Risk mapping

eRiskMapper software has been developed by researchers from the University of Oxford to make predictions about species distributions and/or abundance from sparse point or administration level data, such as field observations and records.

The technique can also be applied to make predictions about the likely future distribution of species (pests, crops, plants etc) under various scenarios of climate change.

eRiskMapper makes possible the production of risk maps by non-expert software users such as biologists, medics and conservationists.

It also allows users to explore their data simultaneously in database, geographical and environmental spaces, thus making clear the linkages between these reference frames.

eRiskMapper selects a subset from a very large set of potential predictor variables (often derived from satellites) that best discriminates the presence/absence or abundance data classes.

Discriminant analysis

This is the first user-friendly package to use non-linear discriminant analysis for modelling species’ distributions.

Additional functionality allows users to explore their data and the model predictions in full.

Linear discriminant analysis was developed by Ronald Fisher, who sought to distinguish three varieties of lilies on the basis of similar measurements taken from each. It assumes similar covariances between the predictor variables.

Importantly, non-linear discriminant analysis allows dissimilar covariances between clusters of points representing presence and absence sites, thus allowing much greater flexibility in the fitting algorithm.

The statistics underlying this are clear, but making it user-friendly to biologists has involved considerable development over the past two decades.

Development

In creating eRiskMapper, the Oxford researchers have developed a series of methods for selecting the predictor variables, deployed a bootstrapping approach to risk mapping and created a new algorithm for risk mapping using polygon rather than just point data.

Text file and other outputs have also been used to make the results clear. The importance of each predictor variable has also been identified.

Licensing

Oxford University Innovation would like to talk to companies that are interested in licensing this technology. Request more information if you would like to discuss this further.