Linking ground-based volunteer observations with moderate-resolution satellite observations of phenology

Complex large-scale environmental issues, such as plant-community response to climate change, require both extensive ground-based observations and remote-sensing platforms suitable for integrating observations and modeling at... [ view full abstract ]

Complex large-scale environmental issues, such as plant-community response to climate change, require both extensive ground-based observations and remote-sensing platforms suitable for integrating observations and modeling at landscape scales. Citizen science offers a promising means of gathering broadly distributed ground-based observations while also enhancing public understanding of science. As part of a larger study to visualize forest adaptive capability under current and projected climates, we are exploring challenges of linking ground-based volunteer phenological observations with continental-scale imagery from the Moderate Resolution Imaging Spectrometer (MODIS). We are partnering with three established citizen science programs to acquire volunteer data of forest trees (National Phenology Network (NPN), Project Budburst and Plant Watch) and have focused on NPN data thus far. We used multiple strategies to assure the quality of both datasets including (1) shifting to adjacent MODIS pixels dominated by forest and (2) limiting NPN observations to those that were accompanied by observations immediately prior to the development of the phenophase. We found moderate correlation between dates of MODIS’ greenness measure and dates of relevant phenophases reported by NPN volunteers for dominant forest species, including sugar maple, dogwood and tulip poplar, as well as for cloned lilacs that have been planted around the U.S. Correlations were weak or absent for other species, such as balsam poplar and quaking aspen, possibly because they are less prevalent in the forest at the observation sites. Our results suggest that citizen science and remote sensing observations can be comparable measures of phenology but will require significant data filtering and would benefit from guiding volunteers in sampling more frequently (particularly before key phenophases) and in selecting large homogenous forest stands. Our data also indicate that some species may better capture the average forest phenological response to the environment, which aligns with satellite measures of landscape phenology.