Introduction

The global estimate for organic carbon (OC) sequestration by no-till (NT) soils is 0.57 Mg ha−1 year−1 (West and Post 2002), while for southern Brazilian subtropical soils it is 0.48 Mg ha−1 year−1 (Bayer et al. 2006). In NT soils, however, OC sequestration should not be taken as the ultimate contributor to the mitigation of global warming. A full assessment of the soil and atmosphere fluxes of the three principal greenhouse gases (CO2, CH4 and N2O) and of the CO2 equivalent costs related to agronomic practices (operations and inputs) is essential to obtain the net global warming potential (GWP) of a system (Robertson and Grace 2004; Six et al. 2004; Mosier et al. 2005) because soil management-related gains achieved by reducing the flux of one greenhouse gas may be offset by increases in the flux of another gas (Six et al. 2004; Gregorich et al. 2005). Besides, a fair comparison between management systems can be achieved using the greenhouse gas intensity (GHGI), calculated by dividing the GWP of a soil management system by the crop yield for the same system (Mosier et al. 2006). Due to the strategic importance that NT-based agriculture has, and will increasingly have, on food production, more studies are needed to investigate the impact of NT management on soil N2O and CH4 emissions, soil OC sequestration, hidden carbon costs of farm operations and net GWP and GHGI, particularly in tropical and subtropical regions where information is still scarce.

The known OC sequestration potential of NT may be offset by increased N2O emission (Gregorich et al. 2005; Li et al. 2005), with N2O being 298 times more potent as a active radiative greenhouse gas as compared to CO2. However, this area is still controversial and further studies are needed to clarify the situation. For example, it has been shown that NT increases N2O emission (MacKenzie et al. 1997; Ball et al. 1999; Baggs et al. 2003), particularly because of higher soil moisture due to mulching of residual plant material and by soil compaction and higher bulk density and lower aeration porosity (Ball et al. 1999), all of which create more favorable conditions for denitrification. Other studies, however, have shown lower N2O emissions in NT soil as compared to CT soil, although the reasons for this are as yet unclear (Chatskikh and Olesen 2007; Malhi et al. 2006; Gregorich et al. 2008; Petersen et al. 2011; Mutegi et al. 2010; Ussiri et al. 2009; Passianoto et al. 2003). Rochette (2008) reviewed the available information and concluded that while emissions were comparable for CT and NT systems in well-aerated soils, there were higher N2O emissions from NT soils in poorly aerated soils, such as the Gleysols investigated in some studies (MacKenzie et al. 1997; Ball et al. 1999).

Studies conducted in the Brazilian Subtropics (Jantalia et al. 2008) and the Brazilian Cerrado scrubland (Metay et al. 2007) have shown similar N2O emissions for NT and CT soils. However, other studies in the Cerrado (Carvalho et al. 2009), and the Subtropics (Escobar et al. 2010) have shown higher emissions in NT soil, while studies carried out in the Amazon (Passianoto et al. 2003; Carmo et al. 2005) have reported lower emissions for NT soils. Studies documenting the full accounting expressed in the net GWP for NT and including gas balances and the carbon costs of agronomic practices are absent or rare for Brazilian conditions, although some studies have already reported the results of gas balances (Carvalho et al. 2009; Metay et al. 2007).

Legume based NT cropping systems have been reported to increase N2O emissions compared to non-fertilized grass-based systems (Gomes et al. 2009). Nonetheless, it would be reasonable to expect lower N2O emissions for grass-based systems due to the immobilization of nitrogen induced by grass residues, which have high ratios of carbon to nitrogen (C:N), lignin to nitrogen and lignin to polyphenol (Gomes et al. 2009; Frimpong and Baggs 2010).

Aerobic soils, either under natural conditions (Dalal and Allen 2008) or subjected to agricultural management (Powlson et al. 1997), are important CH4 sinks because soil methanotrophs use CH4 as source of carbon and energy, oxidizing it to CO2 (Le Mer and Roger 2001). However, methanotrophs are sensitive to soil disturbance (Hütsch 1998), with CH4 oxidation rates reported to be slightly higher in undisturbed NT soil than in ploughed CT soils (Ball et al. 1999; Hütsch 1998; Six et al. 2004).

The hypothesis of this study was that, as compared to CT, NT management in a subtropical Ferralsol can mitigate greenhouse gas emissions by promoting OC sequestration and lowering N2O emission. The objective was to assess the potential contribution of grass-based NT management to the mitigation of global warming in a subtropical Ferralsol, taking into account N2O and CH4 fluxes, soil OC sequestration and the CO2 equivalent costs of agronomic practices for a full accounting of the GWP of NT systems.

