Assessing the distribution and habitat requirements of bat species assemblages by means of bioacoustic, synergetic pattern recognitio

Involved persons

• Martin Obrist (Biology, Coordination)
• Ruedi Boesch (Programming)
• Peter Flückiger (Biology)

  …and many many field enthusiasts…

Problem

The 26 bat species occurring in Switzerland spend a secluded daytime and forage in complete darkness. Investigations in their habitat use and foraging behavior necessitate catching of bats and possible individual marking and tracking (e.g. with radio-transmitters or light-tags). But, all Swiss bats are acoustically conspicuous due to their active orientation system. Echolocation calls are to a large extent species specific, and thus allow to eavesdrop on bat behavior in space and time. Present methods of acoustic species determination employ either high-speed instrumentation tape recorders (very time consuming and expensive) or more readily available electronic aids to make the ultrasound vocalizations audible. Unfortunately, the latter resolves only a limited part of the species and is error prone. As a novel approach, we have developed a pattern recognition system based on synergetic algorithms, that is capable of identifying bat species in nearly real-time. Refinement of the field-portable digital ultrasound recording and analysis will allow standardized, unsupervised and autonomous monitoring of bat foraging activity.

Aim

The purpose of this project is three-fold:

we want to

• field-validate the synergetic pattern recognition system we developed, and calibrate it against other methods of bat monitoring,
• further improve the system to a level appropriate for commercial distribution (including an increased base of reference calls ) and then make it available to fellow scientists and nature conservationists,
• assess species specific habitat use and habitat requirements on a landscape level. We want to identify habitat structures relevant for the foraging and therefore survival of these endangered species. With this insight we will model the species specific, potential distribution, based on habitat use. This may significantly contrast to models based on presently better known roost distribution.

Methods

Verification includes acoustic monitoring with the classifier and with heterodyning bat detectors in parallel to capturing of bats. Data collection will be performed during 10 nights in different habitats, at sites of known bat traffic.

Development/Improvement of a bat detector takes place at the "Institute for Telecommunication and Information-technology" of the "Biel School of Engineering and Architecture" (2 Diploma Theses). Final assembling, software adaptation and commercialization will happen at the WSL.

Habitat use will be assessed by means of acoustic monitoring of bat activity with two detector units (7 microphones each). Each year 16 nights of monitoring (twice at 8 sites) will be spent in parallel at locations in Kt. Solothurn and Kt. Aargau/Zürich. Recording locations will be evaluated regarding intra-site diversity and inter-site comparability. Results of the monitoring will be analyzed on a species level, aiming at spatial mapping of potential species distribution.

State of project

A one-detector unit is presently available. A second system will be acquired in 2000. Software adaptations will enable automatic monitoring in the same year. Development of a new detector and validation of the systems performance in the field should be terminated by the end of the year 2000. Distribution of the product may start in spring 2001. Habitat monitoring is planned for the field-seasons of 2001 and 2002. Intermediate data analysis and software adaptations will be performed over the full project duration. Final analysis and publications are scheduled for 2003.

State of the art

Bats have evolved different types of echolocation calls by adapting the physical signal properties to the requirements of the foraging behavior and environment. Thus, species specificity of echolocation allows to eavesdrop on occurrence and behavior of echolocating bats. Due to the subtleness of call differences, present acoustic monitoring methods only resolve selected species, because they reduce signal information in various ways. Expensive and bulky high-speed tape recorders or digital recording allow to fully record signal content and perform later, time-consuming analysis, including statistical treatment and recognition task. This has been done in a variety of international studies. Synergetic algorithms are novel and we use it the first time to classify animal vocalizations. Pattern recognition algorithms relying on full spectrogram characteristics have not been applied to bat calls. Its use in real-time field monitoring marks a breakthrough in terms of technology and possible standardization.

Most work on the identification of habitat use by bats was done by means of telemetry, focussing on a single species. Acoustics is the method of choice for evaluating foraging of bat species assemblages. The use of synergetic pattern recognition in this task offers a new level of quality for such studies.

Significance of study

The synergetic algorithm we use can theoretically be used for a diversity of pattern recognition tasks, including sound and image analysis. Once the quality in a special sector (bioacoustics) is established, other applications could follow.

Presently lacking deeper knowledge of flexibility or rigidity in habitat use of bats will be improved.

The differential knowledge of habitat use by bats is complementary to the knowledge of roost locations. Together it can enable much more effective conservation strategies. Set in context to other bat species' habitat requirements, an integrated, landscape level approach to bat protection becomes possible. Swiss bats exclusively forage on arthropods, which contribute the majority of species to biodiversity. Thus, the indicatory value of bat foraging habitat for overall biodiversity conservation becomes even more prominent.