EPA/600/R/12/676
Factors Influencing Trends in pH in the Wootton et al. (2008) Dataset
Cheryl A. Brown
Pacific Coastal Ecology Branch
Western Ecology Division
U.S. Environmental Protection Agency
Newport, Oregon
September 3 0,2012
-------
1. Background
Wootton et al. (2008) (herein referred to as "Wootton") analyzed an 8-year dataset (from
2000-2007) collected in a tidepool on Tatoosh Island, Washington and found that this region was
experiencing a rapid decline in pH (annual trend of-0.045 units per year), which was an order of
magnitude greater than that predicted by models relating atmospheric carbon dioxide (€62) to
oceanic pH levels. In this paper, they developed a non-linear regression model that explained
70.7% of the variation in the pH dataset, and found that atmospheric CC>2 level was the only
variable which explained the declining trend in pH. Wootton have continued to measure pH at
Tatoosh Island and have found that the pH levels continued to decline through 2008 - 2010.
In this report, the dataset collected by Wootton at Tatoosh Island during 2000 through
2010 is analyzed to assess whether the apparent rapid decline in the pH data can be explained by
other more localized processes. Specifically, the role of river discharge on pH was examined. In
addition, we analyzed temporal trends in another long-term dataset collected in Yaquina Estuary,
Oregon, to assess whether it is appropriate to extrapolate the long-term temporal trend observed
at Tatoosh Island over a larger region.
Study Area
The pH dataset used in Wootton was collected in the intertidal region of Tatoosh Island,
Washington. Tatoosh Island is located in a region characterized by complex physical
oceanography, which may complicate pH patterns (Figure 1). The oceanography in this region is
influenced by numerous physical factors including complex bathymetry, upwelling (both wind-
driven and tidal), and the presence of large-scale eddies and multiple river plumes. This region
is subjected to strong tidal, estuarine and wind-driven flows. Some of the factors that influence
oceanographic conditions in the vicinity of Tatoosh Island are listed below:
1) Complex bathymetry: There are numerous canyons and banks in the vicinity of Tatoosh
Island, which can influence currents as well as mixing and entrainment of deep water (Figure 1).
2) Upwelling: Wind- and tidally-induced upwelling occur in the vicinity of Tatoosh Island
(Figure 2; Foreman et al., 2008). This enhanced upwelling is a result of upwelling favorable
winds, estuarine flow (river discharge), and tides (Foreman et al., 2008). Deep water advected
upward due to estuarine outflow from the Strait of Juan de Fuca combined with upwelling
associated with canyons equals or exceeds wind-driven upwelling in this region (Hickey and
Banas, 2008). Upwelling in the region can be suppressed by downwelling conditions and low
salinity conditions (such as that caused by the presence of Columbia River or Fraser River
plumes).
-------
Figure 1. Map showing the complex bathymetry in the vicinity of Tatoosh Island and the
proximity of Tatoosh Island to a region of enhanced upwelling (Cape Flattery). Source Foreman
et al. (2008).
3) Juan de Fuca Eddy: There is enhanced upwelling off of Cape Flattery (Figure 1) which
leads to the formation of the eddy. The Juan de Fuca Eddy is an "upwelling center", which
upwells water from deeper depths than classical wind-driven upwelling (MacFadyen et al.,
2008). Waters within this large scale eddy originate from a mixture of sources including the
California Undercurrent, Subarctic, and Juan de Fuca source waters (MacFadyen et al, 2008).
Each water mass has its own unique physical and chemical signatures.
estuarine 6'uff
tidal
isopycnal x . \
mixing displacement mixing \/\\
OO V-~, ^7 4. ^C > * \
Figure 2. Schematic of factors which contribute to the generation of the Juan de Fuca Eddy and
enhanced upwelling off of Tatoosh Island (Foreman et al., 2008).
-------
4) Fraser River: Discharge of the Fraser River can influence circulation in the Salish Sea as
well as the formation of the Juan de Fuca Eddy. Fraser River discharge accounts for
approximately 80% of the freshwater inflow into the Strait of Georgia (Gustafson et al., 2000).
5) Columbia River Plume: At times the Columbia River Plume is advected into the Strait of
Juan de Fuca (Hickey et al., 2009), suggesting that water conditions at Tatoosh Island may also
be influenced by the biology and chemistry of the Columbia River plume water. In addition, the
Columbia River plume influences the biogeochemistry and circulation patterns on the
Washington shelf (Kudela et al., 2010).
spring
downwelling
winds
after 3-5 d
upwelling
winds
Figure 3. Interaction of multiple freshwater plumes on the northern Washington shelf (Hickey et
al., 2009) including: the plume from the Strait of Juan de Fuca (green), the northward plume
from the Columbia River (yellow), and the southwest plume from the Columbia (brown).
Recently upwelled water is indicated by dark blue.
Description ofWootton et al. (2008) Dataset and Analysis
Wootton et al. (2008) measured pH and other physical parameters in a tidepool on
Tatoosh Island. This tide pool was isolated from the coastal ocean when the water level was
below 80 cm above mean lower low water. Physical parameters (water temperature, salinity and
pH) were measured at 30-minute intervals with a Hydrolab DataSonde 4a multi-probe. These
parameters were measured from late spring/early summer through late summer beginning in
2000. Wootton et al. (2008) performed a nonlinear regression analysis to identify important
factors influencing pH levels in this dataset.
In addition to the measured physical parameters from Tatoosh Island, the non-linear
regression analysis also included upwelling index, remotely-sensed chlorophyll a, Pacific
Decadal Oscillation (PDO), and alkalinity. Alkalinity was estimated using the water temperature
and salinity measurement from Tatoosh Island using the empirical relationship of Lee et al.
