Landscape phenology and habitat discrimination: random forests and MODIS time series imagery for E. multilocularis transmission host modelling

International audience Echinococcus multilocularis (Em), a parasitic tapeworm, is responsible for a significant burden of human disease across continental Asia. Here, we look to utilise the increasingly rich availability of satellite remote sensing data to examine the landscape characteristics which are specifically linked to the key disease transmission hosts of EM, Ochotona spp., to better understand host, and therefore potential transmission foci, distributions.A time-series of MODIS 16-day 250m Enhanced Vegetation Index (EVI) satellite data is used to quantify the seasonal vegetation dynamics across a study area in Serxu County, Sichuan Province, China, in relation to the presence of the Em intermediate host Ochotona curzoniae (plateau pika) and Ochotona cansus (Gansu pika) (here merged to Ochotona spp.). A series of derived phenological metrics are analysed using the random forests statistical method to determine the relative importance of seasonal vegetation characteristics. Results indicate negative relationships between Ochotona spp. presence and EVI showing a preference for low-biomass habitats. However, EVI values during green-up and senescence periods are also shown to be important, potentially resulting from improved detectability of low-biomass grassland habitats at these times.Improved detection of Ochotona spp. preferred habitats via time-series EVI imagery offers better understanding of the distributions of this Em host, and the potential for monitoring the changes in Ochotona spp. optimal habitat distributions resulting from landscape change. This could aid the identification of villages at increased risk of infection and how this risk changes over time, aiding the development of preventive strategies.This study illustrates a successful example of using high temporal frequency satellite data in combination with in-situ ecological survey data within an epidemiological context. However with the increasing availability of free satellite data of increasing spatial detail and temporal coverage from Landsat OLI and the Sentinel series of satellites, there is significant potential in applying these methods to other landscape-linked disease transmission hosts both for Em and other pathogens globally.