Detection of gradients of forest composition in an urban area using imaging spectroscopy

Forests play an important role in urban environments by providing a range of ecosystem services. However, forest composition and function in urban ecosystems are ecologically complex because of high heterogeneity of the landscape and long legacies of intensive human management. Accurate quantification and mapping of composition of tree taxa in urban forests using remote sensing require high spatial and spectral resolution imagery. As such, mapping of individual species over large areas is impractical, so instead we map gradients of composition based on canopy traits derived from AVIRIS (Airborne Visible/Infrared Imaging Spectrometer) and lidar. This trait-based approach depicts emergent functional overlaps due to pixel-scale mixing of disparate species or genera found in the region. This provides a basis for understanding forest functioning in urban ecosystems where multiple taxa are functionally and spectrally similar. We used non-metric multidimensional scaling (NMDS) to ordinate multi-dimensional forest composition data, where NMDS scores (NMDS1 and NMDS2) represent compositional gradients. Predictive models of NMDS1 and NMDS2 were built using AVIRIS-derived foliar traits such as nitrogen concentration and lidar structural variables. Gradients of composition were strongly related to AVIRIS-derived traits (R2 = 0.67 for NMDS1 and R2 = 0.47 for NMDS2). Lidar-derived structural variables provided no improvement to prediction of the first NMDS axis estimates, and a minimal improvement (3% more variance explained) on NMDS2. We applied predictive models of NMDS1 and NMDS2 to maps of predictor traits to map NMDS1 and NMDS2. From the NMDS1 and NMDS2 maps, we mapped potential composition patterns and associated uncertainties by genus from the ordination of the field data. Our maps of composition gradients show the potential for data from imaging spectrometers such as HyspIRI to comprehensively characterize the diversity of ecosystems in heterogeneous landscapes. Moderate resolution satellite imagery may not facilitate detection of individual species, but imaging spectroscopy provides a basis to estimate the potential distribution of particular taxa based on assemblages associated with foliar traits that can be derived from imaging spectroscopy. Comprehensive satellite-based mapping will provide basic understanding and spatially explicit information on forest functioning suitable for monitoring and management of urban ecosystems.