Ducks Unlimited Podcast

VO: 00:34

Welcome to the Ducks Unlimited Podcast, RELOADED, where we bring you the best of our past episodes. Whether you're a seasoned waterfowler or curious about conservation, this series is for you. Over the years, we've had incredible guests and discussions about everything from wetland conservation to the latest waterfowl research and hunting strategies. In RELOADED, we're revisiting those conversations to keep the passion alive and the mission strong. So sit back, relax, and enjoy this reload.

Mike Brasher: 01:05

Today, we are again going at it with doctor Jim Nichols discussing in detail adaptive harvest management, what all goes into it, the different components, and how it helps us make better decisions about what our harvest regulations should be. So, Jim, thank you for sharing your time with us and joining us again here on the Ducks Unlimited Podcast.

Jim Nichols: 01:23

Thanks. I really appreciate the opportunity, Mike.

Mike Brasher: 01:25

Okay. Well, Jim, we are trudging on here through these different components of the process, but I think it's important to get all of these components together in one discussion. So, now, with our series of regulatory packages in place, regulatory alternatives, let's move to a discussion of the alternative models of how the system works because just as a reminder, our task here in this harvest management decision is to select the optimal regulatory strategy or alternative, you know, relative to the system, but then that requires us to understand how that system works. And so that's where our models come into play. So talk about the models, the the alternative models of how the the waterfowl population system works and how it responds to to harvest.

Mike Brasher: 02:14

We introduced this previously a bit with the additive versus compensatory mortality than density dependent reproduction, but let's talk about that in a bit more in a bit more detail here as it relates to the where it fits within If

Jim Nichols: 02:29

you don't mind, I'd like to make a real general comment first because every every once in a while you run into or encounter managers and other folks who basically have a sort of a take a dim view of models. Alright? Basically, I don't need no stinking model, I, you know, I know how to make decisions for my area or my system and such. And my claim or my reminder is that when I say the word model, all I mean is something that projects the consequences of the different actions that you can take. In other words, in order to decide on the smartest set of regulatory package for a year, what do we have to do?

Jim Nichols: 03:10

We have to make projections about what each package might do and then ask which one of those comes closest to what we're trying to achieve in our objectives. And so models can live in people's heads, they can live on a computer or anything, but all they are is ways of projecting consequences of actions. And my claim is that if you don't have a model, that's just the same as giving a fair coin to a monkey and let it flip it and pick your management action that way. In other words, you have to have some basis for prediction no matter how much uncertainty is associated with it before you can make a smart or informed decision. Okay.

Jim Nichols: 03:50

So in particular for the adaptive harvest management, there are four basic models that we end up considering and they incorporate different stories about the two important processes that underlie all population growth. And what's that? It's just births and deaths. And so with respect to deaths, we've already talked about this sort of additive mortality hypothesis that basically says for every bird that dies during the hunting season, there's one fewer bird around in the spring to make babies and to come down the next fall and that's just the way it is. And then we also had this compensatory mortality hypothesis that says it's not that simple.

Jim Nichols: 04:32

And actually we end up with more birds than we would have expected under this additive mortality hypothesis. In other words, it's there are a couple of different mechanisms that could make it such it is not nearly as extreme as the additive hypothesis. And so there are two different stories about how harvest rates, different harvest rates translate into different survival rates and survival is a big piece of population growth. So we have those two models of survival. Then we also had two different models of reproductive rate And they have to do with this notion of so called density dependence.

Jim Nichols: 05:14

And the duck folks have been talking about this ever since, well, for example, folks like Walt Krissy and Alex Zubin on the breeding grounds were writing papers and talking about this long, long before folks were talking about adaptive management. But it has to do with the basic idea that if you've got lots and lots and lots of ducks on the breeding grounds such that, you know, maybe maybe such that there are not enough potholes basically for single pairs, for every pair that there is, then what can happen is you can end up having a reduced reproductive rate. In other words, you could have fewer young produced per adult female, maybe some hens don't even breed in a particular year. Whereas in years where density is low, that is you have a lot fewer birds around, then maybe reproductive rate is higher, everybody can find a pond, there's no problem at all with every hen breeding. And so these two different ideas about the degree to which reproductive rate is density dependent.

