Complex network-based time series remote sensing model in monitoring the fall foliage transition date for peak coloration

Vegetation phenological events, especially peak foliage coloration, are among the ecological phenomena that are most sensitive to climate change. Compared to spring seasonally recurring events, fall phenology remains much less understood. Remotely sensed monitoring of fall phenology provides a wealth of opportunities to understand the underlying processes and mechanisms. However, the gradual change of foliage color in the fall season makes it challenging to remotely estimate critical phenological transition dates. Particularly, the transition date for foliage peak coloration cannot be adequately captured via conventional curve fitting-based phenological models. Also the lack of consensus among the conventional models makes it desirable to explore new remotely sensed representations of the fall phenological process. In this study, we developed an innovative complex network-based phenological model, namely “pheno-network”, to estimate the fall foliage transition date for peak coloration. The pheno-network model characterizes the phenological process through analyzing the collective changes of spectral signatures along the temporal trajectory. A network measure, moving average bridging coefficient, is newly designed to estimate the phenological transition date. With Harvard Forest and Hubbard Brook Forest as reference sites, the results demonstrated that the transition date estimated through the devised pheno-network model corresponds well with the peak coloration period of the reference sites. The unique structure of the pheno-network formulated via spectral similarities differentiates the various roles of vegetation spectral signatures at different phenological stages. This study is the first attempt at introducing network science to time series remote sensing in modeling the complex phenological processes of vegetation. The innovative network-based phenological representation shows great potential in improving remotely sensed phenological monitoring and shedding light on the subsequent modeling of vegetation phenological responses to climate change.