Travel Times and Retention


Winds were predominantly from the west to the northwest during the 2009 simulation period between May 14 and June 10, as is typical for the season, but variability in wind direction and speed over this period was still important. The first pulse of larvae, predominantly Lost River sucker larvae, entered the delta between May 18 and May 21, with a peak on May 18. Winds were strong (predominantly 6–8 m/s) and from the west-northwest. The second pulse, comprising both Lost River and shortnose sucker larvae, entered the delta between May 29 and June 1, with a peak on June 1. Winds were weaker (mostly 2–4 m/s) and primarily from the north. Because of this difference in the wind velocity, water currents coinciding with the entry of each pulse into the delta were different: during May 18–21, currents were strong and to the southeast across the southern delta; during May 29–June 1, currents were weaker and more variable in direction (fig. 4). The difference in the water currents during the period associated with the entry of each of the two larval pulses into the delta was reflected in the travel times of particles from the upstream boundary to the outlet. The cohort of the particles inserted at the upstream boundary on May 18 exited the lake at Link River on May 30, 12 days later, whereas the cohort of the particles inserted at the upstream boundary on June 1 exited the lake on June 22, 21 days later.


Comparison of Travel Times by Dispersal Type


Travel times were fastest (1.1–4.3 d) to fixed sites in and near the Williamson River channel on both sides of the delta (table 1, scenario C). Travel time to sites farther away from the Williamson River channel in the northern (USGS 25980) and southern (TNC C) delta was 9.4 and 7.2 d, respectively (table 1, scenario C). Within and near the Williamson River channel, travel times were a function of distance from the upstream model boundary and increased from 1.1 d at TNC A to 8.0 d at OSU U6 at the mouth of the Williamson River. Travel times were slower (10.7–13.9 d; table 1, scenario C) to fixed sites in Upper Klamath Lake northwest of the river mouth, and slower still to sites in Agency Lake (15.2–19.5 d; table 1, scenario C). Compared to travel times to sites in Upper Klamath Lake northwest of the river mouth, faster travel times to sites southeast of the river mouth (8.6–9.9 d; table 1, scenario C) indicate that particles were transported along relatively direct pathways across the southern delta and southeast along the shoreline with the currents.


Travel times to sites in Agency Lake and to sites northwest of the river mouth in Upper Klamath Lake were slower than travel times to sites at comparable or greater distances to the southeast because particles arrived at those sites by way of complex pathways. The travel time to site USGS 25981 offshore of the Williamson River mouth was more than 5 d greater than the travel time to site OSU U6 in the channel at the river mouth (table 1, scenario C). Particles exiting the river mouth head primarily southeast, so that the slower travel times to the site offshore indicate that particles came through the northern delta rather than directly through the river mouth. Finally, the slowest travel time (25.2 d; table 1, scenario C) was for particles captured at Bare Island (OSU U8) and indicated transport by the lakewide gyre.


The difference in travel times simulated by dispersal scenarios A–D varied spatially. At sites reached by relatively straight pathways close to the Williamson River channel (TNC A, B, and D, and USGS 25535; table 1) the maximum difference in travel time between dispersal scenarios was small (≤0.2 d). At other sites in the delta and along the shoreline of Upper Klamath Lake, the difference in travel times simulated by the dispersal scenarios was between 0.7 and 1.5 d. At the mouth of the Williamson River, the difference in simulated travel times was 1.5 d in the channel and 1.3 d offshore. The slowest travel time at the mouth was simulated by scenario C, in which the particles did not drift in the river channel during the day, but dispersed passively at night, and the fastest travel time was simulated by scenario D, in which the particles did not drift in the river channel during the day, but aligned with the current at night. Therefore, the particles traveled fastest when aligned with the strong currents in the river. In contrast, at sites reached by more complex pathways, in Upper Klamath (OSU U8) and Agency Lakes (USGS 25978 and USGS 25979), the maximum difference in the travel times simulated by the four scenarios was 3.9–7.4 d.


The fastest travel to the south (OSU U4) was simulated by scenario D (table 1), in which the movement of particles down the river channel was expedited by alignment with the currents at night, particularly the stronger currents associated with higher streamflow during the first pulse of larvae. Consequently, larvae exited the lake relatively faster, beginning on May 28, in scenario D than in the other scenarios, and fewer total particles were in the domain after that date, although the differences were small (fig. 5A). The lowest retention through both larval pulses, based on R_E/A, was obtained with passive dispersal (scenario A), and highest retention was obtained with active dispersal by random swimming during the entire diel cycle (scenario B, table 1).


