Model-Agnostic, Temperature-Informed Sampling Enhances Cross-Year Crop Mapping with Deep Learning

2025-06-15

AI Summary

Traditional crop-type classification methods rely on fixed calendar-date sampling and same-year labeled data, rendering them vulnerable to interannual phenological variability, with poor generalizability and no quantification of prediction uncertainty—undermining reliability for real-time monitoring. To address this, we propose a model-agnostic time-series resampling framework grounded in growing degree days (GDD), replacing calendar time with thermal time to adaptively focus on phenologically active periods. This work is the first to incorporate ecophysiological GDD principles into remote sensing time-series modeling, enabling cross-year generalization and real-time inference without requiring current-year labels. Furthermore, we integrate an uncertainty calibration mechanism that enhances classification confidence—particularly under small-sample conditions and early in the season (e.g., end of June). Evaluated on multi-season Sentinel-2 data from Switzerland, our method achieves superior accuracy using only 10% of training samples compared to full-data baselines, with significant improvements in both classification performance and uncertainty calibration.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Computer Vision: Remote Sensing / Geospatial AI
• Reasoning under Uncertainty: Graphical Models

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Search and Retrieval-Augmented AI: Agentic search

Abstract

Conventional benchmarks for crop type classification from optical satellite time series typically assume access to labeled data from the same year and rely on fixed calendar-day sampling. This limits generalization across seasons, where crop phenology shifts due to interannual climate variability, and precludes real-time application when current-year labels are unavailable. Furthermore, uncertainty quantification is often neglected, making such approaches unreliable for crop monitoring applications. Inspired by ecophysiological principles of plant growth, we propose a simple, model-agnostic sampling strategy that leverages growing degree days (GDD), based on daily average temperature, to replace calendar time with thermal time. By uniformly subsampling time series in this biologically meaningful domain, the method emphasizes phenologically active growth stages while reducing temporal redundancy and noise. We evaluate the method on a multi-year Sentinel-2 dataset spanning all of Switzerland, training on one growing season and testing on other seasons. Compared to state-of-the-art baselines, our method delivers substantial gains in classification accuracy and, critically, produces more calibrated uncertainty estimates. Notably, our method excels in low-data regimes and enables significantly more accurate early-season classification. With only 10 percent of the training data, our method surpasses the state-of-the-art baseline in both predictive accuracy and uncertainty estimation, and by the end of June, it achieves performance similar to a baseline trained on the full season. These results demonstrate that leveraging temperature data not only improves predictive performance across seasons but also enhances the robustness and trustworthiness of crop-type mapping in real-world applications.

Problem

• Improves cross-year crop mapping with temperature-informed sampling
• Addresses phenology shifts due to interannual climate variability
• Enhances real-time classification without current-year labels

Innovation

• Uses growing degree days for thermal time sampling
• Enhances cross-year crop mapping accuracy
• Improves uncertainty estimation in classification