Greenhouse gas and volatile organic compound emissions of additive-treated whole-plant maize silage: part A—anaerobic fermentation period

Abstract

Background

Silage emits climate- and environment-relevant gases during fermentation and feed-out periods. This trial aimed to determine the unknown carbon dioxide (CO₂), methane, nitrous oxide, ethanol, and ethyl acetate emissions of constant maize silage material over both periods. The results will be published in two consecutive articles (Part A: anaerobic fermentation period, Part B: aerobic storage period).

Methods

The untreated control (CON) was compared with the chemical additive treatment (CHE; 0.5 g sodium benzoate and 0.3 g potassium sorbate per kg fresh matter) and the biological additive treatment (BIO; 10⁸ colony-forming units (CFU) Lentilactobacillus buchneri and 10⁷ CFU Lactiplantibacillus plantarum per kg fresh matter). Barrel silos (n = 4) were connected to gas bags to quantify gas formation during anaerobic fermentation (30 or 135 ensiling days). Glass jar silos (n = 12) were used for laboratory silage analysis.

Results

CHE produced significantly (p < 0.05) less gas (6.7 ± 0.3 L per kg dry matter ensiled material (kg_DM) until ensiling day 14.0 ± 0.0) and ethanol (8.6 ± 1.5 mg kg_DM^–1) than CON did (8.5 ± 0.2 L kg_DM^–1 until ensiling day 19.5 ± 6.4; 12.2 ± 1.5 (mg ethanol) kg_DM^–1). BIO indicates prolonged gas formation (9.1 ± 0.9 L kg_DM^–1 until ensiling day 61.3 ± 51.9; 12.0 ± 2.1 mg kg_DM^–1). CO₂ is the main component of the gas formed. All treatments formed methane and nitrous oxide in small quantities. CON emitted significantly more CO₂eq emissions than BIO and less than CHE (p < 0.05). Additives had no effect on ethyl acetate gas emissions. For BIO, ethanol concentrations in the material (r_S = 0.609, p < 0.05) and gas quantities (r_S = 0.691, p < 0.05) correlate with ethyl acetate gas quantities. All the treatments exhibited decreasing gas and CO₂ quantities, and the dry matter mass increased between ensiling days 14 and 30 (− 0.810 ≤ r_S ≤ 0.442; p < 0.05 to p = 0.20).

Conclusion

Silage generates climate- and environmental-relevant gases during fermentation and silage additives affect this pattern. Gas formation exceeds the fixation potential, and the carbon footprint of silage fermentation is negative.

Graphical Abstract

Introduction

Silage is an essential global feedstuff, with the opportunity to conserve one-time crop yields. The supply of high-quality feed is crucial to feed ruminants resource-efficiently throughout the year. The same applies to biogas plants. The ensiling process includes, among others, the anaerobic main fermentation and aerobic feed-out phase [1]. One of the main objectives is to minimise dry matter (DM), energy, and quality losses to maintain the resource cycle and the nutritional value of harvested plant material in the best possible manner. DM losses in silage are generally accompanied by gaseous emissions [2,3,4,5] or effluent losses.

Losses are partially unavoidable for high-quality silage fermentation, e.g. heterofermentative metabolism of lactic acid bacteria (LAB) [6, 7], but include avoidable losses, too. The latter consists of exceeding activity of undesirable microbes, such as enterobacteria, yeasts, or moulds during anaerobic fermentation or aerobic storage. Several authors provided overviews [7,8,9] concerning losses and management effects, e.g. silage additive (SA) use, packing density, or aerobic stability (ASTA). Köhler et al. [10] reported losses of − 5 to − 15% for farm-scale maize silos during anaerobic fermentation. According to Wilkinson [9], the expected total loss of maize silage production was − 20.6% from field to trough. SA, such as LAB inoculants and organic acids or their salts, can influence microbial metabolism and losses in various ways. This article focuses on the specific group of SA that achieves a prolongation of ASTA through increased acetic acid (AA) production or antimicrobial properties [1, 6, 7, 11, 12].