(2006). Because there was significant diurnal variability in the measured pH and no averaging of
-------
the measured variables was performed, a diurnal term was included in the non-linear regression
analysis. The model formulation in the Wootton analysis was as follows:
pH = a + b * log(CO2) + h * sin(2:r * (cp + time of day / 24)) + u * upwelling + c * log
(chlorophyll a) + T* (water temperature) + d * PDO + k * log(alkalinity) + s * salinity
Data sources for the regional and global variables were as follows: monthly average
atmospheric surface CO2 concentrations measured by the NOAA Earth System Research
Laboratory (http://www.esrl.noaa.gov/gmd/ccgg/trends); monthly average upwelling index at
48°N 25°W (http://www.pfeg.noaa.gov); monthly PDO index (http://jisao.washington.edu/pdo);
and Sea-viewing Wide Field-of-view Sensor (SeaWiFS) estimates of monthly average
chlorophyll concentration (http://coastwatch.pfel.noaa.gov). Measured variables were only
included in the regression analysis if the tide level exceeded 80 cm above mean lower low water.
The dataset, which was analyzed in Wootton et al. (2008), included 19,364 observations after
elimination of outliers and periods with missing data (primary chlorophyll a). Dr. Timothy
Wootton provided the dataset (data from 2000-2007) used in the Wootton et al. (2008) paper, as
well as additional data collected during 2008 -2010.
2. Differences in Sampling between Years
One factor that could have influenced the Wootton analyses was variation in the months
of data coverage between years (Figure 4). This was due to sampling differences between years
in the measurements from Tatoosh Island as well as availability of remotely-sensed chlorophyll a
data. During the majority of the years included in the Wootton dataset (2001-2003, 2005, 2006,
and 2008-2010), there were significant declines in pH during April through September. Figure 5
shows an example of the declining trend in pH during 2002 and 2006. The only exceptions to
this declining trend occurred during 2000 and 2007, when there was a significant increasing
trend in pH (note: during 2000, there are only data from June -August and slope is low).
To control for the differences in sampling intensity and data availability between years,
we re-analyzed the Wootton dataset only using data from June and July, because these months
have data available for almost all of the years (2000-2003, 2005-2007). Data from 2008-2010
were not included in this analysis because no chlorophyll a data are available during this period.
There was a significant declining annual trend in pH of-0.041 in the "June + July" subset (2000-
2007), which is similar to that report in Wootton et al. (2008). The annual trend of-0.045
reported in Wootton et al. (2008) included all pH observations from 2000-2007, including those
that didn't have chlorophyll a available.
Using the model formulation in Wootton with only "June + July" data (note: only July
data are available in 2007), all terms in the model are significant and the overall r2 of the model
is 0.774 (Table 1). Using this abbreviated data set, we found the same general pattern as
Wootton.
-------
Table 1. Best-fit of parameters for model of pH using model formulation in Wootton et
al. (2008) including only "June + July" data.
Parameter
a
b
h
9
u
c
T
d
k
s
Description
Constant, pH
Change in pH with atmospheric CC>2,
pH/ppm CO2
Half amplitude of diurnal productivity
oscillation, pH
Phase shift from midnight of diurnal, h
Effect of upwelling, pH / (metric tons/s/100
m coastline)
Phytoplankton abundance effect,
pH/tmgchir1)
Temperature effect, pW ° C
Pacific Decadal Oscillation, pH/ ° C
Estimate alkalinity, pH / |i mole/ kg
Salinity effect, pH / psu
Value
54.72
-17.55
-0.121
2.375
-0.002
0.217
0.102
-0.072
-0.705
0.009
GR^
-
15.9
40.5
4.5
14.7
36.0
4.8
0.2
0.3
99 9
*GR is the generalized R , which is calculated as: GR = 1 - Z(RSSiuiimodei)/Z(RSSreduced
model)- GR2 represents the reduction in explained variance when a given variable is
removed from the full model (Wootton et al., 2008).
-------
9-
8-
7-
6-
5-
2000 2001 2002 2003 2004 2005 2006 2007 2008
Year
Figure 4. Data used in Wootton analysis by month for each year (n=19364). Note: Months that
did not have chlorophyll a data available are not included in this graph.
9.0-
8.8-
8.6-
8.4-
8.2-
I 8.0^
7.8-
7.6-
7.4-
7.2-
7.0-
o 2002
n 2006
56789
Month
Figure 5. Trend in pH during April through September for 2002 and 2006. Note: There is more
pH data available than that used in the Wootton analysis (see Figure 4) due to gaps in
chlorophyll a data.
-------
3. Effect of River Discharge
Tatoosh Island is located near the Juan de Fuca Strait and circulation in the vicinity of the
island is influenced by discharge of rivers, which flow into the Straits of Georgia and Puget
Sound. The dominant source of freshwater inflow into these systems is the Fraser River, which
flows into the Straits of Georgia. River discharge may be an important factor influencing the
pattern of pH observed at Tatoosh Island. To examine the effect of river discharge on nearshore
pH, river discharge was incorporated into the Wootton model.
A new model was fit to the Wootton dataset using the "June + July" subset of the data, in
which the CC>2 term in the Wootton model was replaced with a term which represents the log of
the Fraser River Discharge. Fraser River discharge data were obtained from Environment
Canada (Fraser River at Hope (08MF005); http://www.wsc.ec.gc.ca/applications/H2O). Low
salinity plumes associated with the Fraser River take about 2 to 4 days to reach the mouth of the
Strait of Juan de Fuca (Hickey et al., 1991). The transport of low salinity water associated with
the Fraser River discharge is modulated by tidal mixing and wind stress, and the plume is most
effectively transported to the coast during neap tides and down strait wind forcing (Hickey et al.,
1991).
