2.1 Measurement site

The CH-Oe2 field site is located in Oensingen, in the canton of Solothurn, Switzerland (47^∘17^′11.1${}^{\prime \prime }$ N, 7^∘44^′01.5${}^{\prime \prime }$ E; 452 m a.s.l.). The crop field has an extent of 1.55 ha with a Fluvisol with 42 % clay, 33 % silt and 25 % sand (Alaoui and Goetz, 2008). The average air temperature (T_A) at the site is 9.8 ^∘C, and the average annual precipitation (Prec) is 1155 mm (Fig. 1, period 2004 to 2016; the diagram was produced in R with the diagwl function of the climatol package). The field has been managed under the regulations of PEP since the late 1990s, featuring a 3-year crop rotation (Table 1). The main crop has been winter wheat, which is usually planted every third year followed by winter barley. The third crop in the rotation was either potato, winter rapeseed or peas. Only between autumn 2006 and autumn 2010 was wheat planted every second year. Before summer crops (potato or peas) were sown, Phacelia only (2009 and 2015) or a mixture of summer oat, Phacelia and Alexandrine clover (2005) was planted. After every rapeseed harvest, a voluntary regrowth of the rapeseed was allowed and the newly grown rapeseed plants were then mulched and incorporated into the soil later in the autumn before wheat was sown. Before the management under PEP started in the late 1990s, the field had an 8-year arable–ley rotation, including 3 years of perennial grass–clover mixture.

Figure 1Climate diagram after Walter and Lieth (1960) for the time period of 2004 to 2016. The monthly average air temperature (T_A) and the monthly total precipitation (Prec) are shown in red and blue, respectively. Average (Avg), average minimum (Min) and average maximum (Max) annual T_A and average annual total (Sum) precipitation are listed at the top of the figure. Note that the scale of the right axis changes above 100 mm.

Download

Management information including dates and type of tillage, sowing dates and seed weights, fertilisation dates and amounts, dates of pesticide applications, and harvest dates and yield (grain and straw) was regularly provided by the farmer (Table 1). Management timing and field conditions were confirmed with webcam images of the field (since 20 May 2005 taken at 10:30, 12:30 and 14:30 CET (UTC + 1 h) and since 1 March 2015 at 09:30, 12:30 and 14:30 CET). In the case of wheat, barley and rapeseed, the moisture content of the harvested grains was reported by the farmer. Cover crops were not harvested and thus ploughed into the soil. No harvest was conducted for the potatoes in 2006; due to a hail storm on 5 July 2006 the potatoes were of very poor quality and therefore left in the ground and later ploughed under. Between 2004 and 2016, solid manure was applied on three occasions (always at the end of the cover crop seasons), whereas liquid manure was applied on five occasions (at the end of wheat, barley and rapeseed seasons). A crop season is defined here as the period between sowing of a crop and sowing of the following crop. Mineral fertilisers were applied during all crop seasons, except for cover crops and the 2016 pea season. Herbicides were applied during all crop seasons, except for cover crops and the potato season. Fungicides were only used in spring 2004 (wheat), 2005, 2012 and 2015 (all barley), and insecticides were applied during the rapeseed season in 2007–2008 and during the 2010 pea season. Grubbing (shallow secondary tillage) was conducted almost every year, ploughing approximately every third year (typical depth of 30 cm), and harrowing was conducted every year since 2010. In 2005, the cover crop was mulched on 9 November. In 2010 and 2016, the cover crop was incorporated into the ground shortly before the next crop was sown without any preceding mulching.

Table 1Management information for all 16 crop seasons (defined as sowing of the current crop to sowing of the following crop) between 2003 and 2016 with crop type, sowing and harvest dates, and yield (G: grain, P: peas, S: straw). Moisture content of the harvested biomass (MC) is given in parentheses. If manure was applied during the crop season, the date, manure type and amount are given as well.

Download Print Version | Download XLSX

2.2 Turbulent fluxes

Since the end of December 2003, eddy covariance (EC) measurements have been made at the site. The eddy covariance measurements consist of three-dimensional wind speed and air temperature measurements with an ultrasonic anemometer (R3-50; Gill Instruments Ltd., Lymington, hampshire, UK) as well as CO₂ and water vapour measurements with an open-path infrared gas analyser (LI-7500; LI-COR, Lincoln, NB, USA) and were recorded at 20 Hz.

