Estuaries and Coasts (2014) 37 (Suppl 1):S91–S110 DOI 10.1007/s12237-013-9765-2
Fortnightly Tidal Modulations Affect Net Community Production in a Mesotidal Estuary Nicholas J. Nidzieko · Joseph A. Needoba · Stephen G. Monismith · Kenneth S. Johnson
Received: 23 October 2012 / Revised: 20 December 2013 / Accepted: 29 December 2013 / Published online: 15 March 2014 © Coastal and Estuarine Research Federation 2014
Abstract Optical in situ chemical sensors enable sampling intervals and durations that rival acoustic techniques used for measuring currents. Coupling these high-frequency biogeochemical and physical measurements in estuaries to address ecosystem-scale questions, however, is still comparatively novel. This study investigated how tides affect ecosystem metabolism in a mesotidal estuary in central California (Elkhorn Slough). Dissolved oxygen measurements were used to estimate the terms in a control volume budget for a tidal creek/marsh complex at tidal timescales over several weeks. Respiration rates were 1.6 to 7.3 g O2 m−2 day−1 ; net community production approached 20 g O2 m−2 day−1 . We found that aquatic NCP integrated throughout the creek complex varied significantly over the spring-neap cycle. The intertidal contribution to aquatic
N. J. Nidzieko () Horn Point Laboratory, University of Maryland Center for Environmental Science, 2020 Horns Point Road, Cambridge, MD 21613, USA e-mail:
[email protected] J. A. Needoba Institute of Environmental Health, Oregon Health & Science University, Mail Code HRC3 - 3181 SW Sam Jackson Park Road, Portland, OR 97239, USA e-mail:
[email protected] S. G. Monismith Department of Civil & Environmental Engineering, Stanford University, 473 Via Ortega, Stanford, CA 94305, USA e-mail:
[email protected] K. S. Johnson Monterey Bay Aquarium Research Institute, 7700 Sandholdt Road, Moss Landing, CA 95039, USA e-mail:
[email protected]
metabolism was net heterotrophic during spring tides and generally in balance during neap tides because spring-tide marsh inundation was limited to nighttime, and therefore the marsh could not contribute any primary production to the water column. At the estuary scale, the fortnightly export of oxygen from the main channel to the intertidal was largely balanced by an advective flux up-estuary. Keywords Biogeochemical cycling · Elkhorn Slough · Tides · Air-sea gas exchange · Primary production · Ecosystem metabolism · Estuarine mixing · Net community production
Introduction The residence time of material entering an estuary is a fundamental component of estuarine ecosystem metabolism (Statham 2012). Estuaries are sites of significant productivity (Cai 2011) and biogeochemical transformations (Smith and Hollibaugh 1993), and therefore the length of time that a parcel of water remains in an estuary affects whether an ecosystem may be a net source or sink of material to the coastal ocean (Nixon and et al. 1996; Dettmann 2001). The rates of these biogeochemical processes, however, can be affected by the same physical processes that control residence time. For example, wind-driven sediment resuspension has been found to increase sediment nutrient fluxes by up to an order of magnitude (Corbett 2010), and tidal sediment resuspension shifts ecosystem processes from the benthos to the water column (Porter et al. 2006, 2010). Determining the interaction between physical and biological processes in estuaries has been hindered by the inherent difficulty of making sustained biogeochemical measurements at the tidal, fortnightly, and episodic timescales over which
S92
both estuarine residence time and geochemical rates vary (Jay et al. 1997). Optical chemical sensors (e.g., Johnson and Coletti 2002; Johnson et al. 2007) have enabled the coupling of biogeochemical and physical measurements at sub-hourly intervals for extended durations, thus permitting observations of the fate of estuarine materials at novel timescales. In this paper, we examine fortnightly changes in estuarine ecosystem metabolism against the paradigm that fortnightly variability is primarily due to tidally driven spring-neap variations in residence time. Physical processes control the residence time of water and materials entering an estuary through a combination of advection, turbulent mixing, and dispersion. Advection refers to transport by mean currents (averaged over either a few minutes or a tidal cycle); these currents may be driven by buoyancy, tides, or wind. Mixing refers to the small-scale turbulent redistribution of material, while dispersion is the large-scale redistribution by time- and space-variable flows, such as transient eddies or chaotic Lagrangian motions (Fischer et al. 1979; Zimmerman 1986; Geyer 1993; Banas and Hickey 2005). In many estuaries, the strength of the estuarine (gravitational) circulation that transports terrestrial material seaward modulates fortnightly with the springneap cycle. Consequently, tides control the residence time of water entering the estuary and produce fortnightly salinity changes (Nunes Vaz et al. 1989; Geyer et al. 2000; MacCready and Geyer 2010). Interpreting changes in geochemical constituents is more difficult because they may be nonconservative, and the physical processes that affect residence time can also affect biogeochemical rates. For example, stronger flows decrease the thickness of the diffusive boundary layer at the sediment-water interface and increase benthic respiration of water column oxygen (Jorgensen and Revsbech 1985; O’Connor and Hondzo 2008). Understanding the nonlinear relationship between physics and biology is essential to advance our understanding on how tidal and weather-band phenomena contribute to changes in estuarine ecosystem metabolism. Ecosystem metabolism is the integrated organismal formation and utilization of organic matter within an ecosystem via the processes of primary production and respiration (cf. Staehr et al. 2012). Gross primary production (GPP) is the autotrophic conversion of inorganic carbon to organic forms; ecosystem respiration (R) is the total oxidation of organic C by both autotrophic and heterotrophic organisms. The balance between gross primary production and total respiration—net community production (NCP = GPP − R)—provides an indication of an ecosystem’s trophic status (Odum 1956; Dodds and Cole 2007; Staehr et al. 2012). The difference between primary production and total respiration is also referred to as net ecosystem metabolism or net ecosystem production. In this paper, we use the phrase ecosystem metabolism qualitatively for narrative
