Highlight
A comparison of multiple phenology data sources for estimating seasonal transitions in forest carbon exchange.
The Ameriflux Tower footprint at the University of Michigan Biological Station
Achievement/Results
The timing and length of terrestrial growing seasons are closely connected to climate, and the effects of climate on plant seasonal activity are qualitatively well known. The extent to which terrestrial ecosystems act as sources or sinks for CO2 is of great interest and a quantitative understanding of vegetation dynamics is needed to predict how terrestrial ecosystem carbon exchange will respond to climate change and climate – biosphere interactions. The timing and length of recurrent plant life cycle events (plant phenology) can be obtained from a variety of data sources. This information can then be compared with direct measurements of CO2 fluxes in order to relate phenology data to ecosystem carbon exchange.
There are currently numerous data sources available for estimating the timing of recurrent plant phenology transitions. This study compares measurements from several phenology data sources to understand the relationship between phenology metrics derived from these data sources and seasonal transitions in net ecosystem exchange (NEE). Regression analysis was used to compare 11 years of NEE from a deciduous broadleaf forest with phenology metrics derived from albedo, fraction of absorbed photosynthetically active radiation (fPAR), Plant Area Index (PAI), and MODIS normalized difference vegetation index (NDVI), enhanced vegetation index (EVI), and leaf area index (LAI) products of various spatial representation. No single source of phenological data was able to accurately describe annual patterns of flux phenology. However, for each transition in NEE (e.g., start of season, transition to net sink, etc.), the metrics from one or more data sources were significantly (p < 0.05) correlated with the timing of these recurring events. LAI-type measurements from satellite- and ground-based instruments were most frequently identified as having the highest performance (R2 ranging from 0.60 to 0.91) for estimating phenology transitions relative to other types of data sources. Mean errors between the estimated and observed date of occurrence ranged from a couple of days to several weeks depending on the data source. The results of this study highlight the relative strengths and weaknesses of each phenology data source for directly estimating seasonal transitions and interannual trends in carbon flux phenology of a deciduous forest.
Overall, the results demonstrate that by using multiple sources of phenology data, it is possible to estimate the timing and interannual variability of several important seasonal transitions in carbon flux phenology (CFP). The use of CFP rather than leaf phenology is beneficial particularly within the context of atmospheric models where vegetation is not represented by its explicit structure, but as a set of parameters that determine the effect of the vegetation on atmospheric processes. This is the case for regional and global land-surface and ecosystem models, which drive the land surface response in many global and regional climate models. In these models NDVI and EVI are used, either as prescribed variables or as a prognostic variable. It is also needed in high resolution models where remote sensing-driven virtual canopies at different phenological states are used for computer-based studies of forest-atmosphere interactions. Future research could go beyond the prediction of leaf-based phenology by more directly investigating changes in ecosystem metabolism linked to climate-vegetation interactions using the numerous available data sources. Furthermore, broad-scale, long-term, multi-site phenology and NEE data can be used to determine combinations of proxy metrics, which when put together, optimize the detection of long term CFP trends.
Address Goals
Engagement of graduate students in the application of multiple areas of expertise to the investigation of an urgent science question that results in new and significant understanding and which improves our ability to meet climate-ecosystem interaction prediction goals.