The eddy covariance data were processed and quality controlled with the software EddyPro (version 6.2.0, LI-COR). Thereby, 30 min averaged fluxes were calculated and the following corrections and filters were applied: high-frequency despiking and a drop-out test (on the raw data) following Vickers and Mahrt (1997), angle-of-attack correction (Nakai et al., 2006), double rotation (Wilczak et al., 2001), lag time compensation via covariance maximisation using a default lag time if a maximum was not attained within a plausible window, density fluctuation correction (Webb et al., 1980), high-pass filter (Horst, 1997), low-pass filter (Moncrieff et al., 2004), and a steady-state test and test for well-developed turbulence conditions (on the processed fluxes). Fluxes were rejected from further analyses when they were outside a physically plausible range (±50 µmol m⁻² s⁻¹). From November 2015 to May 2016, an angle-of-attack filter was also applied, which discarded half-hourly fluxes if the angle of attack was outside the range of −10 to 30^∘ for more than 10 % of the half hours. This additional quality criterion was applied to filter out time periods of an occasional malfunctioning of an anemometer transducer. During times of repair of the R3-50 ultrasonic anemometer, the ultrasonic anemometer was replaced by a model HR-100 ultrasonic anemometer (Gill). CO₂ storage in the air layer below the flux measurement height was calculated according to Aubinet et al. (2001) within EddyPro.

NEE was calculated by adding the half-hourly CO₂ flux and CO₂ storage and subsequently despiked by iteratively removing outliers outside the valid range defined as the mean ± 3 times its standard deviation (SD) (Rogiers et al., 2004) based on a 30-day moving window. NEE was then gap filled and partitioned into gross primary production (GPP) and ecosystem respiration (R_eco) based on Reichstein et al. (2005) using the R software REddyProc by the MPI Jena (Reichstein et al., 2017). Gap filling was done after applying an automatically determined u_* filter (with a threshold ranging between 0.01 and 0.13 m s⁻¹; changed for each crop season). The u_* threshold was automatically determined for each bare soil period and growing period separately within REddyProc by determining the saturation of NEE with u_*. In total, NEE had to be gap filled for 46 % of the half hours.

For the beginning of the first wheat season (October to December 2003), the measurement station was not established yet and therefore no flux data were available. From November 2006 until February 2007, no reliable NEE measurements were available due to a sonic anemometer malfunctioning. Therefore, NEE was estimated for these two time periods in 2003 and 2006–2007 by averaging the gap-filled NEE of the corresponding days of the wheat seasons in 2008, 2010 and 2013 (on a daily basis).

3.1 Carbon budgets over 13 years

Figure 2Daily cumulative net biome production (NBP_cum) between 16 October 2003 and 11 October 2016. The black dashed line connects NBP_cum at the end of each crop season. During time periods with a grey background, the field was bare (from harvest of a crop to sowing of the next crop or from first ploughing after sowing of a crop to sowing of the next crop if the first crop was not harvested).

Download

For all crops, the season (defined from sowing of one crop to the sowing of the following crop) started with a net release of CO₂ until the crop had emerged and became established, after which GPP began to exceed R_eco (Fig. 2). A few weeks before harvesting when senescence started, R_eco exceeded GPP again, resulting in a net CO₂ release. At the point of harvest, C was exported from the ecosystem, which can be seen in most years as a sharp increase in NBP_cum. Organic fertilisation with solid or liquid manure and sowing were C imports into the ecosystem. However, only solid manure applications were large enough C imports to be seen as a sharp decrease in NBP_cum. While the field was bare, it was almost only respiring and therefore NBP_cum increased during these periods. The voluntary regrowth after the harvest of rapeseed (2008 and 2013) resulted in an approximately 2-month-long period of uptake in the autumn of the same years.

For peas, the period of net C uptake was quite short (less than 1 month in contrast to 3 to 4 months for the other crops), which is due to their short growing period as they were peas for canning and were therefore harvested relatively early. The period of net C uptake is barely visible during the pea season in 2016 because the field was flooded due to extensive rain. Cover crops were only growing in the autumn, resulting in a relatively weak CO₂ uptake followed by a relatively long period of net CO₂ loss during winter season.

