Introduction

Small mountainous rivers (SMRs) have been identified for their outsized contribution to worldwide sediment delivery to the ocean (Milliman and Syvitski 1992). For U.S. west coast SMRs that drain to the coastal ocean, consistent river–ocean coherence patterns prevail during winter storms (Kniskern et al. 2011). Because of storm tracks and basin geometry considerations, waves associated with flood-generating storms typically precede river floods by 10–20 h for high-latitude systems and generally persist throughout the duration of the floods. The concept of an oceanic flood (Wheatcroft et al. 1997; Wheatcroft 2000), one in which the river and ocean both respond to the same storm system, describes this behavior and reflects a relatively narrow range of conditions under which sediment is delivered to the coastal ocean (Wheatcroft et al. 2010). In protected embayments without long-lived oceanic swell waves, the persistence and time scales of the oceanic flood change, and water and suspended sediment that might have flooded into an energetic shelf environment may instead empty into a quiescent basin, if the storm system has passed and local wind waves have dissipated. Understanding the timing of delivery, dynamics of transport, and location of deposition of fine-grained sediment is of particular interest because of the habitat-forming role of sediment and the preferential adsorption of contaminants to fine sediment.

In this paper we analyze data collected in three oceanographic-instrument deployments between July 2019 and April 2021 within Bellingham Bay, Washington, USA. We describe the tidal and seasonal hydrodynamics and sediment dynamics of the system to characterize the interplay between fluvial and wave forcing, along with tidal and residual currents, all of which have implications for sediment and contaminant remobilization and transport. Textural, geochemical, and bathymetric observations of the seabed are used to place the time-series observations into broader spatial and temporal contexts showing fine-sediment transport and retention in Bellingham Bay and progradation of the sandy delta. The results are compared to the dynamics of other SMRs throughout the region.

Study Area

Bellingham Bay is a low-mesotidal (mean range 1.7 m, great diurnal range 2.6 m) embayment in the Salish Sea adjacent to the city of Bellingham, Washington, USA (Fig. 1). It is the receiving basin of the Nooksack River, which drains an area of 2050 km\(^2\), and whose origins are the glaciers of Mount Baker and other mountains in the Cascade Range. The mean annual discharge of the Nooksack River is 91 m\(^3\) s\(^{-1}\) and is greatest during storms in November–January, with a smaller peak during the spring freshet in May and June; flows are smallest during August and September. Additional freshwater input comes from Squalicum and Whatcom Creeks, which enter the bay near downtown Bellingham. Their combined discharge is less than ten percent that of the Nooksack River. The annual Noosack River sediment load is 0.86–1.4 Mt y\(^{-1}\) (Czuba et al. 2011; Anderson et al. 2019), making it the second-largest fluvial source of sediment to the Salish Sea south of the Canadian border, and one of the highest-yield rivers given its relatively small watershed. Sediment discharge is greatest during early winter storms and is comparatively smaller during the spring freshet. At Ferndale, Washington (Fig. 1), Nooksack River suspended sediment is composed of approximately 58% material finer than 62.5 \(\mu\)m (Curran et al. 2018).

Fig. 1
figure 1

Map of Bellingham Bay, moorings BEL1, 3, 4, and 5, Nooksack River at Ferndale gauge, and wind station KBLI. Black rectangle in inset shows study area within Washington State; shaded area indicates Nooksack River watershed. Bathymetry from National Oceanic and Atmospheric Administration (2003)

Full size image

Over the Holocene, the Nooksack River has occupied two main distributaries: the modern, minor Lummi River, which enters Lummi Bay and today occasionally carries flood-stage Nooksack River discharge, and the present (since about 1860) distributary system at the northern end of Bellingham Bay (Bortleson et al. 1980). Over the past 160 years, the river has expanded its extensive river-dominated (Galloway 1975) subaerial delta with several distributary channels and corresponding subaqueous delta. The subaqueous delta is about 6 km in width and extends about 2 km longitudinally from the edge of the subaerial delta to the delta topset–foreset rollover. From 1888–1980, the subaerial delta prograded about 1.5 km (Bortleson et al. 1980), representing a rate of 16 m y\(^{-1}\) over that period. Within the main portion of Bellingham Bay, north of Chuckanut Bay, depths are less than 30 m below mean sea level. Early studies (Sternberg 1967) separated Bellingham Bay bed sediments into three major classes: delta platform sands, bay muds, and lag gravels. The well-sorted medium sands are confined primarily to the subaqueous river delta. Homogeneous poorly sorted clayey silt bay muds are seaward of the delta platform, covering the majority of the bay south to and beyond Chuckanut Bay. Lag gravels are confined to regions adjacent to Hale Passage. More recent surveys (Takesue 2020) have largely confirmed the earlier findings.

Bellingham Bay is influenced by southerly (i.e., northward) winds throughout the year. In the winter months of November–February, wind speeds are stronger overall and winds from the north and north–northwest are common in addition to those from the south and south–southeast. Winds produce local wind waves, which are the main source of waves and resultant stress on the seabed in the absence of oceanic swell in this protected embayment.

The region immediately surrounding Bellingham Bay has been impacted by human activity for centuries and has been largely urbanized for over 100 years. In one recent study, all bottom sediment samples collected within the bay were impacted by chemical contamination or other environmental stressors, and nearly two-thirds had some degree of toxicity (Weakland et al. 2013). Low but detectable concentrations of pharmaceuticals and personal care products and perfluoroalkyl substances have been observed in bottom sediments throughout Bellingham Bay, with greatest values near the urban waterfront along the bay’s eastern shore (Long et al. 2013). Benthic communities are also adversely impacted by a variety of stressors across the bay (Weakland et al. 2013). In recent decades, the bay has experienced a dramatic drop in species richness of benthic foraminifera and other taxa, suggesting a deterioration in benthic habitats; the driving cause may be eutrophication and acidification in bottom waters (Weakland et al. 2013; Nesbitt et al. 2015). Historical contamination decreased with the closure of pulp-paper mills and other industries on the Bellingham waterfront; however, contaminants of concern persist in bay sediment at shallow depths (Fawley 2021). In other Salish Sea rivers like the Fraser and Puyallup Rivers, dispersal of fluvial sediment plumes can transport urban and agricultural contaminants to distal locations and ecosystems (Yunker and Macdonald 2003; Takesue et al. 2017).

