Understanding climate-phenology coupling with remote sensing and ground observations

Background/Question/Methods

Plant phenology, the timing of recurring biological events, is one of the most visible fingerprints of climate change. In turn, changes in plant phenology have important implications in natural and social systems such as in agriculture, conservation, and public health. Here we addressed three issues on climate-phenology coupling. First, phenological mismatch has long been hypothesized but rarely quantified on a large scale. It is even less clear how human activities have contributed to this mismatch. We used remote sensing observations (MODIS) to systematically evaluate how plant phenological shifts have kept pace with warming trends at the continental scale, corroborated by independent ground observations. Second, previous models that assume linear climate-phenology relationships often fail to predict phenology accurately. We modeled the nonlinear dynamics of leafing phenology with a state-of-the-art data-driven model, Gaussian process empirical dynamic modeling (GP-EDM). We predicted land surface phenology measured from satellite remote sensing (MODIS) and near-surface camera imagery (PhenoCam). Last, a better mechanistic understanding of flowering and pollen phenology is needed to limit pollen exposure and reduce healthcare costs. We detected the flowering of pollen-producing tree taxa with high spatial resolution satellite imagery (PlanetScope) in several cities in the US and then investigated the relationship with climate.

Results/Conclusions

We gained a quantitative understanding of climate-phenology coupling on the spatiotemporal phenological mismatch, nonlinear mechanisms of phenology, and the drivers of relatively understudied flowering phenology. They represent progress in improving Earth system’s predictability. First, in northern mid-to high-latitude regions over the last three decades, we found evidence of a widespread mismatch between land surface phenology and climate where isolines of phenology lag behind or move in the opposite direction to climate. These mismatches were more pronounced in human-dominated landscapes, suggesting that not even some of the foremost responses in vegetation activity match the pace of recent warming. Second, the GP-EDM was effective in forecasting simulated and observed leafing phenology. These results provide empirical evidence that mechanisms of phenology are highly nonlinear. The model also allows us to detect key environmental drivers of phenology. Last, we employed a novel method of detecting flowers based on tree color change within the spring greenup phase during catkin development, maturation, and senescence. We detected fingerprints of flowering on the tree level and aggregated them to the city level, which was consistent with pollen count data. The large-scale flowering detection then enabled us to infer the influence of temperature and precipitation on flowering phenology.