NBP_cum at the end of each season mostly increased over time. Only during the potato season in 2006, without harvest due to the hail damage and during the crop rotation cycle (wheat, barley and rapeseed) between 2010 and 2013, did the NBP_cum stay almost constant. NBP_cum of the first crop rotation cycle (wheat, barley, cover crop, potato; 2003–2006) was 236 g C m⁻². Between 2006 and 2009, wheat was repeated every second year. During the first 2-year period (wheat, rapeseed), the field was a net source of 302 g C m⁻² and during the second 2-year period (wheat, cover crop, peas) a net source of 396 g C m⁻². During the next full crop rotation cycle (wheat, barley, rapeseed; 2010–2013), the field was close to C neutral (NBP = −22 g C m⁻²), while it was a net source of 748 g C m⁻² during the last crop rotation (wheat, barley, cover crop, peas; 2014–2016). The cumulative net biome production (NBP_cum) for the 16 crop seasons between autumn 2003 and autumn 2016 shows that there was a net C loss of 1674 g C m⁻² over the 13 years of study (Table B2). The field lost on average 129±50 g C m⁻² of C per year (unless stated otherwise, we report mean ± standard error except for soil C and N values, for which mean ± SD is given).

Soil C densities (ρ_C) in the top 12 cm of the field were 0.0355±0.0042 g cm⁻³ (mean ± SD) in 2004 and decreased significantly (p<0.0001) on average by 18.0 % to 0.0291±0.0031 g cm⁻³ until spring 2017 (average over the top 15 cm and over all measurement days in 2017). The bulk density of the same layer increased insignificantly (p=0.25) from 1.16±0.08 g cm⁻³ in 2004 to 1.21±0.14 g cm⁻³ in 2017. The soil C stock decreased significantly (p<0.0001) on average by 775 g C m⁻² in the top 12 cm from 4263±507 g C m⁻² to 3488±374 g C m⁻². At the same time, N stock changes were not significant over the 13 years (372±53 in 2004, 382±44 g N m⁻² in 2017, p=0.19). There were no measurements from deeper soil layers available for 2004. However, measurements in 2017 show that C densities did not vary significantly (adjusted p=0.959) in the top 30 cm (Fig. 3). Ploughing was also done in most years to a depth of 30 cm. If we therefore assume that C stocks changed equally over a depth of 30 cm between 2004 and 2017, the soil C stock decreased in the top 30 cm layer on average by 1980 g C m⁻². This corresponds to an annual average loss of 152 g C m⁻².

The application of slurry caused such a small C input that it was not only invisible in NBP_cum (Fig. 2) but was also not detectable in the soil. Soil C density measurements before and after the application of the slurry in 2017 did not reveal any significant (adjusted p>0.05) changes (Fig. C1). The slurry added only 25.4 g C m⁻² and 4.7 g N m⁻² to the soil. When comparing these numbers to the C and N stock of the top 30 cm of the soil, it can be seen that the C and N input is negligible.

Figure 3Average soil carbon density (ρ_c) for different depth layers on 13 October 2004 and 31 May 2017. Error bars show standard errors.

Download

The field site was clearly a C source, which was also confirmed by the changes in soil C stocks measured at the beginning and at the end of the measurement period. Depending on the measurement method, the field lost 15.7±4.0 % (based on NBP_cum and soil C stock of top 30 cm in 2004) to 18.0±5.3 % (based on soil C stocks in 2004 and 2017, uncertainty is based on standard errors) of C over the 13 years. The differences between the C budget determined by calculating NBP and by measuring C stocks in the soil were remarkably small given that these results are based on two completely independent measurements. The loss strength, however, was likely influenced by the arable–ley rotation, which was used at the field until the late 1990s and which is expected to reach a higher soil C stock than the crop rotation that was used afterwards.