A connection between sediment contamination and oceanographic processes within Bellingham Bay was suggested by Bothner et al. (1980). That study observed more rapid loss of mercury in bottom sediments than would be expected by pore water exchange with the overlying water column alone (Officer and Lynch 1989); the loss was assumed to be associated with sediment resuspension from bottom currents. In this context, understanding the physical and geological oceanographic processes that lead to sediment transport is crucial to determining the fate of contaminants they disperse.

Methods

Time-series Observations

We deployed bottom-mounted moorings at four locations (BEL1, 3, 4, and 5; Fig. 1) within Bellingham Bay during three periods: 30 July 2019–8 September 2019, 19 November 2019–5 February 2020, and 22 January 2021–13 April 2021 (Nowacki et al. 2023). The moorings were equipped with velocity, wave, and water-quality sensors. Nortek Aquadopp acoustic Doppler current profilers (ADCPs) collected 2-min averages of water-column current profiles every 15 min, generally at 0.25 m bin spacing (0.5 m during the second BEL1 deployment and 0.4 m at BEL5). The ADCPs also collected 4.3 min of wave-pressure bursts at 2 Hz every 30 min. YSI EXO multi-parameter sondes collected turbidity and other water-quality parameters 0.15 m above the seabed every 15 min. During the 2021 deployment, we additionally deployed floating near-surface Seapoint turbidity sensors at sites BEL3 and BEL5. Turbidity was converted to SSC via a robust repeated-median regression (Siegel 1982). Additional details on the time-series data collection and processing are available in Nowacki et al. (2023). Three of the moorings (BEL3, 4, and 5) were deployed in 5–7 m water on the outer topset of the Nooksack River subaqueous delta. BEL1 was deployed in 4–7 m depth near the southeastern extent of Bellingham Bay, just north of Chuckanut Bay.

River discharge and turbidity were obtained from USGS gauge 12213100, Nooksack River at Ferndale (U.S. Geological Survey 2023), approximately 10 river km from the mouth; there are no appreciable water or sediment inputs between the gauging site and the mouth. Flow-frequency analysis and corresponding recurrence intervals were computed using PeakFQ software (Veilleux et al. 2013). Measured turbidity at Ferndale was converted to SSC following Curran et al. (2018). Wind speed and direction were obtained from Bellingham International Airport (KBLI), 2.5 km north of the bay. In Sect. 4.1 we consider 20-year records (2012–2021) of Ferndale discharge and KBLI wind speed; both parameters had data coverage in excess of 99% for this period. Ferndale turbidity, which has a shorter period of record and covers about 8 years (41%) of the 20-year window, is also considered in that analysis. As used in this manuscript, floods are defined as discharge events exceeding the 90th percentile (approximately 200 m s\(^{-1}\)) for at least 12 h. The flood-wave propagation time between Ferndale and the river mouth was estimated using the discharge–velocity relationship obtained from data available on the USGS NWIS system. Values typically ranged from 1 to 6 h and were shorter during periods of high discharge.

Water-column Measurements

Near-surface and near-bed water-column SSC samples (Nowacki et al. 2023) were collected at multiple locations in Bellingham Bay on 3 and 6 February 2020 (Nooksack River discharge 274–855 m\(^3\) s\(^{-1}\)), during and immediately following a five-year recurrence interval flood of the Nooksack River (Sect. 3.2). Water samples were collected at BEL1, BEL3, BEL4, and BEL5.

Seabed Samples

Samples of the Bellingham Bay seabed (Takesue et al. 2023) were collected from Western Washington University’s R/V Magister and R/V Zoea on 9–15 February 2020 (Nooksack River discharge 120–235 m\(^3\) s\(^{-1}\)), less than two weeks following the February 2020 flood. Surface sediment from box cores and grab samples were analyzed for \(^{7}\)Be (Takesue et al. 2023), a naturally occurring cosmogenic radionuclide with a half-life of 53.2 d. Atmospheric deposition of \(^{7}\)Be accumulating on land surfaces is transported by precipitation-driven runoff to rivers and their receiving basins. Samples with \(^{7}\)Be activities greater than background atmospheric-supply levels suggests recent (<6 months) delivery, making \(^{7}\)Be a useful tool for establishing a terrestrial source of recently deposited material (Sommerfield et al. 1999).

Eight vibracores were also collected during the same period and were subsampled and analyzed for grain size; \(^{210}\)Pb, a naturally occurring cosmogenic radionuclide with a half-life of 22.3 y; and \(^{137}\)Cs, a product of nuclear-testing fallout with a half-life of 30.1 y. The vibracores were processed by subsampling 2 cm slabs at 10 cm intervals for \(^{210}\)Pb and \(^{137}\)Cs analysis via gamma spectroscopy. More complete details of the methods are presented in Takesue et al. (2023). Down-core activities of excess \(^{210}\)Pb were used to estimate the century-scale sediment-accumulation rate S at core locations (Nittrouer et al. 1979):

$$\begin{aligned} S = \frac{\lambda z}{\ln \left( \frac{A_0}{A_z}\right) }, \end{aligned}$$

where \(\lambda\) is the decay constant of \(^{210}\)Pb and \(A_0\) and \(A_z\) are excess \(^{210}\)Pb activities at two locations in the core separated by a distance z. To compute sediment accumulation with \(^{137}\)Cs, depths of the first detectable values were assigned the year 1954, and peaks were assigned 1963, the year of maximum \(^{137}\)Cs fallout (Delaune et al. 1978). These depths divided by 66 and 57 years, respectively, provided the accumulation rates.

In order to determine sediment accumulation based on the fine fraction of the fluvial sediment load, an estimate of dry bulk density is needed. In the absence of measured bulk density, and because bulk density is related to sediment composition (Flemming and Delafontaine 2000), we determine bulk density with a generalized relationship between porosity and grain size (Hamilton and Bachman 1982), the measured grain-size distribution of the muddy parts of the bay (Takesue et al. 2023), and a mineral sediment density of 2600 kg m\(^{-3}\). The bulk-density range of 500–1100 kg m\(^{-3}\) used in our estimates encompasses the values determined via the above relationship.

