Deep Waters in British Columbia Mainland Fjords Show Rapid Warming and Deoxygenation From 1951 to 2020

Abstract

Many complex fjord systems cross British Columbia's coastline. A 70 year (1951–2020) time
series analysis of temperature, salinity, and oxygen in four such fjords between ∼54 and 50oN (Douglas
Channel, Rivers Inlet, Knight Inlet and Bute Inlet) shows that changes were greatest in deep waters
between the sill and the bottom. In Rivers, Knight and Bute Inlet, the deep water temperature increased
by 1.2–1.3°C over 70 years, up to two times the global average for open ocean waters at corresponding
depths, while salinity increased by 0.1–0.2, and oxygen decreased by 0.4–0.7 mLL−1. The most northern
inlet, Douglas Channel, showed a temperature increase of 0.8°C from 1951 to 2016, while trends in oxygen
and salinity were not statistically significant. An analysis of Apparent Oxygen Utilization suggests that
the deep waters in Douglas Channel are more readily exchanged with the outer coast than the three other
fjords.

Full text

1. Introduction
Fjords are formed by glacial erosion and retreat, leaving behind a terminal moraine or sill at their mouth
(Farmer & Freeland,1983; Pickard,1961). Fjords are fed by rivers and streams that transport freshwater and
land-derived material properties such as dissolved (Oliver etal.,2017) and particulate organic carbon (St
Pierre etal.,2020). Fjords often serve as a gateway between fresh and marine waters for anadromous fish
species such as salmon and eulachon (see Bianchi etal.,2020).
Recent studies have documented changes to the world's fjords. In Chilean Patagonia, a southward displace-
ment of upwelling-favorable winds by 11km year−1 from 2003 to 2016 (Narvaez etal.,2019) could impact
fjord reoxygenation (Silva & Vargas,2014). In Northeast Greenland, coastal freshening from glacial melt
has constrained deep water renewal of Young Sound-Tyrolerfjord (Boone etal.,2018). In Norway, warming
of North Atlantic water has reduced the rate of deep-water renewal and increased deoxygenation of Mas-
fjorden (Aksnes etal.,2019). In British Columbia, anomalously warm water from the 2014 to 2016 marine
heatwave lingered in deep water of Rivers Inlet through at least 2018 (Jackson etal.,2018). Thus, at global
scales, climate change is impacting fjords.
Rivers Inlet, Knight Inlet, Bute Inlet, and Douglas Channel (Figure1) are fjords on British Columbia's main-
land central coast, with glacial watersheds at their heads. Importantly, each fjord has a different connection
pathway to the open ocean—Douglas Channel is a complex series of inlets with two entrances to the open
ocean via Hecate Strait, Rivers Inlet faces the open ocean via Queen Charlotte Sound, Knight Inlet connects
to the open ocean through Queen Charlotte Strait and the Broughton Archipelago, and Bute Inlet connects
to the open ocean through Juan de Fuca Strait and the Strait of Georgia. This is the first time that a time
series of temperature, salinity, and oxygen dating back to 1951 for multiple locations on the BC coast has
been jointly examined in the context of climate change.
Abstract Many complex fjord systems cross British Columbia's coastline. A 70 year (1951–2020) time
series analysis of temperature, salinity, and oxygen in four such fjords between∼54 and 50oN (Douglas
Channel, Rivers Inlet, Knight Inlet and Bute Inlet) shows that changes were greatest in deep waters
between the sill and the bottom. In Rivers, Knight and Bute Inlet, the deep water temperature increased
by 1.2–1.3°C over 70 years, up to two times the global average for open ocean waters at corresponding
depths, while salinity increased by 0.1–0.2, and oxygen decreased by 0.4–0.7mLL−1. The most northern
inlet, Douglas Channel, showed a temperature increase of 0.8°C from 1951 to 2016, while trends in oxygen
and salinity were not statistically significant. An analysis of Apparent Oxygen Utilization suggests that
the deep waters in Douglas Channel are more readily exchanged with the outer coast than the three other
fjords.
Plain Language Summary Glacially carved fjords draw their deep waters from adjacent
offshore ocean basins through a combination of upwelling, cross-shelf transport, tidal mixing and
estuarine circulation related processes. We studied data spanning sevendecades from four such systems
covering four degrees of latitude along the Canadian Pacific and found that fjord deep waters below sill
depth were warming two times faster than their offshore source waters. Changes in temperature were
accompanied by significant trends of decreased oxygen concentrations and increased salinity. Potential
processes explaining these anomalous responses to climate change are discussed.
JACKSON ET AL.
© 2021. American Geophysical Union.
All Rights Reserved.
