Disentangling the effects of climate change, landscape heterogeneity, and scale on phenological metrics

Phenology, the study of the timing of cyclical life history events and seasonal changes, is a fundamental aspect of how individual species, communities, and ecosystems will respond to climate change. Both biotic and abiotic phenological patterns are changing rapidly in response to changing seasonal temperatures and other climate-related drivers, and the consequences of these shifts for individual species and entire ecosystems are largely unknown. Landscape-scale simulations can address some of these needs for better predictions by demonstrating how phenology measures can vary with spatial and temporal grain of observations, and how phenological responses can vary with landscape heterogeneity and climate drivers. To explicitly examine the spatial and temporal scale-dependence of multiple phenology measures, we constructed simulated landscapes populated by virtual plant species with realistic phenologies and environmental sensitivities. This enabled us to examine phenology measures and environmental sensitivities along a continuum of spatial and temporal grains, while also controlling other aspects of sampling design. By relating measures of phenology calculated at a given spatiotemporal grain to average environmental conditions at that same grain size, we are able to determine observed environmental sensitivities for multiple phenological metrics at that spatial and temporal scale. We demonstrate that different phenological events change distinctly and predictably with spatial and temporal measurement scale, opening the way to incorporating scaling laws into predictions. Using plant flowering as our example, we identify that the timing of the beginnings or ends of an event (e.g., First Flower date, Last Flower date), can be especially sensitive to the spatial and temporal grain (or resolution) of observations. Our work provides an initial assessment of the role of observation scale in landscape phenology, and a general approach for incorporating scale-dependence into predictions of a variety of phenological time series.