Silage production leads to the emission of climate-relevant greenhouse gases (GHG) [2, 4, 5, 13, 14] with various global warming potentials (GWP) and other climate- and environment-relevant gases, e.g. volatile organic compounds (VOC), which are precursors of ground-level ozone formation [15,16,17]. The inoculation of ensiling material with heterofermentative LAB can increase DM losses [6] and gas formation during the anaerobic fermentation period [5] due to AA and carbon dioxide (CO₂) production [18]. Chemical additives can decrease DM losses [7] and the formation of VOC during anaerobic fermentation [19, 20]. Both additives can improve ASTA and, therefore, reduce respiratory emissions during the feed-out phase [6, 7]. Ethanol can be used as an indicator of VOC formation patterns, since alcohols contribute to the majority of VOC in silage [15, 16, 21]. Ethyl acetate (EA) is reported to have antibacterial and antifungal properties and may affect microbial metabolism [22, 23]. Furthermore, the high vapour pressure of EA could lead to increased volatilisation into the gaseous phase [20].

According to Schmidt et al. [13], most of the gas produced during anaerobic fermentation is CO₂. The same applies to the aerobic feed-out phase based on respiration pathways. CO₂ can be considered climate-neutral. In the carbon (C) cycle of agricultural resources, photosynthesis converts CO₂ to biomass, which will be converted back to CO₂ in later stages. While photosynthesis is considered a CO₂ sink, the other stages are CO₂ sources. If biomass is degraded to CO₂ during silage storage, those energy-rich C-molecules are unavailable in the later stages of the cycle. Therefore, DM losses during silage storage affect the C retention efficiency of the resource cycle. As far as the authors are aware, quantification of GHG and VOC emissions from anaerobic fermentation to feed-out of constant silage material is lacking in the scientific literature. Former trials examined either emissions of ensiling material during anaerobic fermentation or of ensiled material during the feed-out period. Total quantities could be used to compare the emissions during silage storage with those during the other stages of the cycle or with alternative methods of conserving animal feed. Moreover, a comparison between the carbon footprint (CF) of SA and their effect on silage emissions can be made. Therefore, the CO₂ emissions from silage storage are not classified as climate-relevant emissions but rather as emissions of climate-relevant gases.

Schmidt et al. [13] estimated that silage emissions during anaerobic fermentation are lower than those during animal husbandry. However, others demanded more research to assess the relevance of all silage production stages for VOC—and the same applies to GHG—emissions [15]. Henriksson et al. [24] stated: ‘In-depth knowledge of GHG emissions associated with silage production is, therefore, crucial in mitigating GHG emissions on farm level’. This applies in modern times, to assess the CF of various agricultural food products [25]. However, some studies have reported the opposite behaviour, i.e. a gas fixation and DM increase during anaerobic fermentation, based on unclear biological or chemophysical processes [14, 26].

Previous silage emission research has shown that the activity of microorganisms leads to ongoing gas production and an outwards-directed gas flow from silos [2, 27]. Brazilian working groups assessed gas production by measuring positive pressure in silos [4] or collecting gases in a beaker [28, 29]. Knicky et al. [30] used gas bags to collect silage emissions. Most recently, Krueger et al. [14] published a calculation model to estimate CO₂ emissions during the fermentation process of ensiled maize. An American working group established a model to calculate the emission quantities of ethanol during the ensiling process [31]. However, no working group has conducted trials to verify the calculated data. Shan et al. [32] investigated the ethanol gas emissions of silage-related LAB in broth. Earlier research revealed several measurement and procedural inaccuracies [33, 34]. In one of the most recent studies, gas was sampled in the silo headspace regularly within the first 49 days of anaerobic fermentation [2]. Furthermore, gaseous substances formed during the ensiling process are emitted once the silo is opened [21, 35,36,37].

The quantity of emissions generated during the ensiling period is affected by microbiological activity [5]. In addition, factors, such as plant species [2, 4], the wilting period and DM concentration of the harvested material [2, 38], a delayed sealing time [20, 39, 40], and the use of SA [5, 20, 29], influence metabolite formation. SA are usually considered to ensure high silage qualities and improved ASTA [7, 8, 12]. However, the effect of heterofermentative LAB inoculation depends on the length of the fermentation process [5, 6].

Moreover, recent research has examined the negative pressures within silos [2, 41] and the ability of silage to absorb gas. An overview is given by Schmidt and Vigne [42]. Maize silage was observed to absorb supplied CO₂ and nitrogen (N₂) gas [43]. Empirical data from the Brazilian working group [26, 44, 45] were strengthened by a model for CO₂ absorption and DM build-up [14]. This model has yet to be validated. Schmidt and Vigne [42] expressed the optimistic question: ‘Can silage absorb more carbon than it emits during fermentation?’ This question still has to be answered.

