Variability in storm season intensity modulates ocean acidification conditions in the northern Strait of Georgia

Abstract

Large changes in marine CO₂ chemistry manifest in areas with weakly-buffered seawater where ocean acidification (OA) acts in concert with natural CO₂ additions. These settings can exhibit periods of extreme OA in the form of multiple co-occurring stressors, including calcite undersaturation and low pH. Such conditions were observed in the northern Strait of Georgia, on the northeast Pacific coast, where extreme OA spanned a 3-year period. Here, we utilized an 8-year, highly-resolved record of seawater CO₂ partial pressure and total dissolved inorganic carbon to decompose the drivers of this extreme OA. We find that variability in storm season intensity shaped the extent of conservative mixing and biogeochemical drivers such that manifests of extreme OA arise in this setting. Extreme OA manifested during years with weak storm seasons due to direct and indirect biogeochemical factors and the reduced impact of conservative mixing. This sensitivity to the storm season intensity highlights how vulnerable the northern Strait of Georgia is to subtle changes in environmental forcing and provides some predictive capacity for OA conditions over the coming year. These results illustrate that OA is not a “slow burn” process within weakly-buffered settings, but rather invokes periods of intensification with poorly understood biological implications.

Introduction

Steadily increasing atmospheric carbon content, largely resulting from fossil fuel CO₂ emissions¹, is changing marine carbonate chemistry through a process known as ocean acidification (OA)². Over the industrial era, nominally beginning in 1765, the atmosphere has stored 41% of total emissions while the remaining contribution flowed into the terrestrial (31%) and oceanic (26%) reservoirs¹. This anthropogenic carbon addition to the ocean leads to increases in seawater CO₂ partial pressure (pCO₂) and hydrogen ion content ([H⁺]), and decreases in pH, carbonate ion content ([CO₃²]), and calcium carbonate (CaCO₃) mineral saturation states (Ω). Ω is specific to the mineral phase, typically either aragonite or calcite (Ω_arag or Ω_cal), and is the product of [CO₃^2-] and calcium ion content ([Ca²⁺]) divided by the solubility product of the mineral (e.g., Ω_cal = ([CO₃^2-][Ca²⁺])⁄Ksp_cal)); with calcite being the less soluble and more stable form of CaCO₃. These changes, collectively referred to as OA, are impacting vulnerable marine species and ecosystem services^3,4,5, with implications for the resilience of dependent coastal communities and economies^6,7,8. Rates of OA have been reported in many open ocean time series over the last decade^9,10, and show pH declining at -0.01 to -0.02 units decade^-1 but with large potential for signals to be modulated by natural processes acting in concert with OA^11,12.

OA patterns are inherently linked to the buffering state of seawater. Seawater buffering is the sensitivity of the marine CO₂ system to changes in total dissolved inorganic carbon (TCO₂) and alkalinity. For example, increasing TCO₂ through anthropogenic CO₂ addition alters the ratio of TCO₂ and alkalinity in seawater and reduces the ability for seawater to buffer changes in pCO₂, pH, [H⁺], and Ω^13,14. Weakly-buffered seawater is prone to enhanced OA and exists in portions of the ocean interior¹⁵ where TCO₂ is enriched by organic matter remineralization, and in areas where freshwater addition reduces seawater alkalinity such as the poles¹⁶ and in areas along coastal margins¹⁷, including within large estuaries such as the Salish Sea¹⁸. Enhanced OA patterns reported in these areas of weak buffering include subsurface waters with large decadal changes in pCO₂¹⁵, changes in the seasonal pCO₂ amplitude^19,20, and extended periods of elevated OA conditions associated with climate variability²¹. Enhanced OA conditions due to climate variability occur on near decadal time scales, whereas very little information is available on such manifestations occurring on shorter time scales²². Here, we report on enhanced OA conditions spanning multiple years within the northern portion of a sub-basin of the Salish Sea along the Northwest coast of North America known as the Strait of Georgia.

The Salish Sea is an estuarine network of straits, channels, and inlets semi-enclosed from the open Northeast Pacific by Vancouver Island (Fig. 1A). The Strait of Juan de Fuca, along the southern end of Vancouver Island, is the primary pathway for oceanic water from the Northeast Pacific to enter the Salish Sea²³. On the eastern end of the Strait of Juan de Fuca, the Salish Sea extends south into Puget Sound and north through Haro Strait into the Strait of Georgia (Fig. 1B). The Strait of Georgia is the largest water body within the Salish Sea, being ~ 220 km long, ~ 30 km wide, and extending to a maximum depth of 430 m and with a mean depth of 150 m^23,24,25. This body of water consists of southern and northern basins with a 240 m deep connection through Sabine Channel²⁴ (Fig. 1B).

Fig. 1

Bathymetric maps showing the Salish Sea and Northeast Pacific coast (A), the Strait of Georgia (B), and the area surrounding long-term monitoring station QU39 in the northern Strait of Georgia (C). Also shown are: the Fraser River (grey), rain- (tan), snow- (pink), and glacier-dominated (blue) watersheds that drain into the Strait of Georgia (upper left and right panels); arrows denoting intermediate water inflow to the Strait of Georgia (blue arrows), Sabine Channel, QU39 (red circle) and the Environment and Climate Change Canada Sentry Shoal weather buoy (buoy 46,131, green triangle), and a wind rose displaying Environment and Climate Change Canada Sentry Shoal weather buoy hourly data recorded over the duration of this study.

