Annual climatic and agronomic shifts induce phenotypic plasticity, causing standard deep learning models to fail in high-throughput automated phenotyping tasks, such as alfalfa (Medicago sativa L.) seed maturity assessment. Here, we developed a deep learning-based transfer learning framework to confer climate robustness to such models, validated on a multispectral imaging dataset (365–970 nm) covering five maturity stages across three environmentally distinct years. We designed the Multispectral Spatial Attention Network (MSANet), a hybrid architecture integrating a 3D-CNN backbone with spectral and spatial attention modules to extract complex spatio-spectral features. On single-year data, MSANet achieved 93% classification accuracy, significantly surpassing both traditional Support Vector Machine (77%) and deep learning baselines (e.g., ResNet18, 88%). However, this high intra-year performance did not generalize; direct model transfer to a different year caused accuracy to collapse to 41%, quantifying a profound domain shift. To mitigate this, We proposed an innovative Earth Mover's Distance (EMD)-guided ‘diagnose-adapt-finetune’ framework. This approach utilized EMD to diagnose layer-specific distributional shifts, employed EMD-guided Adaptive Batch Normalization (AdaBN) to align feature statistics across domains, and concluded with a data-efficient, few-shot fine-tuning strategy. The framework restored predictive accuracy to >90% on out-of-domain data using only 100 labeled samples per class from the target year, representing an approximate 90% reduction in annotation costs compared to full supervision. Crucially, the adapted model exhibited remarkable resilience to real-world data imperfections, maintaining stability under scenarios of class imbalance and label noise. Interpretability analyses further indicated that the model learned biologically plausible spectral correlates associated with seed maturation. Our work presents a generalizable methodology for developing environmentally robust phenotyping platforms, offering a promising pathway to enhance the reliability of AI systems in variable agricultural environments.

Physical dormancy (PY) commonly present in the seeds of higher plants is believed to be responsible for the germination failure by impermeable seed coat in hard seeds of legume species, instead of physiological dormancy (PD). In this study, a non-destructive approach involving multispectral imaging was used to successfully identify hard seeds from non-hard seeds in Medicago sativa, with accuracy as high as 96.8%–99.0%. We further adopted multiple-omics strategies to investigate the differences of physiology, metabolomics, methylomics, and transcriptomics in alfalfa hard seeds, with non-hard seeds as control. The hard seeds showed dramatically increased antioxidants and 125 metabolites of significant differences in non-targeted metabolomics analysis, which are enriched in the biosynthesis pathways of flavonoids, lipids and hormones, especially with significantly higher ABA, a hormone known to induce dormancy. In our transcriptomics results, the enrichment pathway of “response to abscisic acid” of differential expressed genes (DEG) supported the key role of ABA in metabolomics results. The methylome analysis identified 54,899, 46,216 and 54,452 differential methylation regions for contexts of CpG, CHG and CHH, and 344 DEGs might be regulated by hypermethylation and hypomethylation of promoter and exon regions, including four ABA- and JA-responsive genes. Among 8% hard seeds in seed lots, 24.5% still did not germinate after scarifying seed coat, and were named as non-PY hard seeds. Compared to hard seeds, significantly higher contents of ABA/IAA and ABA/JA were identified in non-PY hard seeds, which indicated the potential presence of PD. In summary, the significantly changed metabolites, gene expressions, and methylations all suggested involvement of ABA responses in hard seeds, and germination failure of alfalfa hard seeds was caused by combinational dormancy (PY + PD), rather than PY alone.