#25: NorWeST: A Massive Regional Stream Temperature Database

[Dan Isaak]

Lucky region “finds” $10,000,000 of useful data lying around

Hi Everyone,
As we saw with regards to NoRRTN last time, it’s easy to design and
deploy temperature monitoring networks inexpensively these days with
the advent of modern sensor technology (blog #24). Just as
importantly, however, we sometimes also have tremendous resources
available to us in the form of existing measurements if these can be
aggregated into useful forms. So this week we’re introducing NorWeST,
which is a comprehensive interagency stream temperature database being
assembled for Oregon, Washington, Idaho, Montana, Wyoming and bits of
northern Colorado, Utah, and Nevada. Development of this database has
been an ongoing effort the last year (& we still have a ways to go…)
made possible only through the willingness of 100’s of individual
biologists and hydrologists & 10’s of state, federal, tribal, and
private partners to share copies of their data. The NorWeST
compilation effort is phase I of a project funded through one of the
U.S. Fish & Wildlife Service’s Landscape Conservation Cooperatives
that have sprung up across the country the last few years (http://
www.fws.gov/science/shc/lcc.html). Phase II of the project will use
NorWeST to develop a regional stream temperature model for use in
climate vulnerability assessments for cold-water fishes across the
Northwest US. More on that & other potential uses of the database in a
bit, but first a few of the raw statistics…

At present, NorWeST contains digital copies of temperature data from
15,000+ unique stream sites (& still growing daily), some 45,000+
summers of measurements at those sites, and perhaps 45,000,000
individual temperature measurements (graphic 1). As is obvious from
these statistics, we don’t monitor temperatures very long at most
stream sites, averaging only 3 summers/site (and it’s a very skewed
distribution with a median of 1 or 2 years), which limits our ability
to describe long-term trends associated with climate change (blog #’s
10, 11, & 23). We also usually monitor only in the summer and rarely
during other times of the year, which is why we’ve taken to reporting
the number of “summers” as a standard summary metric. Even with the
various imperfections, NorWeST still contains an awesome amount of
data, perhaps the largest compilation of its kind in the world, & the
aquatics community in the Northwest U.S. deserves major kudos for
their dedication to stream temperature monitoring efforts.

I like to call NorWeST a stealth sensor network because it’s been
flying under the radar as it’s slowly & steadily grown the last few
decades while collecting massive amounts of useful data. Yet we didn’t
even know the darn thing existed because the work to measure all those
stream temperatures was spread across hundreds of different people
scattered among dozens of agencies and coordinated monitoring efforts
have rarely been done across agencies or at large spatial scales. If
we had to start from zero and collect all the data in NorWeST again, a
back-of-the-envelope estimate would put the cost somewhere around $10
million. Figure a temperature sensor runs ~$100 & we’d have to put one
at each of those 15,000 stream sites with a small army of field
technicians. Since we’d be collecting primarily summer data, we’d have
to visit each site twice a year (which is another reason full year
monitoring & one site visit/year makes more sense) and be paying for a
lot of gas and vehicle mileage in the process.

As an aside, it’s worth noting that temperature data are just one
example of the proliferation of data that’s becoming available for
streams. As much or more data exists for fish distribution/abundance
surveys and increasing numbers of surveys are being done for genetic
attributes and other water quality parameters. Moreover, the
covariates to predict these biophysical attributes are steadily
improving and oftentimes already available in some form for many
streams (think USGS NHD+ national hydrography layer or satellite
imagery of riparian vegetation or other watershed attributes). We are
entering what some have called the era of “Big Data” wherein modern
sensors, geospatial technologies, and advances in computing power &
memory are making mountains of data available that can be used to
better describe, understand, and manage stream environments (for an
interesting recent perspective on this topic, see “Staying Afloat in
the Sensor Data Deluge” (attached)).