Comparison of Travel Times by Swim Speed


Random swim speeds also influenced travel times and retention. Differences in simulated travel times between random swimming at speeds of 1.2–5.8 BL/s were small (≤0.8 d) at sites near the river channel and in the delta, and ranged from 0.6 to 4.8 d at sites from the river mouth to the north and south along the Upper Klamath Lake shoreline (table 1, scenarios C-1.2 and C-5.8). Stronger swimming led to more dispersal, both into the northern delta and Agency Lake and away from the eastern shoreline of Upper Klamath Lake, and weaker swimming led to less dispersal. These differences in dispersal were most evident in the second pulse of larvae when there was weaker advection by currents. When these particles began to exit the lake at Link River on June 19, those swimming at 5.8 BL/s were more dispersed and therefore, less likely to leave the lake with the eastern boundary current; consequently, there were more particles in the simulation than when swimming was slower, although the differences were small (fig. 5B). Particles swimming randomly at 1.2 BL/s were less dispersed and more likely to leave the lake with the eastern boundary current, and there were fewer particles in the simulation than when swimming was faster, although the differences were small (fig. 5B). The cumulative retention as measured by R_E/A increased with swim speed and was 1.51, 1.38, and 1.11 for swimming at 1.2, 3.5, and 5.8 BL/s, respectively (table 1).


Comparison of Travel Times by Lake Elevation


Lake elevation influenced particle pathways through the delta (fig. 6; appendix A, simulations at three starting elevations). An elevation near full pool resulted in the most particles leaving the Williamson River channel on the southern delta side, many of them soon after entering the delta. These particles entered Upper Klamath Lake southeast of the Williamson River mouth. As a result, few particles were simulated in the northern delta, at sites in Agency Lake, and along the Upper Klamath Lake shoreline northwest of or at the Williamson River mouth (fig. 6C and table 1, scenario C+). Compared to the 2009 observed lake elevation (scenario C), the large number of particles moving quickly through the southern delta in scenario C+ decreased travel time to site OSU U5, on the Upper Klamath Lake shoreline southeast of the river mouth (fig. 1), by 2.3 d, whereas the travel time to site USGS 25976, northwest of the river mouth (fig. 1), increased by 5.8 d. A lake elevation 0.25 m lower than the observed elevation (scenario C-) resulted in more particles staying in the Williamson River channel and exiting the channel to the north, and fewer passing through the southern delta to enter Upper Klamath Lake southeast of the river mouth (fig. 6A). This happened because the connection between the Williamson River channel and the southern delta was restricted at the lower lake elevation, particularly near the entrance to the delta. As a result, travel times to sites at the river mouth and along the Upper Klamath Lake shoreline northwest of the mouth were faster than simulated at the 2009 observed lake elevation (table 1), and fewer particles were retained in the simulation (fig. 5C), reflecting the faster transport into Upper Klamath Lake. Compared to the observed lake elevation, at the lower elevation more particles were captured offshore of the river mouth and at the shoreline sites northwest of the river mouth (table 1). The overall cumulative retention as measured by R_E/A was greatest at the observed lake elevation (1.38), and less for elevations both lower (1.43) and higher (1.53) (table 1).


Comparison of Travel Times by Shoreline Configuration


Changes in the shoreline configuration associated with restoration of the delta between 2007 and 2009 dramatically altered how particles entered and traveled through the delta and Upper Klamath Lake (appendix A, simulations at three shoreline configurations). When the 2007 shoreline was used, particles entered the lake at the mouth of the river as a densely packed kernel. When the northern delta was reconnected, particles left the channel at several points along the northern side, and many took longer and more complicated pathways to Upper Klamath Lake, entering at breaks in the levees northwest of the mouth or from Agency Lake. This resulted in more particles captured at sites along the shoreline northwest of the mouth, and fewer particles at the river mouth, as well as travel times that were consistently longer than before the northern delta was reconnected, as particles took longer routes to Upper Klamath Lake (table 1). These longer routes resulted in more particles retained during much of the simulation (fig. 5D). When both sides of the delta were reconnected, particles could leave the river channel along both sides. Advection under prevailing wind conditions resulted in more particles moving through the southern delta than the northern delta, so travel times to sites southeast of the river mouth decreased between 2008 and 2009 (table 1) as did the number of particles in the simulation (fig. 5D). The overall cumulative retention as measured by R_E/A was lowest for the pre-restoration shoreline configuration (1.75), increased with the first phase of restoration (0.98), and decreased after the second phase of restoration (1.38) (table 1).


Field Data 


Excluding the boundary site at Modoc Point Road, 24,232 individuals of several species were captured in 2009 (table 2). Collectively, suckers constituted only 6.36 percent of the catch and their proportion declined with distance from the river channel, being 10.6 percent of the TNC samples, 5.5 percent of the USGS samples, and 3.1 percent of the OSU samples.