Materials and methods

Site and experimental description

This study was conducted from September 2008 to September 2009 in a field experiment established in the winter of 2005 at the ABC Foundation experimental station located near the town of Castro in the southern Brazilian state of Paraná (24°47′53′′S; 49°57′42′′W, 996 m altitude). The climate is humid subtropical (Cfb, Köppen classification), the average monthly temperature is 23°C in the warmest month (January) and 13°C in the coldest (July) and the mean annual precipitation is 1,400 mm (Caviglione et al. 2000). The soil is a clayey Umbric Ferralsol (IUSS 2006) or Latossolo Bruno (Embrapa 2006), with 384 g kg−1 sand, 177 g kg−1 silt and 439 g kg−1 clay in the 0- to 20-cm depth.

The area was originally under native grassland (Campos Gerais) but around 1960 (no precise record of the date is available) it was converted to annual cropland for commercial wheat (Triticum aestivum L.) and soybean [Glycine max (L.) Merr] cropping under CT, this management system being continued until the establishment of the experiment in 2005. The original aim of the experiment was to assess crop and soil parameters (mainly physical) related to soil use and tillage systems.

The experiment was set up using a split-plot randomized complete block design with four replicates. The main plots (70 m × 10 m) comprised soil use systems in winter, with Italian ryegrass (Lolium multiflorum Lam.) being grazed, hayed or left only as cover crop. Maize (Zea mays L.) was cropped in summer for silage production. The subplots (10 × 10 m) covered seven tillage types, of which we selected conventional tillage (CT) and no-till (NT) plots in the soil use system where ryegrass was used only as winter cover crop. For the CT plots we applied one heavy disking operation (∼15 cm deep) and two leveling disking operations (∼10 cm deep) in spring to incorporate ryegrass biomass before planting maize and again in autumn before sowing ryegrass. We used a heavy disker equipped with 25 inch diameter disks that produced similar effects to ploughing (i.e., a layer cut and inversion, residue incorporation and formation of a compacted pan below the tilled layer). In southern Brazil, this CT management system based on heavy disking up to 15 cm deep represents about 70 % of the area in which soil is tilled (Derpsch et al. 1991). For the NT plots, the ryegrass cover crop was desiccated with a glyphosate-based herbicide, equivalent to adding glyphosate at the rate of 1,200 g ha−1, before planting maize, the same herbicide regime being used to desiccate spontaneous weeds before sowing the ryegrass. The NT systems in this region still depend on the application of herbicide. The NT management implies no tillage operation over the years. Since treatments were based on an annual succession of oat/maize, the cropping system and farm operations were repeated every other year for each treatment.

Maize was planted in October 2008, in rows spaced at 80 cm (seed = 20 kg ha−1 ≈ 60,000 plants ha−1) and a total nitrogen application of 165 kg ha−1 was split-applied using 40 kg ha−1 of 15-30-00 fertilizer with N as NH +4 between seed rows at planting and 125 kg ha−1 of 25-00-25 urea-N as an unincorporated sidedress when the maize reached the V4 stage with four completely expanded leaves. We also applied 80 kg ha−1 P2O5 at planting and 125 kg ha−1 K2O at sidedress. Post emergence herbicide (nicosulfuron, active ingredient = 60 g ha−1) was applied when the maize was at the three-leaf V3 stage. In February, the maize was harvested mechanically for silage when the grains were milky-to-solid as starch. The soil was left fallow for about 2 months and in April 30 kg ha−1 ryegrass seed was sown, in rows spaced 17 cm apart. No fertilizer was applied to the ryegrass cover crop. Lime was applied at a rate of 2 Mg ha−1 every 3 years on the surface of the soil for NT and incorporated by tillage for CT.

Air temperature and precipitation information from September 2008 to September 2009 were collected at the nearest meteorological station (10 km away).

Measurements of N2O and CH4 fluxes from soil

The N2O and CH4 fluxes were measured from 26 September 2008 to 16 September 2009, in the 4th year of the experiment, and included 15 sampling events during maize growth in spring (September–November 2008) and 15 during ryegrass cycle in autumn/winter (April–September 2009). Measurements could not be performed from December 2008 to March 2009 due to technical problems with the chromatograph but since other studies observed that this is a period of background fluxes (Gomes et al. 2009; MacKenzie et al. 1997; Petersen et al. 2011) because no fertilization or tillage was carried out, and because of the intense nitrogen uptake by maize, the information regarding annual gas emissions remained reliable.

