arXiv subjects

Maybritt Schillinger

Publications and source records attributed to Maybritt Schillinger.

7 recordsLinked to original sources

arXiv2026

CORDEX-ML-Bench: A Benchmark for Data-Driven Regional Climate Downscaling -Experiment Design and Overview

Machine learning (ML) has emerged as a cost-effective approach to complement dynamical downscaling for producing high-resolution regional climate projections. However, the absence of standardised training and evaluation protocols, applied consistently across multiple domains, continues to hinder meaningful model intercomparison. We introduce CORDEX-ML-Bench, a benchmark aligned with CORDEX, which constitutes the first phase of a community initiative to advance data-driven downscaling toward operational readiness, and complement future dynamical downscaling efforts under CMIP7. The framework targets downscaled daily maximum temperature and precipitation to ~10 km resolution (20x increase) across three pilot regions; European Alps, New Zealand, and Southern Africa. Using a perfect-model experimental design, we evaluate 40 ML configurations developed independently, spanning traditional ML, convolutional U-Nets, vision transformers, graph neural networks, and generative models based on diffusion, flow matching, and generative adversarial networks. Models are trained under two experimental periods, an empirical-statistical downscaling pseudo-reality (historical period only) and Emulator (historical and future periods) -and are evaluated against a core set of metrics developed specifically for assessing downscaling skill. Generative models consistently outperform deterministic approaches for precipitation, better capturing fine-scale variability and extremes. For temperature, the generative advantage narrows and deterministic architectures remain competitive. Models trained solely on the historical period systematically underestimate future climate-change signals while those additionally trained on a future period perform better. These findings raise concerns about historically trained models widely used in an operational setting, underscoring the need for rigorous extrapolation testing.

physics.ao-ph↗

arXiv2026

Probabilistic storyline attribution using machine learning

A fundamental goal in climate attribution is to estimate how forced climate change contributes to observed extreme weather events. The storyline attribution method compares an observed weather event, conditional on its atmospheric dynamic state (i.e., atmospheric circulation), in the current, 'factual' climate to an event with very similar circulation conditions in a hypothetical, 'counterfactual' climate. However, physical climate models cannot directly transfer these storyline counterfactuals across different climate forcing states. Statistical and machine learning techniques may overcome this limitation; yet, emulating circulation-conditional extreme events under different climate states is challenging. Here, we demonstrate distributional autoencoders (DAEs) as a versatile method for generating climate counterfactuals. They model the full distribution of spatially resolved European temperature fields conditional on the atmospheric circulation state and the mean global warming level. These distributions allow for deriving meaningful conditional probability ratios, which is a particular advantage of the DAE-based storyline approach. We train DAEs on fully coupled climate model simulations and we evaluate the modelled distributions across different factual and storyline-based counterfactual climate model simulations. In an illustrative case study, we revisit the 2003 European heatwave and we generate counterfactuals for a hypothetical `2003-like European heatwave' using ERA5 circulation, which we hypothesize to occur a quarter century (2028) and a half century (2053) after 2003. The conditional intensity would increase from 29.3 {\deg}C in 2003, to 30.3 {\deg}C and 32.1 {\deg}C in 2028 and 2053, respectively and conditional probability ratios would be 2.1 and 3.2 when compared to 2003.

stat.AP↗

arXiv2026

How Can Machine Learning Emulators Best Support Climate Science?

For decades, physics-based climate models have been used to provide insights for climate decision-making. Their application is, however, constrained by significant computational and technical demands. Machine learning (ML) emulators offer a way to reduce these high computational costs; yet, it remains challenging to use ML emulators effectively in climate research. In practice, climate scientists often bypass emulators altogether, and machine learning researchers frequently develop them as methodological showcases without proving their practical utility. The reasons are diverse, ranging from limited accessibility and a lack of specialized knowledge to broader concerns about the physical grounding of ML methods. Here, we discuss limitations and introduce a framework for guiding emulator development, considering both climate science and machine learning perspectives. We argue that designing easy-to-adopt emulators that address clearly defined tasks and demonstrate their reliability is essential. This offers a promising path towards making machine-learning approaches more relevant and usable for applied climate research.

cs.LG↗

arXiv2025

Bridging CORDEX and CMIP6: Machine Learning Downscaling for Wind and Solar Energy Droughts in Central Europe

