The SwissPhenoCam dataset: Country-scale phenology monitoring at the individual tree level

Understanding vegetation phenology at landscape to continental scales is essential for tracking ecosystem responses to climate change, improving biodiversity assessments, and strengthening land-surface models. While satellite remote sensing provides broad spatial coverage, it often lacks the temporal and structural detail required to resolve fine-scale phenological dynamics. In particular, satellite phenology products struggle to capture species-specific phenological responses to climatic variability. Over the past decade, PhenoCam networks have been developed to address some of these limitations. Here, we build on these efforts and introduce a novel Switzerland-scale phenocam dataset capturing both individual- and canopy-level phenological signals. We curate a large collection of high-frequency, high-resolution webcam imagery and use it to monitor expert-annotated regions of interest (ROIs) corresponding to individual trees with known species, as well as tree canopies.

The first iteration of the dataset is based on imagery acquired at 32 sites across Switzerland, spanning the full elevational range of the country. For some sites, observations begin as early as 2010 and extend to the present day. On average, each site has a temporal coverage of six years, amounting to a total of 175 site-years. The dataset currently includes over 5,000 tree-years of observations for individual trees, enabling species-level analyses of phenological variability. For all individual-tree and canopy-level ROIs, we apply automated greenness-based methods to extract green-up and green-down dates, allowing the investigation of species-specific phenological patterns across Switzerland’s climatic gradients, which are strongly structured by elevation. For approximately 1,000 tree-years, expert-annotated phenophase dates are available, providing a unique benchmark for calibration and validation.  We are thus able to report robust phenological transition dates for more than ten tree species in Switzerland over the past decade. The large number of individuals included in our monitoring efforts also allows for cross-sectional comparisons of season length and its variability across species, elevation, and years. 

By making high-frequency phenological observations available at the country scale, the SwissPhenoCam dataset provides a valuable resource for phenology monitoring and supports the development and evaluation of methods for phenological modeling and forecasting. We invite the community to use this dataset to advance understanding of vegetation dynamics in a rapidly changing world.