Estimation of grasslands species diversity from their spectro-temporal heterogeneity using satellite image time series with high spatial and temporal resolutions

Abstract

Grasslands represent a significant source of biodiversity in farmed landscapes because of their plant and animal diversity. However, this biodiversity is threatened by the intensification of agriculture. It is therefore important for ecologists and conservation scientists to monitor biodiversity on large spatial scales. Satellite remote sensing constitutes a useful tool to inform biodiversity over large extents thanks to the broad spatial coverage of sensors. However, until recently, the study of grasslands in fragmented landscapes, such as found in Europe, has been limited because of sensors’ low resolutions. Indeed, grasslands are rather small elements in the landscape which require a high spatial resolution to be detectable. New generation satellites offer new opportunities for grassland’s monitoring because they provide combined high spatial and temporal resolutions images at no cost thanks to the ESA free data access policy. In remote sensing of biodiversity, the Spectral Variation Hypothesis (SVH) [1, 2] assumes that the spectral heterogeneity measured in the image is related to the spatial heterogeneity of the habitat the image represents. The diversity of species being related to the heterogeneity of the habitat [3], the spectral heterogeneity can be used as a proxy for species diversity [1]. In this study, we hypothesis that the grassland’s species differ in their phenology and, hence, that the temporal variations measured from satellite image time series (SITS) can be used in addition to the spectral variations. We propose new spectro-temporal indices derived from SITS as proxies to estimate the species diversity in grasslands. Our method to assess the spectro-temporal heterogeneity is based on a clustering of grasslands using a robust technique [4] suitable for high dimensional data issued from SITS. We tested the method on 192 grasslands from southwest France using an intra-annual time series of 18 SPOT5 satellite images. The results show that our indices – the entropy and the intra-class variability – explain better the variance of the Shannon’s index of grasslands than the commonly used Mean Distance to Centroid [2] that does not require an a priori clustering. However, there is no significant improvement when using the temporal variations in addition to the spectral heterogeneity. The results suggest that the temporal variations measured from SITS may be more related to the effect of management practices in grasslands. Limiting the SITS to a period when no management practices occur may alleviate this effect. We also suggest to extend the SVH to the functional diversity, by using functional traits related to the phenology of species to further exploit the potential of SITS issued from new generation satellites. [1] M. W. Palmer, P. G. Earls, B. W. Hoagland, P. S. White, and T. Wohlgemuth. Quantitative tools for perfecting species lists. Environmetrics, 13(2) :121–137, 2002. [2] D. Rocchini, N. Balkenhol, G. A. Carter, G. M. Foody, T. W. Gillespie, K. S. He, S. Kark, N. Levin, K. Lucas, M. Luoto, H. Nagendra, J. Oldeland, C. Ricotta, J. Southworth, and M. Neteler. Remotely sensed spectral heterogeneity as a proxy of species diversity : Recent advances and open challenges. Ecological Informatics, 5(5) :318 – 329, 2010. Special Issue on Advances of Ecological Remote Sensing Under Global Change. [3] J. Wilson, S. J. Fuller, and P. B. Mather. Formation and maintenance of discrete wild rabbit (Oryctolagus cuniculus) population systems in arid Australia : Habitat heterogeneity and management implications. Austral Ecology, 27(2) :183–191, 2002. [4] C. Bouveyron, S. Girard, and C. Schmid. High-dimensional data clustering. Computational Statistics & Data Analysis, 52(1) :502 – 519, 2007.