COMPUTER VOICE: This conference will now be recorded. 

DEANA CROUSER: All right, good morning, everyone, and welcome to another EcoFOCI seminar series. I am Deana Crouser, co-lead of the seminar series with Heather Tabisola. This seminar is part of NOAA's EcoFOCI biannual seminar series focused on the ecosystems of the North Pacific Ocean, Bering Sea, and US Arctic to improve understanding of ecosystem dynamics and applications of that understanding to the management of living marine resources. 

Since 1986, the seminar has provided an opportunity for research scientists and practitioners to meet, present, and provoke conversation on subjects pertaining to fisheries, oceanography, or regional issues in Alaska's marine ecosystems. Visit the EcoFOCI web page for more information at EcoFOCI.NOAA.gov. We sincerely thank you for joining us today as we continue our virtual series. Our speaker lineup can be found via the OneNOAA seminar series and the NOAA PMEL calendar of events. 

Join at 11:00 AM Pacific time next Wednesday for a final seminar of the series. Did you miss a seminar? Catch up on PMEL's YouTube page. It takes a few weeks for these to get posted, but all seminars will be posted. 

Please check that your microphones are muted and you're not using video. During the talk, please feel free to type your questions into the chat. I'll be monitoring the questions, and we'll address those at the end of the talks. 

Today, I am pleased to introduce Dr. Lauren Rogers, a research fish biologist at NOAA's Alaska Fisheries Science Center where she leads research on early life stages of fish in support of ecosystem-based fisheries management in Alaska. Her research focuses on understanding how fish and their ecosystems are affected by changes in climate. She develops new approaches to further the integration of ecosystem information and to the fisheries management process. 

Lauren was previously a fisheries ecologist at the Natural Capital Project in Stanford University where she developed a climate risk assessment for fish and fishing communities and designed practical tools for mapping and valuing nature's contributions to people. 

Lauren was a postdoctoral research fellow at the University of Oslo, Norway. She earned a PhD in aquatic fisheries science from the University of Washington and a BS in Earth system sciences from Stanford University. 

Today, Lauren will be sharing her talk on how tracking changes in the timing of early life stages of fishes can inform climate-ready fisheries management titled Changing Spring Phenology in the Gulf of Alaska and Implications For Fish and Fisheries. And with that, let's begin. 

LAUREN ROGERS: Thank you. Thanks for the kind welcome, and thank you everyone for being here today. As Deana said, I'm going to be talking about phenology in the Gulf of Alaska, and I'd like to start out by acknowledging the many people who have contributed to this work, in particular, the EcoFOCI and FOCI program throughout the years, without which we wouldn't have the 40 years of data that we can use to really understand ecosystem dynamics in the Gulf of Alaska. 

I'd like to thank Annette Dougherty who has done all the pollock aging work, which has been really key to some of the research I'm talking about today. Ali Deary and the Ichthyoplankton team for their stewardship of the larval data that I'll be presenting. I'd also like to think the MACE program at AFSC, especially Darin Jones and Kresimir Williams, who are collaborators on some of this work, along with Martin Dorn and Cole Monnahan. And then finally also thanking Ben Laurel, all people who have contributed a lot to the thinking and ideas that I'm going to present today. 

So today, I'll be talking about phenology, and when I say phenology what I'm talking about is the seasonal timing of events, and in particular, how this timing might vary from year to year. So some classic examples are the timing of cherry blossoms, the timing of egg laying in birds or migrations, the timing of salmon, return migrations to spawn, or the timing of spring phytoplankton blooms. 

Changes in phenology are a fundamental response to climate change and have provided some of the earliest and strongest evidence of widespread impacts of global warming on Earth's ecosystems. So this example is a really classic one, showing over 800-- or sorry, over 1,000 years of data on the peak bloom day of cherry blossoms in Kyoto. 

And you can see that in the last 100 years-- let's see if I can get my mouse going here, yeah. In the last 100 years, these dates have been progressing earlier and earlier in spring. And in fact, the earliest date on record was just this last March when they recorded the peak day of bloom was in March 26. 

So changes in phenology have been observed across ecosystems and across taxonomic groups. This is just an example showing different marine taxa. And on average, events are shifting earlier, but that's not universally so. The differences here really get to what the underlying mechanisms of change are. 

So typically, we think of temperature as being the main driver of phenology shifts. Springtime is happening earlier as temperatures get warmer, and this makes sense for many species and systems. For instance, in fishes, temperature is really key to growth, to development, to behavior. But other factors are also important, such as potentially precipitation, or winds. 

And the effects of temperature may be different depending on the species. We may see different sensitivities. Some events are getting earlier, whereas others may be getting later. And what this means is that not all events are going to shift in to the same extent or even in the same direction. And that has the potential to lead to significant ecological impacts. 

One example of this is, well, we're already seeing evidence of this, and this is described here as phenological asynchrony. And this is a study that was published a few years ago, which is basically a meta analysis looking at interacting species pairs. This is a little complicated, but I'll try and walk through it. 

So it's basically showing interacting species pairs and their rates of phenological change over the past few decades. We see on the bottom that there are species pairs here where the rates-- they've been showing similar phenological change. So they've been changing synchronously through time. 

But up here we see species pairs where they're quite decoupled, where perhaps some species their are events are happening earlier, whereas they're interacting species, consumer or resource, is shifting later through time. And that's this phenological asynchrony. 