Within this trial, the quantification of emission masses during anaerobic fermentation was determined considering the optional use of SA. The objectives of this article are (1) to examine whether former gas concentration measurements in the silo headspace [2] are combinable with gas quantity collection [28,29,30] to quantify the gas quantities formed; (2) to calculate the GHG, ethanol and ethyl acetate emissions of untreated or treated (biological inoculants or chemical additives) maize silage during varying anaerobic storage periods (duration 30 or 135 days); (3) to assess the temporal changes in gas formation and fixation during the ensiling process; and (4) to determine the chemical and microbiological parameters of the silage (as indicators of ensiling quality) and the emission quantities of climate- and environment-relevant gases.

Methods

Principles of the overarching trial and the two consecutive articles

A trial was conducted to determine the emissions of CO₂, nitrous oxide (N₂O), methane (CH₄), ethanol, and EA as indicators of VOC emissions from constant maize silage during anaerobic and aerobic storage. Constant material means that the forage was filled into silos, where it remained intact and unchanged for the entire trial duration (both storage periods). Forage material treatments were supplemented with SA to affect microbial metabolism. Heterofermentative LAB may lead to a trade-off between increased CO₂ formation during anaerobic fermentation and decreased respiratory losses during the feed-out phase. The impact of chemical SA on VOC gas formation during anaerobic fermentation has yet to be examined. To the authors' knowledge, this trial is the first to determine the emission quantities of constant silage material during all phases of silage storage. The results are to be presented in two consecutive research articles. This article (Part A) describes two sub-experiments (see Fig. 1). Experiment A1 focuses on gas formation and fixation during the anaerobic fermentation period using barrel silos. Furthermore, Part A includes the analysis of chemical and microbial composition of the treatments ensiled in glass jars used in parallel (Experiment A2). The second article (Part B) addresses the emissions during two aerobic feed-out periods and the sum emissions during anaerobic fermentation and feed-out. Furthermore, the second article provides a first step toward balancing SA's CF and effects on emission quantities during silage storage.

Fig. 1

Procedure of the overarching trial and the two consecutive articles. The processing of the treatments (grey boxes) is followed by the gas emission measurements (Article Part A, Experiment A1; blue boxes) and the analyses of chemical and microbial composition (Article Part A, Experiment A2; green boxes) during anaerobic fermentation. After 30 and 135 days of anaerobic storage, two aerobic emission measurement periods (AEMP) follow (Article Part B; yellow boxes). Treatments: treatment containing no additive (CON), treatment containing biological additive (BIO), and treatment containing chemical additive (CHE)

Forage material, silage treatments, and ensiling management

Whole-plant maize (Zea mays; variant SY Werena, Syngenta Agro GmbH, Frankfurt am Main, Germany) grown at the Campus Frankenforst of the University of Bonn (Königswinter, Germany; 50° 42′ 50.1′′ N, 7° 12′ 24.9′′ E) was harvested on 6th October 2021 using a forage harvester with a corncracker (Claas Jaguar 940, Claas KGaA mbH, Harsewinkel, Germany). Chopped material was collected randomly. The theoretical cutting length was set to 11 mm. However, 7% of the particles were ≥ 15 mm, 28% were 10–15 mm, 34% were 6–10 mm, 22% were 3–6 mm, and 9% were < 3 mm long.

