«A Thesis Presented to the Faculty of the Graduate School of Cornell University In Partial Fulfillment of the Requirements for the Degree of Master of ...»
We observed static chamber N2O fluxes from converted soil and adjacent fallow grassland daily from March 27th to April 7th 2013. Hot-moment analysis as well as nonparametric statistics and regression modeling were used to identify topologic and management factors that affect N2O emissions.
Site description and experimental design:
The site for this study was a seasonally wet, 16-acre field near Ithaca, New York USA (42N 28.20', 76W 25.94'). The field is generally flat with slightly undulating topography broken by the remains of an old surface drainage network comprised of shallow swales; soil drainage class ranges from moderately well drained to poorly drained. Aside from occasional mowing or hay harvesting, during the past 50 years the field has laid fallow due to recurring and seasonal wetness that renders the soil unfit for reliable equipment access or row crop production, and it is in this context that the field is classified as marginal. In July 2011 portions of the field were mowed, plowed, disked and seeded with reed canarygrass (Phalaris arundinaceae L., v.
Bellevue) (RCG) for bioenergy feedstock, while other areas were left undisturbed as fallow
fertilizer. In October 2012 the RCG was mowed at a height of approximately 20 cm but was not raked or baled due to low yields resulting from near-drought summer conditions. Because the RCG was left in this manner, the distribution of crop residue was quite variable within those plots; residue was left as it lay after mowing. While the FGL subplots were blanketed with a consistent and substantial layer of partially lodged senescent grass and vegetation, the RCG subplots tended to vary widely in the degree of detritus and crop density, the Low/RCG plots exhibiting patchy, intermittently matted crop residue, crowned vegetation and exposed muddy soil (fig. 1).
Within the field, three plots; A, B and C with dimensions 8.1 x 7.2, 7.9 x 3.4, and 7.9 x
3.8 m respectively, were chosen that allowed side-by-side comparison of adjacent FGL and RCG, and spanned a distinct topological gradient extending from a saturated area. Each saturated area consisted of a shallow depression presented by one of the existing drainage swales, and held several inches of standing water that was sometimes frozen at the surface during the period of data collection. All three plots were mapped as the same soil type (Dalton channery silt loam, thin mantle, 0-3% slopes) (27) and identical sampling designs were applied at each of the three plots. The sampling design was a replicated 2x2 factorial design with topography and crop type as grouping factors (fig. 2). The levels for the crop type class were FGL and RCG, and the levels for the topographic class were High and Low. Each plot contained 4 subplots situated to represent the 4 possible topographic class and crop type combinations.
Data collection and instrumentation:
N2O fluxes were observed by the static chamber method outlined by Parkin and Venterea (28). Chambers were constructed similarly to those reported by Molodovskaya, Oberg (29) using
equipped with a butyl rubber septum for sample withdrawal and a vent tube (length: 203mm, width: 6.4mm) for pressure equilibration, as removable chamber covers. Each cover fit snuggly over the rim of the base with a heavy rubber band (dimensions: 305mm x 25mm x 1.6mm, Dykema Bands, McKees Rocks PA) around the base to seal the surfaces. Overall chamber dimensions were 45 cm high with a circular footprint 36 cm in diameter. Pairs of chamber bases were installed less than 1.33 m apart at each subplot on March 25th 2013 and left in place for the duration of the study. The chamber bases in the low subplots were centered approximately 60 cm back from the standing water/ice while in the high subplots they were placed over what appeared to be the driest local areas. In the RCG subplots where crop residue was quite variable, chambers were placed to capture a representative range and degree of residue and vegetation in that subplot. Bases were installed to a depth of 3 to 5 cm with as little disturbance as possible to the subplot by using a custom circular hole saw to cut a circular notch in the soil. Care was taken to preserve the original quantity, quality and arrangement of vegetation and crop residue within and around the chamber base, and to ensure an adequate seal between the base and the soil.
