Attributing the effects of climate on phenology change suggests high sensitivity in coastal zones

Abstract:

The impact of climate change on spring phenology depends on many variables that cannot be separated using current models. Phenology can influence carbon sequestration, plant nutrition, forest health, and species distributions. Leaf phenology is sensitive to changes of environmental factors, including climate, species composition, latitude, and solar radiation. The many variables and their interactions frustrate efforts to attribute variation to climate change. We developed a Bayesian framework to quantify the influence of environment on the speed of forest green-up. This study presents a state-space hierarchical model to infer and predict change in forest greenness over time using satellite observations and ground measurements. The framework accommodates both observation and process errors and it allows for main effects of variables and their interactions. We used daily spaceborne remotely sensed data from the Moderate Resolution Imaging Spectroradiometer (MODIS) to quantify temporal variability in the enhanced vegetation index (EVI) along a habitat gradient in the Southeastern United States. The ground measurements of meteorological parameters are obtained from study sites located in the Appalachian Mountains, the Piedmont and the Atlantic Coastal Plain between years 2000 and 2015. Results suggest that warming accelerates spring green-up in the Coastal Plain to a greater degree than in the Piedmont and Appalachian. In other words, regardless of variation in the timing of spring onset, the rate of greenness in non-coastal zones decreases with increasing temperature and hence with time over the spring transitional period. However, in coastal zones, as air temperature increases, leaf expansion becomes faster. This may indicate relative vulnerability to warming in non-coastal regions where moisture could be a limiting factor, whereas high temperatures in regions close to the coast enhance forest physiological activities. Model predictions agree with the remotely sensed observations of the enhanced vegetation index. These findings could be used in forest managements for identifying vulnerable forests based on their habitat type and hydrological status.