Skip to main content
NCBI home page
EPA Author Manuscripts logoLink to EPA Author Manuscripts
. Author manuscript; available in PMC: 2023 Jan 7.
Published in final edited form as: Estuaries Coast. 2022 Jan 7;45:1615–1630. doi: 10.1007/s12237-021-01040-y

Effects of Biophysical Processes on Diel-Cycling Hypoxia in a Subtropical Estuary

Melissa S Duvall 1, Brandon M Jarvis 2, James D Hagy III 3, Yongshan Wan 2
PMCID: PMC9727754  NIHMSID: NIHMS1853165  PMID: 36505267

Abstract

In shallow estuaries, fluctuations in bottom dissolved oxygen (DO) at diel (24 h) timescales are commonly attributed to cycles of net production and respiration. However, bottom DO can also be modulated by physical processes, such as tides and wind, that vary at or near diel timescales. Here, we examine processes affecting spatiotemporal variations in diel-cycling DO in Escambia Bay, a shallow estuary along the Gulf of Mexico. We collected continuous water quality measurements in the upper and middle reaches of the Bay following relatively high (> 850 m3 s−1) and low (< 175 m3 s−1) springtime freshwater discharge. Variations in diel-cycling amplitude over time were estimated using the continuous wavelet transform, and correlations between DO and biophysical processes at diel timescales were examined using wavelet coherence. Our results reveal that freshwater discharge modulated inter-annual variations in the spatial extent and duration of summertime hypoxia through its effect on vertical density stratification. In the absence of strong stratification (> 15 kg m−3), vertical mixing by tropic tides and sea breeze enhanced diel fluctuations in deeper areas near the channel, while in shallower areas the largest fluctuations were associated with irradiance. Our findings suggest that processes affecting diel-cycling DO in the bottom layer can vary over a relatively short spatial extent less than 2 km and with relatively small changes in bottom elevation of 1 m or less. Implications for water quality monitoring were illustrated by subsampling DO timeseries, which demonstrates how low-frequency measurements may misrepresent water quality in estuaries where diel-cycling DO is common. In these systems, adequate assessment of hypoxia and its aquatic life impacts requires continuous measurements that capture the variation in DO at diel timescales.

Introduction

Low dissolved oxygen (DO), or “hypoxia,” is a long-standing water quality concern of global significance. The conventional 2 mg L−1 threshold for hypoxia is broadly applied in estuaries and coastal ecosystems to describe DO concentrations at which significant mortality effects occur for aquatic organisms (e.g., Bricker et al. 1999; Vaquer-Sunyer and Duarte 2008). Hypoxic zones have expanded globally over recent decades, fueled by anthropogenic eutrophication and climate change (Diaz and Rosenberg 2008; Rabalais et al. 2010). Given its prevalence and impacts on ecological and socioeconomic systems (Cloern 2001), the physical and biological processes that modulate the development and maintenance of hypoxia in coastal ecosystems have been studied extensively (e.g., Fennel and Testa 2019).

On interannual and seasonal timescales, the formation and expansion of coastal hypoxic zones are often attributed to conditions that increase organic carbon inputs (e.g., warm temperatures, nutrient loading) and decrease circulation of bottom waters (e.g., long water residence time, strong stratification). However, hypoxia can also be intermittent, characterized by fluctuations in DO over a diel (24 h) cycle due to daytime net production and nighttime net respiration (D’Avanzo and Kremer 1994; Beck and Bruland 2000). Less commonly recognized, diel-cycling hypoxia (or diel-cycling DO) can also be sustained by physical processes that vary at or near diel timescales. For example, mixing by diurnal tides can breakdown stratification and enhance diffusion of DO from surface to subpycnocline waters (Simpson et al. 1990). Diurnal tides can also initiate diel-cycling through horizontal advection of water masses with different DO concentrations (Sanford et al. 1990; Nezlin et al. 2009; Beck et al. 2015).

The effects of hypoxia on aquatic life vary between species and are often dependent on the frequency and duration of exposure. For example, Borowiec et al. (2015) and Regan and Richards (2017) demonstrated differences in the behavioral and physiological responses of fish to diel-cycling versus prolonged hypoxia. For some species, diel-cycling can pre-condition individuals and improve hypoxic tolerance by promoting cellular and metabolic modifications (Williams et al. 2019). Diel-cycling may also increase or decrease feeding and growth rates (Dan et al., 2014; Yang et al. 2013) and alter the number and distribution of individuals within an ecosystem (Tyler and Targett 2007; Brady and Targett 2013). Duration and frequency of hypoxic events are particularly important for sessile organisms, such as oysters, that are typically tolerant of hypoxia unless exposure is prolonged (Keppel et al. 2016). Given these complex patterns in both ambient oxygen levels and aquatic life responses, there is growing recognition of the need for ecologically relevant water quality criteria that account for both lethal and non-lethal effects of low DO (Vaquer-Sunyer and Duarte 2008; Hrycik et al. 2017). For example, revised DO criteria for the Chesapeake Bay established habitat-specific thresholds aimed at safeguarding the growth and reproduction as well as survival of various life stages (U.S. Environmental Protection Agency 2003).

Despite its relevance to the protection of aquatic life, there is a limited understanding of diel-cycling dynamics in coastal ecosystems. This is partly because continuous high-frequency water quality measurements remain relatively limited compared to more intermittent survey-based observations of vertical water quality profiles and targeted grab samples. Low-frequency measurements may misrepresent water quality (Lucas et al. 2006), particularly in locations with high amplitude diel-cycling. Little is known about diel-cycling dynamics in subtropical estuaries, which are less studied than temperate estuaries (Wenner et al. 2004; Cloern et al. 2014). The coastal zone of the Gulf of Mexico includes approximately 35 subtropical estuaries that are susceptible to seasonal hypoxia due to common characteristics such as low tidal energy, high solar irradiance, and warm sea temperatures (Murrell et al. 2007). Although episodic and seasonal hypoxia have been recognized in several of these systems (Hagy and Murrell 2007; Park et al. 2007; Xia and Jiang 2015), drivers of diel-cycling hypoxia remain poorly characterized.

This paper aims to improve our understanding of the processes affecting spatiotemporal variations in dissolved oxygen dynamics within Escambia Bay, a microtidal estuary in the northeastern Gulf of Mexico. Because Escambia Bay is shallow, mesotrophic, and often stratified, hypoxia is largely modulated by physical processes. Wavelet analysis has been applied previously to analyze water quality variables in estuaries, where cyclical processes occur at a variety of timescales (e.g., Ganju et al. 2020; Nezlin et al. 2009). Here, we applied a continuous wavelet transform (cwt) to analyze diel variations in physical and water quality parameters over multiple summers to provide additional context on processes and time scales driving hypoxia. Wavelet coherence and phase lag information proved useful in determining dominant forcings of diel-cycling when multiple processes (e.g., tides, sea breeze, solar irradiance) vary at or near diel timescales. The relevance of results to other systems and implications for water quality monitoring is discussed.