Jim Nichols: 06:21

One of them being a fairly extreme high degree of density dependence and the other a much lower degree. And so we had these two different ways in which harvest can translate into survival, two different ways in which the number of ducks in the spring translates into the baby ducks in the fall, and just combine those two survival rate models with two reproductive rate models and it gives us four different models of population growth. So you had the weekly and strongly density dependent reproductive rate with the compensatory and additive mortality rate. And those are the four different models that of our population dynamics for the midcontinent knowledge, which is what we're doing this for. And they were the four models that were actually used.

Jim Nichols: 07:11

And one of the things we always want to be sure of is that those models sort of capture the endpoints, they probably go as extreme as we could possibly ever go. You can't imagine models that come there sort of more extreme than those and these endpoints captured that with the idea that maybe we probably know none of those endpoint models is precisely true all by itself. The truth is probably somewhere in the middle, but for right now, we're just trying to find which one of those models is closest to where the reality is.

Mike Brasher: 07:45

So Jim, do those models project population size in the fall or in the spring?

Jim Nichols: 07:52

Yeah. We end up having to do both. And so the it's interesting because the banding dataset, when we get our survival rates, the actual estimates go from fall to fall. And but when do we get our estimate of how many birds there are? Well, that's in the spring, that's in May.

Jim Nichols: 08:09

And so what we have to do is take into account that sort of a mismatch, if you will, of our survival rates with our population size estimates and we do that in a reasonable way. I mean, I talk about it in detail, but it's basically we do the bookkeeping in a in a correct way. And what we do is we're trying to all the time look at what we have in the spring, and at least until about seven years ago, and then say, okay, what should hunting regulations be this next fall based on the number of birds that we have this spring? So our projections are all the way around to the next spring.

Mike Brasher: 08:49

And so to put a finer point on that, it's what should our hunting regulations be in the fall based on what what we see with regard to population status and habitat conditions right now under the four alternative models, right?

Jim Nichols: 09:04

No, that's exactly right. So each when you're considering what to do in a particular year, each one of your models is gonna make a different prediction. So the models that have compensatory mortality, no matter what harvest rate you set, they're usually going to have a few more birds alive the next spring than the additive mortality hypothesis. And similarly, if you have strongly density dependent reproduction, usually you're gonna have a few more birds left the next fall than you will under the weekly density dependent. So every one of those models makes a different prediction.

Jim Nichols: 09:39

And what that means is if you just set your management regulations according to any one of those particular well, according to any one of those models, you get a different answer, a different set of regulations than you would had you chosen another model. And so the adaptive management process basically has to do with the amount of faith you have not faith, but confidence that you have in the predictive abilities of those different models.

Mike Brasher: 10:06

So we have these four different models and we're talking about a lot of this analytical work that happens in the background. That's some of the bookkeeping that I think you're talking about, but there's a lot of analytical work associated with that where you're actually projecting the effects of a given regulatory decision under these alternative models. So but how do you combine all of that? And that's where you're going with regard to weights and the degree of confidence that you have in these models. So talk about that and how different levels of confidence in these models is incorporated into, you know, the overall, what ultimately gets selected as a regulatory, as the optimal regulatory decision for a given year.

Jim Nichols: 10:43

That's a good question and it gets to sort of the key of adaptive management because the word adaptive in adaptive management, I mean, it could mean a gazillion things, but it actually means that basically we're trying to adapt to new knowledge that we get every single year and hopefully thus be smarter in the regulations we're setting. So how does that knowledge come about? Well, the smartest thing to do in a given year, as I said, we can't just select one of those models because every one of them will give us a different set of different recommendation about which regulations we should use. And so what we do, you can think of it as being sort of an average of what's smartest from all the different models. But how do how do we compute that average?

Jim Nichols: 11:31

Well, it's based on how much confidence we have in each one of the different models. And if we have a lot more confidence in one of the models than another, then it's sort of a weighted average, but it's giving a lot more relevance or weight to one of the models that's doing a good job of predicting what happens next year, then it gives to models that aren't doing a good job of predicting. And so the way the adaptive management program started out in 1995 is we gave each one of the models equal weight. Now why would we do that? Because you as you pointed out, we had, you know, decades of historical data and we actually could have empirically determined sort of a starting point where we gave a certain amount of weight to a little bit more weight to one of the models than another one say.

