Summary


A hydrodynamic model was used to track particles representing individual Lost River and shortnose sucker larvae in Upper Klamath Lake, Oregon. Simulated travel times were used to determine the length of pathways taken by individual particles to sites within the lake, particularly sites on either side of the Williamson River channel in the recently (2007–2008) re-flooded deltaic area surrounding the lowest 5 kilometers of the river, and sites along the eastern shoreline of the lake north and southeast of the pre-2007 river mouth. Travel times generally increased with distance from the Modoc Point Road Bridge (upstream boundary of the simulations), both along the river channel and away from the river channel into the deltaic areas north and south of the channel. Travel times to sites in the northern delta and along the shoreline of Upper Klamath Lake northwest of the mouth of the submerged Williamson River channel generally were slower than travel times to sites at comparable distances from the upstream boundary in the southern delta and along the shoreline of the lake southeast of the submerged channel mouth. This was because the prevailing winds tend to drive water currents to the southeast across the delta, and, therefore, transport to points south is aided directly by currents and is more direct and faster than transport to points north.


Several different scenarios for individual swimming behavior were implemented: completely passive dispersal, random swimming with and without nighttime-only drift in the river channel, and active swimming through alignment with the currents. The fastest travel times were simulated for particles that aligned with the currents, as this expedited transport down the river channel to the lake and across the shallow southern delta to enter the lake southeast of the channel mouth. This behavior also resulted in the fastest exit of particles from the lake and the least retention. This was in contrast with random swimming, which resulted in slower travel times to most sites than swimming aligned with currents. As the strength of random swimming increased, dispersal of particles increased and retention in the lake increased. Overall, however, the sensitivity of particle travel time and retention to different dispersal assumptions was small in comparison to the sensitivity of these quantities to lake elevation and shoreline changes.


Simulations were completed using three shoreline configurations in order to compare the simulated results from 2009 (after restoration of the Williamson River Delta was completed) to simulated results prior to the start of restoration (2007), and partway through restoration when only the northern part of the delta was reconnected to the lake (2008), assuming that all boundary conditions were the same. The purpose of the comparison was to determine the effect of delta restoration on the travel time and pathways of particles representing larval suckers under one complete set of observed boundary conditions. Overall retention was least for the pre-restoration shoreline, and increased with the first phase of restoration when the northern delta was reconnected, because particles were forced to take longer pathways and to undergo more dispersal when leaving the river channel only on the north side. Retention decreased with the second phase of restoration, as particles were able to take “shortcuts” through the southern delta to the lake, but transport overall was slower and retention greater than occurred when the pre-restoration shoreline configuration was used.


Similarly, simulations were completed using three starting lake elevations in order to compare the simulated results from 2009 to simulated results based on the assumption that all boundary conditions were the same, but that lake elevation was either higher (such that it reached full pool at the maximum elevation during the simulation) or lower (by 0.25 meters) than was actually the case in 2009. The purpose of the comparison was to determine the effect of lake elevation, through a reasonable range, on the travel time and pathways of particles representing larval suckers, and ultimately on their retention in the lake, under one complete set of observed boundary conditions. When the lake was simulated near full pool, more particles left the river channel early and took direct paths through the southern delta, so overall retention decreased and travel times, particularly to points south, decreased compared to the simulation using the measured lake elevation. When the lake was simulated at a lower-than-measured elevation in 2009, access to the delta on both the north and the south side of the channel was more restricted than it was under the measured lake elevation, so more particles remained in the river channel and moved relatively quickly into the lake. As a result, overall retention decreased and travel times, particularly to points south, decreased compared to those in the simulation using the measured lake elevation. Retention, therefore, decreased at lake elevations both higher and lower than those measured in 2009, suggesting that retention in the lake is maximized at some intermediate elevation.


Simulated larval lengths as determined by an empirically derived length-at-age relation were compared to measured lengths of larvae caught in three gear types (pop nets, larval trawls, and plankton nets) throughout the Williamson River Delta and along the eastern shoreline of the lake. The purpose was to validate the hydrodynamic and individual-based model used to simulate larval transport and to determine whether comparison with field data could provide guidance as to which dispersal scenarios were more likely to accurately represent larval behavior. Pearson R values for the various dispersal scenarios varied from 0.21 to 0.46; all were significant (p less than or equal to 0.0169), and the highest correlation occurred with simulations obtained assuming completely passive dispersal.


Because gear types probably had their own inherent bias, we parsed the data by gear type and repeated the correlation analysis with simulated data based on the various dispersal scenarios. The comparison with pop net data showed the greatest bias between measurements and simulations (simulated lengths shorter by 1.3–1.4 millimeters [mm]) and the comparison with plankton net data showed the smallest bias (simulated lengths longer by 0.1 mm to shorter by 0.3 mm) across the various dispersal scenarios. Further, we parsed the data by size of fish—small (<13 mm), medium (13 ≤ length < 16 mm) and large (16 ≤ length <19 mm). The highest correlations with simulations were obtained for the medium size class (R=0.07–0.39 and simulated lengths shorter by 0.7–0.9 mm) and the smallest size class (R=0.26–0.31 and simulated lengths longer by 0.5–0.8 mm). By performing the correlation analysis based on particles entering the lake before and after June 2, we roughly parsed the data by species, because Lost River sucker larvae tend to enter the lake prior to shortnose sucker larvae. The correlation with simulated lengths was higher when restricted to particles entering the lake after June 2, indicating that the simulations were better at describing the dispersal of shortnose sucker larvae, that the field sampling was worse at describing Lost River sucker larvae, or both. When all correlation results were considered, there was no compelling evidence that any of the active dispersal scenarios resulted in a better or worse description of larval dispersal than any other scenario, including the passive dispersal scenario.


Our simulations showed that after leaving the delta, particle travel time and retention were affected by wind speed and direction, lake elevation, and simulated behavior, and that a hydrodynamic model in combination with individual-based models of behavior could potentially simulate the complex response of larval transport to the interaction of these factors. The model was validated with moderate success using net catches, as indicated by moderately strong R values that were often significant. Despite the uncertainties in efficiencies and bias in the various gear types, the unknown aspects of larval behavior and mortality, and the fact that the sucker larvae occur at very low densities in the lake and delta, this study, nonetheless, demonstrated that the model can reproduce overall patterns in larval sucker distribution. This approach could be useful for designing sampling strategies, investigating the effects of climate and hydrologic variability on sucker dispersal, and designing scenarios for lake management to maximize larval retention.


First posted January 31, 2014