Effect of winter cover crops on soil nitrogen dynamics in no-till corn systems

I performed a meta-analysis, controlled laboratory, and on-farm experiments to address the objectives of this project. Objective 1 of this project used a meta-analytic approach. The details of the methodology and the findings of the meta-analysis were published in a peer-reviewed journal. In brief, a literature search was performed in the ISI Web of Science database to find articles that compared nitrate leaching between CC and no CC treatments. In total, we found 28 relevant articles that were ultimately included in the analysis. Data were extracted and compiled in an Excel sheet. The multi-level mixed-effects meta-analytic model was performed to investigate the effects of CC on nitrate leaching using the nlme package in R. The effect size of CC on nitrate leaching was calculated as the natural logarithm of the response ratios and the results of the analysis were back-transformed for ease of interpretation. We also performed moderator analysis to investigate the effects of covariates (non-legume category, planting dates, shoot biomass, and relative precipitation) on the mean effect size of non-legumes on reducing nitrate leaching.

To address objective 2, we conducted a 150-d controlled laboratory experiment in a customized air-flow system designed within a CONVIRON E7/2 growth chamber to determine the interactive effects of moisture (Ψ : -10.0, -5.0, -1.5, and -0.03 MPa) and temperature (15, 25, and 35 °C) on C and N mineralization from surface-applied CC residues. Surface soil (0-10 cm) samples were collected from a commercially managed corn field at the USDA ARS Beltsville Agricultural Research Center in Beltsville, MD, USA. Field moist soil was sieved through 4-mm sieve, the roots were manually removed, and the soil was homogenized. The soil was then air-dried, ground, and passed through a 2-mm sieve, then stored at room temperature until the start of the experiment. Cover crop residues used in this experiment consisted of two CC species harvested at different growth stages: (a) late-killed crimson clover (flowering stage), (b) late-killed cereal rye (flowering stage), and (c) early-killed cereal rye (tillering stage). Soil and residues with specific water potential (ψ) were prepared. To prepare microcosms, soils with specific ψ (50 g oven-dried equivalent) were packed in 100 mL plastic beakers (5.0 cm i.d.). Cover crop residues with a given ψ were surface-applied at a rate of 3000 kg ha^-1 on soils with the same ψ. Cover crop residues were composed of both leaves and stem mixed in a definite proportion. An un-amended soil as control was kept for each moisture-temperature combination.

Throughout the 150-d incubation period, headspace gas samples were collected and analyzed for CO₂ and N₂O concentrations using a Varian GC450 gas chromatograph (Agilent Technologies, Santa Clara, CA). The microcosms were also destructively sampled at 15, 30, 60, 100, and 150-d after the initiation of the experiment and analyzed for total inorganic N (NH₄⁺ + NO₃^-) contents. The microcosms were extracted with 500 mL of 1 mol L^-1 KCl (at a 1:10 soil:extractant ratio) after overnight shaking in 1-L plastic bottles. The KCl extracts were analyzed colorimetrically for NH₄⁺ and NO₃^- contents using a SEAL AutoAnalyzer 3 High Resolution (SEAL Analytical Inc., Mequon, WI). Net N mineralized was calculated for each microcosm by summing the total inorganic N contents and cumulative N₂O emissions during that period. To determine the temporal dynamics of C and N mineralization from surface-left CC residues, the total CO₂ emitted and net N mineralized from each residue was fitted to first-order rate kinetic model using nlme package in R . The interactive effects of ψ and T on C or N mineralization from surface-left CC residues was modeled by calculating the residue moisture-temperature reduction factor (MTRF). The MTRF at each ψ and T combination was calculated by dividing the total C or N mineralized in 150-d at a given ψ and T conditions by the maximum value (i.e., total C or N mineralized at optimum conditions of -0.03 MPa and 35 °C). Therefore, MTRF represents the normalized or relative amounts of total C or N mineralized from CC residues when compared at optimum conditions. The MTRF ranged between 0 to 1 and allowed comparison across residue types. Response surface equations were developed relating MTRF to residue environmental conditions (ψ and T) for use in computer simulation models.

