arXiv subjects

Marlene Kretschmer

Publications and source records attributed to Marlene Kretschmer.

7 recordsLinked to original sources

arXiv2026

Disentangling regional impacts of joint teleconnections using causal representation learning

Understanding teleconnections of large-scale modes of climate variability is relevant for seasonal predictability and support a dynamical understanding of climatic changes. While numerical model experiments are the most common approach for investigating counterfactual climate responses, their conclusions are subject to model biases. Data-driven approaches offer a complementary perspective. Deep learning can extract reduced-dimensional patterns but usually lacks causal interpretability, while causal methods can disentangle signals in the presence of confounding yet are typically based on simple indices. Treating dimensionality reduction and causal inference separately thereby risks losing the teleconnection signal of interest. This paper introduces DAG-VAE, a causal representation learning approach that embeds a physics-informed directed acyclic graph in the latent space of a variational autoencoder. Combining deep learning with causal inference, the method jointly learns nonlinear reduced representations of large-scale modes of variability and their causal interactions. We apply DAG-VAE to disentangle the influences of the Pacific and Indian Oceans on the short rains over the Greater Horn of Africa. Trained on seasonal hindcasts, the method identifies dynamically meaningful representations and recovers spatial response patterns consistent with SST-replacement experiments. Trained on reanalysis data, DAG-VAE identifies a different response pattern to direct influence of the tropical Pacific, highlighting potential model biases and the value of DAG-VAE as a complementary, data-driven approach for estimating spatial causal response patterns from observations. Finally, we demonstrate the ability of the method to generate data-driven counterfactuals of extreme short rain seasons, with potential applications for forecast-based early action and scenario planning.

physics.geo-ph↗

arXiv2025

Deep Learning Meets Teleconnections: Improving S2S Predictions for European Winter Weather

Predictions on subseasonal-to-seasonal (S2S) timescales--ranging from two weeks to two month--are crucial for early warning systems but remain challenging owing to chaos in the climate system. Teleconnections, such as the stratospheric polar vortex (SPV) and Madden-Julian Oscillation (MJO), offer windows of enhanced predictability, however, their complex interactions remain underutilized in operational forecasting. Here, we developed and evaluated deep learning architectures to predict North Atlantic-European (NAE) weather regimes, systematically assessing the role of remote drivers in improving S2S forecast skill of deep learning models. We implemented (1) a Long Short-term Memory (LSTM) network predicting the NAE regimes of the next six weeks based on previous regimes, (2) an Index-LSTM incorporating SPV and MJO indices, and (3) a ViT-LSTM using a Vision Transformer to directly encode stratospheric wind and tropical outgoing longwave radiation fields. These models are compared with operational hindcasts as well as other AI models. Our results show that leveraging teleconnection information enhances skill at longer lead times. Notably, the ViT-LSTM outperforms ECMWF's subseasonal hindcasts beyond week 4 by improving Scandinavian Blocking (SB) and Atlantic Ridge (AR) predictions. Analysis of high-confidence predictions reveals that NAO-, SB, and AR opportunity forecasts can be associated with SPV variability and MJO phase patterns aligning with established pathways, also indicating new patterns. Overall, our work demonstrates that encoding physically meaningful climate fields can enhance S2S prediction skill, advancing AI-driven subseasonal forecast. Moreover, the experiments highlight the potential of deep learning methods as investigative tools, providing new insights into atmospheric dynamics and predictability.

cs.LG↗

arXiv2024

Identifying probabilistic weather regimes targeted to a local-scale impact variable

Weather regimes are recurrent and persistent large-scale atmospheric circulation patterns that modulate the occurrence of local impact variables such as extreme precipitation. In their capacity as mediators between long-range teleconnections and these local extremes, they have shown potential for improving sub-seasonal forecasting as well as long-term climate projections. However, existing methods for identifying weather regimes are not designed to capture the physical processes relevant to the impact variable in question while still representing the full atmospheric phase space. This paper introduces a novel probabilistic machine learning method, RMM-VAE, for identifying weather regimes targeted to a local-scale impact variable. Based on a variational autoencoder architecture, the method combines non-linear dimensionality reduction with a prediction task and probabilistic clustering in a coherent architecture. The new method is applied to identify circulation patterns over the Mediterranean region targeted to precipitation over Morocco and compared to three existing approaches, two established linear methods and another machine learning approach. The RMM-VAE method identifies regimes that are more predictive of the target variable compared to the two linear methods, and more robust and persistent compared to the alternative machine learning method, while also improving the reconstruction of the input space. The results demonstrate the potential benefit of the new method for use in various climate applications such as sub-seasonal forecasting, while also highlighting the trade-offs involved in targeted clustering.

physics.geo-ph↗

arXiv2024

Finding the right XAI method -- A Guide for the Evaluation and Ranking of Explainable AI Methods in Climate Science

