Fisheries Biologist Cyril Michel discusses recent predator manipulation studies done in the Delta and efforts to develop a predictive model for predation
Over the past century, populations of salmon in Sacramento-San Joaquin Delta have declined drastically, with at least one population that is locally extinct and the remaining listed as endangered, threatened, or species of concern under the Endangered Species Act. Evidence from long‐term tagging studies suggests that the survival of juvenile salmon during outmigration has a disproportionately large impact on juvenile‐to‐adult return ratios and that low survival while transiting the Delta during outmigration due to predation may be one of the major contributors to the declines of these populations. Predation is a challenge in the Sacramento–San Joaquin River Delta where non‐native predators are known to have substantial impacts on salmonid and other native fish populations; however, resource managers lack the knowledge of the landscape‐scale predator–prey information to mitigate these impacts.
Cyril Michel is a Fisheries Biologist with the University of California Santa Cruz and the team leader for the salmon acoustic telemetry and salmon predation programs at the University of California Santa Cruz. He also has an affiliation with National Marine Fisheries Service Southwest Fisheries Science Center. These two programs are both currently maturing and moving from assessing the spatial and temporal dynamics as well as environmental drivers of juvenile salmon survival and predation risk to the experimental phase with different studies testing ways to manipulate juvenile salmon survival and predation risk on a landscape scale. At a webinar held at the end of August 2020, Mr. Michel discussed the research he and his team are doing on studying salmon predation in the Delta.
Salmon in California
In California, the most robust runs of salmon are the chinook salmon or king salmon species, which are the largest species of salmon on the West Coast. The salmon can weigh up to 130 pounds, although usually, they are much smaller when they are adults, about around 20 to 40 pounds. Salmon support an approximately $100 million recreational industry and $80 million commercial fishing industry in California, although it varies from year to year depending upon run size.
There are two main rivers in the Central Valley, the Sacramento River and all of its tributaries which flows from north to south, and the San Joaquin River, which flows from the south to the north. The two rivers join in the middle, flow out into the San Francisco Bay and out under the Golden Gate Bridge.
Historically, in the 1800s, California’s Central Valley was pristine habitat for chinook salmon spawning and rearing. There were thousands of miles of river for spawning and rearing and plenty of food rich wetlands which supported some of the largest runs of chinook salmon anywhere in the world. However, as California’s population began to grow in the mid to late 1800s, partly due to the Gold Rush, the Delta became increasingly modified by the building of levees and the draining of wetlands for agriculture, which resulted ultimately with 90% of historical wetlands becoming disconnected from the rivers and no longer accessible to salmon and other fish.
Due to the increasing demands for water for both the increasing population and agriculture, dams for water storage and hydropower were built all around the Central Valley, and most of those dams did not provide fish ladders which blocked access for the fish to access the watersheds up above and essentially disconnected the river at many different points. It’s been estimated that about 47% of the historical spawning and rearing habitat became inaccessible, Mr. Michel said.
Many thousands of diversions were built along the Central Valley’s rivers mostly for agricultural purposes; these diversions can remove up to or over 40% of all the water in these rivers every year.
“This is so bad that in certain parts of the watershed, all the water is removed,” said Mr. Michel. “A glaring example is the San Joaquin River, which is the second-largest river in California, and in many years it actually runs dry in a large portion of it due to the amount of water that’s diverted that essentially cuts off that habitat completely for salmon.”
Add to that, over the course of about 120 years, there have been multiple introductions of different fish species, in particular different predatory fish species that eat juvenile salmon. Many of these are formidable predators of salmon, even to the point that the Sacramento-San Joaquin Delta is now considered one of the best fisheries in the world for basses.
The result of all of this is that the populations of chinook salmon in the Central Valley aren’t doing so well, he said. There are three main runs of salmon in the Central Valley: The winter run which is considered an endangered species, the spring run which is considered threatened, and the fall/late-fall complex which is considered a species of concern.
