Efficacy of Portable PIT Detectors for Tracking Long-Term Movement of Colorado River Cutthroat Trout in a Small Montane Stream

Passive integrated transponder technology has provided a valuable method for tracking individual movements of freshwater fishes (Prentice et al. 1990; McCutcheon et al. 1994; Lucas and Baras 2000). Because PIT tags do not require batteries and are relatively small, they provide substantial benefits over larger, battery-powered transmitters (e.g., radio tags) for long-term movement studies and studies of movement in small fishes (Lucas and Baras 2000; Cucherousset et al. 2010). However, the short detection range of PIT tags originally required that fish be physically handled or passed through a fixed and confined point of capture to be scanned for tags (Morhardt et al. 2000).

Passive integrated transponder technology progressed with the advent of portable PIT detectors—mobile reader antenna units that can be used to passively recapture PIT-tagged fish in small stream environments (Morhardt et al. 2000; Roussel et al. 2000). Several studies have reported relatively large read ranges and high detection efficiencies (% of tags recovered) with portable PIT detectors (e.g., Roussel et al. 2000; Cucherousset et al. 2005; Linnansaari et al. 2007). For example, Roussel et al. (2000) reported a read range of up to 1 m and a detection efficiency >80% using a portable detection unit and 23-mm PIT tags, and Linnansaari et al. (2007) reported a read range of up to 90 cm and a detection efficiency of 63–100% using a portable detection unit and 23-mm PIT tags. Other studies suggest that detection efficiency can vary with a number of factors (e.g., fish size, tag size, and environmental conditions; Breen et al. 2009; O'Donnell et al. 2010; Burnett et al. 2013).

Although portable PIT detectors have proven effective for relocating fish over small spatial and temporal scales (e.g., Morhardt et al. 2000; Roussel et al. 2000; Hill et al. 2006), detector performance is less certain over the long term and under natural conditions. Cucherousset et al. (2005) reported detection efficiencies of 69–82% when evaluating the 15-min to 24-h relocation rate of PIT-tagged Brown Trout Salmo trutta and Slimy Sculpin Cottus cognatus in a series of 20–85-m-long, enclosed stream segments. In contrast, Cucherousset et al. (2010) reported detection efficiencies of 0.7–43.1% when evaluating the 11-week relocation rate of PIT-tagged fish (five different species) in a 520-m-long, open stream segment.

In this study, we evaluated the efficacy of portable PIT detectors for tracking long-term movements of Colorado River Cutthroat Trout Oncorhynchus clarkii pleuriticus (CRCT) in an open stream environment. We also examined CRCT movement data for the purpose of informing this evaluation.

METHODS

Study site.

Poose Creek is a small headwater tributary to the East Fork Williams Fork (East Fork) in the Yampa River basin, Colorado (watershed area = 65 km²; latitude = 40º8′42″N, longitude = 107º14′42″W). Colorado River Cutthroat Trout occupy approximately 14.8 km of Poose Creek (25.0 km, including tributaries), from the headwaters down to the confluence with the East Fork (from 3,153 to 2,294 m above mean sea level). Mottled Sculpin C. bairdii is the only other fish species present in Poose Creek. At the time of this study, fish migration in Poose Creek was blocked approximately 9.4 km upstream of the confluence by an impassable culvert at a road–stream crossing. We focused on a 1,712-m-long reach that was bounded on the upstream end by the impassable culvert. Mean wetted width in summer and fall was approximately 3 m. Beaver ponds were present in a 65-m section (4%) of the reach in 2012 but were absent from the reach in 2013. Though discharge data were not available for Poose Creek (base flow was <0.1 m³/s), data were available from a gauging station on the main-stem Williams Fork (Colorado Department of Water Resources Gauging Station WMFKMHCO; watershed area = 1,085 km²). At that site, summer discharge was approximately 17% of average in 2012 and 36% of average in 2013.

Fish capture and tagging.

