Mapping the floristic continuum: Ordination space position estimated from imaging spectroscopy

Abstract

Objective: To present a non-classificatory technique of map representation of compositional patterns of vegetation as no two plant species assemblages are completely alike and gradations often occur. Variation is depicted as continuous fields instead of classes. Location: Murnauer Moos, Bavaria. Methods: The study combined vegetation ecology and remote sensing methods. The gradual representation of compositional patterns was based on techniques of ordination and regression, instead of mapping class fractions. The floristic field data were collected in relevés and subjected to three-dimensional non-metric multidimensional scaling (NMS). The reflectance information corresponding to plots was gathered from remotely sensed imagery with a high spectral resolution. Reflectance values in numerous wavelengths were related to NMS axes scores by partial least squares regression analysis. The regression equations were applied to the imagery and yielded three grey-scale images, one for each ordination axis. These three images were transformed into a red, green, and blue colour map with a specific colour for each position in the ordination space. Similar colours corresponded to similar species compositions. Results: Compositional variation was mapped accurately (R 2 = 0.79), using continuous fields. The results took account of various types of stand transitions and of heterogeneities within stands. The map representation featured relatively homogeneous stands and abrupt transitions between stands as well as within-stand heterogeneity and gradual transitions. Conclusions: The use of NMS in combination with imaging spectroscopy proved to be an expedient approach for nonclassificatory map representations of compositional patterns. Ordination is efficiently extended into the geographic domain. The approach in abandoning pre-defined plant communities is able to reconcile mapping practice and complex reality.

TerraSAR-X dual-pol time-series for mapping of wetland vegetation

“…Regarding the grassland classes, the classification obtained in this study is as accurate as those obtained in studies comparing vegetation and micro-topography using LIDAR data to classify wetland areas (Moeslund et al, 2011). Several studies indicated that vegetation species in wetlands can be identified using hyperspectral data (Schmidtlein et al, 2007). From the perspective of this study, this method could be assessed to classify wetland vegetation at the dominant-species level using TerraSAR-X data.…”

The use of remote sensing in soil and terrain mapping — A review

“…The use of biogeographical gradients in non-linear ordinations combines the information from vegetation ecology and remote sensing methods (Schmidtlein et al, 2007). The approach entails analysing trends present in sets of floristic variables and establishing the nature of interrelationship between them (Armitage et al, 2004).…”

“…Reflectance values have been related to ordination axes by PLSR and the resulting regression equations were then applied on the spectral image. By this method the compositional variation was mapped for an area of 25 ha (R 2 = 0.79), using continuous fields (Schmidtlein et al, 2007). Such biogeographical patterns can be related to precipitation, temperature and soil conditions on a regional scale (Mahecha and Schmidtlein, 2008;Schmidt and Hewitt, 2004).…”

“…SCORPAN includes the same factors as CLORPT but also spatial (cross) correlation of soil properties and presence of spatially autocorrelated errors (McBratney et al, 2003). Several factors of soil formation can be derived from remotely sensed data (Buis et al, 2009;French et al, 2005;Schmidtlein et al, 2007;Singhroy et al, 2003).…”

Hyperspectral remote sensing of peatland floristic gradients

“…The spatial heterogeneity of a peatland surface is such that in practice, these requirements are very difficult to fulfil. Combined ordination and regression approaches have been developed to try and overcome some of the limitations of mapping approaches that require vegetation to be categorised a priori (Schmidtlein, Zimmermann, Schüpferling, & Weiβ, 2007) and the need for homogenous plant coverage at the spatial resolution of the sensor. Typically, ordination methods are used to assign numerical values (i.e., ordination scores) to plot-level species data, which relate to the level of floristic similarity or dissimilarity between plots.…”