And so because interacting species might vary in their rates or in their direction of change, especially the potential for reshuffling of ecological communities and food webs as new winners and losers are sorted out. So I hear somebody who needs to mute their microphone, please. Thank you. 

So today, I'm going to be talking mostly about the timing of spawning and fish early life stages. So the timing of spawning is a key trait for marine fish species. It determines when and where offspring are going to be released into the environment. 

Spawning, in general, is thought to take place-- it's thought to be timed in order to maximize fitness by placing offspring in an optimal-- or in the environment that's most conducive to their survival. And in many cases, that means spawning in time so that first-feeding larvae have sufficient zooplankton prey available, which is usually associated with this rapid increase in production during the spring phytoplankton bloom. And this is really the essence of this match. 

Mismatch hypothesis of Cushing that says that mismatches in the timing of first-feeding larvae and their prey can ultimately determine recruitment success. However, other factors are likely to be important. For instance, spawn too early, in addition to potentially having a lack of prey, larvae may be faced with strong late winter storms. And spawn too late, larvae are going to be small relative to their competitors. They may have delayed juvenile transition time. 

And importantly, they may have less time for growth during that first summer before the onset of winter. And therefore, have fewer reserves to make it through that first winter. And really the optimal timing of spawning is going to vary from year to year. And it's going to depend on the timing of a lot of all these other processes, many of which are temperature dependent, but not all of which are. 

And so these processes and these interactions are really complex, but I think tracking changes in the timing of fish spawning and early life stages and these related ecosystem processes is really the first step towards understanding the drivers and the consequences of phenological shifts for fish and fisheries. 

So that brings me to some of the broader questions I'm going to look at today and just address parts of them, which are, how thermally sensitive are the timing of spawning and the timing of fish early life stages in spring in the Gulf of Alaska? And how does this vary across species? And second, what are some of the consequences of these changes in phenology for species, for ecosystems, and for fisheries management? 

So just a reminder, if you don't have your microphone muted, can you please mute it? Thank you. 

So today, I'm talking about the Gulf of Alaska. This is one of the many large marine ecosystems that the Alaska Fisheries Science Center has stewardship over. And the Gulf of Alaska, in terms of climate, has historically been characterized by climate regimes. So there hasn't necessarily been a very clear long-term warming trend in the Gulf, but rather interdecadal to inter-annual fluctuations, or yeah, inter-annual fluctuations. 

However, in the last 10 years or so, there have been some pretty major changes in the Gulf of Alaska. There were these two very major heat waves, one from 2014 to 2016, and then repeated again in 2019. And these marine heatwaves were record-setting, and they had large-scale ecological and economic impacts. 

So for instance, there were mass seabird die-offs, here shown are common murres, and there were recruitment failures for many species of fish. And economically, the most significant impact was the closure of the federal fishery for Pacific cod in the Gulf of Alaska due to a large scale population crash. And these ecological impacts have been persisting even after the climate has gone back to sort of a more average state. 

And if we look at what's projected in the future, the Gulf of Alaska is projected to warm 1 and 1/2 to four degrees in the next 60 years. And so that's a pretty significant changes upcoming. And understanding how extreme events, like heat waves, as well as long-term warming are going to affect phenology, I think is important for anticipating, for understanding, and for managing for the changing fisheries resources that we're going through. 

So today, I'll talk about a few different studies. First, I'm going to talk about some work I've done looking at changes in pollock spawn timing in the Gulf of Alaska. Second, I'm going to span out and look across species to look at how larval phenology trends have varied and what the drivers are across species. 

I'm going to touch very briefly on what some of the carryover effects of these changes and phenology are for later life stages of fishes. And finally, I'll talk about a study working on looking at how changes in the timing of spawning of pollock are affecting their availability to assessment surveys in the Gulf of Alaska. 

OK, so to start, Gulf of Alaska walleye pollock, this is a story I've told before, and I think many people here today have heard parts of it, so I'm not going to go into great depth, but I'll talk about it because I'm excited about this work, and I think it's relevant. 

So pollock are really a key species for the economy of the Gulf of Alaska, as well as for the ecosystem. They play a notable role as a link between lower and upper trophic levels. And they've also been the focus of decades of research by the EcoFOCI program. 

Gulf of Alaska walleye pollock, to give a little bit of background on their biology, they mature around ages 3 to 4, and they aggregate on spawning grounds in winter prior to spawning. And one of the major spawning grounds in the Gulf of Alaska is Shelikof Strait here, tucked between Kodiak Island and the Alaska Peninsula. 

They're batch spawners, which means that they spawn multiple batches over the course of a few weeks. And the eggs incubate at depth for about two weeks before they hatch. 

Eggs and larvae are subject to drift in the Alaska coastal current and are found in high concentrations here in Shelikof Strait. And then as larvae, a few weeks later, downstream over the Shelikof Sea Valley and [INAUDIBLE] And these areas have been the focus of decades of spring larval surveys conducted by the EcoFOCI program. 

So starting back in 1979, in May or June, we or our colleagues or our predecessors have gone out to sea for a few weeks. We use plankton nets, such as these bongo nets, that I show here on the left, to collect larval fishes of many species, but the focus, historically, was on walleye pollock. 

The larvae are ID'd. They're counted. They're measured for length and, in addition for pollock, they're brought back to the lab. The otoliths are extracted, and they're aged using daily otolith growth increments. 

