Spectral rarefaction: linking ecological variability and plant species diversity

Species rarefaction curves have long been used for estimating the expected number of species as a function of sampling effort and they represent a powerful tool for quantifying the diversity of an area from local (+¦-diversity) to regional scale (+¦- and +¦-diversity). Nonetheless, sampling species based on standard plant inventories represents a cost expensive approach. In this view, remotely sensed information may be straightforwardly used for predicting species rich sites. In this paper, we present spectral rarefaction, i.e., the rarefaction of reflectance values derived from satellite imagery, as an effective manner for predicting bio-diverse sites. We tested this approach in ten biogeographical subregions in Switzerland. Plant species data were derived from the Swiss GÇÿBiodiversity MonitoringGÇÖ programme (BDM), which represents species richness of Switzerland at the landscape scale by a systematic sample of 520 quadrats of 1 km +ù 1 km. Seven Landsat ETM+ images covering the whole study area were acquired. Species and spectral rarefaction were built and results were compared by Pearson correlation coefficient considering several sampling efforts (as measured by the number of sampled quadrats). Local +¦-diversity showed a similar pattern considering the ten biogeographical subregions while +¦- and +¦-diversity showed higher values for regions in the Alpine arc and lower values for plateau regions and Jura mountains on the strength of the higher ecological (and spectral) variability of the former areas. Meanwhile, positive correlations between species and spectral richness values were significant only after a certain amount of area was accumulated, thus indicating a scale dependence of the fit of satellite and species data. With this paper, we introduce spectral rarefaction as an effective tool in quantifying diversity at a range of spatial scales. Obviously, the achieved results should be viewed as an aid to plan field survey rather than to replace it. We propose to use worldwide available remotely sensed information as a driver for field sampling design strategies