Introduction

Global warming resulting from GHG emissions of agricultural origin has been considered as one of the important current environmental impact issues (Robertson et al. 2000; IPCC 2007). Agricultural practices along with domestic fossil fuel combustion have a major impact on the global C and nitrogen (N) cycles, possibly leading to the global temperature increase of up to 0.17 °C per decade since 1950 (Lal 2004; IPCC 2007). The environmental impacts of agricultural activities are diverse, including the contribution of excess nutrients, sediments and pesticides to surface and ground waters, air pollution, and the production of greenhouse gases (GHGs), mainly CO2, N2O and CH4 (Hass et al. 2000). Crop producers are urged to adopt effective management practices to mitigate GHG emissions so that lower the C footprints of agricultural products on the farm (Gan et al. 2011a). Consumers and the public want to know the C cost and are desire to cut C footprint in their choice of food products. There are also tradeoffs between economic yield optimization at current farmgate prices and emissions. As a result, GHG emissions and their effect on global warming have inspired the quantification of the C footprint of many human activities (Hillier et al. 2009). The contribution of specific farming activities during crop production to the overall C footprint has only recently been calculated (Clair et al. 2008; Hillier et al. 2009; Mahmuti et al. 2009; Gan et al. 2011a; Linguist et al. 2011).

It has been demonstrated that a significant part of the GHG emissions related to crop production can be reduced through improved agronomic practices (Mahmuti et al. 2009) that increase yield and generate higher inputs of residue C, which can lead to increased soil C storage (Follett 2001; Lal 2004; Gan et al. 2011a). Examples of such practices include using high yielding crop varieties, timely control of crop diseases, rotation with crops which allocate more C below ground, and judicious nutrient management (Smith et al. 2008; Mahmuti et al. 2009; Gan et al. 2011a). Rotation of annual and perennial crops can reduce GHG fluxes of the cropping system through an efficient use of resources by individual crops (Robertson et al. 2000). Adviento-Borbe et al. (2007) demonstrated that a maize-soybean rotation had a 47 % reduction in CO2 eq m−2 year−1 after switching from continuous maize culture. Van Groenigen et al. (2010) summarised results of 19 studies and state that optimizing fertilizer-N use efficiency under median rates of N input can reduce yield-scaled N2O emissions. Emissions can also be reduced by adopting less intensive cropping systems, which reduce reliance on pesticides and other inputs (Paustian et al. 2004; Gan et al. 2011a). Steen Jensen et al. (2011) reported that there is little evidence to support a direct association between nitrogen fixation and N2O emissions, although the decomposition of nodules and biomass from legume can be a source of N2O emissions (West and Post 2002; Rochette and Janzen 2005; Gan et al. 2011a).

Nitrogen applied in the form of inorganic fertilizers and manure is not always used efficiently by crops (Cassman et al. 2003). Improving crop N uptake efficiency can reduce leaching loss of available N, and hence reduce direct and indirect emissions of N2O generated by soil microbes largely from surplus N (Smith et al. 2008; Ma et al. 2010). Practices that improve N use efficiency include adjusting application rates based on precise estimation of crop needs (i.e., precision farming), using slow release forms of fertilizer, avoiding time delays between N application and N uptake, placing the N more precisely into the soil to make it more accessible to crop roots, and avoiding excess N application (Cole et al. 1997; Ma et al. 2005; Monteny et al. 2006).

Sustainability of crop production partly depends on its C footprint and crop yield (Dubey and Lal 2009). It has been shown that more than 75 % of total C footprint from crop production resulted from N inputs (Hillier et al. 2009). On the other hand, N inputs are essential to sustain high crop yields that are needed to satisfy the increasing global need for food, feed and bio-energy. This indicates that agricultural production should be optimised for high crop productivity and low GHG emissions. It has been proposed to express GHG emissions from agriculture production using a specific emission term (Ma et al. 2010) or on yield-scaled emissions at moderate fertilizer rates (Van Groenigen et al. 2010).