The Strait of Georgia receives the greatest degree of freshwater input relative to the other water bodies within the Salish Sea owing to significant delivery from the Fraser River watershed²⁶, which helps to drive an estuarine circulation pattern oriented toward Haro Strait^24,25,27,28. The fresher surface layer of the Strait of Georgia generally flows south to Haro Strait where the water column is tidally mixed with incoming deep water from the Strait of Juan de Fuca^25,28. An estimated 60% of southward-flowing surface water entering Haro Strait is combined with northward-flowing oceanic water from the Strait of Juan de Fuca and then transported northward as Strait of Georgia intermediate water^24,25. The estuarine return flow of intermediate water is supplied to the Strait of Georgia from Haro Strait year-round²⁸, and transport from there takes roughly 120 days to reach the northernmost portion of the strait²⁴. A secondary source of intermediate water enters from Discovery Passage in the northern Strait of Georgia (Fig. 1B), however this water is largely recirculated from the Strait of Georgia²⁹ and confined to the northeast portion of the basin within the upper 80 m of the water column²⁴. Intermediate water comprises a significant portion of the water column in the northern Strait of Georgia, approximately spanning the 50 to 150 m depth range, and the properties of this water are primarily shaped by the source properties waters entering Haro Strait^28,30.

Despite Haro Strait being the dominant source for intermediate water, there are physical and biological distinctions between the surface layer (< 50 m) of the Strait of Georgia’s northern and southern basins due to the influence of the spring/summer freshet from the Fraser River watershed (Fig. 1B). Model results suggest physical characteristics of the northern basin, such as halocline depth and vertical eddy diffusivity, cluster together and differ from these conditions in the southern basin³¹. Observed phytoplankton biomass and composition patterns also indicate the northern basin is unique from the southern basin³². Lower freshwater input during summer to the northern basin invokes weaker stratification that can be disrupted by summer winds to subsequently increase surface nutrient concentrations that support post-spring bloom increases in diatom abundance³³. Importantly, the patterns of phytoplankton composition observed from spatial surveys of the northern basin were reportedly consistent with time series data from the long-term oceanographic monitoring station QU39^32,33 (Fig. 1C). Considering that variability is expected to be higher in the surface layer and the intermediate water is single-sourced, station QU39 can serve as an important indicator of conditions within the northern Strait of Georgia.

The Strait of Georgia is generally a weakly-buffered body of water due to elevated TCO₂ compared to seawater on the continental shelf with a similar salinity range. Ianson et al³⁰ reported that seawater over the 30–31 salinity range was enriched in TCO₂ by ~ 50 µmol kg^-1 in the northern and southern basins of the Strait of Georgia compared to the Strait of Juan de Fuca, which is more directly connected to the open continental shelf. Evans et al³⁴ also reported TCO₂:alkalinity ratios < 1 spanning the water column at QU39. These low ratios coincided with salinity < 31 and alkalinity < 2100 µmol kg^-1, and similar values to these were not seen below the surface layer at open continental shelf stations surveyed in their study³⁴. These conditions drive the prevalence for aragonite undersaturation (Ω_arag < 1 favouring aragonite dissolution) except within the upper 30 m from spring into autumn when inorganic carbon utilization by phytoplankton is high^30,34,35.

Anthropogenic carbon addition compounds with local TCO₂ enrichment to establish the observed corrosive conditions for aragonite. Estimated anthropogenic carbon content for 2019 in the Strait of Georgia ranged from 35 to 55 µmol kg^{−1 36} and similar values were reported for continental shelf water for 2021¹². Accruing anthropogenic carbon input within this weakly-buffered setting has resulted in aragonite undersaturation emerging in winter surface water beginning around 1950^34,36. While corrosive conditions for aragonite are estimated to have intensified over recent decades³⁴, examinations into the patterns of calcite undersaturation (Ω_cal < 1) have received less attention except within the adjoining fjord Bute Inlet³⁷ and more recently in areas of Puget Sound²². The emergence of extreme OA conditions in the form of calcite undersaturation should be expected given the continued anthropogenic carbon input, and would have broader biological implications than aragonite undersaturation alone. This is because such conditions would impact a wider array of marine life that either utilize aragonite, such as juvenile Pacific oysters⁴, or calcite, such as Dungeness crab³⁸, as well as species that may be sensitive to the co-occurrence of extremely low pH^5,39. Extended periods of extreme OA that deliver multiple modes of potential stressors⁴⁰ could have significant implications for ecosystems and fisheries within the Strait of Georgia.

Here we provide the first report of a multi-year period of calcite undersaturation within the Strait of Georgia using a high temporal resolution 8-year biogeochemical times series from the long-term monitoring station QU39. We use a decomposition analysis on this time series to illustrate that a lack of balance between conservative mixing and biogeochemistry manifests extreme OA conditions within the northern Strait of Georgia. We showcase that these conditions are correlated with variability in storm season intensity, and that they can emerge and abate rapidly from one year to the next. These results highlight the vulnerability of regions characterized by weak buffering to subtle shifts in environmental forcing, provide an analog for what conditions may look like within the northern Strait of Georgia in the near future with continued anthropogenic CO₂ input, and point to the need for an aggressive mitigation strategy that moves away from considering OA as a “slow burn” problem.