Methods

Study Area

Water quality data collected in Escambia Bay were used to evaluate the frequency and duration of hypoxia concurrent with physical forcing measurements. Escambia Bay is part of the Pensacola Bay system (Fig. 1), a 370 km2 estuarine complex in northwest Florida that drains a 18,000 km2 watershed. The largest portion of the watershed is drained by the Escambia River (11,000 km2), which discharges freshwater into Escambia Bay (Fig. 1). Moderate to high discharge (~150–500 m3 s−1) in the springtime forces strong stratification that contributes to the development of hypoxia in the middle reach of Escambia Bay (Hagy and Murrell 2007). Extremely high discharge (> 500 m3 s−1) can displace saltwater completely from the Bay, reducing the area of hypoxia and moving it seaward (Hagy and Murrell 2007). Tides in the Bay are low amplitude, with a mean range of 0.37 m (NOAA 2020). The water level spectrum is dominated by a peak in the diurnal band at K1 (23.9 h) and O1 (25.8 h) frequencies.

Figure 1.

Figure 1

Open in a new tab

Location of Escambia Bay in (a) northwest Florida, USA and (b) Pensacola Bay System composed of Escambia Bay (ES), Pensacola Bay (PB), Blackwater Bay (BB), and East Bay (EA) regions. Red bounding box corresponds to study site shown in (c). Black circles mark location of sampling stations listed in Table 1. Colors in (b) and (c) correspond to bottom elevation (m) relative to mean lower low water (MLLW). Red circle marks location of NOAA station

Observational Data

High-frequency continuous monitoring data were collected at four stations across multiple years to quantify spatiotemporal variations in water quality along and across the main axis of Escambia Bay (Fig. 1). From May to August 2014, measurements were collected in the upper and middle reaches of the bay at P2 and P5, respectively (Table 1). P2 is approximately 9.0 km northwest of P5. From June to August 2016, measurements were again collected at P5, as well as P5M and P5E to resolve cross-estuary gradients in water quality. P5M and P5E are located approximately 1.2 and 2.1 km east of P5, respectively. Bottom elevation increases from P5 to P5E and P2 (Table 1; Fig. 1).

Table 1.

Duration, location, and average depth of WQM sensor at each station surveyed in 2014 and 2016

Year Sampling dates Station description
Name Latitude Longitude Avg. depth
2014 May 7–Aug 28 P5 30.457 −87.132 −3.73 m
P2 30.538 −87.157 −1.64 m
2016 June 2–Aug 28 P5 30.457 −87.132 −3.85 m
P5M 30.456 −87.120 −3.62 m
P5E 30.458 −87.110 −2.92 m
Open in a new tab

WET Labs® water quality monitors (WQMs) were deployed at a height less than 0.5 m above bottom at each station to measure DO, temperature, pressure, salinity, and turbidity. WQMs burst sampled for 1 min at 1 Hz every 30 min. In our analyses, we used the burst-averaged timeseries, which had a sampling frequency (fs) of 2 h−1. From measurements of temperature, pressure, and salinity, we computed timeseries of density (σt). At P5, an additional buoy-mounted WQM fixed approximately 0.5 m below the water surface allowed us to quantify the surface to bottom density difference, Δσt, an approximation of vertical density stratification. In 2016, photosynthetically active radiation (PAR) sensors (ECO-PAR™) were deployed with the WQMs. Furthermore, in 2016, timeseries of vertical DO profiles were resolved using an array of six miniDOT (PME, Inc©) loggers. The loggers sampled DO every 15 min and were deployed every ~0.5 m along a chain that extended upward from 0.5 m above bottom to a subsurface float ~0.5 m below mean low tide. Three HOBO pressure sensors set along the vertical chain were used to confirm miniDOT elevations throughout the deployment.

The WQMs were calibrated by the manufacturer before and after deployment, and instruments were exchanged once during the monitoring period to limit biofouling. No biofouling was visually observed on sensor faces or pump inlets upon retrieval from deployment. Sensor performance and data quality were assessed prior to conducting timeseries analysis, and no data issues relating to biofouling, such as oxygen drift, were encountered.

In our analyses, we examine the relationship between DO and four physical parameters: freshwater discharge, tides, wind, and irradiance. Discharge data for the Escambia River (fs = 4 h−1) was obtained from the U.S. Geological Survey gage near Molino, FL (Site 02376033; USGS 2020). Wind and water level data (fs = 10 h−1) were obtained from the National Oceanic and Atmospheric Administration (NOAA) station (8729840) in Pensacola Bay (Fig. 1; NOAA 2020). Irradiance data (fs = 1 h −1) was obtained from ERA5 (C3S 2020), a climate reanalysis product, for the subregion closest to Escambia Bay (30° 21′ 0″–87° 15′ 0″). All physical parameters were linearly interpolated to the WQM time vector.

Wavelet Analysis

Water quality and physical timeseries were decomposed into the time–frequency domain using a cwt (Daubechies 1990, 1992; Mallat 1989). This method convolves the observed timeseries with a wavelet function that is shifted and dilated to estimate variability at different frequencies through time. Wavelet functions are useful for finding localized, intermittent periodicities in nonstationary geophysical timeseries (Liu 1994; Kumar and Foufoula-Georgiou 1997). For example, the cwt magnitude of DO at period = 24 h provides a measure of DO variability at diel timescales and can therefore be used to analyze changes in diel-cycling amplitude as a function of time. Here, the cwt was obtained using a Morlet wavelet. Values influenced by edge effects (i.e., cone of influence region) were excluded from analyses (Torrence and Compo 1998).

From the cwt, we computed the magnitude-squared wavelet coherence and phase lag between water quality and physical parameters. The wavelet coherence varies between 0 (no relationship) and 1 (perfect relationship) and is a measure of the correlation between two parameters (e.g., DO and water level) in the time–frequency domain. A Monte Carlo simulation was used to estimate the probability value of coherence, deemed significant at α=.05.α=.05. To carry out this simulation, we first computed autoregressive (AR1) coefficients for the input parameters. Next, pairs (n = 1000) of red noise timeseries were generated using the AR1 coefficients and length of the input timeseries. We then calculated the wavelet coherence for each red noise pair, which allowed us to estimate the significance of the original coherence (e.g., between DO and water level) against a red noise process. An overview and MATLAB code were provided by Grinsted et al. (2014).

Results

Physical Parameters

Timeseries of physical parameters included in our analyses are shown in Fig. 2. Peak freshwater discharge in May 2014 (890 m3 s−1) was associated with a historic flash flooding event (Fig. 2a). Maximum discharge during this event was almost two orders of magnitude larger than mean discharge from June to August in both years (~ 75 m3 s−1). Mean water level (relative to mean sea level) in 2016 was 0.1 m higher than in 2014 (Fig. 2b). Mean tidal range was approximately the same in both years, as were the distributions of wind speed and direction. The north–south (N-S) wind vector typically has larger magnitude than the east–west vector and captures most of the wind variance (Fig. 2c, d). It was therefore the primary wind vector utilized in our analyses. Mean peak daily irradiance was also similar across years: 805 and 763 W m−2 in 2014 and 2016, respectively (Fig. 2e).

