The seminar itself is organised in two parts, one after the other, with a short break in between: Part I (30 min) - Break (15 min) - Part II (30 min)
Speakers: Stephane Vannitsem (Part I) and Jonathan Demaeyer (Part II)
Affiliation: Royal Meteorological Institute of Belgium
Abstract
The dynamics of the atmosphere (and of the climate system) as described by the conservation laws of fluid mechanics are known to display the property of sensitivity to initial conditions. This property has considerable impact on our abilities to make predictions at short and medium-range weather time scale, and at seasonal-to-decadal forecast ranges. This implies that weather forecasts or climate predictions are in essence probabilistic problems that should be tackled with appropriate tools. Since the nineties, considerable efforts have been addressed to develop such an information, often through ensemble forecasts based on multiple model integrations starting from different initial conditions. Such an approach is now well settled for weather forecasts, but still in its infancy for seasonal to decadal forecasts.
The first part of the seminar is devoted to introduce the problem of predictability of atmospheric flows, and the impact of both initial condition and boundary condition errors. Since an extensive predictability analysis in comprehensive climate models is impossible in practice, the investigation is performed in a reduced-order coupled extratropical ocean-atmosphere model, introduced some time ago to understand the emergence of low-frequency variability of the atmosphere (Vannitsem et al, 2015) and the impact of Tropical teleconnections to extratropical dynamics (Vannitsem et al, 2021). The analysis reveals (i) the strong impact of the emergence of low-frequency variability on the long-term potential predictability of the atmosphere; and (ii) that the potential of teleconnections in improving the quality of climate predictions could only be realized provided the Tropical dynamics is accurately forecasted.
In the second part of the seminar, a method to construct initial conditions which produce reliable ensemble forecasts at the particularly challenging subseasonal-to-seasonal forecast range is presented. These initial conditions are obtained by perturbing the analysis with random perturbations projected onto the Koopman and Perron-Frobenius operators’ eigenfunctions, which describe the time-evolution of observables and probability distributions of the system dynamics, respectively. In practice, the perturbations are projected on approximations of these eigenfunctions provided by the Dynamic Mode Decomposition data-driven algorithm. The effectiveness of this approach is illustrated in the framework of the above coupled ocean-atmosphere model, and by comparing it to other well-known ensemble initialization methods based on the Empirical Orthogonal Functions of the model trajectory and on the backward and covariant Lyapunov vectors of the model dynamics. Explanations are provided on why this method is effective and could be applied to operational forecasting models.
References:
- Demaeyer, J., Penny, S. G., & Vannitsem, S. Identifying efficient ensemble perturbations for initializing subseasonal-to-seasonal prediction. Journal of Advances in Modeling Earth Systems, 14, e2021MS002828, 2022. https://doi.org/10.1029/2021MS002828
- Vannitsem, S., J. Demaeyer, L. De Cruz, M Ghil, Low-frequency variability and heat transport in a low- order nonlinear coupled ocean-atmosphere model. Physica D, 309, 71-85, 2015. https://doi.org/10.1016/j.physd.2015.07.006
- Vannitsem, S., Demaeyer, J., & Ghil, M. Extratropical low-frequency variability with ENSO forcing: A reduced-order coupled model study. Journal of Advances in Modeling Earth Systems, 13, e2021MS002530, 2021. https://doi.org/10.1029/2021MS002530.
MS Teams link
https://teams.microsoft.com/l/meetup-join/19%3apaSauAIA2ripzigZgVR36alVx9OfBytmUdEXU1wnCfo1%40thread.tacv2/1683708652907?context=%7b%22Tid%22%3a%22e99647dc-1b08-454a-bf8c-699181b389ab%22%2c%22Oid%22%3a%22cd9b84dd-0953-4dca-9967-f30a1a173f7f%22%7d