Deep Waters in British Columbia Mainland Fjords Show
Rapid Warming and Deoxygenation From 1951 to 2020
Jennifer M. Jackson1 , Laura Bianucci2 , Charles G. Hannah2 , Eddy C. Carmack2 , and
Jessy Barrette1
1Hakai Institute, Victoria, BC, Canada, 2Institute of Ocean Sciences, Fisheries and Oceans Canada, Sidney, BC, Canada
Key Points:
• Temperature and oxygen data from
four fjords (Rivers, Knight, and Bute
Inlets and Douglas Channel) from
1951 to 2020 were examined
• Over 70 years, deep water has
warmed by up to 1.3°C and lost up to
0.7mLL−1 of oxygen, with evidence
of accelerated warming since 2016
• Deep warming and deoxygenation
trends are statistically alike in
at least three of the four fjords,
suggesting similar drivers
Supporting Information:
• Supporting Information S1
Correspondence to:
J. M. Jackson,
jennifer.jackson@hakai.org
Citation:
Jackson, J. M., Bianucci, L., Hannah,
C. G., Carmack, E. C., & Barrette,
J. (2021). Deep waters in british
Columbia mainland fjords show rapid
warming and deoxygenation from
1951 to 2020. Geophysical Research
Letters, 48, e2020GL091094. https://doi.
org/10.1029/2020GL091094
Received 2 OCT 2020
Accepted 8 JAN 2021
10.1029/2020GL091094
RESEARCH LETTER
1 of 9

Geophysical Research Letters
2. Data and Methods
2.1. Data
Temperature, salinity, and oxygen measurements have been collected in our study area since 1951 (Fig-
ureS1) through independent programs by the University of British Columbia, Fisheries and Oceans Cana-
da, and the Hakai Institute (TableS1). We examined one station in each inlet—DC (Douglas Channel), RI2
JACKSON ET AL.
10.1029/2020GL091094
2 of 9
Figure 1. Bathymetric map of British Columbia's central coast, highlighting station DC in Douglas Channel, station RI2 in Rivers Inlet, station KN7 in Knight
Inlet, and station BU4 in Bute Inlet. Bathymetric data starting at offshore 200m isobath (from https://www.ncei.noaa.gov) along the black lines are shown in
inset for each station.

Geophysical Research Letters
(Rivers Inlet), KN7 (Knight Inlet), and BU4 (Bute Inlet), which were selected because they were located
landward of the fjord's entrance sill (∼120km from the entrance sill for DC, ∼20km for RI2, and ∼50km
for KN7 and BU4), and in a deep basin (Figure1). In total, 411 CTD profiles were examined—31 from DC,
162 from RI2, 97 from KN7, and 121 from BU4. Since data were collected using different methods, we follow
Chandler etal.(2017) and interpret the temperature, salinity, and oxygen data to an accuracy of 0.1°C, 0.1
salinity units, and 0.1mLL−1, respectively. The temporal distribution of data collected at each station varied
according to observational program (FigureS1; TableS1). For all data, the seasonal cycle was removed as
reported in the supplementary information.
Trendlines for the different variables at each station were computed using ordinary least squares. The trends
were compared against each other using a z-test, where the z-score (z) was calculated as z=(A-B)/sqrt(SE_
A2+SE_B2) where A was the slope of the first trendline, B was the slope of the second trendline, and SE_ was
the standard error of the slope. The null hypothesis (i.e., the assumption that the two slopes are statistically
the same) was rejected at 95% if z>1.96 or z<−1.96.
2.2. Water Types
We define three water types for the inlets. The first is surface water, which is water lighter than a potential
density of 1,022kgm-3. The second is intermediate water, which starts at the base of surface water (typically
less than 10m) and ends at the sill depth, which is (following Pickard,1961 and Wan etal.,2017) 200m in
Douglas Channel, 137m in Rivers Inlet, 64m in Knight Inlet, and 355m in Bute Inlet. The third is deep
water, which starts at the sill depth and ends at the bottom, such that deep water is 201–370m at station
DC, 137–334m at station RI2, 65–513m at station KN7, and 356–632m at station BU4 (Pickard,1961). The
average temperature, salinity, and oxygen (DO) within each water type were calculated for each profile; we
here focus on deep water because this is where the largest changes were observed.
2.3. Calculation of Temperature Impact on Oxygen Changes
Temperature variability affects oxygen concentrations, so the dissolved oxygen saturation (DOsat) was cal-
culated using the TEOS-10 Matlab toolbox (www.teos-10.org) applied to in situ temperature and salinity
data. Afterward, the seasonal cycle was removed as described in the supplemental information. Following
the regional literature (Baker & Pond,1995; Hodal,2010; Lafond & Pickard,1975; Wan etal.,2017), it was
assumed that deep water was renewed annually in each fjord.