Static PVC chambers (Mosier 1989; Parkin et al. 2003) 20 cm high and 25 cm in diameter were deployed on metal bases (collars) to collect air samples. Three metal bases to support chambers were installed for CT and NT plots, within a previously delimitated mini-plot of 2.4 m × 2.4 m. The bases were driven 5 cm into the soil 48 h before the first sample was taken and kept in place continuously, except for tillage and sowing operations. Each sampling began at 9.00 a.m. and lasted for 45 min, with gas samples being collected in 20-mL polypropylene syringes at 15-min intervals (0, 15, 30 and 45 min). The chamber was water-sealed in a gutter surrounding the metal-base. Headspace temperature was monitored and air was mixed with a 12 V fan installed at the top of the chamber. Samples were analyzed in a Shimadzu 2014 Chromatograph, equipped with flame ionization detector (FID) and electron capture detector (ECD). Analyses were carried out within 24 to 36 h of sampling. Emission fluxes were estimated from the angular coefficient of the linear model fitted to describe the increase in headspace gas concentration during the 45-min deployment.

The N2O and CH4 fluxes were calculated using the following equation:

$$ {{f}} = \frac{{\Delta {{Q}}}}{{\Delta {{t}}}}\frac{{{{PV}}}}{{{{RT}}}}\frac{{{1}}}{{{A}}} $$

where, f is the N2O or CH4 flux (μg N2O-N or CH4-C m−2 h−1), Q is the quantity of each gas in the chamber at the sampling moment (micrograms of N2O-N or CH4-C per chamber), P is the atmospheric pressure inside the chamber, assumed as being 1 atm, V is the chamber headspace volume (liters), R is the ideal gases constant (0.08205 atm L mol−1 K−1), T is the inner temperature of the chamber at sampling (K) and A is the surface area of the chamber (m2).

The measured fluxes of N2O and CH4 were converted into mean daily fluxes, integrated along the 218 days of the assessment period and then normalized (extrapolated) to 365 days of 1 year to obtain the total annual emissions.

Soil physical parameters, inorganic N and organic carbon evaluation

For each gas sampling event, three soil subsamples from the 0–5 cm layer were collected within the 2.4 m × 2.4 m mini-plot to determine moisture and inorganic N (NH +4 and NO −3 ). Soil moisture was determined gravimetrically (105°C). Inorganic N was determined by steam-distillation (Mulvaney 1996), after extraction from approximately 5-g soil sample (at field moisture—correction was done later) in 50 mL 1 M KCl solution.

In December 2008, undisturbed soil core samples were collected in the 0–5, 5–10 and 10–20 cm layers, at two points per plot, to determine soil bulk density, total porosity, macroporosity and microporosity. Metal cylinders 5.6 cm diameter and 3.1 cm high were inserted vertically to the middle of each soil layer. Microporosity was assumed as being the volume of water contained in the core sample at after 48 h at 6 kPa in a tension table (Leamer and Shaw 1941). Macroporosity was the difference between total porosity and microporosity. Total porosity was calculated from bulk and particle densities (Danielson and Sutherland 1986), the latter being assumed to be 2.65 Mg m−3. Soil bulk density was determined after drying samples at 105°C. The water-filled pore space (WFPS) was calculated from the total porosity and gravimetric moisture.

A second set of samples were collected for OC assessment, adjacent to the points where core samples were taken. In the 0–5, 5–10, 10–20 cm layers samples were collected with a spatula and in the 20–40, 40–60, 60–80 and 80–100 cm layers with a Dutch auger. Samples were air dried at room temperature, crushed with a wooden roll and stored in plastic pots. To determine the OC concentration, about 20 g was crushed in a mortar to pass a 0.50-mm mesh and 400 mg analyzed by dry combustion in a Shimadzu TOC-VCSH analyzer.

OC stocks were calculated in equivalent soil mass, considering the CT system as the soil mass reference. This procedure normalizes the different effects that management systems have on soil bulk density. The equation presented by Sisti et al. (2004) was used to correct the values. To calculate the equivalent soil mass below 20-cm depth, we assumed that the bulk density was the same of that in 10–20 cm layer, because the opening of a trench to assess bulk density up to 1 m would cause severe damage to the long-term experimental plots.

The OC sequestration rate in NT relative to CT soil was calculated by the difference in OC stock between the two systems divided by the age of the 3.5 years of the experiment (June 2005, when the experiment started, to December 2008, when soil was sampled). This is referred to as the relative sequestration rate rather than the absolute sequestration rate because it does not consider the OC stock at the beginning of the experiment, information on which was not available. Annual OC sequestration rates were calculated for both 0 and 20-cm, the usual sampling depth that includes the tilled layer in CT and the 0–100 cm layers.

