Study Sites and Experimental Design
The experimental sites were located at the Laval University Experimental Farm near Saint-Augustin-de-Desmaures (46°44′ N, 71°31′ W; altitude, 110 m), with mean annual air temperature of 4.4°C and mean annual precipitation of 1231 mm. A poorly drained silty clay (432 g clay kg−1 , 163 g sand kg−1 , 35 g C kg−1 ), mixed frigid dystric eutrudepts (Canadian classification: gleyed melanic brunisol) and a well drained sandy loam (170 g clay kg−1 , 680 g sand kg−1 , 19 g C kg−1 ), mixed frigid typic dystrudept (Canadian classification: orthic dystric brunisol) (Lafl amme and Raymond, 1973) were selected to establish the plots. Before the experiment (2008), both sites were cropped to silage corn (Zea mays L.). All crop residues were removed at harvest, and the sites were moldboard plowed (0.2–0.25 m depth) in the fall. The experimental plots were established in May 2009 as a randomized complete block design and were 5 by 7 m in size; experimental treatments (control [CTL], mineral fertilizer [calcium ammonium nitrate], liquid swine [LSM], liquid dairy cattle [LCM], and solid poultry manure [PM]) were replicated three times on each soil type. The diff erent treatments were surface broadcast in the spring, and all plots were harrowed the same day to a depth of 0.1 m to minimize ammonia volatilization. Manure application and harrowing were performed on 12 and 13 May 2009 on the sandy loam and silty clay soil, respectively. In 2010, operations were performed on 5 May for the sandy loam and 13 May on the silty clay soil. After harrowing, all plots were seeded to spring wheat (Triticum aestivum L., cv. AC Brio) using a nine-row, no-till drill seeder (Great Plains Mfg., Inc., Salina, KS) at 450 seeds m−2 on the sandy loam and at 475 seeds m−2 on the silty clay. Row spacing was 0.18 m. Excluding the CTL, all plots received 90 kg ha−1 available N based on local recommendations (CRAAQ, 2010). For the organic fertilizers (LSM, LCM, and PM), 70 (LCM and PM) to 90% (LSM) of applied N is considered available to cereals when applied in early spring and rapidly incorporated to the soils to minimize N volatilization (CRAAQ, 2010). This coefficient takes into consideration manure N losses due to volatilization and the relative availability of the organic N forms. To simplify field work, an average coefficient of 0.8 was used to calculate field application rates for the three manure types. The mineral N fertilizer was applied as calcium-ammonium-nitrate (CAN), P was applied as triple superphosphate (20 kg ha−1 P2 O5 for the sandy loam; 30 kg P2 O5 ha−1 for the silty clay), and K was applied as potassium chloride (20 kg K2 O ha−1 for both soil types). Application rates for P and K were also based on provincial soil recommendations (CRAAQ, 2010). Harvesting during 2009 occurred on 25 and 26 August for the sandy loam and silty clay, respectively; harvesting in 2010 occurred on 12 and 17 August for the sandy loam and silty clay, respectively.
Manure Collection and Analyses
Raw LCM was obtained from the Deschambault Research Center (Research and Development Institute for the AgriEnvironment) during the spring of 2009 and 2010 and was collected after about 6 mo of storage in the main tank. The LSM came from a commercial farrow-to-finish operation and was collected after about 6 mo of storage by composite subsampling. Fresh PM was obtained from a commercial broiler operation and included bedding material composed of wood shavings. The PM was covered and stored for approximately 1 mo before use. On each application date, a composite sample of each manure type was collected and used for analysis. The LCM and LSM were homogenized using a Polytron (Model PT 3100; Kinematica AG, Littau-Lucerne, Switzerland). Distilled water was added to the PM (10:1, water/PM ratio) to make a slurry solution and homogenized as described above. The diff erent manures were then measured for pH by direct reading with a glass electrode. Dry matter was determined as the weight of materials remaining after drying 100 mL of LCM and LSM and 50 g of fresh PM for 96 h at 55°C. Total C was measured by injecting 50 μL of homogenized LSM and LCM, or the filtered (934-AH; Whatman, Piscataway, NJ) slurry in the case of PM, into an automated combustion C analyzer (Model Formacs; Skalar Analytical, De Breda, The Netherlands). Total N and P concentrations were determined by measuring NH4 + and PO4 3− concentrations in Kjeldahl acid digests (Chantigny et al., 2007) using an automated continuous-flow injection colorimeter (QuickChem 8000 FIA+; Lachat Instruments, Loveland, CO). The mineral N content of the liquid manures or PM slurry was determined by shaking 10 mL of the sample with 50 mL of 1 mol L−1 KCl for 60 min. The extract was filtered through pre washed (1 mol L−1 KCl) Whatman #42 filter paper. The NH4 and NO3 + NO2 concentrations in the extracts were measured using the colorimeter described above.
