SSL4Eco: A Global Seasonal Dataset for Geospatial Foundation Models in Ecology

2025-04-25

AI Summary

Ecological remote sensing faces two critical bottlenecks: severe scarcity of labeled data and a strong bias in existing self-supervised datasets toward anthropogenic regions, neglecting ecologically significant areas; moreover, multi-temporal data are typically sampled on fixed calendar-based seasonal intervals, failing to capture true phenological dynamics. To address these challenges, we propose SSL4Eco—the first global self-supervised learning dataset for multi-temporal Sentinel-2 imagery, explicitly designed around natural phenological cycles. Its core innovation lies in the tight coupling of a phenology-aware sampling strategy with a season-contrastive learning objective, enabling the first phenology-driven pretraining of geospatial foundation models. Evaluated on eight ecological downstream tasks, SSL4Eco achieves state-of-the-art performance on seven, notably improving multi-label classification and regression accuracy. The code, dataset, and pretrained models are fully open-sourced.

Technology Category

• Machine Learning: Unsupervised & Self-Supervised Learning
• Computer Vision: Remote Sensing / Geospatial AI
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Security and Privacy: Data transparency and provenance
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

With the exacerbation of the biodiversity and climate crises, macroecological pursuits such as global biodiversity mapping become more urgent. Remote sensing offers a wealth of Earth observation data for ecological studies, but the scarcity of labeled datasets remains a major challenge. Recently, self-supervised learning has enabled learning representations from unlabeled data, triggering the development of pretrained geospatial models with generalizable features. However, these models are often trained on datasets biased toward areas of high human activity, leaving entire ecological regions underrepresented. Additionally, while some datasets attempt to address seasonality through multi-date imagery, they typically follow calendar seasons rather than local phenological cycles. To better capture vegetation seasonality at a global scale, we propose a simple phenology-informed sampling strategy and introduce corresponding SSL4Eco, a multi-date Sentinel-2 dataset, on which we train an existing model with a season-contrastive objective. We compare representations learned from SSL4Eco against other datasets on diverse ecological downstream tasks and demonstrate that our straightforward sampling method consistently improves representation quality, highlighting the importance of dataset construction. The model pretrained on SSL4Eco reaches state of the art performance on 7 out of 8 downstream tasks spanning (multi-label) classification and regression. We release our code, data, and model weights to support macroecological and computer vision research at https://github.com/PlekhanovaElena/ssl4eco.

Problem

• Addresses scarcity of labeled ecological datasets for remote sensing
• Reduces bias in geospatial models toward high human activity areas
• Improves global vegetation seasonality capture via phenology-informed sampling

Innovation

• Phenology-informed sampling strategy for seasonality
• Season-contrastive objective for model training
• Global multi-date Sentinel-2 dataset SSL4Eco