The annual OC sequestration rate (ΔOC) was used to estimate the net annual CO2 emission or mitigation, instead of the direct CO2 measurements by static chambers. The direct measurement by static chamber considers not only the CO2 from organic matter mineralization, but also that from root respiration, which cannot be distinguished from the former. The approach of using soil ΔOC instead of direct measurement of CO2 is supported by previous studies (Robertson et al. 2000; Robertson and Grace 2004; Six et al. 2004; Mosier et al. 2005).

CO2 equivalent costs of agronomic practices

CO2 equivalent costs associated to farm operations (tillage, sowing, spraying, fertilizer application and harvesting) and inputs (seeds, fertilizers, lime and herbicide) were calculated for the CT and NT systems, considering the particular nature of each of these experimental treatments and coefficients reported in the literature (Zanatta et al. 2007; Lal 2004; West and Marland 2002).

Global warming potential

The GWP of the CT and NT systems were calculated by summing the N2O emission, CH4 emission, soil OC sequestration (assumed to represent the soil CO2 balance) and the CO2 costs of agronomic practices in each system. The CT system was assumed as being the baseline system. Gas emission data were normalized into CO2eq, considering the global warming potential of N2O (298) and CH4 (25) relative to CO2 in a 100-year time horizon (IPCC 2007). We calculated the GWP of the NT system twice, one considering the OC sequestration rate in the 0–20 cm layer, and another considering in the 0–100 cm layer. The difference in GWP between the two systems was used to assess whether NT can mitigate greenhouse gas emissions relative to CT or not.

The greenhouse gas intensity (GHGI) was calculated by dividing the GWP by the maize silage yield, as proposed by Mosier et al. (2006). The silage yield was the average of the 4 years of the experiment (2005/2006 to 2008/2009).

Statistical analysis

Data were submitted to analysis of variance (ANOVA) and significance of differences between CT and NT systems was assessed by P-value of the Fisher F-test.

Results

Precipitation and air temperature

The recorded annual precipitation of 1,740 mm during the study year (21 September 2008 to 21 September 2009 = 5 days before the first air sampling to 5 days after the last) was above the normal rainfall of 1,400 mm. Monthly precipitation was below normal in April 2009 (21 vs 88 mm) and was above normal in January 2009 (296 vs 188 mm) and July 2009 (315 vs 112 mm) (Fig. 1) (Caviglione et al. 2000). In general, for other months, precipitation was normal.

Fig. 1
figure 1

Mean daily precipitation and mean daily air temperature in the experimental site, from 1 September 2008 to 30 September 2009

Full size image

June was the coldest month and its mean temperature of 11.7°C was below the normal 14.0°C (Fig. 1) (Caviglione et al. 2000). February, the warmest month, had a mean temperature of 20.8°C, below the normal 23.0°C.

CO2 equivalent costs

The total CO2 equivalent costs of agronomic practices, given in kg CO2eq ha−1 year−1, were similar for CT and NT, although with a slight tendency of being lower in NT (1,617 vs 1,723) (Table 1). The CT system showed costs due to tillage operations (176.0), whilst NT showed additional costs related to herbicides (65.6 more than CT). The major driver of CO2 costs was N fertilization (786.5), which accounted for 46 % and 49 % of the overall costs in CT and NT, respectively.

Table 1 Annual CO2 equivalent costs related to agronomic practices (operations and inputs) in conventional tillage (CT) and no-till (NT)
Full size table

N2O fluxes

The N2O fluxes, given in μg N m−2 h−1, during the 1st month of measurements, including 3 days before spring tillage in CT (sampling 1) until 1 day after sidedress application of 125 kg urea-N ha−1 to maize (sampling 10), varied from −7 to 27 in NT and from −10 to 38 in CT (Fig. 2a). From 7 to 11 days after spring tillage (samplings 4 and 5), however, CT emitted twice as much N2O as NT (on average, 33 vs 15). One day after sidedress N application (sampling 10), the fluxes averaged a low value of 11 but increased sharply 3 days later (sampling 11), reaching an average of 96 across CT and NT. For the next 3 weeks, until sampling 15, fluxes varied from 80 to 133 in NT and from 94 to 367 in CT, this being the period where the highest emissions were recorded. In the ryegrass cycle (samplings 16 to 30), N2O fluxes were considerably lower than those observed in maize and were similar for CT and NT, except that 1 month after autumn tillage, a significant peak of 69 was observed for CT (sampling 20).

