Meet the frog!

Educational movie

Edit Plugin:img

Image  A 15-min movie featuring the most common species in Central Texas and their calling behavior in the most intense nights of the breeding season.

This movie was produced by Marcos and Cristina Gridi-Papp in 2001.

Watch the mini (lower quality) version now.

Aknowledgements: Supervision: Michael J. Ryan and Carl Gans. Field help: Anne Fowler, Edward Chen, Emily Moriarty, Erika S. Miller and Katja Heubel. Special thanks to: Daniel Porter, Miles Compton, Joe W. Neal, The Romes, The Schapperts, The Crutchfields, Larry Gilbert, David Cannatella, Walter Wilczynski, David Hillis, Robert Dudley, Dan Snodgrass, People of the RyanLab and Ingo Schlupp. Support: Gill Ranch, Horse Thieve Hollow Ranch, Barton Creek Habitat Preserve, The Nature Conservancy, McKinney Falls State Park, Texas Parks and Wildlife, Brackenridge Field Laboratory, Stengl-Lost Pines Biological Station, The University of Texas and CAPES - Brazil


Edit Section
This was a modular computer-based behavior monitoring system developped in 2002. Behavioral data were recorded 24h a day and analyzed in real-time. The results of the analyses were continuously fed into a database, and served over the internet. Listener also incorporated auditing tools that allowed for visual verification of the analyses and performance of the hardware.
Edit Plugin:img
Edit Section


Modular: It could be scaled down to one computer, to monitor a few animals, or it could be expanded to form a cluster monitoring many animals.
Scalable: Listener was assembled with powerful software for high demand networking, databasing and webserving. It could be configured with the same software over a wide range of sizes.
Free: All software used in Listener was free. Surplus computers would suffice for most applications.
Open source: All the scripts and software were open code. One could customize it at will.
Versatile: It was built to monitor sound, but it could be easily adapted to monitor other signals.
Edit Plugin:img
Edit Section

Data flow

Data collection fluctuated with the activity of the animals.
It averaged daily 10.000 new rows of data in early 2003, with peaks of 40.000 rows. Each row contained results for one call. Sixteen call features were measured, describing its spectral and temporal structure.
Listener managed tables with over half a million rows of data without problems.

Besides storing the data, Listener had several scripts for generating summaries and graphs. Those were displayed on dynamic web pages or on CallMon′s monitor, which continuously cycled through graphs and tables to illustrate the status of the system.
Edit Plugin:img

Edit Section


Ten frogs were monitored by ten computers*:

[+] * Each sound card could monitor two frogs and each computer could have several sound cards, but we were accessing the simplest case first.
** The early cages depicted here had excessive sound reflection and their microphones were not directional.

Edit Section


The system was formed by 10 measurement units and 2 servers.
Server2 (S2) was the gateway.
It kept the Firewall, and the WWW server for Listener′s data and website.
It provided dynamic generation of web content with graphical representation of results.
It received level 2 time (synchronized with atomic clocks) and served it to the machines inside Listener.
It checked hourly for integrity and backup of Server1′s database.
Server1 (S1) was the center of the system. It was separate from Server2 to shield the database service from the traffic and risks of the www.
Database server for the other machines inside Listener.
It accessed the hard drives of all measurement units and saved samples of recorded sounds for auditing of the analyses.
Ran the scripts that facilitated visual verification of analyses.
Edit Plugin:img

Edit Section

Measurement Units (M1-M10)

Countinuous recording and analysis of one billion sound samples per machine, per day.
Sound was recorded in alternation between two buffers (60 min long each).
While one buffer was recorded, the other was analyzed.
Analyses consisted of sound detection, recognition and measurement.
The results of the measurements for each sound were sent immediately in binary format to the database server, without temporary storage on disk.
At the end of each buffer analysis, a report on load, performance and time precision was sent to the database server.
Edit Plugin:img

Edit Section


OS: Redhat Linux 7.3 (PCs) or Debian Linux 3.0 (Macs)
Sound: OSS
Acoustic Analysis: Octave (the open alternative to Matlab).
Database: Mysql
WebServer: Apache
Time: Ntp synchronized the clocks of the machines.
Image Edit Plugin:imgImage Edit Plugin:imgImage Edit Plugin:imgImage Edit Plugin:img

Edit Section


Listener ran with 10 measurement units and 2 servers.
Server1: Dell GXpro 200 MHz Pentium, 64 MB ram, 2GB hard disk.
Server2: Power Macintosh 6500 225 MHz, 96 MB ram, 2GB hard disk.
Units6-9: Dell GXpro 200 MHz Pentium, 32 MB ram, 1GB hard disk.
Units10-15: Dell GXM 166 or 133 MHz Pentium, 32 MB ram, 1GB hard disk.

Edit Section


This project was developed in Dr. Michael Ryan and Dr. Carl Gans′ laboratories.
Brackenridge Field Laboratories (BFL).
The University of Texas.
National Science Foundation.


Edit Section

A walker for ants!

Image Edit Plugin:img

The AutoAnt is a fire ant fooler.

Back in 2000, the fire ant biological control project at the University of Texas needed to raise large numbers of parasitic flies for research and biological control of fire ants.

The flies only lay their eggs on the fire ants if the ants are moving. But the ants hide when there are flies around. Researcher would place ants and flies in a closed arena with a single hidding place under a cover. The hidding cover would be moved periodically and the ants would keep walking, even while being attacked by the flies. The trick worked well for a single arena with a few ants, but how could this be done in large scale?

Using an old PC, a parallel port connection, some gears from RadioShack and some home-written software, we created AutoAnt!!!

This was a computer-controlled mechanism that moved the covers in 24 arenas simultaneously at adjustable intervals, allowing for efficient production of fire ant parasitic flies.

The initial prototype worked so well, that we made two replicates of it and never updated the concept any further.

The system ran well for years, producing 500 flies a day, and you might still be able to see it working if you visit Brackenridge Labs at UT.