A Geometry-Aware Multi-Task Framework for Parcel Delineation with Geospatial Foundation Models

Field boundary delineation (FBD) is a foundational task in Earth Observation (EO), supporting a wide range of agricultural and environmental applications. Accurate, parcel-level boundaries enable field-level reporting, water productivity monitoring, and scalable, decision-support systems. However, extracting reliable field geometries from medium-resolution satellite imagery remains challenging, particularly at 10 m resolution where boundaries are thin, low-contrast, and often visually ambiguous. Adjacent parcels can appear similar; supervision data is frequently sparse or inconsistent across regions, and agricultural practices vary widely introducing domain shifts that undermine generalizability. These factors make naïve “extent-only” approaches prone to merging neighboring fields, while “boundary-only” methods often fail to produce closed, stable instances when separators are weak or missing. Geospatial foundation models (FMs), pre-trained on large, multi-modal satellite archives, offer a promising solution by enabling transferable visual representations for EO tasks with limited supervision. Yet their application to geometry-sensitive tasks like FBD remains, to the best of our knowledge, unexplored. 

This work presents a boundary-centric field delineation pipeline that demonstrates one of the first operational deployments of geospatial FMs for parcel mapping using Sentinel-2 imagery. At its core, the model leverages TerraMind, a modality-aware, self-supervised EO foundation model, as the feature encoder. This FM backbone enables the system to learn transferable, generic spatial representations from large-scale EO data. To enhance generalization across regions and seasons, the encoder is explicitly conditioned in both time and space. Temporal context is provided through a Day-of-Year (DOY) sinusoidal embedding, capturing phenological variability and seasonal appearance shifts across acquisitions. Spatial context is introduced via SatCLIP-based coordinate embeddings, which transform geographic patch-center coordinates into rich, location-aware priors using a frozen SatCLIP backbone and lightweight projection.  

Built atop the TerraMind feature hierarchy is a Fractal ResUNet-style decoder that reconstructs fine boundary details while preserving global parcel topology. Operating over a multi-scale pyramidal representation, the decoder reshapes latent token embeddings into spatial maps and progressively upsamples them through skip-connected blocks. This design effectively balances fine-grained localization and broad contextual reasoning—essential at 10m resolution where boundaries are thin and adjacent parcels are visually similar. The model produces three interrelated outputs through a coupled multi-task formulation: a probability map for field extent, a boundary likelihood map capturing separator ridges, and a continuous distance-to-boundary field that encodes interiorness. These outputs are supervised jointly, encouraging geometric coherence across predictions.  

To quantify performance, we evaluate delineation quality on a multi-country European validation set built from parcel-level labels RapidCrops. Across countries, the model reaches boundary and extent IoU in the ~0.75–0.91 range, with higher scores in landscapes dominated by larger, well-separated parcels and lower scores in regions characterized by small fields, weak visual separators, or incomplete ground truth. This variability highlights both the scalability enabled by FM features and the remaining performance ceiling imposed by 10 m resolution and label quality.