Estuaries and Coasts (2014) 37 (Suppl 1):S91–S110
purposes (Howarth and et al. 2013); because we make no estimates of subareal marsh production when the marsh flood plain is exposed at low tide, we use NCP in reference to the apparent metabolic rate of the aquatic community (which includes the effects of emergent marsh vegetation on the water column). When NCP > 0, the water column is net autotrophic; organic material is available for burial or export, and a deficit of dissolved inorganic carbon and/or carbon dioxide is produced. Conversely, a net heterotrophic ecosystem (NCP < 0) has an excess of DIC and requires stored or imported organic matter to maintain its metabolic state (Kemp et al. 1997; Cai 2011). Tracking ecosystem metabolism over decadal time periods can be an important assessment for the impacts of climate variability and anthropogenic activities (Testa and Kemp 2008). Advancing our understanding on how NCP changes on shorter timescales is essential to develop suitable indicators for the effect of ecosystem stressors (McLaughlin and Sutula 2008; Hughes et al. 2011) and resilience to climate change (Statham 2012). Nearly all methods for quantifying ecosystem metabolism have practical limitations that arise in relation to estuarine physics (Gazeau and et al. 2005). Staehr et al. (2012) review four primary measurement methods: (1) direct measurements made from closed containers (bottle incubations), (2) observation of diel changes in dissolved oxygen (DO) within the water column, (3) examination of oxygen isotope fractions in open water, and (4) ecosystem budgets. Reliable metabolic rate estimates can be obtained through incubation bottles, but exclusion of potentially important physical processes (including turbulence and bioturbation) and the inability to account for spatial and temporal heterogeneity has long been recognized at a shortcoming of this method (Kemp and Boynton 1980; Berg and Huettel 2008). The diel cycling method can be used to track water column DO changes through a 24-h cycle; DO draw-down during the night provides a measurement of total respiration, while the increase of oxygen during the day can provide an estimate of photosynthesis (Swaney et al. 1999). Oxygen isotopes are used at different rates during the oxidation and reduction of carbon such that the differential fractionation of oxygen isotopes can be used to calculate integrated respiration and photosynthesis rates (Tobias et al. 2007). Both the diel tracking and oxygen isotope methods, however, can be confounded by strong tidal signals (Howarth et al. 2002; Wankel et al. 2009), and methods have been devised to eliminate this tidal aliasing based on conservative tracers (Swaney et al. 1999). The ecosystem budget method offers the most integrated measurement of NCP by tracking the imports, exports, and change in storage of carbon (and/or nitrogen or phosphorus) within the system (Kemp et al. 1997) though estimates of the relevant advective and dispersive horizontal fluxes
Estuaries and Coasts (2014) 37 (Suppl 1):S91–S110
are required. Such physical transport terms have been computed from water and salt budgets (e.g., Smith 1997; Gazeau et al. 2005; Testa and Kemp 2008) though few studies have attempted to directly measure these horizontal fluxes owing to the difficulty in sustaining observations at the necessary temporal and spatial scales (Jay et al. 1997). Consequently, the timescales investigated through the ecosystem budget method have typically been seasonal to interannual (e.g., Caffrey 2004) and have revealed little about how tidal variations in physical transport affect ecosystem metabolism. Our study combined high temporal resolution measurements of salinity, dissolved oxygen, and currents to observe how ecosystem metabolism varies over fortnightly timescales in response to strong modulations in tidal forcing. We contrast observations between a shallow fringing tidal creek/marsh area and the deeper main channel of Elkhorn Slough, California, a mesotidal estuary with broad upper intertidal salt marsh. Our estimates of NCP are a blend of the ecosystem budget and open-water diel oxygen methods described above, applied in a control volume framework. This control volume method is described in “Control Volume Analyses.” We found that the tidal marsh was a net consumer of oxygen with fortnightly periodicity, switching from net heterotrophy to nearly balanced conditions during neap tides; in contrast, NCP within the main channel had a different phasing. This work included a manipulation of the measurement site to impound water at lower low tide (“Manipulative Experiment”), such that timedependent oxygen production and respiration rates could be estimated and compared against the physical flux measurements. The observed fortnightly periodicity in apparent biogeochemical rates can be attributed to a feedback between the physics and the biology caused by the timing of inundation and solar irradiance. During our study, the upper intertidal was only inundated at night, with the consequence that the marsh can only respire water column dissolved oxygen and does not contribute any oxygen to the water column via primary production at this time of year (“Discussion”).