For this analysis, models were fit using daily Fraser discharge as well as average monthly
Fraser discharge, both of which produced significant discharge effects. Because of the lag
between Fraser river flow and conditions at the mouth of the Strait, model results are presented
using average monthly Fraser discharge (Table 2). An additional reason for expressing river
discharge as a monthly average is so that this comparison is similar to those incorporating
atmospheric CC>2 levels, which is expressed as monthly average. In this model formulation, all
terms were significant and the r2 = 0.795, which is higher than that for the original Wootton
model using "June + July" data (Table 1). This suggests that a portion of the local dramatic
decline in pH observed at Tatoosh Island in the Wootton dataset may be driven by differences in
Fraser River discharge among the years. During periods of low Fraser River discharge, lower
pH levels occurred. During 2000-2007, there is a significant decline in Fraser River Discharge
in the "June+July" data subset (Monthly Average Fraser Discharge = 237936 - 116 * (Year), p <
0.001, r2= 0.04).
Previous studies have demonstrated that oceanography in the vicinity of Tatoosh Island
can be influenced by interactions between the Fraser and Columbia river plumes (Hickey et al.,
2009). Consequently, river discharge data from the Columbia River at Beaver Army Terminal
near Quincy, Oregon (Station 14246900; http://waterdata.usgs.gov/or/nwis) was incorporated
into the regression analysis. The northward transport of the Columbia River plume has been
estimated to be 35 km d"1 (Hickey et al, 2005), which suggests that it can travel from the mouth
of the Columbia to the entrance of the Strait of Juan de Fuca in about 9 days. Similar to the
model including the Fraser River, Columbia River discharge was incorporated as a monthly
average. In this non-linear regression analyses, numerous models were fit to the June and July
subset of the data (see Table 3). Incorporating river discharge into the non-linear regression
analyses resulted in an increase in variance explained and improvement in the AIC score
(Models 2, 4, and 5 in Table 3). Model 5, which incorporated the Fraser and Columbia River
terms into the original Wootton model, explains the most variance and has the lowest AIC score
-------
(Table 3). The best fit of the model parameters for Model 5 are presented in Table 4.
Comparison of the model parameters for river discharge (qf and qc) suggest that the Fraser and
Columbia River are influencing pH in different manners. Model 5 suggests that pH levels
decrease with decreasing Fraser riverflow; while pH levels decrease with increasing Columbia
riverflow. In addition, comparison of GR for these terms suggests that the Fraser River plays a
larger role in pH levels at Tatoosh Island than the Columbia River, and that the atmospheric CC>2
term is comparable to the Columbia River term. During 2000-2007, there is a significant
increasing trend in Columbia River Discharge in the "June+July" data subset (Monthly Average
Columbia Discharge = -430713 + 218 * (Year), p < 0.0001, r2 = 0.05). Figure 6 shows the
relationship between pH and the river discharge terms and the atmospheric CC>2 term using the
parameters from Model 5 (presented in Table 4).
One statistical requirement of the regression analyses is that each of the variables be
independent. Some of the variables included in the regression analysis of Wootton et al. (2008)
as well as in the models presented in Table 3 are non-independent. Most notable is salinity and
alkalinity because alkalinity is calculated from salinity and temperature. This non-independence
is problematic and hinders the partitioning of the amount of variance explained by each factor.
When calculating the GR2, if one term is removed then the variance associated with that term is
shifted to other terms, which are correlated with the removed variable. This is particularly true
for the variables (Fraser River discharge, Columbia River discharge, and CC^), which exhibit
significant trends during the interval of 2000-2007. Even with these limitations, the analysis in
this section suggests that river flow needs to be included as one of the drivers influencing pH at
Tatoosh Island.
Table 2. Best-fit of parameters for model of pH using model formulation in Wootton et
al. (2008) with CC>2 term replaced with Fraser River discharge term (qf * log(average
monthly Fraser discharge)) and including only "June + July" data.
Parameter
a
qf
h
9
u
c
T
d
k
s
Description
Constant, pH
Fraser discharge effect, pH/(m3 s"1)
Half amplitude of diurnal productivity
oscillation, pH
Phase shift from midnight of diurnal, h
Effect of upwelling,
pH / (metric tons sec'VlOO m coastline)
Phytoplankton abundance effect,
pH/tmgchir1)
Temperature effect, pH/ °C
Pacific Decadal Oscillation, pH/ °C
Estimate alkalinity, pH / (|i mole kg"1)
Salinity effect, pH / psu
Value
-8.76
0.648
-0.122
2.395
0.001
0.305
0.074
-0.117
4.02
-0.012
GR2
-
23.7
43.7
0.51
29.0
40.3
29.1
7.1
0.5
-------
Table 3. Comparison of variance explained for different models that incorporated river
flow and Akaike Information Criterion (AIC), which is a measure of goodness of fit of a
statistical model and provides a metric for model selection. For a set of models, the
preferred model is the one with the minimum AIC value.
Model
1: Original Model (see Table 1)
2: Original Model with CO2 replaced with log(Monthly
Average Fraser)
3 : Original Model with CO2 replaced with log(Monthly
Average Columbia)
4: Original Model with CO2 replaced with log(Monthly
Average Fraser) and log(Monthly Average Columbia)
5: Original Model with the addition of log(Monthly
Average Fraser) and log(Monthly Average Columbia)
terms
R2
0.774
0.795
0.734
0.837
0.862
AIC
-20328
-21438
-18506
-24030
-25978
Table 4. Best-fit of parameters for model of pH (model 5 in Table 3) using the
formulation in Wootton et al (2008) with the addition of two river discharge terms (qf *
log(average monthly Fraser discharge) + qc * log(average monthly Columbia discharge)).
