Research in the Kitzes Lab focuses on measuring, understanding, and predicting biodiversity loss on a planet increasingly dominated by human activities. The central question that guides our research is

How does human alteration of natural habitat impact species abundance and diversity at large spatial scales?

Research projects in our lab naturally divide into three complementary themes: bioacoustics, conservation, and spatial macroecology. Many of our projects lie at the intersection of at least two, if not all three, of these areas.


We believe strongly that gaining an accurate understanding of how global change affects biodiversity will require a radical expansion in data on species abundance, diversity, and distributions. Human-driven field surveys, even with the often heroic efforts of citizen scientists, have not been able to scale up enough to cover even a small fraction of a percent of the planet’s surface.

In response to this problem, our lab’s research focuses on the development and use of automated acoustic survey methods for studying biodiversity. We have purchased nearly 3,000 inexpensive field recorders, which we and our collaborators have deployed across more than a dozen field sites around the world. Our lab is heavily involved in developing and training machine learning models that can automatically identify the species present in these field recordings. We are also developing GPS-synchronized field recorders that can be used to locate individual organisms in the field by triangulating their location using sound.

500 Bird Classifiers

We are currently developing automated song classification models for approximately 500 North American breeding bird species. These classifiers are convolutional neural networks (CNNs) that use image classification approaches to identify species-specific vocalizations. Our models are trained using labeled data from public sources that are expanded by various data augmentation routines. These trained models will be released under an open source license, along with a Docker container to allow users to apply these models to their own data sets. This work is financially supported by a Microsoft and National Geographic AI for Earth Innovation Award.

Acoustic Localization

An important limitation of traditional single-microphone acoustic surveys is that they are only able to determine species presence or absence, not abundance. We are currently developing a combination of inexpensive, open source hardware and software that will be able to spatially locate individual singing birds, providing a means of estimating population sizes from automated acoustic recordings. Our recorders are extended versions of the open source AudioMoth that includes a GPS receiver for clock synchronization. Our analysis methods use a simultaneous classification and segmentation process, followed by generalized cross correlation, allowing for hyperbolic sound source localization. We are partnering with Dr. Beth Gardner (University of Washington), Dr. Steve Latta (National Aviary), and Dr. John Wenzel and Luke DeGroote (Powdermill Nature Reserve) to pilot this platform. This work is supported by a National Science Foundation Infrastructure Innovation in Biological Research grant.

Nocturnal Flight Calls

In collaboration with Dr. Phil Taylor (Acadia University), John Kearney, and others, we are developing automated classification models for the nocturnal flight calls of many North American bird species. We are building on prior experiences of the BirdVox project, specifically focusing on increasing the number of species represented and improving accuracy on typical field recordings. This work is financially supported by the Natural Sciences and Engineering Research Council of Canada, Environment and Climate Change Canada, and others.

Frog Calls

In collaboration with several research groups, we are developing pilot classification models to identify calls of frogs of North America, Panama, and Borneo. We have developed several new approaches for this project to complement the convolutional neural networks (CNNs) used in our other projects. These alternative methods include a pulse rate-based classifier and a tone-based filter method.


One of our lab’s primary missions is to support biodiversity conservation through better measurement tools, empirical data, and predictive models. We are particularly interested in the effects of habitat disturbance, including both habitat loss and modification, on species of conservation concern. In our conservation-focused research, as compared to our more general work in spatial ecology, we generally focus on a specific species of concern, inhabiting a single location of interest, and subject to a specific type of disturbance

Within this broad category, we have particular interests in (a) the study of species that are rare or hard to detect, for where the long sampling periods enabled by automated acoustic recorders can play an important role in obtaining accurate censuses, and (b) landscape-scale studies that compare species abundance or diversity across different management regimes or habitat states.

American Prairie

In collaboration with Dr. Bill McShea (Smithsonian Conservation Biology Institute) and his research team, we are conducting several acoustic biodiversity surveys at the American Prairie Reserve in central Montana. Our main focus at the reserve is studying the response of local frog species to changes in grazing management, and we are also supporting projects involving breeding birds and swift fox populations at the reserve.

Ivory-Billed Woodpecker

In collaboration with Dr. Steve Latta (National Aviary), we have provided support to an ongoing project in southern Louisiana that is searching for ivory-billed woodpeckers, a species listed as critically endangered that many believe to be extinct. We have loaned acoustic recorders to this effort and developed classifiers and filters specifically designed to identify ivory-billed vocalizations in field recordings.

California Roads and Agriculture

Our earliest conservation-oriented field work took place in northern California, where we used bioacoustics approaches to study local bat species. We specifically examined the effects of large roads on bat foraging activity and patterns of foraging activity in mixed vineyard landscapes.

Spatial Macroecology

Beyond the particular details of individual species and habitats, we have long been interested in the “universality” of spatial biodiversity patterns. The regularity of patterns such as species-area relationships and species-abundance distributions, which appear to take common forms across communities, suggest the presence of deeply shared similarities among communities.

In the macroecological tradition, we largely focus on the idea that many of these regularities emerge as statistical properties of large, complex systems. Our work focuses on characterizing these regular patterns, describing these patterns quantitatively, relating multiple patterns to each other, and proposing theory and models to explain their emergence. We believe that a statistical mechanical view of ecological communities will ultimately help to explain many observable biodiversity patterns and enable more robust prediction of how they may change under disturbance.

BCI for Birds

There are relatively few high quality data sets available for deep testing of spatial ecological theory. The Barro Colorado Island (BCI) tropical forest plot, and the expanded network of CTFS plots, continue to play an outsized role in our understanding of spatial biodiversity patterns. We are currently using bioacoustics methods to collect two data sets that we hope will play a similar role in future theory testing. The first is a 20 km road transect at Sproul State Forest, in northern Pennsylvania, where we are collecting 3 hours of audio each morning throughout the field season at 200 consecutive points. The second is a 150 recorder grid placed at the Smithsonian Conservation Biology Institute in Front Royal, Virginia. We plan to identify breeding bird species across all recorders at these sites, providing simultaneously fine grain and large extent, temporally-explicit data on bird populations at these locations. These data sets will both be released under open source licenses.

Spatial Scaling

We have pursued several theory-based approaches to predicting patterns involving spatial scaling, which describe how species abundance or diversity change as a surveyed area grows or shrinks. We have published several efforts to upscale species richness from small plot surveys to regional scales using data from Panama, Borneo, and India. We continue to actively investigate the relationship between spatial scaling metrics and species turnover, two patterns that are often considered independently but that we believe are fundamentally linked.

Spatial Scaling and Extinction

Our work on spatial scaling has also focused on more applied questions of investigating species extinction following habitat loss. For this work, we have developed extended frameworks for the species-area relationship, used species-area relationships to estimate extinction debts, and investigated optimal reserve network design using stochastic simulation and optimization approaches.