Ceschia et al. (2010) studied the annual NBP (138±239 g C m⁻² yr⁻¹; they call it net ecosystem C budget) and the annual changes in soil C stocks of the top 30 cm (2.4±4.7 % year⁻¹) of European croplands (averaging over 17 croplands and 41 site years ± SD deviation, between 1 and 5 consecutive years per site). In contrast to our results, their findings were not significantly different from a C neutral budget. However, our results were within the range found by Ceschia et al. (2010). Kutsch et al. (2010) determined an average annual NBP of 95±87 g C m⁻² yr⁻¹ for five crop rotation sites and two monoculture sites. There are a number of other studies on European crop fields with crop rotations that found similar or slightly higher annual losses to what we found in this study (Buysse et al., 2017; Prescher et al., 2010). A modelling approach based on soil stock measurements for European croplands also resulted in comparable average annual C losses of approximately 90±50 g C m⁻² (Janssens et al., 2003). On the other hand, research using a process-based model and soil C inventories (Ciais et al., 2010) and a study combining ecosystem-scale measurements with atmospheric greenhouse gas measurements and an inversion model (Schulze et al., 2009) found an average annual source of 8.3±13 to 13±33 g C m⁻² yr⁻¹ and 10±9 g C m⁻² yr⁻¹, respectively, for croplands. In our study, the management under the regulations of PEP did not result in a neutral C budget or C sink and also not in a significantly smaller average annual loss compared to other European croplands. However, soil N stock measurements showed that the neutral N budget, as required by PEP, was approximately reached.

An uncertainty estimate of NBP calculated with Eq. (1) can be found in Appendix A. In total, the uncertainty adds up to a maximum uncertainty of approximately ±25 % of NBP_cum. Buysse et al. (2017) listed in detail the uncertainties involved in the different NBP terms in their study, which would add up to a maximum uncertainty of 220 g C m⁻² over the 12 years of their study (at NBP = 990 g C m⁻²), corresponding to an uncertainty of 22 %.

3.2 Crop-specific budgets

Figure 4Crop-season-specific cumulative net biome production (NBP_cum) and the main contributing terms: cumulative net ecosystem exchange (NEE_cum), C export by harvest (E_harvest) and C import by fertiliser (I_fertiliser). Each symbol stands for one crop season. Not shown is the C import by sowing (I_sowing), which is negligibly small except for potatoes. It is, however, included in the calculation of NBP_cum. Please note that cover crops were only grown during autumn and winter.

Download

Table 2Average and standard error of cumulative net ecosystem exchange (NEE_cum), C export through harvest (E_harvest) and cumulative net biome production (NBP_cum) in g C m⁻² season⁻² of the five crop types with more than one season. The number (n) of seasons for each crop type is given in parentheses. Please note that cover crops were only grown during autumn and winter.

Download Print Version | Download XLSX

Wheat and barley showed the largest net C uptake from the atmosphere over the crop season and also had the largest C export through harvest (Fig. 4 and Table 2). They were followed by rapeseed, which also had less C exported through harvest. Peas assimilated less C from the atmosphere than the ecosystem released at the same time, and very little was exported from the field during harvest. During winter seasons with cover crops, more CO₂ was also lost to the atmosphere than was taken up by the ecosystem. For all crops, I_sowing was very small to negligible (<15 g C m⁻², Table B1). The application of slurry also resulted in rather small imports of 16 to 25 g C m⁻², whereas solid manure imported 123 to 229 g C m⁻² (Tables 1 and B1).

Taking into account NEE_cum, E_harvest, I_fertiliser and I_sowing, NBP_cum of most crop seasons was positive. Pea seasons showed a substantially larger overall C loss than the other crops. Most other crop seasons ranged between close to zero and 160 g C m⁻². Since the potatoes were not harvested and did not receive a fertiliser application, this resulted in the only season that had an almost neutral C budget. For one season (2007–2008) rapeseed was a weak C source and a very weak C sink in the other season (2012–2013). Cover crops were only on the field from the late autumn until the early spring, when less light was available for growth and conditions were generally colder compared to those of the other crops. Their relatively large C loss to the atmosphere was thus a result of the winter growing season, not of the crop type, and was strongly compensated for by the application of solid manure. Solid manure was always applied at the end of the cover crop seasons. This was done to compensate for the expected C losses during the following pea season, which are often referred to by farmers as consumers of soil organic C. Therefore, it could be argued that the application of solid manure should be attributed to the following pea season instead of the cover crop season. With the crop season defined as the time range between the first ploughing after the harvest of the previous crop to the first ploughing after the harvest of the current crop, all crops (except potatoes and one barley season) would be in a more similar range (peas: 124 g C m⁻² in 2010 and 181 g C m⁻² in 2016; Fig. D1). The attribution of the manure application to the pea season is also discussed in Gilmanov et al. (2014). The reduction of the net C loss during the pea season due to the solid manure application shows that the application of solid manure before the growth of peas is useful to compensate for the loss of C during these seasons, although it can only partly offset the C losses.