Objective 3 of this project was addressed by conducting a series of diurnal experiments in fields and a controlled dry-wet experiment in the laboratory. Diurnal variations in red clover and cereal rye decomposition rates were determined by measuring residue CO₂-C flux at approximately 0:00, 6:00, 9:00, 12:00, 15:00, 18:00, and 21:00 hours. The CO₂-C flux from red clover residues was measured at 4 (June 3-5, 2019), 10 (July 18-20, 2019), and 16 (August 29-31, 2019) weeks after termination. Similarly, the CO₂-C flux from cereal rye residues was measured at five (Jun. 14-16, 2019) weeks after termination using a portable EGM-4 infrared gas analyzer system (PP Systems, Amesbury, MA) equipped with a flow-through closed SRC-2 Soil Respiration Chamber (PP Systems, Amesbury, MA). At the time of CO₂-C flux measurements, CC residues from an equivalent area (~78 cm2) were cut and transferred to a beaker. The chamber tops were placed on beakers and sealed perfectly to measure CO₂-C flux from residues within 300 sec. This method was preferred because it allowed us to capture the actual decomposition rates of surface residues in no-till systems. Following CO₂-C flux measurements, CC residues were transferred into plastic zippered bags, sealed, and brought back to the laboratory to measure residue ψ and residue θ_g within 1-2 h. In addition, we collected surface soil samples (0-5 cm depth, 2-cm diameter probe) from the same place where the residue samples were taken for CO₂-C flux measurements. Measurements (residue and soil θ_g, residue ψ, and residue CO₂-C flux) were replicated four times.

During each study period, air temperature and RH were monitored at 75 cm above the soil surface using a T9602-3-D-1 Humidity & Temperature Probe Sensor (Amphenol Advanced Sensors, Mansfield, TX) and recorded at 10 min intervals using a customized Arduino-based datalogger (company, city, state). A 107-Temperature Probe Sensor (Campbell Scientific, Inc., Logan, UT) was inserted inside the CC mulch layer to measure residue temperature. Soil temperature and volumetric moisture content (θ_v) were measured by installing a time-domain reflectometer TDR-310S probe sensor (Acclima, Inc., Meridian, ID) at a 45° angle beneath the residue layer. Surface CC residue temperature, soil temperature, and soil g were recorded at 10 min intervals using a CR1000 datalogger (Campbell Scientific, Inc., Logan, UT).

The effect of dry-wet cycles (i.e., the effect of rainfall pulse) on residue decomposition was investigated in an 80-d incubation experiment performed in a similar manner as described above. Treatments consisted of a factorial combination of two CC residues (crimson clover or cereal rye) and two residue placement strategies (surface-left or incorporated). A soil-alone control treatment was also included. Microcosms were subjected to four dry-wet cycles in a set of CONVIRON E7/2 growth chambers (Controlled Environment Limited, Winnipeg, MB, Canada) at 25 °C. In each dry-wet cycle, microcosms were dried for 16 d in a drying chamber. At the end of each drying period, 12 mL of double deionized water was quickly added using a pipette to raise θ_g from 0.02 to 0.26 g H2O g^-1. Microcosms were then transferred to an identical high-humidity chamber (hereafter, referred to as a wetting chamber) and maintained at field capacity for 4 d.

The C mineralized from residue treatments was determined by measuring CO₂ fluxes throughout the incubation period in a similar manner as described above. Similarly, three microcosms were destructively sampled at the end of each drying and wetting period and analyzed for inorganic N contents. At the end of four dry-wet cycles, microbial DNA was extracted from both soil and residue samples using the Qiagen DNeasy PowerLyzer PowerSoil and PowerPlant kit (Qiagen, Hilden, Germany), per the manufacturer’s recommendations. Prokaryotic 16S rRNA gene and fungal ITS gene regions were targeted using the primers 515F/806R and ITS1f/ITS2r, respectively. Illumina sequencing was performed according to the 16S metagenomic sequencing library preparation manual (Part number 1504423 rev. B, Illumina, Inc.). Samples were pooled; the amplicon library size was verified using a Bioanalyzer 2100 DNA chip (Agilent Technologies). The library was quantified using qPCR and the final library was run on an Illumina MiSeq using a 600-cycle v3 cartridge.

