Project Title:
Quantitative Identification of Missouri Bat via Acoustic Surveys

Project Description (short):
Echolocation calls have often been used by biologists to identify free flying bats, thereby allowing an investigator to survey an area without ever physically capturing a specimen. However, groups of closely related species often have similarly structured bat calls, and determining the presence of a particular species may not be possible. Accurate identification of free flying bats from their calls is dependent upon a reference library of known calls and a rigorous analytical methodology to differentiate species. This project will entail building a reference library of calls for Missouri bats (i.e., going out and recording their calls) and using statistical and analytical techniques such as discriminant function analysis to refine the analytical capabilities of current identification tools.

Skills needed:
Students should be familiar with basic laptop computer functions and comfortable in a field type setting working long hours into the night. Both students should also be comfortable with the idea of handling live animals; both will receive training in the proper handling of bats. The mathematics student should have completed Linear Algebra (MATH 357) and Statistics (STAT 290) by January 2005, and he or she must be willing to learn to program in Matlab. Additional exposure to statistical techniques (esp. discriminant function analysis) will be a bonus.

Start Date:
Literature review January 2007, Field work in Summer 2007 and Fall 2007

End Date:
January 2008

Mentors:
Prof. Scott Burt (Biology), sburt@truman.edu, Tel: 785-7133
Prof. Jason Miller (Mathematics), millerj@truman.edu, Tel: 785-7430

Accomplishments:

• Bay, Christopher. "Characterizing Bat Species via Their Echolocation Call's Wavelet Transform." Truman State University Student Research Conference. April 2003.
• Gustafson, Katie. "Discriminant Analysis and the Classification of Bat Species." Truman State University Student Research Conference. April 2004.
• Bay, Christopher. "W Wavelet-Based Model for Acoustic Identification of Bat Species." Truman State University Student Research Conference. April 2004.
• Clarke Cooper. "Automated Analysis of Bat Echolocation Calls." Truman Student Research Conference. April 2005.
• "A known call library for the quantitative identification of northeast Missouri bats." (with John Hainline*, Scott Burt, Joshua B. Kelly*, and Rachel O. VanAmburg*.) Annual meeting of the Central Plains Society of Mammalogists. Truman State University, Kirksville, MO. 15 October, 2005.
• "Quantitative identification of northeastern Missouri bats via acoustic and standard surveys." (With Joshua B. Kelly*, Scott Burt, John Hainline*, and Rachel O. Van Amburg*.) Annual meeting of the Central Plains Society of Mammalogists. Truman State University, Kirksville, MO. 15 October, 2005.
• "Quantitative identification of northeastern Missouri bats: survey results." (With Rachel O. Van Amburg*, Scott Burt, John Hainline*, and Joshua B. Kelly*) Annual meeting of the Central Plains Society of Mammalogists. Truman State University, Kirksville, MO. 15 October, 2005.
• Joshua Kelly, Rachel Van Amburg, Jason Miller, M. Scott Burt. "Results of Acoustic and Mist Net Survey of Swan Lake National Wildlife Refuge, Summer 2005." Technical Report.
• M. Scott Burt and Jason Miller. "Quantiative Approaches to Recognizing Bat Species via Acoustic Data." Bats and Caves Roundtable. Preliminary meeting to Missouri Natural Resources Conference. Lake of the Ozarks, MO. 1 February, 2006.
• Rachel VanAmburg and Josh Kelly. "Quantitative Identification of Northeastern Missouri Bats via Acoustic and Standard Surveys" Truman State University Student Research Conference. April 2006.
• MDC Grant for surveying Myotis sodalis near Kirksville region.
• Josh Kelly and Phil Vance. "Accurate Identification of Northeast Missouri Bats from Acoustic Data Alone." Undergraduate Poster Session. Joint Mathematics Meeting, New Orleans, January 2007.
• Kelly, Joshua. "Accurate Identification of Northeast Missouri Bats From Acoustic Data Alone" Truman State University Student Research Conference. April 2007.
• Kelly, Joshua and Vance, Phil. "Descriptive Statistics of Echolocation Calls from Northeast Missouri Bats" Truman State University Student Research Conference. April 2007.
• Phil Vance, Josh Kelly, Scott Burt, and Jason Miller. Poster: "Echolocation call variability among northeast Missouri bats." American Society of Mammalogists. Albuquerque, New Mexico. 6-10 June 2007.

Current Students:

• Joshua Kelly (Mathematics & Computer Science)
• Phil Vance (Biology)
• Ben Hale (Biology)
• Ryan Allen (Computer Engineering - Metropolitan Community College)

Past Students:

• Greg Knese, 2001-2002 (Mathematics)
• Chris Bay, 2002-2004 (Mathematics)
• Katie Gustafson, 2004 (Mathematics)
• Clarke Cooper, 2004-2005 (Computer Science)
• John Hainline, 2005 (Computer Science)
• Rachel Van Amburg, 2005 (Biology)
• Clayton Davis, 2006 (Mathematics)

About Prof. Burt:
Dr. Burt was born and raised in West Texas and recieved his B.S. and M.S. from Angelo State University and Ph.D. from the University of New Mexico. He came to Truman in the fall of 2000 and teaches Introductory Biology, Ecology, Animal Behavior, Conservation and Management of African Mammals, Ecology of Birds in Big Bend National Park, Mammalogy, and Advanced Field Mammalogy.

Throughout his academic career, he has been trained as a field mammalogist and as a natural history collections curator. He has had the great fortune to work with talented biologists in remote locations across the globe, from the Atlantic rainforest of southern Brazil to the grasslands and deserts of Mongolia. These experiences have shaped his reserach philosophy, and he is passionate about conserving biodiversity. He feels that he can contribute to the conservation of mammals by learning more about their past and present distributions, their habitat associations, and their basic natural history.

