Estimating Survival Adjusted for Tag Failure in an Acoustic-Tag Study Using the ATLAS Program

Program ATLAS was developed to analyze the survival of migrating smolts in the Columbia River Basin from studies using acoustic- or radio-tagging data. Active tagging technologies allow increased flexibility in designing survival studies, but also introduce the problem of tag failure. Tag failure can be the result of mechanical or battery failure, and cannot be distinguished from mortality in release-recapture studies. Therefore, the perceived survival estimates (the probability of a fish and tag being “alive" at a detection site) must be adjusted for the probability of tag failure to obtain a bias-corrected survival estimate. 

Program ATLAS has three modes of operation, corresponding to three study designs in use in the Columbia Basin: 1) Single-Release (one release site), 2) Paired-Release (two release sites, one above the other), and 3) Virtual/Paired-Release (a “virtual” release group is formed above a paired-release design).  The single release design estimates survival from release to given detection sites.  The paired-release design is used when the desired placement of detection arrays is not possible, or to control for other mortality effects (handling mortality, etc.).  The virtual/paired-release model is a design to provide better dam-passage survival estimates, by allowing the fish to be released much further upstream.  The third design will be covered in more detail by Dr. Skalski in the talk “Acoustic Telemetry and Compliance with FCRPS” in this session.  This presentation will be a walk-through of ATLAS and its features.

Given release-recapture data, the program can estimate the Cormack-Jolly-Seber probabilities of reach survival and detection at downriver sites.  With the additional data of randomly sampled tags monitored for tag failure, the program will estimate a tag-life failure curve, and provide the tag-life adjusted survival estimates.  Some modeling capabilities are available, such as setting common survival and detection parameters, or removing downstream sites from the analysis.  Other features include mixing plots at each detection site (fish arrival distribution at each site), arrival distributions vs. the estimated tag failure curve, and the estimated probability that a tag was still working at the time fish were in the vicinity of the detection array. 

The latest version of the software, along with a PDF copy of the user manual, is available at www.cbr.washington.edu/paramest/atlas.