At the April meeting of the Delta Stewardship Council, Delta Lead Scientist Dr. Laurel Larsen spotlighted a study funded by the Delta Stewardship Council, State Water Contractors, and the Department of Water Resources that leveraged decades of monitoring data to test competing hypotheses about how combinations of stressors impact the presence or absence of Delta smelt in locations throughout the Delta.
Native fish in the Delta have been dealing with many challenges, including loss of habitat, loss of flows, competition and predation by invasive species, diminished food supply, loss of turbidity that helps them evade predators, and entrainment in Delta pumps.
“Often the science community has grappled with these changes by trying to understand how varying amounts of any one of these stressors impacts one or more life stages of native fish species,” said Dr. Larsen. “These studies contribute valuable information, but they often leave unanswered questions about how interactions between these stressors impact fish or about which are really driving the concerningly low levels of fish populations in the Delta. For example, perhaps there is an apparent relationship between non-native predators and fish populations. But if flows were higher, perhaps this relationship would have negligible importance.”
However, there are big challenges with studying the impact of interacting stressors on fish. For one, a lot of data is needed that spans the full range of possible stressor combinations. Only a few of these synthesis studies have been done, so understanding the impacts of interacting stressors is recognized as a priority science action in the 2022 to 2026 Science action agenda.
In the study, Relations Between Abiotic and Biotic Environmental Variables and Occupancy of Delta Smelt (Hypomesus transpacificus) in Autumn, led by A. Noble Hendrix of QEDA Consulting, decades of data collected through the Fall Midwater Trawl Survey was used to understand how interacting stressors control the presence or absence of Delta smelt in different parts of the Delta in the fall.
Delta smelt typically have one-year life cycles. They are born in the tributaries and migrate downstream to brackish water in the summer, where they feed and grow throughout the fall before migrating upstream in the winter to spawn again. In the fall, Delta smelt reside in the estuarine low salinity zone as they cannot tolerate highly saline waters.
Delta inflows push back the salinity that the tides bring into the Delta, and in drier years, salinity can encroach further inland. When the low salinity zone corresponds with the marshes of Suisun and Cache Slough complex, there’s plenty of good habitat rich with food for them. But in drier years and inflows are less, the low salinity zone moves further upriver, coinciding with agricultural areas in the Northern Delta, which is poor habitat for Delta smelt. Low survival through the fall could be a bottleneck for spawning populations.
“The biological opinion for Delta smelt contains a target value for X2 in the fall; X2 is a measure of how far from the Golden Gate salinity at the bottom of the water column is equal to two parts per 1000,” said Dr. Larsen. “It’s generally thought of as an integrative measure of Delta outflows and Delta salinity. But several recent studies have shown that X2 is only weakly related to Delta smelt abundance, calling into question its appropriateness.”
The researchers used a statistical approach to evaluate how X2 and other measures of stressors for Delta smelt control their fall presence or absence in different portions of the Delta. They utilized a large data set, which presented its own challenges.
“Variables are stressors we can measure like temperature, salinity, and flow,” said Dr. Larsen. “But many of the variables in a comprehensive data set might be related to each other in ways that make it difficult to tease out which variable or variables actually have a direct influence on Delta smelt presence or absence. And sometimes, there are so many variables in the dataset that you could create a model that explains nearly every point it was trained on, but it’s not generalizable and would actually predict other data points that the model hasn’t seen before very poorly.”
So the researchers dealt with the challenge of too much big data by convening many Delta smelt experts and having them generate hypotheses about how multiple variables together might control whether Delta smelt are present or absent at a particular location. Then the researchers translated those hypotheses into alternative statistical models and tested them by how well they predicted Delta smelt presence or absence with a test set of data.
“It’s an innovative approach that isn’t often done in statistical modeling and could be a way forward also to incorporate things like traditional knowledge that usually aren’t incorporated into statistical analyses into these studies,” said Dr. Larsen.
“What the researchers found was really illuminating. In four of the top-performing models, X2 didn’t appear in any of them. Rather, the salinity measured at each particular location was one of the most important variables. So the interpretation is that X2 might not be a good enough predictor of salinity conditions experienced in these specific regions of the Delta.”
“One of the top four models was a much better fit to the data than the others, and it only had temperature and salinity,” continued Dr. Larsen. “But one of the other models also added slight predictive power when used in combination with that model. And that model additionally included water clarity, threadfin shad abundance, and an index representative of predation intensity. Specifically, clear water diminishes the likelihood of finding Delta smelt in particular locations; threadfin shad is thought to compete with Delta smelt, and so that also diminishes the likelihood of finding Delta smelt in particular locations. And then finally, very interestingly, Delta smelt were more likely to be present where the predation index was high, which might just indicate that conditions that favor Delta smelt also favored their predators.”
“Still, the most important variables by far appeared to be temperature and salinity. And it will be very interesting to see whether future biological opinions might take these new findings into account and move away from X2 as the primary regulatory standard.”
During the discussion period, Dr. Larsen further explained that X2, as a measure, doesn’t indicate how salinity is spatially distributed through the Delta or whether salinity is appropriate in places that are good habitats for Delta smelt. “I think what that argues for is that we have a robust monitoring program that provides us real-time information about the spatial distribution of salinity and that monitoring is incorporated into operations in near real-time.”
