Application of the Stream Salmonid Simulator (S3) model to assess fall Chinook salmon (Oncorhynchus tshawytscha) production in the American River, California
Executive Summary Anadromous fish returning to the lower American River are restricted to 36 kilometers of free-flowing river between Nimbus Dam and American River’s confluence with the Sacramento River, California. Salmon in the American River provide an important freshwater recreational fishery. However, annual salmon production in the American River in recent years has been low relative to the mid-1990s (Surface Water Resources, Inc., 2001). To investigate the low production of fall-run Chinook salmon ( Oncorhynchus tshawytscha ), the Bureau of Reclamation requested that the U.S. Geological Survey apply the Stream Salmonid Simulator (S3) model to the population of fall-run Chinook salmon on the American River. The American River was chosen among seven candidate Sacramento Basin rivers for S3 application. The American River was selected because of its management and public interest, recently low anadromous fish production, and rich time series of key demographic data needed for S3 application. Data that were not available, however, were empirical estimates on juvenile salmon habitat suitability in the American River. Therefore, a large component of applying S3 to the American River was devoted to the estimation of juvenile salmon habitat suitability and capacity. This entailed snorkeling the lower American River for 3 weeks in March 2021 during the early out-migration period for juvenile Chinook salmon. These efforts were fruitful and showed that the typically small fish (<55 millimeters) in the American River preferred much shallower depths than predicted by habitat suitability criteria derived from the literature for this population. Having empirical estimates on juvenile salmon in the American River provided a solid foundation from which to simulate the population using the S3 model. The S3 model is a spatially explicit population model that runs on a daily time step to simulate redd superimposition, egg maturation, fry emergence and the subsequent growth, survival, and emigration of juvenile Chinook salmon from the river. The key features of this model relevant to this report include (1) a temperature-dependent bioenergetics model driving daily growth rates; (2) density-dependent dynamics that are influenced by the effect of flow on suitable habitat area; and (3) within-year habitat, river flow, and water temperature effects specific to spawning, egg incubation, and fry, parr, and smolt life stages. We used estimates of spawning escapement and geo-referenced redd locations to quantify the spatial and temporal distribution of female spawners for brood years 2014–19. These estimates of female spawners initiate the simulation of each year’s juvenile salmon emergence and emigration over a spatial domain extending from Nimbus Dam to the river’s confluence with the Sacramento River. Using weekly estimates of juvenile salmon abundance and size (fork length) that passed the Watt Avenue fish trap (river kilometer 14.7), we calibrated the S3 model by estimating three key demographic parameters for each year, y : (1) S y , the average daily survival probability, (2) M 0y , the intercept for density-dependence in movement, representing the average daily probability of remaining in a habitat at zero abundance, and (3) C y , the average daily proportion of maximum consumption. These parameters were obtained by minimizing the Mallow’s distance (Lupu and others, 2017) between distributions of weekly abundances and sizes of fish at the traps and weekly simulated abundances and sizes (by S3). Investigation of model fit showed excellent agreement between simulated annual abundances and the abundance of fish passing the fish trap. However, when we compared weekly abundances at the fish trap, S3 under-predicted peaks and over-predicted troughs in the time series of weekly abundances at the fish trap. Thus, some unknown within-year effects have yet to be identified and incorporated in the S3 model. Identifying these important effects and incorporating them in the S3 model would help explain the lack of fit between estimated and simulated weekly abundances. We estimated parameters for 6 years that included a wide range of female spawner abundances (3,057–10,753) and water year types (Critical–Wet). We contrast our estimated parameters to the corresponding number of female spawners and the water year type for the Sacramento Valley. By happenstance, years having higher annual spawner abundances concurred with Critical to Dry water year types. Estimates of survival trended lower with higher spawner abundances and Critical to Dry conditions. In contrast, the extremely wet water year of 2017 had the lowest M 0y , suggesting less density-dependence in fish movement, and the lowest C y , suggesting lower average consumption in this year. When this high-flow year was excluded, a trend towards higher probabilities of fish remaining in a habitat at low abundance and lower proportions of maximum consumption was apparent from Critical to Wet conditions, but only 5 years of data were included. Except for 2017, daily proportions of maximum consumption were relatively high ( C y > 0.83), suggesting that fish were feeding at reasonably high proportions relative to the expected maximum consumption as defined by the “Wisconsin” bioenergetics model (Stewart and Ibarra, 1991). Survival estimates from fry emergence to outmigration at the Sacramento River confluence were generally low when integrated over time. The highest daily survival probability was S y = 0.93 in 2019, or 50 percent total mortality after 10 days. In contrast, our lowest daily survival probability was S y = 0.74 in 2015, or 95 percent total mortality after 10 days. Consequently, even our highest estimated daily survival probability might be considered low. This is especially true given that S y was estimated over a relatively short distance (<14.7 kilometers) from emergence to the Watt Avenue fish trap. Several factors, including our assumed and relatively high daily egg survival rate of 0.9975, could influence juvenile survival estimates. For example, an egg survival rate of 0.9975 results in 3-percent total mortality after 10 days. Egg mortality estimates used in S3 calibration were approximated from egg survivorship studies in the Yakima River, Washington (Johnson and others, 2012), and remains one of the greater uncertainties in S3 when estimating survival across life stages. By including bona fide estimates of egg survival in S3 simulations, the validity of the S3’s current daily egg survival rate could be assessed specifically for the American River. Tagging studies also could provide S3 with direct estimates of juvenile survival and movement; survival during egg incubation then could be estimated indirectly via model fitting.