Mitigating Nitrous Oxide Emissions from Corn Cropping Systems in the

Mar 21, 2014 - on side-by-side comparisons of common and alternative management practices, especially those pertaining to N-placement, N- timing, and ...
0 downloads 5 Views 1MB Size
Critical Review pubs.acs.org/est

Mitigating Nitrous Oxide Emissions from Corn Cropping Systems in the Midwestern U.S.: Potential and Data Gaps Charlotte Decock*,† University of California − Davis, One Shields Avenue, Davis, California 95616, United States S Supporting Information *

ABSTRACT: One of the unintended nitrogen (N)-loss pathways from cropland is the emission of nitrous oxide (N2O), a potent greenhouse gas and ozone depleting substance. This study explores the potential of alternative agronomic management practices to mitigate N2O emissions from corn cropping systems in major corn producing regions in the U.S. and Canada, using meta-analysis. The use of the urease inhibitor N-(n-butyl) thiophosphoric triamide (NBPT) in combination with the nitrification inhibitor Dicyandiamide (DCD) was the only management strategy that consistently reduced N2O emissions, but the number of observations underlying this effect was relatively low. Manure application caused higher N2O emissions compared to the use of synthetic fertilizer N. This warrants further investigation in appropriate manure N-management, particularly in the Lake States where manure application is common. The N2O response to increasing N-rate varied by region, indicating the importance of region-specific approaches for quantifying N2O emissions and mitigation potential. In general, more data collection on side-by-side comparisons of common and alternative management practices, especially those pertaining to N-placement, Ntiming, and N-source, in combination with biogeochemical model simulations, will be needed to further develop and improve N2O mitigation strategies for corn cropping systems in the major corn producing regions in the U.S.



INTRODUCTION

N2O emissions typically occur in peaks following disturbances such as tillage, fertilization or rewetting, but the magnitude and duration of such N2O emission peaks is very variable.10,11 In addition, N2O emissions show notoriously large spatial variability.12 These characteristics are caused by the large number of processes underlying N2O emissions and the complex interactions between factors driving and affecting those processes.13 This has made it very challenging to predict the outcome of a certain mitigation strategy on N2O emissions. Nevertheless, given the damaging effects of N2O emissions on human and environmental health, there is interest from governments, GHG offset programs, and other environmental interest groups to incentivize agronomic practices that promote the reduction of N2O emissions from U.S. cropland. Several N2O mitigation practices have been proposed and investigated, including reducing N application rates, optimizing the time, place and source of applied N, reducing tillage intensity, and diversifying crop rotations.14−16 Results from such studies have been summarized in various reports and publications, but these summary analyzes have mostly remained qualitative.7,17−20 The objective of the current study is to expose data availability and quantitatively summarize effects of a suite of agronomic management practices on N2O emissions in corn cropping systems in the Midwestern U.S. and South-East Canada

Nitrous oxide (N2O) emissions account for 7% of total U.S. greenhouse gas (GHG) emissions and have been identified as the dominant driver of stratospheric ozone depletion in the 21st century.1,2 In the U.S., agriculture accounts for 75% of total N2O emissions, of which 92% is attributable to soil management practices such as fertilizer nitrogen (N) and manure application.2,3 N2O emissions from soil are mainly mediated by nitrifying and denitrifying microorganisms and are controlled by carbon (C) and N availability, oxygen availability (often approximated by soil moisture content), and soil pH.4,5 Nitrifiers release N2O as a byproduct during the oxidation of NH4+ to NO3−, and are particularly active under aerobic conditions, when NH4+ is abundant. Nitrifiers are also known to reduce NO2− to N2O under oxygen-limited conditions when NO2− pressure is high.6 Denitrifiers produce N2O during the reduction of NO3− to N2 under anaerobic conditions. The process is favored when NO3− concentrations are high. Because denitrifiers are heterotrophic bacteria, denitrification is also stimulated as C availability increases. In addition to preventing high rates of nitrification and denitrification in the first place, N2O emissions could also be mitigated by enhancing the reduction of N2O to N2.7 Nitrous oxide reduction relative to production increases with increasing pH, increasing oxygen limitation, and when C availability is high relative to NO3− availability. Furthermore, abiotic conversions of soil mineral N to N2O have been documented, and could be of particular importance shortly after fertilization.8,9 © 2014 American Chemical Society

Received: Revised: Accepted: Published: 4247

December March 14, March 21, March 21,

12, 2013 2014 2014 2014

dx.doi.org/10.1021/es4055324 | Environ. Sci. Technol. 2014, 48, 4247−4256

Environmental Science & Technology

Critical Review

Figure 1. Map with locations of study sites included in the database. Open and solid circles represent experiments with and without control (no N input) treatments, respectively. LRR stands for USDA Land Resource Region. Figure was created in QGIS version 2.2.0-Valmiera.