Fig. 2
figure 2

a Nitrous oxide and b methane fluxes from a subtropical Ferralsol subjected to conventional tillage (CT) and no-till (NT), in a maize-ryegrass succession. Underlined numbers refers to the sampling identification. *P ≤ 0.10, **P ≤ 0.05, *** P ≤ 0.01, according to F-test. No asterisk Difference not significant

Full size image

After integrating the N2O fluxes over time and extrapolating to the 1-year range, the total N2O emission was 1.26 kg N ha−1 year−1 in NT, which was almost half the emission of 2.42 kg N ha−1 year−1 in CT (P = 0.06) (Table 2). Converting into CO2eq, this difference between systems represented an emission reduction of 0.55 Mg CO2eq ha−1 year−1 in NT compared to CT (Table 2).

Table 2 Nitrous oxide (N2O) and methane (CH4) emissions, organic carbon (OC) stocks, CO2 equivalent costs of agronomic practices (operations and inputs), global warming potential (GWP) and greenhouse gas intensity (GHGI) in a subtropical Ferralsol subjected to NT relative to CT
Full size table

CH4 fluxes

Methane influxes were observed during maize cropping (sampling 1–15), except for two sampling events in CT where fluxes were positive (samplings 3 and 10) (Fig. 2b). For most of the ryegrass cycle (samplings 16–30), CH4 influxes were also observed, but an intense emission peak occurred in May (autumn) reaching 147 μg C m−2 h−1 in CT and 194 μg C m−2 h−1 in NT. Because of this peak, the CH4 annual flux was positive, being 1.15 and 1.08 kg C ha−1 year−1 in CT and NT, respectively, but the difference was not significant (P = 0.90) (Table 2).

Soil physical parameters and inorganic N

Effects of tillage system on soil bulk density, macroporosity and microporosity were restricted to only the 0–5 cm layer, where NT soil showed higher bulk density than CT (1.17 vs 1.09 Mg m−3), lower macroporosity (0.10 vs 0.18 m3 m−3) and higher microporosity (0.45 vs 0.41 m3 m−3) (Table 3).

Table 3 Physical properties in a subtropical Ferralsol subjected to NT and CT systems
Full size table

The WFPS during maize cropping (samplings 1–15) varied from 41 % to 67 % in CT, and from 49 % to 79 % in NT (Fig. 3). A wider variation occurred in the ryegrass season (sampling 16–30), both for CT (29 % to 68 %) and NT (42 % to 92 %). Overall, the WFPS was higher for NT (mean of 65 %) than CT (mean of 52 %). No correlation was found between WFPS and N2O fluxes (data not shown).

Fig. 3
figure 3

Water-filled pore space (WFPS) in the 0–5 cm layer of a subtropical Ferralsol subjected to conventional tillage (CT) and no-till (NT), in a maize-ryegrass succession. Underlined numbers refers to sampling identification

Full size image

The NH +4 concentration in the 0–5 cm layer increased sharply after the sidedress N was applied to maize (sampling 10) but reduced to close to zero 21 days later (Fig. 4a). During this period, the NH +4 concentration peaked at 173 mg N kg−1 in NT (2 days after application, sampling 11) and at 212 mg N kg−1 in CT (6 days after application, sampling 12). The largest differences between these systems were observed 6 and 12 days after application (samplings 12 and 13), although not significant. No correlation was observed between NH +4 concentration and N2O emissions (data not shown), although the NH +4 concentration might have affected the N2O emission at some specific times, such as the peak observed after application of fertilizer (Figs. 2a, 4a).

Fig. 4
figure 4

Inorganic nitrogen concentrations, a NH +4 and b NO −3 , in the 0–5 cm layer of a subtropical Ferralsol subjected to conventional tillage (CT) and no-till (NT), in a maize-ryegrass succession. Underlined numbers refers to the sampling identifications. The differences between tillage systems were not significant according to F-test (P ≤ 0.10)

Full size image

The NO −3 concentration in the soil also increased after application of fertilizer-N, reaching the highest concentrations 12 days after application (79 mg N kg−1 in NT and 66 mg N kg−1 in CT, sampling 13) (Fig. 4b), but this increase was not as great as that observed for NH +4 (Fig. 4a). From sampling 19 to 22, during the ryegrass cycle, the NO −3 concentration tended to be higher in NT, where it reached a concentration of 62 mg N kg−1 (sampling 22). As for NH +4 , the NO −3 concentration show no correlation with N2O emissions (data not shown).