Bathymetric Data

Historical measurements of Nooksack River delta bathymetry were obtained from several sources. Survey data from 1956 (H08320) and 2005 (H11420) were obtained from the National Oceanic and Atmospheric Administration (NOAA) National Centers for Environmental Information Bathymetric Data Viewer (National Oceanic and Atmospheric Administration 2022a). Data from 2019 were obtained from vanArendonk et al. (2021). NAVD88 elevations of the 2019 data were shifted to MLLW, the datum of the earlier surveys, by adding 0.147 m, following the published datums of NOAA Tides & Currents station 9449211, Bellingham, Washington.

Results

Summary of Water Levels, Currents, Waves, and SSC

Depths at the moorings ranged from 4–7 m depending on location, deployment, tidal range, and non-tidal variations in water-surface elevation. Non-tidal residuals of up to 0.5 m were present, resulting from local wind setup and regional water-level trends associated with broad atmospheric pressure changes.

Tidal currents were weak at all moorings despite the relatively large tidal range, consistent with earlier observational (Collias et al. 1966) and modeling studies (Wang et al. 2010; Wang and Yang 2015). Measured current shear generally was minimal given the bin spacing, averaging interval, and current strength, and here we report depth-averaged values. Mean currents over the deployments were 0.10 m s\(^{-1}\) or less at all locations. The 95th-percentile (2.5–97.5%) ranges of flow speeds were 0.01–0.30 m s\(^{-1}\) at BEL1, 0.01–0.18 m s\(^{-1}\) at BEL3, 0.01–0.16 m s\(^{-1}\) at BEL4, and 0.01–0.21 m s\(^{-1}\) at BEL5. The primary tidal constituent (typically M2) velocity amplitude was 0.05 m s\(^{-1}\) or less at each location. Current directions at BEL1 were generally aligned north–south. At BEL3, currents were generally ESE–WNW, while they were E–W at BEL4 and ENE–WSW at BEL5. Wind forcing drove considerable variability around these mean tidal current speeds and directions. When determining the tidal component of the currents via harmonic analysis, the root-mean-squared error of the reconstructed tidal predictions was 0.03–0.08 m s\(^{-1}\). These errors are comparable to the 95th-percentile flow-speed ranges, suggesting processes other than astronomical tides are dominant in Bellingham Bay.

Waves tended to be larger during the winter deployments; mean winter significant wave heights were 1.9–2.3 times greater than summer. Waves were also larger at the delta moorings than at BEL1, consistent with the southerly winds present throughout the year in Bellingham Bay and resultant greater fetch for the delta moorings. Significant wave heights were on average less than 0.5 m at all locations (95th-percentile range 0.05–1.09 m). Median peak wave periods were approximately 3 s. Wave generation was driven entirely by local winds. For winds with a northward component, \(R^2\) values for the relationship between wind speed and significant wave height at BEL3–5 ranged from 0.38 to 0.55.

Surficial seabed sediment at BEL1 was 33% gravel, 58% sand, and 8% mud, BEL3 was 98% mud and 2% sand, BEL4 was 55% sand and 45% mud, and BEL5 was 65% sand and 35% mud. Near-bed suspended-sediment concentrations (SSC) varied by deployment but generally were smallest at BEL1. Values at BEL3–5 were more similar and had 95th-percentile ranges of about 5–250 mg L\(^{-1}\). Near-surface SSC ranged from 5 to 50 mg L\(^{-1}\) at BEL3 and BEL5.

River Floods

Dynamics of Individual Floods

A five-year recurrence interval flood occurred in early 2020, when the Nooksack River discharged approximately 0.16 Mt of sediment between 31 January and 6 February (Fig. 2a), 11–19% of its typical annual load. By comparison, a one-week period in late January typically discharges about 3% of the annual load, and a typical one-week period during peak sediment flux in November discharges about 6% of the annual value. The flood was preceded by strong winds and accompanying waves associated with the storm. Wind-driven flow at BEL3 at speeds up to 0.2 m s\(^{-1}\) dwarfed the tidal velocities present during non-windy conditions (Fig. 2b). Near-bed SSC was elevated during the wavy period, with significant wave heights sometimes exceeding 1 m in the 5–7 m water depths at the mooring locations. By the time peak water and sediment discharge arrived at the river mouth, Bellingham Bay was relatively quiescent, with wave heights less than 0.5 m (Fig. 2c). Despite this large event, almost no water-column signature of the river flood was detected near the seabed at BEL3 along the outer delta topset during the quiescent period (Fig. 2d). Although the small but detectable increases in BEL3 benthic SSC during the quiescent period could be attributed to sediment settling from the plume, these occurred concurrently with moderate (but still smaller than pre-flood) waves.

Fig. 2
figure 2

a: Nooksack River water (Q) and sediment (\(Q_s\)) discharge during a five-year flood event in early February 2020. Vertical gray line shows time of sampling in Fig. 3a. b: Depth-averaged flow velocity at BEL3. b: Significant wave height at BEL3 and BEL4 and KBLI wind speed. d: Near-bed SSC at BEL3 (data unavailable at BEL4). Note minimal tidal velocities and evidence of wind-driven flow at BEL3, correlation between wave height and SSC, and lack of elevated near-bed SSC during river flood

Full size image

No surface time-series data were available during the 2020 flood, but near-surface and near-bed water samples collected during and immediately after the flood show much greater (2.5–19\(\times\)) SSC near the surface than near the seabed (Fig. 3a), suggesting the presence of a stratified river turbidity plume lacking interaction with the bottom. (Near-surface turbidity was measured during the 2021 deployment and is described later in this Section.) The spatial extent of the plume was large, as indicated by elevated near-surface SSC relative to the near-bed values measured at BEL1, 8 km from the river mouth. The decoupling of fluvial processes and benthic marine processes at shallow depths on the outer delta topset as well as at locations distant from the mouth suggests a strongly stratified river turbidity plume that was disconnected from the bottom during the flood, and that near-bed sediment dynamics are strongly related to wave resuspension.