Figure 2.

Figure 2

Open in a new tab

Timeseries of physical parameters for the period of observation in 2014 and 2016. (a) Escambia River discharge (m3 s−1); (b) water level (m) relative to mean sea level; (c) north–south 4-h moving average winds (m s−1). Positive values are from the south; (d) east–west 4-h moving average winds (m s−1). Positive values are from the west; (e) hourly downward surface solar irradiance (W m−2)

The global wavelet spectrum for each parameter shows the time-averaged cwt magnitude as a function of period (Fig. 3). The dominant spectral peak for freshwater discharge occurred around 700 h in both 2014 and 2016 but was an order of magnitude larger in 2014 due to the flood event (Fig. 3a). For water level, wind, and irradiance, the dominant spectral peak was close to 24 h. Tides in the Bay are diurnal, so this agrees with expectations. The cwt magnitude at 24 h is positively correlated with tidal range, and this explains the close agreement among spectra from 2014 to 2016 (Fig. 3b). For wind, cwt magnitude at 24 h depended on diel fluctuations in wind direction and magnitude. The cwt magnitude was highest when northerly winds in the early morning shifted to strong southerly winds by the afternoon. The spectral peak at 24 h suggests that sea breeze is the dominant component of the wind, wherein strong winds blow landward from the Gulf of Mexico in the late afternoon (Fig. 3c). From the magnitude of the peak, we can infer that the sea breeze was slightly stronger in 2016 relative to 2014. Furthermore, irradiance was similar across years, as shown by the close agreement between spectra (Fig. 3d).

Figure 3.

Figure 3

Open in a new tab

(a–d) Time-averaged cwt magnitude as a function of period (h) for each physical parameter. Values within the cone of influence region were excluded from the average. Dotted horizontal line marks period = 24 h

Dissolved Oxygen

Water quality measurements showed large interannual variations in low DO conditions in the bottom layer of the middle reach of Escambia Bay. In 2014, 78% of DO concentration measurements at P5 were less than the conventional 2 mg L−1 threshold, compared to 30% in 2016 (Fig. 4a, c). The lowest incidence of hypoxia was observed in the upper reach of Escambia Bay at P2, where only 6% of measurements were less than 2 mg L−1 (Fig. 4a). Accordingly, mean DO was generally 3.5 times higher at P2 relative to P5 from June to August 2014 (Fig. 4a). However, during the May 2014 flood, high freshwater discharge pushed the salt front seaward and bottom DO at the two stations was similar. In 2016, 20% and 12% of measurements were less than 2 mg L−1 at P5M and P5E, respectively (Fig. 4c). Therefore, as bottom elevation increased from P5 to P5E, the incidence of hypoxia decreased.

Figure 4.

Figure 4

Open in a new tab

Timeseries of (a) DO (mg L−1) and (b) cwt magnitude of DO at period = 24 h in 2014 and (c, d) 2016. Horizontal dashed line in (a, c) marks 2 mg L−1 hypoxic threshold. Distribution of maximum daily DO as a function of cwt magnitude in (e) 2014 and (f) 2016

The timing of diel fluctuations in DO was similar across sampling stations. On an average day, minimum DO occurred in the morning (6–10 AM) and increased throughout the day, reaching peak concentration in the late afternoon (4–8 PM; Figure S1). We investigated the amplitude of diel-cycling DO using the cwt magnitude at 24 h. Prior to computing the cwt, DO timeseries were normalized to zero mean and unit variance, which permitted comparisons across sampling stations. In 2014, mean cwt magnitude was similar at P2 (0.25) and P5 (0.21); however, timeseries of cwt magnitude were almost uncorrelated (Pearson’s r(5331) = −.03, p = .031; Fig. 4b). This suggests that while the average amplitude of diel fluctuations was similar, the biophysical processes controlling diel-cycling varied between the middle (P5) and upper (P2) bay. In 2016, mean cwt magnitude increased from P5 to P5M and P5E (0.35, 0.38, 0.51), indicating higher amplitude diel-cycling at the shallow sites (Fig. 4d). Timeseries of cwt magnitude at P5M was significantly correlated with P5 (r(4085) = .44, p < .001) and P5E (r(4085) = .56, p < .001). On the other hand, timeseries of cwt magnitude at P5 and P5E were weakly correlated (r(4085) = −.11, p < .001). This implies that for a set of forcing conditions, diel-cycling varies across the middle Bay. For example, from July 3 to 6, 2016 large diel fluctuations in DO at P5 coincided with limited diel-cycling at P5M and P5E. In subsequent sections we analyze the effects of physical processes on temporal and spatial variations in diel-cycling.

Temporal Variations in Diel-Cycling Dynamics

Data collected at P5 in 2014 and 2016 allowed us to examine interannual variations in diel-cycling DO for a range of forcing conditions. For both years, the amplitude of diel-cycling increased when south-southwesterly (SSW) winds coincided with large tidal range (i.e., tropic tides; Fig. 5). Onshore winds from the SSW, consistent with sea breeze, are directed up-estuary in Escambia Bay and typically result in reduced vertical shear and stratification (Scully et al. 2005). Tropic tides also enhance tidal mixing and the breakdown of stratification (Simpson et al. 1990).

Figure 5.

Figure 5

Open in a new tab

(a, b) Cwt magnitude at period = 24 h for DO recorded at P5 as a function of median wind direction and tidal range when discharge was less than 150 m3 s−1. Boxplots on the left and right correspond to tidal ranges ≤ 0.3 m and > 0.3 m, respectively. The bottom and top of each box represent the 25th and 75th percentiles, respectively, while whiskers indicate minimum and maximum values

An example of a high-amplitude diel-cycling event is shown in Fig. 6. From July 3 – 6, 2016, surface and bottom layers mixed when strong SSW winds in the afternoon coincided with ebbing tropic tides. In the overnight hours that followed, bottom DO declined in part due to the relaxation of winds and flooding tide. This is consistent with concurrent ADCP measurements that showed decreased flow speeds during flood tide, which likely dampened tidal mixing. In the days following this event (July 8 – 12), DO steadily declined at depth due to weakened sea breeze and equatorial tides, which allowed stratification to reestablish (as indicated by increasing Δσt).

Figure 6.