Partly to mitigate for these stressors, 35 million juvenile salmon are released every year from hatcheries throughout the state – so many that it is estimated that approximately 90% of the salmon caught off of our coast are actually of hatchery origin. Mr. Michel noted that these hatchery salmon themselves may be causing harm to the remaining wild populations in the system.
All of these stressors have likely led to these declines that led to the listing of these different populations, but to really understand the mechanisms of how these populations are being affected requires revisiting the salmon lifecycle. He briefly explained that the adult salmon come in, they spawn in the river, they lay eggs, the juveniles emerge from the gravel and rear in the river for a while, they get bigger, and at some point they make the decision to migrate to the ocean. At which point, they feed a lot more in the productive waters off of California, become much bigger, and finally come back again as adults to spawn again in the rivers.
“It is well known from the literature that this passage of juvenile salmon (or smolts) from the river to the ocean is one of the most perilous stages of this lifecycle, and in California, there’s no exception here,” said Mr. Michel. “Multiple studies since 2007 have used acoustic tags to track the survival of salmon outmigrating through the Central Valley. We implant these acoustic tags into juvenile salmon, we release large batches of fish from the headwaters of these rivers or the spawning-rearing origins of these rivers, and track their migration out through the river system using acoustic receivers to see what their ultimate fate is as they enter the ocean. These studies have shown that we see about 5% survival to the ocean on average, which is much lower than other large West Coast rivers and does not allow for sustainable fisheries.”
There are many possible reasons for why juvenile salmon survival is so low, but for this presentation, Mr. Michel concentrated on the predators. He noted that predation on salmon in California is a fairly big deal, so much so that it was actually a stated directive in the WIIN Act which passed on December 16 of 2016 that there should be research projects into predator removals on the Stanislaus River for the purposes of increasing survival of salmon.
Predator manipulation studies
The researchers wanted to help resource managers answer the questions, how can we reduce predation on juvenile salmon? Can we just simply remove some of these predators? So they took those two questions, and formulated a research question for their first study, which was, are localized predator removals feasible and effective at reducing predation and ultimately increasing salmon survival?
The first study in 2014 and 2015 used boat electrofishing to remove predators from certain portions of a river and relocate them to other portions of the river to see if these manipulations of adding and removing predators had impacts on salmon survival. Mr. Michel explained that boat electrofishing applies electric current to the river which temporarily stuns the fish long enough so they can be netted, put in tanks, and moved them to another location.
The predator manipulation experiment was a Before After Control Impact (or BACI) design that was done in the Lower San Joaquin River in the Delta which is tidally influenced. They split the lower San Joaquin River into three clusters of one kilometer long sites. Each cluster had a control site where they did nothing at all, a removal site where as many predators as possible were removed, and an addition site where predators were added to supersaturate the site. Approximately 2000 predators were removed in the two years of the study. The predators were largely largemouth bass and striped bass both of which are non-native predators in the system.
They wanted to test two things: do removals and additions of predators change predation risk and does that affect salmon survival?
To answer the first question, is predation risk reduced, predation risk must be measured. Their primary tool for this was called a Predation Event Recorder (or PER) which is a floating buoy with a live chinook salmon attached to the bottom. It floats through the water, and when this fish is predated on, a timer and GPS logger gives the exact time and location of the predation event. So the researchers deployed ‘dozens and dozens’ of PERs in a study site over the course of the day, and over time they could start to understand the relative predation risk of different sites and different conditions.
“This is a great tool for us,” said Mr. Michel. “It’s easily repeatable and it’s standardized so we can deploy this at many different sites over the course of a season or several years and have confidence that the measurements are the same.”
To answer the question, have we reduced predation risk, Mr. Michel presented results from the PERs work in 2014 and 2015; the x-axis is the week of sampling and the y-axis is the percent of PERs that were predated on. The blue circles are the control reaches, the light blue triangles are the predator addition reaches, and the orange squares of the predator removal reaches. The black horizontal line represents when the predator manipulations happened and when they removed or added predators.