Using all this information together, we can then back out when these larvae must have been spawned. We basically use the size, the age, the abundance, as well as a mortality correction. We can get to a distribution of hatch dates for these larvae to figure out when were they hatched into a yolk sac larvae. 

From there, knowing how long it takes eggs to develop, depending on what the temperatures are, we can back out when those larvae must have been spawned as eggs. And so doing this for each of the years that we have data for, we can then reconstruct spawn timing for the species going back to 1979. And this is how that looks. 

So this is reconstructed pollock spawn timing using the larval data. And so there's a few things to note here. First of all, again, there isn't really a clear trend towards earlier or later through time. But there is a lot of variability, and that variability appears to be increasing. 

Some of the latest years on record were here in the late 2000s, 2007 to 2009. And the most recent years have act-- one has actually been quite early. And I don't have 2019 on here, but we do have those data. And it is also quite early, similar to 2017. So the mean spawn timing has varied by up to four weeks based on these estimates. 

And so there's a few caveats here. We know we're not sampling the full temporal distribution. We're missing the earliest spawned larvae, and we're missing the latest spawned individuals, just based on our surveys. And we're not able to account for variable mortality that we also know is likely happening. 

But we are pretty confident that we are capturing the inter-annual variability, at least, and that's based on some validation that we have done with independent data sets to look at this. And if you want to read more about that, you can look at the paper that's cited here. 

So that raises the question, what is driving these changes in the spawn timing? And so there's a few factors that we were able to identify that were highly significant. The first is the age of the spawning stock. And so on the left, I'm showing here how when the mean age of the spawning stock is older, the spawn timing was earlier. 

And so you can see that spawning is up to, let's say, two weeks earlier when the spawners are older relative to younger. So that's an effective spawner age. And then we also did find ineffective thermal conditions. So in general, spawning happened earlier when temperatures were warmer, and happened later when temperatures were cooler, which matches what we know for [INAUDIBLE] biology in general. 

However, we did find that this response was nonlinear, and there appeared to be a threshold. And once temperatures got above about 3 and 1/2 to 4 degrees Celsius, there didn't appear to be much of an additional effect on warming temperatures on timing, suggesting there's perhaps a threshold that spawning can perhaps only get so earlier. There's some other processes that are going on, or constraints. 

So what this means is that the phenological response of pollock spawn timing is not necessarily-- it's not so clear cut as just a simple temperature response. There's this other factor, spawner age, that's really important. And there's also this non-linear response, which raises this important question. Are these changes in timing adaptive for the species? Do they confer an advantage? 

And to start to look at that, we can reconstruct the estimated date of first feeding for these fish, or for the offspring here. And so this is the time when prey resources are going to be most critical because larvae are using up their yolk sac reserves. They're becoming dependent on external prey. 

And what we see is a few things. First of most, there's really wide variation within a year as to when these first feeding larvae are arriving in the environment. And that suggests that there is a degree of bet-hedging right, like some-- regardless of when conditions are optimal, there's some spreading the risk over time here. 

But we do see that the date of first-feeding is mostly centered in mid-May, or sorry, in early May. But in some years, especially these years 2007 to 2009, which I pointed out before, these quite cold years, the mean date of first feeding was quite late, as late as late May to early June. And then here in 2017, we're looking at late April. So that's quite a lot of variation in the last decade and a half. 

So then, the question is, how does this variation in timing of first feeding relate to variation in the timing of available prey resources? And that's a really good question and one that I wish I had an answer for, but I don't right now. So I wish I had a lot to say about changes in the phenology of phytoplankton blooms and zooplankton in the Gulf of Alaska. 

Some work has been done on this. I know there's some ongoing work by various different groups on different parts of this. But there are a lot of challenges depending on whether you're using shipboard data, whether you're using satellite data, whether you're using models. And I think this really remains a key research need for climate fishery science in the Gulf of Alaska. 

So unfortunately, I can't give you a clear answer on this part, and I hope to be able to in the future. And I think that this, well, will require a dedicated effort in collaboration with partners, especially PMEL, UAF, people who are already working on these questions, to be able to resolve and synthesize these patterns and drivers of change to be able to then relate them to the larval first feeding date. 

However, I do want to highlight a study that Jens Nielsen did, working with our program a couple of years ago, and this flipped things around a little bit, and instead of thinking about pollock and their prey, thinking about pollock as prey for other species. 

And what Jens was looking at was pollock eggs and whether they are an important prey resource during this time in late winter, which is a prey depauperate time in general. The phytoplankton bloom hasn't started yet. There's not a lot out there to eat if you're a planktivorous consumer. It might be a time when resources are limited. 

And he found that the average peak production date of fish eggs, from pollock eggs from spawning, occurred about 10 to 20 days before the average increase in zooplankton production. And that's shown here on the right. So eggs are this bottom line. And then model-based estimates of when zooplankton production was peaking are provided here in the upper two lines. 

And in most cases, you see that this pollock eggs are actually available at a time that is quite a bit before the zooplankton are available, which means that they might be a really important resource. They're a lipid-rich, nutritious food for planktivorous consumers at a time when otherwise there might not be much available. But in some years when spawning is quite late, these eggs are actually available overlapping in timing with the zooplankton. 