While there have been numerous studies to assess GHG emissions from different agricultural practices and their impacts on global climate change, little is known about the interaction of these practices with crop productivity and C footprints. From an ecological service point of view, the C footprint is defined as a unit of CO2 equivalent gas emitted per unit of grain produced, i.e. kg CO2 eq kg−1 grain. Practices believed to be efficient in reducing the C footprint of crop production include using crop species with low C footprints, diversification of cropping systems, using crops capable of biological N fixation, improving nutrient use efficiency and crop residue management (Gan et al. 2011a), and the timely control of plant diseases and other biotic and abiotic stresses (Mahmuti et al. 2009). There have been very limited studies exploring the C footprint of maize production under variable agronomic practices (Adviento-Borbe et al. 2007). Currently, there is no approved methodology to estimate GHG emissions from the previous crop residue decomposition, either in a continuous monoculture or rotational cropping system. In this study, we assumed that emissions of nitrous oxide from crop residue N occurred in the same year of production. We hypothesize that a continuous monoculture maize cropping system would have a higher C footprint than maize-soybean (Glycine max L. Merr) or maize alfalfa/red clover (Medicago sativa L./Trifolium pratense L.) crop rotations. We also hypothesize that N fertilizer application can be optimized in the cool-season region of Canada to reduce the C footprint while improving crop yield.

The objective of this study was using a 19-year field experiment data to determine the effects of N fertilizer rates and crop rotation on (1) maize grain yield, (2) total GHG emissions on a per ha basis (kg CO2 eq ha−1), and (3) the C footprint on a per kg grain produced basis (kg CO2  eq kg−1 grain).

Materials and methods

Experimental design

In this study, we used data collected from a 19-year field experiment to estimate total GHG emissions and C footprint of different cropping systems. The long-term experiment was carried out at the Central Experimental Farm of Agriculture and Agri-Food Canada at Ottawa, Ontario, Canada (45°22′N, 75°43′W; 79 m above sea level). Crop rotations tested were maize in annual rotation with soybean (SM), maize in annual rotation with forage (FM, alfalfa 1992–2008 or red clover 2009–2010) and continuous monoculture maize (MM). Current high-yielding maize hybrids were used. Perennial alfalfa, planted and used as an annual, was cut and removed from the plot in July and August from 1992 to 1998, but the cut was left on the surface from 1999 to 2008 and then incorporated in the fall each year. Since 2009, alfalfa was substituted for red clover; red clover was planted, cut at the flowering, and the cut was also left on the surface and incorporated in the fall. From 1992 to 2001 (phase I), the plots were fertilized in the spring of the maize year of the rotation with 0 or 100 kg N ha−1 as ammonium nitrate. An additional well-fertilized treatment (200 kg N ha−1) was included in the continuous maize system. In the second phase (2002–2010) of the experiment, each plot was split into two equal halves to implement more N levels. Both phases of the rotation (maize-soybean and soybean-maize, and maize-alfalfa/red clover and alfalfa/red clover-maize) appeared each year to account for annual environmental impact on rotational effect (Ma et al. 2003), and thus total emissions and C footprint of rotational maize could be estimated annually. There was no N application to soybean or forage legumes in the year without maize. During phase I of the study, the experiment consisted of 11 rotation-by-N treatment combinations, and was arranged in a randomized complete block design with three replications. Each plot was 16 m long and consisted of 12 rows of maize with row spacing of 0.76-m, which was planted from the 2nd to the 4th week of May each year. From 2002 to 2010 (phase II), there were four levels of N treatments (0, 50, 100, and 150 kg N ha−1) in the rotational maize and six levels (0, 50, 100, 150, 200, and 250 kg N ha−1) in the continuous maize system. Since 2002, ammonium nitrate fertilizer was changed to urea. The experiment then consisted of 22 rotation-by-N combinations and each plot was 7-m long and 12 rows wide. The experiment followed recommended management practices for weed control (OMAFRA 2002) and a variety of herbicides were used. All plots, including alfalfa/red clover and soybean, were fertilized with adequate P and K fertilizers each year, according to soil test recommendations and with the above different rates of N fertilizer prior to maize planting.

Agronomic data

The central two rows of each plot were harvested for grain yield (reported on a 155 g kg−1 water basis), with a combine fitted with weighing and moisture meter system. For harvest index (HI), 5 plants from each plot were cut at the ground level, and dried at 80 °C in a forced-draft oven to constant weight. After weighing, grain was removed from the ears with a corn sheller (Agriculex Inc., Guelph, ON, Canada), and HI was calculated as the ratio of grain weight divided by total aboveground plant biomass weight. Stover dry matter was then calculated using the grain yield data after moisture correction, taken from a larger plot area, and the corresponding HI from the 5 plant sample. Roots from three plants selected at physiological maturity, were excavated (0–20 cm depth) in each plot, washed and oven-dried in 1993, 1994 and 1995. Stover and root samples were ground and passed through a 1-mm screen. Total nitrogen concentrations of each sample were determined by the Kjeldahl digestion method (AACC 1983) and analyzed on a flow injection autoanalyzer (QuikChem® 8000 Flow Injection Analyzer, Zellweger Analytics, Inc., Lachat Instruments, Milwaukee, WI, USA). Stover total N (STVTN) was the product of stover N concentration (STVNc) and stover dry matter on a per hectare basis. Because root samples were not taken every year, an average root to total aboveground biomass ratio of 7.2 % and a measured mean root N concentration of 4.92 g kg−1 from 3 years were used to estimate root biomass and root TN from each plot in each year.