Explainable artificial intelligence (XAI) methods shed light on the predictions of machine learning algorithms. Several different approaches exist and have already been applied in climate science. However, usually missing ground truth explanations complicate their evaluation and comparison, subsequently impeding the choice of the XAI method. Therefore, in this work, we introduce XAI evaluation in the climate context and discuss different desired explanation properties, namely robustness, faithfulness, randomization, complexity, and localization. To this end, we chose previous work as a case study where the decade of annual-mean temperature maps is predicted. After training both a multi-layer perceptron (MLP) and a convolutional neural network (CNN), multiple XAI methods are applied and their skill scores in reference to a random uniform explanation are calculated for each property. Independent of the network, we find that XAI methods Integrated Gradients, layer-wise relevance propagation, and input times gradients exhibit considerable robustness, faithfulness, and complexity while sacrificing randomization performance. Sensitivity methods -- gradient, SmoothGrad, NoiseGrad, and FusionGrad, match the robustness skill but sacrifice faithfulness and complexity for randomization skill. We find architecture-dependent performance differences regarding robustness, complexity and localization skills of different XAI methods, highlighting the necessity for research task-specific evaluation. Overall, our work offers an overview of different evaluation properties in the climate science context and shows how to compare and benchmark different explanation methods, assessing their suitability based on strengths and weaknesses, for the specific research problem at hand. By that, we aim to support climate researchers in the selection of a suitable XAI method.

cs.LG↗

arXiv2023

Artificial Intelligence for Prediction of Climate Extremes: State of the art, challenges and future perspectives

Scientific and technological advances in numerical modelling have improved the quality of climate predictions over recent decades, but predictive skill remains limited in many aspects. Extreme events such as heat and cold waves, droughts, heavy rain and storms are particularly challenging to predict accurately due to their rarity and non-linear chaotic nature, and because of model limitations. However, recent studies have shown that predictive skill of extremes can be increased when using more sophisticated approaches, indicating that there might be systemic predictability that is not being leveraged. Recently, numerous studies have been devoted to the exploitation of Artificial Intelligence (AI) to study the predictability and make predictions of weather and climate. AI techniques have shown great potential to improve the prediction of extreme events and uncover their links to large-scale and local drivers. Machine and deep learning, causal discovery, explainable AI, are only some of the approaches that have been tested to both improve our understanding of the processes underlying predictability and enhance prediction skill of extreme events. Results are promising, especially for hybrid predictions that combine the AI, which can reveal and exploit unknown spatio-temporal connections from data, and climate models, that provide the theoretical foundation and interpretability of the physical world. On the other hand, challenges are multiple in many aspects, from data curation to model uncertainty and generalizability, to the reproducibility of methods and workflows. A few best practices are identified to increase trust in these novel techniques, and future perspectives are envisaged for further scientific development.

physics.ao-ph↗

arXiv2020

Reconstructing regime-dependent causal relationships from observational time series

Inferring causal relations from observational time series data is a key problem across science and engineering whenever experimental interventions are infeasible or unethical. Increasing data availability over the past decades has spurred the development of a plethora of causal discovery methods, each addressing particular challenges of this difficult task. In this paper we focus on an important challenge that is at the core of time series causal discovery: regime-dependent causal relations. Often dynamical systems feature transitions depending on some, often persistent, unobserved background regime, and different regimes may exhibit different causal relations. Here, we assume a persistent and discrete regime variable leading to a finite number of regimes within which we may assume stationary causal relations. To detect regime-dependent causal relations, we combine the conditional independence-based PCMCI method with a regime learning optimisation approach. PCMCI allows for linear and nonlinear, high-dimensional time series causal discovery. Our method, Regime-PCMCI, is evaluated on a number of numerical experiments demonstrating that it can distinguish regimes with different causal directions, time lags, effects and sign of causal links, as well as changes in the variables' autocorrelation. Further, Regime-PCMCI is employed to observations of El Niño Southern Oscillation and Indian rainfall, demonstrating skill also in real-world datasets.

stat.ME↗

arXiv2018

Detecting causal associations in large nonlinear time series datasets

Identifying causal relationships from observational time series data is a key problem in disciplines such as climate science or neuroscience, where experiments are often not possible. Data-driven causal inference is challenging since datasets are often high-dimensional and nonlinear with limited sample sizes. Here we introduce a novel method that flexibly combines linear or nonlinear conditional independence tests with a causal discovery algorithm that allows to reconstruct causal networks from large-scale time series datasets. We validate the method on a well-established climatic teleconnection connecting the tropical Pacific with extra-tropical temperatures and using large-scale synthetic datasets mimicking the typical properties of real data. The experiments demonstrate that our method outperforms alternative techniques in detection power from small to large-scale datasets and opens up entirely new possibilities to discover causal networks from time series across a range of research fields.

stat.ME↗