Observer Error Structure in Bull Trout Redd Counts in Montana Streams: Implications for Inference on True Redd Numbers

Abstract

Despite the widespread use of redd counts to monitor trends in salmonid populations, few studies have evaluated the uncertainties in observed counts. We assessed the variability in redd counts for migratory bull trout Salvelinus confluentus among experienced observers in Lion and Goat creeks, which are tributaries to the Swan River, Montana. We documented substantially lower observer variability in bull trout redd counts than did previous studies. Observer counts ranged from 78% to 107% of our best estimates of true redd numbers in Lion Creek and from 90% to 130% of our best estimates in Goat Creek. Observers made both errors of omission and errors of false identification, and we modeled this combination by use of a binomial probability of detection and a Poisson count distribution of false identifications. Redd detection probabilities were high (mean ¼ 83%) and exhibited no significant variation among observers (SD ¼ 8%). We applied this error structure to annual redd counts in the Swan River basin to correct for observer error and thus derived more accurate estimates of redd numbers and associated confidence intervals. Our results indicate that bias in redd counts can be reduced if experienced observers are used to conduct annual redd counts. Future studies should assess both sources of observer error to increase the validity of using redd counts for inferring true redd numbers in different basins. This information will help fisheries biologists to more precisely monitor population trends, identify recovery and extinction thresholds for conservation and recovery programs, ascertain and predict how management actions influence distribution and abundance, and examine effects of recovery and restoration activities.

Trout Redd Locations in Two Streams in the Black Hills, South Dakota, USA

“…Although brown trout will build nests in areas of groundwater upwelling, they do not show a distinct preference for such areas [31]. over other fish population survey methods [20], and would be much less expensive than the backpack electrofishing surveys historically used in the Black Hills [28].…”

“…In addition, the relative inexperience of the observers may have influenced the results [20], although any such errors were likely minor given the size of the creeks studied and extensive observer training.…”

“…Each creek section was surveyed for redds five times over a four week period, with redd locations recorded using Trimble (Sunnyvale, California, USA) Global Positioning System units. Only data from confirmed redds were included to minimize potential observer error [20]. The species of fish on or near the redd, if it could be determined, was noted.…”

Factors Influencing the Detectability of Early Warning Signals of Population Collapse

“…Typically, a subset of the population is counted (say, along a transact or within a defined area), with this subset being assumed to reflect the true (total) population abundance (e.g., Franzetti et al 2012; but see Hefley et al 2013). However, in itself, this subsampling generates uncertainty, as chance events such as the movement of organisms, the heterogeneous distribution of individuals, and the skill of the observers influence recorded population abundance (Muhlfeld et al 2006). While there are a variety of sophisticated methods that seek to address the problems raised by spatial subsampling (e.g., Wright et al 2009;Ford et al 2012), the uncertainty generated at the sampling stage can never be fully resolved.…”

The Integrated Status and Effectiveness Monitoring Program : Expansion of Existing Smolt Trapping Program and Steelhead Spawner Surveys : March 1st, 2008 - February 28th, 2009.

“…Studies planned for the Twisp River in 2009 to compare the estimated redd abundance in non-index areas to the actual count should also be conducted in the Entiat and Wenatchee Basins. Furthermore, observer efficiency is a potentially large source of error in conducting redd counts (Dunham et al 2001;Muhlfeld et al 2006). The current methodology should be modified to incorporate estimates of observer efficiency and not only identify, but also quantify sources of error (redd omission or false identification).…”