Spatio-temporal Modelling of Small Mammal Distributions Using Modis Ndvi Time-series Data

This work modelled the spatial distribution of the rodent species that act as hosts in the transmission cycle of the parasitic tapeworm Echinococcus multilocularis. The rodent distribution was modelled in relation to landscape characteristics in four ways, using (1) a Landsat ETM+ derived hard classification, (2) single-image Landsat ETM+ derived NDVI, (3) single-image MODIS 16-day composite NDVI, and (4) time-series MODIS 16-day composite NDVI imagery. The MODIS time-series imagery produced the strongest relationships and explained the highest percentage deviance of the relationships present (up to 41.4%), whereas the hard classification method only explained up to 21.2% of deviance. Single-image NDVI datasets produced poor results, with Landsat ETM+ derived NDVI explaining only up to 11.9% of deviance, and MODIS derived NDVI up to only 8.78%. These results confirm that using time-series NDVI data to model rodent distributions is a valid method, and can offer improved results over single date NDVI and hard classification methods.