Regional-to-continental scale monitoring of phenology using remote sensing with a network of digital cameras: Progress and results from PhenoCam

Background/Question/Methods Regional-to-continental scale monitoring of phenology has recently emerged as an important priority for assessing and understanding ecosystem responses to climate change.  However, sources of in-situ data related to phenology are sparse and of variable quality.  Phenology products based on satellite remote sensing are becoming increasingly available, but uncertainties associated with these products is poorly characterized.   Specifically, the correspondence between what is sensed from space and phenology processes on the ground is poorly quantified. Further, satellite-based measures of phenology show substantial variability across sensors and algorithms.  These uncertainties limit our ability to use satellite data to study phenological responses to climate variability and change.  The objective of this study is to assess the quality and information content of phenology information and products derived from NASA’s Moderate Resolution Imaging Spectroradiometer (MODIS).  To do this, we used “near surface remote sensing” data from a network of digital cameras (PhenoCam: http://phenocam.sr.unh.edu) spanning a range of climate, ecosystem, and landscape types.  Specifically, we analyzed time series of vegetation indices derived from both satellite and camera-based data streams, and assessed commonalities and differences in seasonal phenological trajectories and transition dates (e.g., start, end of growing season) captured by each source of data. 

Results/Conclusions Results show generally good agreement between MODIS and PhenoCam time series for many sites.  Agreement is stronger for sites dominated by deciduous forests and weaker for more arid and conifer locations, where the underlying seasonal cycle is less pronounced.   PhenoCam data are also sensitive to the camera field of view and the sub-region of images used to generate time series.   Camera placement, maintenance, and to some extent calibration also affect the overall data quality.  Preprocessing of camera data (filtering and smoothing) is required to optimize signal-to-noise ratios and fill gaps.  Our results show that data from networked digital cameras provide a useful source of information for monitoring phenology at local scales, for scaling local data to the resolution of satellite data, and for assessing the quality and information content of satellite-derived phenology data sets.  At the same time, more effort is required to refine methods for extracting high quality time series data from digital cameras. Naïvely ignoring these issues can lead to poor results.   As these methods mature, networks of digital cameras such as PhenoCam will become increasingly useful for site-level monitoring as well as satellite validation studies.