arXiv subjects

Thomas Jung

Publications and source records attributed to Thomas Jung.

12 recordsLinked to original sources

arXiv2026

FESOM2-JAX v1.0: a differentiable shadow of the ocean-sea-ice model FESOM2, cast onto GPUs

We present FESOM2-JAX, a Python re-implementation of the Finite-volumE Sea ice-Ocean Model (FESOM2) in JAX. The model retains the unstructured-mesh, cell-vertex finite-volume formulation of the original, runs unchanged from a laptop CPU to 256 GPUs, and is end-to-end differentiable. FESOM2-JAX is a code shadow of the Fortran model: a projection onto the Python ecosystem, translated with large language models and verified kernel by kernel against the original. It is built to lower the barrier to experimentation, from new numerics and parameterizations to gradient-based calibration and hybrid physics-machine-learning components, while remaining close enough to the original so that what is developed in the shadow can be transferred back. In a 1958-2019 hindcast at 1$^{\circ}$ equivalent resolution with identical physics and forcing, the mean states of the JAX and Fortran versions differ from each other by two orders of magnitude less than either differs from observations, and the two runs agree for six decades in global temperature, salinity, heat content, and sea ice. The complete 1$^{\circ}$ configuration fits on a single GPU, a node of four GH200 superchips integrates $\sim$113 simulated years per wall-clock day, and meshes of up to 7.4 million surface vertices ($\sim$5 km) scale to 128 GPUs. What limits the model is communication rather than arithmetic. What the shadow adds to the original is the gradient: a single reverse-mode pass through the full time loop returns the sensitivity of a model diagnostic to a parameter at every mesh vertex, verified against finite differences. To our knowledge, FESOM2-JAX is the first global ocean-sea-ice model of CMIP-class complexity written natively in a differentiable framework, and the first on an unstructured mesh.

physics.ao-ph↗

arXiv2026

CMIP-Forge: An Agentic System that Retrieves, Computes, and Self-Reviews Climate Science

The Coupled Model Intercomparison Project Phase 6 (CMIP6) has generated thousands of peer-reviewed publications documenting model configurations, evaluation procedures, emergent constraints, and projection uncertainties. As the community transitions toward CMIP7, efficiently extracting and operationalizing this unstructured knowledge alongside live data analysis represents a critical bottleneck. Here we present CMIP-Forge, a hybrid retrieval-augmented generation (RAG) and autonomous analysis system that bridges the gap between scientific literature and Earth System Grid Federation (ESGF) data archives. The system pairs a curated corpus of 6,581 CMIP6-related open-access publications (101,828 indexed chunks) with an agentic pipeline in which a tool-augmented worker plans and executes Python workflows over live climate data, while a panel of independent reviewer models audits its methodology end to end. CMIP-Forge introduces a multi-layered Defense-in-Depth architecture that enforces physical and methodological invariants through executable mechanisms: Abstract Syntax Tree (AST) static analysis, audited scientific primitives, and an autonomous adversarial peer-review protocol. We demonstrate the system's capabilities through end-to-end autonomous research pipelines spanning atmospheric teleconnections, ocean dynamics, regional extremes, and global warming projections. An agentic analysis system grounded in peer-reviewed literature, constrained by automated code guardrails, and audited by an independent adversarial review loop can complete complex climate-research workflows autonomously. The same experiments expose concrete failure modes of the review loop (sycophantic regression, REVISE verdicts that are never resolved, and the submission of stub code for review), each diagnosable from the immutable telemetry and provenance record released with the article.

physics.ao-ph↗

arXiv2026

An Ocean Model Ported by a Large Language Model: Experience and Lessons from FESOM2 (Fortran to C to C++/Kokkos)

Large language models (LLMs) can translate and modify source code, and have been shown to do so for codes of different complexity. Whether they can port a complete, production geophysical model to a different language without degrading its physics has not been established. We demonstrate that LLM-assisted code translation can preserve the physics of a complete production ocean model while moving it into a modern performance-portable form. We report our experience using an agentic LLM coding assistant, directed by domain experts, to port the FESOM2 unstructured mesh ocean--sea-ice model (about 74000 lines of core Fortran) first to C and then to C++/Kokkos for performance portability across CPUs and GPUs. We describe the practices that proved necessary, what worked and what did not, and the failure modes that we encountered. Three practices mattered most: translating in two stages that separate reproducing the numerics (Fortran to a clean C reference) from introducing parallelism (C to Kokkos); requiring a strictly literal translation in which the assistant was not permitted to ``improve'' the source; and validating each stage against an acceptance criterion suited to it. The C port reproduces the original Fortran at the level of long-term simulation statistics over five years. The Kokkos port is bit-for-bit identical to the C reference on CPU and statistically close on GPU over multi-year runs. On eddy-rich meshes up to 7.4 million surface vertices a single A100 GPU node runs 1.6--3.7 times faster than a CPU node, reaching the 1-2 simulated-years-per-day required for production integrations. The result is more than a single GPU port: by following a clear validation procedure, an LLM moved a full Fortran ocean model into another language and onto accelerators while preserving its physics in a matter of weeks.