Soil organic carbon

Significant differences in the concentration of OC between tillage systems were observed up to 20-cm depth (Fig. 5). In the 0–5 cm layer, a higher OC concentration occurred in NT than CT (35 vs 31 g kg−1) but the opposite was observed in 5–10 and 10–20 cm layers, where NT showed lower concentrations (Fig. 5).

Fig. 5
figure 5

Organic carbon (OC) concentration along the profile of subtropical Ferralsol subjected to CT and NT after 3.5 years (P values refer to the significance level of the difference, according to F-test)

Full size image

In the 0–20 cm layer, there was a tendency, although not significant (P = 0.36), for greater OC stocks in NT (67.20 Mg ha−1) than in CT (66.49 Mg ha−1) and so an OC sequestration of 0.20 Mg ha−1 year−1 in NT (Table 2). However, when integrating the whole 0–100 cm layer, the OC stocks in NT (234.61 Mg ha−1) were significantly larger (P = 0.01) than in CT (231.95 Mg ha−1), and here the difference leads to an OC sequestration rate of 0.76 Mg ha−1 year−1 in NT assuming CT as the baseline (Table 2).

Global warming potential

The net GWP of CT system was 2.90 Mg CO2eq ha−1 year−1, and that was driven mainly by the carbon costs of agronomic practices (59 %) and N2O emissions (39 %) (Table 2). The NT system emitted 1.26 kg N2O-N ha−1 year−1 and 1.08 kg CH4-C ha−1 year−1 and costs 1.61 Mg CO2eq ha−1 year−1 in agronomic practices (Table 2). Part of those emissions was offset by a sequestration of 0.20 Mg OC ha−1 year−1 (0.73 Mg CO2 ha−1 year−1) in the 0–20 cm layer and hence the net GWP in NT was 1.51 Mg CO2eq ha−1 year−1 (Table 2), indicating a difference of −1.39 Mg CO2eq ha−1 year−1 relative to CT (the negative value implies a mitigation of 1.39 Mg CO2eq ha−1 year−1 in NT compared to CT). However, when considering the OC sequestration in the whole 0–100 cm layer (0.76 Mg ha−1 year−1 = 2.79 Mg CO2 ha−1 year−1), the net GWP of NT dropped to −0.55 Mg CO2eq ha−1 year−1 and the difference to CT was enlarged to −3.45 Mg CO2eq ha−1 year−1 (Table 2).

Considering the maize silage yield in CT (16.95 Mg DM ha−1) and NT (17.33 Mg DM ha−1), the GHGI was 171.1 kg CO2eq Mg−1 silage in CT and considerably lower in NT, where it varied from 87.3 kg CO2eq Mg−1 silage, when considering OC sequestered only in the 0–20 cm layer, to −31.7 kg CO2eq Mg−1 silage, when considering the sequestration in the whole 0–100 cm profile (Table 2).

Discussion

N2O and CH4 fluxes

The N2O flux during most of the assessment period was close to the background emission (≤20 μg N m−2 h−1) and similar in CT and NT (Fig. 2a). However, the greatest differences between these two systems, with higher emissions in CT, became evident in temporary peaks associated mainly to tillage operations and application of fertilizer-N.

The effect of tillage operations was observed in two emission peaks related to CT: a small one at 7 to 12 days after spring tillage (samplings 4 and 5), and another at 29 days after autumn tillage (sampling 20) (Fig. 2a). According to Baggs et al. (2000), these emission peaks are related to a rapid stimulation of microbial decomposition and increased substrate supply to nitrification and or denitrification after incorporation of residues into the soil. Such emission peaks induced by N-mineralization generally occur within 2 weeks after residue incorporation by tillage (Baggs et al. 2000), agreeing with emission peak after spring tillage (7–12 days) but not after autumn tillage, where the emission peak came almost a month later (Fig. 2a), possibly because of the low soil moisture content at the time of tillage and on the week after (samplings 17 and 18, Fig. 3). April 2009 was a month of very low precipitation and possibly the first rain of May (21 mm, at 4 May) (Fig. 1) triggered the N2O emission in CT.

