A preliminary, mouse abundance model was constructed using artificial neural networks and data on trap success rates spanning a ten-year period in neighbouring catchments in Central Queensland. To predict trap success (percentage of traps that catch mice) in June, the model uses trap success rates in December and February, as well as rainfall in January/February and March/ April. With a relatively small number of years in which all these data are available, there is a danger than the network may be ‘overtrained’, whereby the network ‘memorises the data’ rather than minimising the underlying dynamics. The preliminary model will be evaluated in coming seasons. In addition, the data used to construct the Cantrill model will be used to construct an artificial neural network to see if these provide any increased predictive power for southern Queensland.
|Author||Scanlan, J. and Farrell, J.|
|Secondary title||13th Australasian Vertebrate Pest Conference|
|Place published||Wellington, NZ|