Determination of GHG emissions and C footprint

Greenhouse gases emissions for maize production were estimated using a similar approach to Gan et al. (2011a, b). This included estimating CO2 equivalent emissions from (1) the application of synthetic N fertilizers, (2) crop residue (stover and roots) decomposition, (3) production, storage, transportation and application of synthetic N, P and K fertilizers, (4) production of herbicides and pesticides, and (5) various field operations such as tillage, herbicide spraying, seeding and harvesting.

Synthetic N fertilizer applied to a soil can be nitrified or denitrified by micro-organisms, contributing directly and indirectly to N2O emissions (Forster et al. 2007). The amount of direct and indirect emissions is related to the quantity of N applied and the environmental conditions. The magnitudes of these emissions have been extensively studied in Canada (Gregorich et al. 2005; Rochette et al. 2008). Using a large number of observations on measured N2O fluxes from Canadian farmlands, Rochette et al. (2008) developed a simple method for estimating N2O emission factors based on a growing season moisture deficit; a linear function of the ratio of growing season precipitation to potential evapotranspiration as follows:

$$ {\text{EF}} = 0.022{\text{ P}}/{\text{PE}} - 0.0048 $$
(1)

where EF is the emission factor with units of kg N2O-N kg−1 N and P/PE is the ratio of precipitation to potential evapotranspiration during the growing season (1 May to 31 October) based on actual weather data. The upper limit of EF is set at 0.0172 kg N2O-N kg−1 N when the growing season rainfall is equal to or greater than the growing season evapotranspiration rate (Rochette et al. 2008). Mineral N in a form of nitrate in the rooting zone is also subject to leaching. Rochette et al. (2008) developed a method to estimate the fraction of leachable N (FRACLEACH) based on P/PE as follows:

$$ {\text{FRAC}}_{\text{LEACH}} = 0.3247{\text{ P}}/{\text{PE}} - 0.0247 $$
(2)

Equations 1 and 2 showed predominant impacts of weather conditions on soil emission factors of nitrous oxide and leaching of soil nitrate. Therefore, using the method developed by the Intergovernmental Panel on Climate Change adopted for Canadian conditions (IPCC 2006), emissions of N2O from synthetic N applications can be estimated as follows:

$$ {\text{CO}}_{2} \;{\text{eq}}_{\text{SNF}} = {\text{Q}}_{\text{SNF}} \times \left\{ { ( {\text{FRAC}}_{\text{GASM}} \times {\text{EF}}_{\text{VD}})+\,{\text{EF}} + ( {\text{FRAC}}_{\text{LEACH}} \times {\text{EF}}_{\text{LEACH}} )} \right\} \times 4 4 / 2 8\times 310 $$
(3)

where CO2 eqSNF is the total emissions from the synthetic N fertilizer application (kg CO2 eq ha−1), QSNF is the quantity of synthetic N fertilizer applied (kg N ha−1), FRACGASM is the fraction of synthetic N fertilizer that volatilizes as NH3- and NOx−N {FRACGASM = 0.1 kg N kg−1 N} (IPCC 2006), EFVD is the N2O emission factor for volatilized NH3- and NOx-N {EFVD = 0.01 kg N kg−1 N} (IPCC 2006). EFLEACH is the N2O emission factor for nitrate leaching {EFLEACH = 0.0075 kg N kg−1 N} (IPCC 2006), 44/28 is the conversion coefficient from N2O–N to N, and 310 is the global warming potential of N2O.

When urea is applied to a soil, during the hydrolysis, the carbon contained in the urea is often released as CO2 (IPCC 2006). Thus, emissions of CO2 from urea-based N fertilizer were calculated as follows:

$$ {\text{CO}}_{2} \;{\text{eq}}_{\text{SNF - CO2}} = {\text{Q}}_{\text{SNF - UREA}} \times \frac{ 1 2}{ 2 8} \times \frac{ 4 4}{ 1 2} $$
(4)

where CO2 eqSNF-CO2 was the emissions of CO2 from the urea application (kg CO2 eq ha−1), QSNF-Urea was the quantity of urea-based N fertilizer applied (kg N ha−1), 12/28 was the ratio of C to N, and 44/12 was converted C to CO2.

