1. Field methods
Five streams and rivers in Connecticut were monitored for dissolved CO2 and O2: Phelps Brook (PHEL), Hubbard River (HUBB), Nepaug River (NEPA), Farmington River (UNIO), and the Connecticut River mainstem (THOM).
Dissolved pCO2 was measured hourly using showerhead equilibrator systems at A PVC chamber was constructed with gastight Swagelok fittings in 2019. This chamber had an elevated spillover outlet so that water would collect in the bottom ~10 cm and a headspace was created in the top ~20 cm. A mini-typhoon water pump (Pro-Active) sprayed stream/river water into the top of the equilibrator and an air pump pulled headspace samples through a non-dispersive infrared CO2 analyzer (Licor 820/840) and back into the chamber, creating a closed system. The gas analyzers were calibrated with three standards (0 ppm, 400 ppm, and 20,000 ppm), have an accuracy of <3%, and a measurement range of 0 – 20,000 ppm. A Campbell CR800/CR1000 recorded readings from the gas analyzer and temperature in the headspace (Campbell T106). One pCO2 reading was taken per hour after five minutes of equilibration.
Additionally, multiparameter sondes (Eureka Manta Water Probes) were installed at each site to measure dissolved O2, pH, and temperature at least hourly from 2015 until 2019. Sensors were cleaned twice monthly and calibrations were checked quarterly. If dissolved O2 deviated the lessor of 5% or 0.3 mg L-1, the unit was recalibrated using air-saturated water. If pH deviated more than 0.1 pH units, the unit was recalibrated using a three-point calibration. Both the sondes and water pumps were installed ~15-20 cm above the streambed to maintain submersion at low flows.
2. Ecosystem metabolism
Existing ecosystem metabolism models were used to estimate GPP and ecosystem respiration (Hosen et al. 2019). The existing ecosystem models were based on 15-min data (Hosen et al. 2020). However, data from 2019 were collected at 60-minute intervals. To verify that this decreased sampling interval did not substantially alter results, metabolism models were computed using 15-minute interval data from 2015-2017 that was rarefied to 60-minute interval data. For all sites, the R2 of correlations between GPP estimates from metabolism models based on 15-minute and 60-minute interval time series data was greater than 0.95.
3. 48-hr intensive sampling
A follow-up 48-hour intensive sampling of Connecticut River site was conducted on September 6-8, 2020. Grab samples for pCO2, DIC, and 13C-DIC were collected hourly during daylight, while sondes measured dissolved O2, pH, and temperature hourly. For pCO2, duplicate samples were collected with headspace equilibration and analyzed on an SRI Model 8610C Gas Chromatograph with FID; this method is detailed in a previous study (Aho and Raymond 2019). For DIC, duplicate water samples were filtered to 0.22 μm, stored in baked (450 ºC for 5 hr) borosilicate glass vials with PolyCone-lined phenolic caps after flushing and subsequently refrigerated until analysis on a Shimadzu TOC-VCSH. For 13C-DIC, 2-ml water samples were collected in 12-mL gas-tight precombusted glass vials, preinjected with 1 mL of phosphoric acid and flushed with hydrogen gas. Samples were run on a Thermo DeltaPlus XP GasBench, with three known reference materials. Data were reported in terms of deviation from the standard Vienna Peedee belemnite (13CVPDB).
4. Diurnal O2 and CO2 modeling
Diurnal O2 and CO2 dynamics were modeled at hourly timesteps for the Connecticut River mainstem. The model accounted for atmospheric gas exchange, metabolic consumption/production, and carbonate buffering. Diurnal O2 and CO2 dynamics were modeled at hourly timesteps for the Connecticut River mainstem, accounting for atmospheric gas exchange, metabolism consumption/production, and carbonate buffering. For O2, the model accounted for hourly changes in (1) flux between the water surface to the atmosphere and (2) metabolic consumption or production, similar to Stets et al., 2017:
[O2]t = [O2]t-1 + atm exchange + GPPO2,t-1 – ERO2,t-1,
where atmospheric exchange is modeled as
kO2 * ([O2]t-1 – [O2]saturation),
and ER was calculated from the significant regression between daily GPP and ER (p<<0.01) for each GPP input.