“Through this work we did some modeling that indicated that there was no statistical difference in predation risk for the addition and removal sites between the before period and after period, Mr. Michel said. “So in short, we were not able to reduce predation risk.”
To answer the second question, was salmon mortality reduced by the predator manipulations, they implanted acoustic tags into juvenile salmon and released large groups of the juvenile salmon and tracked their migration down the river, measuring survival through each reach. The yellow stars represent where receivers were which would delineate the reaches so they could estimate mortality through these different reaches.
The plot shows the results from one of the study years for the acoustic telemetry work. The x-axis shows the river kilometer going from upstream towards downstream on the right, and the y-axis shows survival per kilometer. The solid symbols represent releases of fish that occurred before the predator manipulations, and the open symbols show the survival of the group released after the predator manipulations. The color coding here shows the removal sites in orange and the addition sites in blue.
“Again, using some models, we were able to determine that there was no statistical difference between before and after treatments, whether in the removal or the addition sites,” said Mr. Michel.
Mr. Michel then reviewed the lessons learned from the study. “The first thing we learned is that localized predator removals don’t appear to work in this system, or at a minimum, that there’s more powerful drivers of predation risk and salmon survival and the signal of the predator removals is being swamped out by these more power drivers. So this was an important realization for us that we really need to better understand these other drivers of predation risk in salmon survival.”
Developing a predictive model
“We could not really explore the larger landscape scale drivers of predation from our 2014 and 2015 work because we really had a limited spatial scale – only about 15 kilometers of the lower San Joaquin River,” he continued. “This is an example of a bigger problem which is there is a disconnect with the spatial or temporal scales at which ecological processes are being studied versus the scale at which management actions are being applied. In the highly managed Sacramento-San Joaquin Delta here, there’s actually a lot of different tools, dials, and levers that managers can use to affect the environment. But in the Delta, the vast majority of the research on predation at least has only occurred in discrete locations, so managers are tasked with using this information from small spatial or temporal scales and using it to make decisions that affect the Delta as a whole.”
For the next phase of work, they decided they needed to take a holistic approach to studying predation and mortality in the Delta. They wanted to develop tools that would do three things:
Identify the sources of mortality that are most important across the estuary,
Use that to predict when and where along the migratory corridor that these fish move through may be becoming hostile,
And to predict the landscape level impacts and success of potential mitigation measures.
The challenge of working on a landscape scale is how to collect finite information on predation that can be reasonably extrapolated to the landscape scale, and so to do this, they employed the Generalized Random Tesselation Stratified protocol or GRTS, which randomly selects field sites while balancing selections across the region of interest. This allows researchers to then defensibly make extrapolations across the landscape.
In 2017, they did a pilot study that focused only on the south Delta. They considered 300 kilometers of waterways, and randomly selected 21 sites in total for the area.
They again used the Predation Event Recorders of the PERs, which are a valuable tool to measure not only relative predation risk, but also predation risk in response to environmental or habitat variables. The other important tool was a Didson Camera, which works much like an optical camera but instead uses sonar to see, and it allowed them to county and measure the predators in the study sites.
They mounted two Didson Cameras to each side of the boat and ran transects along a 1 kilometer study site, and counted the predators that they saw. On the plot, the hatched area represents the scanned area with the Didson Cameras and the black dots are the predators that were identified. They then used the predator densities estimated from this work to see how predator densities impact ultimate predation rates.
In the spring of 2017, they randomly selected 21 sites including three repeat sites; the repeat sites were visited weekly through the season which then allowed them to also look at the trends across the season of predation as well as the spatial trends in predation, which was the goal of the project. They sampled from 3 hours before sunset to 1.5 hours after sunset to amplify the predation signal; in their previous work, they had found that predation rates were highest at sunset, and they wanted to have a lot of predation events so that they can best generate robust relationships between environmental and habitat characteristics in predation.