Figure 6

Open in a new tab

(a) 4-h moving average N-S and E-W winds (m s−1). Positive values indicate winds from the south or west. (b) Water level (m) relative to mean sea level. (c) Density difference (ΔσtΔσt, kg m−3) between the surface and bottom layers at P5. (d) Contour plot of DO (mg L−1) as a function of height above bottom (m) recorded over a two-week period in 2016 at P5. Red triangles correspond to height of miniDOT sensors

In the absence of flooding diel-cycling amplitude was lower in 2014 compared to 2016 for given tidal and wind conditions (Fig. 5). To better understand this discrepancy, we computed the wavelet coherence between water quality parameters (σt, DO) and physical parameters (water level, wind). The coherence between σt and water level at period = 24 h was similar in 2014 and 2016, whereby 39% and 44% of coherence values were statistically significant (Table 2). Near-zero phase lag between σt and water level is consistent with tidal advection, wherein peak density occurs around high tide. If tidal advection modulates bottom DO at P5, we could expect high coherence between DO and water level and a near 12 h phase lag. In 2014, mean ∣phase lag∣ between DO and water level was 10.6 h and 6% of coherence values were significant (Table 2; Fig. 7a). This suggests that the few high amplitude diel-cycles in 2014 may have been due to tides advecting water masses with different DO past the WQM. The infrequency of significant coherence between DO and wind in 2014 suggests that wind mixing did not influence diel-cycling (Table 2; Fig. 7c). In 2016, we can infer from the relatively high coherence and shorter ∣phase lag∣ (9.88 h) that tidal mixing was likely an important driver of diel-cycling (Fig. 7b). The coherence indicates that diel-cycling DO was less correlated with wind than tides in 2016 (Fig. 7d). Overall, higher coherence between DO and both wind and tides was observed in 2016 compared to 2014 due in part to higher amplitude DO diel-cycling in 2016.

Table 2.

Mean ± standard deviation of statistically significant phase lag (∣h∣) at period = 24 h. The percentage of coherence values that were significantly different from red noise at 5% significance level is also reported

Water Level Wind Irradiance
DO 2014 P2 7.39 ± 2.91 (23%) 3.52 ± 2.00 (13%) 5.81 ± 1.70 (27%)
P5 10.6 ± 2.37 (6%) 3.31 ± 2.10 (4%) 3.89 ± 3.35 (8%)
2016 P5 9.88 ± 1.01 (33%) 4.86 ± 2.15 (20%) 7.87 ± 1.84 (32%)
P5M 7.82 ± 1.80 (24%) 2.87 ± 1.51 (29%) 6.07 ± 1.41 (34%)
P5E 8.79 ± 2.47 (30%) 2.61 ± 1.57 (37%) 5.34 ± 1.12 (49%)
σ t 2014 P2 1.72 ± 2.12 (18%) 6.42 ± 3.06 (22%)
P5 1.08 ± 0.63 (39%) 5.65 ± 1.56 (31%)
2016 P5 0.80 ± 0.55 (44%) 4.79 ± 1.84 (33%)
P5M 1.92 ± 1.24 (25%) 6.49 ± 1.95 (36%)
P5E 1.84 ± 1.27 (22%) 6.91 ± 1.84 (35%)
Open in a new tab

Figure 7.

Figure 7

Open in a new tab

Magnitude–squared coherence between (a, b) DO and water level and (c, d) DO and N–S wind at P5 in 2014 (a, c) and 2016 (b, d). Shaded region corresponds to cone of influence. Thick contour line designates 5% significance level against red noise. Red dashed line designates 24 h period

From June to August 2014, limited diel-cycling DO and persistent hypoxia at P5 were attributable to prolonged density stratification that lasted throughout summer, resulting in Δσt greater than 20 kg m−3 (Fig. 8). Enhanced freshwater discharge from the Escambia River in the late spring resulted in broad distribution of low salinity surface waters and enhanced stratification, as previously observed by Hagy and Murrell (2007; Fig. 8). These results suggest that while tropic tides and onshore winds can overcome weak to moderate stratification (Δσt < 12 kg m−3), stratification is more resistant to breakdown when Δσt exceeds 15 kg m−3.

Figure 8.

Figure 8

Open in a new tab

Distributions of temperature (°C) and salinity (ppt) in (a, b) surface and (c, d) bottom layers at P5 in 2014 and 2016. (e) Distribution of the density difference (ΔσtΔσt) between the surface and bottom layers. Positive values indicate higher density in the bottom layer. (f) Distribution of ΔσtΔσt as a function of wind direction when tidal range > 0.3 m, and mean daily discharge is less than 150 m3 s−1

Spatial Variations in Diel-Cycling Dynamics

Using the wavelet coherence, we explored along- and cross-estuary variations in the relationship between physical processes and water quality at diel timescales. Data collected in 2014 showed that physical drivers of diel-cycling varied between upper (P2) and middle (P5) Escambia Bay. At P2, the infrequency of significant coherence between σt and physical processes suggests limited diel variation in σt in the upper Bay relative to the middle Bay (Table 2). From the short phase lag between DO and water level, we can infer that diel-cycling DO at P2 was not due to tidal advection. Coherence and phase information also suggest that wind likely had a limited impact on diel-cycling DO at P2 (∣phase lag∣= 3.52 h; 13% of coherence values significant), but that some diel variation may be attributable to local cycles of net production and respiration (∣phase lag∣= 5.81 h; 27% of coherence values significant).

Data collected in 2016 showed a cross-estuary gradient in diel-cycling in the mid-Bay, wherein mean diel-cycling amplitude increased from P5 to P5E. The coherence between σt and water level showed lower correlation and longer, more variable phase lag at P5M and P5E, which indicated the lessening influence of tidal advection with distance from the channel (Table 2). This is consistent with the observed horizontal density gradient, whereby σt¯¯¯¯¯σt¯ decreased from P5 (15.8 kg m−3) to P5E (12.3 kg m−3). Greater coherence between DO and wind, relative to DO and water level, suggests that wind mixing may play a more significant role in modulating DO at the shallower sites. The reverse is true at P5, where greater coherence between water level and DO suggests that tides are more important. Coherence between DO and irradiance increased from P5 to P5E, while mean phase lag decreased to less than 6 h at P5E (Table 2). If diel-cycling was solely due to local production, we would expect a near-zero to 6 h lag between DO and irradiance (Sanford et al. 1990). For comparison, analysis of the P5 surface data from 2016 showed that 85% of coherence values between DO and irradiance were significant and mean ∣phase lag∣= 3.90 ± 0.88 h (not shown). From the coherence analysis, we can infer that irradiance has a large effect on daytime DO production in surface waters and that the relative importance of biophysical processes responsible for diel-cycling dynamics varies with depth.