Data were collected daily from March 27th 2013 to April 7th 2013. Beginning at 13:11 (+/- 7.5 minutes) to approximately correspond with peak diurnal soil temperature, we conducted simultaneous measurements of gas fluxes at all chambers in all three plots, with one sampler running a rotating schedule around the site. Four gas samples were withdrawn from each chamber at 0:05, 20:00, 40:00 and 60:00 after applying the chamber cover, the entire course lasting approximately 80 minutes for all 24 chambers. During sampling, 20 mL gas samples were withdrawn from the chambers with a syringe, 15 mL of which were injected immediately into 10 mL glass vials that had been sealed with a butyl rubber stopper and pre-evacuated to -90 kPa.
standard gas. Calibration mixes were prepared manually from known gas standards and included 20% oxygen to ensure uniformity in detector response across field samples and calibrations. Gas samples and prepared calibrations were stored together in plastic bags in our lab at room temperature and nitrous oxide concentration determined within 60 days by gas chromatography.
Our gas chromatograph (Agilent Technologies 6890N) is operated with splitless injection and is equipped with a µECD detector at 250°C and a supel-Q plot 30 m capillary column with ultrapure He carrier gas at 2.6 mL min-1. An oven temperature of -22°C was maintained using liquid nitrogen cryocooling to separate the N2O peak. Automatic peak integration was performed with Chemstation™ software. Daily chamber fluxes were calculated using the method outlined by Rochette and Bertrand (30) but without correcting for effects of air humidity (31)(equation 1).
!" Here, is the rate of change in chamber concentration of N2O at t = 0, V is the chamber volume !" (0.02 m3), A is the chamber footprint area (0.07 m2), MN2O is the molecular mass of N2O, and Vm is the molar volume at chamber pressure and temperature.
We used the slope from linear least squares regression of sample concentration vs. time which tends to underestimate flux values compared to a higher order regression, but with less sensitivity to measurement noise.
Soil temperature and moisture at each subplot were observed daily immediately following the conclusion of gas flux observation, typically beginning at 14:45. Soil temperature over the top 2.5 cm was averaged from 3 readings of thermocouple thermometers placed
by time domain reflectometry (TDR) following a similar probe placement pattern as soil temperature. Continuous automatic measurements of precipitation and air temperature were recorded by a tipping-bucket rain gauge (model 3665R, Spectrum Technologies) and a HMP45A/D temperature probe (Vaisala Group) linked to a datalogger at the site. Above-ground biomass (vegetation and crop residue) was collected from within each chamber on April 13th
2013. The total wet biomass from each chamber was weighed and a subsample was weighed and dried at 105° C for 16 hours as recommended by Undersander (32), then reweighed to calculate total dry biomass for each chamber. Sample calibration, flux calculation and data handling were performed using “R” software (version 2.14.1), as were statistical analysis and modeling.
Method detection limit estimation:
To estimate the detection limit of the N2O flux observations as determined by the static chamber method followed by gas chromatography, we used the Monte Carlo technique. We first selected the initial samples (time = 0.083 min) from the data set. These initial samples are theoretically very close to the ambient concentration of atmospheric N2O with little spatial variation, the average concentration of the initial samples was 0.303 ppm. The standard deviation of concentrations of the initial samples was calculated for each gas chromatograph sequence and the average standard deviation of all the GC runs was found to be 0.011 ppm.
We then generated 1000 values from the standard normal distribution in matlab™ and multiplied each random value by the standard deviation (0.011 ppm) from the previous step, then added the mean concentration value (0.303 ppm). These simulated concentrations were split into groups of four and assigned times based on the sampling interval for the chamber method used in this study. For each group of four simulated concentrations, the slope of the linear regression was
was computed. Finally, the standard deviation of the simulated fluxes was found to be 1.4 * 10-7 g N2O m-2 min-1. Multiplying by a factor of 1.96 gives the 95% confidence interval and this was calculated as +/- 2.7 * 10-7 g N2O m-2 min-1, or +/- 10.2 µg N2O-N m-2 hr-1.
Data handling and statistics:
As has been observed in other studies (12, 16), our fluxes were non-normally distributed.
