Do cover crops enhance soil greenhouse gas losses during high emission moments under temperate Central Europe conditions?

Measured parameters comprised cover crop dry matter and C/N ratio, greenhouse gas emissions (CO₂, N₂O), soil mineral N (NO₃^- and NH₄⁺), dissolved organic carbon, soil moisture and soil temperature (Table 1).

Table 1

Overview of measured parameters, methods and sampling dates

Tabelle 1. Überblick über die gemessenen Parameter, die Methoden und Beprobungstermine

Aboveground biomass samples were taken on 16^th November 2012 from one square meter, pre-dried at 50°C and thereafter dried for 24 hours at 105°C to constant dry weight. After weighing, the samples were milled and analyzed for carbon and nitrogen content using a Carlo Elba CNS Elemental analyzer.

Soil greenhouse gas emissions were measured on three dates. The first sampling was done in late autumn after a short frost period (16—18^th November 2012) terminating most plants in order to capture immediate effects from dead leaf and stem addition to soil before winter. The second sampling was done during a period of temporary soil thawing over winter, while the third sampling date coincided with snow melting in March 2013. These sampling periods were considered hot moments (fresh residue addition, freezing-thawing, snow melt with high soil water content) for potential gas emissions. In a sub-set (mustard, legume mixture, fallow; fertilized plots), at the humid (high emission) site, the Pötting measurements were continued until the end of June to obtain comparative values of main crop related emissions (tillage, sowing, N-fertilization). All measurements were done using the closed chamber method (e.g., Rochette et al., 1992) with soil inserted PVC cylinders (10 cm height, 20 cm diameter, insertion depth 3 cm). Cylinders were closed airtight with a lid containing a rubber septum for gas sampling via a syringe (30 ml sampling volume). Gas samples were taken immediately after closing the chamber (initial concentration), after 15 and 45 minutes and injected in evacuated cups. Analysis of trace gas concentration (CO₂, N₂O and CH₄) was done by gas chromatography with a 6890 N GCSystem (Agilent, Santa Clara, CA, USA) connected to an automatic headspace-sample injection system (Dani HSS 86.50 Dani, Cologno Monzese, Italy). Gas flux (µg resp. mg m^-2 h^-1) was then calculated from the change in gas concentration over time.

Soil parameters related to the greenhouse gas formation were measured in the upper soil layer (0—5 cm) at the time of gas sampling. Soil physical variables comprised water content and temperature (Vitel Hydra Probe). Soil chemical parameters were mineral N (NO₃, NH₄; ÖNORM L1091) and dissolved organic carbon (DOC). DOC was determined for selected sampling points only due to the limited analytical capacity. A mixed sample (mixture of three replicate samples; 10 samples per site) of each treatment was taken immediately after the cover crop vegetation period first gas sampling date) expecting most evident differentiation in DOC. A second set of DOC analysis was made for the measurement dates with the highest N₂O emissions at each site, that is, 1^st February 2013 at Lichtenwörth and 7^th March 2013 at Pötting. Again, a mixed sample for each treatment was analyzed. Analysis was done by infrared spectroscopy after the extraction of the samples by deionized water (Brandstätter et al., 1996).

Samples were taken along a line transect with three replicate sampling points in each plot equally spaced at a 2-meter distance. All parameters were sampled at the same position with only small shifts in the longitudinal direction in case of destructive sampling. We first tested spatial correlation (trends) between the data points via autocorrelation analysis. If no spatial correlation was identified, the three samples per plot were considered random replicates and analysis of variance was performed to test for significant treatment effects. In case of spatial trends, further geo-statistical methods would be applied for data analysis.

Generally, the auto-correlograms did not reveal any spatial trend in the measured parameters as exemplified by Figure 2 for N₂O. In case of greenhouse gas emissions, it is well known that emissions are frequently related to soil micro-sites (e.g., Parkin, 1987) with small scale variability. Thus, all data could be treated as spatially independent and tested for treatment effects by analysis of variance. Analysis of variance was performed using a mixed model with measurement dates as repeated measure and an unstructured covariance model for the repeated factor (Piepho et al., 2004).

Figure 2

Another analytical method we applied was joint regression. This technique originates from plant breeding to reveal the distinct response of varieties in different environments (e.g., Annicchiarico, 1997). Environments are characterized via the overall mean of all treatments and are plotted against the single treatments. For this purpose, the two sites were used together thereby defining six “environments” (2 sites × 3 measurement dates) from low to high emission potential. Regression analysis was applied to single soil cover emissions versus the average environmental emission potential.

