#40: Crowd-sourcing a BIG DATA regional stream temperature model for high-resolution climate vulnerability assessments
Dan Isaak
Everyone’s data, everyone’s skin in the game…
Hi Everyone,
So last time we described the potential that BIG DATA holds, this time we’ll demonstrate its power. Recall that ridiculously large NorWeST stream temperature database compiled from data collected by hundreds of biologists and hydrologists from dozens of agencies over the last two decades across the Northwest U.S. (blog #25)? That same database that contains some 45,000,000 hourly temperature recordings at more than 15,000 unique stream sites and would require some $10,000,000 to replicate if we were to collect it from scratch. Yes, that one (graphic 1).
Well, a small tribe of us temperature aficionados have been working diligently on NorWeST the last few years to surmount the technical challenges associated with organizing that much data, feeding it through the spatial statistical stream network models to extract unbiased parameter estimates (blog #’s 27, 28, and 29), and using those estimates and the model’s enhanced predictive abilities to generate accurate stream temperature climate scenario maps. And I’m happy to say that the day has finally come wherein most of those challenges are now behind us and we’re in the process of developing and distributing a regionally consistent set of stream temperature scenarios. Similar to earlier pilot work that involved application of the spatial models to smaller temperature databases (blog #7), the model predictions are very accurate (r2 ~ 0.90; RMSE ~ 1.0˚C) and provide the fundamental underpinnings for “smart maps” based on hundreds or thousands of observations (graphic 2; blog #26). Different from the earlier pilot work, however, are that the smart maps can now be generated at scales large enough to be relevant for conservation planning, and the maps are easily accessible and usable as geospatial data (i.e., ArcGIS shapefiles) distributed through a NorWeST website (graphic 3). We still have a lot of work to do over the next year to complete the stream temperature scenarios across the region (graphic 4), but at project completion, an accurate, consistent datum will exist to facilitate applies-to-apples comparisons and strategic climate vulnerability assessments across a 5 state region and more than 300,000 stream kilometers.
Because NorWeST addresses a significant need, it is already serving as a catalyst for a variety of related research and monitoring projects that are being developed in parallel, and which will ultimately compose an integrated system that sums to something larger than the individual pieces (graphic 6; I like to think of it as a “stream intranet” but that’s the subject of a future blog post). As illustration of the potential synergies, one project is co-registering a massive fish survey database compiled earlier by Wenger & colleagues (blog #33) with NorWeST historical temperature scenarios to develop regionally consistent thermal niche definitions (graphic 7). With stream temperature criteria based on these niches, it is then trivial to accurately map the distribution of thermally suitable habitats for historic conditions and to play “what-if” games with future scenarios (graphic 8). Moreover, subtraction of the historic thermal habitat map from a future map highlights habitats that are most vulnerable to loss as the climate warms & thermal thresholds are exceeded for a species (graphic 9). It becomes similarly straightforward to identify key historical fish survey sites for resurveys to describe the rate at which fish distributions are shifting as we discussed last time (blog #39). And we could also map where thermal constraints on upstream invasions would be ameliorated with future warming, or predict where and how long thermally isolated headwater populations may persist (graphic 10).
So a scientifically sound, accurate and consistent means of mapping stream temperature scenarios could go a long way toward providing a basis for strategic prioritization of efforts to conserve and restore habitats, populations, and aquatic biodiversity threatened by climate change this century (graphic 11). There will be many places and populations that, despite our best intentions & significant investments, we simply won’t be able to maintain in desired historical states. There will be other places and populations where what we do will make a big difference in effecting the outcome. We want to spend as much of our time and limited resources in the latter areas rather than the former.
But even with better information to inform conservation decisions, the choices we face in the next few decades will often be difficult and at times painful. So perhaps the most significant thing about NorWeST—beyond the useful data summaries and model outputs, analytical advances it has spurred, and sexy geospatial maps—is that it’s a crowd-sourced approach to doing science that could significantly enhance natural resource stewardship (graphic 12). Crowd-sourcing lets us leverage massive amounts of useful information from existing datasets at relatively low cost, which is always useful, but especially so during times of shrinking budgets. And because NorWeST is built from the collective efforts of hundreds of people that have worked for more than 60 natural resource agencies across the region during the past 20 years, literally everyone has some skin in the game & is engaged in the process. And it’s going to be regional & local communities of managers, researchers, decision makers, anglers, and the like that will ultimately have to come together in future years as the decisions are made about where, and what, work is done to conserve the species and fisheries that we care about.
Until next time, best regards,
Dan
For more details regarding NorWeST, visit the project website (http://www.fs.fed.us/rm/boise/AWAE/projects/NorWeST.html) or see this recent article in the Great Northern Landscape Conservation Cooperative newsletter (http://greatnorthernlcc.org/features/streamtemp-database).