#33: Part 1, Fish distribution shifts from climate change: Predicted patterns
Dan Isaak
The Sky is Falling!...(maybe?)
Hi Everyone,
So after dealing with time last time in fish phenologic responses to climate change, we’ll look in the spatial dimension this time out. More specifically, we’ll look at how fish species distributions are predicted to shift in response to climate change and what empirical support there may be that distributions are already shifting. It’s a big deal if a species we’re used to having in a certain stream or lake disappears from that system (especially if we’re investing significant resources to keep it there as is the case with some protected species), so as you might imagine, this topic has sparked much interest in the fisheries literature over the last two decades. The topic is large enough, in fact, that we’re going to split it into a mini-module of 3 – 4 related blogs. Today we’ll focus on bioclimatic model predictions of changes in fish distributions this century. By my count, there are some 23 peer-reviewed manuscripts that have developed these models and used them to predict the future (full bibliography in graphics 1 – 3). These date back to the pioneering work of Don Meisner with eastern brook trout in the late 1980’s/early 1990’s and since then we’ve been applying similar approaches to a host of fishes in different parts of the globe.
For background, these bioclimatic models are really just another type of habitat suitability model that many of us may be more familiar with. A mathematical relationship is built that links the occurrence of a fish species to a set of habitat characteristics, but in bioclimatic models, we also throw in a few climatic predictors like temperature or stream flow to compliment the static habitat features that are traditionally used in suitability models. Once the relationship between the climate predictors and the historical fish distribution is described, the values associated with the climate predictors are adjusted to represent future climate scenarios and the associated distribution of suitable habitats remapped. The only other differences between the bioclimatic models and traditional habitat models then is the larger spatial scales that are usually addressed with bioclimatic models & maps are generated to show the model’s habitat predictions spatially. A key assumption embedded in most bioclimatic models is that the habitat niches of species are conserved (i.e., constant through time), even if the distribution of these niches shifts in space. The most obvious example of this is the assumption that fish distributions are delimited by critical temperature isotherms (e.g., a temperature where it’s too warm for a species to survive) and distributions will track these isotherms as they shift in response to warming (graphic 4).
There are many good examples of these sorts of models, but here we’re highlighting a paper by Rieman and colleagues that predicts the distribution of bull trout habitats across the Columbia River Basin in the Northwest U.S. The first component of this study simply asked whether a relationship existed between climate patterns and the historical distribution of the fish. So longitudinal stream survey data were compiled for 76 streams from biologists across the river basin and the lower elevation limit of bull trout in these streams was modeled relative to latitude and longitude to describe spatial trends (graphic 5). The bull trout distribution dropped in elevation as one went north and west across the basin—closely matching the spatial trends in mean annual air temperature, so it was concluded that climate did indeed affect the fish’s distribution. The second part of the study then simply remapped the locations of habitats by adjusting the elevation of the critical isotherm that delimited the historic bull trout distribution (5 ˚C) to represent future warming scenarios of +0.6 ˚C, +1.6 ˚C, etc. (graphic 6). Two things are apparent from the latter exercise; 1) even a relatively minor amount of warming could result in a significant loss of thermally suitable habitat (0.6 ˚C ~ 20% reduction) for this species, and 2) habitat losses would not be uniformly distributed. For the same 1.6 ˚C warming, for example, some areas were predicted to lose 70% of thermal habitat whereas others were predicted to lose only 20%. One thing to keep in mind with regards to a species like bull trout is that it has a very cold thermal niche, so often occurs as high in streams as is possible and lacks upstream refugia. Other species might be able to retreat upstream and not experience a net loss of thermal habitat…assuming, that is, nothing is in the way to block those movements.