Inferences from the coherence analysis were supported by the distribution of cwt magnitude as a function of wind and tide (Fig. 9). Strong sea breeze and tropic tides amplified diel-cycling at P5 in 2016 (Fig. 9a). However, at P5E, the highest amplitude diel-cycling occurred when weak sea breeze coincided with equatorial tides (Fig. 9c). Given the high coherence and shorter lag between DO and irradiance, local production at or near P5E may force diel-cycling during weak wind and tidal forcing. Across periods with varying tidal stage and range, measurements from P5E show that average turbidity increased to 7.54 NTU and average PAR decreased to 26.9 μmol photons m−2 s−1 during strong sea breeze (> 6 m s−1) compared to periods with less wind, when mean turbidity was 5.79 NTU and mean PAR was 35.9 μmol photons m−2 s−1. Decreased PAR would be expected to result in decreased daytime production, while increased wind would be expected to increase air-sea gas exchange, both of which would reduce the magnitude of diel-cycling.

Figure 9.

Figure 9

Open in a new tab

(a–c) Cwt magnitude at period = 24 h in 2016 as a function of tidal range and wind speed when median wind direction is 180–225°. Boxplots on the left and right correspond to tidal ranges ≤ 0.3 m and > 0.3 m, respectively. For each tidal range group, boxplots on the left and right correspond to low (L; ≤ 6 m s−1) and high (H; > 6 m s−1) wind speeds, respectively

Discussion

Dissolved Oxygen Dynamics in Escambia Bay

Biological processes are an essential driver of hypoxia in aquatic ecosystems and in most cases are largely responsible for long-term trends in relation to nutrient loading (Diaz and Rosenberg 2008). However, physical processes strongly modulate DO dynamics and may dominate variability at some spatial and temporal scales (e.g., Park et al. 2007; Scully 2013). Our findings suggest that biophysical processes affecting diel-cycling DO vary across the bay, including over short spatial extents on the order of 1 to 2 km and with relatively small variations in bottom elevation of 1 m or less. The influence of water depth on wind driven mixing and light attenuation is an important factor for understanding the relative contributions of biophysical processes to diel-cycling dynamics.

In upper Escambia Bay (P2), bottom DO is generally higher due to frequent freshening by discharge from the Escambia River. Limited diel-cycling DO at the upper Bay site is consistent with low-frequency freshwater forcing with associated high turbidity (10.2 NTUs) and low algal biomass (Murrell et al. 2009). The mean ∣phase lag∣ of 5.81 h between DO and irradiance with 27% coherence values significant (Table 2) may be indictive of effects of benthic algae in the shallow upper Bay. In deeper portions of the middle Bay, stratification strength and mixing intensity strongly influence temporal variations in bottom DO. For example, strong stratification in 2014 persisted throughout the summer and was resistant to breakdown, which dampened diel-cycling and prolonged hypoxia. Under weak to moderate stratification, diel-cycling DO intensifies and can be driven by a combination of tidal and wind mixing and, to a lesser extent, tidal advection. Our results show reduced tidal influence with distance from the main channel and enhanced wind mixing and local DO production in shallower water. While other factors, such as temperature, influence DO dynamics and ecosystem metabolism, DO diel-cycling due to daytime net production and nighttime net respiration is driven by the daily timing of solar irradiance. The highest amplitude diel-cycling occurred in shallower portions of the middle Bay with mean ∣phase lag∣ of 5.34 h between DO and irradiance and 49% coherence values significant (Table 2), suggesting strong association with light availability and local production. The coincidence of high surface-layer production and sea breeze in the afternoon may increase the magnitude of turbulent oxygen flux through the pycnocline.

It should be noted that ascribing patterns of variability to biophysical processes is often difficult when multiple processes (e.g., sea breeze, diurnal tides, irradiance) vary at similar timescales, as is the case in Escambia Bay. This is a challenge because when two timeseries vary near the same frequency, they could be highly correlated yet unrelated. For example, if DO varies at diel timescales due only to advection of hypoxic water by diurnal tides, we would expect high coherence between DO and water level and a consistent 12-h phase lag. We would expect that a percentage of the coherence between DO and irradiance will still be significant even though the processes are unrelated because they both vary near the same frequency. However, the phase lag between DO and irradiance would be highly variable and longer than expected if DO were solely attributable to local production. This is in part due to the fact that while irradiance and tides vary at similar frequencies, they do not vary at the same frequency. Therefore, while peak DO production occurs at a similar time each day, peak DO due to tidal advection varies daily depending on the timing of high and low tides.

This study used wavelet coherence to examine coupling between water quality parameters and biophysical processes, which performed better in this application than previously applied correlation metrics. For example, correlation coefficients (Iriarte et al. 2015; Villate et al. 2013; Kim et al. 2020) and cross-spectra (Sanford et al. 1990) often fail when the statistical properties of the timeseries or correlation between two timeseries change over time. Given that wavelets are localized functions, the wavelet coherence is useful for finding intermittent correlations between two timeseries in the time–frequency domain. Given that factors affecting DO diel-cycling vary in time and space, multiple processes should be considered when trying to determine potential forcing mechanisms from coherence and phase lag information. Combined with increasing collection and availability of continuous water quality data, wavelet analysis is a useful analytical approach that can improve our understanding of spatial and temporal water quality dynamics in coastal ecosystems.

Broader Relevance to Pensacola Bay and Other Ecosystems

The dynamic interactions between biophysical processes affecting DO patterns in Escambia Bay are likely important in other areas of the Pensacola Bay system. Key characteristics of Pensacola Bay like shallow depth, variable freshwater inflows, density stratification, and periodic wind forcing are shared by many estuaries. The effect of wind and tidal mixing on bottom DO has been documented in other shallow microtidal systems, including Gulf of Mexico estuaries like Mobile Bay and Corpus Christi Bay and Atlantic coast estuaries like Pamlico River estuary and Neuse River estuary (Applebaum et al. 2005; Park et al. 2007; Lin et al. 2008; Borsuk et al. 2001). On the other hand, very large and relatively deep estuaries like Chesapeake Bay are overall less dynamic. Stratification is seasonally stable with deeper portions of the Chesapeake Bay resisting turnover even during hurricane events (Cho et al. 2012). A comparison can also be made with nearshore waters of Louisiana, whereby connectivity with the continental shelf hypoxic zone of the Gulf of Mexico is facilitated by wind-driven bottom water upwelling on timescales of several days to weeks (Jarvis et al. 2020). Although the Pensacola Bay system is microtidal, our results show that larger amplitude tropic tides increased vertical mixing in areas close to the tidal channel. Whereas previous studies have noted that low-amplitude tides contribute to more stratification and hypoxia than larger amplitude tides (Hagy and Murrell 2007; Park et al. 2007), our analysis illustrates that tides still modulate stratification strength and associated hypoxia.