Dr. Burt has an active research program at Truman. Since his arrival, he has mentored three M.S. students and at least eight undergraduates in a multitude of projects. His students have studied the effects of disturbance on small mammal community structure, the ectoparasites of bats, morphological variability in shrews and rodents, bobcat ecology, hantavirus in Missouri rodents, habitat preferences for voles, echolocation call structure of bats, and prey selection of bats. He currently has three graduate students and two undergraduates involved in research, and over ten additional students that assist in the day to day activities in the mammal collection. He and his students regularly attend scientific conferences (regionally and nationally) to present the results of their work, and currently three papers are in review for publication, with three others in preparation.

About Prof. Miller:
Jason earned his B.A. in mathematics at Saint Olaf College, ventured to North Carolina to get his Ph.D. in mathematics, and headed back north to find real winters. He loves the outdoors (esp. paddling and hiking), music, food, and his Macintosh computers. When he has spare time, he likes to read fiction or essays. He has been at Truman for seven years, teaching mathematics and mentoring mathematics and computer science students in research. He is currently interested in applying his knowledge of the mathematical sciences to answer questions that are important to biologists.

Project Description (long):
In the past, to census bat populations, biologists relied on nets and traps to capture bats for positive identification. Recently, ultrasonic bat detectors have become a popular method of recording echolocation calls fromfree-flying bats. Features of the search-phase echolocation calls are often regarded as species-specific, and these features are use to infer the species of bat from its recorded echolocation call. Many advocate that inventories of bat communities be conducted using a combination of mist nets and acoustic surveys to inventory bat communities. Whether quantitative methods to infer bat species from acoustic survey data, or whether qualitative methods should be used is a matter of debate.

There are three methods commonly used for acquiring ultrasonic calls: heterodyne, time-expansion, and frequency division. Though time-expansion detectors acquire data with the most detail, they are expensive. Many field biologists who survey bat populations use the inexpensive Anabat II system (Titley Electronics Pty. Ltd., Ballina, NSW), which uses a frequency-division approach to acquiring data. The system employs a zero-crossing preprocessor produces time-frequency data for a call's harmonic of greatest amplitude, and displays the time-frequency sonogram of a call as it is recorded. The user trades detailed data for speed and affordability. This causes problems when trying to distinguish between the calls of two closely related species of bat, especially when one (e.g. the Indiana bat, Myotis sodalis) is a federally endangered species, and the other (e.g. the Little Brown bat, Myotis lucifugus) is very common. Confusing one for the other in an acoustic survey is a serious problem that can have unpleasant consequences for land managers.

Believing that robust and accurate quantitative methods give vastly more people access to reliable analysis of survey results, the proposed project will improved quantitative methods based on the Anabat II system.

RESEARCH OBJECTIVE: Of the fifteen native species of bat in Missouri, two are listed as endangered by the U.S. Fish and Wildlife Service, and one has not been observed in the state since 1971. In response to recent calls for help from those charged with protecting the Indiana bat, we will use an interdisciplinary approach to improve quantitative methods for identifying bat species from data gathered by the Anabat II bat detector.

The research group will: 1) conduct surveys of regional bat populations using netting techniques, 2) build a library of known echolocation calls from Missouri bats; 3) use the data to identify call characteristics that will improve species identification via statistical methods such as discriminant function analysis.

METHODOLOGY: Known bat calls will be recorded after hand identification of bat using nets or traps. Bioluminescent light sticks will be affixed dorsally to identified bats, and the bat will be released. Color variations in the light sticks will allow the identification of individual bats should they return to the vicinity; they will be targeted by the Anabat II system and their echolocation calls will be recorded. Each summer, fifty search­phase call sequences for each species of bat known to be resident in defined collection regions will be added to the library.

The library will catalog all calls acquired in both Anabat II format and ASCII time-frequency format. Each data file will also be cleaned to remove non-bat noise, non-search-phase call sequences, sequences with superimposed calls, and sequences with fewer than five consecutive chirps. Cleaning will occur manually using Anabat II's visualization and filtering package, Analook.

These data will then be used to create the discriminant function analysis model for species identification. This model will serve as the baseline for accuracy comparisons. It will be implemented using Matlab 6 or SAS.

Species-specific characteristics in the data will be identified and used to improve species identification rates in a discriminant function analysis model. This problem will be approached using several methods: using techniques belonging to advocates of the qualitative identification approach to identify these characteristics, as has; they could use discrete wavelet transforms of the data to seek species-specific characteristics; they could use continuous wavelet transforms to do the same. In the latter case they might investigate the deep scale-space structure of a call sequence to see if it contains any species specific characteristics.

SIGNIFICANCE: Discriminant function analysis models correctly identify the endangered Indiana bat fromits echolocation call 85% of the time. With our new model, we aimto improve the accuracy of the identification to 95%. Other members of the Myotis genus found in northern Missouri include the Little Brown bat (Myotis lucifugus) and the northern long-eared bat (Myotis septentrionalis), each of which has identification accuracy rates of only 80% using current technology; we aim to increase these accuracy rates to better than 90%. Data collected and quantitative methodologies will be freely shared. Truman State University's mammal collection will soon be available via the Internet; the bat call library would be made available in the same way. Where there are opportunities to develop quantitative methods based on open source tools (e.g. using Octave in place of Matlab) the research teams will do so and make those tools available under an open license such as the Gnu Public License.