Assessment of Remotely Sensed Land Surface Phenology Data for North America: Inter-comparison and Forecasting

Land surface phenology is widely used as a diagnostic of ecosystem response to global change and influences seasonal scale fluxes of water, energy, and carbon between the land surface and atmosphere. While many data sets related to plant phenology have been collected at specific sites or in networks focused on individual plants or plant species, a robust framework in classifying the ground phenological events to compare with remotely sensed phenology is lacking. In this research, we develop a methodology to compare the latest Collection 5 500m MODIS phenology data (cardinal dates corresponding to greenup, maturity, senescence and dormancy) with all available ground observed data for North America for the period 2001-2006. Inter-comparison results show that the MODIS phenology product performs well in capturing the “earliest events” corresponding to each of the phenological phases (e.g. budburst, first leaf, first leaf color change) of plant growth in a community. Of all the ground phenological stages, the leaf greenup events are best correlated to the MODIS greenup onset. Establishing confidence in the MODIS greenup onset data through our inter-comparison exercise and exploiting the long-term nature of the data set (2001-present) led us to build a seasonal phenology forecast model within the Terrestrial Observation and Prediction System (TOPS) framework. The available MODIS greenup data serves as a training data to our climate based phenology forecast model. We parameterize the empirical model for each 500m pixel based on the relationship between local climate provided by TOPS and MODIS greenup. As a test case, we have predicted the greenup for 2010 spring by using the seasonal climate forecast data from NECP Climate Forecast System (CFS) to our forecast model. The phenology forecast portal is now available online, which provides a unique platform for community participation in monitoring local phenology forecast events and uploading observed phenological events through the web-based interface.