Study Area

Round Butte Dam is a 134-meter (m) high and 402-m wide, rockfill embankment dam. At the base of the dam is a powerhouse containing three Francis turbine units capable of generating a total of 324 megawatts (fig. 2). During normal operations, all water passing the dam goes through the powerhouse intake, located approximately 200-m upstream of the dam. An emergency spillway with a single radial gate is located to the west of the dam but is not used during normal dam operations. The 1,619-hectare (ha) Lake Billy Chinook has a storage capacity of 65,991 hectare-meters (ha-m), but a useable storage capacity of 33,797 ha-m. The forebay elevation typically fluctuates ±1 m during normal weather and project operations.

The above-water fish collection structure of the SWW (fig. 3) is approximately 27 × 46 × 18 m in size and has water drawn from the surface of the reservoir, that is directed to the power intake via a vertical flow conduit with a maximum inflow of 6,992 ft³/s. Discharge herein is presented in cubic feet per second (ft³/s), according to the local reporting conventions of both the U.S. Geological Survey (USGS) and PGE. In total, the SWW structure is 83 m in height. Fish enter either collector flume (9.6 m wide × 12.2 m deep), move past dewatering screens, and then pass directly into a pipe that delivers them into the fish transfer facility located alongside the SWW. At the fish transfer facility, fish are sorted and identified before being trucked and released into the Deschutes River downriver of the Project. The SWW collector is placed near the north shore of the reservoir and over the turbine intake. In 2021, from March 15 to June 15, the SWW followed a nighttime generation operation protocol whereby the Project attempted to maintain a nightly minimum flow of 4,500 ft³/s to increase attraction flow during periods of active smolt migration. For the duration of the study, the SWW was scheduled to operate with two discharge intervals changing approximately every two weeks: (1) Traditional condition, seven hours at a target of 4,500 ft³/s from 9:00 p.m. to 4:00 a.m., and (2) Delayed condition, seven hours at a target of 4,500 ft³/s during 11:00 p.m. to 6:00 a.m. Additional information on SWW operations and performance may be found in Smith and others (2022), Pyper and others (2017), and Portland General Electric Company and The Confederated Tribes of the Warm Springs Reservation of Oregon (2015, 2018).

Dam Operations and Environmental Conditions

Project discharge, forebay elevation, and water temperature data were summarized to document the environmental conditions that fish experienced from March 20 to June 21, 2021. Hourly data were collected and provided by PGE. Water temperature data were collected at a water quality monitoring station located in the forebay, and turbidity measurements were collected from the tailrace. Tributary inflow data were collected at the U.S. Geological Survey streamgages on the Deschutes and Crooked rivers near Culver, Oregon, (U.S. Geological Survey, 2021a, b), and the Metolius River near Grandview, Oregon (U.S. Geological Survey, 2021c). Data were summarized using hourly observations, but mean daily values were used to increase plot clarity.

Imaging Sonar Deployment and Data Processing

Adaptive resolution imaging sonar systems were used to collect data of fish movements immediately upstream of the SWW entrances. The sonars were operated at 1.8 MHz, with an operating range of 2–12 m. The sonars were attached to a pole-mounted platform located between the collector entrances and lowered to a depth of 3.5 m (fig. 4).

The imaging sonars collected data continuously at the SWW between March 20 and June 21, 2021 to coincide with the peak outmigration timing of salmonid smolts. Data collection was interrupted only when the SWW was non-operational because of maintenance or when equipment malfunctioned. All data collected were stored to hard drives for archival and subsequent processing. Over the duration of the study, 94 dates with complete 24-hour sampling periods were included in the analysis.

The image recognition and tracking programs used for identifying fish targets at the entrances used four types of software to prepare and process the collected acoustic signals. Initially, ARISFish software (Sound Metrics, 2018) removed acoustic noise from the collected files. Then, files were parsed to time sequential images using MATLAB software (The Math Works, 2019). Next, computer vision software, OpenCV (2020) was used to locate fish target positions in the sequential images. Finally, target tracking algorithms developed in Python software (Python Core Team, 2020) were applied to the located targets to determine individual fish behavioral and morphometric data.