However, wind and tidal mixing of surface and bottom layers are limited during periods of strong stratification that follow periodic flooding, which occurred in 2014 and is a signature characteristic of northeastern Gulf Coast estuaries (Turner 2001). In the middle reach of Escambia Bay, the average surface to bottom density gradient (Δσt) in 2014 was 12.8 kg m−3 and peak Δσt was greater than 22.0 kg m−3. This surpasses vertical gradients documented in other systems, such as Newport Bay, Narragansett Bay, and Neuse River estuary, where Δσt rarely exceeds 15 kg m−3 and is typically less than 10 kg m−3 (Nezlin et al. 2009; Codiga 2012; Buzzelli et al. 2002). During flood events, high freshwater discharge from the Escambia River displaces saltwater completely from upper Escambia Bay (Hagy and Murrell 2007). This enhances stratification in the middle Bay and differs from observed spatial patterns in estuaries with low inflow, like Newport Bay, where stratification is strongest near the head of the estuary (Nezlin et al. 2009). In relation to the rest of the Pensacola Bay system, CTD casts collected from 2009 to 2012 show that vertical density gradients are typically stronger in Escambia Bay than Blackwater Bay and East Bay. Therefore, eastern portions of the Pensacola Bay system may be more dynamic and less susceptible to prolonged hypoxia.

Aliasing Diel Variability in Long-term Monitoring Trends

Assessment of hypoxia and its associated ecological effects requires measurements that capture the variability in DO within an ecosystem. Therefore, the frequency at which DO is sampled should be higher than the dominant frequency of variability. However, high-frequency variability is often not considered in monitoring plans, and evaluation of water quality is often based on intermittent survey-based sampling. While less frequent sampling may be appropriate when DO varies at longer timescales, it may misrepresent water quality in estuaries with diel-cycling due to high-frequency aliasing. Here, we demonstrate bias in long-term water quality trends introduced by high amplitude diel-cycling. Subsampling routines were determined from water quality criteria established by the state of Florida (rule 62–302.533; Online Resource 2), which are based on mean percent (%) saturation at daily, weekly, and monthly timescales. Timeseries of DO, salinity, and temperature collected at each sampling station were used to quantify % DO saturation. Average weekly and monthly values were computed using 3 days of diel data, as specified by the criteria. For each averaging period, a minimum (maximum) value was assessed using the 3 days with the lowest (highest) mean % saturation. This provided a range of possible % saturation values one could obtain by subsampling DO.

At all stations, the 10% allowance threshold for daily values less than 42% saturation was exceeded (Fig. 10a, b). Attainment of 7- and 30-day criteria varied by station, averaging period, and diel days included in the mean. For the 7-day criterion, minimum mean values based on days with the lowest % saturation would not attain the criterion at any of the stations. Depending on the start date of the averaging period, maximum mean values based on days with the highest % saturation could satisfy the 7-day criterion at P2, P5M, and P5E (Fig. 10c, e). For the 30-day criterion, P5 in 2014 does not meet the criterion regardless of diel days included in the mean (Fig. 10d). For all other stations, maximum mean values do satisfy the criterion, while minimum mean values do not (Fig. 10d, f).

Figure 10.

Figure 10

Open in a new tab

(a, b) Empirical cumulative distribution function of daily mean % DO saturation in 2014 and 2016. The vertical dashed line marks the 42% saturation threshold, and the horizontal dashed line marks the 10% allowance threshold. (c, d) Timeseries of 7-day and 30-day mean % DO saturation in 2014 and (e, f) 2016. For each start date of the averaging period, minimum (small circles) and maximum (large circles) values are shown. Horizontal dashed lines designate the 51 and 56% saturation thresholds for 7- and 30-day criteria, respectively

The difference between minimum and maximum mean values generally increased with length of averaging period and diel-cycling amplitude. In 2016, the average difference between minimum and maximum 30-day means based on diel data was 51, 63, and 56% at P5, P5M, and P5E (Fig. 10f). This difference would be larger if means were based on grab samples, particularly if the samples were collected around the same time of day as is common practice. For example, average 30-day mean % saturation at P5 was 17 versus 99% when computed from the 10 lowest or highest samples collected on 10 different days, as specified by the rule (Online Resource 2). This discrepancy is larger at stations with high amplitude diel-cycling such as P5E, where the average 30-day mean was 27 versus 132% when computed from the 10 lowest or highest samples.

This exercise demonstrates the potential for intermittent sampling schemes to misrepresent average weekly or monthly water quality due to high amplitude diel fluctuations in DO. In Pensacola Bay and similar systems, sampling schemes that resolve diel variability will provide a more informative characterization of water quality patterns. Time–frequency analysis using wavelets applied to an initial data collection can inform future monitoring efforts by identifying key processes modulating DO.

Conclusions

Continuous water quality monitoring during summer 2014 and 2016 allowed us to examine processes affecting bottom DO at different locations within a shallow, subtropical estuary. The continuous wavelet transform helped illuminate changes in diel-cycling amplitude over time, and the wavelet coherence illustrated correlations between DO and several biophysical processes occurring at similar diel timescales. Freshwater discharge modulated inter-annual variations in the spatial extent and duration of summertime hypoxia through its effect on vertical density stratification. Diel fluctuations in bottom DO were attributable to local cycles of net production and respiration as well as mixing by diurnal tides and sea breeze. Our results show that the processes affecting diel-cycling DO can vary over a relatively short spatial extent and with relatively small changes in bottom elevation. Given the physical characteristics of the Pensacola Bay system, diel-cycling DO likely pervades much of the estuary. Future monitoring efforts should focus on continuous sampling that resolves variance in DO at diel timescales since the ecological effects of hypoxia are likely frequency dependent.

Supplementary Material

Supplement1
Supplement2

Acknowledgements

This research was supported in part by a postdoctoral appointment to the U.S. Environmental Protection Agency (USEPA) Research Participation Program administered by the Oak Ridge Institute for Science and Education (ORISE) through an interagency agreement between the U.S. Department of Energy (USDOE) and the USEPA. ORISE is managed by Oak Ridge Associated Universities (ORAU) under DOE contract number DE-SC0014664. We acknowledge the significant contributions from many individuals involved in long-term monitoring efforts of Pensacola Bay. We thank the contributions of many USEPA staff for their efforts at producing quality data in the field and lab.

Funding

This study was funded, reviewed, and approved for publication by the USEPA, Office of Research and Development, Center for Environmental Measurement and Modeling, Gulf Ecosystem Measurement and Modeling Division.

Footnotes

Disclaimer

The views expressed in this article are those of the authors and do not necessarily reflect the views or policies of the USEPA, USDOE, or ORAU/ORISE. Use of trade names or commercial products does not constitute endorsement by the USEPA.

