Automating Phenological Stage Detection from Citizen Science Images for Plant Phenology Monitoring

• DOI: 10.5194/egusphere-egu25-9921
• Publication Date: 2025-03-14T22:37:01Z
• Authors: Negin Katal; Michael Rzanny; Susanne Tautenhahn; Patrick Mäder; Hans Christian Wi...; David Boho; Jana Wäldchen
• Keywords: Citizen Science

Plant phenology, the study of seasonal events in plants' life cycles such as budburst, flowering onset, leaf-out, fruit ripening, and senescence, is intrinsically linked to climatic conditions plays a crucial role ecosystem processes like carbon nutrient cycling. Due its ecological importance, many countries have established phenological monitoring networks based on systematic protocols. However, declining volunteer participation recent decades has raised concerns about continuity these invaluable datasets.Advancements technology, machine learning, smartphone accessibility spurred development plant identification apps. These apps enable users identify species without prior botanical knowledge, generating vast datasets occurrences.This investigates potential applying learning citizen science-derived image data for monitoring. By utilizing pre-trained deep model, we extracted relevant features classified 39 species-specific phenostages nine common Germany using Support Vector Machine (SVM) classifier. Our model achieved an impressive overall accuracy 96%, enabling automated annotation over 600,000 occurrence images from Flora Incognita app into corresponding stages.With this approach, not only did capture additional fine-granular phenostages, flower bud unripe stages, which are less commonly resolved traditional network datasets, but also observed interannual variability each phenostage across different years. This demonstrates feasibility integrating opportunistic science schemes. addressing challenges posed by participation, method significantly enhances temporal spatial resolution offering innovative opportunities phenology research.