Regarding the potential applications of NorWeST, that’s where the true
value of the database lies and ultimately dwarfs that $10 million
valuation. First, temperature monitoring efforts in the Northwest can
now potentially be done much more efficiently (graphic 2). Nothing is
more efficient than simply sharing copies of existing data rather than
going to the field to collect redundant information, so by the end of
the year, the plan is to post the NorWeST database to its own website
where those datasets we have permission to disseminate can be
downloaded. With that information available, regional planning efforts
might be undertaken to implement a strategic monitoring design that
builds most efficiently from existing data collection efforts (blog #4
& #24). Even in the absence of any centralized & formal monitoring
plan, however, many of us will still be out there monitoring stream
temperatures for a variety of reasons & to help facilitate those
efforts, we’ve developed high-resolution maps showing where within
each state data do & don’t exist (archived here at the project website
(http://greatnorthernlcc.org/technical/stream-temp). Armed with this
spatial information, we can all be more thoughtful about where new
data are collected & my guess is we’ll choose more often to measure
temperatures in streams where information is lacking. Simple self-
organizing principles, therefore, will help fill current spatial gaps
in regional coverage.

Second, NorWeST will facilitate more temperature research & accelerate
the pace at which we’re learning about thermal regimes in streams. The
experiences I’ve had doing temperature research in the past suggest
that developing & organizing the database often takes at least as much
effort as the analysis & preparation of a manuscript. Those
preliminary steps will be eliminated after NorWeST is made publicly
available and interested parties need only develop useful research
questions to guide queries of the database before proceeding to the
analysis stage.

Third, as mentioned above, we’ll build a regional stream temperature
model using all the data in NorWeST so that a means exists of making
consistent predictions of thermal conditions across all the ~350,000
kilometers of perennial streams in the Northwest (blog #7). These
predictions will be made for a variety of historical and possible
future scenarios to facilitate better climate vulnerability
assessments for aquatic organisms. A common limitation of previous
such assessments has often been the use of air temperature as a
surrogate for stream temperatures, despite the fact that these
variables may be weakly correlated in complex mountainous landscapes
like those characterizing much of the Northwest (graphic 3). As such,
we’ve often had poor representations of the most fundamentally
important variable when it comes to understanding climate & climate
change effects on stream biotas. The hope is that much of the previous
imprecision will be eliminated by developing standardized “maps” of
stream temperature predictions so it’s possible to see how species,
aquatic communities, and other forms of biodiversity relate to thermal
characteristics across the region. The other hope is that because
these maps are developed from everyone’s data & will cross
jurisdictional boundaries, they may facilitate better coordination of
conservation efforts among agencies (graphic 4).

At a time when everyone’s budget is shrinking, aquatic resources are
under greater duress, & we’re perpetually being asked to do more with
less, developing comprehensive databases like NorWeST & extracting
useful information from them is one easy way of making everyone’s
dollars go a lot farther and working more efficiently. Not to say that
there aren’t issues with these aggregated databases that are often
characterized by varying degrees of spatial non-randomness,
clustering, and gaps in coverage, but as long as the stream attributes
being measured are assessed accurately & the location of the survey is
georeferenced, then there’s going to be some useful information there.
Moreover, new analytical techniques, developed specifically for data
on stream networks (which we’ll explore in future blogs), not only
deal with this sort of non-randomness, but actually use it to their
advantage to improve the predictive accuracy of models while yielding
unbiased parameter estimates and providing access to a host of
previously impossible types of analyses for stream data. It all sounds
too good to be true, but so does finding $10,000,000 of useful data
lying around…

Until next time, best regards.

Dan

p.s. One last good question is how the heck one goes about using the
entire mass of data from NorWeST in a single model? Well, just as is
the case with collecting that much data, one person doesn’t do it. To
quote a famous politician, it literally “takes a village” & we have
our own small village of 17 or so people scattered across 5 states and
2 countries working on this project in various capacities. About half
our village is composed of scientists, but most of the real work is
being done by database engineers, GIS/spatial analysts, technicians,
webpage designers, statisticians, etc. that are digitally connected in
a virtual office as we build the infrastructure for the regional
model. I’d grappled with finding a name to describe this thing we’re
doing & then ran across a recent paper by Michener & Jones (attached).
In this paper, they coined the term “Ecoinformatics” to describe the
rapid evolution of ecology toward a more open, interdisciplinary,
collaborative, and data-rich endeavor in response to the challenges
and opportunities of the day. And so, as best I can figure, we’re
doing an Ecoinformatics approach to stream temperatures as we attempt
to better understand and predict how climate change may affect aquatic
resources across the Northwest.