References

  1. Applebaum S, Montagna PA, and Ritter C. 2005. Status and trends of dissolved oxygen in Corpus Christi Bay, Texas, U.S.A. Environmental Monitoring and Assessment. 10.1007/s10661-005-3111-5. [DOI] [PubMed] [Google Scholar]
  2. Beck MW, Hagy JD III, and Murrell MC. 2015. Improving estimates of ecosystem metabolism by reducing effects of tidal advection on dissolved oxygen time series. Limnology and Oceanography: Methods. 10.1002/lom3.10062. [DOI] [Google Scholar]
  3. Beck NG, and Bruland KW. 2000. Diel biogeochemical cycling in a hyperventilating shallow estuarine environment. Estuaries. 10.2307/1352825. [DOI] [Google Scholar]
  4. Borowiec BG, Darcy KL, Gillette DM, and Scott GR. 2015. Distinct physiological strategies are used to cope with constant hypoxia and intermittent hypoxia in killifish (Fundulus heteroclitus). Journal of Experimental Biology. 10.1242/jeb.114579. [DOI] [PubMed] [Google Scholar]
  5. Borsuk ME, Stow CA, Luettich RA, Paerl HW, and Pinckney JL. 2001. Modelling oxygen dynamics in an intermittently stratified estuary: Estimation of process rates using field data. Estuarine, Coastal and Shelf Science. 10.1006/ecss.2000.0726. [DOI] [Google Scholar]
  6. Brady DC, and Targett TE. 2013. Movement of juvenile weakfish Cynoscion regalis and spot Leiostomus xanthurus in relation to diel-cycling hypoxia in an estuarine tidal tributary. Marine Ecology Progress Series. 10.3354/meps10466. [DOI] [Google Scholar]
  7. Bricker SB, Clement CG, Pirhalla DE, Orlando SP, and Farrow DRG. 1999. National estuarine eutrophication assessment: effects of nutrient enrichment in the Nation’s estuaries. Silver Spring, MD: NOAA, NOS, Special Projects Office and the National Centers for Coastal Ocean Science. [Google Scholar]
  8. Buzzelli CP, Leuttich RA, Powers SP, Peterson CH, McNinch JE, Pinckney JL, and Paerl HW. 2002. Estimating the spatial extent of bottom-water hypoxia and habitat degradation in a shallow estuary. Marine Ecology Progress Series 230: 103–112. [Google Scholar]
  9. Cho KH, Wang HV, Shen J, Valle-Levinson A, and Teng YC. 2012. A modeling study on the response of Chesapeake Bay to hurricane events of Floyd and Isabel. Ocean Modelling. 10.1016/j.ocemod.2012.02.005. [DOI] [Google Scholar]
  10. Cloern JE 2001. Our evolving conceptual model of the coastal eutrophication problem. Marine Ecology Progress Series. 10.3354/meps210223. [DOI] [Google Scholar]
  11. Cloern J, Foster S, and Kleckner A. 2014. Phytoplankton primary production in the world’s estuarine-coastal ecosystems. Biogeosciences. 10.5194/bg-11-2477-2014. [DOI] [Google Scholar]
  12. Codiga DL 2012. Density stratification in an estuary with complex geometry: Driving processes and relationship to hypoxia on monthly to inter-annual timescales. Journal of Geophysical Research. 10.1029/2012JC008473. [DOI] [Google Scholar]
  13. Copernicus Climate Change Service (C3S). (2020). ERA5: fifth generation of ECMWF atmospheric reanalyses of the global climate. Copernicus Climate Change Service Climate Data Store (CDS), accessed Nov 25, 2020, at https://cds.climate.copernicus.eu/cdsapp#!/home. [Google Scholar]
  14. Dan X, Yan G, Zhang A, Cao Z, and Fu S. 2014. Effects of stable and diel-cycling hypoxia on hypoxia tolerance, postprandial metabolic response, and growth performance in juvenile qingbo (Spinibarbus sinensis). Aquaculture. 10.1016/j.aquaculture.2014.02.025. [DOI] [Google Scholar]
  15. Daubechies I 1990. The wavelet transform, time-frequency localization and signal analysis. IEEE Transactions on Information Theory 36: 961–1005. [Google Scholar]
  16. Daubechies I (1992). Ten lectures on wavelets. CBMS-NSF Regional Conference Series in Applied Mathematics. Philadelphia: Society for Industrial and Applied Mathematics (SIAM). [Google Scholar]
  17. D’Avanzo C, and Kremer JN. 1994. Diel oxygen dynamics and anoxic events in an eutrophic estuary of Waquoit Bay Massachusetts. Estuaries 17 (1): 131–139. [Google Scholar]
  18. Diaz RJ, and Rosenberg R. 2008. Spreading dead zones and consequences for marine ecosystems. Science. 10.1126/science.1156401. [DOI] [PubMed] [Google Scholar]
  19. Fennel K, and Testa JM. 2019. Biogeochemical controls on coastal hypoxia. Annual review of marine science. 10.1146/annurev-marine-010318-095138. [DOI] [PubMed] [Google Scholar]
  20. Ganju NK, Testa JM, Suttles SE, and Aretxabaleta AL. 2020. Spatiotemporal variability of light attenuation and net ecosystem metabolism in a back-barrier estuary. Ocean Science. 10.5194/os-16-593-2020. [DOI] [Google Scholar]
  21. Grinsted A, Moore JC, and Jevrejeva S. 2014. Application of the cross wavelet transform and wavelet coherence to geophysical time series. Nonlinear Processes Geophysics. 10.5194/npg-11-561-2004. [DOI] [Google Scholar]
  22. Hagy JD III., and Murrell MC. 2007. Susceptibility of a northern Gulf of Mexico estuary to hypoxia: An analysis using box models. Estuarine, Coastal and Shelf Science. 10.1016/j.ecss.2007.04.013. [DOI] [Google Scholar]
  23. Hrycik AR, Almeida LZ, and Höök TO. 2017. Sub-lethal effects on fish provide insight into a biologically-relevant threshold of hypoxia. Oikos. 10.1111/oik.03678. [DOI] [Google Scholar]
  24. Iriarte A, Villate F, Uriarte I, Alberdi L, and Intxausti L. 2015. Dissolved oxygen in a temperate estuary: the influence of hydro-climatic factors and eutrophication at seasonal and inter-annual time scales. Estuaries and Coasts. 10.1007/s12237-014-9870-x. [DOI] [Google Scholar]
  25. Jarvis BM, Lehrter JC, Lowe LL, Hagy JD III, Wan Y, Murrell MC, Ko DS, Penta B, Gould RW Jr. 2020. Modeling spatiotemporal patterns of ecosystem metabolism and organic carbon dynamics affecting hypoxia on the Louisiana Continental Shelf. Journal of Geophysical Research: Oceans. 10.1029/2019JC015630. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Keppel AG, Breitburg DL, and Burrell RB. 2016. Effects of co-varying diel-cycling hypoxia and pH on growth in the juvenile eastern oyster. PLoS One. 10.1371/journal.pone.0161088. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Kim YH, Son S, Kim H-C, Kim B, Park Y-G, Nam J, and Ryu J. 2020. Application of satellite remote sensing in monitoring dissolved oxygen variabilities: A case study for coastal waters in Korea. Environment International. 10.1016/j.envint.2019.105301. [DOI] [PubMed] [Google Scholar]
