Scaling and Evaluation of Ecosystem Carbon Uptake Through Integration of Multi-Scale Remote Sensing with AmeriFLUX Field Observations
Scott
Ollinger, University of New Hampshire, scott.ollinger@unh.edu
(Presenting)
Mary
Martin, University of New Hampshire, mary.martin@unh.edu
Marie-Louise
Smith, USDA Forest Service, marielouise.smith@unh.edu
Andrew
Richardson, University of New Hampshire, andrew.richardson@unh.edu
Lucie
Plourde, University of New Hampshire, lucie.plourde@unh.edu
Jenkins
Julian, University of New Hampshire, julian.jenkins@unh.edu
David
Hollinger, USDA Forest Service, davidh@hypatia.unh.edu
Spatial patterns of carbon assimilation in terrestrial ecosystems are influenced by two important classes of vegetation variables: those related to canopy structure (e.g. foliar biomass and leaf area index) and those related to photosynthetic capacity (e.g. leaf nitrogen and pigment concentrations). Although EOS-era remote sensing instruments have greatly improved assessment of vegetation productivity, focus has been concentrated on detection of LAI and related structural attributes that are used in models designed to be driven by these variables. Information on plant traits related to photosynthetic capacity has lagged far behind and is typically only available through use of biome-specific look-up tables that lack important sources of spatial variability.
Here, we report on an investigation that examines the degree to which carbon assimilation in forest ecosystems can be related to both local and regional variation in canopy nitrogen. Field measurements collected at a diverse array of forested research sites within the AmeriFlux network have been combined with hyperspectral remote sensing data from the airborne AVIRIS and spaceborne Hyperion instruments. Resulting coverages of canopy nitrogen concentrations have been used to relate tower-based estimates of carbon assimilation capacity to canopy N for the local landscapes surrounding each tower. Results to date indicate a significant positive relationship between canopy N and GPPmax that cannot be attributed to co-variation with LAI. Because existing methods of canopy N detection are labor intensive and are restricted to small landscapes, a parallel investigation involves developing generalizeable canopy N detection methods that would enable more widespread application of these results. Results of this effort indicate that a single PLS regression equation can accurately predict canopy N concentrations at independent sites covering a wide range of site types. Further, we show that a substantial fraction of the variation in canopy N can be related to spectral features available from broad-band sensors.