Thomas Ludwig

1. End-User Development (EUD) represents a crucial shift toward empowering end users to actively participate in technology creation, adaptation, and extension, democratising software development be...

Kilometer-scale Earth system models are essential for capturing local climate change. However, these models are computationally expensive and produce petabyte-scale outputs, which limits their utility for applications such as probabilistic risk assessment. Here, we present the Field-Space Autoencoder, a scalable climate emulation framework based on a spherical compression model that overcomes these challenges. By utilizing Field-Space Attention, the model efficiently operates on native climate model output and therefore avoids geometric distortions caused by forcing spherical data onto Euclidean grids. This approach preserves physical structures significantly better than convolutional baselines. By producing a structured compressed field, it serves as a good baseline for downstream generative emulation. In addition, the model can perform zero-shot super-resolution that maps low-resolution large ensembles and scarce high-resolution data into a shared representation. We train a generative diffusion model on these compressed fields. The model can simultaneously learn internal variability from abundant low-resolution data and fine-scale physics from sparse high-resolution data. Our work bridges the gap between the high volume of low-resolution ensemble statistics and the scarcity of high-resolution physical detail.

Kilometer-scale Earth system models are essential for capturing local climate change. However, these models are computationally expensive and produce petabyte-scale outputs, which limits their utility for applications such as probabilistic risk assessment. Here, we present the Field-Space Autoencoder, a scalable climate emulation framework based on a spherical compression model that overcomes these challenges. By utilizing Field-Space Attention, the model efficiently operates on native climate model output and therefore avoids geometric distortions caused by forcing spherical data onto Euclidean grids. This approach preserves physical structures significantly better than convolutional baselines. By producing a structured compressed field, it serves as a good baseline for downstream generative emulation. In addition, the model can perform zero-shot super-resolution that maps low-resolution large ensembles and scarce high-resolution data into a shared representation. We train a generative diffusion model on these compressed fields. The model can simultaneously learn internal variability from abundant low-resolution data and fine-scale physics from sparse high-resolution data. Our work bridges the gap between the high volume of low-resolution ensemble statistics and the scarcity of high-resolution physical detail.

This study addresses an unexplored intersection of vocational education and training (VET), motor skill learning, craftsmanship and the architecture, engineering and construction (AEC) industry as well as expertise sharing. Acquiring proficiency in craftsmanship involves not only theoretical knowledge but also a multitude of physical skills, such as chiselling, which are honed through repetition. Mastering these manual skills and combining them with knowledge of materials, such as wood, is essential for an effective executing of craftmanship. We developed an augmented reality (AR) prototype through the design case study approach to design a support for beginners, especially apprentices, in woodworking education where the connection between craftspeople and material is important for the end results. The prototype was evaluated by 28 apprentices inside and outside a woodwork shop setting. While participants praised the prototype’s 3D visualisations and interactive models, challenges in compatibility with woodwork shop environments were evident. Our findings highlight the importance of on-site evaluations in the craftsmanship context and the significance of direct interaction with real world materials. We provide a set of seven design guidelines derived from our empirical findings to inform the development of AR-based learning applications in craftsmanship. Our discussion opens up the question of the role of interactive 3D models for collaborative learning. The research extends AR’s impact in craftsmanship and AEC industry. Our study underscores the importance of expert knowledge in handling certain materials and the potential of AR to reshape work practices and visualisations, contributing to expertise sharing in this domain.