Parameter
a
b
qf
qc
h
9
u
c
T
d
k
s
Description
Constant, pH
Change in pH with atmospheric CO2, pH/ppm
CO2
Fraser River discharge effect, pH/(mJ s"1)
Columbia River discharge effect, pH /(m s" )
Half amplitude of diurnal productivity
oscillation, pH
Phase shift from midnight of diurnal, h
Effect of upwelling,
pH / (metric tons sec'VlOO m coastline)
Phytoplankton abundance effect,
pH/Cmgchir1)
Temperature effect, pH/ °C
Pacific Decadal Oscillation, pH/ °C
Estimate alkalinity, pH / (|i mole kg"1)
Salinity effect, pH / psu
Value
34.59
-13.85
1.414
-0.447
-0.126
2.413
0.003
0.2539
0.068
0.004
1.581
-0.002
GR2
-
15.8
37.8
18.2
55.0
8.2
44.4
45.5
1.8
1.5
1.6
10
-------
Q.
9.0-
8.5-
8.0-
7.5-
pH = 4.289 + 1.092 * log(Fraser1414 /Columbia0447)
r2 = 0.36, p < 0.0001
3.3
3.4
3.5
3.6
3.7
3.8
3.9
log(Fraser1414/ Columbia044')
Q.
9.0-
8.5-
8.0-
7.5-
pH = 50.169+1.232 * (-13.85 * log (CO2))
r2 = 0.21, p< 0.0001
-35.75
-35.70
-35.55
-35.50
-35.65 -35.60
b*log(CQ2)
Figure 6. Relationship between pH and a) sum of the Fraser and Columbia River discharge
terms and the b) CC>2 term in Model 5. Logarithmic identities have been used to transform the
two river discharge terms into one.
In order to be able to utilize the additional years of pH data available (2008-2010),
additional regression models were fit excluding the chlorophyll a term using two different
subsets of data. The first data subset included "June+July" data, while the second data subset
included "June+July+August" data both for the period of 2000-2010. The month of August
could be included in this analysis, because missing chlorophyll a caused most of the gaps in
August data presented in Figure 4. The river discharge terms were included using monthly
average river discharge as well as monthly average discharge with a one month lag. The model
used for the regression analysis was:
pH = a + h * sin(27r * (cp + time of day / 24)) + u * upwelling + T* (water temperature) +
d * PDO + k * log(alkalinity) + s * salinity +T. additional terms
The additional terms included in each model are presented in Table 5. Using the
"June+July" subset, the inclusion of Fraser and Columbia river discharge terms explained more
variance than the inclusion of atmospheric CO2 term. The model with the most variance
explained included river discharge and atmospheric CO2 terms. For the "June+July+August"
11
-------
subset, the model which included river discharge but excluded the CO2 term explained less
variance than the model with CC>2 term. As previously discussed, the models which included
river discharge and CC>2 terms explained the highest amount of variance. Including a one month
lag in the Columbia River term resulted in an increase in variance explained. Similar to the
results presented in Table 4, the Fraser and Columbia River discharge terms had opposing signs,
suggesting that these rivers influence pH in different manners.
Table 5. Model formulation for the "June+July" and "June+July+ August" subsets for
data from 2000 -20 10.
Additional terms included in each model
R2
AIC
"June+July" for 2000-20 10
b*log (CO2)
qf * log(average monthly Fraser discharge)+ qc *
log(average monthly Columbia discharge)
qf * log(average monthly Fraser discharge)+ qc *
log(average monthly Columbia discharge) +b*log(CO2)
0.719
0.817
0.845
-2955
-10729
-13813
"June+July+ August" for 2000-2010
b*log(C02)
qf * log(average monthly Fraser discharge)+ qc*
log(average monthly Columbia discharge)
qf * log(average monthly Fraser discharge)+ qc*
log(average monthly Columbia discharge) +b*log(CO2)
qf * (average monthly Fraser discharge)+ qc * (average
monthly Columbia discharge) +b*log(CO2)
qf * log(average monthly Fraser discharge)+ qf *
log(average monthly Columbia discharge with 1 month lag)
+b*log(CO2)
0.740
0.663
0.802
0.817
0.830
-8082
-1165
-15311
-17448
-19456
Figure 7 shows pH as a function of temperature and salinity for the "June+July+August"
subset of the data from 2000-2010. Low pH levels primarily occur at high salinities and cool
temperatures (> 32 psu and < 11 °C). There are a few observations with low pH at low salinities
but these occurrences are infrequent. Figure 7 also shows characteristic temperature and
salinities for different water masses, which are present in the vicinity of Tatoosh Island (Juan de
Fuca and California Undercurrent from Mackas et al. (1987) and Fraser River plume from
Masson (2006). The Fraser River plume source water has variable temperature and salinity
characteristics that vary seasonally (Masson, 2006). This figure demonstrates that water in the
vicinity of Tatoosh Island is a mixture of these different source waters; however, the peak
salinity values in the Wootton dataset appear to be too high.
Comparison of salinity data from Tatoosh Island and with mooring data (Stations El and
E2) in close proximity (http://clover.ocean.washington.edu/ECOHAB_Data_rep/ECOHABa-
l.htm) showed that at times the salinity at Tatoosh is higher than expected for this region.