Results and discussion

Emergence and abatement of calcite undersaturation

Observations from long-term monitoring station QU39 indicated that freshwater discharge to the Strait of Georgia had a pronounced impact throughout much of the water column. In the surface layer, salinity reached a minimum each year in early summer, coinciding with the spring freshet that is predominantly sourced from the Fraser River watershed (Figs. 1B, 2A,B). Freshwater discharge in autumn has greater relative contributions from the combination of many glacier-, snow-, and rain-dominated watersheds surrounding the Strait of Georgia (2,312 watersheds; Figs. Figs. 1A and 2A; Supplementary Text: Estimating freshwater discharge from watersheds draining into the Strait of Georgia; Supplementary Tables 1—4), and surface salinity was markedly reduced during autumns with particularly high discharge (Fig. 2B). Autumn discharge is more episodic relative to the spring–summer freshet, albeit autumn discharge in both 2016 and 2021 was sustained at high levels for longer periods of time owing to recurrent atmospheric rivers^41,42. Salinity declined each spring within intermediate water, but this seasonal reduction was enhanced following the high autumn discharge in 2016 and 2021 (Fig. 2B). Notably, the 30.5 isohaline deepened below 150 m approximately 3 to 4 months after the start of these wet autumn periods (Fig. 2B). This timing was consistent with the notion that intermediate water is supplied from Haro Strait²⁴, and the degree of change in salinity reflects how heavily influenced intermediate water characteristics are to changes in the surface layer²⁸.

Fig. 2

Strait of Georgia time series from March 18, 2015 to December 19, 2023. (A) Freshwater discharge into the Strait of Georgia (m³ s^-1 × 1000) with total discharge, Fraser River watershed discharge, and the combined discharge from glacier-, snow-, and rain-dominated watersheds (Fig. 1A) shown as black, red, and blue, respectively. (B) Salinity through the water column at QU39 with the 30.5 isohaline highlighted. (C) Ω_cal through the water column at QU39. The black line in (C) denotes Ω_cal < 1, while the magenta line denotes pH_T < 7.52. (D) Wind speed cubed (m³ s^-3) and the cumulative north–south wind stress (N m^-2) beginning each year on September 1 from the Environment and Climate Change Canada Sentry Shoal weather buoy in the northern Strait of Georgia (Fig. 1C). Red and green dots denote the start and end of the storm season, respectively.

The occurrence of calcite undersaturation within intermediate water was a striking feature of the northern Strait of Georgia (Fig. 2C). Corrosive conditions for calcite manifested between late summer and autumn over the 50 to 150 m depth range each year, but from 2018 through 2020 the volume of corrosive water greatly exceeded the mean of the time series (Table 1). The minima in Ω_cal during this period was 0.73 and coincided with a pH_T (pH on the total scale and reported throughout this study relative to in situ conditions) of 7.46. While numerous studies have highlighted corrosive conditions for aragonite in the Strait of Georgia^30,34,35, there has been little focus on the emergence of corrosive conditions for calcite despite the potentially broader ecological significance⁵. Calcite undersaturation is a more typical feature of Northeast Pacific basin water near 350 m⁴³, although has been reported within the adjacent Bute Inlet³⁷ and within areas of Puget Sound²². Observations from Hood Canal in Puget Sound show persistent and seasonally-intensifying calcite undersaturation with most corrosive conditions in July²². Calcite undersaturation was also observed to extend to the main basin of Puget Sound during a “CO₂ storm” in September 2017, and this was attributed to discharge anomalies that occurred earlier in the year²². The time series from the northern Strait of Georgia was similar in that there was seasonality to the manifestation of calcite undersaturation, but differed in that most extreme conditions were not the result of a single event, or “storm”, but spanned multiple years. Expansive and extremely corrosive conditions emerged during the summer of 2018 and lasted into the following winter, and then this seasonal manifestation repeated in 2019 and 2020 before the vertical extent and intensity of corrosive conditions for calcite began to abate (Fig. 2C). The Ω_cal conditions seen over this period were so extreme that pH_T < 7.52, levels known to impact decapod larval survival³⁹, spanned significant portions of the water column for extended periods of time (Fig. 2C and Table 1). The emergence and then abatement of these extreme conditions occurred from one year to the next, deviated from the slow secular trends in Ω_cal and pH_T expected from anthropogenic CO₂ uptake^11,12,36 and instead appeared to coincide with changes in storm season intensity (Fig. 2D and Table 1).

Table 1 Storm season characteristics, annual mean percentage of the water column with Ω_cal < 1 and pH_T < 7.52, and particulate organic carbon (POC; µmol kg^-1) from 5 m integrated over the year from 2016 to 2023.

Wind forcing in the Strait of Georgia is highly channelized³⁴, blowing predominantly from the southeast during the storm season that begins in autumn and lasts into spring (Fig. 1C). Owing to the consistent directionality in the winds, the cumulative sum of the north–south wind stress over the storm season serves to quantify its intensity (Fig. 2D); a storm season with a higher cumulative wind stress (CWS) has a much greater intensity than a storm season with a low CWS. Years with strong storm seasons tend to occur during cool phases in the Pacific Decadal Oscillation and El Niño Southern Oscillation (Supplementary Text: On the relationship between climate indices and cumulative north–south wind stress in the northern Strait of Georgia; Supplementary Figs. 1 and 2; Supplementary Table 5). Notably, years that began with an intense storm season coincided with a lower annual mean percentage of the water column with Ω_cal < 1 (Table 1 and Fig. 3), and tended to have a lower particulate organic carbon (POC) integrated over the year (Table 1), compared to years with weaker storm seasons that manifested high annual mean percentages of the water column corrosive to calcite. This apparent relationship between storm season intensity and the annual mean percentage of the water column exhibiting calcite undersaturation was statistically significant (Fig. 3). The mechanisms linking storm season intensity with the extent of the water column corrosive to calcite in the northern Strait of Georgia was explored through a decomposition analysis of thermodynamic, conservative mixing, and biogeochemical drivers^44,45.