Crop residue N, including the N in stover (measured each year) and roots (estimated from average values derived from measurements) at harvest, was also considered as N source for nitrification and denitrification, contributing directly and indirectly to N2O emissions. Similar to synthetic N fertilizer application, emissions from crop residue decomposition were calculated as follows:

$$ {\text{CO}}_{ 2} \;{\text{eq}}_{\text{CRD}} = {\text{Q}}_{\text{CRD}} \times \left\{ {{\text{EF}} + ( {\text{FRAC}}_{\text{LEACH}} \times {\text{EF}}_{\text{LEACH}} )} \right\} \times 4 4 / 2 8\times 310 $$
(5)

The quantity of crop residue N (QCRD) was calculated using the above- and belowground crop residue biomass multiplied by its respective N concentration. Crop residues usually decompose during winter or spring of the following year, but the exact time of the decomposition cannot be determined precisely. Therefore, all N contained in the crop residues were assumed to be released in the same year of production.

It is known that the Haber–Bosch process combining N2 with hydrogen gas (H2) into ammonia (NH3) which is used to make fertilizer is energy and emission intensive. Lal (2004) conducted an extensive literature review on emissions from manufacturing fertilizers and reported an emission factor of 4.8 kg CO2 eq kg−1 N, 0.73 kg CO2 eq kg−1 P2O5 and 0.55 kg CO2 eq kg−1 K2O from production, transportation, storage, and transfer of fertilizers to farm fields. In this paper, we estimated emissions from the production of N, P and K using the amount of N, P2O5 and K2O applied on a per hectare basis multiplied by the average emission factors (Lal 2004). Herbicides and fungicides were used in growing maize. Although emission factors for each individual pesticide are not available at the present time, we assumed that the emissions during processes of production, transportation, storage, and field application were similar among pesticides within a similar category. Thus, an average emission factor of 23.1 kg CO2 eq kg−1 was used for herbicides and 14.3 kg CO2 eq kg−1 for fungicides, based on the active ingredient of a fungicide or herbicide product (Lal 2004). The emissions associated with miscellaneous farming operations such as seeding, spraying, and crop harvesting were estimated using a factor of 14 kg CO2 eq ha−1 for planting, 5 kg CO2 eq ha−1 for herbicide and fungicide spraying, and 37 kg CO2 eq ha−1 for harvesting (Lal 2004). Total GHG emissions were calculated as CO2 equivalent and expressed as kg CO2 eq ha−1. The C footprint of maize production (grain yield measured for individual plots each year) was expressed as kg CO2 eq kg−1 grain.

$$ {\text{C}} {\hbox{-}} {\text{footprint}} = {\text{Total GHG emissions }}\left( {{\text{kg}}\;{\text{CO}}_{ 2}\;{\text{eq}}\;{\text{ha}}^{ - 1} } \right)/{\text{grain yield }}\left( {{\text{kg}}\;{\text{ha}}^{ - 1} } \right) $$
(6)

Statistical analysis

Statistical analyses of data were performed with the MIXED procedure of SAS (Littell et al. 2006). Due to differences in N levels across the entire experimental period, the data across years were considered for two phases (phase I = 1992–2001 and phase II = 2002–2010) where N and rotation were considered fixed effects, year, replication and interactions random effects. Phases were introduced and herewith used to reflect the change in N fertilizer levels since 2002. For each phase of the study, analysis of the data was first run on a balanced rotation design, and then a separate analysis was performed on the data including continuous maize only so that all the N treatment levels could be compared. The MANOVA option of the GLM procedures was used to calculate partial correlation coefficients between different variables and thus major variables contributing to C footprint of maize production could be identified. Quantitative relationships between total GHG emissions (or C footprint) and fertilizer application were established for each rotation, using the linear regression procedures of SAS. Differences between treatment means were considered significant if P ≤ 0.05.