To remove the effect of temperature and solubility from DO changes, apparent oxygen utilization
(AOU=DOsat–DO) was calculated, which represents the change in oxygen since a water parcel was last
in contact with the atmosphere (e.g. Pytkowicz,1971; Broecker & Peng,1982), and assumes the surface
oxygen concentration was DOsat (e.g., Ito etal.,2004). AOU estimates biological processes, with positive
values indicating aerobic remineralization that consumes DO and negative values indicating photosynthe-
sis that produces DO. Mixing and advection can affect DO and DOsat, so impact the value of AOU (Koeve &
Kähler,2016). To distinguish between in situ biological processes and mixing, we follow Pytkowicz(1971)
and plot AOU against salinity for the deep waters of the fjords; a linear relationship indicates that changes
in AOU result from mixing rather than local remineralization/photosynthesis. For example, if two water
masses are mixed in the absence of local biological processes, the final values of both AOU and salinity are
simply a weighted average of the source waters’ AOU and salinities. In an AOU versus S plot, AOUfinal and
Sfinal fall on top of the line that connects water masses 1 and 2 end points. However, if there is in situ rem-
ineralization, the final AOU will change at a different proportion than S and will introduce nonlinearity in
the AOU versus S space.
3. Results
Deep waters in all four fjords revealed statistically significant trends, based on the trendline, from 1951 to
2020 of increasing temperature (Figure2a), resulting in decreased DOsat (Figure3b). In Rivers, Knight and
Bute Inlet, salinity significantly increased (Figure2b) and DO significantly decreased (Figure3a). Only in
Bute Inlet did AOU significantly increase (Figure3c). Table1 shows several metrics, including slopes, p
JACKSON ET AL.
10.1029/2020GL091094
3 of 9

Geophysical Research Letters
values, determination coefficient, and the total change from 1951 to 2020 (based on the trendlines). Deox-
ygenation, salinity, temperature, and DOsat trends are statistically indistinguishable at 95% confidence for
the three southern inlets.
The Douglas Channel time series ended in 2016 and in 2020 at the other inlets. To determine the impact
of the shorter time series, we examined temperature in Rivers Inlet from 1951 to 2016 (Knight and Bute
Inlets had large data gaps through the 1990s) and found that trend lines were almost identical to Douglas
Channel (Figure2a), suggesting similar warming in Rivers Inlet and Douglas Channel from 1951 to 2016.
In addition, the slope of the temperature line was steeper in Rivers Inlet (1951–2020) than Douglas Channel
(1951–2016; Figure2a and Table1), suggesting that warming has accelerated since 2016. We acknowledge
that the large data gaps in Rivers Inlet are a potential source of error; however, z-tests show that the slope
of the lines are statistically similar in all inlets, suggesting that the trend in Rivers Inlet is real. Further,
boxplots of temperature and oxygen at decadal scale were constructed in all four inlets (FiguresS3 and S4)
and they showed that, based on a z-test, the trends calculated from the decadal median temperature and
oxygen were statistically similar to the trends shown in Figures2 and 3 in all inlets, except for DO in Rivers
Inlet (TableS2).
Hypoxic conditions, here taken as oxygen concentrations below 2mLL−1 (e.g., Diaz & Rosenberg,2008),
have been observed in the deep water of Rivers and Bute Inlet but not in Knight Inlet or Douglas Channel.
In Rivers Inlet, hypoxic deep water was observed in March 2009, 2014 to 2017, and 2019 and in April 2014,
2016, and 2019. In Bute Inlet, available data shows hypoxic deep waters in June 1998, July 1999, and March
and July 2018. These two inlets show the overall lowest mean DO concentrations (Figure3a), which com-
bined with the observed trends in deoxygenation, leads to the observed hypoxic conditions.
AOU is positive at all four stations (Figure3c), suggesting that consumption of DO, via biological process-
es, is occurring. The increasing AOU indicates that DO changes are larger than DOsat changes (i.e., DO
JACKSON ET AL.
10.1029/2020GL091094
4 of 9
Figure 2. (a) Temperature and (b) practical salinity in deep water at DC (Douglas Channel), RI2 (Rivers Inlet), KN7 (Knight Inlet) and BU4 (Bute Inlet)
from 1951 to 2020. The seasonal cycle was removed as described in the supplementary material. The dashed lines represent the linear best fit for each station
(statistics are reported in Table1). The dotted orange line represents the linear best fit for station RI2 from 1951 to July 2016, the same length of time as the DC
time series.