Colorado River Cutthroat Trout were captured by backpack electrofishing during single passes of the study reach in June and October of 2012 and October of 2013. Total length and wet mass of CRCT were measured to the nearest millimeter and gram, respectively. All CRCT ≥120 mm in length that did not already have a PIT tag (applicable only in October) were anesthetized with MS-222 (tricaine methanesulfonate) and implanted with a 23-mm half-duplex (HDX) PIT tag (Biomark, Boise, Idaho). Tags were injected into either the peritoneal cavity or dorsal musculature of fish. Differences between PIT tag treatments were addressed in a companion study (authors' unpublished data).

Read range analysis.

To understand the effectiveness of detection, we evaluated the read range of our portable PIT detectors (i.e., reader antenna configurations) before using them in the field. In 2012, we tested and used Biomark's commercially available BP portable antenna and portable FS2001 reader. In 2013, we tested and used Biomark's commercially available BP portable antenna and portable HPR Plus reader (which was equipped with an internal GPS). Read range was measured as the maximum distance from the antenna at which a 23-mm HDX PIT tag was detected. Read range was evaluated in three detection planes (axes in space relative to the antenna loop: horizontal, 45°, and vertical) and two tag orientations (cylindrical axis alignments relative to the antenna loop: vertical and horizontal), and at 10 equally spaced locations around the circular antenna. Maximum detection distance was measured five times for each of the 60 combinations of detection plane, tag orientation, and antenna section. We compared read range among detection planes, between tag orientations, and among antenna locations using one-way ANOVA and compared read range between detectors using a paired t-test. All calculations and statistical analyses were performed in R (R Development Core Team 2013) at α = 0.05.

Detection.

Locations of PIT-tagged CRCT were determined with a single portable PIT detector during one-pass, foot-based surveys in July–October 2012 and June–September 2013 (n = 15). At approximately 2-week intervals (mean = 13 d; range = 7–29 d), a one-to-three-person crew worked from the downstream end of the study reach (meter marker 0) to the upstream end of the study reach (meter marker 1,712), surveying all water in between. The exceptions occurred in 2012, when on several occasions the beaver ponds (from meter marker 1,110 to meter marker 1,175) were not surveyed due to the difficulty of detecting PIT-tagged fish in open and relatively deep water. To cover the entire stream channel, the single operator moved the antenna both left and right and up and down in the water column. The operator scanned above the water surface where the habitat was shallow and rocky, and extended the antenna beneath and around features where the habitat was complex (e.g., log jams and undercut banks). When a PIT-tagged CRCT was detected, the date and time, tag number, location (as determined with GPS), and any relevant comments were recorded. Single passes of the study reach typically began midmorning and took 5–6 h to complete. Passive integrated transponder tags that were detected in the same locations repeatedly without evidence of movement or a fish host were flagged as “suspect.” Tag numbers remained on the suspect list unless active movement of that tag was observed or a trout bearing that tag was recaptured. Efforts were made to flush fish from cover and to recover tags from the streambed. At the end of each season, detection histories associated with suspect and expelled and recovered tags were excluded from further analyses.

Efficiency evaluation.

We evaluated portable PIT detector performance based on the detection rate of PIT-tagged CRCT during foot-based surveys. Detection efficiency E was calculated for each sampling event i by following a modified version of the methods in Cucherousset et al. (2010), expressed as

(1)

where N_i is the number of PIT-tagged CRCT present in the study reach during the ith sampling event (number of PIT-tagged CRCT recaptured during the subsequent October electrofishing event) and D_i is the number of October recaptures detected during the ith sampling event. This efficiency calculation accounts for permanent emigration, tag loss, and mortality by basing N_i on only the number of live, tag-bearing fish present in the study reach at the end of the season (Cucherousset et al. 2010). This baseline efficiency calculation does not, however, account for temporary emigration of fish from the study reach. To address this possibility, we modified the conditions of the model and ran a second iteration of the efficiency equation, where N_i was based on the number of PIT-tagged trout both recaptured in October and detected during survey event i + 1, and D_i was based on the number of PIT-tagged trout both recaptured in October and detected at survey events i and i + 1. This modified version of equation (1) minimized the effects of temporary emigration by reducing the time lag between the events underlying N_i and D_i. To test for differences in detection efficiency between years and between alternative efficiency calculations, we constructed a generalized linear mixed-effects model with efficiency as the response variable, year and calculation as fixed-effect predictor variables, and sampling event as a random effect.