Methodology: Field Setting and Equipment Elkhorn Slough is located at the center of the Monterey Bay coast (Fig. 1a). The estuary is short relative to the tidal excursion (the horizontal distance traversed by a tidal current), shallow relative to the tidal range, and wide relative to the main channel width (Table 1). These scales describe an estuary where the water column is closely tied to the benthos, tidal mixing is important, and horizontal exchange between the estuary and ocean is fast. Flushing times in the lower estuary are a few days, while the upper estuary
S93
flushing times approach 2 weeks (Nidzieko and Monismith 2013). We deployed a suite of instrumentation in a small tidal creek, northwest of the entrance to Parson’s Slough from June 4–July 21, 2009 (Fig. 1b). The creek was 8 m wide at the measurement location; the channel meandered roughly 300 m into the marsh interior (Fig. 1c), which consisted of bare mud panne (evident in Fig. 1d) at an elevation of 1.5 m mean lower low water (MLLW) and marsh plants (predominantly pickleweed, Salicornia virginica) with canopies around 1.6 m MLLW (Caffrey et al. 2002a). The measurement site elevation was approximately 0.4 m MLLW, so the creek went dry during spring tide lower low water (LLW); the muddy bottom of the creek remained flooded during neap tide low water. We deployed a small tripod with a Seabird conductivity-temperature-depth (CTD) recorder and an Aanderaa oxygen optode, both at 10 cm above the bed. Nortek acoustic Doppler velocimeters (ADVs) on the tripod leg measured velocities at 10, 25, and 40 cm above the bed. A bottom-mounted 2-MHz Nortek acoustic Doppler profiler (ADP) was located adjacent to the tripod to measure velocity throughout the water column; a 6-cm blanking distance and 10-cm bins were used with ensemble averages of 2 min. This instrument array was used to measure fluxes into and out of a theoretical control volume encompassing the creek/marsh complex for calculating NCP. To estimate NCP in the upper estuary, we used hourly hydrography observations from the L02 mooring (Fig. 1b, 36◦ 50 27.2 N, 121◦ 44 51.7 W) of the Land/Ocean Biogeochemical Observatory (LOBO, Johnson et al. 2007; Jannasch et al. 2008). This site is 6.9 km along channel from the ocean, situated in the main channel at an elevation of −1.9 m MLLW; temperature, salinity, and DO were measured from a surface float at 0.4 m in depth at hourly intervals. Data quality was assessed for each instrument following manufacturer-specified protocols. The oxygen optode was factory calibrated for individual sensing foils; drift was monitored in the laboratory before and after deployments following the recommended procedure for a two point calibration (Aanderraa Data Instruments 2007). Briefly, 100 % saturated DI water was obtained by bubbling water with an aquarium pump until air-water equilibrium was reached; very low DO water was created by adding 5 g of sodium sulfite to 500 ml of DI water in a beaker. Meteorological measurements were acquired from the rooftop station at Moss Landing Marine Laboratories (MLML in Fig. 1b). This station was 5 km from the field site, at an elevation of 12.2 m MSL. A direct comparison between winds measured over water at L02 showed that the MLML observations were representative of wind within Elkhorn Slough (Nidzieko 2009). Given the tidally variable sea surface elevation and uncertainty associated with different topographic influences at the MLML site versus
S94
Estuaries and Coasts (2014) 37 (Suppl 1):S91–S110
a
c
d
b
e
f
Fig. 1 a Location of Elkhorn Slough and b The observational sites. The L02 site (dark circle) is a LOBO mooring in the main channel. The location of the tidal creek is denoted with a box. c Detailed bathymetry of tidal creek/marsh complex used for control volume calculations. The horizontal extent of the control volume is shown with a dark line. The portion of the creek flooded behind the sill, at approximately 0.65 m MLLW, is shown in dark purple; light yellow-filled contours
indicate marsh plain flooded at 1.5 m MLLW. The measurement location is shown with a triangle, adjacent to the sill location. d Aerial image from September 29, 2009. e Wetted surface area (As ), volume (V), and cross-sectional area of the tidal creek (Ac ) as a function of tide height. f Schematic control volume sketches. Am is the theoretical area through which the over-marsh flux occurs for tides above 1.5 m MLLW
our field site, the recorded wind speed at this elevation was assumed to be equivalent to wind speeds at 10 m for the purpose of calculating gas exchange coefficients. In addition to wind speed, we assumed that measurements of solar irradiance from this station were representative of conditions at our field site, though there could have been short periods where the coastal marine layer resulted in greater insolation at our site compared to the measurement location.
crest relative to the surrounding bathymetry. The mean depth of this impoundment region was 0.2 m. Water was impounded behind the sill at LLW for nearly all tides; the exception were neap tides when LLW elevation was higher than the sill crest. The instrumented tripod was located 10 m upstream from the sill, so that measurements could be made when water was impounded. The period of quiescent, impounded water around lower low tide enabled the calculation of R at night and NCP during the day following correction for wind-driven surface gas exchanges (Nixon and Oviatt 1973; Needoba et al. 2012). The details of these calculations are described in “Control Volume Analyses.”