Fig. 3
figure 3

a: Near-surface and near-bed SSC collected at BEL1, BEL3, and BEL4 on 3 February 2020 and at BEL5 on 6 February 2020. Numbers indicate SSC in mg L\(^{-1}\). Near-surface SSC was 2.5–19 times greater than that near the bed. Symbols at mooring locations have been offset for display purposes. b: Seabed samples analyzed for \(^{7}\)Be during the 9–15 February 2020 cruise showing detectable \(^{7}\)Be at eight locations. Contours show depth in meters

Full size image

Approximately one year later, in late February 2021, a flood that increased water discharge from about 70 m\(^3\) s\(^{-1}\) to more than 250 m\(^3\) s\(^{-1}\) resulted in peak river sediment-discharge rates of 0.12 t s\(^{-1}\) (Fig. 4). This flood was about four times smaller than the February 2020 flood but was nevertheless nearly three times greater than the mean Nooksack River discharge. During the flood peak, significant wave heights at BEL5 were small, 0.3 m or less, but were bookended by larger heights approaching 1 m immediately prior to and following maximum discharge, similar to the patterns in the February 2020 flood. In other words, the energetics of the river’s receiving basin were distinct from those of the river and its watershed, despite a system composed of a small mountainous river with a watershed spatial scale of about 65 km and a diminutive, protected basin.

During the flood, near-bed SSC recorded by the delta moorings increased from about 10 mg L\(^{-1}\) to greater than 100 mg L\(^{-1}\) during the wavy periods before and after the flood, but remained low, around 10 mg L\(^{-1}\), throughout the flood (Fig. 4b). In contrast, near-surface SSC was minimal during the waves but approached 100 mg L\(^{-1}\) in response to the flood and associated sediment-laden surface plume spreading across the bay. Despite relatively energetic wind-wave conditions before and after the flood, wave resuspension was not sufficient to mix sediment through the full water column, evidenced by low SSC measured at the near-surface sensors. Consistent with the disparity between SSC measured at the near-bed and near-surface sensors, resultant sediment fluxes were also uncoupled. Unit-width fluxes were less than 5 g s\(^{-1}\) m\(^{-2}\) during the non-wavy and low-discharge periods, but peaked at 10–15 g s\(^{-1}\) m\(^{-2}\) for both parameters during the wavy (near-bed) and high-discharge (near-surface) periods (Fig. 4c).

Fig. 4
figure 4

a: Nooksack River water (Q) and sediment (\(Q_s\)) discharge during a minor flood in late February 2021. b: Significant wave height at BEL3–5 and KBLI wind speed. c: Near-surface and near-bed SSC at BEL5. d: Near-surface and near-bed sediment flux at BEL5. Compare axis limits in (a), (b) and (c) to those in Fig. 2. Patterns in (c) and (d) are qualitatively similar at BEL3, particularly for the near-bed dynamics

Full size image

Integrated Effects of River Floods

Like the patterns observed in the February 2020 and February 2021 floods, correlations between near-surface SSC and Nooksack River discharge and between near-bed SSC and waves were evident throughout the deployment (Fig. 5).

SSC increases at Ferndale were correlated with elevated near-surface SSC at BEL5, which also exhibited higher-frequency, tidal-scale variability (Fig. 5a). The peak correlation is obtained with a lag of about 13.5 h between Nooksack River SSC and near-surface SSC at BEL5 (Fig. 5b). The lag results partially from the transport time required between the Ferndale gauge and the mouth, estimated to be 1–6 h using the discharge–area relationship at the Ferndale gauge. Additional components of the lag include transport from the mouth to BEL5 via integration of the tidal cycle and wind conditions present during or temporally adjacent to high-SSC time periods.

In contrast, and in a manner consistent with the individual floods already discussed, benthic SSC was much more strongly correlated with wave conditions (Fig. 5c). At all benthic sites, there was minimal correlation between near-bed SSC and Nooksack River SSC or discharge; it was instead strongly correlated with local wind-generated waves. The response in SSC to wave forcing was rapid: the greatest cross-correlation occurred at a lag time of 0.25 h (Fig. 5d).

The benthic SSC response to wind waves was nearly instantaneous (less than one hour) and spatially local, while the near-surface SSC response for the river forcing was both longer in time (hours to days), representing the range of river floods draining the Nooksack River watershed, and larger in space, reflecting the multiple kilometers traveled by the river plume to the mooring sites.

Fig. 5
figure 5

a: Near-surface SSC at BEL5 (blue) and Nooksack River (orange) from late January to mid March 2021. A 12-h running average has been applied to the raw BEL5 data (gray), revealing a strong correlation between Nooksack River and near-surface SSC. b: Cross-correlation between near-surface BEL5 SSC and Nooksack River SSC measured at Ferndale showing greatest correlation at a lag of 13.5 h (BEL5 lags Nooksack River). c: Near-bed SSC (blue) and significant wave height (orange) at BEL5. d: Cross-correlation between near-bed BEL5 SSC and BEL5 significant wave height showing greatest correlation at a lag of 0.25 h