Station El is located at the entrance of the Strait of Juan de Fuca approximately 11 km from
Tatoosh Island. Station E3 is located at in the Juan de Fuca Eddy region about 55 km from
Tatoosh Island. Using data from these moorings, Hickey et al. (2009) demonstrated that the
Columbia River plume entered the Strait of Juan de Fuca in 2005 and 2006. During 2003, 2005,
12
-------
and 2006, the peak salinities at Tatoosh were about 0.5 - 1.0 psu higher than salinities at depth
(120 - 250 m depth) at Stations El and E3. Comparison of salinity data from Tatoosh Island and
Station El showed that in late May when river flow was at its peak, the salinity was reduced at
Station El in the upper portion of the water column and at Tatoosh Island (Figure 8). During
these conditions, the salinity at Tatoosh Island was similar to surface values observed at Station
El. During the time of the identified intrusion of Columbia plume water into the Strait of Juan
de Fuca, low salinities were experienced at Tatoosh Island (Figure 8; May 20-30, 2005),
indicating that the Columbia plume influenced salinities in the Tatoosh Island dataset. As the
river flows declined, salinities at Tatoosh Island increased and became more similar to water at
deeper depths. However, at times the salinity at Tatoosh Island exceeded the salinity at depths of
100-200 m, indicating that there is a problem with salinity data from Tatoosh Island. Assuming
that the peak salinities are about 0.4 to 1.0 psu high, Figure 7 suggests that low pH water in the
vicinity of Tatoosh Island is a mixture of Juan de Fuca and California Undercurrent water. This
error in the Tatoosh Island salinity data hinders the identification of interannual variations in
contributions of source waters at Tatoosh Island. Comparison with water temperature data from
Station El shows that the water temperatures measured at Tatoosh Island were similar to the near
surface (1 m depth) temperatures measured in the Strait of Juan de Fuca (Station El; Figure 9).
However, in July and August, at times water temperatures at Tatoosh Island were more similar to
water temperatures at depths of-20-30 m at Station El (Figure 9).
16-
14-
10-
6-
Fraser River Plu
(Fall)
Fraser River Plui
(Summer) *$
Fraser River Plume
(Spring)
California Undercurrent
25 26 27 28 29 30 31 32 33 34 35
Salinity (psu)
Figure 7. pH as a function of temperature and salinity for the "June+July+August" data subset
(data from 2000-2010). Also shown are temperature and salinity characteristics of different
water masses which occur in the vicinity of Tatoosh Island.
13
-------
a
2
I
34-
32-
30-
28-
26-
24
o.
Columbia River Plume Instrusiori
Identified by Mickey et al. (2009)
Tatoosh Salinity
E1 (4 m)
E1 (14m)
E1 (245 m)
E3(120m)
Fraser
Columbia
4/1/2005 5/1/2005 6/1/2005 7/1/2005 8/1/2005 9/1/2005
Date
Figure 8. Time-series of salinity at Tatoosh Island and Stations El and E3 and Fraser and
Columbia River discharge.
16-
15-
14-
2«-i
-------
4. Interactions Among River Discharge, Chlorophyll a, and pH
Biological factors such as water column chlorophyll a can strongly influence pH levels
through photosynthesis and respiration. In the Wootton analysis, they found that there was a
positive relationship between chlorophyll a (using monthly average data from SeaWIFS) and pH
levels, with chlorophyll a explaining about 14% of the variation. Figure 10 shows the
relationship between Fraser River discharge and pH for subsets of observations with low
chlorophyll a levels (< 4 jig I"1) and high chlorophyll a (> 8 jig I"1). When chlorophyll a is low,
there is a relationship between pH and river flow; however, when the chlorophyll a is high, pH
levels are not influenced by river flow. This suggests that there may be interactions among river
flow, chlorophyll a, and pH.
9.6-
Q.
9.4-
9.2-
9.0-
8.8-
8.6-
8.4-
8.2-
8.0-
7.8-
7.6-
Chl a < 4 |jg r
Chi a > 8 |4g r1
0 2000 4000 6000 8000 10000 12000
Fraser River Discharge (m3 s"1)
Figure 10. Relationship between pH and Ratio of river discharge when chlorophyll a levels are
relatively low (< 4 |ig I"1) and high (> 8 |ig I"1). Plot created using data from 2000-2007 and all
months and no averaging performed.
5. Comparison to Long-Term Dataset from Yaquina Estuary, Oregon
If the rapid decline in pH observed in the Wootton dataset is an indicator of ocean
acidification, then this decline should be present at other locations in the region. Trends in pH
observed in the Wootton dataset were compared to those from a long-term dataset collected at
the mouth of the Yaquina Estuary, Oregon, which is located approximately 425 km south of
Tatoosh Island. Previous studies have demonstrated that water quality in the Yaquina Estuary is
closely coupled to near shore upwelling dynamics, indicating that data from this location are
representative of the coastal ocean (Brown and Ozretich, 2009; Brown and Power, 2011). In
addition, the Yaquina Estuary is located in a region of simpler offshore bathymetry than Tatoosh
Island, and is not influenced by large river systems.
15
-------
Although the time period of the Yaquina dataset (2004-2011) is different than in the
Wootton dataset (2000-2010), there is an overlap of about seven years between the 2 datasets.
Additionally, pH was measured using comparable methods to Wootton. A long-term decline in
the flood-tide pH is not evident in the Yaquina Estuary dataset during the period of 2004-2011
(Figure 11). Similar to the Wootton dataset, there is considerable interannual variability between
years in the Yaquina dataset; however, there is no correlation between interannual median pH
levels observed in the Wootton dataset and the Yaquina dataset.
To compare the coupling between pH and coastal upwelling, we examined the
relationships between flood-tide pH and integrated alongshore wind stress (see Brown and
Ozretich (2009) for details on calculation), and Bakun Index (daily and monthly) using data from
2007 as an example. In Yaquina, there is a strong coupling between integrated wind stress and
pH. There is a sigmoidal relationship between flood tide pH and integrated alongshore wind
stress (r2 = 0.62; Figure 12). Although there appears to be a sigmoidal relationship between
Bakun Indices and flood-tide pH, this relationship is not significant for either the daily or
monthly data (not shown). There are significant correlations between daily and monthly Bakun
Index and flood-tide pH at the Yaquina Estuary (Pearson Product Moment Correlation, p<
0.001); however, these indices explain a small amount of the variation in pH (r2 = 0.12 - 0.14;
Figure 13). Consequently, at Yaquina Estuary, integrated alongshore wind stress is a good
predictor of flood-tide pH, while the Bakun Indices are poor predictors.
