#37: Part 1, Monitoring to detect climate effects on fish distributions: Sampling design and length of time
Dan Isaak
What will it take?
Hi Everyone,
So simple question. If one were to design a monitoring strategy for measuring the effects of climate change on fish populations, how would it be done? Well, since climate’s everywhere & affecting biological processes all the time, that’s probably too broadly phrased. Let’s constrain the question, therefore, to describing the effects of climate change on the spatial distribution of a population or species because that’s what we’ve most recently focused on in the bioclimatic model blogs (#’s 33, 34, 35). Given that one of the primary assumptions in those models is that distributions are delimited by critical isotherms, a good start would be focusing where population boundaries are thermally mediated (typically at the margins of a distribution) and things are either a bit too warm or too cold to be prime habitat. In these areas, we’d want to estimate the location where a population boundary actually occurs & one straightforward way of doing this is to sample a series of sites along a transect & determine the presence/absence of a target species relative to the local temperature gradient (graphic 1). Key to this sampling design is covering a wide range of temperatures that exceed some portion of the target species’ thermal niche (i.e., either too warm or too cold). And…sampling multiple sites where a species didn’t occur along a transect due to thermal limitations (somewhat ironically, one needs to sample where a species isn’t to most reliably estimate where it is (at least for statistical purposes)).
Simple enough, but transect sampling to determine the location of thermally-mediated species boundaries is a very different sort of sampling than we’ve traditionally done, which has more often consisted of trend monitoring of abundance at sites that were often initially chosen for having lots of the target species. For a variety of reasons, those traditional abundance sites may be some of the last places we’d expect to see the biological effects of climate change (although data from these sites still are invaluable assets because of their long timespans & we’ll discuss ways of teasing a bioclimate signal from these in part 4 of this mini-module). That said, the transect/boundary approach isn’t necessarily that different from another type of traditional sampling design that has involved longitudinal stream surveys. Lots of these studies have been published, but they’ve typically been done for a variety of reasons other than climate change assessments. A good example is the study by Rieman & colleagues (attached) that estimated how far upstream invasive brook trout have pushed the distributions of native bull trout. Longitudinal samples of bull trout populations were done in 12 streams with & without brook trout and statistical comparisons made between the two types of streams to measure the difference in boundary locations of the bull trout distributions. Extending this traditional sampling design to make it work for climate change assessments requires only adding a temporal dimension by resurveying the same sites periodically, deriving new estimates of the population boundaries, and comparing those estimates to measure the amount of shift (graphic 2). Tingley & Bessinger (attached) wrote a great general ecology paper on the subject a few years ago that explores the statistical nitty-gritty & other considerations in these sorts of assessments (graphic 3).
Two important questions that will come up in bioclimatic monitoring designs are how frequently sites should be resampled? & how long will it take for a distribution shift to occur that could be attributable to climate change? For the former, a rule of thumb might be to not sample at frequencies less than the generation time of the critter being studied (for those interested, Morris et al. Ecology 89:18-25 provide a nice synthesis regarding how generation time affects species’ climate tracking abilities). And for relatively short-lived critters like most fishes, it makes sense & ensures that some inter-generational dispersal/colonization processes occur between sampling efforts that are relevant to determining the locations of population boundaries. So, for example, with the efforts I’m involved in to monitor distributions of inland trout, we’re resurveying our sites once each decade and plan to keep doing it as long as we need to. Which leads to question 2—How long will it take for a distribution shift to occur? It would be nice if we had at least a general idea so we could be sure to sample long enough that the effects of climate change were truly assessed…and because hiking electrofishers up steep mountain trails to torture fish with electricity isn’t easy…and will get harder I suspect each decade. So will we be doing those hikes for another decade, 2 decades, or will I be an 80 year old with a bad back and still no end in sight? And if it might take many decades, then it’s probably smart to know before my back gives out so I can be working to secure the funding necessary to continue the monitoring, and more importantly, engaging a younger scientist with a stronger back to continue the work after I hang it up.
But back to the question at hand—How long? In the absence of good empirical data from fish case histories that describe the details of multi-decadal range shifts, it’s difficult to address the question directly. But we can make an educated guess through extension of what was discussed last time in the “velocity” blog (#36). That discussion focused exclusively on the rate at which isotherms shifted during climate change but ignored the natural, short-term variability that would accompany these shifts. And that short-term variability (i.e., inter-annual and decadal timescales) is huge relative to the tiny annual increments that comprise the global warming signal. So the other important aspect of the Isaak & Rieman paper from last time addressed this variability question and calculated how long it would take an isotherm to move a statistically significant distance while exhibiting short-term variation (graphic 4). It was reasoned that biological distribution shifts from climate change wouldn’t be expected until that isotherm shift had occurred because short-term thermal “noise” would mask the warming signal at short timesteps.
Turns out that that calculation can be made with the same set of equations that are used for the velocity calculations but rather than going through another mind-numbing trigonometry session, we’ll just skip ahead to the results (graphic 5). As we’d expect, the faster an isotherm shifts & the less short-term variability in stream temperatures, the quicker a statistically significant shift occurs. But we also have to constrain the range of isotherm shift rates considered (x-axis in graphic 5) to something that’s realistic for current and anticipated stream warming rates; and associate that with specific stream slope & lapse rate conditions. That done, the calculations suggest it will take 2 – 6 decades for a stream isotherm to shift to a statistically different place along a stream’s course. This time span varies slightly for streams with different slopes & lapse rates, but only slightly—so in most instances it seems we’re looking at a fairly prolonged period over which thermal shifts manifest. Moreover, fish distribution shifts would be expected to lag behind these shifts, which would make these estimates conservative.
If that’s truly the case, then I’d argue it’s going to be game over if we start implementing new bioclimatic monitoring programs now & then wait multiple decades before getting the all-important estimates of fish distribution shift rates. We simply don’t have the luxury of waiting given the rate of global temperature increase. Indeed, the final data are in for 2012 and the U.S. just set a new record for the warmest year (by a lot) since direct instrumental records were initiated 117 years ago in 1895 (graphic 6). We need good estimates of fish responses in the next few years so they can be used to validate, refine, and improve the bioclimatic models to the point that decision makers have the scientific backing they need to start making tough calls. Until those biological estimates exist to “prove” that it’s happening to fish, I suspect we’ll continue to suffer from an “inertia of inaction” regarding tough choices because the “climate thing” will seem too abstract. But at the same time, until we’ve documented the biological effects and used them to calibrate the bioclimatic model projections, we run the risk of making poor decisions and misallocating scarce conservation resources if the model predictions prove to be inaccurate.
So a rock & a hard place be where we be at present. If only we could fish-warp back in time somehow to set up the necessary monitoring transects several decades ago, then we’d be in business. Actually, there is a way & we’ll talk about it next time….
Until then, best regards,
Dan