Full size image

Wind Forcing

Given the minimal tidal velocities in Bellingham Bay, wind forcing and freshwater input are important factors in setting circulation patterns bay-wide. Particularly during northward winds, when fetches are greater at the delta mooring locations, subtidal depth-averaged flow directions were anti-correlated with the direction of wind forcing. As exemplified by observations at BEL3, on the eastern flank of the delta topset, flow was southward for northward winds (Fig. 6a), suggesting return flow from water-level setup at the north end of the bay. Similarly, flow was northward during southward winds. These patterns were observable throughout the deployments and are reminiscent of similar wind-driven flows seen in long lakes (e.g., Csanady 1973), where wind-driven setup drives return flow. Although there was considerable scatter (\(R^2=0.31\)), regression of the tidally averaged winds against the tidally averaged flow velocities at BEL3 produced a slope of approximately 0.004. In other words, each 1 m s\(^{-1}\) increase in north–south wind velocity resulted in a 0.004 m s\(^{-1}\) increase in north–south flow velocity. Given an M2 tidal constituent of no greater than 0.05 m s\(^{-1}\), even moderate winds can produce flows comparable to the tidal forcing. As has already been seen, wind-generated waves were effective at resuspending sediment near the bed, and the larger waves generated by larger fetch during northward winds produced the greatest benthic SSC values (Fig. 6b). Near-bed SSC did not appreciably increase for southward winds given the minimal fetch at BEL3 for these wind conditions. Resultant southward sediment fluxes were greatest at BEL3 during periods of strong northward winds and were oriented along isobaths toward the SE (Fig. 6c). Moderate northwestward sediment fluxes were generated by south–southwestward winds, even though the SSC response to winds was minimal for these wind directions. The distribution of wind conditions is key to assessing the importance of these patterns (Fig. 6d). Using an arbitrary threshold of 5 m s\(^{-1}\) to indicate windy conditions, 22 percent of all measured non-zero wind speeds were greater than 5 m s\(^{-1}\) and northward. In contrast, 9 percent of all winds were greater than 5 m s\(^{-1}\) and southward. Thus, northward moderate-to-strong wind events are most representative of the resuspension and return-flow transport process illustrated by Fig. 6.

Fig. 6
figure 6

Plots of a: BEL3 depth-averaged flow velocity, b: near-bed SSC, and c: near-bed sediment flux, and d: KBLI wind binned by E–W and N–S wind velocity during the deployments. Arrows in (a) and (c) show flow direction. Particularly for strong northward winds, SSC, flow, and sediment flux are enhanced in a southward direction. Note log color scales in (b)–(d)

Full size image

Sediment Accumulation and Export

Short-term Terrigenous Deposition

Geochemical analyses of bed sediment collected 7–14 days after the February 2020 flood show evidence of recent terrestrial sediment input to the bay. Significant \(^{7}\)Be activity (values greater than three times the measurement uncertainty) observed in eight of 29 seabed samples indicated terrestrial input within \(\sim\)6 months of early February 2020 (Fig. 3b). Samples with significant activity were generally located near the urban waterfront and along the delta foreset. Two of the eight samples with significant activity were located within the central bay at 20–25 m water depth. Despite the highly stratified nature of the turbidity plume and general lack of a detectable increase in benthic SSC at the instrument locations during the February 2020 river flood, the presence of \(^{7}\)Be in the eastern portion of the bay indicates terrigenous material settles out of the plume and is deposited on the seabed. Although \(^{7}\)Be activity represents an integration of terrigenous delivery over the preceding several months, the large magnitude of the February event likely makes it the dominant source.

Long-term Fine-sediment Accumulation

The short-term terrigenous sedimentation indicated by \(^{7}\)Be is corroborated by longer-term sedimentation indicated by excess \(^{210}\)Pb. The naturally occurring radioisotope \(^{210}\)Pb enables assessment of century-scale sediment accumulation on the seabed within Bellingham Bay. Accumulation rates were calculated from vibracores collected during the February 2020 cruise, shortly after the large flood event (Figs. 7 and S1). Excess \(^{210}\)Pb is associated most strongly with fine-grained sediment, and at all locations, cores consisted of 90% or greater mud at all depths, except for a coarse layer (66% sand) at 50 cm depth in core VC08 (Fig. S3). Average accumulation rates across the muddy central bay as determined at the eight cores ranged from 0.45 to 1.45 cm y\(^{-1}\). The highest calculated rate was on the outer delta foreset, and the second greatest was at a site near the Bellingham industrial waterfront. All of the rates were considerably greater than the range of local sea-level rise (SLR) estimates computed at NOAA tide gauges at Cherry Point (20 km NW of Bellingham Bay), Friday Harbor (40 km SW), and Port Townsend, Washington (70 km S) (National Oceanic and Atmospheric Administration 2022b). Accumulation rates calculated using \(^{137}\)Cs were 0.17–2.14 cm y\(^{-1}\) (Fig. S2). The two highest rates computed with \(^{137}\)Cs were at VC01 and VC09, the same as those determined using \(^{210}\)Pb.

Fig. 7
figure 7

Sediment-accumulation rates (cm y\(^{-1}\) \(\pm \, \sigma\)) derived from \(^{210}\)Pb geochronology at eight sites within Bellingham Bay. The range of SLR rates (cm y\(^{-1}\)) derived at three nearby NOAA tide gauges is shown for comparison. Marker size and color correspond to rates. Contours indicate depth in meters

Full size image

The Nooksack River discharges 0.86–1.4 Mt y\(^{-1}\) of sediment, about 58% of which is fine-grained. Assuming equivalent deposition across the 70 km\(^2\) area of Bellingham Bay north of Chuckanut Bay and a dry-sediment density of 500–1100 kg m\(^{-3}\) results in a yearly mud deposition of 0.06–2.3 cm (7.1–11.6 kg m\(^{-2}\)). These values bracket those computed using \(^{137}\)Cs and \(^{210}\)Pb geochronology. The equivalent volume- and mass-accumulation rates for the geochronological estimates are \(6.1\times 10^5\) m\(^3\) y\(^{-1}\) or 0.31–0.67 Mt y\(^{-1}\) using the above bulk density values and represent about 40–60% of the yearly fine-sediment discharge of the Nooksack River. Taken at face value, and assuming the \(^{137}\)Cs and excess \(^{210}\)Pb activities represent only the fine-sediment dynamics, these percentages suggest about half of the fine-grained Nooksack River sediment is transported out of Bellingham Bay and carried to other locations within the Salish Sea. Given the sluggish tidal currents in Bellingham Bay, waves, wind-driven currents, and gravitational plume spreading that can convey sediment to more energetic environments like Hale Passage (Fig. 1) likely are responsible for this export.