Finally, regression analysis was performed to reveal any causal relations of greenhouse gas emissions versus soil (water content, temperature, N content, DOC) and plant (C/N ratio, N concentration) parameters.

All statistical analyses were done using SAS 9.1 with PROC MIXED for analysis of variance and PROC REG for regression analysis. For autocorrelations, the SAS time series viewer was used.

Figure 3 shows the cover crop dry matter and C/N ratio at the two sites. There was a significant interaction between site and cover crop species for dry matter, while C/N ratio showed only significant site and species main effects.

Figure 3

Cover crops achieved a relatively high dry matter yield at both sites compared to literature values (e.g., Lütke Entrup, 2000). Mustard and cress-mungo-radish mixture had higher dry matter yield at Lichtenwörth compared to the mixtures of mustard-phacelia and Alexandrian clover-Mungo-phacelia. At Pötting, on the contrary, mustard had the lowest dry matter yield, while all the other cover crops were at a similar level. This was obviously related to the later sowing date of mustard. Dry matter yield of both mustard and cress-mungo-radish mixture differed significantly between sites.

C/N ratio was lowest for mustard and highest for the mustard-phacelia mixture. The legume containing mixture (Alexandrian clover-Mungo-phacelia) was between these two variants. In average, the sub-humid site showed a significantly higher C/N ratio of biomass compared to the humid site.

Figure 4 shows the boxplots of the measured N₂O and CO₂ emissions. It reveals the average emission potential for each date and site as well as the high variability of soil emissions that has been described in several studies (e.g., Partkin, 1987; Rochette et al., 1991).

Figure 4

N₂O emissions were generally low. The median of values was near zero in all cases. The highest N₂O flux at Pötting was registered in March 2013 with a mean value of 6.3 μg m_-2 h^-1 and a peak of 31.4 μg m^-2 h^-1. At Lichtenwörth, the highest emissions were measured in February 2013 with a mean of 5.4 μg m^-2 h^-1 and a maximum of 46.6 μg m^-2 h^-1. Average CO₂ emissions were higher at Pötting (15.1 mg m^-2 h^-1) compared to Lichtenwörth (7.9 mg m^-2 h^-1). At Pötting, the measurement date showed stronger influence on the average CO₂ flux with a clear depression in February 2012. At Lichtenwörth, on the contrary, the variation between the measurement dates was low (± 1.3 mg m^-2 h^-1). Table 2 shows the ANOVA results for the influence of main and interaction effects on the measured greenhouse gas emissions.

Table 2

Results of mixed model analysis of variance (F-statistics and p-value of main effects and interactions) for greenhouse gas emissions. Significant effects (in case of significant interactions, highest order interaction) are marked in bold.

Tabelle 2. Ergebnis der Varianzanalyse (gemischtes Model; F-Werte und p-Werte der Haupteffekte und ihrer Wechselwirkungen) für Treibhausgasemissionen. Signifikante Effekte (bei signifikanten Interaktionen, die Interaktion höchster Ordnung) sind fett gekennzeichnet.

N₂O emissions differed significantly between measurement dates and cover crop species. Interestingly, there was no significant difference between the average N₂O emissions at the two sites. CO₂ emissions showed significant interactions of date with site and cover crop species. Species effects (N₂O main effect, CO₂ date × species interaction) are shown in Figure 5.

Figure 5

The species effect for N₂O emissions clearly indicated that monocropping a brassica species resulted in significantly higher nitrous oxide losses compared to the other treatments, which did not differ significantly among each other. Concerning measurement dates (cf. Figure 4), the first measurement in November (0.9 μg m^-2 h^-1) at the end of cover crop growth was significantly lower as compared to the measurements in February (3.7 μg m^-2 h^-1) and March (4.3 μg m^-2 h^-1) that did not differ among each other.

Species effects for CO₂ (Figure 5b) revealed the highest emissions for the cover cropped plots at the first measurement date. Mixtures of mustard-phacelia and cress-mungo-radish showed a significant increase of CO² losses in March, following the lowest emissions in February. Fallow had no significant differences between dates, while mustard and clover-mungo-phacelia were at a similar level in February and March. On average, all the cover cropped plots had significantly higher CO₂ emissions compared to the fallow in November and February. In March, the CO₂ flux was similar for all treatments except the mustard-phacelia mixture that had a significantly higher value.