Results

Both crop rotation and N treatments had significant effects on grain yield, stover biomass, grain and plant N concentrations, and N uptake each year, except for those in 1992 when the experiment was first initiated. Over the last 19 years, grain yield of maize with the 0 N treatment was reduced to <30 % of the original yield (from 6.6 Mg ha−1 in 1992 to 1.95 Mg ha−1 in 2010) under continuous monoculture maize (MM), or to 73 % of the original yield (from 8.1 Mg ha−1 in 1992 to about 5.9 Mg ha−1 in 2010) under SM. In comparison, maize grain yield of the FM rotation with 0 N yielded 8–10 Mg ha−1 each year, similar to that of well fertilized (200 kg N ha−1) continuous MM. In general, grain yield of maize responded to N fertilizer rates up to 150–200 kg N ha−1 in continuous culture or up to 100 kg N ha−1 in the SM system. A portion of this agronomic data was published in Ma et al. (2003).

Climatic conditions and general trends of treatments effects on total GHG emissions and the C footprint

The growing season rainfalls over the 19 cropping years varied from 337 mm in 1997 to 707 mm in 2006 while the potential evapotranspiration rates ranged from 525 mm in 2000 to 641 mm in 1999 (Fig. 1a). As a result, both FRACLEACH and EF varied from 17 % in 1997 to 39 % in 2006, and from 0.0084 kg N2O-N kg−1 N in 1997 to 0.0172 kg N2O-N kg−1 N in 1995, 1996, 2000, 2003, 2006, 2009 and 2010, respectively (Fig. 1b). Accordingly, total GHG emissions and the C footprint were largely affected by the dominant weather conditions, particularly the growing season rainfall (Ma et al. 2005; Rochette et al. 2008).

Fig. 1
figure 1

Precipitation, plant evapotranspiration (PE), emission factor (EF) and the fraction of leached N (FracLEACH) during the field experiment (1992–2010) at Central Experimental Farm of Agriculture and Agri-Food Canada, Ottawa, Ontario. a precipitation and plant evapotranspiration; b emission factor (EF) and the fraction of leached N

Full size image

Across the 19 years of the experiment, the highest total GHG emissions of 2176 kg CO2 eq ha−1 occurred in 1995 for phase I, and of 2659 kg CO2 eq ha−1 in 2003 for phase II. The highest C footprint of 0.435 kg CO2 eq kg−1 grain and 0.756 kg CO2 eq kg−1 grain were recorded, in 2000 and 2002 for the two phases, respectively. Among the three rotation systems, maize following a forage legume (FM) always produced the highest total GHG emissions (P < 0.01) (Tables 1 and 3), but had the lowest C footprint (P < 0.01) (Fig. 2). In contrast, continuous maize (MM) had the lowest total GHG emissions (Table 3) and the highest C footprint (Fig. 2). Total GHG emissions and the C footprint both increased with increasing N application in each cropping system (Fig. 2). There were significant rotation-by-N and year-by-N interactions for total GHG emissions, and year-by-rotation-by-N on the C footprint during phase II of the study (Table 1).

Table 1 Analysis of variance of the main effects of N fertilizer, crop rotation and their interactions on the GHG emissions from crop residue decomposition (CresidE), total GHG emissions (Temit) and the carbon footprint (Cprint) of maize during the 1992–2001 and 2002–2010 phases of the experiment
Full size table
Table 2 Partial correlation coefficients of the GHG emissions from crop residues decomposition (CresidE), total GHG emissions (Temit) and the carbon footprint (Cprint) with different components of maize biomass as affected by N treatment and rotation with a forage legume or soybean from 1992 to 2010
Full size table
Table 3 Total GHG emissions and the contributions of N related factors to total GHG emissions in a 19-year field study on maize production
Full size table

Continuous maize monoculture

In the continuous maize treatment, three (0, 100 and 200 kg N ha−1) and six levels of N (0, 50, 100, 150, 200 and 250 kg N ha−1) were used during phase I (1992-2001) and phase II (2002–2010), respectively. Fertilizer N application significantly (P < 0.01) affected emissions from the crop residue (stover and roots) decomposition (CresidE), the total GHG emissions and the C footprint, and there always were year-by-N interactions for these variables during both phases of the study (Table 1).