We compared these patterns to those in the Wootton dataset during the same year. There
were very weak correlations between Bakun Indices (daily and monthly) and pH in the Wootton
dataset (Pearson Product Moment Correlation, r = 0.01 and 0.07, respectively). The coupling
between the north-south wind stress (as well as daily Bakun upwelling index) and flood tide pH
in the Yaquina Estuary is much stronger than that observed in the Wootton dataset (Figures 12
and 13).
16
-------
7.0 -, , , , , , , , , , , , r
5/1/2004 5/1/2005 5/1/2006 5/1/2007 5/1/2008 5/1/2009 5/1/2010 5/1/2011
Date
Figure 11. Interannual patterns in flood-tide pH at Yaquina Estuary collected during May
through September of each year. Linear regression through flood tide pH is presented (y =
94.95 + 4.19 x 10'5 * x, r2 = 0.02, p < 0.0001)
17
-------
8.4-,
8.2-
£ 8.0-
I 7-8:
rr 7.6-
7.4-
7.2-
-VMndStress
pH
5/1/2007
8.4-,
8.2-
8.0-
S" 7.8 -|
7.4-
7.2-
I
6/1/2007
Upwslling Conditions
7/1/2007
Date
. /• •
. •«
-40
-20
I
8/1/2007
\
20
I
3.
20 g.
-20
3.
9/1/2007 »,
^
40
Integrated Alongshore Wnd Stress (m2 s1)
Figure 12. Comparison of flood-tide pH at Yaquina Estuary and integrated alongshore wind
stress. Wind stress (calculated using data from Station 46050 (http://www.ndbc.noaa.gov) with a
lag = 4 days; see Brown and Ozretich, 2009 for details). There is a sigmoidal relationship
between integrated alongshore wind stress and flood-tide pH (r = 0.62).
18
-------
,- -150
5/1/2007
9.0-
8.8-
8.6-
8.4-
8.2-
8.0-
6/1/2007
7/1/2007
8/1/2007
9/1/2007
7.8-
b) Tatoosh
5/1/2007
I
6/1/2007
--100 |
o'
I--50 o
en
ko °-
-100 I
- 150 I
200
7/1/2007
Date
8/1/2007
9/1/2007
Figure 13. Time-series of flood-tide pH and Bakun Index (daily and monthly) at a) Yaquina and
b) Tatoosh Island.
The monthly Bakun index used in the Wootton statistical model is a crude representation
of wind-driven upwelling. Even for a region where there is close coupling between alongshore
wind stress and pH, the monthly Bakun Index does not capture the variability in flood-tide pH
observed at the Yaquina Estuary. In addition, pH levels at Yaquina Estuary appear to be more
closely coupled to upwelling than those at Tatoosh Island, suggesting that that data from the
Yaquina Estuary may be a better representation of ocean conditions.
19
-------
6. Summary
Analyses presented in this report suggest that a portion of the rapid decline in pH
observed at Tatoosh Island may be related to differences in river discharge between the years,
and a component of the decline in pH may reflect localized conditions rather than a large-scale
decline in nearshore pH. The regression analyses in Section 3 shows that Model 2, which
replaces the atmospheric CO2 term with a Fraser River term, explains more variance than the
model formulation in Wootton et al. (2008). This suggests that this regional driver may have
contributed to the rapid decline in pH; and provides an alternate hypothesis. Inclusion of
atmospheric CC>2 and river discharge effects in the Wootton statistical model result in
improvement in the variance explained, suggesting that atmospheric CC>2 may explain a portion
of the decline. Further work is needed in identifying the dominant factor driving the declining
pH at Tatoosh Island due to the covariability of multiple variables in the regression model.
Others have also suggested that the decline in pH observed in the Wootton dataset was related to
local conditions. For example, Feely et al. (2010) suggested that the recent rapid pH decline
observed by Wootton is probably explained by a combination of factors including enhanced
upwelling of waters off the Washington coast resulting from changes in regional ocean
circulation as well as a smaller contribution from ocean acidification.
The analyses in this report suggest that pH levels at Tatoosh Island may be influenced by
the interaction of the Fraser and Columbia River plumes. Previous studies have documented that
water quality conditions on the shelf in the vicinity of Strait of Juan de Fuca are influenced by
the interaction of these river plumes (Hickey et al., 2009). River plumes may influence pH
levels through numerous mechanisms. They modify the physical and biogeochemical
environment by influencing circulation patterns, supply of nutrients, and the formation of
phytoplankton blooms (Hickey et al., 2009). Hickey and Banas (2008) suggested that outflow
from the Strait of Juan de Fuca may be the dominant control on regional nutrient supply and
phytoplankton growth in the California Current System, and outflow from the strait may be more
important to nutrient input in the region than coastal upwelling (Hickey et al., 2009).
The regression analysis in this report suggests that the Fraser River has a stronger
influence on pH levels at Tatoosh Island than the Columbia River and declining Fraser River
flow appears to result in the transport of more low pH water into the vicinity of Tatoosh Island.
