Newly published paper: Using Bayesian hierarchical models to estimate coho salmon escapement to a river with limited data

The first chapter of my doctoral dissertation was recently published in the Canadian Journal of Fisheries and Aquatic Sciences. The article entitled Coho salmon escapement and trends in migration timing to a data-poor river: estimates from a Bayesian hierarchical model has recently been uploaded to the CJFAS website in publication format.

As fisheries management agencies consider shifting towards more ecosystem based approaches to managing fisheries, they need to manage species which have historically garnered limited interest from fishers and researchers. As such, there is generally limited information on the basic ecology of these species in the systems which are being managed. In the Chignik salmon fishery on the Alaska Peninsula, fisheries management and harvest has historically focused on sockeye salmon. The system also supports a population of coho salmon, but they are not managed or directly targeted for harvest due to low economic value and logistic factors. However, previous research (Ruggerone and Rogers 1992) has estimated that juvenile coho salmon consume over half of the emerging sockeye salmon fry in the rearing lakes annually, presenting a potential predation bottleneck to the productivity of the sockeye salmon fishery. Therefore, there is increasing interest in managing the coho salmon population for increased harvest in order to reduce this predation pressure on sockeye salmon.

As there has been limited interest in coho salmon historically, there are limited data available to estimate the numbers of coho salmon returning to the system each year. Sockeye salmon are enumerated at a seasonal weir. However, due to a later spawning migration time, coho salmon are just starting to return to the system when the weir is removed for the season. Therefore, only the beginning of the coho salmon run is counted. In this paper, co-author Daniel Schindler (UW) and I used a Bayesian hierarchical modeling approach to estimate the number of adult coho salmon returning to the system in years for which limited data are available. The Bayesian hierarchical model structure assumes that there is a river-level mean escapement date, migration duration, and escapement size, and that the peak escapement date, migration duration, and escapement size in any given year are drawn from a distribution around these river-level means. This allows the model to use years with more escapement data to inform years with less data available. Our estimates of escapement were more precise in years for which more daily escapement counts were available, and less precise when fewer data were available, relying more heavily on the historical mean values than the few observations in those years. The Bayesian hierarchical model structure also provides estimates of uncertainty around the annual escapement estimates.

Additionally, we examined the trends in peak escapement timing over time and in relation to broad-scale environmental conditions. We found that coho salmon escapement is negatively correlated with PDO index, being earlier in positive PDO years, and that it is getting later over time. However, the significance of these trends depends on the assumptions made about the shape of the spawning migration arrival timing. If we assume normally distributed arrival timing, only the relationship with PDO was significant. If we assume a gamma distributed arrival timing with a long descending limb, only the relationship with time is significant.

Overall, our results have implications for the management of any future coho salmon fishery that may be implemented. The escapement estimates allow for the calculation of escapement goals, under either single-species or multi-species management frameworks. Further, knowing the productivity of the coho salmon populations allows us to simulate the fishery dynamics under different harvest scenarios, as well as under different environmental and economic conditions. Such simulations are important to provide stakeholders with knowledge about the viability of alternative harvest strategies for their fishery. The relationships of peak escapement timing with time and environmental conditions can aid managers and fishers with in-season decisions about when to allow fishing and when the run has likely peaked. Finally, the precision of our annual escapement estimates (or lack thereof in data poor years) highlights the importance of monitoring data if coho salmon populations are to be effectively managed.


Walsworth TE and Schindler DE (In press) Coho salmon escapement and trends in migration timing to a data-poor river: estimates from a Bayesian hierarchical model. Canadian Journal of Fisheries and Aquatic Sciences. Accepted July 25, 2015. DOI: 10.1139/cjfas-2014-0554.


Newly published paper: Ecological niche models to predict effects of restoration activities along a desert river corridor

Species conservation efforts are often limited by funding for restoration activities, and time in the face of decreasing populations. Therefore, it is critical that restoration activities be prioritized such that the greatest benefit to target species and communities can be achieved. In a paper recently published in Transactions of the American Fisheries Society, co-author Phaedra Budy (Utah State University) and I describe a modeling effort to predict where different restoration activities would have the greatest benefit for imperiled native species in the San Rafael River (Utah).

Flannelmouth sucker, bluehead sucker, and roundtail chub are three imperiled species native to the Colorado River Basin. Populations of each of these species have declined dramatically in the last century in the face of habitat and flow alteration, and invasive species establishment. The states of the upper Colorado River Basin have signed agreements to conserve these three species, but in the face of numerous widespread threats, management agencies need to know where they should focus management and restoration efforts.

