Spectral reflectance characteristics of arctic vegetation

Paul Budkewitsch, Canada Centre for Remote Sensing, Natural Resources Canada, now at Aboriginal Affairs and Northern Development Canada. Pak-Yau Wong: Canadian Museum of Nature. Karl Staenz: Canada Centre for Remote Sensing, Natural Resources Canada, now at University of Lethbridge. Robert Hitchcock: Prologic Systems Ltd, now at DRDC. Éric Gauthier: MIR télédétection Inc., now Agriculture and Agri-Foods Canada

RGB image of the region of interest

RGB image of the region of interest, a 1.7 x 1.7 km area in northern Baffin Island. Vegetation appears in green in this false-colour image and is concentrated mainly along a network of small valleys in the lower left portion of the image.

 

Spectral characteristics of willows, mosses and sedges are sufficiently different in the range of 440 to 2450 nm that they can be distinguished in hyperspectral data, or imaging spectrometer data sets.  To test the capability of imaging spectroscopy for mapping the distribution of different plant communities, a PROBE-1 airborne hyperspectral data cube collected in July of 1999 over a sparsely vegetated area of northern Baffin Island near Arctic Bay, Nunavut, was processed to isolate spectrally distinct end members. The Iterative Error Analysis unmixing method employed is data dependent and two vegetation end members of different spectral characteristics (Type A and B) were identified within the region of interest.

 

Type A is representative of a mesic heath, dominated by arctic willow (Salix arctica), whereas Type B is recognised as a moss-sedge meadow in a meso-hydric zone where mosses are mainly Ditrichum fleixcaule and sedges are made up of several species of Carex.  Field investigation of these two vegetation end members was carried out where they occurred.  The use of imaging spectrometer data is shown to be effective for identifying and mapping some types of arctic vegetation. Although no attempt was made to conduct a complete botanical survey of the study area, a good understanding of the spectral end members obtained in our analysis was aided by ground-based spectral measurements of single plant species.

Averaged field spectra for three different plant varieties, see text version.
text version
Averaged field spectra for three different plant varieties, (a) arctic willow (Salix arctica), (b) green moss (Ditrichum flexicaule), and (c) sedge (Carex sp.,). Spectra are plotted off set from one another for clarity. Horizontal scale lines are spaced at 10% intervals.

The purpose of this work is two-fold.  First, ground spectral measurements of a few common arctic vegetation types, appropriate to the 1-10 m scale of detection by methods using imaging spectrometer data, are presented and their salient features discussed.  Second, results from mapping the principal vegetation types in imaging spectrometer data is demonstrated for a small region of interest  in the Canadian High Arctic. 

Examples of arctic vegetation as described in the text below.

A:  Prostrate arctic willow (Salix arctica).  Ground covering beneath the new plant growth is mainly leaf litter from the previous year. B:  Green moss (mainly Ditrichum flexicaule) forming a 6-8 cm thick, dense mat on moist soil.  C:  A relatively dense growth of sedges (Carex sp.) share the ground cover with lesser amounts of Salix arctica, Eriophorum sp., other vascular plants and an under carpet of several bryophyte species.

Recognising the density and spatial distribution of arctic plant communities is essential for many environmental baseline studies and can provide data for detecting change, estimating biomass, and monitoring the overall ecological health of different regions.  The use of imaging spectrometer data can be effective for identifying and mapping some basic types of arctic vegetation.  Since the spectra collected by imaging spectrometers covers several square metres on the ground, the spectral response is often a function of a mixture of different types of plants.  Using spectra from ground-based measurements of single plant species, or linear combinations of associated vegetation, one is able to obtain a better understanding spectra from imaging spectrometer data in these environments. Data analysis can therefore greatly benefit from a spectral library of known vegetation types.