Imaging sonar technology cannot distinguish individual fish that have entered and exited the field of view multiple times; therefore, the following analysis assumes that each fish track, not the fish, were independent observations. Additionally, it is not possible to definitively identify the species of fish observed; however, if discrete group sizes were present and identifiable through other methods (in other words, collection), informed inferences of species can be made. Despite these limitations, imaging sonar can be used to collect data that would otherwise be unattainable.

Summary statistics of fish targets derived from signal processing (mean length, direction, speed, and location in beams) were imported into SAS software (SAS Institute, 2012), for subsequent proofing and to merge imaging sonar data with the environmental data. Data were proofed to eliminate non-valid records or records that did not provide measurable morphometric or behavioral data. Target datasets were then exported for additional statistical analysis using R software (R Core Team, 2014).

Dam Operations and Environmental Conditions

A combination of upper basin reservoir water storage for irrigation and lower than average rainfall and snowmelt resulted in reduced instream flows during March and early April than were normally experienced during the typical spring seasonal pattern on the Deschutes, Crooked, and Metolius Rivers (fig. 5). However, by about late-April, instream flows returned to near-median levels for the remainder of the observation period. The median hourly SWW discharge was 2,392.9 ft³/s (range 42.9–6,101.4 ft³/s; fig. 6; table 1). The spillway was not opened during the study, and all water passed through the SWW, with a peak daily discharge of 3,956.0 ft³/s occurring on April 28, 2021. The reservoir elevation was maintained at a constant daily elevation of nearly 1,944.6 feet above National Geodetic Vertical Datum of 1929 for the entirety of the study, with a 1.6-foot fluctuation (table 1). Water temperature generally increased through the study period, peaking at 19.0 degrees Celsius on June 4, 2021, whereas the lowest water temperature of 7.9 degree Celsius occurred on April 13, 2021 (table 1; fig. 6). The SWW was non-operational because of mechanical failure between March 24 and April 5. Fish observed during this period were not included in the general analysis; however, abundances and distributions are reported.

The target nighttime generation operation protocol discharge of >4,500 ft³/s was not constantly achieved for the entirety of the study because reduced reservoir inflow and the requirement of the Project to maintain a constant reservoir elevation. However, SWW discharges between 4,000 and 4,500 ft³/s were generally attained during the nighttime planned operation discharge. Although the SWW was scheduled to operate with two discharge intervals (Traditional and Delayed) changing approximately every two weeks, insufficient tributary inflow resulted in inconsistent operation timing. The resulting reduced-discharge operations caused imbalanced discharge intervals of 29 days of Traditional and 53 days of Delayed night generation operation timing.

Fish Abundance

Data from the imaging sonars indicate that abundances of both smolt and bull trout-size fish near the entrances of the SWW varied while the migration season progressed (fig. 7). Abundance of fish observed on both imaging sonars generally were lowest during late-March and early-April, then trended upward to a peak abundance of 4,708 fish on May 10, before decreasing to a low of 240 fish on June 6. Fish observations were largely dominated by smolt-size fish; however, the daily abundance trends of adult bull trout-size fish were low (range 5–103 daily) and generally corresponded with that of the smolt-size fish.

In total, 120,564 fish observations were recorded, with nearly twice the number of fish observed at the south entrance (n=77,749) than at the north entrance (n=42,815; table 2). The Chinook and sockeye smolt-size fish group (95–190 mm) was the predominant size class observed, constituting 67 percent of the total observations. The steelhead and kokanee fish-size groups represented 22 and 9 percent of observations, respectively, whereas the adult bull trout-size group constituted 3 percent of total observations.

