#23: New studies describe historic & future warming rates in northwest U.S. streams

[Dan Isaak]

Climate’s Cooking Along in Northwest U.S. Streams…

Hi Everyone,
So even though we in the northwest US have enjoyed a few years that
are more “average” in terms of snowpack and temperatures recently as
regional climate cycles do their thing (blog #12), the trends
associated with long-term climate change are still in motion and it’s
important not to take our eyes off the ball. To keep abreast of
emerging science regarding one of the fundamental climate-aquatics
issues, that of rising stream temperatures (blogs #3 - #15), this week
I wanted to highlight two new studies. The first describes how streams
across the northwest US have been warming the last three decades; the
second predicts how warm they could become by in the 21st Century.

Regarding the former and historical trends, there are numerous
published accounts of streams warming the last several decades (blog
#10) but these usually come from one or a few sites on larger rivers
because these are where the longest temperature monitoring records
exist (graphic 1). Hydroelectric dams, reservoirs, and urbanization
alter the thermal regimes of most large rivers these days, however, so
it’s unknown whether trends in these systems are indicative of trends
in unregulated streams, which often constitute the majority of stream
kilometers throughout regional networks. Isaak & colleagues recently
addressed these issues by querying the USGS National Water Information
System database (website: http://waterdata.usgs.gov/nwis/) to extract
temperature data on all streams and rivers in the northwest US that
had been monitored most of the 30 years from 1980-2009. Their query
turned up records from 18 sites at USGS gage locations where at least
21 years of daily temperature measurements were available (graphics
2). Unfortunately, 11 of those 18 records were on regulated streams,
so only 7 were from unregulated streams, although these were spread
across four states.

The data are limited but Isaak & colleagues could identify several
important historical patterns in the monitoring records, including: 1)
warming trends at both regulated and unregulated streams, with trends
at unregulated sites being relatively consistent as these streams
responded similarly to regional climate forcing. Inter-site
variability was much higher among regulated sites, which precluded
detection of a common, statistical trend across all sites; 2) warming
trends at unregulated sites were most pronounced during the summer
season (0.22 °C/decade) and smaller when averaged over the full year
(0.11 °C/decade); and 3) seasonal temperature trends at unregulated
stream sites closely tracked seasonal air temperature trends at nearby
climate stations, which included a statistically significant cooling
trend during the spring period (graphic 3). Perhaps the strongest
conclusion from this study, however, is that long-term temperature
monitoring efforts have been inadequate. A tiny number of sites
currently exist on unregulated streams where historical climate trends
can be directly described through examination of a monitoring record.
Data from sites like these are much of what we have to provide
unambiguous evidence of warming across approximately 350,000
kilometers of fish-bearing streams, despite, by some estimates,
several billion dollars being spent on fish & wildlife restoration
efforts in this region the last 30 years.

Hmmm…we should probably make sure we’re not in the same boat 30 years
from now. One would hope we’ll eventually be swimming in long-term
records since full year monitoring is so easy and inexpensive with
modern temperature sensors (blog #3, #4, #8). That said, we don’t have
the luxury of waiting another 30 years for trends at new monitoring
sites to manifest, so it’s also good to know that a lot of useful
information can be leveraged out of even short-term temperature
records as described in previous blogs (#7, #14; graphic 4). The most
important thing is to simply start collecting stream temperature data
in places where it doesn’t already exist &/or to pick a few sites and
monitor these for as long as possible (then hand them off to someone
else when you retire or move to a new job).

Our second study by Mantua & colleagues provides a good example of how
short-term temperature records can provide inference over much longer
time-spans. This study looked at thermal regimes in many of the big
salmon rivers across the state of Washington and how these could
respond to future climate forcing during the thermally stressful
summer period. The model linking historical air temperatures to river
temperature monitoring data was similar to one we described earlier
(blog #14; graphic 4). Once the historical linkage was made, an air
temperature increase representing an IPCC climate scenario (A1B, B1,
etc) at a specific future date was simply substituted into the model
to predict the future river temperature. Their results suggest that
many of Washington’s salmon streams could experience temperature
increases of 1°C – 2°C by mid-century, and 2°C – 4°C by the 2080’s
(graphic 5). Mantua & colleagues also describe how the duration of
thermally stressful temperatures for salmon is expected to increase at
specific river locations in the future. That answer, of course,
depends a lot on a stream’s historic thermal regime because a cold
stream has to warm up more than a warm stream before it’s stressful to
fish. Many of the warmest streams within a region are the largest
rivers because they occur at the lowest elevations and also often have
reservoirs that exacerbate warming. That’s readily apparent in the
temperature model projections (graphic 6), where thermally stressful
temperatures already exist along the mainstems of major rivers like
the Snake and Columbia, which also act as important migratory
corridors for adult salmon that return from the ocean each year. Mid-
to late-century projections for already warm rivers like these have
the annual duration of stressful temperature increasing from
historical averages of 2 – 3 weeks to 5 – 10 weeks (graphic 6).