So this I think raises an interesting question, which is could it be that predation might be actually higher on these eggs? Consumption of the eggs might be higher when they're spawned early relative to the zooplankton bloom. So it's a way of flipping the question around and thinking about ecosystem effects in a different way. 

So to summarize this first part of the work I've talked about, found that pollock spawn timing is varied by up to about four weeks over the past four decades. And the timing depends on temperature and also on the spawner age, which means that changes in timing are likely to be asynchronous with other species because spawner age is something that's going to be really particular to pollock in this case. 

And although we haven't identified all of them here, there are probable impacts on the population through match-mismatch with their prey, as well as through predation timing. 

So from there, I'm going to move on and talk about looking more broadly at larval phenology trends across species, asking how sensitive is early life stage phenology to thermal conditions? 

So I talked about these spring surveys, but walleye pollock aren't the only species we sample in the Gulf of Alaska. We actually have a fair amount of information on a lot of other species as well. 

And so here I'm showing a figure from Miriam Doyle's work summarizing the larval stage abundance by month in the Gulf of Alaska. And the red and purple colors indicate higher abundance. And you can see going through time here for a number of different species. There's pollock, this dark purple. But there's a lot of species that are available to our surveys in this springtime period. 

And so the question is, what can we infer about changes in phenology from these nearly 40 years of spring surveys that we have? We don't necessarily have otoliths or age data from these species, so we can't reconstruct their spawn timing. However, we do have larval size or larval length data. 

And I'm just going to ask one more time. I hear somebody with their microphone not muted. If you could please mute, that would be helpful. Thank you. 

So what can length distributions tell us about phenology? So here I'm showing length distributions for Pacific Sand Lance caught in our surveys in the springtime from some years here, 1979, for instance, through to 2000. And if you look at these length distributions over the years, we can see in some years, such as 1999, see there's out here, they were quite small. And in other years they were larger or more spread out, like 1998 here, the previous year. 

And so the trick here is that some of this variation in size is due to variation in our sampling dates. So later surveys, we're going to catch larger and older larvae. And even within a survey year, that difference from week one to week three of the survey, we're going to catch-- there's going to be a fair amount of growth that we need to be able to account for. 

And so that was the first step in this work, was to correct for variation in sampling date in order to estimate the mean size on a given date. Sort of took an empirical approach to this, used a modeling approach to basically account for the day of year, and then use that model to predict mean length on May 25 for each year. 

And I guess I won't really go through the details of this, but what we come out with is, here on the right, here's for Pacific Sand Lance, for instance, estimates of mean larval length on May 25. And I think those two years I pointed out earlier were 1998 and 1999. '99 was relatively small, and '98 was relatively large, even after accounting for variation and sampling dates. 

And so here there's quite a bit of variation in mean larval length, and the question is, what's driving that variation? Is it related to temperature? So there's going to be a number of things that are going to affect larval length on a given date. Spawn timing is going to be important, effects how old the individual is. Growth is also going to be important, and there could be an aspect of selective mortality in there as well. 

So we can't, without having the age data, we can't really separate out these different things. However, temperature is going to be directly affecting, especially, spawn timing, as well as growth or at least it's likely going to be. So the question then is, are these changes in mean larval length related to temperature? 

And the simple way to do that would be to take those mean estimates I showed in the last plot and relate back to temperature. And in case, we can see that there is actually a clear positive relationship where, in years where temperatures are warmer, larval length was longer. So they have advanced in their larval development on a given date in years when it's warmer. 

To do this in really a more robust statistical framework, to actually test this, we do a slightly different approach, but we get to the same point. And so this is showing that there is a significant positive effect of April sea surface temperatures, in this case, on size obtained by Pacific Sand Lance larvae by the end of May. 

And so how does that look like if we look across species? So here I'm showing the same type of plot but, in this case, for 20 different species that we sample in the Gulf of Alaska. And these are taxa that we do consistently catch in our spring surveys, but note that some of them are still really not very common. And our surveys aren't necessarily well-suited for capturing many of these species, depending on where they are, depending on their timing, when we expect to see them, when they're spawned, their larval development, and also depending on year. 

However, despite that, we did detect a significant positive effect of temperature on larval size in 15 of the 20 species shown here. And notably, also we did not detect any significant negative effects, which shows here that this advancement in larval development in warmer years is common, although it's not necessarily universal. 

And unfortunately, for some of these species, for instance, Northern Rock Soles and Butter Soles, we have just relatively few data to know if this is just because we don't have enough data for this species, or if, in fact, there's something else going on. But it does seem to be a relatively common pattern to see this advancement in phenology. 

So why does larval size in late May matter? Is this really a relative metric? It's not exactly looking at phenology, right? It's looking at size obtained on a given date. And it likely reflects on a number of important processes. 

For instance, if larvae obtained a larger size on a given date, it's likely going to affect their timing of juvenile transition. It's going to affect the timing of when they're able to switch prey from eating very small prey to eating larger and more diverse prey. They're going to have larger jaws, be more competent swimmers. 

It may also affect the timing of when they're able to evade predation, as well as there could potentially be carryover effects of these early changes in phenology or of size on date onto site at settlement, for instance, size at end of summer and even into the following years. 

So I'll touch really briefly on this last part here, which is looking at carryover effects. And the species that we have the most data to do this for is looking at Pacific cod, and this is work with Ben Laurel. And Ben Laurel and his team have been conducting a beach scene surveys on Kodiak Island, targeting juvenile age-0 Pacific cod dating back to the mid 2000s. 

