Species-specific leaf phenology of deciduous trees captured by digital cameras
Plant phenology is typically observed via either ground-based observations on individuals or remote sensing of land surface vegetation. Little attempt has been made to integrate these two sources of phenological information. In part this relates to the challenge of interpreting information from multiple data sources collected at different spatial scales using different observational protocols. As an intermediate step we employed digital cameras, which span an area with enough spatial resolution to identify temporal changes in individual deciduous tree species canopies with continuous observations. The challenge is how these camera images relate to field observations: what phenological information is captured by digital cameras, how do they compare with field observations, and do the metrics from those images provide comparable species-specific phenological responses to the environment? To answer these questions, we set up time-lapse cameras to take daily photos of deciduous forest tree species canopies in nine sites with parallel observations on leaf phenology of eight tree species twice a week in spring and autumn in five sites over three years. Time series of color indices from camera images were analyzed for each tree species and compared with field observations in southern New England.
Color indices generated from camera images represent seasonal changes of deciduous trees species over the growing season. However, the indices do not adequately capture the timing of bud burst. The starting point of increasing green color index in spring was later than onset of leaf unfolding. However it did match the peak timing of leaf unfolding. The green color index increased as the leaf area expanded in spring. Decreasing green color index during autumn matched the slow process of leaf coloration and leaf fall in autumn. Thus we suggest using the starting point of increasing green color index as proxy of leaf unfolding time or the start of growing season, and to use end point of decreasing green color index as proxy of end of growing season. Consistent differences of growing season length among deciduous tree species (e.g. shorter growing season length of maples than oaks) were found. Interestingly these differences were mostly attributed to the large differences in the end of growing season in autumn among species. Our study improves the understanding of phenological information in deciduous trees caught by digital cameras and provides insights in relating the information from field to remote sensing.