In total, 1,670 PERs were deployed over 6 consecutive weeks during the spring; Mr. Michel noted that the spring of 2017 was a historically wet year in California, which means any relationships generated as part of the study were likely only be representative of what wet years are like in California.
The plot shows the relative predation rate. Each one of the hollow dots represents a daily predation rate, or what percent of the PERS that were deployed were predated upon; the x axis shows the date of the sampling; and the y axis, shows the percent preyed upon. The black dots represent a weekly mean of the repeat sites only; the repeat sites were visited each week which allowed them to keep the spatial variability constant and look at the seasonal trends as well.
“What we see is that the first four weeks of the study, predation rates were relatively constant and in the last two weeks, we saw a large increase in predation,” said Mr. Michel. “More interesting is when we started to relate different habitat and environmental variables to see if we can predict what predation rates are like. We used a whole suite of habitat and environmental variables that we believed would have influence on predation. That included temperature, dissolved oxygen, turbidity, river depth, but as part of that also the bottom roughness which was the coefficient of variation in depth as well as the bottom slope. We also measured and used submerged aquatic vegetation, predator densities as estimated from the Didson cameras, water velocity which was estimated from the PERs speed through water, distance to shore, and time to sunset.”
On the slide, the plot on the top left shows the path of the PERs and the red dots are where the predation events were recorded. The plot on the top right shows the predators as identified by the Didson Cameras. The bottom plot shows some of the habitat feature variables that were collected; there was submerged aquatic vegetation on the edge, levees along the shore, and deep scour holes in the river. They then used a Cox Proportional Hazard Model to see how the different environmental and habitat variables influence predation risk.
“We used every possible combination of those habitat and environmental variables to build multiple different models as we wanted to see which model had the most ability to explain the dynamics in predation risk,” Mr. Michel said. “Of all those models that we tested, the top model had four covariates in it. And so I’m going to walk you through the response plots for those four different variables.”
The response variable in the following plots is the predation hazard ratio which is the factor change in predation risk. The first important covariate is the distance to the nearest predator; in the plot, distance to the nearest predator is on the x-axis; if there’s a short distance to the nearest predator, that means there are high predator densities and if there’s a long distance to the nearest predator, that means there are low predator densities. For the predation hazard ratio, a “1” means essentially no change in predation risk, which is shown by the dotted line; if the distance to the nearest predator is short, there is an increase in predation risk.
“Here, this is about 1.7 so there is a 1.7-fold increase in predation risk over mean conditions when the distance to the nearest predator is short and vice versa when the distance to the nearest predator is far, we have a .5 so a 50% reduction in predation risk over mean conditions,” said Mr. Michel. “The take home here is predator density does matter for predation risk, maybe unsurprisingly.”
Bottom roughness is an important as well such that when the bathymetry is more complex, there’s a higher predation risk. Time to sunset is very important as was predicted. They knew from experience that predation risk is highest near sunset and indeed the model showed the highest predation risk was about a half-hour after sunset, which was a three-fold increase over mean conditions. The most important predictor of predation risk was temperature. At the higher end of the measured temperatures, which was 20 degrees Celsius, there was a predation risk of four-fold increase in predation risk over mean conditions.
The next step of the study was to assess predation risk across the landscape by extrapolating what they had learned and try to make predictions of what predation risk would be through space and through time. In order to that, the researchers had to collect the four covariates at the same scale.
“So for temperature, we interpolated temperature from 27 existing temperature gauges and we interpolated them across the landscape,” said Mr. Michel. “For bottom roughness, we used a digital elevation map. Time to sunset was fixed at a mean value. Predator densities were a little trickier. Fortunately, one of the coauthors Chris Loomis, as part of his thesis, developed a predator submodel, which is a model that predicts predator densities based on environmental habitat features. As part of this work he found that sinuosity, number of submerged aquatic vegetation patches, and bottom roughness was the best predictor of predator densities, so using the submodel, we can also estimate what predator densities are on the landscape. Taking all of these different components and adding them together, we can then try to predict predation risk upon the landscape.”