Geophysical Research Letters
decreases faster than DOsat; see Table1); these trends are significant in the three southern inlets (at 95%
confidence with p<0.05 in Bute Inlet; Rivers and Knight Inlets would be significant at 90% with p<0.1),
suggesting that decreased oxygen solubility due to warmer waters is not the only driver of deoxygenation in
these three fjords (more detail in Section4.2).
A plot of AOU versus salinity shows different roles of water mass mixing versus in situ remineralization
in the different fjords (Figure4; see details about this kind of plot in Section2.3). The relationship is sig-
nificantly linear in Douglas Channel (r2=0.62, p<<0.01); however, in Rivers and Knight Inlet, a linear
regression explains little variability (r2=0.15 and 0.19, respectively, both with p<<0.01), while it explains
none in Bute Inlet (r2≈0.00, p=0.64). Bute has a stronger relationship between temperature (and thus, DO-
sat) and salinity, whereas DO versus salinity dominates the AOU slope in the other three fjords (FigureS5).
4. Discussion
Multidecadal time series of deep-water properties in three different British Columbia mainland fjords have
shown, despite data gaps, statistically significant increases in temperature and salinity and decreases in
oxygen, while a fourth, Douglas Channel, showed significant trends in temperature only. The fact that the
warming and deoxygenation trends are statistically similar in the three southerly fjords infers that common
processes are driving the changes; we note, however, that differences exist between the northern Douglas
Channel and the three southern fjords.
4.1. Mechanisms for Warming
A recent assessment of 2004–2019 global ocean data by Johnson etal., (2020) found that waters at 150m
(about the depth of source water) are warming on average 0.1ºC per decade. At Ocean Station Papa in the
northeast Pacific Ocean, using data from 1956 to 2011, Freeland(2013) reported a warming trend of 0.05ºC
JACKSON ET AL.
10.1029/2020GL091094
5 of 9
Figure 3. As in Figure4 but for (a) dissolved oxygen (DO), (b) oxygen saturation (DOsat), and (c) apparent oxygen utilization (AOU) in deep water (all in
mLL−1). Horizontal gray line in top panel highlights the hypoxic threshold (2mLL−1).

Geophysical Research Letters
per decade at 200m, which increased to 0.08ºC per decade at 150m when the timeseries was extended to
2018 (Cummins & Ross,2020). Within deep water, we observed a warming rate of 0.12°C per decade in
Douglas Channel, 0.19°C per decade in Rivers Inlet, 0.18°C per decade in Knight Inlet, 0.20°C per decade
in Bute Inlet. This rate is 1–2 times the global average found by Johnson etal.(2020), about 3 times
the warming rate reported by Freeland(2013), and twice the warming rate reported by Cummins and
Ross(2020). Why, then, are the deep waters in these fjords warming at a greater rate than the open
ocean?
Upwelling causes deep-water renewal in all four inlets (Baker &
Pond, 1995; Hodal, 2010; Lafond & Pickard, 1975; Wan et al., 2017).
Upwelled water travels from the NE Pacific to Rivers Inlet along the
1,026kgm−3 isopycnal (Hare etal.,2020) and both offshore and deep Riv-
ers Inlet water were anomalously warm following the 2014–2016 marine
heatwave (Jackson etal., 2018; Scannell etal.,2020), which accelerat-
ed the warming trend after 2016 (Figure2a and Table1). Rivers Inlet
deep water is significantly denser than waters in Bute or Knight Inlet,
yet the fact that some of the warmest deep waters in Bute and Knight
Inlet were observed after 2017 suggests that, similar to Rivers Inlet, the
source waters for Bute and Knight Inletalso warmed. Once deep water
enters a fjord and is trapped behind the sill (Farmer & Freeland,1983), it
is modified only by slow processes such as vertical eddy diffusion (Paw-
lowicz, 2017) or yearly deep-water renewal events. Since the offshore
JACKSON ET AL.