Fig. 3

The relationship between storm season CWS (N m^-2) and the annual mean percentage of the water column with Ω_cal < 1 (circles). The relationship between storm season CWS and Ω_cal < 1 was statistically significant at p = 0.05 and the linear fit and statistics are shown in black.

Characteristics of the storm season include start date, duration, CWS, and total discharge. The mean and standard deviation (STD) for all years is also provided for each term.

Decomposition of Ω _cal seasonal cycle

To illustrate the decomposition analysis, we first applied it to a composite seasonal cycle of Ω_cal (Fig. 4A). The composite of Ω_cal, based on 8 years of data (Fig. 2C) spanning a wide range in inter-annual forcing (Supplementary Fig. 1), showed Ω_cal < 1 first emerging in July over the 50 to 150 m depth range before corrosive conditions for calcite intensified and extended to the seafloor by October. Corrosive conditions for calcite were apparent through January from 100 m to the seafloor before beginning to abate near the start of February. An Ω_cal anomaly (∆Ω_cal; Fig. 4B), computed by subtracting a mean vertical profile from the composite seasonal cycle, was decomposed to evaluate the variance associated with seasonal changes in temperature (Fig4c,D); ∆Ω_cal,T) and salinity (Figs. 4E,F); ∆Ω_cal,S), changes in TCO₂ and alkalinity due to conservative mixing (Fig. 4G; ∆Ω_cal,mix), and biogeochemistry (Fig. 4H; ∆Ω_cal,BGC)⁴⁴. Increased temperature and decreased salinity lead to positive values in ∆Ω_cal,T and ∆Ω_cal,S, respectively, which represent improved Ω_cal conditions albeit the magnitude of these components was small over the range of observed temperature and salinity (Fig. 4D,F). Conservative mixing of TCO₂ and alkalinity had a far more significant impact on ∆Ω_cal, although in a counter-intuitive way. Owing to a steeper slope in the regional TCO₂-salinity relationship relative to the slope of the alkalinity-salinity relationship over the observed salinity range (Supplementary Text: Conservative relationships for alkalinity and TCO₂; Supplementary Fig. 3), a reduction in salinity due to conservative mixing caused a decrease in the TCO₂:alkalinity ratio that drove positive values in ∆Ω_cal,mix (Fig. 4G). Periods of lower salinity in the surface and intermediate layers corresponded with positive ∆Ω_cal,mix and occurred from June to August and November to February, respectively. Conversely, increased salinity resulted in negative ∆Ω_cal,mix values, which represent worsened Ω_cal conditions, and these occurred in the surface layer from September through May and below 150 m from September to March. ∆Ω_cal,mix generally opposed the ∆Ω_cal,BGC term, and the latter exhibited negative values within the intermediate layer beginning in May (Fig. 4H). The negative ∆Ω_cal,BGC signal intensified from July into October, when it also extended to the surface layer until March. Within the intermediate layer, the ∆Ω_cal,BGC term was greater in magnitude and opposite in sign to the ∆Ω_cal,mix term from May to March. Finally, an error term showed that the sum of these components accounts for nearly all of the variability in ∆Ω_cal (Fig. 4I). The large signals in ∆Ω_cal result from shifts in the balance between ∆Ω_cal,mix and ∆Ω_cal,BGC.

Fig. 4

Decomposition of the Ω_cal seasonal cycle in the northern Strait of Georgia. (A) The composite seasonal cycle of Ω_cal. (B) Ω_cal anomaly of the composite seasonal cycle (∆Ω_cal). (C) Annual composite of temperature (°C). (D) Temperature component of ∆Ω_cal (∆Ω_cal,T). (E) Annual composite of salinity. (F) Salinity component of ∆Ω_cal (∆Ω_cal,S). (G) The mixing component of ∆Ω_cal (∆Ω_cal,mix). (H) The biogeochemical component of ∆Ω_cal (∆Ω_cal,BGC). (I) The error term.

Decomposition of interannual Ω _cal variability

Armed with the understanding that the balance between ∆Ω_cal,mix and ∆Ω_cal,BGC shapes ∆Ω_cal in the northern Strait of Georgia, we recomputed ∆Ω_cal by subtracting the composite seasonal cycle (Fig. 4A) from the Ω_cal time series (Fig. 2C), and subsequently decomposed ∆Ω_cal to examine the influence of interannual variability on ∆Ω_cal,mix and ∆Ω_cal,BGC (Fig. 5). Between 2018 and 2020, when extremely corrosive conditions for calcite were observed (Fig. 2C), ∆Ω_cal exhibited negative values throughout much of the water column (Fig. 5A). During this period, ∆Ω_cal,mix was more neutral in the 50 to 150 m depth range than was seen in other years, whereas ∆Ω_cal,BGC was predominantly more negative (Figs. 5B,C). Average ∆Ω_cal,mix and ∆Ω_cal,BGC values over the 50 to 150 m depth range showed the weakened influence of conservative mixing relative to biogeochemistry in shaping ∆Ω_cal during this time (Fig. 5D). Conversely, 2017 exhibited predominantly positive ∆Ω_cal with coincident periods of positive ∆Ω_cal,mix and ∆Ω_cal,BGC. ∆Ω_cal from 2021 through 2023 had an increasingly positive tendency that tracked larger decreases in the magnitude of ∆Ω_cal,BGC compared with ∆Ω_cal,mix. This comparison reveals that years exhibiting extreme Ω_cal conditions within intermediate water occur when ∆Ω_cal,mix and ∆Ω_cal,BGC are out of balance with ∆Ω_cal,BGC being the dominant term.