physics.ao-ph↗

arXiv2026

A Hierarchical Multi-Agent System for Autonomous Discovery in Geoscientific Data Archives

The rapid accumulation of Earth science data has created a significant scalability challenge; while repositories like PANGAEA host vast collections of datasets, citation metrics indicate that a substantial portion remains underutilized, limiting data reusability. Here we present PANGAEA-GPT, a hierarchical multi-agent framework designed for autonomous data discovery and analysis. Unlike standard Large Language Model (LLM) wrappers, our architecture implements a centralized Supervisor-Worker topology with strict data-type-aware routing, sandboxed deterministic code execution, and self-correction via execution feedback, enabling agents to diagnose and resolve runtime errors. Through use-case scenarios spanning physical oceanography and ecology, we demonstrate the system's capacity to execute complex, multi-step workflows with minimal human intervention. This framework provides a methodology for querying and analyzing heterogeneous repository data through coordinated agent workflows.

cs.AI↗

arXiv2026

Enhanced Accessibility for Mobile Indoor Navigation

The navigation of indoor spaces poses difficult challenges for individuals with visual impairments, as it requires processing of sensory information, dealing with uncertainties, and relying on assistance. To tackle these challenges, we present an indoor navigation app that places importance on accessibility for visually impaired users. Our approach involves a combination of user interviews and an analysis of the Web Content Accessibility Guidelines. With this approach, we are able to gather invaluable insights and identify design requirements for the development of an indoor navigation app. Based on these insights, we develop an indoor navigation app that prioritizes accessibility, integrating enhanced features to meet the needs of visually impaired users. The usability of the app is being thoroughly evaluated through tests involving both visually impaired and sighted users. Initial feedback has been positive, with users appreciating the inclusive user interface and the usability with a wide range of accessibility tools and Android device settings.

cs.CY↗

arXiv2025

Climate Knowledge in Large Language Models

Large language models (LLMs) are increasingly deployed for climate-related applications, where understanding internal climatological knowledge is crucial for reliability and misinformation risk assessment. Despite growing adoption, the capacity of LLMs to recall climate normals from parametric knowledge remains largely uncharacterized. We investigate the capacity of contemporary LLMs to recall climate normals without external retrieval, focusing on a prototypical query: mean July 2-m air temperature 1991-2020 at specified locations. We construct a global grid of queries at 1° resolution land points, providing coordinates and location descriptors, and validate responses against ERA5 reanalysis. Results show that LLMs encode non-trivial climate structure, capturing latitudinal and topographic patterns, with root-mean-square errors of 3-6 °C and biases of $\pm$1 °C. However, spatially coherent errors remain, particularly in mountains and high latitudes. Performance degrades sharply above 1500 m, where RMSE reaches 5-13 °C compared to 2-4 °C at lower elevations. We find that including geographic context (country, city, region) reduces errors by 27% on average, with larger models being most sensitive to location descriptors. While models capture the global mean magnitude of observed warming between 1950-1974 and 2000-2024, they fail to reproduce spatial patterns of temperature change, which directly relate to assessing climate change. This limitation highlights that while LLMs may capture present-day climate distributions, they struggle to represent the regional and local expression of long-term shifts in temperature essential for understanding climate dynamics. Our evaluation framework provides a reproducible benchmark for quantifying parametric climate knowledge in LLMs and complements existing climate communication assessments.

cs.CL↗

arXiv2024

Robustness of AI-based weather forecasts in a changing climate

Data-driven machine learning models for weather forecasting have made transformational progress in the last 1-2 years, with state-of-the-art ones now outperforming the best physics-based models for a wide range of skill scores. Given the strong links between weather and climate modelling, this raises the question whether machine learning models could also revolutionize climate science, for example by informing mitigation and adaptation to climate change or to generate larger ensembles for more robust uncertainty estimates. Here, we show that current state-of-the-art machine learning models trained for weather forecasting in present-day climate produce skillful forecasts across different climate states corresponding to pre-industrial, present-day, and future 2.9K warmer climates. This indicates that the dynamics shaping the weather on short timescales may not differ fundamentally in a changing climate. It also demonstrates out-of-distribution generalization capabilities of the machine learning models that are a critical prerequisite for climate applications. Nonetheless, two of the models show a global-mean cold bias in the forecasts for the future warmer climate state, i.e. they drift towards the colder present-day climate they have been trained for. A similar result is obtained for the pre-industrial case where two out of three models show a warming. We discuss possible remedies for these biases and analyze their spatial distribution, revealing complex warming and cooling patterns that are partly related to missing ocean-sea ice and land surface information in the training data. Despite these current limitations, our results suggest that data-driven machine learning models will provide powerful tools for climate science and transform established approaches by complementing conventional physics-based models.

