Phenology, measured as the timing of recurring biological events and their environmental drivers, is highly sensitive to changes in environmental conditions and can vary widely across landscapes. Current technologies for observing plant phenology are either (1) ground-based for small-scale, high precision but labor-intensive measurements or (2) remote sensing-based for large-scale but low spatial resolution measurements, often too coarse to detect species and community level responses. We present a novel approach for detecting phenological events across North America utilizing freely-available Internet-connected cameras. We provide methodologies for detecting changes in ‘greenness’ and determining the timing of phenological events from webcam images. In addition, we compare the quality and precision of our detection with that of remote sensing MODIS (Moderate Resolution Imaging Spectroradiometer) products, currently used for large-scale environmental monitoring. Images from over 1,100 georeferenced webcams were collected twice-daily from February 2008 – February 2009. In a subset of 32 cameras, images were segmented into different vegetation types (i.e. evergreen, deciduous, understory). Greenness signals specific to vegetation types were then calculated using a per-pixel transformation of ‘Excess Green’. By fitting a double sigmoid function to the greenness time series, we were able to estimate dates of spring onset and senescence.
Results/Conclusions
We successfully detected the spring on-set of green-up from our webcam images. In addition, we identified a trend in the timing of spring on-set across a latitudinal gradient that differed between deciduous, evergreen, and understory vegetation types. Both webcam and satellite green-up signals had high day to day variability. Problems with using webcams included: cameras were subject to move, malfunction or go offline without warning. Automatic white balance posed additional variability. These data quality issues are comparable to sensor malfunction, cloud cover, and varying atmospheric conditions that decrease data quality in satellite-based signals. Webcams, however, have an advantage over remote sensing images in that they can capture more frequent images (minutes to daily), have much greater spatial accuracy and can potentially be used to detect organism, species and community level responses. Also, they are less sensitive to cloud cover, which completely masks green-up signals in satellite imagery. Thus, they represent an untapped, yet highly valuable resource, for large-scale ecological and environmental monitoring.