The application of urea-N at 125 kg ha−1 to maize significantly increased N2O emission 3 days later (sampling 11), both in CT and NT (Fig. 2a). The effect of fertilizer-N in increasing N2O emission has been reported widely (Bouwman et al. 2002; Baggs et al. 2003; Zanatta et al. 2010) and attributed to nitrification or denitrification induced by the rapid and great increase in inorganic N concentration in soil. A significant increase in NH +4 concentration occurred in both CT and NT after fertilization (sampling 10) (Fig. 4a), and this possibly caused the rapid increase in N2O emission (Fig. 2a, sampling 11). In this case, nitrification was possibly the process driving N2O emission, once the substrate for this process (NH +4 ) was present in a much higher concentration than the substrate for denitrification (NO −3 ) (Fig. 4a,b). However, the greatest difference between CT and NT came afterwards (sampling 13 to 15), where the response of N2O emission to N fertilization differed considerably in these two systems. The explanation for the lower N2O emission in NT was possibly a partial immobilization of fertilizer-N, thus reducing nitrification. The broadcasted N was applied in direct contact to the aboveground residue of ryegrass on soil surface or to root residue in a few centimeters below surface, both with high C:N ratios (>25:1) (Kuo et al. 1997), which could easily stimulate immobilization. The considerably lower NH +4 concentration in the 0–5 cm layer in NT compared to CT in samplings 12 and 13 is strong evidence of N immobilization in NT. For NO −3 , the concentration of which was much lower than that of NH +4 during this period, this was not observed and concentrations were comparable among the two tillage systems (Fig. 4b). Several studies have reported the existence of a “immobilization layer” 3–6 cm below the surface of NT soils (Dieckow et al. 2006; Cochran et al. 1980) that could contribute to reducing nitrification or denitrification after N application on soil surface or within the 0–5 cm layer, and thus to reduce N2O losses in NT soil. Gomes et al. (2009) also reported the possibility of N immobilization to reduce N2O emission, after observing lower N2O emissions (and sometimes even uptake) in NT crop rotation systems that included black oat (Avena strigosa) (C:N ratio of 33) as cover crop. Additionally, these authors suggest that immobilization would be useful to reach a better synchrony between inorganic N availability in soil and plant N uptake.

No correlation was observed between N2O emission rates and NH +4 concentrations (data not shown), although the NH +4 concentration might have affected N2O emission at some specific times, e.g., the peak observed after fertilizer application (Figs. 2a, 4a). Part of the reason for this lack of correlation can be related to the delay between the presence of NH +4 in soil and its effect in N2O emission. The highest NH +4 concentrations in soil (samplings 11 and 12, Fig. 4a) were observed to precede the N2O emission peak (samplings 14 and 15, Fig. 2a) by about 10–14 days.

Results from the study of Rochette (2008) point toward comparable emissions in CT and NT when soils are well aerated. In our study, however, the WFPS was always higher in NT than in CT. This suggests a less aerated condition in NT but, as reported earlier, this was not expressed in higher N2O emission in NT than CT. The lack of correlation between WFPS and N2O emission rates thus shows that soil moisture was not as important a controlling factor of N2O emissions. In the similar subtropical conditions of Southern Brazil, Jantalia et al. (2008) also found no relationship between WFPS and N2O emission.

Regarding CH4 fluxes, soil served as a sink during most of the assessment period, but the significant emission in May (Fig. 2b) turned the overall annual flux into positive emission, both in CT and NT (Table 2). No plausible explanation could be offered for this unexpected CH4 emission peak. In a study in the Brazilian Cerrado, Metay et al. (2007) also observed that soil was a CH4 source, either in CT or NT, although slightly lower in NT (0.25 vs 0.40 kg CH4-C ha−1 year−1). Usually, agricultural aerobic soils are regarded as important CH4 sinks (Powlson et al. 1997) and CH4 oxidizing activity is reported to be higher in NT than in CT (Hütsch 1998). So, the positive annual CH4 emission in this study (Table 2) should be interpreted with caution, since it might have been skewed because of the unexpected and significant emission peak of May. Despite this, the observed CH4 emission did not represent more than 5 % of the N2O emission, in CO2eq units.

Organic carbon stocks

The total OC stock in the 0–20 cm layer was not significantly different (P = 0.36) between CT (66.49 Mg OC ha−1) and NT (67.20 Mg OC ha−1) (Table 2), suggesting that inversion and OC redistribution occurred within the tilled layer of CT soil (Angers and Eriksen-Hamel 2008; Angers et al. 1997). Nonetheless, there was a tendency towards higher stock in NT soil.