Colorado River Cutthroat Trout range analysis.

The range of CRCT was analyzed in two steps. First, we snapped GPS coordinates of fish detection to the nearest 5-m segment of the study reach. This step converted coordinates in space to meter markers along the stream course (while GPS-borne variability could have exceeded 5 m, small deviations were inconsequential to our evaluation of portable PIT detectors). Next, we calculated the range (variously defined as “home range,” “dispersal,” and “total range”) of individual fish according to a common methodology whereby range is the distance between the upstream-most and downstream-most locations of detection (Young 1996; Gresswell and Hendricks 2007; Alexiades et al. 2012). Recapture locations from electrofishing were included in the range analysis; however, original capture and release locations were omitted from the analysis to avoid introducing any potential bias of a tagging effect.

RESULTS

Detection Efficiency

Detection efficiency did not differ between years (P = 0.058) nor between versions of the efficiency calculation (P = 0.733; Figure 1). Overall mean detection efficiency was 34% in 2012 (baseline = 31 ± 4% [mean ± SE], modified = 39 ± 8%) and 45% in 2013 (baseline = 49 ± 6%, modified = 40 ± 8%). Within years, detection efficiency varied between certain events. For example, efficiency was 22% (95% CI = 1–43%) in mid-June of 2013 and 72% (95% CI = 50–94%) in late August of 2013.

DISCUSSION

Our findings are generally consistent with findings of earlier studies with regard to the read range of portable PIT detectors. For example, like several others (e.g., Roussel et al. 2000; Cucherousset et al. 2005; Linnansaari et al. 2007), we found that the maximum detection distance of portable detectors was greater when the PIT tag was oriented vertically than when the PIT tag was oriented horizontally. Our observed maximum read ranges with 23-mm PIT tags (32–56 cm) were short of the 90–100-cm read ranges reported by Roussel et al. (2000), Hill et al. (2006), and Linnansaari et al. (2007), but on a par with the 37-cm read range reported by Morhardt et al. (2000).

Our detection efficiencies with portable PIT detectors were generally lower than those derived in closed systems and during studies of immobile or less mobile targets, and similar to those derived during other studies of salmonids in open systems. For example, Linnansaari et al. (2007) reported 63–100% detection efficiency when recapturing hidden PIT tags, whereas Cucherousset et al. (2010) reported 39–43% detection efficiency when recapturing PIT-tagged Atlantic Salmon S. salar and Brown Trout in an open study reach. Our results suggest that if a PIT-tagged CRCT was present in the study reach, the probability of detecting that individual was, on average, 34% in 2012 and 45% in 2013. That read range differed between years while detection efficiency did not suggests that efficiency is influenced by additional factors (e.g., the adjustment of survey techniques to read range). We recommend testing and knowing the read range before using portable PIT detectors to track fish movement.

Comparison of alternative efficiency calculations suggests that temporary emigration had no significant effect on detection efficiency. If CRCT were absent from, or moving in and out of, the study reach in the months leading up to the October recapture events, we would expect to see higher efficiencies from the modified calculation than from the baseline calculation. Further, we would expect the negative bias associated with the baseline calculation to generally decrease as the time span between detection events and recapture events grew shorter. While we found no difference between alternative methods of calculating efficiency, our modification reduces the likelihood that temporary emigration results in underestimation of the true detection efficiency.

