Utilizing public Internet-connected cameras for a cross- continental plant phenology monitoring system

Development of inexpensive, automated ground methods is necessary to advance precision plant phenology monitoring across large spatial extents. We propose the use of free, publicly available, Internet-connected cameras, often associated with nonscientific monitoring, to monitor plant phenology at continental scales. We provide a methodology to detect changes in vegetation greenness and determine timing of spring and fall events from over 1100 public cameras across North America from February 2008 - 2009. Manual image segmentation facilitated spring detection for both deciduous and understory vegetation occurring within a single camera view. Deciduous spring green-up was highly correlated with visual ground truths, despite signal noise introduced by varying image exposure and automatic color correction. Compared to co-occurring satellite remote sensing products, public cameras had an equivalent or higher ability to detect spring with fewer days lost to cloud cover.