Applications of Underwater Video for Imperiled Fish Species Population Monitoring

Abstract

Common freshwater fish sampling methods (e.g., seining and electrofishing) are inherently invasive and often not appropriate for imperiled species. Visual observation methods provide a noninvasive alternative for population monitoring. Despite increasing popularity, the use of underwater video to monitor imperiled fishes is relatively unexplored. We evaluated the efficacy of underwater video to estimate occurrence and abundance of the imperiled Comanche Springs Pupfish Cyprinodon elegans using both point observations and time intervals (surveys). We deployed camera traps at sites within major habitat types (pool, canal, and ciénaga) of Balmorhea State Park, Texas, United States in March and October 2019 (seasons). We detected Comanche Springs at all occupied sites in both seasons when viewing ~30 min of video. The species was detected at 80% of occupied sites when viewing ~10 min and ~5 min of video in March and October, respectively. Comanche Springs Pupfish detection probability was higher in October, with no variability among habitat types. On average, cumulative species detection probability was >0.9 with fifteen 60-s surveys. However, species detection probability of a single survey ranged from 0.02 to 0.62 (mean = 0.14). Although there was no variation between seasons or among habitats, variation was high in the detection of the maximum Comanche Springs Pupfish count among sites even with observations every 5 s. Individual capture probability from a repeated-count abundance model was less variable than species detection probability (0.01-0.33) and generally low (mean = 0.06). Site absolute abundance was generally comparable among major habitats, but with higher uncertainty with increasing maximum count. Our study provides a comprehensive assessment of underwater video for imperiled fish species population monitoring. The findings show a trade-off between processing effort and information loss and limitations associated with imperfect detection and individual capture common to any fish sampling method.