Metadata

Description
Site description
Our study site (25.780859°, -80.726742) was located in ridge and slough wetlands of the southern Water Conservation Area 3A (WCA-3A) about 1.6 km North of Tamiami Trail with Everglades National Park to the South (FIG. 2). We selected three sampling locations, all within 20 m of each other, two within a single slough and one in the adjacent sawgrass (Cladium jamaicense) ridge (FIG. 2). The slough was ~50 m wide; the ridge ~30 m wide; and the ridge sampling point was located ~3 m into the ridge. The high-SAV slough site contained water lily (Nymphaea odorata), spikerush (Eleocharis cellulosa), and bladderwort (Utricularia foliosa and U. purpurea), and the low-SAV slough site was dominated by few water lily (Nymphaea odorata) (FIG. 2). A benthic layer of unconsolidated floc – a heterogeneous mixture of algal and detrital organic matter – generally is found covering the peat of the sloughs and the ridges to a thickness that varies from 0 to 25 cm thick (McVoy et al. 2011, Pisani et al. 2015). The thickness of the layer may be thinner on the sawgrass ridges than in the sloughs, but on ridges, the thickness can be obscured by visible fragments of dead sawgrass leaves (C. McVoy, personal observation).

Dissolved oxygen and temperature
We deployed dissolved oxygen (DO) loggers (D-Opto SDI-12 Optical DO Sensor, ENVCO Global, New Zealand) by attaching them to vertical tripods adjusted every 60 d to mid-depth of the water column level at the two slough locations (~15 to 45-cm depth) and one ridge location (~5 to 30-cm depth). We positioned DO sensors facing upward and above benthic substrates to ensure that measurements were made in the water. When the DO sensor in the ridge site was above the water during peak dry periods, we were able to detect and exclude these data recorded as supersaturated %DO. The ENVCO Global loggers recorded dissolved oxygen concentrations and temperature at 15 min intervals at each location from 14 February 2014 to 23 March 2016. Sites were visited every 60 d to download data, remove any fouling of the sensors, check calibration, and vertically reposition the DO loggers at mid-depth in the water column, if necessary. Prior to deploying DO loggers, each logger was calibrated in air-saturated water (water that had been aerated with an airstone for 20 min). Upon retrieval from the field, DO loggers were tested for potential drift assessed by recording DO and temperature in air-saturated water and adjusted as necessary.

Meteorological and Stage Measurements
At the research platform, photosynthetically active radiation (PAR) (μmol m2 s-1) was measured continuously at approximately 3 m above the water surface and averaged every 15 min (Apogee Model SQ-100, Logan, Utah, USA). Barometric pressure was recorded every 15 min (WXT-520, Vaisala, Helsinki, Finland). Stage in the slough was recorded at 15 min intervals using a level transducer installed in a PVC tube at a set depth (KPSI Model 500 SDI-12, 0-4 PSI range, 0.05% accuracy, TE Connectivity, Schaffhausen, Switzerland) and converted to water depths (cm) using the peat elevations of the bottom of the slough and surface of the ridge. Due to proximity (less than 20 m) and the absence of any hydrological barriers, it was assumed that stages at both slough locations and at the ridge location were identical. Wind speed (m s-1) at approximately 3 m above water surface was measured using a sonic anemometer (Vaisala WXT-520).

Surface Water Physicochemistry
In October 2014, during the peak of the wet season when water depths were highest, we recorded depth-specific PAR using triplicate measures at 15- to 20-cm depth intervals from the surface of the water to the surface of the soil in each habitat. We generated PAR attenuation curves for each site to quantify the change in light from the surface with increasing water depth. We used these discrete depth-specific PAR attenuation measurements to help interpret modeled estimates of metabolism from continuous water depth and surface PAR measurements.

Diel Dissolved Oxygen and Aquatic Ecosystem Metabolism
Oxygen flux was estimated by fitting the following model to the oxygen data (Van de Bogert et al. 2007, Hall et al. 2015)
O_i= O_(i-t)+ (GPP × 〖PPFD〗_t)/(z × ∑ PPFD)+ (ER × ∆t)/z+K(O_s- O) × ∆t (1)
where O2 at time i is equal to O2 at the previous time (i-1) plus time-step-specific rates of GPP and ER, z is water depth, and K is air-water gas exchange coefficient per unit time (based on the reaeration flux K(Os - O), and the difference between dissolved O2 and O2 at saturation for a given temperature and barometric pressure). Os is saturated oxygen concentration, estimated as a function of barometric pressure and water temperature (Garcia and Gordon 1992), and Os - O is the saturation deficit. PPFDt is photon flux density during the time interval t (µmol photons m-2 s-1). In this model, ER is a negative O2 flux because O2 is being consumed. The time-step () is the measurement interval of logged O2 data for each DO sonde. We assumed that GPP was a linear function of light (Van de Bogert et al. 2007) and that ER was constant throughout the day. During our measurements, we rarely observed DO saturation (see below Results). Net aquatic productivity (NAP) was measured as the change in DO after accounting for diffusive exchange with the atmosphere or reaeration (K) (Hagerthey et al. 2010).
NAPt = GPPt -ERt (2)
Kt =0.5e0.15*wt (3)
In aquatic ecosystems, Kt is generally modeled as a function of wind speed (wt) (Cole and Caraco 1998, Hagerthey et al. 2010, Caffrey 2004, Solomon et al. 2013), although studies have shown that commonly used wind speed/gas exchange relationships may overestimate the gas transfer velocity in wetland ecosystems (Ho et al. 2018).
To estimate GPP and ER, we fit the model of diel whole-stream metabolism to the measured oxygen data by finding estimates of GPP and ER that minimized the negative log-likelihood of the model to the data using function nlm() in R (Hall et al. 2015). We compared modeled estimates and fits based on the minimized log-likelihood estimate (LLE). We used the following exclusion criteria for modeled GPP and ER: we eliminated positive estimates of ER, repeated values for subsequent dates, and used LLE (values > 600) combined with the determination of realistic estimates of GPP and ER based on the distribution of all modeled values. Exclusion of data based on these criteria accounted for 3-20% of daily estimates of GPP and ER (ridge: n = 97 of 525; high-SAV slough: 122 of 597; low-SAV slough: 18 of 611), leaving n = 428 (ridge), n = 475 (high-SAV slough), and n = 593 (low-SAV slough) diel estimates (total N = 1496).