Manipulative Experiment In order to measure biological sources and sinks, a sharpcrested sill was installed across the creek entrance on June 29, 2009 (Fig. 2). This manipulation eliminated horizontal fluxes for a portion of the tidal cycle. The sill was constructed of polymer sheet pile that was driven manually into the soft mud. A tarp fastened over the sill prevented erosion due to spillover at the end of ebb tide. The elevation of the sill crest was roughly 0.65 m MLLW, with a height of 0.4 m above the headwater bottom. The dark purple fill in Fig. 1c indicates the elevation of the sill
Table 1 Characteristic scales of Elkhorn Slough Length Depth Channel width Total width Mean tidal range
10 km 3.5 m 100 m 270 m 1.7 m
Estuaries and Coasts (2014) 37 (Suppl 1):S91–S110
a
high tide
b
low tide
S95
c low tide, sill installed tripod
Fig. 2 Tidal creek study site at a High tide and b Low tide. c Following the installation of sharp crested sill
Observations: Tidal and Diel Variations in Elkhorn Slough The motivation for this study was to improve the understanding on how physical processes contribute to biological variability, such as the chl a time series in Fig. 3d, where semidiurnal, diurnal, and fortnightly tidal signals are superimposed on a phytoplankton bloom within the upper estuary (the chl a maxima at L02 occur with falling water and peak prior to or at low tide). A confounding factor in deciphering such a signal is the small difference in the periodicity of tidal and diel variations that are superimposed on one another. We attempt to illustrate this with observations from the field site. Salinity in the main channel shows how tidal dynamics alone affect scalar concentrations in Elkhorn Slough, with
variations evident at tidal, fortnightly, and monthly periods (Fig. 3a). Tidal advection of the along-channel salinity gradient was evident in the large salinity deviations relative to the tidal mean trace. Elkhorn Slough is an inverse estuary in the dry summer months (Nidzieko and Monismith 2013), and so there was a negative relationship between salinity and tide height (evident in the symbol sizes in Fig. 3c). The variation in the low tide salinity maxima were caused by monthly and semimonthly tidal range modulations that occur along the US west coast (Fig. 3i). Salinity at low tide increased throughout neap tides as salt was concentrated within the estuary via evaporation; greater exchange during spring tides brought in comparatively fresher ocean water (at 33.5) and decreasing salinity. Changes in the tidal excursion also contribute to the observed variations: lower tides bring saltier water from farther up-estuary, while higher tides push ocean water farther in, partially masking changes due to tidal mixing. The salinity observations in the tidal creek closely mirrored those in the main channel (Fig. 3b). The salinity in the creek was lower than in the channel (Fig. 3c) because the tidal creek location was seaward of the L02 site. Oxygen in the main channel reflected some of the same dynamics as salinity, with a few notable exceptions (Fig. 3d). As with salinity, there was a pronounced tidal signature relative to the mean trace. Oxygen minima occurred at low tide (contrast the reversed symbol size distributions between panels c and f of Fig. 3) and were coincident with salinity maxima (Fig. 3g), indicating that dissolved oxygen decreased with distance into the estuary. The time-averaged oxygen trace, however, has a fortnightly signal that is much less pronounced than the fortnightly salinity variation. In general, there is a draw-down of DO throughout neap tides followed by an increase of DO during spring tides; this variation may be attributed to the longer residence time of upper Slough water during neap tides. There is a slight phase shift with regard to where in the fortnightly spring-neap cycle these subtidal changes occur: oxygen changes several days before salinity changes, implying that biological factors contribute to dissolved oxygen variations in upper Elkhorn Slough. Oxygen in the tidal creek (Fig. 3e) had much greater diel swings than in the main channel (Fig. 3f), higher maxima, and lower minima. These can be attributed to the shallower depths in the creek than in the channel, which can more readily buffer the diel cycle. These large diel swings are also evident in the dissolved oxygen-salinity relationships, which show considerable variability around mid-tide in contrast to the more linear oxygen-salinity trend in the main channel (contrast Fig. 3g, h).
S96
Estuaries and Coasts (2014) 37 (Suppl 1):S91–S110
36
35
35
34
34
4
8 0
DO
3 2
0
e creek oxygen
33
34
5
f oxygen
4
4
3
3
2
35 36 channel
37
(µM÷100)
2 1
0
160 167 174 181 188 195 day in 2009
160 167 174 181 188 195 day in 2009
0
1
3 2 channel
4 1500
5
g channel high tide
4
oxygen (µM÷100)
oxygen (µM÷100)
5
160 167 174 181 188 195
1
1
35 34
creek
chl a
5 16
(µg/L)
d channel oxygen & chl a
(µM÷100)
5
c salinity
36
33
160 167 174 181 188 195
37
b creek salinity
creek
36
33
(µM÷100)
37
a channel salinity
low tide
3 2 1 0 33
h creek
i tide height
4 2 (m MLLW)
37
3 2 1
34
35 36 salinity
37
0 33
1000
& irradiance 500 (W/m2) 0
1 0
34
35 salinity
36
37
160 167 174 181 188 195 day in 2009
Fig. 3 Time series and property-property plots for salinity and dissolved oxygen. For the time series, the thin lines show instantaneous values; thick lines show tidal average mean. For property-property plots, the color of the symbol indicates solar irradiance, as indicated in panel i; the size of the symbol indicates the tidal height, with increasing symbol size indicating higher water. a Salinity in main channel (at L02); b Salinity in tidal creek; c creek salinity against channel
salinity, line shows 1:1 dependence; d Dissolved oxygen in the channel, with the inset showing chl a in the channel; e Dissolved oxygen in the creek; f creek oxygen against channel oxygen; g Channel oxygen against channel salinity; h Creek oxygen against creek salinity; i Water level record in channel showing color-coded solar irradiance. The water level record in i is shown below the panels a and b, as well as in some subsequent figures, with an arbitrary vertical axis
These observations show that both tidal and diel processes contribute substantially to dissolved oxygen variations in Elkhorn Slough. For example, hypoxic conditions observed at a fixed location may be produced locally through nighttime respiration or may be advected down the estuary with low tide. A purely tidal
signal would show no variation in oxygen with irradiance; a purely biological signal would show no variation with tidal salinity changes. In the following sections, we employ a control volume approach to better elucidate the interaction between the physical and biological processes.