Our results for winter wheat and winter barley are comparable to what was found in Europe for these crop types (Ceschia et al., 2010). There are very few studies looking at rapeseed or peas. For winter rapeseed (in Germany) and peas (in France), Ceschia et al. (2010) reported values of NEE = −306 and 278 g C m⁻², E_harvest=560 and 98 g C m⁻², and NBP = −2 and 375 g C m⁻², respectively, including only one season per crop type. In our study rapeseed assimilated less C in both seasons and less C was also exported with the harvest; however, NBP was again comparable. For peas, NEE was comparable to NBP because the export with the harvest was much smaller than for all other harvested crops. This could be related to the fact that the peas cultivated at CH-Oe2 were peas for canning, which are harvested when they are still relatively small. We are not aware of a study having investigated the C budget of potatoes that does not use data from our own site. The results of our potato season should not be considered representative for regular potato seasons due to the hail damage, which had major impacts on the management and the growth of the plants and resulted in no harvest. In our study, applying solid manure to the cropland was found to import substantial amounts of C to the ecosystem, while the import through liquid manure was very small. For a variety of European croplands, Ceschia et al. (2010) found that organic fertilisation tended to lower the C budget even though respiratory losses can slightly increase (less than 10 %) in the first month after the application of solid manure (Eugster et al., 2010).

3.3 The effect of cover crops

During the winter seasons with cover crops, there was always a net C loss. This loss, however, could have been larger not having a crop on the field at all. Having a crop on the field allows for C uptake through photosynthesis; however, autotrophic respiration (by the plants) and heterotrophic respiration (by providing more soil C matter to decompose) will also be enhanced. Depending on whether photosynthesis or respiration is enhanced more, a cover crop may be beneficial in the context of the C budget. In order to assess the benefit of having a cover crop, the CO₂ exchange of the field without a crop (i.e. bare field) was modelled with the SPA-Crop model. All other terms of NBP were kept constant since the cover crop was not harvested. SPA-Crop captures the CO₂ exchange from the harvest of the previous crop until the start of the cover crop growth quite well (Fig. 5). In contrast to tropical regions (Powlson et al., 2016), where climate during cover crop seasons is not a limiting factor, the field experienced a net loss of C during the cover crop seasons due to the less favourable climate (colder and less light) on the Swiss Plateau in autumn. Nevertheless, in all three seasons, the field with a cover crop is overall a smaller net C source than the bare field, even though the NEE_cum difference covers a large range of 11 to 163 g C m⁻². The cover crop seems to be clearly beneficial (GPP increases larger than R_eco increases) in reducing C losses during fallow periods. Furthermore, substantial amounts of C are introduced into the soil by incorporating the biomass at the end of the season when the field is prepared for the next crop. Ceschia et al. (2010) report that the voluntary regrowth of seeds and weeds after the harvesting of winter wheat at Avignon in the season 2005–2006 also reduced the C losses. In a recent review by Chenu et al. (2018) the use of cover crops was discussed. Similar to our findings they conclude based on a number of different studies that the use of cover crops is beneficial for soils because it results in higher soil organic C stocks compared to their absence. The result on cover crops at CH-Oe2 shows that the regulations of PEP requiring a cover crop during fallow periods improved the C budget of the field.

Figure 5Daily average NEE of the three cover crop seasons (a) 2005–2006, (b) 2009–2010 and (c) 2015–2016, displaying measured data with cover crop and modelled data with a bare field. NEE was measured with an EC system, while modelled NEE was simulated with the model SPA-Crop. Vertical lines indicate sowing, tillage and mulching dates. Numbers in the top right corner of each subfigure are the cumulative NEEs of the field with cover crop in black and bare field in brown.