  28. Kumar P, and Foufoula-Georgiou E. 1997. Wavelet analysis for geophysical applications. Reviews of Geophysics. 10.1029/97RG00427. [DOI] [Google Scholar]
  29. Lin J, Xu H, Cudaback C, and Wang D. 2008. Inter-annual variability of hypoxic conditions in a shallow estuary. Journal of Marine Systems 73: 169–184. [Google Scholar]
  30. Liu PC 1994. Wavelet spectrum analysis and ocean wind waves. In Wavelets in geophysics, ed. Foufoula-Georgiou E and Kumar P, 151–166. San Diego: Academic Press. [Google Scholar]
  31. Lucas LV, Sereno DM, Burau JR, Schraga TS, Lopez CB, Stacey MT, Parchevsky KV, and Parchevsky VP. 2006. Intradaily variability of water quality in a shallow tidal lagoon: mechanisms and implications. Estuaries and Coasts. 10.1007/BF02786523. [DOI] [Google Scholar]
  32. Mallat SG 1989. A theory for multiresolution signal decomposition: The wavelet representation. IEEE Transactions on Pattern Analysis and Machine Intelligence. 10.1109/34.192463. [DOI] [Google Scholar]
  33. Murrell MC, Campbell JG, Hagy JD III., and Caffrey JM. 2009. Effects of irradiance on benthic and water column processes in a Gulf of Mexico estuary: Pensacola Bay, Florida, USA. Estuarine, Coastal and Shelf Science. 10.1016/j.ecss.2008.12.002. [DOI] [Google Scholar]
  34. Murrell MC, Hagy JD III, Lores EM, and Greene RM. 2007. Phytoplankton production and nutrient distributions in a subtropical estuary: importance of freshwater flow. Estuaries and Coasts. 10.1007/BF02819386. [DOI] [Google Scholar]
  35. National Oceanic & Atmospheric Administration (NOAA). 2020. Water and meteorological data available on the World Wide Web (NOAA Tides and Currents), accessed Aug 12, 2020, at https://tidesandcurrents.noaa.gov/.
  36. Nezlin NP, Kamer K, Hyde J, and Stein ED. 2009. Dissolved oxygen dynamics in a eutropic estuary, Upper Newport Bay, California. Estuarine, Coastal and Shelf Science. 10.1016/j.ecss.2009.01.004. [DOI] [Google Scholar]
  37. Park K, Kim CK, and Schroeder WW. 2007. Temporal variability in summertime bottom hypoxia in shallow areas of Mobile Bay, Alabama. Estuaries and Coasts. 10.1007/BF02782967. [DOI] [Google Scholar]
  38. Rabalais NN, Díaz RJ, Levin LA, Turner RE, Gilbert D, and Zhang J. 2010. Dynamics and distribution of natural and human-caused hypoxia. Biogeosciences. 10.5194/bg-7-585-2010. [DOI] [Google Scholar]
  39. Regan MD, and Richards JG. 2017. Rates of hypoxia induction alter mechanisms of O2 uptake and the critical O2 tension of goldfish. Journal of Experimental Biology. 10.1242/jeb.154948. [DOI] [PubMed] [Google Scholar]
  40. Sanford LP, Sellner KG, and Breitburg DL. 1990. Covariability of dissolved oxygen with physical processes in the summertime Chesapeake Bay. Journal of Marine Research 48: 567–590. [Google Scholar]
  41. Scully ME 2013. Physical controls on hypoxia in Chesapeake Bay: A numerical modeling study. Journal of Geophysical Research, Oceans. 10.1002/jgrc.20138. [DOI] [Google Scholar]
  42. Scully ME, Friedrichs C, and Brubaker J. 2005. Control of estuarine stratification and mixing by wind-induced straining of the estuarine density field. Estuaries. 10.1007/BF02693915. [DOI] [Google Scholar]
  43. Simpson JH, Brown J, Matthews J, and Allen G. 1990. Tidal straining, density currents, and stirring in the control of estuarine stratification. Estuaries 13 (2): 125–132. [Google Scholar]
  44. Tyler RM, and Targett TE. 2007. Juvenile weakfish Cynoscion regalis distribution in relation to diel-cycling dissolved oxygen in an estuarine tributary. Marine Ecology Progress Series 333: 257–269. [Google Scholar]
  45. Torrence C, and Compo GP. 1998. A practical guide to wavelet analysis. Bulletin of the American Meteorological Society 79: 61–78. [Google Scholar]
  46. Turner RE 2001. Of manatees, mangroves, and the Mississippi River: Is there an estuarine signature of the Gulf of Mexico? Estuaries 24 (2): 139–150. [Google Scholar]
  47. U.S. Environmental Protection Agency. 2003. Ambient water quality criteria for dissolved oxygen, water clarity, and chlorophyll a for the Chesapeake Bay and its tidal tributaries (EPA 903-R-03-002). Washington, D.C.: U.S. EPA. [Google Scholar]
  48. U.S. Geological Survey (USGS). 2020. National Water Information System data available on the World Wide Web (USGS Water Data for the Nation), accessed Aug 12, 2020, at https://waterdata.usgs.gov/nwis/.
  49. Vaquer-Sunyer R, and Duarte C. 2008. Thresholds of hypoxia for marine biodiversity. Proceedings of the National Academy of Sciences of the United States of America. 10.1073/pnas.0803833105. [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Villate F, Iriarte A, Uriarte I, Intxausti L, and de la Sota A. 2013. Dissolved oxygen in the rehabilitation phase of an estuary: Influence of sewage pollution abatement and hydro-climatic factors. Marine Pollution Bulletin. 10.1016/j.marpolbul.2013.03.010. [DOI] [PubMed] [Google Scholar]
  51. Wenner E, Sanger D, Arendt M, Holland AF, and Chen Y. 2004. Variability in dissolved oxygen and other water-quality variables within the National Estuarine Research Reserve System. Journal of Coastal Research. 10.2112/SI45-017.1. [DOI] [Google Scholar]
  52. Williams KJ, Cassidy AA, Verhille CE, Lamarre SG, and MacCormack TJ. 2019. Diel cycling hypoxia enhances hypoxia tolerance in rainbow trout (Oncorhynchus mykiss): Evidence of physiological and metabolic plasticity. Journal of Experimental Biology. 10.1242/jeb.206045. [DOI] [PubMed] [Google Scholar]
  53. Xia M, and Jiang L. 2015. Influence of wind and river discharge on the hypoxia in a shallow bay. Ocean Dynamics. 10.1007/s10236-015-0826-x. [DOI] [Google Scholar]
  54. Yang H, Cao Z, and Fu S. 2013. The effects of diel-cycling hypoxia acclimation on the hypoxia tolerance, swimming capacity and growth performance of southern catfish (Silurus meridionalis). Comparative Biochemistry and Physiology Part A: Molecular & Integrative Physiology. 10.1016/j.cbpa.2013.02.028. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplement1
NIHMS1853165-supplement-Supplement1.pdf (224.7KB, pdf)
Supplement2
NIHMS1853165-supplement-Supplement2.pdf (32.2KB, pdf)

RESOURCES