Estuaries and Coasts (2014) 37 (Suppl 1):S91–S110
S97
Control Volume Analyses A control volume mass balance can be used to interpret the sources of the variability in the hydrography
∂ ∂t
c dV = − cu · n dA V A time rate of change advective flux in and out of mass within control volume of control volume through control surface
c is the mass concentration of the tracer of interest, and the angle brackets denote a time average. The left-hand side of the equation tracks time change of mass integrated throughout the control volume V; these changes are caused by the terms on the right-hand side: the total advective and diffusive fluxes through the control surface with cross-sectional area A at the boundary of the control volume and internal sources and sinks S. The velocity vector u is positive outward (in the n-direction) such that a positive velocity produces a negative change in mass; the down-gradient diffusive flux governed by Dn may represent turbulent and/or molecular diffusivity. We describe in this section the measurements and assumptions required to evaluate (1). The horizontal extent of the assumed control volume is shown by the solid line in Fig. 1c. The movement of material into and out of the tidal creek and surrounding salt marsh and mud panne can be simplified into fluxes through three surfaces (shown schematically in Fig. 1f), the area of which
observations by separating the importance of advective, diffusive, and internal processes. The generic equation for a control volume defined by an arbitrary control surface is as follows:
∂c + Dn dA ± S . ∂n A sources & sinks diffusive flux in and out of control volume
all vary tidally: (1) the creek flux through a cross-sectional area Ac ; (2) the over-marsh flux through a cross-sectional area Am , which delineates the control volume marsh from the main channel and surrounding marshes; and (3) the surface gas exchange flux through the wetted surface area As . The border between the marsh and the upland habitat could define another horizontal flux face through which terrestrial material could be delivered to the marsh, but we selected a tidal creek with no appreciable watershed and conducted the experiment during the dry summer months. Likewise, the benthic flux across the sedimentwater interface, also with surface area As , could be calculated with appropriate measurements; we place this control surface far enough into the mud that the vertical flux of dissolved oxygen is zero. Benthic processes therefore contribute via the source/sink term. These control surfaces can be used to evaluate (1) for dissolved oxygen as follows:
⎫ ⎧ (a) change in (e) marsh (f) creek ⎪ (b) observed (c) derived (d) surface ⎪ ⎪ ⎪ storage ⎪ ⎪ ⎪ NCP NCP ⎪ creek flux over-marsh flux gas exchange ⎪ ⎪ ⎪ N ⎬ ⎨ ⎪ {V c}tN − {V c}t1 + + ksurf As [csat. − c] + kcVmarsh exp(kt) + Vcreek P (I0 ) uAc c vAm cL02 = ⎪ ⎪ T ⎪ ⎪ ⎪ i=1 ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎭ ⎩
(1)
(2)
i
fluxes/sources/sinks summed between successive LLW from time t1 to tN t
The symbols are used as follows : t c cL02 V As
time: T = N t defines a roughly 25-h window consisting of N records at t = 10-min intervals. dissolved oxygen concentration measured in the creek dissolved oxygen measured at L02 flooded volume of creek/marsh complex, including tidal creek (Vcreek ) and marsh habitat (Vmarsh ) wetted surface area of control volume
Ac Am u v ksurf
cross-sectional area of tidal creek entrance (at the location of measurements) cross-sectional area of the plane along the edge of the control volume cross-sectionally averaged velocity flowing through Ac , positive towards main channel average over-marsh velocity flowing through Am surface gas exchange piston velocity, with units of meters per second
S98
csat. k P (I0 )
Estuaries and Coasts (2014) 37 (Suppl 1):S91–S110
saturation dissolved oxygen concentration of water removal rate coefficient, with units of per second net community production per unit volume in the tidal creek, which varies as a function of solar irradiance at the water surface I0 .