Jim Nichols: 12:21

Well, the reason had to do with the other the other big rationale for using adaptive management and that is taking into account different people's belief, different stakeholders in their beliefs. And the minute you would have put a whole lot of you know, some more weight on one of the say on the compensatory mortality model and an additive mortality model, you would have had a group of folks saying, hey, wait a minute, you cheated, you started out the system with way more weight on your model. And so no wonder my model doesn't look as though it's, we don't have much confidence anymore. And so there was since because of that motivation that the importance of being able to treat everybody's ideas fairly, in this case ideas about how the system responded to to management actions, each one of the models was started out with equal weight. So what that meant is that essentially we had just a garden variety average to base our initial year, the 1995, '96 hunting regulations on.

Jim Nichols: 13:23

And then after that, what happened is that you go ahead and you make your predictions and take your average and go ahead and establish your regulations for that year. And then what happens? Then you implement those regulations, in the next spring you get an idea of what actually happened. What do I mean by that? Well, what does the population look like now?

Jim Nichols: 13:46

And then you compare not any sort of an average, but you compare the population size the next spring that was predicted by each one of your different models separately. And you say, one of these did best? Which one of these came closest to what we actually saw? And basically the model that does best, you increase your degree of confidence in that model and you decrease your degree of confidence in the models that didn't predict as well. And so now when I make my decision, say in 1996, now we had model weights that were a little bit different from everybody having the same weight.

Jim Nichols: 14:25

And so now we gave a little more weight to one model than another, but we didn't do it because one particular state had a water state biologist who had a more compelling argument or who was had more political clout, we did it because that was what the data suggested in terms of predictive ability of the different models, which is what we want our models to be able to do is predict things. And so these model weights, these degrees of confidence in the models are a big deal and they're established strictly by the historic ability of models to predict what next year's population size ought to look like given this year's population size and given that the and given the regulations that you select.

Mike Brasher: 15:12

Is it fair to say that model weights can be interpreted as those change as one develops more weight through time, then that is sending a signal that that hypothesis, that model is the one in which we can have greater confidence with respect to how the system works and how Harvest interacts with it?

Jim Nichols: 15:33

Yeah. That's exactly right. And in fact, if you were lucky enough to have one of those models sort of be dead on, which is extreme I would be extremely lucky if it just was almost a perfect description of reality, then what would you expect to happen? You'd expect basically the model weight to go to one for that particular one and to go to zero for for the other models in the model set. And I'm not silly enough to believe that we were lucky enough to capture nature exactly in any one of those models, you almost can't by definition, but that's the way it works.

Jim Nichols: 16:06

And once again, now twenty five years later, there has been fair amount of separation among those those four different models.

Mike Brasher: 16:14

Do you recall off the top of your head? We don't have to go with the exact weights that are assigned to them, but do you recall how those how those alternative hypotheses, alternative models are falling out?

Jim Nichols: 16:25

Well, right now I do. Yeah, and what I can't remember is each year's dynamics and what happened each year, but yeah, so right now it turns out that the two strongly density dependent reproductive rate models kind of fell out. In other words, populations can't respond as well with respect to reproductive rate is indicated by those strongly density dependent models. And so the weights for both those, meaning we have strongly density dependent reproduction with both compensatory mortality and with additive mortality, the survival piece. And so both those model weights are approaching zero, like they're certainly less than, I don't know, least less than 3% or 2%, they're darn low.

Jim Nichols: 17:15

And the two remaining models then that have pretty high weight are the ones that say reproductive rate is fairly weakly density dependent. So it is density dependent, but the relationship is fairly weak. So we have a fair amount of confidence in that reproductive rate component. But the two survival rate components then are what we have variation in now. So the additive mortality hypothesis with weekly density dependent reproduction is the one that has the highest weight right now.

