Temporal variability in the thermal requirements for vegetation phenology on the Tibetan plateau and its implications for carbon dynamics

Abstract

Static thermal requirements (T _req ) are widely used to model the timing of phenology, yet may significantly bias phenological projections under future warming conditions, since recent studies argue that climate warming will increase T _req for triggering vegetation phenology. This study investigates the temporal trend and inter-annual variation of T _req derived from satellite-based spring and autumn phenology for the alpine and temperate vegetation on the Tibetan Plateau from 1982 to 2011. While we detected persistent warming in both spring and autumn across this time period, we did not find a corresponding long-term increase in T _req for most of the study area. Instead, we found a substantial interannual variability of T _req that could be largely explained by interannual variations in other climatic factors. Specifically, the number of chilling days and fall temperature were robust variables for predicting the dynamics of T _req for spring onset and autumn senescence, respectively. Phenology models incorporating a dynamic T _req algorithm performed slightly better than those with static T _req values in reproducing phenology derived from SPOT-VGT NDVI data. To assess the degree to which T _req variation affects large-scale phenology and carbon cycling projections, we compared the output from versions of the Terrestrial Ecosystem Model that incorporated static and dynamic T _req values in their phenology algorithms. Under two contrasting future climate scenarios, the dynamic T _req setting reduced the projected growing season length by up to 1–3 weeks by the late twenty-first century, leading to a maximum reduction of 8.9 % in annual net primary production and ~15 % in cumulative net ecosystem production for this region. Our study reveals that temporal dynamics of T _req meaningfully affect the carbon dynamics on the Tibetan Plateau, and should thus be considered in future ecosystem carbon modeling.

http://ift.tt/2cpHwoJ