In this paper, we fit random forest models biotic and abiotic variables measured at sampling locations to determine the factors most limiting to each of the native species. We then used data from a longitudinal habitat survey of the lower 64 km of river to predict the effect of habitat restoration and non-native species removals at different locations along the San Rafael River. Expanding areas of high quality habitat was predicted to result in greater benefits for the native species than improving isolated patches of habitat. Additionally, the greatest benefit to the native species occurred when non-native species were removed to below about 10% of their current abundance. Non-native species present sources of predation and competition to the native species, and as such can limit the number of native species a habitat unit could support. In fact, our models predicted that habitat restoration without non-native species removals could reduce native species abundances in certain reaches, likely due to increased non-native species following habitat restoration.

Overall, this study highlighted the importance of considering both biotic and abiotic drivers of abundance and persistence of threatened species. Only considering abiotic drivers can lead to unexpected and even negative restoration results, wasting limited time, money, and opportunity. Additionally, river-scale ecological niche models can describe systems at the scale at which endemic species interact with their environment, and can allow for managers to obtain spatially-explicit information at the scale at which restoration activities will occur. Further, we recognize and discuss the importance of restoring the processes that shape river ecosystems (e.g., natural flow regime) in order to ensure the long-term success of any restoration strategy.

Timothy E. Walsworth & Phaedra Budy (2015) Integrating Nonnative Species in Niche Models to Prioritize Native Fish Restoration Activity Locations along a Desert River Corridor, Transactions of the American Fisheries Society, 144:4, 667-681

Update on third Alaskan field season

It has been a long time since my last update here.  Since then, I have completed my third field season in Alaska.  It was a good, if rather uneventful field season this year.  The previous winter’s snowfall was well below average, and the signs of this were obvious as we arrived in June.  There was far less snow on the surrounding peaks, and the rivers were all running much lower than normal for that time of year.  The Chignik, Black, and Alec Rivers were all at levels normally seen in late July at the beginning of June, and they continued to drop throughout the summer.  Part of the reason for this was the excellent weather we had for much of the season.  While Chignik in the summer is usually grey, wet and cool, we had weeks on end of sunshine and warm temperatures.  It certainly didn’t feel like Alaska, and I was wishing I had packed more warm weather clothing at times.

Field work

We continued much of the same field work from the previous two summers, including maintaining the long-term data set for the Alaska Salmon Program, collecting juvenile coho salmon diets, sockeye salmon otoliths from post spawn carcasses, and Dolly varden for gut size analysis.  Sampling for the long-term dataset took us from Black Lake to the Chignik Lagoon, sampling fishes on beaches with a seine deployed from a skiff, sampling water chemistry, lake productivity, and zooplankton biomass from stations in the middle of the lakes, measuring river discharge throughout the watershed, and sampling the lakes with tow nets (small, top-water trawls pulled between two boats) at night to examine the condition of sockeye salmon at the end of the growing season.


Sam pulls in a beach seine on Chignik Lake.

Beach seine sampling is always interesting, as the species composition of our catches changes depending on the time of year, lake, and location within the lake.  You can begin to appreciate the seasonal movements and behaviors of the different species by seeing what you catch and where throughout the season.  This year, we caught many more pond smelt than usual in our beach seines early in the year, but they were nearly absent in August.  We also noticed that many juvenile coho were preying upon pond smelt early in the year, capitalizing on the high densities of this prey item.


Measuring a coho salmon in Clark Bay on Chignik Lake.

Examining coho salmon diets throughout the summer provides insight not only into what the coho are eating, but also what the aquatic invertebrates and other fishes of the lakes are doing at different times of year.  The composition of coho diets tracks the active and emerging insects, as well as the fish populations that are spawning or emerging from the gravel.  Much of the diets we examined this year were full of chironomid pupa, though caddis adults were common for a 2-3 week period near the beginning of July.  Pond smelt were commonly seen in the diets until August.

We also collected our fourth year of adult sockeye otoliths this season.  This project is aiming to expand on our findings in our recent paper (Walsworth et al. 2014) in which we found that there is a wide diversity of juvenile lie-history behaviors represented in a single population of sockeye salmon that survive to spawn.  In a future paper we are going to examine how the success of different behaviors changes over across seasons.  Is spending more time in Black Lake better in year A, but not in year B?  What are the environmental conditions that may influence such patterns?  These collections are pretty fun, as the river corridors come alive when the salmon are spawning and dying.  Bears are in the rivers and riparian areas, gulls take over beaches, and eagles perch in branches overhanging the rivers, waiting for a meal.


Sockeye salmon carcasses collected in the Alec River.

