Plant biodiversity between history and genomics: studying, promoting and digitalizing the historical herbarium collections

Herbaria are among the few infrastructures that preserve biological material across centuries at continental scale, enabling direct genomic comparisons between past and present populations. However, reliable temporal inference from herbarium specimens depends on a chain of linked decisions, including taxonomic stability, sampling design, specimen choice, and DNA-appropriate analytical workflows. This thesis develops an integrative pipeline that treats these dependencies as a single system, moving from identity, to data readiness, to sequencing triage, to temporal population genomics, with the aim of improving the reliability, efficiency, and interpretability of historical genomics in Alpine plants. Using the Alpine endemic Favratia zoysii as a focal system, the thesis addresses both methodological and conceptual challenges inherent to long-lived perennials, where genomic responses to recent demographic change may be delayed. First, it stabilizes taxonomic identity through critical examination of historical material and formal typification. Second, it quantifies herbarium “data readiness” at scale via a Herbarium Density Index, exposing temporal and spatial biases in Alpine collections. Third, it develops an image-based machine-learning predictor to triage specimens for destructive sampling, increasing sequencing efficiency while reducing unnecessary damage. Finally, it implements a temporal genomics workflow tailored to degraded herbarium DNA and emphasizes power-aware interpretation of both detected and null results. Together, the thesis provides a blueprint for historical genomics that integrates taxonomy, digitization, machine learning, and population genomics, while explicitly accounting for uncertainty, sampling limits, and biological time lags.