It is not surprising that the Fraser River may influence pH levels in the Wootton dataset,
particularly since peak Fraser River discharge typically occur during June and river flow declines
through late summer (Halverson and Pawlowicz, 2008) and there is considerable interannual
variability in Fraser River flow (Li et al., 1999). Masson (1996) demonstrated that the influence
of the Fraser River plume on water masses varies seasonally in response to river flow and during
the spring and summer the contribution of Fraser River plume water reaches 50%-70% near the
entrance of the straits of Juan de Fuca. The majority of the low pH levels occur at relatively high
salinities, suggesting that low pH river plumes are not the cause of the declining pH at Tatoosh
Island.
Fraser River discharge may modulate the amount of upwelling of deep ocean water in the
vicinity of Tatoosh Island through numerous mechanisms, such as through the formation of the
Juan de Fuca eddy and stratification suppressing upwelling. MacFayden et al. (2008) estimated
20
-------
the contribution of various water masses to the Juan de Fuca eddy. They suggested that when the
Fraser River discharge is at its peak and the eddy is less developed, Juan de Fuca source water is
the dominant water mass in the region. Juan de Fuca source water is a mixture of Fraser River
plume water and deep estuarine inflow (composed of California Undercurrent water). As the
Juan de Fuca eddy becomes more developed, the contribution of California Undercurrent source
water increases. Variations in Fraser River discharge may thus be influencing the relative
contribution of these source waters in the vicinity of Tatoosh Island. In addition, the decline of
pH observed in the Wootton dataset from April through September is consistent with the increase
in California Undercurrent source water as the eddy becomes more well developed (MacFayden
et al., 2008). Water properties with the Juan de Fuca Eddy vary seasonally and interannually
(MacFayden et al., 2008).
The regression analysis in Section 3 suggests that the Columbia River explains less
variance in pH and influences pH in a different manner than the Fraser River. The Fraser River
would be expected to have a stronger influence on water properties near Tatoosh Island because
it is discharges into a constrained water body (Straits of Georgia), whereas the Columbia River
discharges into the Pacific Ocean and the transport of its plume water is dependent upon
conditions on the shelf. The Columbia River plume has been found to travel northwards during
the spring and summer and at times the plume enters the Strait of Juan de Fuca (Hickey et al.,
2009); however, the position of the plume is variable depending upon wind conditions. During
the summer, it is located north of the Columbia River mouth approximately 50% of the time and
is typically within 30 km of shore (Hickey et al., 2005); Frequently, the Columbia River plume
travels as far north as La Push, Washington, which is about 55 km south of Tatoosh Island
(Hickey et al., 2005). The regression analysis suggested that low pH levels at Tatoosh were
associated with high Columbia River flows. The Columbia River plume is important at a
regional scale and has been referred to as a "regional bioreactor" which has biogeochemical
impacts on the coastal ocean, including enhanced primary productivity and biomass (Kudela et
al., 2010). Studies of other large river plumes have shown that plume produced organic carbon
is exported to subsurface waters and the degradation of this organic matter results in enhanced
acidification of the subsurface waters (Guo et al., 2012; Cai et al., 2010). We speculate that a
similar mechanism may be occurring on the Washington shelf. The relationships between river
discharge and pH levels at Tatoosh Island may be improved by incorporating time-averaging or
lags in the river discharge, since we would not expect daily variations in river discharge to
immediately influence pH levels at Tatoosh Island. This may be particularly true for the
Columbia River, which may be influencing pH through biogeochemical processes occurring on
the shelf and its transport will be influenced by wind conditions on the shelf.
There are several factors which hinder the analysis of the Wootton dataset, including
issues with the salinity data, limitations in chlorophyll a data, and limitations in Bakun upwelling
index. Peak salinities in the Wooton dataset are 0.5-1 psu higher than mooring data at depths of
100-200 m. Temperature and salinity data can be used to identify source waters, such as the
contribution of Juan de Fuca versus California Undercurrent waters. However, the limitations of
the salinity data in the Wootton dataset prevent such analyses. In addition, salinity errors may
have influenced the results for the salinity term in the regression analysis.
21
-------
The Wootton analysis used monthly-average chlorophyll a data from SeaWIFs. Due to
the retirement of SeaWIFs, chlorophyll a data are not available for 2008 - 2010. In a previous
paper, Pfister et al. (2007) compared daily average chlorophyll a data from satellite observations
to in situ fluorescence readings in a tide pool on Tatoosh Island. Pfister et al. (2007) found that
satellite chlorophyll a estimates were poor predictors of chlorophyll a at Tatoosh Island
(Figure 9). This remotely-sensed chlorophyll a dataset is probably best used to describe large
differences in chlorophyll a patterns, such as sustained phytoplankton blooms in the region and
interannual differences. The pH levels at Tatoosh Island are probably more influenced by
chlorophyll a levels in the close proximity of the island.
The pH dataset from the Yaquina Estuary does not show evidence of a rapid decline in
pH similar to that at Tatoosh Island. Using the dataset from the Yaquina Estuary, it was
demonstrated that the monthly Bakun Index explains a small amount of variation in
nearshore/estuarine pH levels, even for a system in which pH levels are closely coupled to wind-
driven upwelling. Through further analyses, it may be possible to develop a better indicator for
upwelling at Tatoosh Island. The Bakun Index provides a metric for wind-driven upwelling.
However, in the vicinity of Tatoosh Island there are other forms of vertical transport, which
cause upwelling of deep water. Previous studies have suggested that upwelling associated with
canyons and with estuarine outflow from the Strait of Juan de Fuca equal or exceed upwelling
associated with wind-driven upwelling (Hickey and Banas, 2008). The Bakun Index would not
capture these forms of vertical transport.
In conclusion, the best way to examine whether the decline in pH is related to physical
processes may be to use a numerical model of the region (including river discharge and wind
forcing), and compare years with differences in river discharge to see if there are differences in
the relative amount of upwelling in the vicinity of Tatoosh Island or the relative contribution of
different water masses. Previous studies have demonstrated that climate change will influence
the discharge of rivers in the Pacific Northwest. The analyses presented in this report suggest
that climate change might influence nearshore pH levels through alterations in river discharge,
modulating the degree of coupling between the nearshore and upwelling of deep low pH water.