The date × site interaction for CO₂ emissions can be inferred from Figure 4. Sites differed significantly in November (Pötting 20.8 mg m^-2 h^-1, Lichtenwörth 7.0 mg m^-2h^-1), while being similar in February (7.6 and 9.2 mg m^-2 h^-1) and March (16.9 and 7.4 mg m^-2 h^-1). At Pötting, the emissions in February were significantly lower compared to November and March; while at Lichtenwörth, the CO₂emissions were at a similar level at all measurement dates. In March, the variability of CO₂ values in Pötting was particularly high.

It can be hypothesized that the treatment differences become more evident when the average emission potential is high. Figure 6 compares the increase of single treatment emissions with increasing environmental emission potential (joint regression). For N₂O, the application of organic N-fertilizer resulted in higher emission over the whole range of emission levels. For CO₂ emissions, no differences in average level and trend could be found between the N-treatments. Comparing the soil cover treatments, there was a clear difference of mustard that leads to substantially higher emissions compared to the other treatments in high emission environments. For the other treatments, confidence bands were overlapping over the whole range of emission levels.

Figure 6

In case of CO₂ emissions, joint regression did not reveal a clear distinction among the treatments. Still, it is evident that fallow has the lowest CO₂ emission and differentiation to the cover cropped soil tends to increase with increasing emission potential.

In order to understand the role of causal factors for greenhouse gases formation, we determined regression relations of CO₂ and N₂O with soil chemical (mineral N, DOC; Figure 7) and soil physical (water content, temperature; Figure 8) parameters. For significant relations, regression equations, R² and significance levels are given in Table 3. Among the soil chemical factors, a slight significant relation was found between CO₂ and soil mineral N at Lichtenwörth. At both sites, the N₂O emissions were not influenced by soil mineral N. DOC and soil water content were found as a significant causal factor for N₂O emissions at both sites. At the semi-arid site Lichtenwörth, increasing soil water content also resulted in higher CO₂ emissions.

Figure 7

Figure 8

Table 3

Regression relations between soil chemical and physical parameters and greenhouse gas emissions

Tabelle 3. Regressionsbeziehungen bodenchemischer und bodenphysikalischer Parameter mit Treibhausgasemissionen

Soil temperature influenced neither CO₂ nor N₂O emissions under the given temperature ranges during the cover crop period.

Finally, we tested if either cover crop biomass or quality of residues (C/N, N) were directly related to any of the two greenhouse gasses. For this purpose, we pooled together the cover crop data from both sites resulting in 8 data pairs (4 cover crops × 2 sites). A significant relation was found between the N content of biomass and N₂O emissions in February (Figure 9; R² = 0.50, p = 0.05). Although significant, the regression has to be interpreted with care due to the clustering of observations at two extremes with the high C/N ratio mustard variants at the upper end and the other species with similar C/N values at the lower end.

Figure 9

Our results demonstrated that even during hot moments of N₂O emissions between late autumn and early spring, average fluxes were low. For better comparison, Figure 10 provides a time series of the subsequent emissions between March and end of June 2013 at the humid (high emission) site for three soil cover treatments (fallow, mustard, non-brassica mixture of clover-mungo-phacelia). This period included N-fertilization (pig slurry + urea; 162 kg N ha^-1), tillage (ploughing) and sowing of the main crop (maize).

Figure 10

Comparing Figure 4a and 10a, it is obvious that the main N₂O losses are related to operations during the main crop, particularly tillage and N-fertilization. Here, an N₂O peak as high as 2316 μg m^-2 h^-1 (N-application + tillage, 87.6% water filled porosity, 23.0°C soil temperature) occurred. A second peak of 874.4 g m^-2 h^-1 was measured on 2^nd June following an intense rainfall with soil at 80.4% of water filled porosity, soil temperature of 11.5°C and mineral N in the topsoil (0—5 cm) of 36.7 kg ha^-1. On the contrary, the highest emission to be considered an immediate spillover effect of cover cropping at this site was 31.5 μg m^-2 h^-1in a mustard plot in March 2013. This is in agreement with Petersen et al. (2011) who also reported N₂O values near zero during the cover crop period, while values increased to 900 μg m^-2 h^-1 until May in the subsequent vegetation period. Several studies reported high N₂O fluxes following N-fertilization (e.g., Arcara et al., 1999) and rainfall events saturating the topsoil (e.g., Jacinthe and Dick, 1997). Dobbie and Smith (2001) showed that a temperature increase between 5 and 12°C strongly stimulated N₂O emissions, while the temperature effect became lower when further increasing soil temperature from 12 to 18°C. Thus, soil temperatures at soil thawing (February) and snow melting (March) were probably still suboptimum for microbial activity catalyzing denitrification. Furthermore, soil mineral N levels in the top soil in February (Pötting: 26.7 kg ha^-1; Lichtenwörth: 26.1 kg ha^-1) and March (Pötting: 34.4 kg ha^-1; Lichtenwörth: 37.8 kg ha^-1) were substantially lower compared to the values registered during the main cropping period (up to 179 kg N ha^-1). From this comparison, it can be concluded that N₂O emissions following cover crops are negligible compared to emissions during the main cropping period mainly due to environmental constraints (low soil temperature and N-content). We also notice here that N₂O emissions from April onwards (cf. Figure 10a) did not show a distinctive trend between the cover cropped and fallowed plots, which is in agreement with other studies (e.g., Parkin and Kaspar, 2006).

