Evaluating and optimising performance of multispecies call recognisers
for ecoacoustic restoration monitoring
Monitoring the effect of ecosystem restoration can be difficult and time
consuming. Autonomous sensors, such as acoustic recorders, can aid
monitoring across long time scales. This project successfully developed,
tested and implemented call recognisers for eight species of frog in the
Murray-Darling Basin. Recognisers for all but one species performed well
and substantially better than many species recognisers reported in the
literature. We achieved this through a comprehensive development phase,
which carefully considered and refined the representativeness of
training data, as well as the construction (amplitude cut-off) and the
similarity thresholds (score cut-offs) of each call template used.
Recogniser performance was high for almost all species examined.
Recognisers for C. signifera, L. fletcherii, L. dumerilii, L. peronii,
and C. parinsignifera all performed well, with most templates having ROC
values (the proportion of true positive and true negatives) over 0.7,
and some much higher. Recognisers for L. peronii, L. fletcherii and L.
dumerilii performed particularly well in the training dataset, which
allowed for responses to environmental watering events, a restoration
activity, to be clearly observed. While slightly more involved than
building recognisers using commercial packages, the workflows ensure
that a high quality recogniser can be built and the performance
fine-tuned using multiple parameters. Using the same framework,
recognisers can be improved in future iterations. We believe that
multi-species recognisers are a highly effective and precise way to
detect the effects of ecosystem restoration.