Jim Nichols: 17:47

And I think that's, you know, close to I should look, I should know this off the top of my head. I think it's about two thirds of the yeah, sixty, sixty five, seventy percent or something, and the one with compensatory mortality, but weekly density dependent reproduction is maybe half of that, so it's about a third or something. I think that's the way they're shaking out now. And so basically, we have very little confidence now in the strongly density dependent reproductive model, but there's still some uncertainty about the survival component of that with right now more weight associated with the additive mortality hypothesis, saying it looks like that's a little bit closer to reality in the sense that we're better able to make predictions with that model. Those predictions are a little closer than the ones that say mortality is completely compensatory.

Jim Nichols: 18:42

The one thing I will say is that I think this iterative testing of model based predictions each year, so now we've had 25 iterations, I think that's probably a stronger way to build inferences, in other words a better way to learn stuff than would be a single analysis even for those twenty five years. And so I think it's about the it's a really smart way to try to learn, but as I said, there's still some uncertainty with respect to two of those models still have non negligible weight and so it absolutely has not been completely resolved. And what probably that means is that reality is somewhere in between the two, those two endpoint hypotheses and maybe a little bit closer to one than another. Things have they have not been totally resolved, but yet on the other hand, I do think we're making decisions now based on that it have gotten better with time and that are better than we were doing at the beginning of the, of the program.

VO: 19:52

Stay tuned to the Ducks Unlimited Podcast sponsored by Purina Pro Plan and Bird Dog Whiskey after these messages.

Mike Brasher: 20:39

There are a couple of remaining topics here that I want to discuss before we close out this episode, and we'll try to keep these a bit brief. One of the things, Jim, is just how we characterize the state of the system, because we're making decisions for a given year relative to what we're observing with respect to the waterfowl population and habitat conditions. So talk about that. What were chosen as variables that were ultimately used to assess the state of the system and what's the importance of those in this entire process?

Jim Nichols: 21:08

Well, the key thing for the state of the system is the, I think the key so folks talk about state variables. And the key thing I think is the population size itself. Well, why is that? Because that's, you know, when you apply harvest rate, the population size, you know, the number of ducks that are harvested very much depends on population size, how many birds come down with fall flight. And so with respect to our objective and with respect to how our system is sort of doing, the duck population is doing well or not, basically we have to pay attention to how many birds there are.

Jim Nichols: 21:46

So to me that's the key characterization of the state of the system. And you mentioned ponds and that's another one that ends up being important. We don't have to keep track of it, but it's a smart thing to keep track of. So a long time ago, and I know the first quantitative paper I read about this was by again an early waterfowl pioneer named Walt Crissy, who looked at age ratios of ducks and noted what every duck person who certainly ever worked in the Prairie Provinces or North Dakota or whatever sort of knew, and that is years when there were lots of ponds, you had lots more baby ducks than years when you didn't have lots of ponds and potholes that were basically filled with water. And so he established a quote, Chrissy established a quantitative relationship and it was really really strong, in other words, it explained a fair amount of the year to year variation in age ratio.

Jim Nichols: 22:46

You told me how many ponds were wet that year and I could do a pretty good job of telling you how many young mallards it would be per adult now produced in the the spring and available in the fall. And so it was such a dominant environmental covariate that we thought we'd be it would be silly to ignore it. In other words, it's something that really helps us predict how many birds are going to be around in the fall at the beginning of the hunting season. Now there's all kinds of other environmental variables that likely influence that same thing. Habitat variables for example.

Jim Nichols: 23:26

But we haven't found any that are strong enough, that we're smart enough to be able to measure at a continental scale and that do a good enough job of predicting, so we say we'll include them in our prediction. So in other words, when let me digress real quickly. If you have an environmental variable that's an important predictor, in this case pond number, but in another case it could be something to do with habitat around the ponds, then you have two choices when you're modeling it. One is to incorporate that as a predictor in your model and actually include it as a variable that you pay attention to and monitor also. The other is just to say, well, I know in any given year, my prediction is going to be characterized by noise.