The highest N treatment often produced the greatest crop residue while the lowest crop residue occurred with the 0 N treatment across all years (data not shown). The CresidE was greater at high rather than at low N levels, but the magnitude of increase in CresidE due to N application varied from year to year with the highest CresidE in 2000 (average of 434 kg CO2 eq ha−1) and 1996 (average of 430 kg CO2 eq ha−1), and the lower values of 175 and 213 kg CO2 eq ha−1 in 2002 and 2007, respectively (Fig. 3). CresidE represented only a small portion of total emissions for maize production from various sources, and rainfall amounts and its distribution during the growing seasons exhibited a large impact on the relative contribution of N-related factors to total GHG emissions (Fig. 1, Table 3). Total GHG emissions under continuous maize production increased linearly with increasing fertilizer N rates (Table 3, Fig. 3), with an average of 16.4 kg CO2 eq ha−1 for each kg N applied (Fig. 2). The highest levels of both total GHG emissions and C footprint were found in the 200 kg N ha−1 treatment during phase I and with the 250 kg N ha−1 treatment during phase II, respectively, while their lowest values were noted at 0 kg N ha−1 across all years (Fig. 3). Without N application, CresidE accounted for 39–57 % of total GHG emissions while other sources (tillage, pesticides, planting and harvesting etc.) accounted for the remainder (Table 3). On average, the C footprint for the 0 N treatment was 0.10 kg CO2 eq kg−1 grain in phase I and 0.127 kg CO2 eq kg−1 grain in phase II. In comparison, the highest C footprint of 0.454 and 0.583 kg CO2 eq kg−1 grain were noted for the highest rates of N application, respectively during the same phases.

Fig. 2
figure 2

Effects of N fertilizer rates on emissions from crop residue N, total GHG emissions and C footprint of maize in rotation with a forage legume (FM) or soybean (SM), compared to those under continuous maize monoculture during 2002–2010. The error bars indicate the SEM

Full size image
Fig. 3
figure 3

Effects of N fertilizer rates on emissions from crop residue N, total GHG emissions and C footprint of continuous maize monoculture during 2002–2010. The error bars indicate the standard error of the mean

Full size image

Maize in rotation with a legume

When data were analyzed as a balanced rotation-by-N application experiment, year, rotation and N fertilizer application significantly (P < 0.01) affected both CresidE and total GHG emissions in both phase I and phase II, and affected the C footprint in phase II of the experiment (Table 1). There were significant interactions (P < 0.001) between year and N treatments for all three variables in phase II, but only C footprint showed a significant interaction (P < 0.05) during phase I (Table 1). The year-by-rotation-by-N three way interaction occurred only for C footprint during phase II of the experiment (Table 1).

Emissions from crop residue decomposition differed greatly among rotation systems. With increasing N fertilizer rates, CresidE for SM was similar to that of FM, but significantly lower than that of MM (Fig. 2). Across the 19 years, there were significant differences in CresidE among rotation treatments in 84 % of the years at the 0 (Fig. 4) and the 100 kg N ha−1 (Fig. 5). Higher CresidE occurred in FM than in both SM and MM in 58 % of the years. Both FM and SM were statistically similar but had higher CresidE than MM in 26 % of the years (Figs. 4, 5). Total GHG emissions and the C footprint were found to be significantly affected by crop rotation (Figs. 4, 5) in 84 and 63 % of the years, respectively. Both FM and SM resulted in higher total GHG emissions than MM in 21 % of the years. In general, increasing N fertilizer rates greatly increased CresidE, total GHG emissions and the C footprint (Fig. 5). However, there were some exceptions; SM had a higher C footprint than MM or FM in year 2000 and a lower C footprint than MM or FM in 2008 (Figs. 4 , 5). In 2000, the maize hybrid ‘Pioneer 38W36’ was injured by the residue of the herbicides LexoneDF 75 (metribuzin) and Pursuit (imazethapyr) applied to the soybean in the previous year, and as a consequence, maize following soybean in 2000 produced the lowest yield and correspondingly the highest C footprint (Figs. 4 and 5). If the data in 2000 were considered as outliers and omitted, the SM system with 100 kg N ha−1 applied to maize produced, on average over the 18 years, similar grain yield as MM fertilized with 200 kg N ha−1 with 42 % lower total GHG emissions and a 41 % lower C footprint (Fig. 6). On average, responses of grain yield to increasing N rates were quadratic in nature, and there were often minimal differences in yield between 150 and 250 kg N ha−1 under continuous MM culture. With rotational cropping systems, either SM or FM, grain yields of N rates between 100 and 150 kg N ha−1 often not differed statistically (Fig. 6), and the economic optimum rate of N frequently obtained at 100 kg N ha−1 or less. In contrast, total GHG emissions and the calculated C footprints increased linearly with fertilizer N rates in all systems (Figs. 2 and 6).