Long-term Coarse-sediment Accumulation

To assess the fate of the fluvial sand fraction, we use historical bathymetric surveys spanning 63 years to assess delta progradation (Fig. 8). The repeat surveys indicate 400 m of progradation at the 5-m isobath over 63 years, corresponding to a rate of 6.3 m y\(^{-1}\). Progradation between 1956 and 2005 was greatest at the east-central portion of the subaqueous delta, but there was little progradation at this location between 2005 and 2019. In contrast, there was relatively less progradation from 1956 to 2005 at the west-central delta but greater progradation in the later time period. Beginning in 2006 and continuing through 2011, a large scale avulsion of the distributary channels from eastern distributaries to those on the western delta took place (Northwest Hydraulic Consultants 2015), which is the likely driver of the change in progradation in the different time periods. A bathymetric cross-section along the west-central delta shows the character of this progradational foreset and the increasing rate of growth in the first two decades of the 21st century (Fig. 8).

Using the annual Nooksack River sediment-discharge values and a 42% coarse fraction results in 0.36–0.59 Mt y\(^{-1}\) of sand export. Assuming a sand density of 1600 kg m\(^{-3}\) equates to a yearly sand export of \(0.23\times 10^6\)–\(0.37\times 10^6\) m\(^3\). Distributed along the 6 km-wide delta front and to 10 m water depth, the sand flux corresponds to a lateral progradation rate of 3.8–6.1 m y\(^{-1}\), values which compare favorably to the repeat surveys. This suggests that a large majority of the coarse material accumulates in the Nooksack River subaqueous delta.

Fig. 8
figure 8

a: Locator map of Bellingham Bay showing area of (b) and transect shown in (b) and (c). b: 5-m depth contours along the Nooksack River subaqueous delta foreset showing progradation from 1956 to 2019. Dotted line indicates transect shown in (c). c: Bathymetry profiles along cross-shore transect showing delta progradation and foreset steepening

Full size image

Comparing the February 2020 Flood to Long-term Sediment Accumulation

Using the suspended-sediment grain-size distribution at Ferndale, an estimated 0.09 Mt of fine material and 0.07 Mt of sand was discharged during the early 2020 flood. If the fine material were spread evenly across Bellingham Bay, a mud layer about 0.12–0.26 cm in thickness would result. The estimated thickness from the 2020 flood is 14–30% of the mean yearly mud deposition determined via excess \(^{210}\)Pb and compares favorably to the estimated 11–19% of the annual sediment load discharged during this event at the Ferndale gauge (Sect. 3.2).

Similarly, an equal distribution of coarse material from this event along the delta front would represent about 0.8 m of lateral delta progradation. In this context, the early 2020 flood represents about 13% of the yearly delta progradation rate computed above and is consistent with the 11–19% of the annual sediment load of this storm as measured at the Ferndale gauge.

Discussion

Relevance of Waves and River Floods to Near-bed Sediment Dynamics and Fate

Because of the observed incoherence between fluvial-source and receiving-basin energetics, riverine floods maintain a subtle relationship to near-bed sediment and contaminant dynamics in Bellingham Bay. The river turbidity plume is strongly stratified, and although gravitational settling of sediment out of the water column onto the seabed occurs, this typically is not reflected in a meaningful increase in near-bed SSC at the instrument locations (Fig. 4). Given the bathymetry and grain-size distributions of the delta and Bellingham Bay, riverine sediment is strongly associated with the dynamics and morphology of the system, but in a way that is not immediately apparent with our time-series observations at the event to seasonal scale.

Wave energy is generally not sufficient to resuspend sediment to the surface even on the relatively shallow delta topset and waves are not relevant to near-surface sediment transport at water depths between 4 and 7 m, where our instruments were located. Near-bed SSC dynamics, however, are largely controlled by wave processes. Indeed, for BEL5 \(H_s\) greater than 0.3 m, each wave-height increase of 0.1 m drives an additional 9 mg L\(^{-1}\) near-bed SSC on average. Similarly, every increase of 10 mg L\(^{-1}\) at Ferndale corresponds to a 1 mg L\(^{-1}\) increase in near-surface SSC at BEL5 when incorporating a time lag of 13.5 h. At the scale of an individual flood, this approximately tenfold scaling between near-surface and Ferndale SSC can be seen in Figs. 2 and 3.

To investigate the relationship between waves and benthic SSC further, we estimate a critical bed stress from wave action sufficient to resuspend sediment following Madsen (1994). To determine the critical bed stress, we compute the binned median SSC as a function of the combined wave–current shear velocity \(u_{*wc}\). A noticeable increase in SSC starts at a \(u_{*wc}\) value of approximately 0.05 m s\(^{-1}\) at BEL3–5. This value is exceeded about 8% of the time at BEL3, and 15% of the time at BEL4 and BEL5, suggesting a cumulative sum of 30–50 days per year of wave-induced resuspension at those sites. Larger wave events resuspend more material but occur more rarely. For example, a wind event generating significant wave heights of about 1.0 m results in \(u_{*wc}\) values of roughly 0.1 m s\(^{-1}\) at the mooring locations and increases SSC by about 100 mg L\(^{-1}\). Events of this magnitude occurred approximately one percent of the time during the deployments, or about 3.8 d per year when considered as an annual rate. The same southerly winds that are most effective at generating the waves that resuspend sediment on the delta topset also drive southward return flows and may act as a process to prevent long-term accumulation of fine sediment on the delta topset.

The optical sensors used in this study are most sensitive to fine-grained sediment, and as such our benthic observations primarily represent the fines in suspension on the delta topset; sands likely are transported largely as bedload. Our near-surface observations represent the fine-sediment in suspension in the river plume. Given the stratified nature of the turbidity plume, these fines could be those ephemerally deposited on the topset and resuspended by waves and transported into deeper water via return flow following flood events. Without a continued supply of fines (i.e., between floods), a lower wave–turbidity relationship could develop over time as the pool of fines is winnowed. Another alternative is that fluvial fines bypass the delta topset as they disperse, and the wave–turbidity response at the topset sites represents remobilized sands.

Our data suggest that settling of sediment is relatively slow on the outer delta topset (Fig. 4). Wave resuspension is frequent (Fig. 5) and the most frequent and strongest resuspension events are driven by southerly winds that result in additional southward sediment flux away from the delta topset (Fig. 6). Combined, thse processes result in efficient dispersal of fine sediment away from the delta topset and accumulation in deeper bay waters (Fig. 7). These processes also efficiently sort bottom sediments and result in a sandy delta topset and muddy bay floor elsewhere, except in environments like Hale Passage, where high currents prevent accumulation of fine sediment.

