The UMR CECI (previoulsy URA SUC 1875 from 1999 to December 2015) is a research unit that associates CERFACS (Centre Européen de Recherche et Formation Avancée en Calcul Scientifique, located in Toulouse, France) and CNRS (Centre National de la Recherche Scientifique). CERFACS hosts four interdisciplinary teams, both for research and advanced training, that comprise physicists, applied mathematicians, numerical analysts, and software engineers. Among the four CERFACS research teams (Aviation and Environment (AE), Climate Modelling and Global Change (GLOBC), Computational Fluid Dynamics (CFD), Parallel Algorithms (ALGO), only the first two (AE and GLOBC) are part of the CECI. However, the CECI teams have strong cross-disciplinary interactions with the other CERFACS teams.
The UMR activities are structured by scientific themes rather than by scientific teams as to mimic the CERFACS research strategy. There are 2 core themes and 3 transverse themes :
- CLIM (Climate Variability and Change)
While the CECI core historical climate research activities, namely Climate Variability, Predictability and Climate Change, were still very active (with in particular, major contributions to the Coupled Model Intercomparison Exercise version 5, CMIP5, of the World Climate Research Program-WCRP, the continuation of detection and attribution studies focusing on climate variables different from air temperature, e.g., tropical sea surface salinity, and studies devoted to the past and future changes of the atmospheric annular modes), new research axes have been started within the Climate Variability and Change theme. In particular, the themes of air-sea interaction at small spatial scale and chaotic ocean low-frequency variability have been started as well as that of the influence of high horizontal model resolution on climate variability and predictability. A strong activity regarding the understanding and subsequent reduction of coupled climate model biases has also started with a primary focus on the Tropical Atlantic. Understanding and predicting impacts of climate change remained one of our main activities with a focus on France projections of temperature, precipitation and stream flows changes as well as those of extreme temperature events and heat waves. In addition to the assessment of global climate model results, a variety of downscaling methods have been developed and/or used to refine the regional or local projections with an appropriate uncertainty range. CECI has also pursued research to assess the potential benefits of using process-based metrics to reduce the structural or model uncertainty in European climate projections for the 21st century. These activities have enabled a large number of new fruitful collaborations with other national (such as LGGE and the DRAKKAR group) and international (Institut Català de Ciències del Clima IC3, Universities of Reading and Toronto) teams and a strengthening of the existing ones such as those with our long-term partners CNRM/GAME and IPSL.
- AE (Aviation and Environment)
The theme of Aviation and Environment, which was started in 2004 within CERFACS and CECI, has seen the continuation of its fast growing development for the past five years. The main objective of this research activity is to understand and simulate chemical and radiative atmospheric impacts of aviation at the various scales from the aircraft near field to the global atmosphere. Major progress has been achieved in the themes of near-field wake chemistry and contrail physics and dynamics such as a pioneering study of the first simulation of the diffusion regime with an early transition of a contrail into young cirrus that accounts for the impact of background atmospheric turbulence and with radiative transfer. A new activity concerned the evaluation of the impact on the stratospheric composition of solid rocket boosters that eject high quantities of gas and particles. Finally, there was also a substantial model numerical scheme development activity such as that of a solver for atmospheric chemistry models.
- CPL (Coupling Tools)
The development, maintenance and transfer of coupling tools towards the scientific national and international communities have always been at the core of the CECI activities since the pioneering work on the OASIS and OpenPALM couplers in the early and late nineties, respectively. These activities have been intensively pursued and developed for the past five years with major milestones such as the development of a new fully parallel and scalable version of the OASIS coupler, OASIS3-MCT, and the extension and use of OpenPALM to a large variety of multi-physics coupling and assimilation suites that benefited from the inclusion of a new parallel remapping library for coupling fields. It is also worth noting the strong and continuous implication of CECI engineers and researchers in regular and sustained advanced training on these two couplers offered to an ever-growing list of users with very different backgrounds.
- ASSIM (Data Assimilation)
While the ASSIM perimeter initially concerned only oceanography, other domains such as atmospheric chemistry, hydrology and hydraulics and other physics applications such as wild fires have now been tackled. It is worth noting that important advances in theoretical and algorithmic aspects of data assimilation have been pioneered by CECI such as a new method for modelling correlation functions based on an implicitly formulated diffusion equation or the design of a new and very efficient minimization algorithm formulated in observational rather than model-space. As with atmospheric chemistry, major developments have been performed in order to allow assimilation of a variety of ozone or other chemical species satellite measurements. A new research axe on regional air-quality has also been initiated in the framework of the MACC suite of European funded projects to fulfill the needs of European air-quality monitoring. With regard to hydrology and hydraulics, CECI has been investigating and developing the use of variational and filtering data assimilation techniques to improve mono and multidimensional hydrodynamics modelling for rivers and lakes with real and operational applications such as flood forecasting. With regard to wild fires, CECI has developed FireFly data driven simulator constrained by assimilation of airborne observations of the fire front position with a Kalman Filter algorithm. The SUC has also been strongly active in the development, maintenance and dissemination of full assimilation systems such as NEMOVAR (for oceanographic applications) or VALENTINA (for atmospheric chemistry) in close cooperative relationships with other research centres (ECMWF, CNRM/GAME) or consortium (NEMO). It is worth recalling that the data products derived from these systems have wide-ranging applications for a large french and international community.
- HPC (High-Performance Computing, Big Data and Uncertainty Quantification)
The last transverse and emerging theme is related to HPC, Big Data and uncertainty quantification (UQ). The need to adequately and explicitly simulate small-scale processes is driving a strong increase in model spatial resolution for many of our applications. The probabilistic nature of climate prediction, and consequently the need for a large ensemble approach, is also contributing to an increasing need of massively parallel computing resources. No need to have a crystal bowl to predict that this trend is going to continue at an even faster pace than now. Thus we have begun to investigate several issues that will need to be solved to have efficient and reliable very high-resolution climate models on the most recent (and future!) HPC system architectures (with thousands and/or several ten thousand of processors, and even ultimately millions). In a near future, the road towards Exascale computing will encompass even more radical changes in the physical codes we are using every day, on both algorithmic and numerical aspects. Our gut feeling is that we already need to anticipate and get prepared for this computing revolution. Another activity in this theme is the Big Data activity. Given that the amount of data generated by climate modelling exceeds, and by far, that of all other disciplines, designing fast and robust diagnostic tools to analyse these very large data sets is becoming a priority as well as the development of compression algorithms, metadata protocols and web-based data portals for climate data sets. The last and emerging aspect is that of uncertainty quantification. While UQ has always been, to some extent, present in climate related studies (with uncertainties related to model parameters, initial and boundary conditions), CECI has recently pioneered the use of a surrogate model to reduce the cost of an ensemble-based data assimilation algorithm using a polynomial chaos decomposition technique with real application in forecasting wildfire spread and hydraulics. Extension of this kind of approach to other domains (such as hydraulics and hydrology) is underway.