10.1029/2020GL091094
6 of 9
DC RI2 KN7 BU4
Temperature (ºC/decade) Slope=0.12±0.035 Slope=0.19±0.027 Slope=0.18±0.022 Slope=0.20±0.010
p=0.0025 p<0.0001 p<0.0001 p<0.0001
r2=0.30 r2=0.26 r2=0.40 r2=0.83
Δ=0.8ºC Δ=1.3ºC Δ=1.2ºC Δ=1.3ºC
Salinity (salinity units/decade) Slope=0.01±0.026 Slope=0.02±0.008 Slope=0.02±0.009 Slope=0.03±0.003
p=0.7019 p=0.0315 p=0.0364 p<0.0001
r2=0.01 r2=0.04 r2=0.05 r2=0.48
Δ=0.2 Δ=0.1 Δ=0.2
Oxygen (mLL−1/decade) Slope=−0.07±0.047 Slope=−0.11±0.043 Slope=−0.07±0.023 Slope=−0.08±0.019
p=0.1424 p=0.013 p=0.0053 p<0.0001
r2=0.09 r2=0.05 r2=0.13 r2=0.19
Δ=−0.7mLL−1 Δ=−0.4mLL−1 Δ=−0.6mLL−1
Oxygen saturation (mLL−1/decade) Slope=−0.02±0.005 Slope=−0.03±0.004 Slope=−0.03±0.004 Slope=−0.03±0.002
p=0.0005 p<0.0001 p<0.0001 p<0.0001
r2=0.38 r2=0.27 r2=0.39 r2=0.82
Δ=−0.1mLL−1 Δ=−0.2mLL−1 Δ=−0.2mLL−1 Δ=−0.2mLL−1
Apparent oxygen utilization (mLL−1/decade) Slope=0.05±0.047 Slope=0.07±0.042 Slope=0.04±0.000 Slope=0.05±0.018
p=0.2781 p=0.0671 p=0.0734 p=0.0051
r2=0.05 r2=0.03 r2=0.06 r2=0.10
Δ=0.4mLL−1
Note. Changes observed were positive for temperature (warming), salinity (salinification), and AOU and negative for oxygen (deoxygenation). Slopes and their
errors are shown in units of the variable (ºC for temperature, mLL−1 for oxygen) per decade. Statistically significant relationships, here defined as when the
p-value for a linear relationship is less than 0.05, are highlighted in bold. In the latter cases, the total change Δ from 1951 to 2016 or 2020 (based on the trendline)
is given.
Table 1
Regression Slopes (With Standard Error) and Associated Metrics (p-value and r2 Value) for Time Series From 1951 to 2020 at Deep Waters of the Four Stations
Figure 4. Apparent oxygen utilization versus salinity in deep waters of
the four fjord stations (color scales as in previous figures). The dashed
lines represent the linear best fit for each station. Data at BU4 from 2017
onwards is highlighted in black symbols, and a linear fit without these data
is shown by the dotted line for BU4 only.

Geophysical Research Letters
source of upwelled water remained anomalously warm through at least 2018 (Jackson etal.,2018), fjord
deep water continued to be anomalously warm. Furthermore, winter outflow winds cool fjord waters (Mac-
Neil,1974), and winter continental temperatures are warming by more than 2°C per century (BC Ministry
of Environment, 2016), so it is possible that continental warming has meant less cooling of fjord water.
4.2. Mechanisms for Deoxygenation
The DO changes we observe in fjord deep waters are like those observed in the northeast Pacific Ocean. Cum-
mins and Ross(2020) found a deoxygenation trend from 1956 to 2020 of about 0.5mLL−1 (0.09mLL−1per
decade) at 150m in the open ocean Station Papa, which is similar to the deoxygenation we observed in fjord
deep waters (0.07–0.11mLL−1 per decade (Table1). It is important to note that oxygen over the continental
slope of the northeast Pacific has oscillated from low concentrations in the 1950s to high concentrations
in the 1980s and back to low concentrations in the 2010s (Crawford & Peña,2016), and thus these results
should be interpreted within the context of multidecadal oscillations. Waters from about 150m in the north-
east Pacific are the source for deep water in Rivers Inlet (Jackson etal.,2018); thus, it is likely that much
of the deoxygenation observed in deep fjord waters are associated with changes in source waters from the
northeast Pacific. Other processes that can impact oxygen concentrations in fjord deep waters are examined
below.
The comparison of changes in DOsat from 1951 to 2020 (0.2mLL−1 for the three inlets) to those of DO (0.8,
0.5, and 0.6mLL−1 for Rivers, Knight, and Bute Inlet) suggests that temperature-dependent solubility effects
(either locally or at the source waters) represent 28%, 42%, and 39% of the total DO change in each of those
three inlets, respectively. Since the fjord deep water is warming faster than offshore water, it is likely that the
decreased solubility in coastal fjords, which enhances deoxygenation, plays a more important role in these
coastal waters than offshore (e.g. Crawford & Peña,2016).