Differences in the height of CO₂ flux between the period from late autumn to early spring and the subsequent period until early summer were less pronounced compared to N₂O. This shows the substantially stronger response of N₂O to hot spots and hot moments compared to CO₂ leading to higher spatial and temporal variability (Parkin, 1987; Yao et al., 2010). Following frost termination of cover crops, the highest average CO₂ emissions were measured in November upon fresh residue addition to the soil (21.3 mg m^-2 h^-1), while the highest average value during the main crop time series was 52.7 mg m^-2 h^-1 on 16^th June 2013. A first peak of CO₂ was coincident with the N₂O peak (5^th May 2013) following tillage and N-fertilization. In this case, the emissions from fallow were significantly lower, which indicates an N-fertilizer induced mineralization push of cover crop residues. The second emission peak for CO₂ on 16^th June was registered two weeks later compared to N₂ O, coincident with very high soil temperature (26.9°C) and lower water content (18.4%). Generally, the main driver of CO₂ emissions (R² = 0.63; p < 0.001) was soil temperature.

Cover cropping is expected to reduce nitrous oxide emissions via plant uptake of soil mineral N as well as microbial N-immobilization during residue decomposition. Most studies comparing different cover crop treatments focused on legumes versus non-legumes assuming that the latter contribute more to mitigation of greenhouse gas emissions due to higher depletion of soil NO₃^- and lower residue C/N ratio (e.g., Rosecrance et al., 2000; Baggs et al., 2000; Flessa et al., 2002; Basche and Miguez, 2012; Sanz-Cobena et al., 2014). However, availability of easily decomposable C from cover crop residues can also cause higher emissions (Mitchell et al., 2013). Concerning CO₂, Lal (2011) attributes cover crops a potential to sequester C in agricultural soil. Still most cover crop residues are allocated to the easily degradable soil C pool and are readily mineralized (Puget and Drinkwater, 2001).

Cover crop treatments in our experiment were mainly species mixtures except for mustard as a predominant species in Central Europe. Mixtures combine single species environmental benefits and increase biodiversity within the main crop monocultures (e.g., Creamer et al., 1997; Snapp et al., 2005). Our results clearly showed a distinctive behavior of mustard with a significant increase of N₂O emissions during the high emission moments between late autumn and early spring. In the subsequent main crop vegetation period, there was no evidence of enhanced emissions from mustard plots (cf.Figure 10). In a study of Baggs et al. (2000) in Scotland, on the contrary, no enhanced emissions from mustard were registered. In this study, the N₂O measurements were performed after cover crop incorporation in April. The authors reported comparatively dry soil conditions and attributed N₂O formation to nitrification. Furthermore, mustard residues in their experiment had larger C/N ratio (30.3:1) compared to our study (15.5:1). Petersen et al. (2011) found slightly higher emissions after freezing-thawing events from plots with fodder radish compared to fallow, which they attributed to residue mineralization and higher soil NO₃ contents during warmer periods over winter. Similar to our study, the general emission level before the main crop vegetation period was low.Velthof et al. (2002) also reported peak N₂O losses for brassica species (broccoli, mustard, white cabbage, Brussels sprouts) compared to cereals and sugar beet leaf. They related these differences mainly to distinct C/N ratio and water-soluble N content of residues. Interestingly, sugar beet leaf with similarly low C/N ratio still had low N₂O emissions. Velthof et al. (2002) indicated a C/N ratio < 30 as a threshold for enhanced denitrification losses. In our case, all residues had C/N ratios < 26:1 with lowest value for mustard. The clearly distinctive behavior of mustard suggests that also biomass quality, that is, glucosinolates in brassicaspecies, had further enhanced N₂O losses when all other conditions (low C/N ratio, high water filled pore space) are conducive to nitrous oxide formation. Glucosinolate content is about 1% of fresh weight, with biochemistry and concentration varying between plant organs and with plant age (Fahey et al., 2001). During decomposition of the mustard glucosinolate sinigrin, allyl-isothiocyanate and SO₄^2- are formed. Production of glucose, subject to fermentation under anaerobic conditions, further lowers the soil oxygen content. In this process, H₂S is formed, decreasing the soil redox potential and acting as a bactericide, thereby providing additional substrate for anaerobic decomposition (Fahey et al., 2001; Mentler, 2015, personal communication). Thus, several aspects in this process (oxygen consumption, lowering of rH, and addition of easily decomposable substrate) drives denitrification beyond the level found for other residues under similar environmental conditions. Mixtures including brassica species however did not show enhanced emissions. This is probably explained by either their lower seeding density within mixtures or differences in plant senescence.