We used a hot-moment approach similar to that of Molodovskaya, Singurindy (11) and Corre, vanKessel (12) in recognition of the extreme temporal and spatial variability that is typical of N2O emission patterns. This approach identifies outliers and qualifies them as hot-moments;
instances of sudden biogeochemical response to multiple driving factors. This approach can elucidate the link between field scale factors and sites of intense microbial action in the soil that produce N2O. We followed the procedure used by Corre, vanKessel (12) to qualify hot-moments as individual flux observations that exceed a threshold set as the median plus 3 times the interquartile range for the dataset.
In addition, we used the Kruskal-Wallace rank sum test and Wilcoxon’s rank-sum test (both non-parametric tests that can be used with skewed data) to identify significant differences in fluxes among the four topographic position – crop type combinations. Finally, we removed outliers from the dataset and constructed a least-squares linear regression model from the remaining data. No log-transformation of the data was performed.
Fluxes (fig. 3A): Chamber N2O fluxes for the study period ranged from -9.8 to 21.3 µg N2O-N m-2 hr-1 except for one observation of 77.6 µg N2O-N m-2 hr-1 at chamber 2 at the low topographic position with RCG of plot B (BLT-2) on April 4th. Exploratory data analysis showed
Figure 3: Daily fluxes for each treatment with the hot-moment indicated by an arrow (A), air temperature and soil temperature for each treatment (B), precipitation (mm/hr) and soil water content (% VWC) for each treatment (C) over the full course of the study. The hotspot is indicated by open circles with a cross. High indicates the high topographic position, Low indicates the low topographic position. FGL indicates fallow grassland, RCG indicates reed canarygrass.
Temperature (fig. 3B): The record of hourly mean air temperature at the site shows a diurnal cycle with significant day-to-day variation. A multi-day cold spell beginning on April 1st and lasting till April 4th drove soil temperatures downward. This trend was followed by an abrupt change to warmer weather, and the soil temperatures at all subplots rose. Daily maximum soil temperatures across subplots ranged from -0.5 to 9.6°C with clear differences in trends between subplots, most notably between the RCG and FGL subplots, the RCG generally warming to a greater degree than the FGL and more variable from day-to-day (fig. 4, 5).
Precipitation and soil moisture (fig 3C): All snowcover had melted by the beginning of the study period, but some soil at the site remained frozen during the study; by the final observations on April 7th almost all subplots had completely thawed. Light to moderate precipitation occurred during the first half of the observation period. The precipitation affected soil moisture immediately, with some decrease evident during the drier second half of the observation period. There were clear differences in soil moisture between subplots which ranged from 27.4 to 74.0% VWC, the low subplots clearly wetter than the high.
Above-ground biomass (fig. 4, 5): The quantity of above-ground biomass within each chamber varied from 9.0 to 62.9 g dry biomass with differences evident between subplots in the RCG compared to those in the FGL, the RCG clearly exhibiting reduced biomass. An inverse relationship was observed between above-ground biomass and mean daily maximum soil temperature, and between above-ground biomass and the range of daily maximum soil temperature, illustrating that soil temperature is strongly influenced by the insulating nature of the biomass.
Total N2O emitted from all chambers during observation was 43.8 µg N2O-N.
Approximately 25% of this total, 10.7 µg N2O-N, was from the single most active chamber, BLT-2, hereby called the hotspot. The emissions from chamber BLT-2 on April 4th, hereby called the hot-moment, were 5.1 µg N2O-N, approximately 12% of the total from all 288 individual observations. A similar pulse of N2O flux did not occur in the nearest chamber, BLTwhich was located at the same subplot, 1.0 m away from BLT-2 (fig. 3A). This pattern of isolated but intense emissions is similar to the observations of Molodovskaya, Singurindy (11), Wagner-Riddle, Furon (16), Corre, vanKessel (12) and others who have reported high temporal and spatial variability of N2O emissions from agricultural fields, and warrants special attention given to both this chambers location and micro-environment, and to the time of maximum flux.