The central question guiding research in the Kitzes Lab is how are species distributed across complex landscapes and how do human impacts drive these distributions? This question is one of the oldest in ecology and is foundational to research in fields as diverse as population biology, community ecology, landscape ecology, wildlife biology, and conservation biology.
For as long as this question has been asked, however, efforts to answer it have been limited by the fundamental challenge of observing rare and hard-to-detect species and events. Despite the explosive growth of ecological data due to volunteer efforts and citizen science, global biodiversity data today remain biased towards common, charismatic, and easily-observable species and communities.
To help address this challenge, our lab focuses our research in terrestrial bioacoustics, which combines automated biodiversity sensors and deep learning/AI models to dramatically increase our ability to understand the ecology of rare and hard-to-detect species. Our group both creates methods and tools to enable this research and applies our methods to previously unanswerable questions that span natural history, conservation, and spatial ecology.
Our group’s research focuses heavily on the development and application of bioacoustic methods for surveying biodiversity, including research on autonomous sound recorders, machine learning methods, and hierarchical statistical models. We have purchased over 3,000 inexpensive field recorders (AudioMoths), which we and our collaborators have used to record over 1 million hours of audio in the field. Our group is also heavily involved in the development of machine learning classifiers for wildlife vocalizations (using both neural networks and signal processing approaches) and the exploration of statistical frameworks for interpreting the resulting classifier output. Some of our currently active projects include:
- Developing and maintaining OpenSoundscape, our lab’s open source software for bioacoustic analysis.
- Training machine learning classifiers for many species of birds (particularly eastern forest birds, nocturnal species, and secretive marsh birds) and frogs (temperate and tropical).
- Exploring variations of occupancy models and protocols for expert annotation to draw robust, efficient research conclusions from uncertain machine learning classifier output.
- Designing an inexpensive and open source hardware and software platform for acoustic localization, which will allow us to spatially locate individual singing birds in the field, ultimately providing a means of estimating population sizes from automated acoustic recordings.
Natural History, Conservation, Ecology
Our methodological work provides a foundation for our research on natural history, conservation, and ecology across a range of species and landscapes. We are particularly interested in the effects of habitat disturbance, including both habitat loss and habitat restoration, on species of conservation concern. Some of our currently active projects include:
- Conducting long-term monitoring for Golden-winged Warbler, Wood Thrush, and Cerulean Warbler (along with other focal species) in Dynamic Forest Restoration Blocks in central and eastern Pennsylvania.
- Using automated acoustic methods to survey rare, endangered, and potentially recovering frog populations at field sites across North and South America.
- Using bioacoustics data to go beyond measuring only species presence by using audio recordings to survey ecological processes such as reproduction, territory settlement, and predation.
- Develop and use new bioacoustics datasets to drive new developments in spatial ecological theory, with a particular focus in macroecological patterns of species aggregation and turnover