And have noticed that there's quite a lot of variation year to year in the size at settlement in mid-July of age-0 Pacific cod. And we were wondering and have the question, could that variation inside a settlement be related to the variation that we were seeing in larval size obtained by late May? 

And in fact, we did see there was a pretty clear relationship here, shown on this plot, where larval size in late May is a reasonably good predictor of size at settlement, suggesting that these differences in size carry over at least until that May-- or sorry, July settlement period. 

The questions then, and these are the subjects of ongoing work, which is, are there consequences of this for over-winter survival? And do these size effects continue to carry over into older ages, for instance, size at age-1 or size at age-2? And preliminary work, which I'm not going to show, does suggest that this is the case. And this could have some real implications for thinking about the stock assessment and having early indications of changes in size at age, which can be pretty significant. 

So to summarize that bit of the talk here, they've known that most species showed significant advancement in their phenology, or larger size obtained by late spring, with warmer temperatures. And these effects may carry over to later life stages. At this point, we can't distinguish between the relative contributions of temperature dependent growth or spawn timing to these changes in size. But regardless, these are likely to have implications for recruitment success through a lot of the mechanisms that I suggested earlier. 

That brings me to the last bit here, which I'm excited to talk about. And this is a project looking at how changes in pollock spawn timing, that I talked about earlier, may be affecting their availability to assessment surveys in the Gulf of Alaska. 

So to get into this, it helps to know a little bit about how stock assessment works, and I know this is a very broad audience that attends these seminars, so bear with me here. Regardless of which side you're coming from, the side part doesn't-- you don't bother yourselves with stock assessment or this type where you're very deeply involved with this, you'll probably both find issue with what I'm saying here. 

But basically, the way stock assessment models work here is that you have multiple data that comes in from multiple surveys. You have estimates of biomass, in this case, from the Shelikof acoustic trawl survey, from a bottom trawl survey, from the ADFG trawl survey. 

These different estimates of biomass get combined with data on age, on weight, on length, as well as information on fisheries catches. We go into this age structured stock assessment model, which tries to make sense of all this information. And based on that estimate the size of the stock, more or less, and importantly, estimate where the size of the stock is relative to these management thresholds. And based on that, recommendation is made for how much harvest should be allowed. So there you go. There's my 30-second stock assessment overview. 

In the last few years, specifically from 2017 to 2019, there's a large divergence in these survey trends. And so this winter Shelikof acoustic trawl survey, which I'll be talking about today, had record high biomass, and biomass estimates that were higher than had been seen in the previous few decades. Whereas, the trawl surveys had near record low biomass, shown down here. 

And this really large discrepancy in these biomass estimates and differences in trends really increased the uncertainty in the stock assessment. The model didn't really know what to make of these differences, and just cut a line through the middle. And there's been a lot of talk about what are potential mechanisms that could be driving these diverging trends in biomass estimates. 

It turns out for the winter Shelikof survey, that these unusual biomass estimates, shown again here on the left, and here's the best that the model could do to fit those data. Here is the line. These weren't the only unusual survey results. So when the survey team was presenting on their-- after they got done with a survey, they also presented these unusual maturation results. 

And so they found that a relatively high proportion of fish that were caught during the survey were already in spawning or spent stage. And so here we have over 90% of males and over 30% of females who are already spawning are spent at the time of the survey. And this was very unusual and was flagged by the base team as being unusual and not what they typically look for in their assessments, or sorry, in their surveys. 

So this caused me to wonder whether changes in spawn timing could be affecting the relative availability of pollock to this winter acoustic survey. So how does that look like? This winter acoustic survey, it's a pre-spawner survey. So it's designed to occur prior to the peak in spawning. 

It's concentrated here in Shelikof Strait in Sea Valley, the known high-density spawning location that I talked about earlier. And the general thinking is that pollock aggregate on the spawning grounds and then rapidly disperse after spawning. And so you can imagine then, depending on when spawning is happening relative to the survey, there may be a slightly different, or significantly different, proportion of the spawning stock that's actually available to be detected by the survey gear. 

And so to look at this we first thought, well, let's see how survey timing has varied relative to the spawn timing that we've estimated from the larval data. And so that's what I'm showing here. In the red are the estimated spawning dates reconstructed from the larval survey data that I showed earlier. 

And in orange are the dates of the Shelikof acoustic survey. And there have been additional surveys in this, especially back in the 80s and early 90s. They often had multiple surveys. This is just showing the timing of the surveys where the data were used for the stock assessment. And therefore relevant for the work that we're doing here. 

So you can see that there's been, in some years, there's quite some overlap between when we think spawning is happening and when the survey is happening. And in other years, again, pointing out these unusual 2007 to 2009 years, there was quite long time lag in between the timing of the survey and when we think peak spawning was happening. 

If we turn that into time series. You can see it varies from somewhere around 15 days of a, I'll call it a mismatch, but it's just a measurement of the difference between the median spawn date and the middle of the survey date. So that's varied from about 15 days up to 45 days. And so that's quite a bit of difference in variation in the past few years. 

So the question then is, is that related to changing availability for the survey? So how do we measure availability of the survey? We have to do proxies for this. But one approach is to look at residuals from the assessment model fit to the survey data as a potential indicator. So this is like, how is the survey data consistent or inconsistent with what the assessment model thinks should be happening based on all the other data that is informing it. 