Fig. 5

Decomposition of the interannual ∆Ω_cal variability in the northern Strait of Georgia. (A) ∆Ω_cal from the full time series. (B) ∆Ω_cal,mix. (C) ∆Ω_cal,BGC. (D) Averages of ∆Ω_cal,mix (blue) and ∆Ω_cal,BGC (red) over the 50 to 150 m depth range.

To understand the drivers for a lack of balance between ∆Ω_cal,mix and ∆Ω_cal,BGC, we return to the strong relationship between the annual mean percent of the water column with Ω_cal < 1 and storm season intensity (Fig. 3) and consider how these decomposed terms may be directly impacted by the storm season. Both of these terms had significant correlations with the storm season intensity; however, this was only the case for ∆Ω_cal,mix after 2023 was excluded (Fig. 6A). Notably, the total freshwater discharge during the storm season also did not correlate with storm season intensity unless 2023 was excluded (Fig. 6B). 2023 was an anomalous year in that the storm season was intense with persistently high wind speeds but freshwater discharge was low (Fig. 2A and Table 1) so an increase in ∆Ω_cal,mix was not seen like during typical stormy years. This deviation did not impact the relationship between the annual mean percent of the water column with Ω_cal < 1 and storm season intensity because ∆Ω_cal,BGC exhibited a large increase during 2023 (Fig. 6C).

Fig. 6

Storm season intensity versus primary components of ∆Ω_cal decomposition and associated factors. (A) Storm season CWS (N m^-2) and annual mean ∆Ω_cal,mix. (B) Storm season CWS (N m^-2) and storm season discharge (km³). (C) Storm season CWS (N m^-2) and annual mean ∆Ω_cal,BGC. (D) Storm season CWS (N m^-2) and the difference between alkalinity and TCO₂ integrated over the upper 50 m over the storm season (Storm season / 50 m integrated Alk-TCO₂; µmol kg^-1).

The relationship between ∆Ω_cal,BGC and storm season intensity was robust for all years, with ∆Ω_cal,BGC becoming increasingly positive during strong storm seasons likely resulting from increased CO₂ outgassing to the atmosphere. During strong storm seasons, frequent periods of high wind speeds combined with surface ocean pCO₂ values that greatly exceed atmospheric levels would lead to large CO₂ fluxes from the Strait of Georgia to the atmosphere⁴⁶. Large outgassing fluxes would act to reduce the surface layer TCO₂ content by reducing the concentration of aqueous CO₂ with no effect on alkalinity. We have parameterized this by integrating the difference between alkalinity and TCO₂ over the upper 50 m, and then averaging these values over the storm season. While the relationship between this parameter and storm season intensity is not statistically significant, the data suggest there is an impact on surface ocean CO₂ content from intense storm seasons (Fig. 6D). Storm season outgassing likely explains the decrease in surface seawater pCO₂ levels observed in high-resolution time series beginning in January that occurs well ahead of the spring bloom^33,34.

The factors related to storm season intensity that shape ∆Ω_cal,mix and ∆Ω_cal,BGC implies that intense storm seasons lead to greater modification of the Strait of Georgia surface layer through higher freshwater discharge that enhances conservative mixing and greater outgassing that ventilates surface layer CO₂ content. This modified surface layer would then enter Haro Strait, where it would be vertically mixed and subsequently refluxed back to the northern Strait of Georgia as a modified intermediate layer. Upon reaching the northern Strait of Georgia, this modified intermediate layer would exhibit ∆Ω_cal,mix and ∆Ω_cal,BGC values that support favorable Ω_cal levels well into the year, as was seen in 2017 (Fig. 2C). Weak storm seasons would typically have less freshwater input and a lower potential for CO₂ ventilation (Fig. 6A,C) leading to ∆Ω_cal,mix and ∆Ω_cal,BGC values that are near neutral or negative, ultimately supporting the manifestation of extremely corrosive conditions for calcite. Freshwater discharge anomalies during strong storm seasons, like seen in 2023, highlight ∆Ω_cal,BGC as the leading term in shaping ∆Ω_cal. Owing to these factors, the intensity of the storm season parameterized using CWS serves as a good predictor of and potential warning for extreme Ω_cal conditions in the coming months (Fig. 3).

