Seasonal temperature change in temperate forests is known to trigger the start of spring growth, and both interannual and spatial variations in spring onset have been tied to climatic variability. Satellite dates are increasingly being used in phenology studies, but to date that has been little effort to link remotely sensed phenology to surface climate records. In this research, we use a two-parameter spring warming phenology model to explore the relationship between climate and satellite-based phenology. We employ daily air temperature records between 2000 and 2005 for 171 National Oceanographic and Atmospheric Administration weather stations located throughout New England to construct spring warming models predicting the onset of spring, as defined by the date of half-maximum greenness (D-50) in deciduous forests as detected from Moderate Resolution Imaging Spectrometer. The best spring warming model starts accumulating temperatures after March 20th and when average daily temperatures exceed 5 degrees C. The accumulated heat sums [heating degree day (HDD)] required to reach D-50 range from 150 to 300 degree days over New England, with the highest requirements to the south and in coastal regions. We test the ability of the spring warming model to predict phenology against a null photoperiod model (average date of onset). The spring warming model offers little improvement on the null model when predicting D-50. Differences between the efficacies of the two models are expressed as the 'climate sensitivity ratio' (CSR), which displays coherent spatial patterns. Our results suggest that northern (beech-maple-birch) and central (oak-hickory) hardwood forests respond to climate differently, particularly with disparate requirements for the minimum temperature necessary to begin spring growth (3 and 6 degrees C, respectively). We conclude that spatial location and species composition are critical factors for predicting the phenological response to climate change: satellite observations cannot be linked directly to temperature variability if species or community compositions are unknown.