Potential for Reworking at the Seasonal Scale

To assess sediment-reworking potential in the Nooksack River–Bellingham Bay system, we consider monthly mean values of water discharge Q, sediment discharge \(Q_s\), and wind speed for winds with a southerly component. Q peaks in November and remains high through January; elevated values associated with the spring freshet eventually drop considerably in July (Fig. 9). For these monthly means, maximum Q (in November) is 3.4 times greater than the minimum (in September). \(Q_s\) reaches a dramatic peak in November, more than double the value of the next-highest month, and maintains elevated values only through March; \(Q_s\) is relatively low during the spring and summer and reaches its annual minimum in August. For \(Q_s\), the maximum value is about 13 times greater than the minimum, a difference four times that of Q. Southerly wind speed is variable throughout the year, but is greatest in October–March in conjunction with fall and winter storms, and reaches its minimum in August and September. The maximum monthly mean wind speed is only 1.5 times greater than the minimum. Because the Nooksack River drains a small watershed, the standard deviation relative to the mean values plotted in Fig. 9 are rather large (see Fig. 9 caption).

Fig. 9
figure 9

Monthly mean values of Nooksack River water discharge Q (overall mean \(112\pm 93\) m s\(^{-1}\)), Nooksack River sediment discharge \(Q_s\) (overall mean \(29 \pm 111\) kg s\(^{-1}\)), and wind speed for winds with a southerly component (overall mean \(4.3\pm 2.2\) m s\(^{-1}\)). Sediment discharge peaks in late fall. Strong winds rework this flood material during the rest of the winter and spring, when sediment discharge is lower

Full size image

Patterns emerge in the seasonality of these three parameters, which are closely related to sediment transport. First, in the Nooksack River and other small-to-moderate Salish Sea river systems, Q is not a suitable surrogate for \(Q_s\), as the seasonal hysteresis in the water–sediment relationship is significant (Anderson et al. 2019). This indicates the importance of direct measurements of turbidity rather than relying on Q–\(Q_s\) rating curves. Although Q is elevated throughout the spring, \(Q_s\) is less important during this time, particularly compared to the late fall and early winter when \(Q_s\) (and Q) is greatest. \(Q_s\) is greatest in the early fall, when rain falls on a landscape without snow cover, thereby mobilizing sediment downstream (Downing 1983) and resulting in a high ratio of \(Q_s\) to Q. This behavior—relatively lower \(Q_s\) compared to Q during the spring freshet and summer periods—is distinct from that observed on the nearby Skagit River, for instance, where high SSC from glacier melt can drive greater \(Q_s\) during summer periods when Q is typically low (Curran et al. 2016).

Like Q, southerly wind speed—which is often associated with the same storms that ultimately drive the increases in Q—is high throughout the winter and spring. As a result, the sediment discharged by winter floods can be reworked by months of subsequent wind waves before the largest sediment inputs arrive again in the late fall of the following year. Southerly winds drive reverse flows induced by water level set-up at the northern bay margin (Fig. 6); this, coupled with wind-wave resuspension, could contribute to southward export of bed sediment, both new (fluvial) and pre-existing (associated with legacy contaminants), from Bellingham Bay. This reworking at the seasonal scale is distinct from the dynamics at the scale of an individual storm, which is considered in the next section.

River–ocean Coherence at the Storm Scale

Within western Washington, the largest floods are caused by autumn and winter precipitation associated with atmospheric rivers (Neiman et al. 2011). Indeed, the dominant storm landfall region on the Pacific coast of North America is on the coast of Cascadia and British Columbia (Bromirski et al. 2017). The coherence of river and ocean conditions has been noted along the U.S. west coast: rivers debouching to the open ocean often do so into an energetic ocean environment (Kniskern et al. 2011). In contrast, the lack of swell in protected basins like Bellingham Bay means that the short timescales of wind and wave forcing result in low coherence of river and ocean (bay) conditions at the storm scale. If a river delivers sediment into a quiescent basin, that material will potentially be preserved preferentially if another mechanism does not rework it.

On the Nooksack River, the coherence between the river and its receiving basin is time-shifted at the storm scale, in contrast to those dynamics on the open coast. Nooksack River floods are typically delivered to a more quiescent receiving basin than the basin’s conditions prior to the flood (Fig. 10). For the 100 largest water floods and 50 largest sediment floods between September and March, winds were strongest 16–30 h prior to the flood peak for water floods and 12–20 h for sediment floods (Fig. 10a). The lags are shorter for sediment floods because the sediment flood wave arrives earlier than the water flood wave (e.g., Fig. 2). Of the water floods, 73% had winds speeds faster 24 h prior to the peak than during the peak itself.

Wind speeds were faster during the pre-flood period (Fig. 10b). Twenty-four hours prior to the flood peak, wind speeds were on average 3.2 m s\(^{-1}\) faster than during the peak. The less energetic period during the debouchment of water and sediment to Bellingham Bay suggests that flood sediment may be able to settle more efficiently on the seabed and potentially form part of the geologic record. Indeed, the presence of \(^{7}\)Be at multiple locations around the bay (Fig. 3) and accumulation rates in excess of SLR (Fig. 7) indicates that flood material reaches depth.

Considered in terms of \(u_{*wc}\), a 3 m s\(^{-1}\) decrease in wind speed, say from 8 m s\(^{-1}\) prior to the flood to 5 m s\(^{-1}\) during the flood, results in a 37% decrease in the median \(u_{*wc}\), from 0.068 to 0.043 m s\(^{-1}\). Depending on the specifics of a given storm, this reduction could make the difference between a wave climate sufficient to resuspended sediment and one that enables sediment deposition during a flood (Sect. 4.1).

