Species in a dynamic landscape – a spatio-temporal analysis using MODIS time-series

Abstract

Can satellite imagery in conjunction with a meta population model provide further relevant parameters to improve the knowledge about broadscale spatial biodiversity patterns in West Africa? For a better understanding of the spatial distribution and the explaining variables of biodiversity, a correlation between broadscale, longterm environmental data and species' habitat requirements is indispensable. High temporal remote sensing provides this information globally with sufficient spatial resolution. However, the explanation is limited to coarse spatial patterns, also due to restrictions in manpower for ground based assessments of additional environmental parameters. Here we focus on ecologically elevant products which are derived from remote sensing imagery, i.e. habitat fragmentation and habitat loss, land cover dynamics and change. These parameters are crucial for ecological research and resemble environmental indicators which have already been proven to be important for species assemblage, richness and population survival. The use of high temporal and moderate spatial resolution satellite imagery (e.g. MODIS) allows the construction of timeseries using several higher level products, such as the Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), Leaf Area Index (LAI), or Land Surface Temperature (LST). These datasets provide information on phenology and, compared for several years, land cover change and temporal anomalies. Moreover, using the quality layer of MODIS, quality assessments of the respective remotely sensed data can be conducted and unwanted distortions like clouds, haze or shadow can be masked. With this approach the quality of other temporal datasets (e.g. NOAA AVHRR NDVI) is evaluated. In contrast to monotemporal data, multitemporal variables provide information regarding land cover dynamics and transformations. Considering the historical archives of the NOAA AVHRR sensor dating back to the early 1980ies, past events can be taken into account to complement the analysis of present land cover patterns. Based on these data sets land cover information can be derived and validated. Meta population models using algorithms by Hanski and Ovaskainen were implemented in the Geographic Ressource Analysis Support System (GRASS) and applied on MODIS timeseries. The required species settings for habitat requirements and dispersal capabilities were provided by zoologists for each species. The model provides a value for the importance of each habitat patch for spatial distribution of the species or functional group in question. Patches can thus be classified as possible sources, sinks or stepping stones for dispersers in a metapopulation. Using MODIS timeseries imagery with an appropriate temporal and spatial resolution for the taxa under investigation the annual and inter annual dynamic of the meta population value of each patch are analysed. By applying meta population models on timeseries satellite data it is feasible to delineate areas of high importance for the persistence of populations in a land scape.