#39: Part 3, Monitoring to detect climate effects on fish distributions: BIG DATA regional resurveys
Dan Isaak
Hi
Everyone,
So as we saw last time, there are ways to do the time travel necessary for understanding how climate change is affecting fish distributions, and to develop that understanding in the next few years. Streams with good historical transects are out there & will be important sources of information, but they are also relatively rare against the backdrop of 100,000’s of stream kilometers in regional networks. Scattered across those same networks, however, are hundreds, sometimes even many thousands, of fish survey sites our “army in the woods” has been busily sampling in previous decades (graphic 1). If we could somehow tap into that wealth of data to extract information regarding the climate signal, then we’d really be in business.
Well, we can. And all it requires is doing what the Beever did with the pika. That is, doing what Eric Beever & colleagues did via broad regional resurveys of pika populations across the Great Basin in the western US (graphic 2; paper attached). They simply revisited habitats where pika populations had been documented in previous decades, determined whether populations still existed, and then examined the frequency of extirpations relative to local climatic conditions. To get accurate measures of local climatic conditions, they deployed temperature sensors within the pika habitats (e.g., blog #9) and reconstructed long-term trends through linkages to nearby weather stations (e.g., blog #14). What they found was that 9 of 25 historical populations had been extirpated and that measures of local climate (i.e., average summer temperature, # days exceeding threshold temperatures, etc.) were good indicators of pika persistence. In particular, those pika populations that were already near important thermal thresholds during historic surveys proved to be most vulnerable to recent extirpations.
So as you’ve probably noted by now, pika are not fish but rather small, adorable, rabbit-like mammals that live on the side of mountains rather than in streams. Warm and fuzzy yes, but nowhere near as cool and slimy as our favorite animals, so modifications of Beever’s pika study are needed for aquatic applications. First, we generally have poorer historical documentation and distinct population boundaries to work with for stream fishes than is the case for pika. Where we do have those things and extirpations have occurred, the complex of contributing factors generally has more to do with species invasions, excessive harvest, and habitat degradation than recent climate change. So it’s tough to do something as clean-cut climate-wise as Beaver’s pika study. But we do have those masses of fish surveys from previous decades to play with. And if per the general predictions from the bioclimatic models are right (blog #33), we’d expect that as it gets warmer, and species distributions shift to track thermally suitable areas, then there should be a disproportionate number of extirpations at sites that have been historically warm for some species (graphic 3). Similarly, we’d expect a disproportionate number of colonizations at new sites that had previously been too cold for some species.
These will be subtle patterns that take decades to manifest given the incremental shifts of isotherms shifts relative to short-term variability (blog #’s 36 and 37), and as we saw last time with transect resurveys (blog #38), there will be other factors such as invasive species & habitat degradation to confound the climate signal. But a BIG DATA approach that resampled hundreds of sites strategically placed along the margins of thermally mediated boundaries would allow the signal to be extracted from the noise using appropriate statistical techniques. In my view, getting good estimates of the rates at which fish population boundaries are shifting due to climate change is presently the most important question in the climate-aquatics world. Doesn’t matter whether we get those estimates through a BIG DATA regional resurvey, or transect resurveys, or yet some other means, as long as we get them. That being the case, today we’re initiating a global contest, much like the X prize, to see who can provide the first conclusive estimates of these shift rates (graphic 4; fine print disclaimer: unlike the X prize, there is no monetary compensation for this prize, only the eternal esteem & recognition by your colleagues as one famous fishy person).
One challenge with the BIG DATA approach is targeting the site resurveys as precisely as possible. Brute force resampling of all historical sites, which could run into the thousands, would be too expensive in most cases so criteria to guide sampling efforts efficiently are needed. Focusing on older sites (e.g., original surveys > 15 years ago) with faster climate velocities (i.e., lower stream slopes) would be a good start as these areas are most likely to show changes. Also importantly, however, we’d want sites meeting those criteria to be located near thermally mediated species boundaries because we wouldn’t expect distribution shifts to occur where temperature hasn’t been a limiting factor. Identifying thermally limiting sites requires some means of inference about local temperatures at all the candidate sites. Crude surrogates like elevation could be used, or one could take Beever’s approach and deploy sensors within the fish sites. But the former is very imprecise and doing the latter for hundreds of sites across a large area is a logistical nightmare because of the multiple trips it requires to each site (1: install sensors, 2: sample fish populations, 3: retrieve sensors) and it wouldn’t provide the temperature information at the study outset for planning purposes.
No, the best option would be having a consistent set of stream temperature “maps” (blog # 26) derived from accurate models against which all historic fish survey sites could be referenced. Simple plots of species occurrence relative to a consistent temperature baseline would immediately reveal sites near thermal boundaries (graphic 5), while also providing the significant side benefit of describing realized thermal niches for various species (which are critical for assessing climate vulnerability and making projections). Unfortunately, regional stream temperature models capable of providing the necessary historic climate scenarios don’t yet exist in most places. But efforts to develop such models are underway and next time out, we’ll highlight how something called “crowd-sourcing” is being used to develop a BIG DATA stream temperature model in the Northwest US that can facilitate regional fish resurveys, among a good many other things…
Until then, best regards,
Dan