CO₂ emissions were readily increased by cover crops upon frost killing in late autumn and thereafter decreased until spring. Also, Magid et al. (2004) demonstrated quick release of N and C from the cover crop residues after incorporation even at low temperatures (3°C). Taking the bare soil treatment as reference for CO₂ respiration, CO₂ release from the cover cropped plots was 313.7%, 103.1% and 28.7% above fallow in November, February and March respectively. Mixtures of mustard-phacelia and cress-mungo-radish had a significant depression of CO₂ fluxes in February before the onset of spring mineralization. Their wider C/N-ratio probably slowed down the decomposition process to a larger extent under lower soil temperatures and microbial activity in February. We notice here that CO₂ emissions from cover cropped plots do not allow inferring on potential soil carbon storage from the added plant material. Lal et al. (1998) estimated carbon sequestration from cover crops between 100–300 kg C ha^-1 yr^-1 depending on residue quality (Puget and Drinkwater, 2001) and interaction with soil aggregation (Kong et al., 2005). Concerning fallow, the measured CO₂ emissions in any case are losses from the soil organic carbon pool, which are not counterbalanced by any previous inputs.

The conflicting results on cover crop induced N₂O emissions are explained by their different impact on either of the causal factors for nitrous oxide formation: in case of dominant reduction of soil NO₃-(uptake, immobilization) and/or drying of the top soil by plant water extraction, a reduction of N₂O fluxes compared to fallow can be expected. On the contrary, whenever the dominant cover crop effects are addition of easily decomposable carbon, quick N release from fresh residues and/or higher top soil moisture under surface mulch residues, a temporal increase of losses can be caused. In this context, the inter-annual variability in weather conditions adds complexity to the cover crop effect. The overall effect will be reduced in years with adverse autumn conditions, leading to lower cover crop biomass compared to this study. Warmer spring temperatures than in this experiment may enhance an early cover crop turnover. In sub-humid Eastern Austria, the influence of spring-drought with subsequent wetting of the soil can constitute a special case for residue turnover to be studied.

As expected, our results showed a positive effect of soil water content on N₂O emissions. A strong increase of losses, exceeding the background emissions of 5 μg m² h^-1, was observed at > 60%water filled porosity. Soil mineral N did not show a significant relation to N₂O fluxes. At the time of highest N₂O emissions (Lichtenwörth February; Pötting March), the soil N content was significantly lower in bare soil (11.4 kg N ha^-1; 22.3 kg N ha^-1) compared to the cover cropped plots (24.9 kg N ha^-1; 37.4 kg N ha^-1). Still, this does not explain the distinctive behavior of mustard, which did not differ from other cover crops in soil mineral N. The highest R² among the tested causal factors for N₂O was found for DOC. Also, Mitchell et al. (2013) showed that mineralizable C input from cover crop residue can control N₂O emissions in case of sufficient availability of NO₃^-. Foi the distinctive behavior of mustard, furthermore, the higher biomass N-content of residues (cf. Figure 9) and their chemical composition has to be taken into account.

CO₂ emissions were significantly related to soil mineral N at Pötting and soil water content at Lichtenwörth. CO₂ and N release occur within the same process of organic matter mineralization. Therefore, it is difficult to prove if the relation between soil mineral N and CO₂ is causal (enhanced microbial activity via otherwise limiting soil N), or coincidental (two parameters expressing the same process). The relation to soil water content at Lichtenwörth indicates that the average saturation level at this site was still favorable for aerobic processes. The equally positive relation to N₂O points to the role of anaerobic microsites in a soil with otherwise sufficient air-filled porosity at larger representative elementary volume.