The next step was to break up the landscape into 1-kilometer segments, in total, over 300 1 kilometer-long segments. They then used the model to predict what predation risk was at a day step and 1-kilometer scale resolution.
One thing they found was that predation risk changes depending upon the time of year. Early in the season, predation risk was fairly low across the system due to cold water temperatures, but as the spring progressed and water temperatures become warmer, there were increases in predation risk to the point where by late May, predation risk skyrockets.
Mr. Michel said it was informative to see how it changed through the season as it allowed them to see the spatial resolution of predation risk. “To highlight that a little bit, I zoomed in on a couple of different spots. There is the lower portion of the south Delta, there’s a lot of different areas that have differing predation risk, and near the upper portion of the south Delta, we see that predation risk seems mostly homogenous except for one particular site that has very high predation risk which happens to be the Head of Old River site.”
He then turned to the remaining results and why it might be important to salmon management in California.
One of the ways this work can be applied is to objectively determine areas of high predation risk. In the Central Valley, there are predation hotspots that are discussed, but there isn’t necessarily a lot of work that has gone into identifying them objectively, but this work might be a way to do that.
“We make these day scale predictions of predation risk across the landscape, and we can look at the sites that happen to have higher than average predation risk throughout long portions of the season,” Mr. Michel said. “On this plot here, the x-axis is the percent of daily hazard ratios that are above 1, which means above average, so these are essentially sites that have at least 50% of the season or more estimated to be higher than average. I ranked them by how bad they are, and the worse site up here has 100% of the days of the season has predation risk above 1, so higher than average. And that site is the Head of Old River site, which is interesting for many reasons.”
For one, the Head of Old River site spurred the original predator removal project because it was known to be an area with high predator densities and they were asked to research if removing these predators would help. Also, it is a known area where predators are a problem, but it was not studied as part of this study, but the model was effectively able to estimate this known outlier without ever visiting the site so that gave the researchers some confidence that the model wasn’t completely wrong.
The work could also be insightful to managers by making it possible to consider the predation risk along the migratory corridor. “If you consider the migratory corridor of a fish, you have to consider that a fish needs to move through multiple 1-kilometer cells to move out into the ocean, and so we can start to look at the cumulative impacts of these 1-kilometer cells along the fish’s migration and use that to make a prediction of what survival might be on any given day if you were to migrate through all of these cells,” said Mr. Michel.
The red line on the map shows the migration of a fish through all the different cells, and the plot shows the result on a day time-step. “For example, on April 1st, if a fish were to transit all of these different cells with their different predation risks, what would be their predicted outmigration survival for those fish. We can look at the seasonal changes in outmigration survival as a result of the changing predation risk, and what we see for this spring of 2017, was that through Delta survival was relatively good during the late March early April period, but then as temperatures increase, especially into early and late May, survival really goes to zero.”
Mr. Michel reminded that these are just predictions based on their models and it still needs to be ground-truthed, but it is good to put this in context of what actual salmon were doing that year. The Kodiak Trawl is operated on the San Joaquin River on a near-daily basis, so they know the relative abundance of salmon entering the Delta on any given day.
The bar chart shows the catch per unit effort from the trawl and so represents salmon entering the Delta. “What we see is in the spring of 2017, the majority of the salmon were entering the Delta during the worst time, when our model was predicting total outmigration survival of near zero,” he said.
He then proposed three hypothetical management scenarios that could be informed by this work:
Scenario #1: Earlier hatchery release. Weather and hydrology could be used to build predictive models to inform hatcheries about good times to release fish for survival. This work could have resulted in fish being released earlier and these fish benefitting from much better conditions during their outmigration survival and ultimately better survival to the ocean.
Scenario #2: Mitigate hotspots. For example, the Head of Old River was an outlier for predation risk, so if the Head of Old River site was mitigated so the predation risk was similar to other sites around it, what would that do to total outmigration survival? They ran a model scenario, the results of which are shown on the graph. The red line is the status quo outmigration survival estimates, and if the Head of Old River site is mitigated to be similar to the others, the purple line indicates what outmigration survival would be.