Warmer waters accelerate aerobic remineralization (i.e., the decomposition of organic matter by oxy-
gen-consuming bacteria), leading to a stronger sink of DO by bacteria (e.g. Brewer & Peltzer,2017; Pomeroy
& Wiebe,2001). The remineralization rate (r(T)) can be estimated if we assume a typical Q10 formulation for
this process (e.g. Peña etal2016; Segschneider & Bendtsen,2013),
rT rQ
TT



0 10
0
10
where Q10 is the rate change for a 10K temperature increase and r0 is the remineralization rate at a reference
temperature T0. If we apply the above equation to two temperatures, T1 and T1+ΔT, we find that Q10ΔT/10
represents the fractional change due to a temperature change ΔT. Typical values of Q10 vary between 1.5 and
3 (e.g., Lomas etal.,2002; López-Urrutia etal.,2006; Laufkötter etal.,2017), with Q10=2 a conservative es-
timate of the temperature sensitivity (Segschneider & Bendtsen,2013). Applying the change in temperature
found in the three southern inlets (1.2°C–1.3°C), we estimate that the effect of warming on remineralization
rates is around 9% (varying between 5% and 15% if we use a Q10 of 1.5 or 3, respectively). This suggests that
increased remineralization under observed warming conditions have had a quantifiable effect on deoxygen-
ation in fjord deep waters.
Increased primary production and rainout of organic matter to the deep water can also enhance the total
amount of remineralization, but unfortunately only Rivers Inlet has a useful chlorophyll timeseries, starting
in 2014. The analysis of a 2014–2018 time series of vertically integrated chlorophyll from 5 to 50m depth
shows a decline in the magnitude of the spring bloom in this station (Jackson etal, under revision). This
suggests increased rainout is not a mechanism for deoxygenation in the deep waters. However, this short
chlorophyll time series prevents us from making a sound conclusion in the context of our longer DO time
series.
Our analysis of AOU versus salinity following Pytkowicz(1971) highlights some differences between Doug-
las Channel and the southern inlets. Specifically, changes in AOU in Douglas Channel are primarily due to
changes in the relative proportion of the source waters; in contrast, local processes are likely more impor-
tant in the three inlets. The larger range of salinity in Douglas Channel suggests that source waters experi-
ence less mixing by the time they reach the DC station compared with the stations in the southern fjords.
JACKSON ET AL.
10.1029/2020GL091094
7 of 9

Geophysical Research Letters
Further processes that require examination include the intensity and frequency of upwelling, and changes
in deep water residence times and ventilation (e.g., Erlandson etal.,2006; Pawlowicz,2017). Understanding
these processes are important areas for future research.
5. Conclusions
An analysis of deep water from three British Columbia mainland fjords (Rivers, Knight and Bute Inlet) has
found that from 1951 to 2020, temperature and salinity increased while oxygen decreased. In these three
fjords, temperature warmed up to two times faster than the global average of waters at similar depths and
these changes can be attributed to warming of offshore waters that are the source for fjord deep water via
upwelling, and possibly continental warming that changed outflow winds. Decreased oxygen in offshore
waters can also explain the lower oxygen in fjord deep water though it is likely that enhanced remineraliza-
tion from warming further reduced oxygen in the fjords.
Douglas Channel was different. Although temperature warmed at the same rate as Rivers Inlet until at
least 2016, oxygen did not decrease at a statistically significant rate. An analysis of AOU versus S suggests
that deep waters in Douglas Channel are more readily exchanged with the outer coast than the three other
fjords.
Data Availability Statement
All data used in this project are managed by CIOOS and can be found at https://catalogue.cioos.ca/en/
harvest/cioos-pacific
References
Aksnes, D. L., Aure, J., Johansen, P.-O., Johnsen, G. H., & Gro Vea Salvanes, A. (2019). Multi-decadal warming of Atlantic water and
associated decline of dissolved oxygen in a deep fjord. Estuarine, Coastal and Shelf Science, 228, 106392. https://doi.org/10.1016/j.
ecss.2019.106392
Baker, P., & Pond, S. (1995). The low-frequency residual circulation in Knight Inlet, British Columbia. Journal of Physical Oceanography,25,
747–763. https://doi.org/10.1175/1520-0485(1995)025<0747:TLFRCI>2.0CO;2
Bianchi, T. S., Arndt, S., Austin, W. E. N., Benn, D. I., Bertrand, S., Cui, X., etal. (2020). Fjords as Aquatic Critical Zones (ACZs). Earth-Sci-
ence Reviews, 203, 103145. https://doi.org/10/1026/j.earscirev.2020.103145
Boone, W., Rysgaard, S., Carlson, D. F., Meire, L., Kirillov, S., Mortensen, J., etal. (2018). Coastal freshening prevents fjord bottom wa-
ter renewal in Northeast Greenland: A mooring study from 2003 to 2015. Geophysical Research Letters, 45, 2726–2733. https://doi.
org/10.1002/2017GL076591
Brewer, P. G., & Peltzer, E. T. (2017). Depth perception: The need to report ocean biogeochemical rates as functions of temperature, not
depth. Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences, 375, 20160319. https://doi.
org/10.1098/rsta.2016.0319
British Columbia. Ministry of Environment (2016). Indicators of climate change for British Columbia, 2016 update, Victoria, BC: eBook.
https://www2.gov.bc.ca
Broecker, W. S., & Peng, T. H. (1982). Tracers in the sea, Lamont-Doherty Earth Observatory. NY, U.S.A: Palisades.