Jim Nichols: 24:11

And if an environmental factor isn't important enough, I'm just going to let it be part of that noise. It doesn't mean that I don't think it's important, and it doesn't mean that it doesn't exist. It just means that either I can't measure it well enough or it's not important enough for me to actually include in my modeling effort. It's easier to just think of it as being extra variation and I'll just deal with that variation by itself. And so for example, that was one of the first things that was asked about after the sort of adaptive harvest management models were developed, there was a lot of emphasis on this idea of what about habitat around these ponds that we're counting?

Jim Nichols: 24:52

What about cases where after a dry year somebody will plow right down to the edge of the pothole of the pond, as opposed to something where I have a pond where it might have dense nesting cover around it, there's got to be that's got to be a big deal, right? And so a postdoc came and worked with the management office and me for three different for three years pulling all the ag statistics he could, trying to get one of those agricultural statistics from from Canada and The US that would be a good predictor of age ratio along with the number of wet ponds. And there was just none of those that ended up coming out really, really that explained a lot of the variation. Now it doesn't mean for a minute that those kinds of things aren't important, it just meant that the combination of their relative importance and our ability to measure them from some large scale sort of agricultural monitoring program, the combination of those was just such that it didn't make sense to put them into the models. So so basically that's we had one important predictor, environmental predictor, number of ponds, and we knew how to monitor those while we're hanging out of airplane windows counting ducks, can count ponds as well.

Jim Nichols: 26:09

And the idea was to thus incorporate that into our predictions. And there has been there have been periodic searches for other covariates that are equally important that should be put in there. And up to now, it's not as though none of them are important at all, but it's that they haven't risen to the level where we can both measure them and they're sufficiently predictive that they've gone into these models.

Mike Brasher: 26:34

Jim, we've the Internet age is is pretty neat in a number of ways. One of which is the fact that highly technical reports and all sorts of information is readily available to pretty much anyone that knows how to use a computer and search on targeted terms. And as a result of that, we have a lot of savvy waterfowl hunters, a lot of people that are interested in the resource, that are interested in how harvest regulations are set, that have accessed some of the adaptive harvest management reports. Those are produced every year, and in those reports, there's a matrix which it's maybe an optimal decision matrix. I'm not sure if that's the technical term for it, but that's what I'm calling it here.

Mike Brasher: 27:15

But on one axis, it has mallard breeding population size, on the other, it has pond count, And in each of the cells within that matrix, it has what has been identified through an evaluation of these models that we've talked about and the assignment of the the current application of weights relative to each of those models. It it indicates when each of those cells what is the optimal regulatory decision for a given year relative to, you know, where we are relative to population size and pond counts. How did that come about, I guess, is the simplest way of asking the question. Is that something where you thought, alright, we need something that enables us to easily communicate what is going to be selected in a given year as the optimal regulatory package? Or or or is there something else more fundamental to that?

Mike Brasher: 28:09

You know, because it's basically a graphical depiction of what is the the optimal regulatory decision given to state variables.

Jim Nichols: 28:19

Right. So you described it perfectly, so I won't won't bother to re describe it and I'll say that I view that as sort of the fundamental output of a decision process. And so I'll mention one thing real quickly. So we've talked about these three components already of any kind of smart decision process and certainly about Adaptive Harvest Management. You got to have objectives, you got to have some actions, in our cases such of regulations that you decide on and you've got to have some means of projecting the consequences of those different actions to see which one is smartest and those are the models.

Jim Nichols: 28:56

We also have to have a monitoring program, which is one that again has been established for waterfowl folks had foresight a long long time ago and established these May population surveys for example and as well as the substantial banding program. So we need monitoring and as you said, in this case the monitoring allows us to estimate that key variable in the population, how many adult ducks are there in the spring in the breeding population. And it also allows us to estimate this key environmental covariate the ponds. And then the last, so monitoring is a component of a smart decision process. And the last component is a decision algorithm.

Jim Nichols: 29:37

It's something that says, okay, you give me all this stuff. You tell me what my objectives are, what actions I have to choose between, how I project the consequences of those actions, and you tell me something about my system looks like now. Now what's the smartest thing to do? In this case, what the the matrix you see is actually an output of this stochastic dynamic programming. It's basically an optimization that says given that the system looks like this and these are my alternatives and this is my objective, these are the smartest things to do.

