SAMA : Statistics for Analysis, Modeling and Assimilation
SAMA develops various tools in order to solve problems IPSL has to face in its activities. Here are some particular examples :
- Inverse problems : Here is a typical example : radiometers on board satellites measure infra-red radiation emitted from Earth to outer space. This radiation mainly depends on the vertical temperature and humidity profiles in atmosphere thickness. How to reconstruct these profiles using the measurements from space based radiometers ?
- Data assimilation : born from the need of providing suitable initial conditions for numerical weather forecasts. The goal of data assimilation is to combine the available observations (nowadays about 20 millions per day) in order to define the flow state the best we can. The problem is even harder than it seems because observations are distributed in time, and the flow evolution must be directly taken into account within the assimilation process.
- Extreme events statistics (floods, storms, heat waves, ...) : What are the statistical recurrence of these events ? How are these laws are modulated by climate change ? How the knowledge of these laws can help to predict these events ?
The different problems are not separated at all, but are more or less linked. For example, data assimilation is itself a certain kind of inverse problem.
- Neural networks: a powerful tool to solve non-linear inverse problems.
- Different kinds of data assimilation algorithms (Kalman filter, variational algorithms).
- Adjoint models: the basic tool of variational data assimilation algorithms. They are providing an efficient solution to many optimization and sensitivity problems.
SAMA activity is centered on seminars, workshops, work teams and software development.