Download

3.4 Solid manure can at least partly compensate for the C losses

The more frequent use of solid manure could compensate at least partly for the C losses of the crop field and decrease or prevent the loss of soil fertility. Assuming the same average C loss rate for the future but without any organic fertiliser application (also no slurry), the average annual loss would be 174 g C m⁻². The average C concentration in solid manure at CH-Oe2 was 440 g kg⁻¹ dry mass (Table E1). Based on these numbers an annual manure application of approximately 15.8 t ha⁻¹ would compensate for the C losses without any further slurry applications if we assume no increase in R_eco. The regular application of solid manure could also reduce the amount of mineral fertilisers applied to the field because substantial amounts of N, phosphorus pentoxide (P₂O₅), potassium oxide (K₂O) and magnesium (Mg) would be supplied by the solid manure (for N approximately half and in all other cases close to the needs as given by the fertilisation plan of the Landwirtschaftliche Beratungszentrale Lindau LBL, 2005, averaged over all crop seasons). The application of compost instead of solid manure should be considered if not enough solid manure is produced by the farm. We estimate that 28.6 t ha⁻¹ of compost would be needed to compensate for the average annual C losses assuming that the net fluxes of compost are similar to manure. Also, in the case of compost, large fractions of the N, P₂O₅ and K₂O needs would be met. On the other hand, Mg would be overfertilised. This is, however, only a rough estimate because the composition of compost and manure can vary substantially. Furthermore, the manure amount needed to compensate for C losses should be seen as a lower limit because several studies in Switzerland have shown that the C loss reduction can be much less than the C input through manure (Leifeld et al., 2009; Maltas et al., 2018; Oberholzer et al., 2014), which also likely applies to compost. Including ley in the crop rotation could also be considered to compensate for C losses. According to Maltas et al. (2018) green manure or cereal straw application can also be effective measures to prevent or reduce soil degradation, while solid manure has the highest C loss reduction efficiency (compost was not included in the study).

Switzerland's nationally determined contribution (NDC) to the reduction in greenhouse gas emissions lists zero emissions from non-forest lands like croplands (NDC, 2017). Therefore, the C losses should be reduced from a climate change point of view. The use of organic fertilisers could help in coming closer to meeting the goal. In the case of CH-Oe2, the grains, peas and potatoes were not used to feed animals on the same farm. However, straw produced on the field at a rate of 78 g C m⁻² yr⁻¹ (1013 g C m⁻² in total during the 13 years of measurements) is used on the farm. If this straw had been added back to the field (either directly or included in solid manure), it could have compensated for a fraction of the C losses over the 13 years. Ammann et al. (2007) studied the C exchange of the neighbouring grassland managed by the same farm. Intensive management of the grassland fertilised with liquid manure (mixture of cow dung and urine) from the same farm resulted in a significant uptake of C. Because the grassland was a C sink it could have been considered to apply the manure to CH-Oe2 instead to counteract the higher C loss of the arable field. Therefore, we assume that there is a potential to decrease the field's C losses substantially by increasing the application of the farm's own solid manure to the field. In order to determine if the application of manure would improve the greenhouse gas budget of the cropland as listed by Switzerland's NDC, it would require a complete life cycle assessment which goes beyond the scope of this study.

Table 3Nutrient requirements and input: average annual need based on the fertilisation plan (Landwirtschaftliche Beratungszentrale Lindau LBL, 2005), the input of the same nutrients with the annual application of 15.8 t ha⁻¹ yr⁻¹ of solid manure (based on the average concentrations of solid manure given in Table E1) and the annual application of 28.6 t ha⁻¹ yr⁻¹ of compost (Landwirtschaftliche Beratungszentrale Lindau LBL, 2005), corresponding to an approximate C_org input of 174 g m⁻² yr⁻¹. For N an efficiency of 60 % was assumed for manure and for compost as required by the regulations of PEP in the case of farmyard manure (Amaudruz et al., 2014).

Download Print Version | Download XLSX