Terms on the right-hand side are evaluated from 10-min observations and then summed over the averaging period T (with the exception of (2 (c)), which is derived from a tidal average). For this study, the time averaging is conducted over a ∼25-h window between successive times of LLW. This time period is most appropriately referred to as a lunar day (National Ocean Service 2000), but for brevity, we use tidal to refer to this averaging window. A diel 24-h averaging period aligned with the solar day cannot be used because the progression of the semidiurnal tide (at ∼24.84 h) relative to the sun would introduce a fortnightly periodicity where the averaging window aligns with progressively higher water. As we describe in the “Disscussion ,” the precession of the solar cycle relative to LLW in a mixed, mainly semidiurnal tidal regime results in an annual periodicity; over the duration of the experiment, our averaging window remains closely aligned with the solar cycle. Aligning the average with lower low water is essential so as to track the difference between flood and ebb fluxes; the inequality in the tide prohibits using a shorter tidal period. As and V were determined from 1-m horizontal resolution bathymetry obtained via light detection and ranging (LIDAR) (Fig. 1e, f). The cross-sectional area Ac was validated against the LIDAR dataset by physical measurement of the bathymetry at the onset of the experiment. Am was not needed explicitly, as it was inferred as part of the volume flux correction (c, below). Terms (2 (b)) through (2 (d)) are physical fluxes; the sum of terms (2 (e) and (f)) collectively comprises the total NCP, which, as described above, includes both benthic and pelagic metabolic contributions. The change in storage (2 (a)) for the tidal creek was found to be negligible, and so the physical terms and biological terms must balance one another. These terms were calculated as follows. (a) Change in storage (the total dissolved oxygen present in the water within the control volume) was calculated as the difference of V c observed at subsequent lower low tides (in contrast to the summation used for the other terms); once the sill was installed, low tide was taken to be the time at which the water elevation reached the crest of the sill. When the creek drains completely during spring tides, V → 0. For the impounded water and during neap tides, the storage measurement assumes negligible concentration gradients within the creek at low tide, such that the observed concentration is representative of all of the water within the creek. Estimating dc/dx ≈ 200 μM/6 km =
0.03 μM/m from the main channel gradient, the error in the low tide oxygen concentration is on the order of Ldc /dx ≈ 9 μM, where L is the along-creek distance of 300 m. At normoxic conditions, the storage estimate is in error by less than 10 %. (b) Advective fluxes in the tidal creek cross section were calculated as concentration c times depth-averaged velocity u times the creek cross-sectional area Ac (Fig. 4c); this assumes no lateral or vertical concentration gradients within the cross section of the creek. Lateral uniformity is a good assumption because of the narrow tidal creek width (8 m). Vertical uniformity is a reasonable assumption because the estuary is weakly stratified during the summer months, and the 50 m of shallow tidal creek leading from the main channel landward to the measurement site provides a source of mixing to eliminate any vertical gradients on flood tide. This advective flux only accounts for flow directly within the cross section of the creek. The high tide volume of the control volume is roughly 104 m3 (Fig. 1e); an instantaneous volume flux of order 0.5 m3 s−1 over half a semidiurnal tidal period of 6.21 h (Fig. 4a) yields a tidal prism of 1.1 × 104 m3 , such that the majority of water entering and leaving the control volume is conveyed through the tidal creek cross section. Phase lags in water velocity across the marsh on ebb allow some of the water that enters from the main channel and/or the adjoining marsh to exit via the tidal creek, resulting in a negative tidal average volume flux (Fig. 4d). This residual is an order of magnitude smaller than the instantaneous flux, and so we estimate an error of 10 % for this term. (c) The tidal average advective flux over the marsh (through Am ) was determined via conservation of mass based on the difference between the volumetric change of storage between successive LLW (V /T ) and the measured, tidal volumetric flux uAc through the creek cross section (Fig. 4a, d). The change in volume between tides (not shown) was 2 orders of magnitude smaller than the creek volume flux, and so the inferred over-marsh volume flux vAm must be positive (into the control volume) to balance the negative creek volume flux (Fig. 4g). The base of the vegetated marsh elevation was approximately 1.5 m MLLW, and so the over-marsh flux increased in magnitude proportional to the elevation of higher high water (HHW) above this threshold elevation. This mean over-marsh flow carries salt and oxygen from the main channel and adjacent marsh into the control volume. High-water observations at L02 were assumed to be representative of the salinity and dissolved oxygen content of the water entering over the marsh from the main channel and the neighboring marsh to the north. Multiplying
c
mmol/s
200
0
0
-200
-50
-2
-400
kg/s
2
e
10 mmol/s
d
0
0
g
h
kg/s
2
0 160 167 174 181 188 195 day in 2009
j
kg/s
2
0
mmol/s
0
f
-10
-2
-0.1 0.1
400
50
0
0.1
b
a
mmol/s
2
S99
kg/s
tidal avg. total over-marsh tidal avg. thru-creekinstantaneous thru-creek m3/s m3/s m3/s
Estuaries and Coasts (2014) 37 (Suppl 1):S91–S110
0
i 10 0
k 10 0
160 167 174 181 188 195 day in 2009
160 167 174 181 188 195 day in 2009
Fig. 4 Calculation of horizontal fluxes. a–c Instantaneous volume, salt, and dissolved oxygen fluxes, respectively, measured through the tidal creek. d–f Tidal averaged flux, with negative indicating flux out of the control volume. To ensure volume continuity, there must be an unmeasured volume flux across the marsh into the control volume; the sum of d and g is zero. Multiplying the volume flux in g times HHW
channel salinity and dissolved oxygen produces the inferred overmarsh fluxes in h and i, respectively. j The sum total of the measured and inferred salt fluxes is near zero. k The total horizontal oxygen flux indicates that the creek/marsh system respires oxygen during spring tides, tending to zero or a small production of oxygen during neap tides
the tidal average over-marsh volume flux times L02 high tide salinity provides an over-marsh salt flux that balances the creek tidal average salt flux to better than 10 % (Fig. 4h, j); this computation was not particularly sensitive to the choice of salinity (e.g., using the tidal average mean concentration at L02 instead of the high tide value; Fig. 5a) and indicates that the signal is driven primarily by the volume flux. In contrast, the inferred over-marsh oxygen flux does not balance the creek oxygen flux (Fig. 4i, k), indicating the importance of biogeochemical transformations within the control volume. The over-marsh oxygen flux includes a correction for the respiration of oxygen that occurs prior to oxygen entering the control volume from the adjacent marsh; this correction is based on the rate determined in (e), below, and was applied to the high water L02 oxygen value. The effect of different oxygen concentrations has a large impact on the spring tide flux magnitude, ∼ 30 %, but is of little consequence for neap tide fluxes which are small (Fig. 5b).
