SwissCrop25: A National Multi-Year Benchmark for Operational Crop Mapping

2026-08-10

AI Summary

Existing crop mapping datasets struggle to jointly evaluate model performance across multi-year generalization, fine-grained classification, and cropland identification. To address this gap, this work introduces a nationwide Swiss benchmark dataset spanning seven growing seasons, integrating Sentinel-2 time-series imagery, phenological features derived from daily temperature, 73 crop classes, and five non-crop land-cover categories. A leave-one-year-out cross-validation protocol is designed to simulate real-world deployment conditions. Experiments on joint cropland delineation and crop classification using spatiotemporal architectures—including U-TAE, TSViT, and Galileo—demonstrate that TSViT achieves the best performance, surpassing U-TAE by 12 percentage points in macro mIoU. Domain-specific models consistently outperform general-purpose remote sensing foundation models, and incorporating phenological information significantly enhances inter-annual robustness, while also revealing pronounced seasonal trade-offs in model performance.

Abstract

Operational crop mapping requires models that generalise across years, resolve fine-grained crop taxonomies, and distinguish cropland from surrounding landscapes. However, existing crop mapping datasets enable evaluation of these requirements only in isolation. We therefore introduce SwissCrop25, a national-scale crop mapping benchmark dataset spanning seven growing seasons (2019-2025). SwissCrop25 combines Sentinel-2 time series, daily temperature observations, a fine-grained 73 crop taxonomy including grassland management types, and 5 explicit non-crop land cover classes. To evaluate realistic deployment conditions, we define a leave-one-year-out protocol with joint cropland delineation and crop classification for benchmarking representative crop mapping architectures. Evaluating U-TAE (convolutional temporal-attention model), TSViT (transformer-based spatio-temporal model), and Galileo (EO foundation model) reveals differences between architectures hidden by conventional benchmarks. In this setting, domain-specific models outperform Galileo, with TSViT achieving the best overall performance and a 12 pp macro-mIoU advantage over U-TAE. SwissCrop25 also exposes substantial interannual distribution shifts and shows that incorporating temperature-derived phenological information improves robustness. Finally, in-season evaluation reveals a trade-off between models, with U-TAE performing better early in the season and TSViT gaining an advantage later through improved rare-class discrimination. SwissCrop25 provides a challenging testbed for evaluating crop mapping systems under realistic operational conditions and is publicly released at https://huggingface.co/datasets/EOA-team/SwissCrop25 .

Problem

• crop mapping
• benchmark dataset
• multi-year generalization
• fine-grained taxonomy
• cropland delineation

Innovation

• multi-year benchmark
• fine-grained crop taxonomy
• leave-one-year-out evaluation
• phenological information
• operational crop mapping