Optimising acoustic monitoring for frogs in the Koondrook-Perricoota (2018)
Funding
Forestry Corporation of NSW, $29,623
Investigators/researchers
Dr Amelia Walcott, Dr Andrew Hall, Associate Professor Skye Wassens & Associate Professor Dale Nimmo
Research Theme(s)
Environmental Water
Description
Environmental water has been strategically delivered to the Koondrook-Perricoota Forest, an ecologically and culturally significant ecosystem in the Murray River, to improve and sustain wetland health. Gaining insight into how water-dependent communities respond to flow management is critical to making evidence-based natural resource management decisions.
Assessment of calling activity by frogs can yield useful information on both long term changes in species richness, as well as annual variability in community composition, species richness and breeding activity in relation to wetland inundation and environmental watering. A cost and time effective way of monitoring frog calling activity over long and continuous time frames is to use automated audio recording units.
However very large datasets can quickly accumulate and the subsequent data extraction process is complex. For large audio datasets, automated audio recognition is the most feasible method but this method requires significant time and expertise to produce accurate and meaningful insight. There are a number of technical challenges and ecological knowledge gaps that need to be addressed before automated audio datasets can be considered to produce consistent, reliable data suitable for the evaluation of the responses to environmental watering events.
NSW Forestry Corporation has collected a large audio dataset (since mid 2015 daily five minute recordings from 20 sites within the Koondrook-Perricoota Forest) with the objective to monitor frog call responses to the delivery/management of flow regimes. Manual extraction of a small subset of the audio dataset has previously been undertaken but ceased due to the excessive time restraints of this data extraction method.
The aim of this project is to gain a more comprehensive understanding of the dataset and improve the efficiency of the audio monitoring design. Its objectives are to:
- Develop species specific call recognition models
- Use the species call models to scan the entire dataset and extract data on a site by site basis
- Validate the extracted dataset
- Compare the manually extracted dataset with the automated recognition dataset
- Evaluate detection probability for each of the resident frog species given the current recording regime and develop statistical models to test probabilities under alternate modelling designs
- Provide recommendations on monitoring design to optimise sampling effort
Outcomes
The knowledge gained from this project is expected to be used for a more robust assessment of the relationship between frog calling activity and water management.
Contact
Dr Amelia Walcott email
CSU Albury-Wodonga
May 2018