(d) Surface gas exchange was calculated as a piston velocity that varied both as a function of wind speed and tidal current speed ksurf = kwind + kflow (Needoba
a
b 0.2
RMS using: tidal avg S = 0.60 high tide S = 0.51
15
RMS using: tidal avg O = 4.5 high tide O = 6.3 high, adj. = 5.6
10
0.1
5 0 0 −0.1
160
188 174 day in 2009
−5 160
188 174 day in 2009
Fig. 5 Sensitivity of horizontal fluxes to choice of scalar concentration. a High tide salinity (closed squares) has a slightly smaller RMS total than using the tidal average salinity (open circles). b Total oxygen flux computed using high tide concentration (open squares), the high tide concentration corrected for respiration (closed squares), and tidal average concentration (open circles). RMS values are indicated in the inset in each panel
M÷100 m2 x 10 4
mm/s
S100
0.1
Estuaries and Coasts (2014) 37 (Suppl 1):S91–S110
a piston velocity
ksurf wind
0 2 0 5
b wetted surface area c dissolved oxygen concentration saturation
measured
0
mmol/s
100 50 0
−50 mmol/s
d gas-exchange
e
10
0 −10
160
167
174 181 day in 2009
188
195
Fig. 6 Components of the gas exchange surface flux. a Piston velocity, including contribution from wind speed (light red fill) and water flow speed (dark red fill). b Wetted surface area, derived from hypsographic curve in Fig. 1e. c Dissolved oxygen concentration, measured (light blue) and at saturation (dark blue). d Instantaneous surface flux. e Tidal averaged surface flux
et al. 2012). Briefly, kwind is a Schmidt number that is proportional to wind speed (Ro and Hunt 2006); √ kflow = uD/H , where D is an empirically derived diffusion coefficient (Wilke and Chang 1955) and H is the water column depth. kflow was up to 5.1×10−5 m/s, and kwind was up to 9.2×10−5 m/s; typical values were 10–20 % of these maxima (Fig. 6a). The piston velocity was multiplied by the time-variable wetted surface area As (Fig. 6b) and the difference between the measured and saturation oxygen concentrations (Fig. 6c). The low-frequency shift from supersaturated to undersaturated dissolved oxygen controlled the sign of the tidal average surface flux. The large tidal average flux on day 171 and instantaneous fluxes from days 170 to 171 were due to strong winds generated by a lowpressure system that were roughly twice the magnitude of the typical diurnal sea breeze. (e) Marsh NCP (comprising the habitats above 1.5 m MLLW) was estimated by representing upper intertidal inundation as a plug flow reactor in which we estimated the concentration change between flood and ebb (Fig. 7a). In a plug flow reactor, the loss or gain of a nonconservative tracer with initial concentration c0 can be modeled by an exponential function c(t) = c0 exp(kt) (this is the solution to dc/dt = kc with initial condition c(t = 0) = c0 , where
the first-order rate process kc takes the place of the source/sink term in Eq. (1)). For every incoming fluid parcel (defined by the 10-min sample interval), the residence time τ spent in the control volume was found by estimating an excursion position based on a quasi-Lagrangian predictor-corrector method (Table 2, Fig. 7b). cflood = c0 was the dissolved oxygen concentration entering on flood tide; cebb = c(τ ) was interpolated based on the exit time τ of that fluid parcel, determined when the excursion distance crossed zero. Concentrations were corrected for atmospheric exchange. As should be expected in a short tidal embayment, parcels ebbed at roughly the same stage height at which they entered on flood (Fig. 7c). k was found via regression of log(cebb /cflood ) = kτ from nighttime observations where both the water elevation was above 1.35 m and the nighttime flood/ebb duration exceeded 3.33 h (Fig. 7d, g). Because only nighttime tides above the marsh elevation were used, we assume this removal rate to be indicative of marsh respiration, calculated for each i-th (10-min) observation as kci Vmarsh,i exp(kt). The exponential model is justified because the change in dissolved oxygen was found to vary with cflood (Fig. 7e). Physically, this method requires the assumption that the water column is well-mixed vertically but that the fluid parcel does not mix appreciably with other neighboring fluid parcels; a similar regression with salinity shows that the rate of conservative tracer mixing (calculated as a removal rate) is more than 2 orders of magnitude smaller than the biological processes (Fig. 7f). The DO removal rate coefficient derived from the marsh residence times for 23 tidal cycles totaling 356 observations was k = −2.1 × 10−5 s−1 (R 2 = 0.70, p < 0.001, Fig. 7g). For a mean water column depth over the marsh of 0.3 m, and an average oxygen concentration of 185 mmol m−3 , this yields a respiration rate of 5 g O2 m−2 day−1 . This is somewhat lower than typical subareal respiration rates reported for S. virginica-dominated marshes. Using plants from Suisun Marsh (San Francisco Bay, CA), Pearcy and Ustin (1984) report respiration rates on the order of 30 g O2 m−2 day−1 ; similarly, based on a CO2 uptake rate of 0.95 mg CO2 g dry wt−1 h−1 (plants from Sapelo Island, GA, in Antlfinger and Dunn 1979) and above-ground biomass densities of 2 kg dry wt m−2 (from two southern California estuaries, in (Zedler et al. 1980), subareal marsh respiration is 35 g O2 m−2 day−1 . Standing biomass of salt marsh plants do not peak until August (Zedler et al. 1980), which may partially explain our observed lower rates in June and early July. Additionally, much of the marsh interior consists of bare mud panne, which would have
