Accurate multi-day forecasting of floating-object trajectories on the ocean surface is critical for applications ranging from search-and-rescue to environmental tracking. This task remains however challenging due to the complex interplay of influencing factors such as ocean currents and winds. In this work, we frame trajectory prediction as a denoising task and present Conditional Diffusion models for Trajectories (CoDiT), which adapts the denoising diffusion framework, originally developed for image synthesis, to the problem of trajectory forecasting. CoDiT generates realistic trajectory forecasts, conditioned on heterogeneous context data: ocean currents and winds from reanalysis products, bathymetry, and the initial position. We train and evaluate CoDiT on two global, specialized datasets focusing on the open ocean and coastal regions, using GPS trajectories from the Global Drifter Program as ground truth. We compare CoDiT rigorously against various baselines, including a convolutional neural network that predicts velocity fields, and physical forecasts generated directly from the current and wind fields. Quantitative evaluations show that CoDiT achieves the lowest position error across both datasets and all forecast horizons, and the best probabilistic forecast quality among all methods, as measured by the energy score. Notably, in the coastal setting, CoDiT is the only method to surpass the naive persistence baseline in position error.
Christian Donner, Shirin Goshtasbpour, Emanuele Dalsasso et al.· Machine Learning: Earth· 0 citations
Forecasting crop growth across agricultural landscapes is important for improving the productivity, resilience, and operational management of farming systems. In this work, we investigate whether Earth observation time series and meteorological drivers can be used to predict future canopy development at country scale. We focus on winter wheat and formulate crop growth prediction as forecasting future leaf area index (LAI) trajectories beyond the last available Sentinel-2 observation. We evaluate this task on a multi-year dataset which spans the entire country of Switzerland, containing over 20 million pixel-level Sentinel-2-derived LAI time series paired with meteorological variables. Because cloud cover and revisit gaps leave LAI supervision sparse, models fit the few valid (cloud-free) LAI observations yet oscillate implausibly between them, producing trajectories no real canopy could follow. We introduce a lightweight unimodal shape regulariser which improves trajectory plausibility with negligible loss in accuracy. We compare deep learning sequence-to-sequence (Seq2Seq) models with classic machine learning baselines and show that Seq2Seq models generalise well across years, achieving $\mathrm{R}^2$ above 0.8 and consistently outperforming conventional approaches. Together, these results demonstrate that remote sensing and weather-driven sequence modelling can learn crop growth dynamics at landscape scale. S
Dominik Senti, Mehmet Ozgur Turkoglu, M. Volpi et al.· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.