And so here we can see, again, these recent 2017, 2018, and 2019, where the assessment model just couldn't fit these arc estimates, and here we have this a big mismatch, or a large residual between the model estimate and the survey estimate. And so we use these residuals, these differences between the survey and the model, as a potential indicator of changes in pollock availability to this survey. 

So looking at these things together, we can see that it does appear that changing spawn timing relative to survey timing affects the availability of pollock to this survey. And so on the bottom here, I'm showing that difference between the spawning timing and the serving timing that we reconstructed here, as I said, ranging from about 15 days to 45 days. 

And it appears that that is related to these survey residuals, so this difference between what the survey was measuring and what the model's best guess of what the biomass was. In particular, the survey estimates tended to be high relative to the model or have these positive residuals in years when the survey was closer in timing, or later, relative to peak spawning. 

And this suggests here that estimates of spawn timing, here, relative to survey timing, can explain some of this misfit of the assessment model to the survey data. 

However, the problem here is that we don't actually have larval surveys every year to be able to come up with a metric that we can use to explain those differences. In recent years, especially since 2010, we have only had these larval surveys in odd years, which means that we're missing information on the even years. So we need to develop an alternative metric that we can use in these cases. 

And we have a couple of options here. We could use a statistical modeling approach to predict spawn timing using temperature and age of the spawning stock based on the model that I presented in the first part of this talk. And there's some value to doing that, for sure, and I think it might have reasonable performance. 

But a different approach would be to use information from this Shelikof survey directly. And that's what we looked into. So I mentioned earlier that the MACE program has always used maturation data as an indicator of relative timing for the survey, and that was what came out is really unusual in these recent years. And we can see that here. 

So if we look at the proportion of fish that are already spawning or spent during the survey, shown in blues, here we can see in recent years, there's been this real peak. There have been a lot more fish spawning or spent than in the 2000s here and the early 1990s. 

And so we can then-- what if we try and use that as an actual metric of relative timing of the survey relative to spawning? If we do that, we see that it's actually, really, I thought, surprisingly closely related to the timing-based metric that I presented earlier, which is looking at the days of mismatch. Here we're looking at the proportion of females that were in spawning or spent stages during the survey. 

So it's showing that in years where the survey is more closely timed to spawning, which is this-- let's see, my labels are a little bit off here. I apologize there. I think what I've done here is I've switched these, is exactly what I've done. So sorry about that. 

But the relationship would look similar. But in days-- in years when the survey is more closely timed with spawning, there's a higher portion of fish that were already spawning during the survey. So that's intuitively makes sense, but I wasn't expecting the data, actually, these two independent sources of data, to actually show this so well. 

So that means that we can then use this maturation-based estimate of relative timing to explain this misfit of the assessment model to the survey data. And we see that, in fact, historically, the biomass estimate from the survey's been high relative to the model when more females are spawning or spent during the survey. And it's a reasonably clear relationship here, I think. 

So where does that bring us to? What this, I think, shows is that we've taken this problem and provided a mechanistic explanation for why the survey trends diverged. And we provided multiple different lines of evidence that it's this variation in spawn timing relative to the survey that are driving differences in availability of pollock to the survey, and therefore, differences in the biomass estimates. 

The next steps for this work are to actually incorporate these time series as catchability covariates directly on the stock assessment model in order to account for these changes in availability. And that's work that's ongoing right now with Martin Dorn and with Cole Monnahan. And I'm excited to see where this goes in the next few months. 

And what we want to look at there is to see, does incorporating this information, does it help to reduce uncertainty? The model, it certainly helps provide us with an explanation for what's going on. And does that help us to-- or does it affect in any way the biomass trends or the estimates, because that would be important? 

Finally, one additional thing we want to look at is exploring whether there's potential differences in the age-specific availability of pollock to the survey? And if you think back to those original results I showed, where the spawn timing seems to vary with age, we might expect that there's also differences in the availability of older versus younger fish to the survey depending on when exactly they're spawning. 

And there's a lot of unknowns with regard to the biology there. For instance, it could be that fish arr-- the young fish spawn later, but they arrive on the spawning grounds just at the same time. They just hang out there longer, in which case, it wouldn't affect the availability of the survey, what we're trying to develop approaches to look at this, because I think that would also be an important finding. 

So what I've done here is give an example of how changing phenology, in this case, the timing of spawning, is affecting our ability to accurately survey and assess fish stocks, and then have provided a way forward for dealing with this that's based on having this mechanistic understanding of the underlying process with the spawning behavior, in this case. 

And I think as climate continues to change, we're going to see increasing numbers of cases where there is changes in depth distribution of fishes. We know we're already seeing changes in the geographic distribution-- we're seeing this really clearly in the Bering Sea-- that's affecting the availability of fish stocks to our surveys. 

And as climate changes, I think it's going to be increasingly important that we're tracking and understanding changes in phenology, changes and spawning dynamics, changes in migratory behavior, these distribution shifts, and staying on top of this as an agency to make sure that our surveys remain effective, and that they're measuring what we think they're measuring. And I'd say we already have a good track record of doing this. 