Direction of Travel

The predominant directions of travel for observed fish differed by entrance but were generally similar between the night generation operation timing conditions and the fish-size groups (table 3). All fish size groups observed with the imaging sonar units at both the north and south entrances had travel paths that were generally circular (figs. 8–11), with significant (Rao P<0.05) primary directions of travel that were directed toward the center of the SWW under both the Traditional (from 9:00 p.m. to 4:00 a.m.) and Delayed (from 11:00 p.m. to 6:00 a.m.) night generation operation timing conditions. The primary directions of travel of the steelhead, Chinook, and sockeye smolt-size fish groups were more directed toward the north entrance during the Traditional night generation operation timing condition, but more directed toward the south entrance during the Delayed night generation operation timing condition; however, rose plots had generally similar directional headings among groups (figs. 8, 9). The primary directions of travel of the kokanee-size fish groups were more directed toward the north entrance during the Delayed night generation operation timing condition, and more directed toward the south entrance during the Traditional night generation operation timing condition, with similar rose plot headings (fig. 10). Directions of bull trout-size fish observed at the north entrance were generally circular, whereas fish observed at the south entrance had primary directions of travel directed toward the center of the collector (fig. 11). Low concentration parameter (κ) values (range 0.38–0.95) for all criteria reflect a reduced concentration of the distribution toward a mean direction.

The general direction of travel of the three smolt-size fish groups near the entrance to the SWW depended on fish group, entrance, and sometimes discharge (fig. 12). Near the north entrance, the general directions of travel (either toward or away) of the Chinook and sockeye smolt-size fish group were generally toward the entrance when discharge through the SWW was less than 2,000 ft³/s with no strong effect indicated by Cramér’s V (table 4). However, when discharge was increased to greater than 2,000 ft³/s, the general direction of travel became near 50 percent, either toward or away from the entrance, but with a weak effect indicated by Cramér’s V. The directions for kokanee-size fish were like the Chinook and sockeye smolt-size fish group, with directions generally toward the entrance when discharge through the SWW was less than 2,000, and the general direction of travel became near 50 percent, either toward or away from the entrance when discharge was greater than 2,000 ft³/s, with low Cramér’s V scores indicating weak effects for travel direction (table 4). For steelhead smolt-size fish, the general directions of travel at all discharge rates were toward the entrance, with moderate effect values (Cramér’s V>0.34) at all discharge rates greater than 1,000 ft³/s (table 4).

Near the south entrance, the percentage of the Chinook and sockeye smolt-size group directed toward the south entrance was 42 to 54 percent with the highest rates observed at discharges between 2,000 and 3,000 ft³/s; however, all effects were weak as indicated by Cramér’s V (table 4). The direction of travel for the kokanee-size group was toward the south entrance at discharges less than 3,000 ft³/s, and away from the entrance at discharges greater than 3,000 ft³/s, with weak effects as indicated by Cramér’s V scores (table 4). The steelhead smolt-size group was directed toward the entrance at discharges greater than 2,000 ft³/s; however, Cramér’s V scores were low and indicate weak effects (table 4).

Fish Swimming Velocity

The speed at which fish traveled at the entrances of the SWW generally differed between the fish-size classes but were similar between the two entrances, except for bull trout-size fish (table 5; fig. 13). For example, at the north entrance, the mean swimming velocity of Chinook and sockeye smolt-size fish was 0.32 meters per second (m/s; interquartile range [IQR]=0.29), 0.34 m/s (IQR=0.30) for kokanee-size fish, 0.33 m/s (IQR=0.31) for steelhead-size fish, and 0.31 m/s (IQR=0.29) for bull trout-size fish. These differences in fish swimming speed were significant between fish sizes (ANOVA; F_{3, 142,811}=20.27, P<0.001) with mean swimming speeds varying irrespective of size groups. Results of the post hoc Tukey HSD test indicate that the swimming speeds of the Chinook and sockeye smolt-size group and the bull trout-size group were not significantly distinguishable from each other (P=0.28), nor was the steelhead smolt-size group and the kokanee size group and (P=0.12); whereas the remaining groups had significantly different swimming speeds from each other (P<0.001).

At the south entrance, the mean swimming velocity of fish were generally similar to the speeds of fish observed at the north entrance, with Chinook and sockeye smolt-size fish velocities at 0.33 m/s (IQR=0.30), 0.33 m/s (IQR=0.29) for kokanee-size fish, 0.36 m/s (IQR=0.32) for steelhead smolt-size fish, and 0.40 m/s (IQR=0.36) for bull trout-size fish. These differences in fish swimming speed also were significant between fish sizes (ANOVA; F_{3, 77,741}=170.60, P<0.001). Results of the post hoc Tukey HSD test indicate that the swimming speeds of the Chinook and sockeye smolt-size group and the kokanee-size group were not significantly distinguishable from each other (P=0.32), whereas all other fish size groups had significantly different swimming speeds (P<0.01).

