Site description

We collected soil using a shovel from the Ap horizon (0−20 cm) of an arable grass field at the W.K. Kellogg Biological Station (KBS) in Hickory Corners, MI (42°41’02” N, −85°37’34” W). KBS soils are mixed, mesic Typic Hapludalfs of co-mingled Kalamazoo and Oshtemo series (Crum and Collins, 1995) developed on glacial outwash with intermixed loess (Luehmann et al., 2016). Soil collected in September 2019 for this experiment was from the Kalamazoo series, which are well-drained fine-loams (43% sand, 38% silt, 19% clay) with ~2% total C (Grandy and Robertson, 2006) and a pH of 7.2 (Robertson et al., 1993). Average annual precipitation at KBS is 1005 mm, with an average of 1.3 m of snowfall per year, and mean annual temperature is 10.1°C (Robertson and Hamilton, 2015). The site was in various corn-soybean-wheat rotations for the prior 40 years and before that, corn-soybean-small grain rotations for at least 60 years.

Experimental design

To examine the influence of initial soil moisture on the pre-assay loss of CO₂ during dry-down, we pre-wet recently collected soil to three different initial water-filled pore space (WFPS) levels: 30, 50, and 70%. Then we measured gravimetric soil moisture (GSM) and CO₂ loss while soil was air-drying, after which we re-wet them and measured the 24-h CO₂ pulse by standard methods (Robertson et al., 1999; Franzluebbers et al., 2000).

Laboratory analyses

After collection, soil was sieved through a 4-mm mesh and mixed. We measured GSM and calculated the target volumetric water content (VWC, g H₂O cm^-3 soil) for each treatment following Eq. 1 (Elliott et al., 1999):

VWC = WFPS/100 * (1− SBD/2.65)                                                                                                                                     (1)

where soil bulk density (SBD) is 1.5 g soil cm^-3, previously assessed(Robertson, 2016). Then we divided VWC by SBD to obtain a target GSM and thereby determined the amount of water to add to the field-moist soil (11% WFPS; GSM = 0.032 g H₂O g^-1 dry soil). We then weighed 40 g of soil into each of 75 polyethylene cups. Each cup was randomly assigned to an initial WFPS treatment (30, 50, or 70%), for a total of 25 replicates per treatment. We added sufficient deionized water to each cup to achieve the target initial WFPS and stirred to evenly distribute water. After soil was wet and stirred in the cups, the contents of each cup were transferred to a labeled paper bag. The soil was spread evenly across the bottom of the bag, and the top portion of the bag was removed to increase air flow. Afterwards, the soil was immediately weighed and set on a laboratory bench to air-dry.

Immediately after wetting, as well as 1, 3, and 8 days later, we assessed GSM and CO₂ loss rates for five replicates per initial WFPS treatment. GSM, which was determined after drying the soil at 105°C for 24 h, stabilized at 1.5% in the air-dried soil (Fig. 1a), but did not reach zero even when soil was completely air-dry. Because soil in all initial WFPS treatments were air-dry by day 3, with CO₂ loss rates close to zero, we terminated GSM and CO₂ measurements after day 8.

CO₂ loss rates at each sampling interval were measured by placing 10 g of soil into a 235 mL mason jar fitted with a gas-sampling septum. Then we sampled 5 mL of headspace from each jar at 4 intervals (0, 0.5, 1, and 2 h), injected it into an evacuated 3 mL exetainer (Labco Limited, Lampeter, Wales, United Kingdom), and replaced the jar headspace with laboratory air. CO₂ samples were analyzed within 24 h using a LI-820 CO₂ Gas Analyzer (LI-COR Biosciences, Lincoln, NE, USA).

On day 15 we re-wet the remaining five replicates of air-dried soil from each initial WFPS treatment to 50% WFPS (Franzluebbers et al., 2000). We then assessed subsequent 24-h CO₂ pulses by sampling headspaces at 0, 2, 4, 8, and 24 h. 

2.4 Statistical analyses and correction factor

CO₂ pulses were calculated as the positive slope of the linear regression of CO₂ concentrations through time after accounting for headspace dilution, and then converted to a standardized rate using the ideal gas law. In 17 of 75 cases, we omitted one of the four data points within a jar, which were clear visual outliers. In two cases, we rejected jars with leaks. CO₂ loss rates during the dry-down period were analyzed with a two-way analysis of covariance (ANCOVA), where initial WFPS treatment and days elapsed since wetting (Day) were factors and GSM at the time of sampling was a covariate. Additionally, a one-way analysis of variance (ANOVA) was used to determine whether initial WFPS treatment significantly affected the 24-h CO₂ pulses upon re-wetting the air-dried soil.

We also calculated a correction factor to account for pre-assay CO₂ loss prior to the 24-h CO₂ pulse assay. To calculate the total amount of CO₂ loss during dry-down for each initial WFPS treatment, we calculated a best-fit exponential decay curve:

Y = α^βX + θ                                                                                                                                                                                (2)

where Y = daily CO₂-C loss and X = length of dry-down period, until soil was air-dry (i.e., immediately after wetting through day 3). Total C loss was equivalent to calculating the area under the curve, using bootstrapping to allow for error estimates.

Because we used sacrificial sampling, we could not calculate standard deviation or standard error in the usual way. Instead, we used a bootstrapping approach in which we computed predicted values for CO₂ losses (Ŷ_i) and residuals (e_i = Y_i – Ŷ_i). All zeroes for CO₂ losses were set to 1 for fitting the regression because an exponential decay curve can approach but never attain 0 and because 1 was lower than any value we observed. Then we created a bootstrap sampling of residuals specific to each dry-down interval (0, 1, or 3 days), sampled randomly from each interval with replacement, and added randomly sampled residuals to predicted values (Y_i^* = Ŷ_i + e_i^*) for each dry-down interval (after Hesterberg, 2015). Residuals were bootstrapped 10,000 times to derive multiple estimates of coefficients for the exponential decay curve (α, β, and θ). We also integrated under the curve 10,000 times to get an error estimate (i.e., coefficient of variation) associated with the total amount of pre-assay CO₂ loss during dry-down.

Then we divided the total CO₂ loss by three days to obtain the daily rate used to calculate a correction factor following Eq. 3:

CF = (daily CO₂ loss during dry-down / 24-h CO₂ pulse after re-wetting) + 1                                                                   (3)

The correction factor for each treatment was then multiplied by each replicate’s 24-h CO₂ pulse following re-wetting. Finally, we verified that the correction factors worked by conducting a one-way ANOVA to determine whether initial WFPS treatment still had an effect on the corrected pulses. For all analyses, we confirmed that assumptions of normality and homogeneity of variance were not violated.