Chandler, P. C., Foreman, M. G. G., Ouellet, M., Mimeault, C., & Wade, J. (2017). Oceanographic and environmental conditions in the Dis-
covery Islands, British Columbia. DFO Canadian Science Advisory Secretariat Research Document, 071(XIII), 51. https://waves-vagues.
dfo-mpo.gc.ca/Library/40653043.pdf
Crawford, W. R., & Peña, M. A. (2016). Decadal trends in oxygen concentration in subsurface waters of the Northeast Pacific Ocean. Atmos-
phere-Ocean, 54(2), 171–192. https://doi.org/10.1080/07055900.2016.1158145
Cummins, P. F., & Ross, T. (2020). Secular trends in water properties at Station P in the northeast Pacific: An updated analysis. Progress in
Oceanography, 186, 102329. https://doi.org/10.1016/j.pocean.2020.102329
Diaz, R. J., & Rosenberg, R. (2008). Spreading dead zones and consequences for marine ecosystems. Science, 321, 5891926–5891929.
https://doi.org/10.1126/science.1156401
Erlandsson, C. P., Stigebrandt, A., & Arneborg, L. (2006). The sensitivity of minimum oxygen concentrations in a fjord to changes in biot-
ic and abiotic external forcing. Limnology and oceanography, 51(1, part 2), 631–638. https://doi.org/10.4319/lo.2006.51.1_part_2.0631
Farmer, D. M., & Freeland, H. J. (1983). The physical oceanography of fjords. Progress in Oceanography, 12(2), 147–219. https://doi.
org/10.1016/0079-6611(83)90004-6
Freeland, H. J. (2013). Evidence of change in the winter mixed layer in the Northeast Pacific Ocean: A problem revisited. Atmosphere-Ocean,
51(1), 126–133. http://dx.doi.org/10.1080/07055900.2012.754330
Hare, A., Evans, W., Pocock, K., Weekes, C., & Giminez, I. (2020). Contrasting marine carbonate systems in two fjords in British Colum-
bia, Canada: Seawater buffering capacity and the response to anthropogenic CO2 invasion. PLoS One, 15(9), e0238432. https://doi.
org/10.1371.journal.pone.0238432
Hodal, M. (2010). Net physical transports, residence times, and new production for Rivers Inlet (MSc thesis, p. 71). BC: University of British
Columbia. https://doi.org/10.14288/1.0053224
Ito, T., Follows, M. J., & Boyle, E. A. (2004). Is AOU a good measure of respiration in the oceans? Geophysical Research Letters, 31(17),
L17305. https://doi.org/10.1029/2004GL020900
JACKSON ET AL.
10.1029/2020GL091094
8 of 9
Acknowledgments
Funding for this research was provided
by the Tula Foundation (J. M. Jackson,
and J. Barrette) and Fisheries and
Oceans Canada (L. Bianucci and C. G.
Hannah). We thank the scientists and
crew from the University of British Co-
lumbia, Fisheries and Oceans Canada,
and the Hakai Institute for collecting
these data. Keith Holmes created Fig-
ure1. We gratefully acknowledge that
this work took place on the traditional
territories of the Gitga'at, Wuikinuxv,
Tlowitsis, Da'naxda'xw Awaetlala, and
Homalco First Nations. This research
is not associated with any known
conflict of interest. We thank the editor
and two reviewers for their construc-
tive comments, which improved this
manuscript.