tide (m MLLW)
(µM) dis. oxygen
(m - mllw)
0
2
0 0
C
200
ebb
Sebb
ΔC
2
τ
Cebb
1 0
-10
-5 0 5 time (hours)
2
300µM 100
0 -50
C
e
100
0.5
f
34.5
S 0
33.5
-0.5
1.6 1.4
Cebb
-0.4
1.1
C -0.2 0 0.2 -creek velocity (m/s)
d oxygen & tide height
100
300µM
ksalt = -2.0 x 10 -7 s-1
1 koxygen = -2.1 x 10 -5 s-1
0.9 0.8 0.7
1
0.6 0 160
167
174 181 day in 2009
188
Fig. 7 Estimate of upper intertidal respiration rate. a Example salinity (dark blue line), dissolved oxygen (light blue line), and tide height (filled region) traces for the tide starred in panel d. The thicker portion of the salinity and oxygen traces show period where criteria for nighttime high water are met (see text). The time is relative to the entry of a parcel of water into the tidal creek. The exit concentration of that parcel is indicated with the gray arrow, as derived from panel b. b Estimated excursion distance for parcels of water entering on flood tide. The starred tide in d is shown as an example. c Stage-velocity diagram of entering and exiting fluid parcels. The dark gray line corresponds to the same fluid parcel as in a and b. The velocity is shown with
lower respiration rates. Caffrey (2004) reports on respiration rates in the range of 5 to 10 g O2 m−2 day−1 for intertidal mudflat regions in Elkhorn Slough, which is more consistent with our rate. An additional source of uncertainty is how subareal respiration rates are manifest when the salt marsh plants are submerged; the subareal respiration rates likely represent an upper bound, as diffusive transfers into the water would be slower than in air. (f) Tidal creek NCP (the low to middle intertidal below 1.5 m MLLW) was estimated as a function of irradiance by impounding water behind a sill at LLW (Fig. 2c). By impounding water, the generic control volume Eq. (1) is reduced to the following:
300µM
-100
2 4 time (hours)
c
1.8
Δ oxygen
34
400
b
200
Δ salinity
S
400
Cebb/C
salinity
35
tide height (m MLLW)
a
S101
excursion (m)
Estuaries and Coasts (2014) 37 (Suppl 1):S91–S110
∂ ∂t
C dV = V
A
∂c dA ± S, ∂n
(3)
where the diffusive fluxes are only in the vertical direction. The air-sea surface diffusive flux can be readily removed (d, above); the observed change in oxygen
0.5
g 0
2
4 6 τ (hours)
8
flood positive for consistency with the flux terms; in Eq. (2), negative velocities are into the control volume. d Tidal trace showing nighttime oxygen concentrations for tide heights above 1.35 m MLLW. Black portion of tidal trace and black banding along time axes show periods with zero solar radiation. e A logarithmic decay model was selected because observed rates of dissolved oxygen decrease were dependent on the initial concentration. The initial concentration cflood is colorcoded, and the difference is defined as oxygen = cebb − cflood (f); salinity did not show this same dependence on initial concentration. f Decay rate derived from change in concentration and residence time. The thick gray lines are linear regressions of all 24 tides
Table 2 Quasi-Lagrangian predictor-corrector method to estimate marsh excursion distance in Fig. 7 (1) (2) (2a)
(2b) (2c) (2d)
Dn
195
(2e)
Record initial position: xp (t = 0) = x0 = 0 Iterate WHILE xp >= 0 Estimate velocity at current position, using velocity at creek entrance u0 = u − u xL0 Estimate new position at t + t x ∗ = x0 + u0 t Estimate velocity at estimated new position u1 = u − Lu x ∗ Calculate new position x1 = x + 12 (u0 + u1 )t Record new position and reiterate x0 = x1 , t = t + t xp (t) = x0
Pseudo-code tracks position x of water parcel p entering through tidal creek at time t = 0. Cross-sectionally averaged velocity measured in the creek is u; the distance into the marsh L = 400 m
S102
Estuaries and Coasts (2014) 37 (Suppl 1):S91–S110
can be assumed to be driven by benthic and water column biological processes to provide an estimate of NCP for the tidal creek. If the rate of mass transfer throughout the water column depth H is fast relative to the rate of molecular diffusion across the diffusive boundary layer δ, then the benthic fluxes can be represented as a combined source or sink term with the water column processes, similar to the calculations for (e). More formally, we require TH