Above we discuss the direct impacts from storm season intensity on ∆Ω_cal,mix and ∆Ω_cal,BGC; however, ∆Ω_cal,BGC is subject to additional modification outside of the storm season when organic carbon production in the surface layer and its subsequent supply to intermediate water is high. Notably, weaker storm seasons tend to be shorter in duration and coincide with years that exhibit higher POC loads (Table 1), and a higher supply of POC to intermediate water would increase inorganic carbon content through bacterial remineralization. Conversely, the intense storm season of 2017 coincided with low annually-integrated 5 m POC (Table 1), consistent with an observed late spring bloom³³ and positive annual mean ∆Ω_cal,BGC (Fig. 6C). While no significant relationship existed between annually-integrated POC measured at 5 m depth and storm season intensity, the data do suggest that weak storm seasons generally support higher POC loads (Table 1). Once this material is exported and respired, the TCO₂ produced would exacerbate unfavorable ∆Ω_cal,BGC conditions established over the weak storm season by lower CO₂ ventilation; effectively leading to a 1–2 punch combination that supports the manifestation of extreme OA conditions.

The annually-integrated 5 m POC content can be used to estimate the respiratory TCO₂ signal and showcase that the magnitude of this signal does not need to be large to enhance OA in such a weakly-buffered system. POC in the northern Strait of Georgia originates primarily from net primary production with nearly 26% respired in the water column⁴⁷. Assuming that our 5 m POC measurements generally represent conditions over the upper 10 m of the water column with negligible concentrations below this depth and that the organic to inorganic carbon remineralization ratio is 6⁴⁸, then we estimate a range of respiratory TCO₂ input between 35 and 67 µmol kg^-1. This closely brackets the current anthropogenic CO₂ load and is consistent with the amplitude of TCO₂ variability within the intermediate layer (2047–2102 µmol kg^-1 at 100 m). Notably, the difference in respiratory TCO₂ signal between years, such as between 2017 and 2018 with the lowest and highest POC content, respectively (Table 1), was only about 20 µmol kg^-1. This difference in carbon input is ~ 40% of the contemporary anthropogenic CO₂ load and would accrue over the next 23 years based on current rates of addition¹², over which time the extreme Ω_cal conditions observed now may become the norm. This simple calculation combined with the results from the decomposition highlight that: (1) the northern Strait of Georgia is highly vulnerable to very subtle changes in inorganic carbon content, (2) direct and indirect effects from weak storm seasons can compound with the current anthropogenic CO₂ load to manifest extremely corrosive conditions for Ω_cal, and (3) the conditions observed between 2018 and 2020 may serve as an analog for the near-future state of Strait of Georgia intermediate water.

Our assessment here highlights that, while anthropogenic CO₂ uptake by the ocean is a slow process¹², when this addition is combined with natural processes acting within weakly-buffered environments, such as the Strait of Georgia, rapid transitions to extremely corrosive conditions can occur. OA is not a slowly emerging process within these settings like it may have been in the past under lower anthropogenic CO₂ conditions. We show that extreme OA can emerge quickly and last for years with currently unknown biological impacts. Variability in storm season intensity played a dominant role in manifesting these conditions in the northern Strait of Georgia, and this result provides predictive ability to estimate the onset of extreme OA over the coming months. This relationship between storm season intensity and extreme OA conditions would have been very difficult to resolve with a lower-resolution and shorter time series. Support must be maintained for such highly-resolved long-term marine CO₂ system time series as well as enhanced to develop similar time series stations in other coastal regions. Future work should evaluate the implications of storm season variability more broadly within the Strait of Georgia from a modeling perspective, as well as utilize the predictive relationship with storm season CWS to interrogate the ecosystem for biological responses that could elucidate how marine life will fare under perennially corrosive conditions for calcite in the near future.

Methods

Observations at oceanographic station QU39

Oceanographic surveys to station QU39 took place every week for conductivity-temperature-depth (CTD) and oxygen sensor profiling and every two weeks for discrete sample collections using Niskin bottles. The CTD data handling and quality control procedures are described elsewhere (https://dmphub.uc3prd.cdlib.net/dmps/https://doi.org/10.48321/D129FB2E20). Discrete samples were collected for determining the marine CO₂ system from 12 depths spanning the water column (nominally 0, 5, 10, 20, 30, 40, 50, 75, 100, 150, 200 and 250 m) using 350 mL amber soda-lime glass bottles and filled with care to avoid introducing bubbles into the sample. Seawater samples for marine CO₂ system determination were preserved immediately after collection with 80 µL of saturated mercuric chloride and stored in the dark at room temperature until analysis. Seawater samples for POC were drawn from a Niskin bottle into 1 L sample bottles and processed as described below.

Analytical methods

Seawater samples were analyzed for TCO₂ and pCO₂ generally within 6 months of collection using a Burke-o-Lator TCO₂/pCO₂ analyzer at the Hakai Institute’s Quadra Island Ecological Observatory. The TCO₂ measurement was conducted with a non-dispersive infrared gas analyzer (LI-COR LI840A) following seawater acidification and flow-balanced gas stripping, and the pCO₂ measurement as achieved by headspace gas equilibration. Detailed protocols for sample handling, processing and calibration are provided by Campbell et al⁴⁹. Description of the data quality control and uncertainty assessment are provided in the Supplementary Information (Supplementary Text: TCO₂ and pCO₂ measurement quality control and uncertainty assessment; Supplementary Figs. 4—7; Supplementary Table 6).