Geophysical Research Letters
Jackson, J. M., Johnson, G. C., Dosser, H. V., & Ross, T. (2018). Warming from recent marine heatwave lingers in deep British Columbia
fjord. Geophysical Research Letters, 45, 9757–9764. https://doi.org/10.1029/2018GL078971
Johnson, G. C., Lyman, J. M., Boyer, T., Cheng, L., Domingues, C. M., Gilson, J., etal. (2020). Ocean heat content [in “State of the Climate
in 2019”]. Bulletin of the American Meteorological Society, 101(8), S140–S144. https://doi.org/10.1175/BAMS-D-20-0105.1
Koeve, W., & Kähler, P. (2016). Oxygen utilization rate (OUR) underestimates ocean respiration: A model study. Global Biogeochemical
Cycles, 30, 1166–1182. https://doi.org/10.1002/2015GB005354
Lafond, C. A., & Pickard, G. L. (1975). Deepwater exchanges in Bute Inlet, British Columbia. Journal of the Fisheries Research Board of
Canada, 32, 2075–2089. https://doi.org/10.1139/f75-246
Laufkötter, C., John, J. G., Stock, C. A., & Dunne, J. P. (2017). Temperature and oxygen dependence of the remineralization of organic
matter. Global Biogeochemical Cycles, 31(7), 1038–1050. https://doi.org/10.1002/2017GB005643
Lomas, M. W., Glibert, P. M., Shiah, F. K., & Smith, E. M. (2002). Microbial processes and temperature in Chesapeake Bay: Current rela-
tionships and potential impacts of regional warming. Global Change Biology, 8, 51–70. https://doi.org/10,1046/j.1365-2486.2002.00454.x
López-Urrutia, Á., Martin, E. S., Harris, R. P., & Irigoien, X. (2006). Scaling the metabolic balance of the oceans. Proceedings of the National
Academy of Sciences of the United States of America, 103, 8739–8744. https://doi.org/10.1073/pnas.0601137103
McNeil, M. R. (1974). The mid-depth temperature minimum in B.C. Inlets, Vancouver, B.C.: University of British Columbia. Retrieved from
https://ocean.library.ubc.ca/collections/ubctheses/831/items/1.0053207
Narvaez, D. A., Vargas, C. A., Cuevas, L. A., Garcia-Loyola, S. A., Segura, C., Tapia, F. J., & Broitman, B. R. (2019). Dominant scales
of subtidal variability in coastal hydrography of the Northern Chilean Patagonia. Journal of Marine Systems, 193, 59–73. https://doi.
org/10.1016/j.marsys.2018.12.008
Oliver, A. A., Tank, S. E., Giesbrecht, I., Korver, M. C., Floyd, W. C., Sanborn, P., etal. (2017). A global hotspot for dissolved organic car-
bon in hypermaritime watersheds of coastal British Columbia. Biogeosciences, 14, 3743–3762. https://doi.org/10.5194/bg-14-3743-2017
Pawlowicz, R. (2017). Seasonal cycles, hypoxia, and renewal in a coastal fjord (Barkley Sound, British Columbia). Atmosphere-Ocean,
55(4–5), 264–283. https://doi.org/10.1080/07055900.2017.1374240
Peña, M. A., Masson, D., & Callendar, W. (2016). Annual plankton dynamics in a coupled physical–biological model of the Strait of Geor-
gia, British Columbia. Progress in Oceanography, 146, 58–74. https://doi.org/10.1016/j.pocean.2016.06.002
Pickard, G. L. (1961). Oceanographic features of inlets in the British Columbia mainland coast. Journal of the Fisheries Research Board of
Canada, 18(6), 908–999. https://doi.org/10.1139/f61-062
Pomeroy, L. R., & Wiebe, W. J. (2001). Temperature and substrates as interactive limiting factors for marine heterotrophic bacteria. Aquatic
Microbial Ecology, 23, 187–204. https://doi.org/10.3354/ame023187
Pytkowicz, R. M. (1971). On the apparent oxygen utilization and the preformed phosphate in the oceans. Limnology and Oceanography,
16(1), 39–42. https://doi.org/10.4319/lo.1971.16.1.0039
Scannell, H. A., Johnson, G. C., Thompson, L., Lyman, J. M., & Riser, S. C. (2020). Subsurface evolution and persistence of marine heat-
waves in the northeast Pacific. Geophysical Research Letters, 47, e2020GL090548. https://doi.org/10.1029/2020GL090548
Segschneider, J., & Bendtsen, J. (2013). Temperature dependent remineralization in a warming ocean increases surface pCO2 through
changes in marine ecosystem composition. Global Biogeochemical Cycles, 27(4), 1214–1225. https://doi.org/10.1002/2013GB004684
Silva, N., & Vargas, C. A. (2014). Hypoxia in Chilean Patagonian Fjords. Progress in Oceanography, 129, 62–74. https://dx.doi.org/10.1016/j.
poceam/2014.05.016
St. Pierre, K. A., Oliver, A. A., Tank, S. E., Hunt, B. P. V., Giesbrecht, I., Kellogg, C. T. E., etal. (2020). Terrestrial exports of dissolved and
particulate organic carbon affect nearshore ecosystems of the Pacific coastal temperature rainforest. Limnology and Oceanography, 65,
2657–2675. https://doi.org/10.1002/lno.11538
Wan, D., Hannah, C. G., Foreman, M. G. G., & Dosso, S. (2017). Subtidal circulation in a deep-silled fjord: Douglas Channel, British Colum-
bia. Journal of Geophysical Research, 122, 4163–4182. https://doi.org/10.1002/2016JC012022
JACKSON ET AL.
10.1029/2020GL091094
9 of 9