Department of Defense (DoD) lands are critically important to migratory bird species as breeding sites, wintering sites, and as migratory stopover sites.  Often the majority of individual bird detections in surveys are acoustic and many of the birds noted are never seen.  Because acoustic detection plays such a prominent role in avian population monitoring, the use of automated acoustical recording instruments and signal detection and classification software has the potential to lead to improved monitoring of bird populations on DoD lands and elsewhere. Specifically, such techniques may enable more extensive sampling, improved estimates of the birds counted and missed, and improved estimates of the area surveyed.

Specific objectives of this project included:

    • ground-based acoustic censusing of species that vocalize infrequently
    • documenting variation in calling activity to improve the accuracy of all acoustic censuses and the value of historical data sets.
  2. Developing the critical hardware and software components for a network of acoustic detectors to monitor flight calls of nocturnally migrating bird species to document species-specific stopover use on and around DoD installations.

Technical Approach

The Cornell Bioacoustics Research Program has developed two interactive sound analysis software packages: eXtensible Bioacoustics Tool (XBAT) and Raven. Both programs incorporate interactive sound visualization, measurement, and annotation tools. Enhancements were made to both programs under this project to improve their utility for tasks such as detecting and classifying sounds from species of interest on DoD lands.


The following enhancements of XBAT’s data template detector were implemented:

  1. The ability to run multiple templates simultaneously was added, vastly improving processing speed.
  2. The use of rejection templates was added reducing false detection rates by an order of magnitude.
  3. Batch detection capability was implemented allowing the user to specify an arbitrary number of recordings to process them sequentially making it possible to use computing time efficiently.
  4. The band-limited energy detector (BLED) displays were enhanced with a diagnostic display that visualizes results of the several intermediate steps in the detector algorithm, enabling the user to rapidly and efficiently configure the detector for improved performance.  This enhancement enables the user to make targeted improvements to detector performance in a few minutes that previously could have taken hours of trial and error.

The following are enhancements made to Raven analysis software:

  1. Architecture used in Raven 1.2.1 was enhanced to achieve a high level of modularity and extensibility.
  2. A BLED plug-in was implemented using Raven Pro’s new plugin architecture.  The BLED typically processes these recordings at over 200 times real-time speed, so that an eight-hour recording is completely processed in slightly more than two minutes. This capability is crucial to the use of Raven as a real-time monitoring tool.



This project has made extensive use of software to automate processing of tens of thousands of hours of recordings, focusing on BLEDs as a means to extract flight calls and other vocalizations of interest as rapidly as possible. Research conducted through this project has improved the accuracy of bird abundance measures extracted from DoD field survey efforts and large historical data sets. The extended spatial and temporal scale of these acoustical monitoring techniques enable the monitoring of rare or otherwise cryptic species and demonstrates the feasibility of a network to monitor nocturnal migration traffic by species by identifying the birds from the distinctive characteristics of their flight calls.