Canopy traits in remote sensing of floristic and functional vegetation patterns
Final Report Abstract
The reflectance signal of vegetation stands is determined by optical traits. Inter-correlations between these biochemical (pigments, leaf water and dry matter content) and biophysical (leaf and canopy structure) properties enable a remote-sensing based analysis of compositional and structural vegetation patterns. The present project aimed to answer the following questions related to the role of optical traits in remote sensing of vegetation: a) How do canopy optical traits contribute to statistical relations between floristic vegetation patterns and reflectance? b) What are the relations between optical traits and functional traits of a vegetation stand? These questions were addressed in multiple case studies in open, temperate ecosystems. Question a) was addressed with pre-existing time series of spectral data sampled in three open, temperate ecosystems. A statistical inversion of the leaf and canopy optical properties model PROSAIL was used to retrieve optical trait values for each field plot and date from the spectral data. The species composition of the plots was described through a gradient analysis. We analyzed relations between the extracted floristic gradients and the retrieved optical traits as well as the spatio-temporal variation of these relations. Finally, the predictive power of the retrieved optical traits for an explanation of the variation in the compositional data was assessed. As benchmark, we used the predictive power of the original spectral data. Further, we developed a statistical approach that enables a robust assessment of the importance of individual wavelengths in regression models. This approach is based on an ensemble of different regression techniques and allows to draw conclusions on the contribution of optical traits to statistical relations between spectral data and compositional patterns. Question b) was addressed for the example of pollination types in the Ammer catchment. We sampled optical traits and information on pollination types in the field and complemented these data with information from trait data bases. Regression models and correlation analyses enabled to analyze relations between pollination types and optical traits as well as between pollination types and canopy reflectance. Problems encountered were mostly related to issues during data acquisition (unsuitable land use, structural damage to the institute building that affected the trait analysis) and led to some changes in the selection of study sites and the acquisition of trait data. Despite these problems, the project aims could be fulfilled. The results of both studies show that: I) almost all tested optical traits contributed to statistical relations between canopy reflectance and compositional and functional vegetation patterns, II) these contributions are subject to considerable spatio-temporal variation (analogue to the relations to different wavelength regions), III) optical traits are better predictors of compositional and functional vegetation patterns than canopy reflectance, and IV) ensemble modeling techniques allow for a robust assessment of important spectral wavelength regions (and thus indirectly of optical traits).
Publications
- (2014). Mapping the local variability of Natura 2000 habitats with remote sensing. Applied Vegetation Science 17, 765-779
Feilhauer H, Dahlke C, Doktor D, Lausch A, Schmidtlein S, Schulz G, Stenzel S
(See online at https://doi.org/10.1111/avsc.12115) - (2015). Multi-method ensemble selection of spectral bands related to leaf biochemistry. Remote Sensing of Environment 164, 57-65
Feilhauer H, Asner GP, Martin RE
(See online at https://doi.org/10.1016/j.rse.2015.03.033) - Mapping pollination types with remote sensing. Journal of Vegetation Science Vol 27 Issue 5, September 2016, Pages 999-1011
Feilhauer H, Doktor D, Schmidtlein S, Skidmore AK
(See online at https://doi.org/10.1111/jvs.12421)
