Greenhouse gas and soil sampling
The net terrestrial biosphere-atmospheric exchange of CO2, CH4, and N2O was measured using closed, static chambers (Figure A1 in Supplemental Materials). Gas samples were collected on approximately a monthly basis from June to September, 2019. At each site, six chambers were placed along the study reach, on level ground, about 1 to 2 m from the stream bank-full margin, in order to capture an area within the zone of influence of the stream (Gregory and others 1991). Chambers (diameter of 30.5 cm, height of 23 cm) made of grey PVC pipe were permanently inserted about 10 cm below the soil surface at least 10 days (on average 23 days) prior to the first gas sampling to reduce the effects of soil and root disturbances.
At each site, the chambers were stratified according to local groundwater discharge conditions, in groundwater discharge (DIS) and non-groundwater discharge (ND) microsites, in order to account for some of the high spatial variability associated with GHG fluxes from soils (Vidon and others 2015). These DIS areas, or discrete riparian inflow points, occur when upland-originating groundwater converges and discharges in a depression in the topography of the riparian zone (Kuglerová and others 2014). The DIS areas were identified in Arcmap 10.6.1 using a 1 m digital elevation model (DEM) and flow accumulation modelling using a channelization threshold of 1 ha. This modelling process assumes that topography and gravity control water movement, and that the groundwater flow path follows the ground surface (Kuglerová and others 2014). The DIS areas were then confirmed with field observations of topography, wetness, and hydrophilic vegetation.
During gas sampling, a PVC lid was placed on top of the chamber and headspace air samples were taken at 0, 15, 30, and 45 minutes after closure. Headspace air samples of 20 mL were taken from a rubber septa sampling port in the middle of the lid using a 23 gauge needle and a 50 mL syringe after pumping 20 mL of the headspace gas twice to facilitate mixing. The gas sample was then injected into a pre-evacuated 12 mL exetainer (LabCo Ltd., Lampeter, Wales) until over-pressurized. After gas sampling, air temperature of the headspace was recorded and two ambient air samples were taken for reference.
Sampling was performed between 9:15 and 16:30 h to capture peak fluxes and reduce the effects of diurnal variation (Parkin and Venterea 2010). Given the difficulty of sampling 54 chambers in one day, three sites were sampled per day over three days, in randomized order within treatment, each month. Gas samples were analysed on a 7890A gas chromatograph (Agilent Technologies Inc., CA, USA) equipped with a flame ionization detector and an electron capture detector (Agilent Technologies Inc., CA, USA).
On each gas sampling date, volumetric soil moisture (± 3%) was recorded at each chamber using a GS3 ProCheck portable probe (Decagon Devices, Inc., Washington, USA) by using the mean of three readings, each no more than 0.5 m from each chamber, at a depth of 5.5 cm from the LFH layer. The depth to the water table was also measured on each sampling date by blowing into a thin tube attached to a meter stick lowered into a well until the bubbling noise of the groundwater was heard, and corresponding depth on the meter stick was noted. The well was made of perforated PVC pipe wrapped in landscape fabric installed at least 40 cm deep into the soil. Based on the groundwater table level data from the wells installed at each chamber, eight DIS areas were re-classified after the fact in cases where the well was dry for at least 80% of the sampling occasions. Over the entire sampling period, continuous soil temperature readings (± 0.5°C) were taken at 1 hr intervals using iButton® dataloggers (DS1992L- Thermochron and DS1923- Hygrochron, Maxim Integrated Products, USA) buried about 10 cm below the soil surface. At each site, one ibutton was buried adjacent to a chamber in a DIS and ND area, respectively, and an additional ibutton was buried at the most upstream chamber, 0.5 m and 1.5 from stream bank-full width, respectively. Air temperature (± 0.04°C) was measured at each site using two HOBO U23 Pro v2 data loggers (Onset Computer Corporation, MA, USA) located 0.5 m and 1.5 m from the stream bank-full width at each study reach.
Statistical analysis
Gas flux rates were calculated by linear regression of gas concentrations over time. Each time series was evaluated for goodness of fit by visual inspection (Collier and others 2014). Additional quality control measures included visual inspection for abnormally high and low values outside the range of reported riparian emissions, as well as Cook’s Distance statistical test (Zuur and others 2007). In sum, these quality control measures resulted in the removal of 11% of flux rate data points for CO2, 26% for CH4, and 34% for N2O. Using the ideal gas law, the flux rate was converted to µmol, and then the molecular mass was used to translate this value into µg or mg. These equations are described by Collier and others (2014). For statistical analysis, we used the software R3.6.1 (R Core Team 2020). We used linear mixed effects (LME) models to evaluate the forest management effects on GHG fluxes using the “glmmTMB” package (Brooks and others 2017). The models included the AR(1) autoregressive covariance structure to account for temporal autocorrelation and repeated measures (Kravchenko and Robertson 2015). The model residuals were tested for normality and homogeneity of variance. When the model residuals violated the assumption of normality, they were transformed and tested again. In the case of models with N2O flux rate as the response variable, the flux data were log-transformed. We used a post hoc Tukey’s HSD test to compare pairwise differences in treatment levels. For all statistical analyses, significance was accepted at p < 0.05.
Results
Environmental variables
Mean soil temperature (June to September) was, on average, the lowest at the reference sites (13.6 ± 1.5 °C; mean ± SD ), intermediate at the buffer sites (14.4 ± 1.3 °C), and highest at the no buffer sites (14.9 ± 1.3 °C) (Figure 2). Maximum and mean daily air temperature followed the same trend, with a mean temperature of 14.5 ± 2.0 °C, 14.8 ±2.2 °C, and 15.3 ± 2.6 °C, at the reference, buffer, and no buffer sites, respectively. Mean soil temperature was similar at the groundwater discharge (DIS) areas (14.1 ± 1.6 °C) compared to the ND areas (14.5 ± 1.5 °C).
Mean soil moisture (measured as volumetric soil water content) was highest at no buffer sites (Figure 3A). The mean soil moisture from June to September was 43.3 ± 14.4%, 45.5 ±16.7%, 50.4 ± 14.3% at the reference, buffer, and no buffer sites, respectively. In terms of local groundwater conditions, mean soil moisture was significantly higher in the DIS areas at 55.4 ± 11.4% compared to 41.6 ± 15.1% in the ND areas, across all treatments.
Mean depth to the groundwater table was the lowest at the no buffer sites, and was significantly lower than the buffer and reference sites (Figure 3B). The mean depth to the groundwater table (June to September) was 28.7 ± 9.6 cm, 26.2 ± 12.7 cm, and 15.9 ± 8.9 cm at the reference, buffer, and no buffer sites, respectively. In terms of local groundwater conditions, mean depth to the groundwater table was significantly lower at the DIS areas (20.0 ± 12.6 cm) compared to the ND areas (27.4 ± 10.1 cm), across all treatments. There was no statistically significant difference in daily average soil temperature and soil moisture between treatments.