Objective 4 of this project was addressed by conducting multi-state inter-site litter bag decomposition studies in farmers' fields. We established a network of CC adopters across the mid-Atlantic and Southeastern US states. At each farm site, the experiment was conducted in a strip-plot design with CC and no CC as adjacent treatments. There were two subplots within each strip-plot. Farmers were required to leave a strip of bare soil at the time of CC planting or terminate a strip of CC within 10 d after its planting using herbicide. All other CC and cash crop management decisions were dictated by the farmers. They were free to plant any CC species and mixture composition and have flexibility in their preferred CC planting and termination dates/methods. The only requirement to get enrolled in this project was to leave the CC mulch layer on the soil surface and plant subsequent cash crops using no-till methods. In total, CC decomposition studies were performed in 105 no-till corn fields during 2016-2019 across five states (Fig. 1): Georgia (GA), North Carolina (NC), Delaware (DE), Maryland (MD), and Pennsylvania (PA).

Decomposition kinetics of the aboveground CC biomass was determined using nylon mesh litter bags (26 by 60 cm dimensions, 1-mm mesh size). Just prior to termination, CC biomass was collected by clipping two separate 1 m² areas in each of the two subplots. The fresh weight of CC biomass from each subplot was recorded and was subsequently used for filling 12 litter bags. In total, 24 litter bags were prepared at each farm site which provided four sets of six litter bags. Out of six litter bags, one bag was retained on the same day and then dried at 60 °C for at least two days to determine residue θ_g. The remaining five bags were deployed on the soil surface using landscape staples and were collected at approximately 14, 30, 60, 90-120 d after CC termination, and finally at corn harvest at physiological maturity. Retrieved litter bags were oven-dried at 60 °C for at least two days and the weights were recorded after scraping any soil adhered to the outside of the bags. Residues from each litter bag were ground to pass through a 1-mm sieve. Soil contamination was determined by ashing 1-g subsample of ground residue at 550 °C for 4 h. The residue weights were then adjusted to an ash-free basis. We used non-linear mixed-effect models to model CC residue decomposition as a function of the proportion of initial ash-free CC dry weight remaining in litter bags over time. A single-pool first-order exponential decay model was fitted separately for each site-year using the R nlme package to determine residue decay rate constants (i.e., k-values).

A 1-g subsample of ground residue from each bag was analyzed for tissue C and N concentrations using dry combustion via Leco TruMac CN Analyzer (LECO Corporation, St. Joseph, MI). A 5-g subsample from time 0 bag from each subplot was analyzed for residue C chemistry [carbohydrate, holo-cellulose (cellulose plus hemicellulose), and lignin concentrations] via near infra-red reflectance spectroscopy (NIRS) using a scanning monochromator (model 6500; FOSS NIRSystems, Silver Spring, MD). Soil texture was determined using the hydrometer method. Climatic data were imported from the in-house weather API (Application Programming Interface) developed for this on-farm project. Our weather API contains hourly weather data generated by the North American Land Data Assimilation System Phase-2 (NLDAS-2) and hourly rainfall data generated by Multi-Radar Multi-Sensor System (MRMS). The hourly data were summarized to calculate cumulative rainfall, the number of rainy days, mean daily air temperature, and mean daily RH.

Simple linear regressions were fitted between k-values and individual controlling factors to show their independent effect on CC residue decomposition. To account for non-normality and heteroscedasticity of the residuals, the k-values were naturally log-transformed before performing regression analysis. We performed a correlation analysis to assess collinearity among factors controlling k-values. Three stepwise multiple linear regression models were fitted to best predict log-transformed k-value: model I included only intrinsic factors [latitude, soil (sand and clay content), and climate (cumulative rainfall, number of rainy days, mean daily air temperature, and mean daily RH)]; model II included only extrinsic or management factors [CC quantity (biomass) and quality (C:N, residue carbohydrate, holo-cellulose, and lignin concentrations, and lignin:N)]; and model III included both intrinsic and extrinsic factors. The best subset of predictors among intrinsic and extrinsic factors in predicting k-value was identified using the Bayesian Information Criterion (BIC) penalty for each added predictor in the model. Models with the lowest BIC value were selected. Finally, commonality analysis was performed to assess the relative contribution of the factors selected via stepwise multiple linear regression procedure in explaining the model total variance, i.e. R². In commonality analysis, the contribution of each regressor was assessed in terms of unique effects and in the presence of other independent variables, i.e. common effects.