Developing distribution models

Phytophthora data
Distribution data for Rhododendron and bilberry are available via the National Biodiversity Network Gateway.  Although the coverage for bilberry is likely to be good at the 10km square level, it is too common to have been recorded in greater detail.  Similarly, coverage for rhododendron is probably less consistent, even at the 10km square level, because non-native species have not been consistently recorded.  Unfortunately data at 10km square level is not sufficient for establishing detailed impact or identifying areas at risk. 


The distribution data, together with a number of other datasets, such as altitude, weather and landcover Rhododendron and Vaccinium data overlapwere fed into a ‘Maxent’ model to create modelled distributions of the species with a precision of 1km2 or better.  In order to be able to test the fit of the models, they were developed using a random selection of half of the data, and tested with the other half.  The distributions developed produced a fit of 94% for Rhododendron and 93% for bilberry.


These models can then be used to look at the likelihood of co-occurrence of the two species.  Different thresholds can be chosen for different purposes, for example to guide inspections or control measures – the map on the left shows a threshold where the predicted probability of both Rhododendron and bilberry is greater than or equal to 30%.  Threshold maps are important to help make decisions on where control measures are most important, and depending on the control chosen, where that control might be sensitive.