physics.ao-ph↗

arXiv2024

Emerging AI-based weather prediction models as downscaling tools

The demand for high-resolution information on climate change is critical for accurate projections and decision-making. Presently, this need is addressed through high-resolution climate models or downscaling. High-resolution models are computationally demanding and creating ensemble simulations with them is typically prohibitively expensive. Downscaling methods are more affordable but are typically limited to small regions. This study proposes the use of existing AI-based numerical weather prediction systems (AI-NWP) to perform global downscaling of climate information from low-resolution climate models. Our results demonstrate that AI-NWP initalized from low-resolution initial conditions can develop detailed forecasts closely resembling the resolution of the training data using a one day lead time. We constructed year-long atmospheric fields using AI-NWP forecasts initialized from smoothed ERA5 and low-resolution CMIP6 models. Our analysis for 2-metre temperature indicates that AI-NWP can generate high-quality, long-term datasets and potentially perform bias correction, bringing climate model outputs closer to observed data. The study highlights the potential for off-the-shelf AI-NWP to enhance climate data downscaling, offering a simple and computationally efficient alternative to traditional downscaling techniques. The downscaled data can be used either directly for localized climate information or as boundary conditions for further dynamical downscaling.

physics.ao-ph↗

arXiv2024

Recent global temperature surge amplified by record-low planetary albedo

In 2023, the global mean temperature soared to 1.48K above the pre-industrial level, surpassing the previous record by 0.17K. Previous best-guess estimates of known drivers including anthropogenic warming and the El Nino onset fall short by about 0.2K in explaining the temperature rise. Utilizing satellite and reanalysis data, we identify a record-low planetary albedo as the primary factor bridging this gap. The decline is caused largely by a reduced low-cloud cover in the northern mid-latitudes and tropics, in continuation of a multi-annual trend. Understanding how much of the low-cloud trend is due to internal variability, reduced aerosol concentrations, or a possibly emerging low-cloud feedback will be crucial for assessing the current and expected future warming.

physics.ao-ph↗

arXiv2021

A cavity-based optical antenna for color centers in diamond

An efficient atom-photon-interface is a key requirement for the integration of solid-state emitters such as color centers in diamond into quantum technology applications. Just like other solid state emitters, however, their emission into free space is severely limited due to the high refractive index of the bulk host crystal. In this work, we present a planar optical antenna based on two silver mirrors coated on a thin single crystal diamond membrane, forming a planar Fabry-Pérot cavity that improves the photon extraction from single tin vacancy (SnV) centers as well as their coupling to an excitation laser. Upon numerical optimization of the structure, we find theoretical enhancements in the collectible photon rate by a factor of 60 as compared to the bulk case. As a proof-of-principle demonstration, we fabricate single crystal diamond membranes with sub-$μ$m thickness and create SnV centers by ion implantation. Employing off-resonant excitation, we show a 6-fold enhancement of the collectible photon rate, yielding up to half a million photons per second from a single SnV center. At the same time, we observe a significant reduction of the required excitation power in accordance with theory, demonstrating the functionality of the cavity as an optical antenna. Due to its planar design, the antenna simultaneously provides similar enhancements for a large number of emitters inside the membrane. Furthermore, the monolithic structure provides high mechanical stability and straightforwardly enables operation under cryogenic conditions as required in most spin-photon interface implementations.

quant-ph↗

arXiv2020

Properties of the LWR model with time delay

In this article, we investigate theoretical and numerical properties of the first-order Lighthill-Whitham-Richards (LWR) traffic flow model with time delay. Since standard results from the literature are not directly applicable to the delayed model, we mainly focus on the numerical analysis of the proposed finite difference discretization. The simulation results also show that the delay model is able to capture Stop & Go waves.

math.NA↗

arXiv2019

Spin Measurements of NV Centers Coupled to a Photonic Crystal Cavity

Nitrogen-vacancy (NV) centers feature outstanding properties like a spin coherence time of up to one second as well as a level structure offering the possibility to initialize, coherently manipulate and optically read-out the spin degree of freedom of the ground state. However, only about three percent of their photon emission are channeled into the zero phonon line (ZPL), limiting both the rate of indistinguishable single photons and the signal-to-noise ratio (SNR) of coherent spin-photon interfaces. We here report on the enhancement of the SNR of the optical spin read-out achieved by tuning the mode of a two-dimensional photonic crystal (PhC)cavity into resonance with the NV-ZPL. PhC cavities are fabricated by focused ion beam (FIB) milling in thin reactive ion (RIE) etched ultrapure single crystal diamond membranes featuring modes with Q-factors of up to 8250 at mode volumes below one cubic wavelength. NV centers are produced in the cavities in a controlled fashion by a high resolution atomic force microscope (AFM) implantation technique. On cavity resonance we observe a lifetime shortening from 9.0ns to 8.0ns as well as an enhancement of the ZPL emission by almost one order of magnitude. Although on resonance the collection efficiency of ZPL photons and the spin-dependent fluorescence contrast are reduced, the SNR of the optical spin read-out is almost tripled for the cavity-coupled NV centers.

quant-ph↗