Jim Nichols: 30:13

Now initially that might sound like a real easy thing like an optimization, but these so called dynamic is in that in that term, stochastic dynamic process. And the reason for the term dynamic means that it's not a one time optimization. Notice my if I just had a single year and somebody said maximize harvest, well, would I do? I'd take as many I'd open up the season to, you know, you could blast on the breeding grounds, you can do anything you wanted to because that was my objective is just number of dead ducks that year. But with a process like ours, we've already been very clear, we've got to leave enough ducks next year so that they can maintain a population size, maintain their population as well as allow hunters to have access to birds the next year.

Jim Nichols: 31:05

And so that aspect of an optimization program problem makes it a lot more difficult than a one time problem. In other words, we've got to do optimization that allows us to do something smart for this year, but then to leave the system state such that it's in good shape for us so we can make a smart decision the next year as well in terms of as opposed to having a system that hardly had any ducks in. So it's a difficult optimization problem, but basically if you think about it, that's the ultimate thing we sought to do in our decision process anyway, right? This is decide what the regulations package should be this year. And all it's doing is saying, you have to tell me two things about this year before I tell you what's smartest.

Jim Nichols: 31:51

You have to tell me how many birds I started out with in the spring, and you have to tell me how wet things were in the spring, how how good habitat conditions were for reproduction. And you tell me those things, and I'll tell you what the smartest thing to do this year is. So that that's sort of the yeah. I view that matrix you talk about as the fundamental output of this decision process.

Mike Brasher: 32:12

A lot of our listeners may be thinking, holy cow, why all this detailed discussion about adaptive harvest management? For crying out loud, we have had liberal regulations for the last twenty five years. Why do we need all this high-tech analysis and all this detailed thinking? Well, we're gonna touch on that a bit in the next episode, and so stay tuned for that. We also have a bit more information to cover, quite a bit more information to cover on the next episode.

Mike Brasher: 32:38

One of the things, Jim, that I want to lead off with on the next episode relates to the monitoring component of of adaptive harvest management. We've touched on that a number of times here already, but I won't lead off this next episode talking about why it's so important, talking about the key data streams, sort of emphasizing those again. Then we're also going to just reflect on we're going to offer some concluding thoughts on our next episode about the progress that has allowed us to make on a number of fronts. We're going to talk a bit about the fact that we have had 25 of liberal harvest regulations. I'm going to ask Jim to kind of explore why do we think that is.

Mike Brasher: 33:18

Do we just get lucky or is there some magic inherent in that has enabled this? So we have a bit more to discuss if you're willing to rejoin us. So you you up for that? Another episode, Jim?

Jim Nichols: 33:29

Oh, you bet. It's easy for me. I just feel sorry for your listener having to to go through some of this, but if it's yeah. I'm absolutely up for it.

Mike Brasher: 33:37

Well, that, Jim, thank you so much for joining us and encourage you, the listener, to to check back with us on the next installment of this Harvest Management episode. So thank you very much, Jim.

Jim Nichols: 33:46

Oh, thank you, Mike.

Mike Brasher: 33:49

Another very special thanks to our guest on today's episode, Doctor. Jim Nichols. We appreciate his time and his expertise during retirement, taking time out of his retirement schedule on these days to to share with us some of his expertise and some of his knowledge in in this field. As always, we thank our producer, Clay Baird, for the great work he does on these podcasts. And to you, the listener, we thank you for your time and support of this podcast and for your support of Wetlands Conservation.

VO: 34:16

Thank you for listening to the DU Podcast sponsored by Purina Pro Plan, the official performance dog food of Ducks Unlimited. Purina Pro Plan, always advancing. Also proudly sponsored by Bird Dog Whiskey and Cocktails. Whether you're winding down with your best friend or celebrating with your favorite crew, Bird Dog brings award winning flavor to every moment. Enjoy responsibly.

VO: 34:38

Be sure to rate, review, and subscribe to the show and visit ducks.org/dupodcast. Opinions expressed by guests do not necessarily reflect those of Ducks Unlimited. Until next time, stay tuned to the Ducks.