Metastability, atmospheric midlatitude circulation regimes and large-scale teleconnection: a data-driven approach
Dmitry Mukhin, Roman Samoilov, Abdel Hannachi · 2024
Read Metastability, atmospheric midlatitude circulation regimes and large-scale teleconnection: a data-driven approach by Dmitry Mukhin, Roman Samoilov, Abdel Hannachi (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-06933.txt.
Introduction
Dmitry Mukhin, Roman Samoilov, and Abdel Hannachi build a data-driven method to find lasting weather regimes in the mid-latitude atmosphere. Low-frequency change in the air is non-linear and chaotic, which limits forecasts. It forms into long-lived flow patterns that recur from time to time. The patterns span a hemisphere and are known as weather regimes. To model them stays hard.
The team builds on metastability, which sees regimes as sets of states where the air lingers for a long time. Their method builds an evolution operator through a hidden Markov model of air dynamics. Graph-based splits then cut the transition matrix into weakly linked blocks. Those blocks are the sets tied to regimes. A non-linear kernel method helps make the structure workable. Better regime finds aid forecasts weeks ahead.
Catalog record
| CREATOR | Dmitry Mukhin, Roman Samoilov, Abdel Hannachi |
|---|---|
| YEAR | 2024 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 56 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2412-06933
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Metastability, atmospheric midlatitude circulation regimes and large-scale teleconnection: a data-driven approach Dmitry Mukhin, Roman Samoilov, Abdel Hannachi Abstract The low-frequency variability of the mid-latitude atmosphere involves complex nonlinear and chaotic dynamical processes posing predictability challenges. It is characterized by sporadically recurring, often long-lived patterns of atmospheric circulation of hemispheric scale known as weather regimes. The evolution of these circulation regimes in addition to their link to large-scale teleconnections can help extend the limits of atmospheric predictability. They also play a key role in sub- and inter-seasonal weather forecasting. Their identification and modeling remains an issue, however, due to their intricacy, including a clear conceptual picture. In recent years, the concept of metastability has been developed to explain regimes formation. This suggests an interpretation of circulation regimes as communities of states in which the atmospheric system remains in their neighborhood for abnormally longer than typical baroclinic timescales. Here we develop a new and effective method to identify such communities by constructing and analyzing an operator of the system’s evolution via hidden Markov model (HMM). The method makes use of graph theory and is based on probabilistic approach to partition the HMM transition matrix into weakly interacting blocks – communities of hidden states – associated with regimes. The approach involves nonlinear kernel principal component mapping to consistently embed the system state space for HMM building. Application to northern winter hemisphere using geopotential heights from reanalysis yields four persistent and recurrent circulation regimes. Statistical and dynamical characteristics of these circulation regimes and surface impacts are discussed. In particular, unexpected high correlations are obtained with EL-Nino Southern Oscillation and Pacific decadal oscillation with lead times of up to one year. 1 Introduction Humans have, since the dawn of civilization from the Babylonian to present day, sought to predict the weather far in the future. The theory of modern mathematical models of nonlinear dynamical systems show that weather cannot be forecast accurately beyond 1-2 weeks ahead. However, this is not the full story. In fact, the atmosphere exhibits complex variation on a wide spectrum of spatio-temporal scales ranging from synoptic to decadal and longer time scales originating from internal nonlinear dynamics and also from the boundary conditions such as sea surface temperature (SST) variation. These sources can contribute to extending the chaotic predictability limits to include subseasonal-to-seasonal (S2S) timescales. Low-frequency variability (LFV) of the atmosphere, spanning spatio-temporal scales greater than typical synoptic and baroclinic timescales is poorly understood and is difficult to predict. Extratropical LFV, in particular, involves recurring and persistent large-, e.g. hemispheric-scale structures. These structures of atmospheric circulation are often referred to as teleconnection patterns or regimes and are known to have profound impact on weather predictability on intraseasonal and longer timescales [ 1 , 2 ] . With such scales, the ensuing circulation anomalies strongly modulate the tracks and intensity of synoptic disturbances in the atmosphere, operating like waveguides, and hence determine the long-term behavior of weather over large areas. The modeling and analysis of the mid-latitude climate is closely related to recurring circulation regimes. Both the identification and the understanding of their dynamics continue to be a debatable issue due to the complexity and nonlinearity of the problem beside the unsettled concept of regimes. Quasi-geostrophic theory of large-scale atmospheric circulation shows that LFV is essentially the result of nonlinear dynamics [ 3 ] . Experimental analyses based on quasi-geostrophic barotropic and baroclinic models reveal the high-dimensional nature and chaotic dynamics of the LFV atmospheric flow [ 4 ] . There is suggestion in the literature [ 5 ] that in an infinite set of saddle-invariant manifolds of a chaotic attractor, the most stable of those manifolds form the skeleton of the dynamics, and sporadic adherence of the system state in their neighborhood can cause regime persistence. This nonlinear perspective leads to the understanding of regimes as metastable regions in the system’s phase space where the system slows down. To consistently adopt this paradigm of persistent regimes, an extra requirement on regime detection methodology has to be added, namely, it should account for temporal relationships of observed states rather than just their similarity in a selected projection of phase space. A number of existing methods [ 6 ] for identifying recurrent patterns, e.g. based on PDF decomposition or partitioning networks of similarity, do not satisfy this requirement. An appropriate and elegant way to address this issue is to construct and analyze an evolution operator of the system using hidden Markov model (HMM) [ 7 , 8 ] . A Conventional Markov chain is a stochastic model based on ’observed’ states of the chain. A HMM [ 7 , 9 ] is an extension of the Markov chain by including other ( hidden ) states, which generate the observed sequence. Because of the finite and short-memory representation of the system, HMM is a consistent method for detecting metastability as well as identifying the basins of metastable states. Essentially, with HMM nonlinear effects, like metastability and trace of multiple equilibria, can be captured by linear transfer operator pretty much like dynamic mode decomposition (DMD) [ 10 ] . HMM has been used to study atmospheric regimes via spectral analysis of the Markovian operator [ 8 , 11 ] . Inspired by ideas from [ 8 ] , we elaborate a new methodology for constructing and analyzing HMM aimed at detecting persistent and recurrent circulation regimes determined as metastable communities of hidden states. We derive and use a method of partitioning the HMM transition probability matrix into blocks, providing significantly low probabilities of transitions between them. The method is based on maximizing a cost-function for a given partition and has a clear probabilistic meaning. In fact, the approach finds that partition with abnormally high mean probability that the state will remain within its community, i.e. persistent. We show that this optimization problem effectively reduces to the Newman iterative algorithm, widely used in graph theory for graph decomposition [ 12 ] . We pay special attention to the significance test, which rejects the null hypothesis that the detected communities could have been obtained by chance from a sample of a randomized process with similar spectral properties. Another key part of the methodology is the choice of an embedding space, i.e. a set of dynamic variables for dimension reduction and model construction. This is of paramount importance because it determines the projection of the phase space in which the model operates. The dimension of this space is desired to be as low as possible to ensure stable results during the HMM learning process, but at the same time it should be informative enough for regime detection and identification. Kernel principal component analysis (KPCA) [ 13 , 6 ] is considered here, which belongs to the class of nonlinear data transformation, and has been shown to enable a reliable and robust separation of states based on similarity measure in a space of just few leading components. Recent works [ 14 , 15 ] used it to study recurrent patterns of the mid-latitude atmospheric circulation. In this manuscript, KPCA serves as a prepossessing step for the HMM to get both recurrent [ 1 ] and persistent circulation flows. The paper is organized as follows. In Section 2 we describe