Fig. 4
figure 4

Comparison between crop rotations and monoculture system in emissions from crop residue N, total GHG emissions and C footprint of maize production during 1992–2010. FM-0, SM-0 and MM-0 represent maize following a forage legume (alfalfa or red clover), maize following soybean and continuous maize monoculture, respectively each with 0 N fertilizer application. The error bars indicate the standard error of the mean

Full size image
Fig. 5
figure 5

Comparison between crop rotations and monoculture system in emissions from crop residue N, total GHG emissions and C footprint of maize production during 1992–2010. FM-100, SM-100, MM-100 and MM-200 represent maize following a forage legume (alfalfa or red clover), maize following soybean and continuous maize monoculture, respectively each with 100 (and 200 in the case of MM) kg N ha−1 application. The error bars indicate the standard error of the mean

Full size image
Fig. 6
figure 6

Means of grain yield, estimated total GHG emissions and C footprint across the 19-year period of the study, as affected by N fertilizer rates in each rotation system. Note N rates of 50 and 150 kg ha−1 in all systems and 250 kg ha−1 in MM system occurred from 2002 to 2010. The error bars indicate the standard error of the mean

Full size image

Correlations among different variables

Partial correlation analyses showed that both CresidE and total GHG emissions were positively correlated with grain yield, stover biomass, and crop residue N. These parameters were negatively correlated with HI (Table 2). The C footprint was negatively correlated with grain yield and stover biomass in phase I and phase II, and with CresidE and total GHG emissions in phase II. This is of critical importance as it indicates that the C footprint can be reduced while simultaneously increasing the yield of maize, one of the most important field crops worldwide.

Discussion

In this study, our emphasis was given to the GHG emissions derived from N (either in production of the fertilizer, or N application to the soil or crop residue N decomposition) and the potential impact of management practices on GHG emissions and C footprint of maize production. Results from this study demonstrated that high N application increased both crop residue N and the CresidE, with the highest values resulting from the highest N rate. This suggests that while N application increases both the uptake and concentration of N in plant components, grain yield may be limited by the capacity of resource capture (i.e., net assimilation rate) and translocation efficiency under high tissue N concentrations. High tissue N concentrations due to higher N supply were also reflected in the total N content in crop residue. This could be due to either a high N uptake and/or a poor N remobilization from the stover to the grain during the grain filling period. Both CresidE and total GHG emissions were highest at 250 kg N ha−1 in our study. Our results are in general agreement with other studies in which GHG emissions were largely associated with inorganic N application (Hillier et al. 2009). As a result, the C footprint of maize, expressed as CO2 eq kg−1 grain under a monoculture system was the highest at 250 kg N ha−1. This suggests that while the C footprint of crop production is largely related to N application, it also depends on the ability of the crop to convert that N into grain yield. Our study is in agreement with the common view that the C footprint of food crop production can be reduced by good agronomic measures that improve crop yield with optimum or moderate N application and improved N use efficiency (Cole et al. 1997; Monteny et al. 2006; van Groenigen et al. 2010).

With 0 kg N ha−1 GHG emissions resulting from crop residue decomposition constituted, on average, about 37 % of the total GHG emissions for MM, 49 % for SM and 57 % for FM with the remainder coming from additions of P and K fertilizers, herbicides, and various farm operations. The GHG emission contribution from crop residue decomposition reflected the annual variations in grain yields of maize because GHG emissions from P, K and herbicide applications and field operations generally stayed constant and independent of weather conditions. Relative contribution of GHG emissions from synthetic N fertilizer production, storage and transportation linearly increased with N rates. The narrower range of this contribution indicated a dominant influence on the emissions regardless of crop yields and weather conditions. Relative soil GHG emissions resulting from synthetic N application, were only slightly different among the different crop rotations and N rates, but exhibited tremendously annual variations due mainly to weather conditions that controlled EF and FRACLEACH (i.e. rainfall and evapotranspiration).