The warming Mantua & colleagues project in their scenarios generally
implies higher rates than those described by Isaak & colleagues but
that fits with most climate model projections that have air
temperature increases accelerating during the first half of this
century. Either way, it’s likely to keep getting warmer for the rest
of our lives and at least the next dozen or so fish generations
barring some unforeseen breakthrough in clean energy to power the
world’s economies in the next decade. Understanding the implications
of that warming for fish, identifying key population thresholds well
in advance of their exceedance, and developing means of strategically
allocating scarce conservation resources to maximize their impacts are
going to be vitally important. Distributions of many native fish
populations in the northwest are already fragmented and/or seasonally
stressed by temperatures that are too warm, so any additional warming
will only exacerbate these constraints (as we’ll see in the biology
module).

Until next time, best regards,
Dan

[DaveP]

Hello Dan and others,

I always enjoy these thought-provoking posts.

I have every faith in the recently published projections, which are
very sobering in terms of potential impacts on aquatic systems.
However, some caution may be needed in looking at historical trends.
Many of the data sets that document increasing trends in temperature
phenomena range from around the 1940s to the present, including some
of the data presented here, encompass a cool PDO followed by a warm
PDO. So if you draw a straight line through the data, your eye
believes the straight line. But if you look more closely, you can
(perhaps) see a step change between the two phases near the 1977
boundary. The same thing has always concerned me about snowpack
data. I'm a little surprised that the "skeptics" have not called this
out. I admit that I tend to look for this PDO connection in all kinds
of natural resource data, so may be biased.

Assuming that we are now in a cool PDO phase again, we will see if the
numbers decrease and how that might change the time series -- and
there might be some explaining to do. If they continue to increase,
then we have a strong confirmation of general warming.

Dave Peterson
USFS-PNW Station
Seattle, WA

.

[Dan Isaak]

Hi Dave,
The PDO & climate cycle point is a very salient one and was a primary
motivation for why I estimated the trends in the historical trend
paper just published in Climatic Change based on reconstructions
rather than using raw trends. None of the other stream temperature
studies I'd seen previously had dealt with the climate cycle issue,
which is a huge potential confounding factor in stream temperature
time series that are relatively short and/or of the same length as a
climate cycle like the PDO that so strongly affects things in the
northwest US. In particular, the paper by Kaushal et al. 2010 that
appeared in Frontiers in Ecology & the Environment & got a lot of
national attention made no attempt to address the climate cycle issue
& when I started using some of the same datasets but getting very
different estimates I became concerned. The difference was that the 30
year periods we were looking at were offset by 2 - 3 years & that made
a large difference in the warming estimates because it affect the
start/stop points in the datasets relative to the PDO cycle you
describe in 1977.

So rather than estimating a trend by looking at the regression slope
from the raw temperature data, I thought it was more valid to
reconstructed the trend by linking interannual variation in stream
temperature to discharge and air temperatures measured at nearby
climate stations using a multiple regression. Once that link was made,
I then used the smoothed trends in air temperature and discharge at
those climate stations over the previous 30 and 55 year periods,
respectively, to filter out the large effect PDO has on discharge in
particular. Granted, the historical estimates then depend on the time
periods that were used to do the smoothing but the approach does
provide better answers that correct for climate cycles and also
missing years of data, which is another issue in the limited time
series that are available. To be more comprehensive, I could have also
used a 55 year air temperature period to include another historical
scenario but chose the 30 year period for air temperatures because
that's when air temperatures around the globe started showing pretty
consistent warming and that period also better capture the
acceleration in air temperatures that is occurring and would be lost
in the use of a longer period.

Dan