NP4.4Linking Models and Data: Prediction, Verification, and Intercomparison
|Convener: Valerio Lucarini | Co-Conveners: Alexander Feigin , Jürgen Kurths|
The session is concerned with the development and application of empirical approaches to modeling and prediction of natural processes and with the definition of smart methods for using models and interpreting their results. The Earth is an extremely complex system with a large number of feedbacks and nonlinearities, whose behavior represents a broad spectrum of temporal scales. Predictive understanding of Earth system´s variability is a key goal of geophysical research. It is essential to emphasize that in spite of progress in first principal modeling of natural phenomena, predictive abilities of these models are fundamentally restricted by the Earth system complexity: inherent sensitivity to initial data, model parameters, and the complex nonlinearities and feedbacks involved, present a substantial and persistent obstacle to developing robust prediction methods based on first principal model usage.
At the same time a shortness of greater part of natural time series restricts seriously capabilities of purely empirical modeling.
Because of this the main intention of the Session is to bring together specialists who use empirical models and who use first principles knowledge, from mathematical statistics and statistical physics as well as from complex systems science to present and combine different modern approaches for a better description of processes in the Earth system and to study their predictability.
Specific topics include but are not limited to the following:
• Data-driven (empirical) approaches to modeling climate variability: stochastic empirical models, qualitative and quantitative forecast, optimal model selection, etc.;
• Cooperation of empirical and first-principal modeling; improving empirical models by analysis of time series generated by first principal models, etc.;
• Use of dynamical systems and statistical mechanical methods based on unstable periodic orbits, covariant Lyapunov vectors for studying climate dynamics and constructing reduced order models;
• Relating forced and free fluctuations by use of methods based on fluctuation dissipation theorem and response theory;
• Efficient expansions of space-distributed data: from EOFs and its extensions to spatio-temporal and/or nonlinear decompositions;
• Complex network approach to investigation of teleconnection patterns and their stability studying;
• Empirical modeling of extreme events and studying of their predictability;
• Model inter comparison in terms of prediction on different time scales