References
Brown, C. A. and R. J. Ozretich. 2009. Coupling between the coastal ocean and Yaquina Bay,
Oregon: Importance of oceanic inputs relative to other nitrogen sources. Estuaries and
Coasts, DOI 10.1007/sl2237-008-9128-6.
Brown, C.A. and J.H. Power. 2011. Historic and recent patterns of dissolved oxygen in the
Yaquina estuary. Importance of anthropogenic activities and ocean conditions.
Estuarine, Coastal and Shelf Science 92, 446-455.
Cai, W.-J. X. Hu, W.-J. Huang, M.C. Murrell, J.C. Lehrter, S.E. Lohrenz, W.-C. Chou, W. Zhai,
J.T. Hollibaugh, Y. Wang, P. Zhao, X. Guo, K. Gundersen, M. Dai and G.-C. Gong.
2011. Acidification of subsurface coastal waters enhanced by eutrophication. Nature
Geoscience. 4: 766-770. doi:10.1038/NGEO1297
Feely, R.A., S.R. Alin, J. Newton, C.L. Sabine, M. Warner, A. Devol, C. Krembs, and C. Maloy.
2010. The combined effects of ocean acidification, mixing, and respiration on pH and
carbonate saturation in an urbanized estuary. Estuarine, Coastal and Shelf Science, 88,
442-449.
22
-------
Foreman, M.G.G., W. Callendar, A. MacFadyen, B.M. Hickey, R.E. Thomson, and E. Di
Lorenzo. 2008. Modeling the generation of the Juan deFuca Eddy. Journal of
Geophysical Research, 113, C03006, doi:10.1029/2006JC004082.
Guo, X. and Others. 2012. Carbon dynamics and community production in the Mississippi River
plume. Limnology and Oceanography, 57(1), 1-17
Gustafson, R.G., W.H. Lenarz, B.B. McCain, C.C. Schmitt, W.S. Grant, T.L. Builder, and R.D.
Methot. 2000. Status review of Pacific hake, Pacific cod, and walleye pollock from
Puget Sound, Washington. U.S. Dept. of Commerce, NOAA Tech. Memo., NMFS-
NWFSC-44, 275 p.
Halverson, MJ. and R. Pawlowicz. 2008. Estuarine forcing of a river plume by river flow and
tides. Journal of Geophysical Research, 113, C09033, doi: 10.1029/2008 JC004844.
Hickey, B.M., and N.S. Banas. 2008. Why is the northern California Current so productive?
Oceanography, 21(4), 90- 107.
Hickey, B.M., S. Geier, N. Kachel, and A. MacFadyen. 2005. A bi-directional river plume: The
Columbia in summer. Continental Shelf 'Research, 25, 1631-1656.
Hickey, B., R. McCabe, S. Geier, E. Dever, and N. Kachel. 2009. Three interacting freshwater
plumes in the northern California Current System, Journal of Geophysical Research, 114,
COOB03, doi: 10.1029/2008JC004907.
Hickey, B.M., R.E. Thomson, H. Yih, and P.H. LeBlond. 1991. Velocity and temperature
fluctuations in a buoyancy-driven current off Vancouver Island. Journal of Geophysical
Research, 96(C6), 10507-10538.
Kudela. R.M., A.R. Horner-Devine, N.S. Banas, B.M. Hickey,T.D. Peterson, R.M. McCabe, EJ.
Lessard, E. Frame, K.W. Bruland, D.A. Jay, J.O. Peterson, W.T. Peterson, P.M. Kosro,
S.L. Palacios, M.C. Lohan, and E.P. Dever. 2010. Multiple trophic levels fueled by
recirculation in the Columbia River plume. Geophysical Research Letters, 37, LI 8607.
Lee, K., L.T. Tonge, F.J. Millero, C.L. Sabine, A.G. Dickson, C. Goyett, G.-H. Park, R.
Wanninkhof, R.A. Feely, and R.M. Key. 2006. Global relationships of total alkalinity
with salinity and temperature in surface waters of the world's oceans. Geophysical
Research Letters, Vol. 33, L19605, doi:10.1029/2006GL027207, 2006.
Li, M., A. Gargett, and K. Denman. 1999. Seasonal and Internannual Variability of Estuarine
Circulation in a Box Model of the Strait of Georgia and Juan de Fuca Strait. Atmosphere-
Ocean 37(1), 1-19
MacFadyen, A., B. M. Hickey, and W. P. Cochlan. 2008. Influences of the Juan de Fuca Eddy
on circulation, nutrients, and phytoplankton production in the northern California Current
System, Journal of Geophysical Research, 113, C08008, doi:10.1029/2007JC004412.
MacFadyen, A., B.M. Hickey, and M.G.G. Foreman. 2005. Transport of surface waters from
the Juan de Fuca eddy region to the Washington Coast. Continental Shelf Research, 25,
2008-2021.
Mackas, D. L., K. L. Denman, and A. F. Bennett. 1987. Least squares multiple tracer analysis of
water mass composition. Journal of Geophysical Research, 92(C3), 2907-2918.
Masson, D. 2006. Seasonal water mass analysis for the Straits of Juan de Fuca and Georgia.
Atmos.-Ocean, 44(1), 1-15.
Wootton, J.T., C.A. Pfister, and J.D. Forester. 2008. Dynamic patterns and ecological impacts
of declining ocean pH in a high-resolution multi-year dataset. Proceedings National
Academy of Sciences of the United States of America, 105(48), 18848-18853.
23
------- |