There's a couple of really nice studies that have come out in the last couple of years. This first one by Dan Nichol looking at how Yellowfin Sole availability to the Eastern Bering Sea trawl survey has varied. And that's another interesting story that shows how important temperature and spawning dynamics are. And that if we can understand those processes and account for them in the stock assessment, we can explain some of the otherwise unexplainable variation in biomass estimates. 

And then another example here from Cecilia O'Leary that just came out recently that shows how we can use model-based approaches to account for changes in distribution shifts of walleye pollock in the Bering Sea and how that's affecting survey estimates. So there's some nice examples of this work on-going, but I think this is going to be increasingly important as climate changes, continues to change. 

So that brings me to the end of my talk here. To summarize, I've described a few studies that show how the timing of spawning larval development are sensitive to temperature. And I think what this means is that phenological change can be expected with future climate change. And as of now, the consequences of these changes are uncertain. 

However, given that for fish, the majority of lifetime mortality happens in those earliest life stages, I think we can certainly expect that there are going to be consequences for recruitment and for population dynamics of these stocks. 

And then finally, I think monitoring and understanding changes in phenology, I think, is going to be increasingly important as we prepare for and adapt to the effects of climate change on our fisheries and on our ecosystems. And with that, thank you, and I'd be happy to take any questions. 

HEATHER TABISOLA: Thank you, Lauren. Oh, I'm back on. I always love your presentations. You just make everything so clear. It's fun to listen to, so thank you for that. It is 10:51, so we have about nine minutes to ask questions of Lauren. So please type them in the chat. We will read them out, or if you want to just ask her directly, please, just type your name in the chat, so at least we have a queue to go through. 

We've all been home for a long time, and sometimes it's nice just to ask people directly versus through a chat screen. So Libby and Mike and Abigail McCarthy all say, thanks, Lauren. Great talk. Anybody have questions for Lauren? Jim, Jim Ianelli. Hi, Jim. Oh, he's chatting something else. 

LAUREN ROGERS: I think you have a side conversation going on there. 

HEATHER TABISOLA: Totally!

JIM IANELLI: I hit enter. Sorry, can you hear me? 

HEATHER TABISOLA: Uh-huh. 

JIM IANELLI: Yeah, sorry. Great talk. Just curious on the mean age and the timing of spawning. Could maybe ask it more clearly. If the population is mainly made up of young spawners and the survey knew that in advance, should they change their schedule? 

LAUREN ROGERS: That's a good question, and I've thought about this with respect to could we forecast when spawn timing is going to happen? And I've done a little bit of work looking at that, and yeah, we can do an OK job of forecasting that. That being said, I think the reality of trying to move a survey is really hard. And there's always uncertainty up until the moment you actually get on the ship as to when you're going to be able to leave. 

So I think logistically that's just not such a practical thing to do, and that's why I think it's good to be doing the kind of work we're doing here, which is to basically come up with ways of correcting for those potential changes or potential differences post hoc by using these catchability covariates. 

HEATHER TABISOLA: Roger is on and says, thanks for the great presentation, Lauren. So many interesting questions and needs. Keep up the great work. Roger, it's nice to see you, have you join us. Libby has a comment. Libby, you can go ahead. 

LIBBY: OK, thank you. Great talk, Lauren. Very exciting stuff. And I just wanted to highlight a couple of things that really jumped out at me from the program perspective. So I really appreciate that you highlighted the need for more work on phenology of zooplankton and phytoplankton in the Gulf of Alaska. That's just the kind of thinking we need to do as we strategize for the future. 

And then I'm super excited about this collaboration with MACE and with the assessment team on the effects of spawn timing on survey catchability of pollock in the Gulf of Alaska. I think that looks like great potential to incorporate some of FOCI's data directly into the stock assessment, in terms of this catchability covariate. 

And I really loved the quote part where you said that we need to "track and understand." I think that's really key to what FOCI does and what we need to do for the agency. So good job. Thank you. 

LAUREN ROGERS: Thanks, Libby. Yeah, it's been really fun. The collaboration with MACE and with stock assessment, has been-- it's fun. I came at this from thinking about the biology and the ecology, and then all of a sudden realized, oh, wait. This actually could also really matter for our ability to assess these stocks. And putting the pieces back together has been a fun process. 

PARTICIPANT: I have a question. Hey, Lauren, going back to the beginning of your talks, the match-mismatch is one of the Holy Grails of fisheries for many years. Do you think, based on what the research you've been doing, that we can come up with a potential for match-mismatch, or even try to view this in our samples? Or do you think we're still pretty far away from being able to talk about match-mismatch and your class success? 

LAUREN ROGERS: It's a good question. I think-- and you and I have obviously talked about this quite a bit with regard to the zooplankton side of things. And it's really hard when you just have a single snapshot of the zooplankton community in time each spring to reconstruct what that means in terms of overall timing and how that might be changing. 

So we can get creative with that. We can use some modeling approaches to look at that. I think there's always going to be a fair amount of uncertainty just based on the limitations of the way that we can sample. I think there may be some larger trends that might come out. For instance, I know some of Jens' work has found that the timing of primary production, for instance, seems to have responded in quite an opposite way in some of these really extreme recent heat wave years than the timing of spawning has. 

And those extremes might provide us with interesting case studies that we can look at, even if we don't have enough resolution in our data to be able to pick it up on an every year basis. 

PARTICIPANT: Thanks, and I have just one more question. If we sample at the wrong time, or continue to miss these key points, how damaging is it, do you think, in terms of management to not be able to track these changes or to have a sampling time that really doesn't match up with what we want to see? 