“It does increase outmigration survival during the early portion of the spring, but come May, water temperatures are so warm that essentially all the sites have high predation risk, so just fixing that one site doesn’t really fix the problem and total outmigration survival is still poor,” said Mr. Michel.
Scenario #3: Manage temperature. Another scenario is imagining if temperature can be managed in the Delta. Mr. Michel noted that while we might have control over when hatchery fish move through the Delta, we don’t really have control over when the wild fish are going to move through and they are the fish we’re most concerned about, so if we could increase survivorship of salmon through this May period when a lot of fish seem to be moving through, it could be great for wild fish.
He then proposed a hypothetical management scenario where temperatures are decreased by one degree. The purple line represents what the predicted outmigration survival would be on a day time step if water temperatures were decreased Delta wide by 1 degree through the spring season.
“What we see is it actually doubles survival on any given day, although doubling survival when survival is already near 0 doesn’t do much, but it does help with this little bump down here when fish might be outmigrating, but there’s still this early and late May period which has near zero survival,” he said.
Future management implications
Mr. Michel said they need to repeat the study in other major regions of the Central Valley; so far, it’s just been the south Delta but could be performed in other parts that are of interest. They also need to do the study across more water year types; 2017 was a wet year, so drought or average years also need to be studied. Then, build a predictive model that uses weather and hydrology forecasts to build a predation risk model that has a predation risk prediction forecast for three to five days out, for example.
“We really need to pair these studies with smolt physiology studies to also understand the prey vulnerability dynamics in response to local conditions,” he said. “Because of the fact that these PERs have a fish that’s tethered to the PER, there’s not a lot of prey side dynamics that are allowed, so we’re really starting to study the predator side of the predator-prey relationship. We also need to compare the predation mortality predictions that we generate to actual survival data from these acoustic tag salmon, and that will assist us in ground-truthing our model, but I think it will also allow us to determine the contribution of predation to mortality. Finally, I think it will further develop the mechanistic understanding of variations in survival, whether they are really predation or other things.”
How PERs might be used in other systems
Lastly, Mr. Michel suggested ways in which PERs could be useful for studies in other system.
One other way they have been using PERs is to investigate point sources of predation or contact points. “If you release enough PERs over time to the river, you eventually get full spatial coverage of these PERs, and then you generate a heat map of predation risk from this PER data while standardizing for the amount of PER time spent in each cell, and you can figure out where the predation hot spots are. For example, it shows a predation hot spot right behind a water diversion on the river here.”
They have also used PERs to see if artificial illumination might be increasing predation risk. To do this, they added an artificial light source which cast a large beam of light across a section of the river, and then deployed dozens and dozens of PERs to find out where the predation events were happening. The white dots show the predation events and the heat map shows that the predation risk was highest near the light source.
Another application of the PERs is to add cameras to them to identify the predator. In the 2014-2015 season, they added GoPros to all of the PERs and identified 285 striped bass predations and 36 largemouth bass predations. The plot shows the distance to shore of the predation event and the predations per effort; the red bars are largemouth bass predation events and the blue bars are striped bass predation events.
“What we see is the largemouth bass predation events tend to occur nearshore while the striped bass predation events tend to occur further off shore, which is not a big surprise to people familiar with the system,” said Mr. Michel. “What’s interesting too is you can parse this out by predation events occurring below 20 degrees Celsius versus above 20 degrees Celsius; there’s also an important dynamic there where these largemouth bass are fairly inactive at these lower temperatures but they tend to wake up above 20 degrees Celsius and there’s a lot more predations in this nearshore littoral zone.”
QUESTIONS & ANSWERS
Question: With respect to the reduction of predation risk, how do you know enough removals occurred? Perhaps risk was not changed due to predators are more abundant than removals were able to compensate for?