Our results suggest that detection efficiency can vary between sampling events. These differences could be explained by surveyor efficiency. We used several different crew combinations, with varying degrees of experience. An alternative explanation for the differences in detection efficiency is that fish were more susceptible to detection during certain times of the year than during others. We saw that in 2013 detection efficiency was lowest in June (when flows were relatively high and CRCT were potentially redistributing following the spawn) and highest in August (when flows were relatively low and few fish were moving [we observed three movements >15 m in August]). In their evaluation of portable PIT detectors, O'Donnell et al. (2010) observed that discharge influenced the probability of detecting PIT-tagged Atlantic Salmon and that the most and least experienced surveyors generated the highest and lowest detection probabilities, respectively. Keeler et al. (2007) observed that variation in electrical current to the antenna (i.e., the energy required to power PIT tags) influenced the detection rates of PIT-tagged Slimy Sculpin. We did not formally evaluate within-year variation in efficiency or any of the possible explanations for between-event differences in efficiency. Nevertheless, we suggest that both the conditions of the study area and fish behavior can influence detection efficiency and that efficiency is likely to peak when the search area is relatively small (e.g., discharge is low) and the targets are relatively available (e.g., fish are not burrowed in the streambed).

We identified at least two limitations of using portable PIT detectors to track the movement of fish. First, portable PIT detectors were effective in detecting fish only during low-flow, ice-free seasons. At Poose Creek and other high-elevation streams, the ice-free, low-flow period typically amounts to a short time window between mid-June and mid-October. Even if fish could be effectively detected through ice (Linnansaari et al. 2007), snow depth would exceed the detector read range. Because we were unable to monitor fish movement from late October through early June, we failed to capture movements associated with CRCT spawning migrations (April–June). Young (1996) observed, for example, that the median weekly movement of CRCT in the North Fork Little Snake River basin (Wyoming) peaked in mid-June and declined thereafter (i.e., postspawn). In short, portable PIT detection may not be a practical way to track the movements of spring-spawning fish in montane streams, and collection of data only during summer months could result in underestimation of fish movement.

Second, the relatively short read range of portable PIT detectors and the time requirement of surveying stream segments dictated that only a relatively small study reach could be surveyed repeatedly. This likely led to an undersized study area and bias against capturing long-range movement (Gowan et al. 1994), a notion supported by comparison of our data with other CRCT movement data. We found that the median range of PIT-tagged CRCT in Poose Creek was 15 m, whereas Young (1996) found that the median range of radio-tagged CRCT in the North Fork Little Snake River basin was 233 m, and B. W. Hodge, K. D. Battige, and K. B. Rogers (unpublished data) found that the median range of radio-tagged CRCT in a stream near Poose Creek was on the order of several km (B.W.H., K.D.B., and K.B.R., unpublished data). In our study, the range of CRCT was capped at 1.7 km. That 40% of our PIT tags were never detected when the mean detection efficiency was 34–45% suggests that some PIT-tagged CRCT moved more than 1.7 km downstream and thus emigrated from our scope of detection. Use of a stationary antenna at the downstream end of the study reach would have allowed us to monitor the ingress and egress of PIT-tagged CRCT but would have provided incomplete estimates of fish movement. We were unable to install an antenna at the downstream end of the reach because of the remote location and difficulty of providing a continuous power source to an antenna.

Our study contributes to a growing body of information on the use of portable PIT detectors. We are among the first to evaluate the efficacy of portable PIT detectors over the long term (>1 year) with PIT-tagged fish in a relatively large (>1-km), open study area. In addition, we offer a modification to an existing efficiency equation (Cucherousset et al. 2010) that should reduce the bias resulting from temporary emigration on calculations of detection efficiency. Finally, our inclusion of movement data and comparison with other species-specific observations highlights potential drawbacks of using portable PIT detectors to track the movement of relatively mobile fish species (e.g., salmonids) in montane streams.