The OC concentration in layers below 20-cm depth was slightly higher in NT than CT and when integrating the whole 0–100 cm profile, the total OC stock was significantly higher (P = 0.01) in NT than CT (Table 2). The difference of 2.66 Mg OC ha−1 accumulated in NT up to 100-cm during the 3.5 years of experimental duration denotes therefore an OC accumulation rate of 0.76 Mg ha−1 year−1 (2.79 Mg CO2 ha−1 year−1), and the potential of the NT system to contribute to global warming mitigation by soil OC sequestration (Franzluebbers 2010; Dieckow et al. 2009; Zanatta et al. 2007; Vieira et al. 2009; Sá et al. 2009). The combination of root maize contribution to C input (Eghball and Maranville 1993) and the benefits of NT in allowing deep root development (lack of subsurface compacted layer) might explain OC sequestration. This accumulation rate falls within the range of 0.5 to 1.5 Mg OC ha−1 year−1 for NT soils in subtropical Brazil, according to estimates by Boddey et al. (2010), after considering OC stocks measured up to 100 cm depth. Therefore, the OC assessment up to 100-cm depth seemed to be a more appropriate methodology to quantify the effects of tillage systems and the benefits of NT (Boddey et al. 2010; Sisti et al. 2004; Diekow et al. 2005).

Global warming potential

There are concerns about the possibility of N2O emission offsetting the gains of soil OC accumulation in NT (Li et al. 2005; Gregorich et al. 2005), but this was not the case in the current study. NT showed a significant contribution to mitigating global warming, both by lowering N2O emission and by promoting soil OC accumulation, so that the net GWP in NT (−0.55 Mg CO2eq ha−1 year−1) was −3.45 Mg CO2eq ha−1 year−1 lower than that of CT (2.90 Mg CO2eq ha−1 year−1 in CT). The OC sequestration represented 84 % of this reduction in GWP, being thus the main driving factor of mitigation in NT.

For major US cropping systems, Del Grosso et al. (2005) estimated net GWP values of 1.58, 1.06 and −0.55 Mg CO2eq ha−1 year−1 under CT, NT and for native systems, respectively. The same trend had already been observed by Robertson et al. (2000), in a study in Michigan, where GWP was higher in CT and almost neutral in NT (1.14 vs 0.14 CO2eq ha−1 year−1). In our study, we found a negative GWP for NT (−0.55 Mg CO2eq ha−1 year−1) when considering OC changes in the 0–100 cm, which indicates true mitigation of this system in acting as a net sink of greenhouse gases.

The particularities of the cropping system of this study, based only on grass-species of maize and ryegrass (cover crop), could have contributed to the lower emission results for NT. The root additions by maize (shoot was removed for silage) and ryegrass shoot and root surely contributed to OC accumulation in NT soil and since both species demand appreciable amounts of N and produce high C:N residues that potentially immobilize inorganic N in NT soil, the chances of N being prone to losses as N2O are lower throughout the year.

In spite of that, most of the 0.59 Mg CO2eq ha−1 year−1 emitted as N2O in NT (Table 2) was possibly related to N fertilizer application, as seen by the flux increment after sidedress (Fig. 2). This, together with the fact that almost half of the 1.61 Mg CO2eq ha−1 year−1 of costs in NT was also due to N fertiliser (Table 1), highlights the significant role of production and application of synthetic N in causing gaseous emissions, which reinforces the necessity of establishing best management strategies for this nutrient element.

With respect to the GHGI, this is an indicator of how much a management system emits to produce one unit of grain or dry matter (Mosier et al. 2006). In this study, maize was not for grain but for silage production, and so dry matter of silage was used to calculate GHGI. To produce 1 Mg silage, the CT system emitted 171.1 kg CO2eq ha−1 year−1 (Table 2), while nearly half of that was emitted in NT (87.3 kg CO2eq ha−1 year−1), considering the OC sequestration in 0–20 cm. When considering the OC sequestration in the whole 0–100 cm, instead of emitting, NT mitigated 31.7 kg CO2eq ha−1 year−1 per each megagram of silage and thus can be considered a sustainable system from the point of view of balancing agricultural production with global warming mitigation.

A final point to be addressed, however, is that the GWP reduction of −3.45 Mg CO2eq ha−1 year−1 in NT compared to CT will not remain constant indefinitely but may possibly decrease over time. This is because the soil capacity to store carbon under a certain management condition is finite, so that the accumulation rate decreases over time. However, Six et al. (2004) indicated a reduction of N2O emission in soils under long-term NT management (>10 years), which could counterbalance the OC accumulation decrease. More effort should be placed therefore on elucidating this time-dependency of tillage effects on greenhouse gas balance in agroecosystems.

Conclusions

Our findings show that, relative to CT, NT management in this subtropical Brazilian Ferralsol can reduce the emission of N2O from soil, disagreeing with several previous works showing the contrary. Besides, our findings confirmed that NT is an efficient management system that promotes soil OC sequestration up to 1 m deep. After considering N2O and CH4 fluxes, soil OC changes and equivalent C costs of agronomic practices, we conclude that NT has a lower GWP and GHGI relative to conventionally tilled soil, and contributes to mitigating greenhouse gas emissions.