Forage legume after harvest retained most of its fixed N in the soil, benefitting the following maize crop, whereas the majority of the N fixed by the soybean in SM was removed with the harvested grain. Our results indicate that maize in rotation with a forage legume or soybean significantly increased the CresidE and total GHG emissions in almost all of the cropping years. The greater increase in GHG emissions in maize production following a forage legume or soybean suggests that total GHG emissions largely depend on the content of the available soil N from any source, either synthetic or biologically fixed N for the bacterial processes of denitrification and nitrification (Bouwman and Boumans 2002). While crop rotation demonstrated an increase in total GHG emissions, both FM and SM systems showed a large reduction in the C footprint of maize production with an average of 34 and 22 %, respectively compared to maize monoculture. The higher reduction in the C footprint of maize production in the FM rotation compared to the SM rotation can be explained by the higher maize grain yield after forage than soybean. The higher yield of maize following a forage legume can be explained by higher N levels in the soil and the plant (Ma et al. 2000; Ma et al. 2003). However, our results suggest that while both FM and SM cropping systems reduce the C footprint of maize production, maize following alfalfa or red clover appeared to be a most efficient rotation system in reducing the C footprint of maize production (Robertson et al. 2000; Paustian et al. 2004; Rochette and Janzen 2005). Reduction in the C footprint of maize production in rotation with a legume was due to the legume N credit and the resulting greater maize yield. Further reduction in the C footprint of maize following a legume was associated with an improved N uptake because of the N released from mineralization of organic N (Ma et al. 2003). Similarly in western Canada, Gan et al. (2011b) noted a 17 % lower C footprint of durum wheat (Triticum durum L.) production following a legume crop than following a cereal crop.

Our study currently examines only half of the GHG emissions and C footprint from a maize-legume rotation because we did not calculate these parameters for the legume half of the rotation. However, the treatments to the maize were made on an annual basis and the sustainable production of maize is the focus of our study. We feel confident that our conclusions for GHG emissions and the C footprint for maize are valid and it is our intent to determine the GHG emissions and C footprints of the legume part of each rotation which would then permit us to calculate these parameters on a bi-annual cumulative basis.

Fertilizer N application significantly increased maize stover biomass either in rotation with a legume or in a monoculture system. Higher stover biomass production in the FM and SM rotations with added N fertilizer can be explained by higher available soil N and higher N uptake as shown in the content of N in stover and crop residue (Yusuf et al. 2009). This can be further explained by a significant positive association (P < 0.05) between grain yield and N content in stover or crop residue. Our results indicate that N fertilizer increased crop residue N content across all rotation systems showing the highest N content in crop residue at 150 kg N ha−1 in almost all the cropping years. However, total GHG emissions and the part of GHG emissions from CresidE increased with N application across rotation systems. The 150 kg N ha−1 treatment had the highest GHG and CresidE in almost all the years of the study. The results obtained here indicated that the amount of total emissions in FM and SM with 100 kg N ha−1 treatment was slightly higher than MM with the same N level. Our results also indicate that FM with 100 kg N ha−1 treatment decreased total GHG emissions and CresidE by 40 and 3 %, respectively compared to MM with the 200 kg N ha−1 treatment. In contrast, SM with the 100 kg N ha−1 treatment showed a reduction in total GHG emissions and CresidE by as much as 42 and 21 %, respectively relative to MM with the 200 kg N ha−1 treatment.

The C footprint was the highest at 100 kg N ha−1 during phase I and 150 kg N ha−1 during phase II (when 150 kg N ha−1 was the highest N rate), across all cropping systems. However, FM and SM with 100 kg N ha−1 treatment showed a reduction in C footprint by 17 and 8 %, respectively compared to MM with the same N level. Moreover, the FM and SM systems with 100 kg N ha−1 treatment demonstrated a reduction in C footprint by as much as 46 and 42 %, respectively relative to continuous maize with 200 kg N ha−1 treatment. As responses of maize grain yield to fertilizer N rates are generally quadratic (Ma et al. 2005), and GHG emissions increased linearly with increasing N rates, the highest grain yields and the lowest C footprints were observed most frequently in a SM or FM system at 100 kg N ha−1 fertilizer. Therefore, our study suggests that crop rotation options which reduce GHG emissions without yield penalty are available for Canada and temperate regions.

Conclusions

In this long-term field study, high rates of N fertilizer often increased maize grain yield, but also increased the total GHG emissions and the C footprint of maize production across all the rotation systems. Overall, total GHG emissions of maize production were highly correlated with the amount of N fertilizer, the N content of the stover and crop residue and the emissions from the crop residue. When maize was grown in rotation with a forage legume or soybean the C footprint (kg CO2 eq kg−1 grain) decreased. The magnitude of the reduction in the C footprint of maize production in rotation with a forage legume or soybean differed with varying levels of supplementary N application provided in the maize year of the rotation. Provided that neither field-level emissions nor footprint influences farmgate prices, results from this study showed that moderate fertilization of 100 kg N ha−1 produced the most economic grain yield and relatively lower GHG emissions and a lower C footprint than when the maize crop was grown in monoculture with 200 kg N ha−1. Increased grain yield and reduced C footprints in maize production can be achieved through appropriate N application and rotation with a forage legume or soybean.