Dolly varden sampling for a gut size analysis is always a fun sampling activity, as we sample them with fly-rods.  Fishing for science!  We are examining the timing of gut capacity changes in response to salmon egg subsidies, and how these differ across the landscape.  The Dolly fishing was excellent again this year.  I often worry that the Dolly fishing is so good in Alaska that it will ruin fishing for me anywhere else.  So far, that hasn’t happened, though I catch far fewer fish when I am not in Chignik.


Dolly varden caught in the Alec River.

Salmon runs

The past three years have seen 3 very solid fishing seasons, including two of the highest sockeye harvests in the history of the fishing district.  Fishermen were able to make some money and plenty of fish were allowed to escape into the rivers to produce the next generation and feed the many different predators that rely on the eggs, brains, and carcasses of the sockeye to grow and survive, including bears, gulls, eagles, Dolly varden, juvenile salmon, flies, and many others.  While each of these users of the salmon resource has had access to plenty of fish in recent years, this year was forecast by both the Alaska Dept. of Fish and Game and by our research group to be quite a bit more modest.

The early run of fish that spawn in Black Lake tributaries turned out to be much smaller than even the forecasts predicted, and the fishery remained closed throughout the entirety of the run.  Even when the fishery eventually opened during the second run, the daily harvest numbers remained quite low.  Speaking with several of the fishermen, it sounded like some were having trouble paying off their fuel bills with the low harvest, while others were still able to scratch out a decent season.

Due to how Alaska’s salmon fisheries are managed, the ecosystem users (bears, gulls, etc.) still received a similar amount of salmon resource.  Alaska’s salmon fisheries are managed for escapement goals.  The local manager sees how many fish have entered the river at a given point in time and determines whether they are on pace to hit their goal.  If they are not on pace to hit the goal, they have the authority to close the fishery, and if they are on pace to make their goal with surplus, they can open the fishery to harvest that surplus.  This type of management seeks to maintain future returns, even if it means reduced harvest within a single season.


Along with the low water and meager salmon runs, we also saw very few bears this year.  Usually we can count on seeing several bears every day we are out, particularly late in the year when we go to Black Lake.  This year we saw probably only 50 bears all summer, and maybe only 15 unique bears.  Colleagues at the other field stations in Alaska reported similar low bear sightings.  I wonder where they all were?  Did they have low survival over the winter with the low snow pack and heavy rains?  Were they higher in the watersheds?  Were they on the coasts?  I wonder if they will be back in more normal numbers next year.

Back to office work

Since my return from Alaska, I have been back in the office working on analyses and writing.  I am hoping to submit two papers to review soon, and am feverishly working on my next chapter in preparation for a presentation.  Until next time…

Diverse juvenile life-history behaviors

Sockeye salmon demonstrate extraordinarily diverse life-history strategies, with individuals balancing trade-offs in growth and reproductive opportunity with risks of predation and competition throughout their life-cycle.  This diversity is widely appreciated at the scale of years, with different populations spending more or less time in freshwater and saltwater before spawning.  However,  life-history diversity within single populations has been less well studied.


Our paper examining the diversity of juvenile life-history behaviors within a single population of sockeye salmon has recently been published in Ecology of Freshwater Fish.  We examined the life-history behaviors of adult sockeye salmon that had survived to spawn in the Alec River.  Otoliths (ear bones) provide a chronological record of an individual’s age, growth, and the environmental conditions to which it was exposed.  We examined the microchemical signatures of the otolith to determine how much growth was accumulated in the different rearing lakes of the Chignik River watershed.

Black Lake, in the upper watershed, is a shallow, warm and highly productive lake that supports rapid growth for juvenile sockeye salmon.  All sockeye salmon juveniles that rear in Black Lake eventually migrate downstream to Chignik Lake, a deep, colder and less productive lake that also supports different populations of sockeye salmon from the lower watershed.  Previous research found that earlier migrants to Chignik Lake were smaller, and in poorer condition, suggesting that these individuals would be less likely to survive to reproduction.  However, our analyses determined that Black Lake contributed to only about half of the juvenile freshwater growth on average for this population, and that 47% of individuals accumulated at least half of their juvenile growth in Chignik Lake or Chignik Lagoon.  Additionally, we detected an unexpected life-history behavior in which juveniles from the upper watershed move downstream quickly to the estuarine Chignik Lagoon and rear in brackish to saltwater for the summer prior to migrating back upriver to freshwater to overwinter.

Overall, the analysis revealed the broad diversity of life-history behaviors that contribute to a single spawning population of sockeye salmon.  In certain environmental conditions, one of the strategies may be more successful than the others, but it may fare less well under alternative environmental conditions.  The presence of these diverse viable life-history behaviors may buffer the population against extreme poor conditions in any single rearing environment.