Reliable regional climate information is essential for assessing the impacts of climate change and for planning in sectors such as renewable energy; yet, producing high-resolution projections through coordinated initiatives like CORDEX that run multiple physical regional climate models is both computationally demanding and difficult to organize. Machine learning emulators that learn the mapping between global and regional climate fields offer a promising way to address these limitations. Here we introduce the application of such an emulator: trained on CMIP5 and CORDEX simulations, it reproduces regional climate model data with sufficient accuracy. When applied to CMIP6 simulations not seen during training, it also produces realistic results, indicating stable performance. Using CORDEX data, CMIP5 and CMIP6 simulations, as well as regional data generated by two machine learning models, we analyze the co-occurrence of low wind speed and low solar radiation and find indications that the number of such energy drought days is likely to decrease in the future. Our results highlight that downscaling with machine learning emulators provides an efficient complement to efforts such as CORDEX, supplying the higher-resolution information required for impact assessments.

stat.AP↗

arXiv2025

EnScale: Temporally-consistent multivariate generative downscaling via proper scoring rules

The practical use of future climate projections from global circulation models (GCMs) is often limited by their coarse spatial resolution, requiring downscaling to generate high-resolution data. Regional climate models (RCMs) provide this refinement, but are computationally expensive. To address this issue, machine learning (ML) models can learn the downscaling function, mapping coarse GCM outputs to high-resolution fields. Among these, generative approaches aim to capture the full conditional distribution of RCM data given coarse-scale GCM data, which is characterized by large variability and thus challenging to model accurately. We introduce EnScale, a generative ML framework emulating the full GCM-to-RCM map by training on multiple pairs of GCM and corresponding RCM data. It first adjusts large-scale mismatches between GCM and coarsened RCM data, followed by a super-resolution step to generate high-resolution fields. To efficiently model the high-dimensional output, the super-resolution step employs a novel class of sparse local stochastic layers. Both steps employ generative models optimized with the energy score, a proper scoring rule. Compared to state-of-the-art ML downscaling approaches, our setup reduces computational cost by about one order of magnitude. EnScale jointly emulates multiple variables -- temperature, precipitation, solar radiation, and wind -- spatially consistent over Central Europe. In addition, we propose a variant EnScale-t that enables temporally consistent downscaling. We establish a comprehensive evaluation framework across various categories including calibration, spatial and temporal structure, extremes, and multivariate dependencies. Comparison with diverse benchmarks demonstrates EnScale(-t)'s competitive performance and computational efficiency, offering a promising approach for accurate and temporally consistent RCM emulation.

physics.ao-ph↗

arXiv2025

Enforcing tail calibration when training probabilistic forecast models

Probabilistic forecasts are typically obtained using state-of-the-art statistical and machine learning models, with model parameters estimated by optimizing a proper scoring rule over a set of training data. If the model class is not correctly specified, then the learned model will not necessarily issue forecasts that are calibrated. Calibrated forecasts allow users to appropriately balance risks in decision making, and it is particularly important that forecast models issue calibrated predictions for extreme events, since such outcomes often generate large socio-economic impacts. In this work, we study how the loss function used to train probabilistic forecast models can be adapted to improve the reliability of forecasts made for extreme events. We investigate loss functions based on weighted scoring rules, and additionally propose regularizing loss functions using a measure of tail miscalibration. We apply these approaches to a hierarchy of increasingly flexible forecast models for UK wind speeds, including simple parametric models, distributional regression networks, and conditional generative models. We demonstrate that state-of-the-art models do not issue calibrated forecasts for extreme wind speeds, and that the calibration of forecasts for extreme events can be improved by suitable adaptations to the loss function during model training. This introduces a trade-off between calibrated forecasts for extreme events and calibrated forecasts for more common outcomes.

stat.AP↗

arXiv2023

Separating internal and externally-forced contributions to global temperature variability using a Bayesian stochastic energy balance framework

Earth's temperature variability can be partitioned into internal and externally-forced components. Yet, underlying mechanisms and their relative contributions remain insufficiently understood, especially on decadal to centennial timescales. Important reasons for this are difficulties in isolating internal and externally-forced variability. Here, we provide a physically-motivated emulation of global mean surface temperature (GMST) variability, which allows for the separation of internal and external variations. To this end, we introduce the ``ClimBayes'' software package, which infers climate parameters from a stochastic energy balance model (EBM) with a Bayesian approach. We apply our method to GMST data from temperature observations and 20 last millennium simulations from climate models of intermediate to high complexity. This yields the best estimates of the EBM's forced and forced + internal response, which we refer to as emulated variability. The timescale-dependent variance is obtained from spectral analysis. In particular, we contrast the emulated forced and forced + internal variance on interannual to centennial timescales with that of the GMST target. Our findings show that a stochastic EBM closely approximates the power spectrum and timescale-dependent variance of GMST as simulated by modern climate models. Small deviations at interannual timescales can be attributed to the simplified representation of internal variability and, in particular, the absence of (pseudo-)oscillatory modes in the stochastic EBM. Altogether, we demonstrate the potential of combining Bayesian inference with conceptual climate models to emulate statistics of climate variables across timescales.

physics.ao-ph↗