It's been existentially-- as a crisis for myself, I'm really concerned that I'm extrapolating data from very limited snapshots and making pronouncements that may or may not be correct based on the fact that we have limited sampling. Can you comment on that? 

LAUREN ROGERS: Yeah, I think it's really a challenge, especially if you don't have-- Let's say we had a few years where we have really extensive sampling through the spring, and we would know what an average year would look like. You could then lay-- in other years, if you just had a snapshot, you could sort of lay that on this long-term average, and that would tell you something. Or if you had a lot of knowledge about exactly what the processes and rates are, you can reconstruct some of that even with just a simple snapshot. 

But we have that for some processes, for some species, but we don't have it for all of them, which does mean that, in many cases, we do have this problem of knowing, is abundance really high this year, or did we just sample at a slightly different time so we're catching things at a different part of the cycle? 

And I think that's something we need to be thinking about with all of our surveys, both our early life history surveys, our lower trophic level surveys, but then obviously also with our broader fish assessment surveys that we do at the center. And I think that time aspect of things, it comes in everywhere. It comes in if we're looking at body condition. 

How does body condition change from year to year? Well, you better make sure that you're also correcting for differences, for instance, in the time of year that you're sampling, because there can be a really big difference if you're looking in June versus if you're looking in August. The seasonal timing, it comes up everywhere. And hopefully, we're remembering and have the tools to be able to account for those changes. 

PARTICIPANT: Great. Thanks, Lauren. Excellent talk, and stimulating, as always. Thank you. 

HEATHER TABISOLA: So there are a few more comments or notes in the chat, and I'll read them quick. Sorry, excuse me. But it is also 11 o'clock, so if people want to jump off, please, feel free. This is your free time to go. So let's see, so Carol was on here, also said, enjoyed the talk. Colleen, really exciting stuff. 

Cole Monnahan, said note in the assessment we estimate a time varying catchability that partially reflects the proportion of the stock available to the survey, so less than 100%. 

LAUREN ROGERS: Can I comment on that real quick? 

HEATHER TABISOLA: Totally, yeah. 

LAUREN ROGERS: It does, and I think it would be-- I think what's interesting is that, that's just empirical, right? And it's this necessarily smooth process. And I think what we're seeing, if part of the mechanism behind this change in catchability is the spawn timing, that's not something that necessarily changes smoothly from year to year. 

And so I think we're missing our ability to catch some of that variation. But seeing how this plays out actually on the stock assessment, I'll be really interested to see, can we keep that smooth time-varying catchability term in there and include this covariates. And that's where I'm really glad, Cole, that you're working on this, because I don't know how to do that. 

HEATHER TABISOLA: Carol did ask, based on that, what would be the ideal design for sampling? 

LAUREN ROGERS: Ideally, you'd just have unlimited resources and be able to go out there and not just get one snapshot in the spring. It would be great-- this is where moorings can come in really handy, right, because they are out there all the time. 

No, you don't get the spatially-- you don't get the spatial coverage, but if you can track, even in one place, what the seasonal cycle is looking like, and that works for a lot of the things we're interested in. But we need some new technology to be able to get to all of it. 

But I think that is certainly an important tool given that we can't do monthly surveys with any resourcing that we're anticipating to have in a long time from now. [LAUGHS]

Thanks. 

HEATHER TABISOLA: Thanks, Carol. OK, Lauren, do you have a couple more minutes? It's 11:01. Or do you have a meeting after this? 

LAUREN ROGERS: I've got a few minutes. If people want to stay and ask questions, that's fine. If people want to leave, bye. Thank you everyone for coming. 

HEATHER TABISOLA: Jim, Patrick, and Cole are all very invested here. [LAUGHS] So Jim just says, welcome to our world, using snapshots to pretend we know something, which made me laugh while Dave was talking. 

Patrick, looks like he has a question, and give me one second, Patrick. Cole also said, agree, but just to note, we do not assume 100% of stock is surveyed, so missing a consistent timing due to portion is handled, in a sense. 

Patrick, go ahead and unmute if you want to ask Lauren. Maybe he's still-- let me double check that he's still here, and maybe I can unmute him for him. [INAUDIBLE] 

LAUREN ROGERS: Hmm. Patrick, if you want to type it in or if you want to give me a call later, we can chat. I'd be glad to. 

HEATHER TABISOLA: Patrick, you're unmuted, so not sure what's happening there. 

LAUREN ROGERS: Or I can see you at pick up at kids time later. [LAUGHS]

HEATHER TABISOLA: Matt said, excellent talk. If I understand the nonlinearity of response to temperature presents a challenge to any correction. Comment? Thanks. 

LAUREN ROGERS: I think if we know what that nonlinearity looks like, and if we have a good estimate of it, then it's no different from if it were a linear response. But it does present a challenge, in that, if that's going to be different for each species, you'd have a lot of data to be able to estimate. 

HEATHER TABISOLA: All right, I think with that, Patrick, I'll let you reach out to Lauren directly. And yeah, thanks, everybody, for joining us. We have one more seminar in this fall series, and that will be Phyllis Stabeno will be presenting next week, same time, same link. We'll all be here, so thank you, everybody, for joining us. And thank you so much, Lauren, for taking your time today. 

LAUREN ROGERS: Thanks.