Additionally, I started collaborating with the World Bank Group in Washington DC, USA, were I presented my results on extreme events in Africa in October 2022. No website has been developed for the project yet.
This study responded to the World Climate Research Programme’s (WCRP) Grand Challenge on Weather and Climate Extremes by addressing the key attribution question of what are the contributors to observed extreme events and to changes in the frequency and intensity of the observed extremes? Many extreme events are closely linked to the variability of large-scale atmospheric circulation. However, the role of circulation in the change of probabilities of extreme weather and climate events has not yet been analysed in a unified framework encompassing both tropics and extratropics. The project ANDANTE made significant contributions to our goal to advance such understanding and disentangle the dynamic and thermodynamic components of the anthropogenic contribution to single-variable and compound extreme events (e.g. combinations of heat wave and drought). Furthermore, produce multi-method multi-model datasets contain climate information about a spectrum of extreme events that can provide key information for hazard and risk assessment critical for resilience strengthening and adaptation planning around the world.