When observations of fish were compared between the north and south entrances, swimming velocities were statistically different for each of the fish-size groups (at the α=0.05 level) but were unlikely to be biologically significantly different for the smolt-size fish groups. For Chinook and sockeye smolt-size fish, mean swimming velocities were 0.01 m/s different between the two entrances but were significantly different (t; _79,423=6.19, P<0.001). Mean swimming velocities for kokanee-size fish were 0.01 m/s different between the two entrances and were significantly different (t; _10,705=3.44, P<0.001). Swimming velocities for steelhead-size fish were slightly faster at the south entrance (0.36 m/s) than at the north entrance (0.33 m/s) and were significantly different (t; _26,605=9.77, P<0.001). Bull trout-size fish had the greatest range in mean swimming velocities with a speed of 0.40 m/s at the south entrance, and 0.31 m/s at the north entrance, and were significantly different (t; _3,819=10.65, P<0.001).

Timing of Detection

The counts of fish observed with the imaging sonars at the SWW entrances differed by night generation operation timing and discharge. During the Traditional night generation operation timing (from 9:00 p.m. to 4:00 a.m.), counts of fish from the Chinook and sockeye smolt, steelhead, and bull trout-size fish groups were lowest during the daytime periods when discharge rates were reduced (fig. 14). Counts of kokanee-size fish were highest at 4:00 p.m. immediately preceding the increase in discharge. After about 6:00 p.m., counts of steelhead and Chinook and sockeye smolt-size fish groups began increasing about two hours after the initial increase in discharge and increased through the night. After the decrease in flow at about 3:00 a.m., smolt-size fish counts began declining, before declining sharply at 5:00 a.m. Bull trout-size fish were most abundant at 5:00 a.m. Counts of fish from all size groups were lowest at 8:00 a.m.

During the Delayed night generation operation timing (from 11:0 p.m. to 6:00 a.m.), trends of counts for fish were generally like those observed during Traditional night generation operation timing (fig. 15). Chinook and sockeye smolt, and bull trout-size fish groups were lowest during the daytime periods when discharge rates were reduced. Counts of the steelhead and the Chinook and sockeye smolt-size fish groups increased through the night, with maximum observations occurring at 5:00 a.m., before rapidly decreasing to reduced observations during periods of daytime and low discharge rates. Counts of kokanee-size fish were highest at 2:00 p.m., immediately preceding the increase in discharge, and remained relatively stable through the night, before declining at 6:00 a.m. Bull trout-size fish were most abundant at 5:00 a.m., during the period when smolt-size fish were most abundant. Counts of fish from all size groups were lowest at 8:00 a.m. The effects of daylight on fish counts during the Delayed night generation operation timing became more evident while the outmigration season progressed (fig. 16). Detections of smolt-size fish were greatest at 5:00 a.m., before declining rapidly an hour later during March and April, when sunrise occurred later in the morning. And, during May and June, counts of fish declined rapidly at 5:00 a.m. because daylight began earlier in the morning.

Fish Track Density

The spatial distributions of fish positions near the entrances of the SWW suggest some differences in distribution depending on the fish size group. The fish location point density data include greater than 2.8 million location points from 120,564 individual tracks that were recorded by the imaging sonars. Point density data indicate that positions for Chinook and sockeye smolt-size fish observed on the north sonar were primarily near the area between the SWW entrances, with lesser concentrations of positions on the outsides of the SWW structure (fig. 17). However, positions for Chinook and sockeye smolt-size fish observed on the south sonar were nearly spread across the entirety of the south entrance. Positions for kokanee and steelhead smolt-size groups were similar at the north and south entrances with both fish groups primarily observed near the area between the SWW entrances, with fewer detections on the far north and far south ends of the SWW (figs. 18, 19). The bull trout-size group was primarily observed near the area between the SWW entrances, like detections of the kokanee and steelhead smolt-size groups (fig. 20).