Fig. 10
figure 10

Wind-speed statistics for the 100 largest Nooksack River water floods and 50 largest sediment floods occurring between September and March using the 20-year dataset (eight years for \(Q_s\)). a: Fraction of wind speeds faster before floods than those during the flood peak for a range of hours prior to each flood. b: Two-dimensional histogram of wind-speed difference before and during water floods by hours before flood peak (sediment-flood histogram not shown). Lines show mean and standard deviation of the wind-speed difference for Q and \(Q_s\). Wind speeds were faster 12–36 h prior to each flood peak for a majority of floods

Full size image

Mechanisms Leading to Retention or Export of Fluvial Sediment

Nearly all of the coarse material and about half of the fine fraction of sediment discharged by the Nooksack River can be accounted for within Bellingham Bay (Sect. 3.4). Given roughly equal coarse and fine fractions, this suggests about 75% total sediment retention within the system. This high rate of retention may enable sufficient sediment accretion in local habitats in response to sea-level rise as fluvial sediment is deposited over pre-existing bay sediment. Other fluvial dispersal systems within the Salish Sea export greater proportions of sediment than are retained within their respective subaqueous deltas. On the Elwha River, for example, although most coarse material is retained in nearshore areas, more than 90% of fine-grained river sediment is lost via energetic tidal currents in the Strait of Juan de Fuca, that river’s receiving basin (Gelfenbaum et al. 2015; Warrick et al. 2015). On the Skagit River, 39% of the total (fine and coarse) sediment load can be accounted for in deposits near the river mouth, suggesting 61% export (Grossman et al. 2020). Here we consider the processes that can lead to sediment retention or export in protected embayments sharing features similar to Bellingham Bay.

In a number of other systems of the protected, meso–macrotidal Salish Sea, tidal currents are the key factor in dispersing fluvial sediment delivered to the estuary; the general absence of meaningful tidal currents in Bellingham Bay makes it an outlier among these other systems. For example, intense tidal currents in the Nisqually River delta in the southern Salish Sea rework that system’s delta-front morphology (Barnhardt and Sherrod 2006). In a fashion similar to the Nisqually, sediment discharged from the Skagit River, the largest water and sediment source to the Salish Sea in the U.S. (Czuba et al. 2011), is deposited and reworked on its extensive tidal delta by vigorous tidal currents. There, fine-grained material is rapidly dispersed by strong tidal currents just offshore of the delta (Webster et al. 2013; Grossman et al. 2020). The subaqueous delta of the Elwha River is tidally dominated, and sediment-transport capacity is predictable and persistent regardless of river dynamics. If tides are strong enough, this transport capacity may be sufficient to preclude long-term mud deposition, which leads to patchy sediment deposits in zones of weak currents (Gelfenbaum et al. 2015; Eidam et al. 2016). On the Puyallup River, the correlation between river conditions and receiving-basin characteristics extends to phasing of the tide. There, low tides may be necessary to produce hyperpyncnal sediment flows in order to ensure vigorous transport of coarse sediment to the edge of the river-mouth platform of this system (Mitchell 2005).

River discharge is, naturally, another important control on sediment delivery and retention. Sediment delivery to coastal wetlands near the mouth of the Stillaguamish River is strongly controlled by river discharge, both via routine storms and also from catastrophic geomorphic events which modulate delivery to the coastal zone (Nowacki et al. 2020). Despite the existence of coastal storms capable of resuspending and transporting sediment, delivery to the coastal zone in that system is dominated by fluvial delivery.

Wind waves are a further mechanism to resuspend benthic sediment and make it available for transport. Near Hood Canal’s Skokomish delta, weak currents (Noble et al. 2006) and short fetch from the south likely result in minimal reworking of its subaqueous delta. The large fetch to the north may allow resuspension of fine and potentially coarse material during strong northerly wind events.

Finally, geomorphic events like large debris flows, lahars, and landslides that add a notable supply of different-sized sediment (e.g., Tucker et al. 2014; Nowacki et al. 2020) might modify some of the relationships found here.

In most river systems delivering to protected embayments, a balance among fluvial input, tidal currents, and wave resuspension determines the system’s sediment retention or export. As seen in this work, Bellingham Bay is distinct from others in the Salish Sea: despite a large tidal range, its weak tidal currents are insufficient to resuspend sediment regularly. With a moderate wave fetch and moderate-to-strong winds, sediment is easily resuspended by waves. Subtidal return flows generated by the wind-driven water-level setup are comparable to and often exceed the magnitude of the tidal currents and are important in transporting the remobilized material.

Conclusion

Bellingham Bay, the receiving basin of the Nooksack River, a U.S. west coast small mountainous river, is subject to large fluvial sediment loads, moderate waves, and weak tidal currents, which lead to sediment retention within the system. Reworking of benthic sediment by local wind waves and sediment transport via wind-driven flows are important to long-term accumulation of sediment in the system and remobilization and redistribution of legacy contaminants from historical activities. Stratification of the turbidity plume results in dramatically different suspended-sediment concentrations and transport dynamics in near-surface and near-bed waters. Despite small drainage and receiving-basin areas, fluvial sediment input was often decoupled from wind-wave reworking because of a time lag between the arrival of the storm and its resultant flood: the Nooksack River discharges into a protected embayment that has such short time scales of wind-wave generation that the storm may have passed by the time the flood arrives. In the absence of meaningful tidal currents, wind-driven counter flows provide important circulation forcing. For example, during strong northward winds, a southward reverse flow coupled with wind-wave resuspension could export sediment temporarily deposited on the delta to deeper locations within Bellingham Bay or potentially outside of the bay. The wave climate on the outer delta topset is great enough to redistribute the sizable fine fraction delivered by this high-yield river.

Sediment delivered during a five-year flood event in mid winter produced a theoretical deposit of 0.12–0.26 cm across the muddy portion of the bay, and a potential 0.8 m of lateral delta progradation. At the decade-to-century timescale, 0.8–1.1 cm y\(^{-1}\) of fine-grained sediment is deposited in the deeper portion of the bay, and the lateral delta progradation rate is about 6 m y\(^{-1}\). These values represents about 75% of the annual sediment load delivered by the Nooksack River. The high rate of retention of fluvial sediment in the Nooksack River–Bellingham Bay system may benefit local wetland, intertidal, and subtidal habitats in the face of SLR and affect bay sediment-quality via addition of fluvial sediment to pre-existing bay sediment.