When the number of suckers caught at fixed sites was greater than 10, there was not an obvious progression in length from shorter to longer with distance from the upstream boundary (table 3). Different gears fished at sites near each other sometimes were similar in median larval sucker lengths (OSU U6 and USGS 25981 differed by <0.1 mm), but also could differ substantially (TNC B and USGS 25535 differed by 2.0 mm). A comparison of the range in the size of fish caught at pop net site TNC B (12–17 mm, fig. 7) to the range in size of fish caught at plankton net site USGS 25535 (10–14 mm, fig. 8), located near each other, suggests differences in gear efficiencies or sampling protocols, such that pop nets catch fewer fish shorter than 12 mm, and plankton nets catch fewer fish longer than 14 mm. Similarly, the range in size at USGS 25981 (12–14 mm), located offshore from larval trawl site OSU U6, where the range in size was 10–20 mm (fig. 9), further suggests that plankton nets catch fewer relatively “large” fish than pop nets.


Model Validation


At pop net fixed sites, larvae were longer than would be expected from ages of simulated particles on the basis of the length-at-age relation by 1.1–2.2 mm (table 3, fig. 7). At the 2 plankton net fixed sites where at least 10 fish were caught, the larvae were shorter than simulated particles by 0.4 to 1.6 mm (table 3, fig. 8). At larval trawl fixed sites where at least 10 fish were caught, larvae were longer than simulated particles by as much as 0.8 mm or shorter than simulated particles by as much as 0.5 mm (table 3, fig. 9). At individual sites without regard to gear type, median lengths were up to 2.2 mm longer than simulations (fig. 7, TNC D), essentially comparable to simulations (fig. 9, OSU U6), or shorter than simulations by as much as 1.6 mm (fig. 8, USGS 25981).


Given that gear efficiencies likely differ, simulated lengths are most appropriately compared to lengths in net catches among the same gear types, rather than across gear types. The simulated and measured differences in median length between a given pair of sites of the same gear type were usually of the same sign but differed in magnitude. At pop net sites, simulated and measured differences in median lengths between TNC A and TNC B were 0.3 and 0.8 mm, respectively; between TNC A and TNC C were 1.3 and 1.5 mm, respectively; and between TNC A and TNC D were 0.7 and 1.8 mm, respectively (fig. 7; table 3). At plankton net sites, simulated and measured differences in median lengths between USGS 25535 near the channel and USGS 25981 offshore from the Williamson River mouth were 2.0 and 0.9 mm, respectively (fig. 8, table 3). At larval trawl sites, simulated and measured differences in median lengths between OSU U6 and OSU U5 were 0.4 and 1.7 mm, respectively; between OSU U6 and OSU U4 were 0.5 and 1.5 mm, respectively; and between OSU U5 and OSU U4 were 0.1 and -0.2 mm, respectively (fig. 9, table 3).


The dispersal scenario that best matched larval lengths depended on how the data were parsed. When all data were combined, the correlation between simulated and measured lengths was highest for passive dispersal (R=0.46, scenario A,table 4). When correlations were made by gear type, plankton and pop net data correlated most highly with passive dispersal (scenario A) (R=0.76 and 0.36, respectively), but larval trawl data correlated most highly with active dispersal by random swimming (scenario B) (R=0.49). Scenario C (active dispersal by random swimming with nighttime-only drift in the Williamson River) resulted in the least overall bias (-0.3, -1.3, and 0.0 mm for larval trawl, pop net, and plankton net data, respectively). Across all dispersal scenarios, correlations with plankton net data spanned the highest range from R=0.30 to R=0.76, and the bias between the simulated and measured lengths was smaller than in the combined dataset, ranging from +0.1 mm to -0.3 mm.


When the data were analyzed separately for particles inserted into the domain before and after June 2, the correlation was significant across all dispersal scenarios for particles inserted after June 2 (R between 0.23 for scenario D and 0.53 for scenario A), when drift was dominated by shortnose suckers and advection by currents was relatively weaker (table 4). Correlation was lower across all dispersal scenarios for those particles inserted prior to June 2, when drift was dominated by Lost River suckers and advection by currents was relatively stronger (highest R=0.45 for scenario A). When correlations were made by size class, consistently positive correlations (R between 0.26 and 0.31) resulted across all dispersal scenarios for the smallest-size class (fish length <13 mm). Significant correlations resulted for the middle-size class (fish length ≥13 and <16 mm) across three of the four dispersal scenarios (highest R=0.39, scenario A). The lowest correlations were in the largest-size class (fish length ≥16 and ≤19 mm).


First posted January 31, 2014