Soil and Grain Sampling and Analyses
Soil sampling commenced the day after seeding and was performed twice in the first week, weekly for the next 3 wk, and every second week until early October. The top 0.15 m of soil was collected with a stainless steel step-probe soil core sampler (0.025 m diam.). Three subsamples were taken from each plot and combined into one composite sample. The sample was rapidly brought to the laboratory, and large soil aggregates were manually broken and mixed to homogenize the sample. Soil mineral N concentration was determined by extracting 25 g of field-moist soil with 125 mL of 1 mol L−1 KCl the day of collection. Soil slurries were shaken for 1 h on a reciprocal shaker (120 strokes per min), centrifuged (3000 × g for 10 min), and passed through filter paper (Whatman #42) prewashed with 1 mol L−1 KCl. The extracts were frozen (−20°C) until analyzed for NO3 + NO2 (hereafter called NO3 ) using the colorimeter described for manure analyses. Volumetric soil water content was measured using threebar 0.15-m time domain reflectometry probes at 0.075 m depth. Soil bulk density was measured in the 0- to 0.15-m depths using the cylinder method (Culley, 1993). Soil water content and bulk density were used to calculate waterfi lled pore space (WFPS) assuming a soil particle density of 2.65 g cm−3 . Soil temperature was measured using copper constantan thermocouples installed at 0.075-m depths and a handheld thermocouple thermometer (model 600–1040; Barnant, Hayward, CA). Grain yield was determined in each plot by harvesting a 1.62 by 5 m area using a small plot combine (Wintersteiger, Austria). To determine total N content, a 25-g subsample was dried by ventilation (3 d at 55°C) and milled with a Cyclone Lab Sample Mill (UDY Corp., Fort Collins, CO) to pass a 1-mm screen. Total N content of the subsample was measured using a dry combustion analyzer (LECO CNS-1000; Leco Corp., St. Joseph, MI). Grain N yield was calculated by multiplying the N concentration in the grains by grain yield. Grain and grain N yields are not presented in this paper and were used only to calculate yield-based N2O emissions as described below.
Gaseous Emissions
Soil N2 O and CO2 fluxes were measured the same day as soil sampling and repeatedly until mid-October using non flow through, non–steady state chambers (Rochette and Bertrand, 2008). Two rectangular acrylic frames (0.15 m by 1.00 m by 0.14 m height) were inserted between crop rows to a depth of 0.1 m in every experimental plot and left undisturbed for the duration of the experiment. On every sampling date, the frames were tightly fitted to a vented and insulated acrylic chamber (0.2 m in height) covering the same area as the frame. Air samples (20 mL) were collected from the chamber headspace at 0, 8, 16, and 24 min after chamber deployment using a syringe through a rubber septum and immediately transferred to 12-mL pre-evacuated glass vials (Exetainers; Labco, High Wycombe, UK). The gas samples and quality control samples were analyzed within 10 d using a gas chromatograph (Model 3800; Varian Inc., Walnut Creek, CA) equipped with an electron capture detector (N2 O) and a flame ionization detector (CO2 ). The soil-surface N2 O–N and CO2 –C fluxes for the sampling dates were calculated using equations proposed by Rochette and Bertrand (2008), and the flux for each plot was estimated using the mean flux of the two chambers. Although fluxes were calculated as N2 O–N and CO2 –C, for simplicity flux is herein referred to as N2O and CO2.
Data and Statistical Analyses
Daily N2O and CO2 flux rates for dates between samplings were calculated using linear interpolation, and cumulative N2 O and CO2 emissions were calculated as the sum of all daily flux rates for the entire growing season (May through October) as well as for each of three seasons (spring [seeding to 15 June], summer [15 June to 2 wk after harvest], and autumn [2 wk after harvest to 31 October]). One N2O flux measurement was identified as an outlier using Grubb’s test (Grubbs, 1969), and, although it is included in the figures, it was excluded from the statistical analysis. This outlier is discussed in more detail below. The treatment effects on cumulative N2O and CO2 emissions and seasonal mean N2O fluxes were tested separately for each soil type using the MIXED procedure of SAS (Version 8; SAS Institute, Cary, NC) with years as repeated measures and block as a random factor. Contrasts were used to test for differences between the control and treated plots, between the CAN and manure treatments, between the solid (PM) and liquid manures, and between liquid manures. When appropriate, data were log transformed. Stepwise linear regression (PROC STEPWISE) was calculated for each soil type separately using seasonal N2O fluxes as the dependent variable and seasonal NO3 exposure, CO2 flux, soil WFPS, summer (set up as an indicator variable), and the two-way interaction terms between these variables. The indicator variable “summer” was used to differentiate between the summer season and the other two seasons because we believe that the size and activity of the crop during this period would provide sufficient competition for resources that would alter the response of N2 O to the environmental variables used in the regression model. The CO2 flux was used as a proxy for C availability (Petersen et al., 2008). Differences were considered statistically significant at P < 0.05, although variables with P < 0.1 were also examined.
