WCRP Workshop on
Improving climate models and projections using observation
MIT, Cambridge MA, USA June 12 – 14, 2023
A three-day workshop is planned to take place between June 12 and June 14, 2023. The workshop will focus on data assimilation and artificial intelligence/machine learning (AI/ML) approaches in the context of Earth system reanalysis and climate model improvement. The workshop will be hosted by MIT in Cambridge, Massachusetts, and will be open but limited in number of participants.
Recent model developments have led to substantial improvements in the simulation of the Earth system and its subcomponents. Further improvement is expected from enhanced representation of physical processes associated with increased model resolution. However, climate models will always be prone to some biases, and are expected to produce climate features and variability which will differ from the real world.
To further improve the simulation capability of climate model with respect the real world as well as their predictive capabilities, Climate Data Assimilation attempts to optimally integrate climate models with climate observations to produce a balanced and coherent present day climate estimate in support of initialized predictions. Beyond this specific goal, climate data assimilation is used to provide a best possible description of the past climate, thereby optimizing the use and interpretation of limited climate observations. Previously, multiple atmospheric reanalyses and more recently also ocean or coupled ocean/sea ice reanalyses are prominent examples of such efforts. The field is now targeting a more comprehensive approach towards coupled Earth System reanalysis, e.g., to improve the description of changes of energy, water and carbon in the system and, thus, to allow for more consistent analyses of climate variability and change.
Beyond this more traditional reanalysis application, in the future arguably the most important aspect of climate data assimilation and ML might become optimizing model parameters to mitigate model biases and thereby improving the model’s skill in simulating the observed climate and enhancing model predictability. This will allow for further improvements in understanding climate variability and feedback mechanisms as well as in predictability on seasonal-interannual to decadal scales and beyond. Accomplishing physical and dynamical consistency in the assimilation and estimation process is key for reaching its goals. Embedding this approach in an ensemble setting will allow to account for natural variability and chaotic aspects and provide useful uncertainty estimates while observing and simulating the Earth system.
The workshop will be organized 25 years after the inception of the ECCO assimilation effort at MIT. It will consist only of a plenary session featuring keynote talks on all aspects of future Earth system reanalysis and climate model improvements and related aspects of coupled data assimilation and machine learning techniques. As outcome a white paper is anticipated that will guide WCRPs efforts on coupled data assimilation and that can be used as input into WCRPs Open Science Conference, held in Kigali during Oct 23 – 27, 2023.
The local hosts will be Raffael Ferrari, Chris Hill and Carl Wunsch. The science organizing team is being led by Detlef Stammer.