Walsworth TE, Schindler DE, Griffiths JR, Zimmerman C (2014) Diverse juvenile life-history behaviors contribute to the spawning stock of an anadromous fish population. Ecology of Freshwater Fish. DOI: 10.1111/eff.12135

Sockeye carcass collections

(Apologies for the long delay between posts)

Summer is turning to autumn here in Chignik Lake.  The air is taking on a chill and the berries are starting to ripen.  This also means that the rivers are low, the salmon are spawning, and the bears are feeding on the salmon in the shallow water.

Photo by Nick Sisson

Photo by Nick Sisson

The combination of salmon dying after spawning and being preyed upon by bears results in the banks of the rivers and streams in the area to be covered in dead salmon carcasses.  These carcasses, while pungent, can actually provide us with an incredible amount of data.DSC_0065

One research project we are working on is examining the otoliths (ear bones) of adult salmon to determine their habitat use as juveniles.  This may sound like a strange way to look at juvenile habitat use.  A more common approach to studying juvenile habitat use is to study, well, juvenile salmon and observe their behaviors.  However, one limitation of this approach is that is nearly impossible to follow the ultimate fate of individual juveniles, particularly for organisms with low survival rates such as fish.  When we sample otoliths from spawning salmon carcasses, we know their ultimate fate – they survived to reach the spawning grounds.

A cross-section of an otolith looks similar to a cross-section of a tree.  Material is laid down as a fish grows, with periods of slow growth making dark rings (usually in winter, an annual marker).  The otoliths also incorporate the chemistry of the water in which the fish resides.  Thus, the otolith contains a record of the water chemistry throughout its life.  If the water chemistry of different habitats varies in a consistent manner, we can determine what habitat an individual used throughout its life.  For the Chignik system, Black Lake, Chignik Lake, and the marine environment have significantly different water chemistry and we can determine which habitat an individual was located in at any point during its life.  From this, we can determine which habitats contribute most to the growth of individual fish.

We are currently drafting a manuscript describing the first round of results, which demonstrate that, even within a single population, there are a diversity of viable juvenile behavioral strategies that contribute to the spawning population.

Fishing for science

One of the research projects we are collecting samples for this summer is examining the response of Dolly Varden gut size to the pulse of sockeye eggs that are delivered to rivers each summer.  In a recent study*, Armstrong and Bond demonstrated that Dolly Varden in the upper Alec River eat very little during the year before the arrival of spawning salmon, and then gorge themselves on eggs.  They found that not only does the amount of food in their stomachs increase after salmon are in the streams, but that the size of the fish’s gut increase relative to their body weight.  They have a physiological response to increase their capacity to take advantage of the annual pulse of sockeye eggs to their streams.  This summer we are investigating the timing of this increase in gut capacity.

To collect our samples, we drive our jet boat far up winding salmon spawning streams, pull out our fly rods, and try to land a few fish.  Image

Not a bad way to spend an afternoon working, before returning to the lab to measure our sampled fish.  The Dollies are pretty skinny at this time of year, but will be much fatter and more colorful (as they enter their spawning period) during our later samples.

* Armstrong JB and Bond MH. 2013. Phenotype plasticity in wild fish: Dolly Varden regulate assimilative capacity to capitalize on annual pulsed subsidies. Journal of Animal Ecology.

Finally, on to the fun stuff

After a week of opening up camp, soldering water pipes, hauling gear up river, putting boats in the water, soldering more water pipes, mounting motors, and soldering even more water pipes, we have finally gotten down to some sampling in the last few days!  Today, the wind relented enough to allow us to get out on Chignik Lake and haul some beach seines.


(Oh, and it was sunny again.  Five days in a row now.)

We collect these samples regularly throughout the summer and they have been collected since the late 1950’s.  This long-term data set allows us to examine trends in abundance, size and condition of the fishes inhabiting the watershed.  Our research has particularly been focused on the size and condition of juvenile sockeye salmon, an economically, ecologically, and culturally valuable species in the region.

It feels good to be getting out and sampling again!

Back for more!

Well, it’s June again, and like the salmon I have returned to my field home at Chignik Lake.  This is my second year at the Alaska Salmon Program field station, which has been here since the late 1950s.  This year we are undertaking a number of interesting projects, including continuing our examination of the within-population diversity of juvenile behavioral strategies employed by sockeye salmon, investigating how coho salmon change their diets in response to sockeye salmon fry density, and working with colleagues at UW to extend a really cool study they recently published demonstrating Dolly Varden increase their gut capacity to take advantage of the pulse of sockeye salmon eggs that are available for a short period each summer.  Be sure to check the blog for updates on what we are up to as the summer progresses.