Seawater samples for POC from 5 m depth was measured by filtering 1000–2000 mL of seawater through pre-combusted (4 h at 475 °C) 25 mm Whatman glass fiber filters (GF/F 0.7 nominal pore size μm). To remove particulate inorganic carbon, the filters were then acidified with 2.5 mL 1 M HCl and after 30 s were rinsed with ~ 5 mL filtered seawater. The filters were then stored at -20 ˚C until they were dried at 60 ˚C for 24 h, weighed and sealed in tin capsules for shipment and analysis at the University of California Davis Stable Isotope Facility. Analysis was performed using either a Vario EL Cube or Micro Cube elemental analyzer. POC concentrations were provided in μg L^-1 and converted to μmol kg^-1 utilizing seawater density from CTD measurements at the sample collection depth.

Calculation of annually-integrated POC

To calculate annually-integrated POC, data from 2015 through 2018, which were collected weekly, were downscaled to bi-weekly values to match the following years of the timeseries. This was done by applying a linear fit to the data using a 14-day window, with the derived values over the large spring 2020 Covid-19 gap removed from the final dataset. Annually-integrated POC was determined through a cumulative sum of the bi-weekly concentrations over each calendar year. Data from 2023 was not available due to issues with sample analysis.

Definition of storm season intensity

We used hourly wind speed and direction measurements taken at the Environment and Climate Change Canada Sentry Shoal weather buoy to estimate the intensity of the storm season. The cube of the wind speed, representing the energy input to the surface ocean, and wind stress were computed from the hourly measurements, with the latter utilizing the formulation for the drag coefficient from Smith et al⁵⁰. Cumulative wind stress was computed from hourly wind stress values over a 365-day period starting on September 1^st of each year so as to highlight the storm season. The start of storm season was defined as the first day after September 1 that experienced near gale force wind speeds (13.9 m s^-1 according to the Beaufort Wind Scale, red dots in Fig. 2). The storm season termini were defined as the inflection point in the cumulative wind stress curves that indicates a change toward predominantly southward winds (green dots in Fig. 2). Storm season intensity was quantified using the cumulative wind stress between the beginning and end of the storm season (Table 1).

Estimation of daily discharge

We estimated daily discharge (2015 through 2023) from all land in the drainage area (264,839 km²; Fig. 1), with separate estimates for the Fraser River (232,137 km², 88% of area) and for small coastal watersheds (32,703 km², 12% of area). Small coastal watershed estimates were further broken down into glacierized (10,752 km², 4.1% of area), snow mountain (13,979 km², 5.3% of area), and rain (7,971 km², 3.0% of area) type watersheds, which are known to support contrasting streamflow regimes in this region⁵¹. We compiled discharge data from all of the available near-outlet gauges in the Water Survey of Canada (WCS) database (HYDAT) using a combination of the tidyhydat package in R⁵² and the WSC Real-time Web Service (https://wateroffice.ec.gc.ca/services/real_time_links_e.html). However, most of the 2,312 small coastal watersheds were ungauged and all gauges were located some distance upstream of the watershed outlet. As such, prediction procedures were required to account for ungauged areas (Supplementary Text: Estimating freshwater discharge from watersheds draining into the Strait of Georgia).

For gauged watersheds, we estimated discharge from ungauged areas (downstream of gauges) by simple area-based scaling using the most representative gauge (Supplementary Table 1). The ungauged areas ranged from 0.6% to 15% of the total watershed. For watersheds with multiple near-outlet gauges, we summed the discharge from all available gauges (Supplementary Table 1).

For ungauged watersheds, we predicted the total discharge from each of six watershed types using the mean specific discharge from representative gauged watersheds of the same type or the most similar watershed type with a gauge (Supplementary Table 2). Watershed types and quantitative watershed characteristics (e.g., glacier cover, mean annual precipitation, precipitation as snow) were taken from Giesbrecht et al.^51,53. Watersheds with highly regulated flow (Campbell) and atypical flow regimes (Squamish) were omitted from the training set because they are not representative of most watersheds. Three gauges had large data gaps that were filled before scaling and summing for the region (Supplementary Table 3). All four rain type watersheds were summed together, given the limited extent of three of these watershed types (Supplementary Table 4).

Definition of composite annual cycles

Composite annual cycles were computed for temperature and salinity measured by the CTD and from observed and derived (Ω_cal) marine CO₂ system variables determined from discrete seawater samples. Data from these two sources were handled in largely the same way, except for slight differences in the gap-filling and bin-averaging procedures. Annual composites were produced by first plotting parameters as a function of day-of-the-year and depth, and then filling the data gaps present at the start and end of the year such that a composite could be produced that spanned the entire year (Supplementary Text: Definition of composite annual cycles; Supplementary Fig. 8). The gap at the start of the record was filled by duplicating the last two days of data and moving these duplicated data to the start of the record. The gap at the end of the record was larger, and filled by duplicating the first 15 days of data and moving these duplicated data to the end of the record. For Ω_cal, 20 days of data were duplicated because of the greater sparsity in discrete sample profiles compared to CTD profiles (237 discrete sample vs 554 CTD profiles). This approach was justified because the degree to which variables differed between mid-December and early January was minimal. Using the gap-filled record, data were subsequently bin-averaged using a 40-day averaging window with a 1-day time step. For temperature and salinity from the CTD profiles, data within the 40-day averaging window were averaged vertically at 1-m resolution. For Ω_cal, data within the averaging window were averaged vertically according to the Niskin bottle target depth. These composite values were depth- and time-matched with the observed values, and subsequently used to compute anomalies as the difference between observed and composite values.