calculate N-surplus as the difference between N applied and aboveground N-uptake,24 but data limitation precluded this approach here. Furthermore, information on the following environmental characteristics was recorded: field site, agroecological region, longitude, latitude, soil order, percent sand, percent clay, percent soil organic carbon (SOC), pH, bulk density, soil texture group, mean annual precipitation, mean annual temperature, aridity index, aridity class, potential evapotranspiration, minimum temperature, maximum temperature, precipitation, irrigation, precipitation plus irrigation, nitrate exposure,25 average water-filled pore space (WFPS), duration of measured N2O flux in days, and measurement period (year or growing season). For the agroecological regions, Land Resource Regions (LRR) as defined by the USDA for observations in the U.S. and Ecozones as defined by the Ecological Stratification Working Group (Canada) for observations in Canada were chosen. Both LRRs and Ecozones confine areas with similar ecological and agricultural characteristics (Figure 1). More details on data sources and categorization of environmental characteristics can be found in the Supporting Information (SI). Data coverage is visualized in Figure 1 and Table 1, and the full database is available in SI.

through meta-analysis. Corn is selected as a focus crop, because it covers the greatest proportion of U.S. cropland and receives more N than any other U.S. crop.21 The geographic area of interest confines a large corn-growing region with relatively comparable climate, environmental conditions, and agronomic management. Implications of findings from this meta-analysis are discussed in the context of current trends in agricultural management practices as reported by the United States Department of Agriculture National Agricultural Statistics Service (USDA NASS).



MATERIALS AND METHODS Relevant studies on N2O emissions in corn cropping systems in the Midwestern U.S. and South-East Canada were identified through a literature search in March 2012 using the Web of Science (keywords ‘N2O’ and ‘corn’ or ‘maize’), and through personal communications with researchers in the field. For inclusion in the database, N2O emissions had to represent at least a full growing season, covering emissions across furrows and seedbeds, and capturing anticipated N2O emissions peaks after tillage, fertilization and wetting.22 Only data from field experiments was included. In addition to growing-season or annual N2O emissions (in kg N2O−N ha−1 yr−1) for each of the field-treatment-year combinations in each study, data for a suite of ancillary variables pertaining to management practices and environmental characteristics were recorded. With respect to management practices, variables included in the database were N-rate (continuous variable, kg N ha−1), N-placement (broadcasted or banded), N-timing (split or single N application; before or after planting), N-source (solid manure, liquid manure, anhydrous ammonia = AA, urea ammonium nitrate = UAN, urea = U, polymer coated urea = PCU, urea ammonium nitrate with nitrification and urease inhibitors, or urea with nitrification and urease inhibitors), nitrification inhibitor (none, DCD = dicyandiamide, ECC = encapsulated calcium carbide, or nitrapyrin = 2-chloro-6-(tricholoromethyl) pyridine), tillage (tilled or no till), rotation (corn soybean = CS, or continuous corn = CC) and irrigation (irrigated or rainfed). When available, yield and grain N-content at harvest were recorded, and used for the calculation of N-surplus (i.e., N applied minus N removed by crop harvest, in kg N ha−1). For data-interpretation, it was assumed that variation in harvest index is small relative to yield variation.23 It is preferable to

Table 1. Summary of the Number of Studies and Observations That Report or Provide Sufficient Information for the Determination of Cumulative N2O Emissions, Fertilizer-Induced Emissions (FIE), Crop Yield, Crop N Export (Harvest Removal), and Year-Round (As Opposed to Growing-Season) Cumulative N2O

number of studies number of observations

cumulative N2O

FIE

corn yield

crop N export

year-round measurements

48

18

33

7

10

548

258

417

88

74

For management practices for which sufficient pairwise comparisons between alternative treatments were available, meta-analyses were performed using the natural logarithm of the response ratio (lnR) as effect sizes: ln R = ln(XA /XB) 4248

dx.doi.org/10.1021/es4055324 | Environ. Sci. Technol. 2014, 48, 4247−4256

Environmental Science & Technology

Critical Review

Where XA and XB are the mean values for cumulative N2O emissions in treatment A and treatment B, respectively. Such analyses were achievable for manure versus synthetic N application, continuous corn versus corn-soybean rotations, no-till versus tilled systems, conventional urea versus polymercoated urea, and no inhibitors versus the urease inhibitor NBPT (N-(n-butyl) thiophosphoric triamide) in combination with the nitrification inhibitor DCD. Heterogeneity and publication bias were evaluated, fixed versus random effects models were tested, effect sizes were weighted by the pooled variance or the number of observations per study, and parametric versus nonparametric confidence intervals were constructed. (More details in SI).26,27 The final results selected for presentation in this paper are conservative and based on a random effects meta-analytic model, using the pooled variance as a weighting function, and with 95% confidence intervals that are generated nonparametrically by bootstrapping (1000 iterations). For studies that included control treatments (i.e., treatments without fertilizer N application), the natural logarithm of the fertilizer-induced emissions (FIE) was used as an effect size in meta-analytic moderator analyses. The natural logarithm of FIE is calculated as follows:

different agronomic practices and/or environmental variables. Furthermore, the data included in meta-analytic moderator analysis with FIE as the dependent variable overlaps with, but is not identical to, data used for side-by-side comparisons. Reoccurring significant effects for side-by-side comparisons and moderator analysis of FIE therefore strengthen confidence in the validity of such results. The relationship between N-rate or N-surplus and N2O emissions was assessed using regression analysis. Linear and exponential models were fit to the data and parametric 95% confidence intervals were constructed. Effects of N-rate and Nsurplus on N2O emissions for the whole data set were evaluated, as well as for each agroecological region separately. The R project tool for statistical computing was used for the regression analyses. Finally, in order to place results from the meta-analysis in a broader context, trends in adoption of various agronomic management practices over time (1996−2010) in the focus area were assessed based on survey data from the USDA’s Annual Agricultural Resource Management Survey (USDA-ARMS).



RESULTS AND DISCUSSION Reducing N-Rate and N-Surplus. Up until now, reducing N rate has been postulated as the most robust and promising strategy to mitigate N2O emissions from arable soil.30 Nitrogenrate has been identified as a strong predictor for N2O emissions,31 and a linear relationship between N2O emissions and N input rate has been adopted by the Intergovernmental Panel on Climate Change (IPCC) to quantify N2O emissions from soil in national greenhouse gas inventories.32 The linear relationship has been challenged in recent years, as an increasing number of studies found an exponential response of N2O emissions to increasing N-inputs, especially when Nrates exceed plant-N demand.14,33,34 Across all data included in this study, the exponential model fit the data slightly better than the linear model, but both model fits were relatively poor (Table 2). The exponential model was almost identical to the linear model at N-rates below 200 kg N ha−1. The slope of the linear model equaled 0.017 ± 0.03, implying that, on average,

⎛ X − Xcontrol ⎞ ln(FIE) = ln⎜ A ⎟ ⎝ NA − Ncontrol ⎠

Where XA and Xcontrol are the cumulative N2O emissions for treatment A and the control treatment, respectively; NA and Ncontrol are the N-rates applied to treatment A and the control treatment, respectively. The variable FIE normalizes N2O emissions from individual observations for differences in Nrates and background emissions between observations and studies,28 and can therefore be used for interstudy comparisons. In this study, effects of agronomic management practices and environmental variables on FIE were assessed using metaanalytic moderator analyses. Random effects meta-analytic models were used for continuous moderator variables and mixed effects meta-analytic models for categorical variables. An evaluation of the normal quantile plot and weighted histogram indicated that ln(FIE) was normally distributed and that publication bias was unlikely (see SI). Meta-analytic moderator analysis is similar to analysis of variance in the sense that average effects of each category of the moderator variable on the dependent variable can be assessed, but meta-analytic moderator analysis takes into account weighting functions to evaluate effects of the moderators. The weights reflected the number of observations per study. In this way, bias from studies with more observations compared to studies with a small number of observations was controlled for. In meta-analysis, variability in the data is referred to as heterogeneity, and the parameter Qm estimates the amount of heterogeneity explained by the moderator. By dividing Qm by the estimate for the total heterogeneity (Qt), the percentage of heterogeneity that is explained by a particular moderator can be determined (further referred to as Qm/Qt).29 All meta-analyses in this study were performed in MetaWin© Version 2.1 (Release 5.10). In addition to the evaluation of the normal quantile plot and weighted histogram, all results were qualitatively evaluated for potential bias. This was necessary, because the representation of management practices and environmental variables by FIE (258 observations, Table 1) was unbalanced. The unbalanced nature of the data set disabled testing for interactions between

Table 2. R-Squared and p-Values Associated with Linear and Exponential Model Fits for the Response of N2O to N-Rate and N-Surplus in Different Agroecological Regionsa response curve

linear model

agroecological region N-Rate full database LRR G LRR L LRR M LRR N Atlantic Maritimes mixed wood plain N-Surplus full database LRR G LRR M mixed wood plain

exponential model

R2

p-value

R2

p-value

# obs.

# sites

0.06 0.3 0.36 0.2 0.10 0.03 0.02