Deep generative vision models are now able to synthesize realistic-looking satellite imagery. However, the possibility of hallucinations prevents their adoption of risk-sensitive applications, such as generating materials for communicating climate change. To demonstrate this issue, we train a generative adversarial network (GAN, pix2pixHD) to create synthetic satellite imagery of future flooding and reforestation events. We find that a pure deep learning-based model can generate photorealistic flood visualizations but hallucinate floods at locations that are not susceptible to flooding. To address this issue, we propose to condition and evaluate generative vision models on segmentation maps of physics-based flood models. We show that our physics-conditioned model outperforms the pure deep learning-based model and a handcrafted baseline. We evaluate the generalization capability of our method to different remote sensing data and different climate-related events (reforestation). We publish our code and dataset which includes the data for a third case study of melting Arctic sea ice and >30 000 labeled HD image triplets—or the equivalent of 5.5 million images at$128 \times 128$pixels—for segmentation guided image-to-image (im2im) translation in Earth observation. Code and data are available at github.com/blutjens/eie-earth-public.

Recent progress in self-supervision has shown that pre-training large neural networks on vast amounts of unsupervised data can lead to substantial increases in generalization to downstream tasks. Such models, recently coined foundation models, have been transformational to the field of natural language processing.Variants have also been proposed for image data, but their applicability to remote sensing tasks is limited.To stimulate the development of foundation models for Earth monitoring, we propose a benchmark comprised of six classification and six segmentation tasks, which were carefully curated and adapted to be both relevant to the field and well-suited for model evaluation. We accompany this benchmark with a robust methodology for evaluating models and reporting aggregated results to enable a reliable assessment of progress. Finally, we report results for 20 baselines to gain information about the performance of existing models.We believe that this benchmark will be a driver of progress across a variety of Earth monitoring tasks.

Requirements are rarely static, and are ever more likely to evolve as the development time of a system stretches out and its service life increases. In this paper, we discuss the evolution of requirements for the U.S. spacesuit, the Extravehicular Mobility Unit (EMU), as a case study in the need for flexibility in system design. We explore one fundamental environmental change, using the Shuttle EMU aboard the International Space Station (ISS), and the resulting EMU requirement and design changes. The EMU, like most complex systems, faces considerable uncertainty during its service life. Changes in the technical, political, or economic environment cause changes in requirements, which in turn necessitate design modifications or upgrades. We make the case that flexibility is a key attribute that needs to be embedded in the design of long-lived systems to enable them to efficiently meet the inevitability of changing requirements after they have been fielded.

Cognitive-performance and motor-performance activities in multi-task, high-workload environments were assessed during astronaut performance in space flight and in isolation. Data was collected in microgravity on the International Micro-gravity Laboratory (IML) space shuttle mission (STS-42), and the Canadian Astronaut Program Space Unit Life Simulation (CAPSULS) mission offered an ideal opportunity to collect data for individuals in extreme isolation to complement the space flight data using similar hardware, software, and experimental protocols. The mental workload and performance experiment (MWPE) was performed during the IML-1 space flight mission, and the memory processes and motor control (MEMO) experiment was performed during the CAPSULS isolation mission. In both experiments, short-term exhaustive memory and fine motor control associated with human-computer interaction was studied. Memory processes were assessed using a Sternberg-like exhaustive memory search containing 1, 2, 4, or 7 letters. Fine motor control was assessed using velocity-controlled (joystick) and position-controlled (trackball) computer input devices to acquire targets as displayed on a computer screen. Subjects repeated the tasks under two conditions that tested perceptual motor adaptation strategies: 1) During adaptation to the microgravity environment; and 2) While wearing left-right reversing prism goggles during the CAPSULS mission. Both conditions significantly degraded motor performance but not cognitive performance. The data collected during both the MEMO experiment and the MWPE experiments enhance the knowledge base of human interface technology for human performance in extreme environments.