Mr. Michel: “It was unclear to us from the onset how many predators needed to be removed to have an effect, and that kind of spawned a new set of questions.”
Question: What percent of the resident population in each removal zone was reduced during the electrofishing? Did you resample predator abundance in the removal, control, and addition zones?
“Yes we did,” said Mr. Michel. “I believe we felt it was around 50% of the predators were removed at the time of the predator removals. At the end of the talk there is a link for the paper where you can look at the real numbers, because I’m trying to remember. We did try to estimate what proportion of the predator population were removed during these efforts.”
Question: Is predation primarily an introduced species issue or a water management issue?
Mr. Michel: “I think that’s up for debate. I would say the fish community in the Delta is more hostile to salmon nowadays then it was when it was strictly native fish communities, but on the flip side, I think the bird community was much larger historically. Historical accounts say that the Central Valley skies were black during certain times of the year, there was so many birds, so perhaps back in the day, avian predation was a bigger deal.”
Question: Based on your article, 262 of 1167 PERs were predated on. Adding your diet content analysis from earlier electrofishing work, would you agree that migratory juvenile salmon is not the majority diet of these predatory fish?
Mr. Michel: “I think that is accurate. Especially in the south Delta where there are much lower densities of salmon than the north Delta or even less than the upper river portions, I think salmon are these transients that are moving through and they are only there for a few months a year, so I think these large predator populations are being supported by other forage and the salmon are just kind of guest visitors. What’s interesting in the south Delta, in particular, is actually these large populations of non-native predators are being supported largely by large populations of non-native prey, including crawfish and Mississippi silverside and other non-native or invasive prey items.”
Question: It looks like there’s a lot of variability. Would a larger sample size perhaps show some trends? Was avian predation an issue, too?
“Those plots that I showed indicated a lot of variability in predation rates, but that is also a function of how we plotted it there. We do feel like we collected sufficient information using the PERs to address the central question, but also if the signal of a predator removal is really low, for example, if we expect predation risk to only change by 1 to 5% which is maybe still something worth striving for, yes at those levels, maybe we wouldn’t detect changes.”
With respect to avian predation, in the portion of the river we were working, avian predation does not appear to be a big problem. We rarely saw any fish-eating birds around, but this is also an untapped area of research in California that more research will start thinking about.”
Predation Event Recorders (PERs)
Question: Could a PER be used to measure predation on salmon parr?
“Yes, I believe so and we’re attempting that this week,” said Mr. Michel. “We’re taking it to the next step which is we are attempting to attach salmon fry to PERs and we’re developing that now. We’re doing some work in Redding. If any of you have been to Redding, there is landmark here called the Sundial Bridge. It produces quite a bit of artificial illumination on a very important part of the Sacramento River where these endangered winter-run fry emerge from the spawning redds, so we’re going to look into if the illumination coming off the bridge is influencing predation risk on these salmon fry. We’ve had to change our game to figure out how to attach these much smaller and more delicate fry-sized fish, we’re talking 35 to 55 mm sized fish to these PERs so more on that soon.”
Question: The PERs don’t identify predator species. Is it possible that the focus of the predator removals, striped bass and largemouth bass, are inordinately represented in your study and other less representative species such as catfish account for far more predation than the study is able to capture?
“PERs are only going to sample the area where you set them to sample and our PERs were floating PERs sampling the upper water column and catfish are more bottom oriented fish, so it is possible that we’re not really capturing the dynamics driven by predators that are more bottom oriented,” said Mr. Michel. “I think future studies could try to target that part of the river as well, but it’s worth mentioning that the literature seems to indicate that when salmon are migrating through the Delta, they are migrating through the central channel and the upper water column, so in a sense, we might be doing a good job of sampling in the areas where the salmon are most likely to be found.”
Question: How much do PERs cost to use? How much does it cost to implement?