For CO2, the model tracked hourly changes in the DIC pool:
[DIC]t = [DIC]t-1 + atm exchange of CO2 – GPPDIC,t-1 + ER DIC,t-1,
where atmospheric exchange is modeled as
kCO2 * ([CO2]t-1 – [CO2]saturation).
The model used CO2SYS to determine [CO2]t from [DIC]t and [Alk]t based on carbonate equilibria. Following Stets et al., 2017, the model held alkalinity constant if the change in the DIC pool could be attributed to CO2 evasion or CO2-supported metabolism:
Alkt = Alkt-1,
when kCO2 * ([CO2]t-1 – [CO2]saturation) - GPPCO2,t-1 + ER CO2,t-1 < [CO2]t-1.
However, at moderate to high levels of GPP, [CO2]t was often less than the net change during mid-day hours (i.e., the sum of atm exchange and consumption from GPP). In this case, the model assumed direct uptake of HCO3-, which resulted in a decrease in alkalinity because in natural waters (excluding acidic, organic-rich freshwaters [Abril et al. 2015]), alkalinity is largely HCO3-.
Alkt = Alkt-1 - kCO2 * ([CO2]t-1 – [CO2]saturation) - GPPCO2,t-1 + ER CO2,t-1 + [CO2]t-1,
when kCO2 * ([CO2]t-1 – [CO2]saturation) - GPPCO2,t-1 + ER CO2,t-1 > [CO2]t-1.
Based on a significant regression between daily GPP and ER (p<<0.01), an associated ER was modeled for each of these GPP levels.
Median pH, [O2], and pCO2 in the Connecticut River mainstem at 1:00 for July 20, 2019 to September 24, 2019 were used as initial conditions. Initial [DIC] and alkalinity were modeled from initial pH and pCO2 using CO2SYS and appropriate constants (Millero 1979; Dickson and Millero 1987; and Lee et al., 2010 and the NBS pH scale)(Lewis and Wallace 1998). A constant temperature (25 C) and K600 (5 m d-1) were used. Hourly gas exchange was modeled from gas-specific kO2 and kCO2, which were converted from K600 with the appropriate temperature-dependent Schmidt number and an exponent of ½ (Raymond et al. 2012), and the molar difference between aqueous O2 and CO2 concentrations and atmospheric equilibrium concentrations, assuming atmospheric conditions of 20.95% and 0.425% for O2 and CO2, respectively and Henry’s law solubility constants (Weiss 1974). Daily GPP values were partitioned into hourly rates according to average solar insolation compared to total daily solar insolation for each hour of the day from July 20, 2019 to September 24, 2019. Daily ER (modeled from the correlation between daily GPP and ER, p<<0.01) was partitioned equally for each hour of the day.
References:
Abril, G., S. Bouillon, F. Darchambeau, and others. 2015. Technical Note: Large overestimation of pCO2 calculated from pH and alkalinity in acidic, organic-rich freshwaters. Biogeosciences 12: 67–78. doi:10.5194/bg-12-67-2015
Aho, K. S., and P. A. Raymond. 2019. Differential Response of Greenhouse Gas Evasion to Storms in Forested and Wetland Streams. J. Geophys. Res. Biogeosciences 649–662. doi:10.1029/2018jg004750
Hosen, J. D., K. S. Aho, A. P. Appling, and others. 2019. Enhancement of primary production during drought in a temperate watershed is greater in larger rivers than headwater streams. Limnol. Oceanogr. lno.11127. doi:10.1002/lno.11127
Hosen, J. D., K. S. Aho, J. H. Fair, and others. 2020. Source Switching Maintains Dissolved Organic Matter Chemostasis Across Discharge Levels in a Large Temperate River Network. Ecosystems. doi:10.1007/s10021-020-00514-7
Lewis, E., and D. W. R. Wallace. 1998. Program Developed for CO2 System Calculations. ORNL/CDIAC-105.
Raymond, P. A., C. J. Zappa, D. Butman, and others. 2012. Scaling the gas transfer velocity and hydraulic geometry in streams and small rivers. Limnol. Oceanogr. Fluids Environ. 2: 41–53. doi:10.1215/21573689-1597669
Weiss, R. F. 1974. Carbon dioxide in water and seawater: The solubility of a non-ideal gas. Mar. Chem. 2: 203–215.