Decomposition of calcite saturation state variability

We use a linear Taylor series decomposition⁴⁴ to examine how changes in temperature, salinity, and marine CO₂ system variables drive variability in Ω_cal. An Ω_cal anomaly, ∆Ω_cal, was computed for every Ω_cal value:

$${\Delta\Omega }_{\text{cal}}={\Omega }_{\text{cal}}-{\Omega }_{\text{cal},0},$$

(1)

where ${\Omega }_{\text{cal},0}$ was a reference value. For the decomposition of the seasonal cycle, ${\Omega }_{\text{cal}}$ was the seasonal composite while the reference value, ${\Omega }_{\text{cal},0},$ was the mean vertical profile computed from the seasonal composite. Subtracting a mean vertical profile from the seasonal composite produced an anomaly that preserved the seasonal differences. For the decomposition of the interannual ${\Omega }_{\text{cal}}$ variability, ${\Omega }_{\text{cal}}$ was the observed time series and ${\Omega }_{\text{cal},0}$ was the seasonal composite. For this case, seasonality was removed and the remaining variability resulted largely from interannual variability. ${\Delta\Omega }_{\text{cal}}$ is a result of the sum of the perturbation effects from changes in temperature, salinity, marine CO₂ system variables, specifically TCO₂ and alkalinity, with the addition of an error term that represents unidentified processes and nonlinearities in the marine CO₂ system⁴⁴:

$$\Delta \Omega_{cal} = = \Delta \Omega_{cal,T} + \Delta \Omega_{cal,S} + \Delta \Omega_{cal,TCO2} + \Delta \Omega_{cal,Alk} + \varepsilon$$

(2)

Following the approach of Rheuban et al⁴⁴, changes in TCO₂ and alkalinity are both shaped by conservative mixing and biogeochemical processes such that these terms can be written as:

$$\Delta \Omega_{cal} = \Delta \Omega_{cal,T} + \Delta \Omega_{cal,S} + \Delta \Omega_{cal,mix} + \Delta \Omega_{cal,BGC} + \varepsilon$$

(3)

where ${\Delta\Omega }_{\text{cal},\text{T}}$, ${\Delta\Omega }_{\text{cal},\text{S}}$, ${\Delta\Omega }_{\text{cal},\text{mix}}$, and ${\Delta\Omega }_{\text{cal},\text{BGC}}$ were perturbation terms for ${\Delta\Omega }_{\text{cal}}$ from variations in temperature, salinity, conservative mixing and biogeochemical processes, respectively, and $\varepsilon$ is the error term. ${\Delta\Omega }_{\text{cal},\text{T}}$ was computed using reference values and observed temperature:

$${\Delta\Omega }_{\text{cal},\text{T}}=f\left(\text{T},{\text{S}}_{0},{\text{TCO}}_{\text{2,0}},{\text{Alk}}_{0}\right)-{\Omega }_{\text{cal},0},$$

(4)

where ${\text{S}}_{0}$, ${\text{TCO}}_{\text{2,0}}$, and ${\text{Alk}}_{0}$ were the reference values. Similarly, ${\Delta\Omega }_{\text{cal},\text{S}}$ was computed using observed salinity:

$$\Delta {\Omega }_{{{\text{cal}},{\text{S}}}} = f\left( {{\text{T}}_{0} ,{\text{S}},{\text{TCO}}_{2,0} ,{\text{Alk}}_{0} } \right) - {\Omega }_{{{\text{cal}},0}}$$

(5)

Conservative mixing of alkalinity in the northern Strait of Georgia was highly linear across a broad salinity range, whereas TCO₂ appeared to have two distinct relationships with salinity (Supplementary Text: Conservative relationships for alkalinity and TCO₂; Supplementary Fig. 3). For the salinity range above 24 that encompassed > 98% of the data from QU39, the slope of the TCO₂-salinity curve was ~ 1.5 × that of the curve for the < 24 salinity range. The TCO₂-salinity curve for the salinity < 24 data was more similar to the alkalinity-salinity curve. These relationships were used to determine the conservative mixing term:

$$\Delta {\Omega }_{{{\text{cal}},{\text{mix}}}} = f\left( {{\text{T}}_{0} ,{\text{S}}_{0} ,{\text{TCO}}_{{2,{\text{S}}}} ,{\text{Alk}}_{{\text{S}}} } \right) - {\Omega }_{{{\text{cal}},0}}$$

(6)

where ${\text{TCO}}_{2,\text{S}}$ was defined using the salinity > 24 relationship. The biogeochemical term representing a number of processes that shape the marine CO₂ system, most important of which on sub-seasonal timescales being organic matter production and respiration, was estimated as a difference from ${\Omega }_{\text{cal}}$ calculated with the conservative mixing relationships:

$$\Delta \Omega_{cal,BGC} = f\left( {T_{0} ,S,TCO_{2} ,Alk} \right) - f\left( {T_{0} ,S,TCO_{2,S} ,Alk_{S} } \right)$$

(7)

$\varepsilon$ is calculated by summing the above terms and subtracting the result from ${\Delta\Omega }_{\text{cal}}$:

$$\varepsilon = \Delta {\Omega }_{{{\text{cal}}}} - \left( {\Delta {\Omega }_{{{\text{cal}},{\text{T}}}} + \Delta {\Omega }_{{{\text{cal}},{\text{S}}}} + \Delta {\Omega }_{{{\text{cal}},{\text{mix}}}} + \Delta {\Omega }_{{{\text{cal}},{\text{BGC}}}} } \right)$$

(8)