Mr. Michel: “A lot of the work we’ve done in the Central Valley looking at salmon survival uses acoustic telemetry or coded wire tags which is different kind of tag, and both of those systems are very, very expensive although very informative. PERs are dirt cheap in comparison so I think our average PER probably cost, we’ve had different variations over the years but I say could make a PER for about $300 a piece and you can probably only make 10 to 15 and just recycle those and deploy them over and over again, so that’s what we tend to do. Fairly cheap and easy to deploy.”
Question: Did you lose many PERs?
“We’ve only ever lost two,” said Mr. Michel. “One sunk and the other one escaped and we never found it again. We’ve had variations of PERs, we’ve had some PERs that had real-time tracking so we can actually find them again if they escape our view, and we’ve had other PERs that only had passive tracking so we download the GPS data after the fact, and if we lose sight of those, they are really gone.”
Question: Can you speak a little bit about the significance of these results relative to the low flow conditions in 2014 and 2015, that is, during the drought?
Mr. Michel: “In 2014-2015 when we attempted those predator removals, that was right in the middle of a historic drought here in California and I think that played a large role in the fact that we did not measure a response to our predator removals. That’s what I insinuated earlier, which was that the noise created by this drastically changing environment swamped out the signal of the predator removals, and what I mean by a drastically changing environment is, due to this drought, water temperatures in the Lower San Joaquin River were abnormally high for the springtime to the point where we actually had to end one of our study seasons early because the salmon were not surviving when they were attached to our PERs just due to poor water conditions.”
“So that’s an important point, and perhaps studying predator removals should be done again during more normal conditions, and I know there’s some work going on in a major tributary here in California in the Stanislaus where they are looking into predator removals again, so more research to come on that, I’m sure.”
Question: Have you looked at the water temperature record along the outmigration route in flood versus non-flood years, wet versus dry?
Mr. Michel: “I’ve looked at that and I’m familiar with it. I haven’t yet attempted to use our predation risk model to make predictions in out of sample years. I could easily take our predation risk model and estimate what predation risk might have been in a drought year, but I kind of feel like it’s not totally appropriate until we collect more data in different year types, so I haven’t really taken that plunge, but certainly the conditions in 2017 were fairly good for salmon so I can only imagine that our predation risk model would predict some pretty poor outmigration survival in years such as 2014 or 2015.”
Question: If the water temperature is high and salmon survival is low in May-June, what are the consequences for CV fall chinook life histories, given this is the primary smolt emigration period?
Mr. Michel: “I think there’s some good research that’s come out of different labs .. that when winter temperatures are high, especially during this period of May and June when historically lots of fish would outmigrate, we’re essentially disconnecting the migratory corridor. The migratory corridor just doesn’t work anymore, and we’re in essence kind of lopping of this life history diversity so the tag end of the historical outmigration window is being lopped off which means less diversity which means less ability to handle other trends and dynamics.”
“For example, ocean entry is considered another important period for salmon and when you enter the ocean might be really important to your ultimate survival based on how much productivity for example in the ocean. If fish are no longer able to exhibit some of these historic migration trends because of the migratory corridor is essentially cut off, then these fish might have issues dealing with pertubations later on in their list cycle.”
Question: Was flow a factor considered in your initial predation risk model development?
Mr. Michel: “Yes, flow was definitely an important factor. Flow drives everything in a river, so we included flow in preliminary runs and then we realized that really the way flow manifests itself at the resolution we’re looking at is through water velocity so we included water velocity in the modeling efforts and surprisingly, it did not come up as one of the top covariates, so we’re still trying to understand that. It could just be it was wet year and in a dry year, water velocity might be much more important.”
Question: What do we do with all the removed fish? Since relocation within the same general area is not a solution, as demonstrated by your removal translocation work.
Mr. Michel: “Removing predators, in this system at least, is controversial, and a lot of large, powerful fishing advocacy groups really enjoy fishing for striped bass and largemouth bass. So when you talk about removing these predators, it can really cause problems, and that’s part of why we designed the study as we did, because we saw value in having a supersaturated predator site for one, but also realized that it would be less controversial to just relocate predators than to remove them completely.”