In a second paper, Wenger and colleagues add several dimensions to the approach taken by Rieman. A huge fish database consisting of 10,000 unique survey locations was compiled from managers and researchers across the Rocky Mountains in the western U.S. (graphic 7). The occurrence probabilities of four trout species in this database were then modeled as a function of physical habitat characteristics, the occurrence of competitor species, and climatic attributes. Climate was represented not only by air temperature predictions from a global climate model (GCM), but also by a stream hydrology model linked to the GCM scenarios and tailored to predict flow changes within individual stream segments across the Rocky Mountains (described in Blog # 20; flow metrics archived online at: http://www.fs.fed.us/rm/boise/AWAE/projects/modeled_stream_flow_metrics.shtml). Here again, large reductions in the amount of future habitat for trout species were predicted to occur but there were also big differences among species in the amount of anticipated change. Habitats for brook trout, a fall-spawning species introduced to the Rocky Mountain region from the eastern U.S., were predicted to decline most (77%) by the year 2080 (graphic 8). Temperature increases, combined with an increasing frequency of winter flooding that negatively affected embryo and juvenile survival, were hypothesized to account for this large reduction. Rainbow trout habitats, in contrast, were predicted to decline only by 35%, in part due to a warmer thermal niche than the other species, but also because their spring spawning period renders them less vulnerable to alterations in winter flooding.
The main take home from these two studies and others like them is just how disruptive warming and other habitat perturbations associated with climate change may be for species distributions. Moreover, different things are likely to happen to different species for different reasons in different parts of their range. Even though climate change imposes a relatively consistent set of air temperature and precipitation changes across broad areas, complexity emerges from the interaction of these changes with local habitats and biology. Next time out we’ll take a closer look at the empirical evidence supporting whether the distribution shifts predicted by the bioclimatic models have been occurring. We have, after all, been predicting these changes with dozens of models now for more than 20 years, & surely, with all that’s on the line, someone has bothered to check whether the predictions are right…right?
As we’ll see, however, space may turn out to be the final fishy frontier…
Until next time, best regards,
Dan
Dave Peterson
BCE models have also been used recently to project the effects of climate change on tree species distributions and bird species distributions. In this case, the use of BCE models is much less defensible than for fish, and they have been severely criticized in the literature, because the projections are almost certainly highly inaccurate. But BCE models are easy to construct, and the results (MAPS!) make the future appear to be seductively easy to predict. Resource managers like the maps because they are easy to interpret, not being aware of the assumptions behind them or of the availability of other predictive tools that may be harder to use by more accurate. Scientists who are producing these BCE maps (atlases they sometimes call them) for tree species without thoroughly stating assumptions and caveats are doing a real disservice to the resource management community.
As a forest ecologist, I don't know anything about process modeling for fish species, but it seems like it deserves a mention relative to the simplistic approach of BCE modeling. Combining different modeling approaches may or may not produce a "better" answer, but it allows a broader scientific perspective on projections of species distribution.
Dan Isaak
As far as I'm concerned, the jury is still very much out on how accurate many of the bioclimatic niche models for fish ultimately prove to be. Future blogs will examine the observed empirical support in fish population trends that the models predict, but as we'll see, there is far more modeling and predicting that has gone on the last 20 years than there is measuring biological responses and validating. At some level we have to have these models for strategic planning purposes but we have to make sure they're based on very rigorous science before we start reallocating conservation resources based on their outputs. That said, I think their general conclusions, that there are likely to be big changes and spatial variability to those changes, is correct, but we'll need better precision in future models to understand the local details such that managers can use them to make project level decisions.
Though there are significant challenges to the bioclimatic approaches in general, I suspect that fish will ultimately prove to be one taxa for which accurate predictions can be made because they are constrained to linear networks (&have less wiggle room than terrestrial species), have relatively short generation times (so should track isotherm shifts much more quickly than trees), and limited ability to evolve warmer thermal tolerances. Factors other than temperature will accelerate or dampen the response to isotherm shifts, but I think the general pattern will be there once we start looking in the right places.
For me, the biggest uncertainty fish biologists won't be able to resolve anytime soon is what the future warming trajectory is. The climate modelers ultimately have to answer that one, but right now generally can't tell us whether temperatures will be 1 C warmer at midcentury or 3 C warmer & that swamps much of the uncertainties in the fish world.
Dan