The forage material was split into equal parts to prepare the three silage treatments. The control treatment (CON) was not supplemented with a silage additive. The chemical additive treatment (CHE) was supplemented with the additive Kofasil Stabil (Addcon GmbH, Bitterfeld-Wolfen, Germany) to improve the ASTA. The resulting dosage was 0.5 g of sodium benzoate per kg fresh matter (kg_FM) and 0.3 g of potassium sorbate per kg_FM. The biological additive treatment (BIO) was supplemented with 1.0 × 10⁸ colony-forming units (CFU) of Lentilactobacillus buchneri per kg_FM and 1.0 × 10⁷ CFU of Lactiplantibacillus plantarum per kg_FM of silage. In detail, the original concentrations of the bacteria in the additive (SILA-BAC RAPID REACT Maize Combi, Pioneer Hi-Bred International Inc., Johnston, Iowa, USA) were obtained as follows: Llb. buchneri ATCC PTA-2494 7.0 × 10¹⁰ CFU g^–1, Llb. buchneri NRRL B-50733 3.0 × 10¹⁰ CFU g^–1, Lpb. plantarum DSM 18112 5.0 × 10⁹ CFU g^–1, and Lpb. plantarum ATCC 55942 5.0 × 10⁹ CFU g^–1. LAB are considered bacterial strains for short ensiling periods. Both additives were applied according to the manufacturers’ dosage recommendations using a cleaned backpack sprayer, while the material was mixed regularly.

In Experiment A1, the silage treatments (n = 4) were packed into high-density polyethene barrels (34.8 L maximum volume). The barrel-specific silage material was retained within the barrels throughout the anaerobic fermentation (Experiment A1) and aerobic measurement period (Article Part B). Consequently, constant material was employed for both emission trials. The filling was performed in layers to ensure even filling and packing density. In detail, silage was loaded onto perforated plastic intermediate plates (diameter of 27.5 cm; Fig. 2), which were positioned on small stands into plastic barrels and had 63 drill holes (diameter of 10 mm). Underneath these intermediate plates was a 3 L volume of gas space (hereafter referred to as floor space) to ensure adequate ventilation of the silage during the following aerobic emission measurements (Article Part B). The silage used to fill each barrel had a mass of 10,206.6 ± 7.6 g_FM for CON, 10,200.3 ± 1.3 g_FM for BIO and 10,198.9 ± 0.9 g_FM for CHE; a volume of 28.8 L for all silos; and a resulting packing density of 150.60 ± 0.08 kg_DM m^–3. During packing, a temperature logger (Testo 174 T, Testo SE Co. KGaA, Titisee-Neustadt, Germany) was positioned in the centre of the silage material. The silage was filled in such a volume that a gas space (3 L in volume, hereafter referred to as head space) remained above it when the barrel cover was put on and closed with a clamping ring. Both the barrel (at the level of the floor space) and the cover were equipped with two hose connections each (Sect. “Measurement of silage emissions”).

Fig. 2

Maize silage barrels and set-up used in the anaerobic emission measurements

On the harvest date, the silage barrels of the CON treatment were filled first. Afterwards, the CHE forage and the BIO forage were treated, and the silos were packed. All 12 silage barrels were closed and sealed simultaneously. This should ensure (a) that the silage is exposed to oxygen for the same length of time and (b) that the ensiling process starts simultaneously.

The barrels were transported to the Institute of Agricultural Engineering (University of Bonn) and stored indoors (for ambient temperatures, see Sect. “Ambient air and silage temperatures”). During anaerobic storage, the barrels were regularly checked, and emission measurements were carried out (Sect. “Measurement of silage emissions”). After 30 days of ensiling, 6 barrels (2 of each treatment) were opened for aerobic emission measurements (Fig. 1 and Article Part B). Anaerobic emission measurements were taken with the remaining 6 barrels until the second aerobic measurement period started on day 135.

In Experiment A2, in parallel with the barrel preparation, the same silage variant material was added to glass jars (maximum volume 1.8 L; J. Weck GmbH u. Co. KG, Wehr-Öflingen, Germany). The jars were filled with 0.602 kg_FM up to a volume of 1.65 litres. This corresponds to a packing density of 155.1 kg_DM m^–3 and is therefore at a similar level as the barrel silos. The 36 jars (n = 12 per treatment) provided the material for the chemical and microbiological analyses (Sect. “Laboratory analysis of the silage material”), so the silage barrels in Experiment A1 were unaffected during the ensiling process.

Laboratory analysis of the silage material

Material samples were collected before the CON barrels were packed for chemical and microbiological analysis on the harvest date. In Experiment A2, the ensiled material was collected for analysis on ensiling days 2, 14, 30, and 135 (Fig. 1). Samples for the chemical analyses were stored at – 18 °C immediately after sampling; samples for microbiological analyses were stored at 4 °C and analysed on the same day.

Silage barrels and glass jars were weighed during the anaerobic fermentation period on ensiling days 0, 2, 14, 30, and 135. The silo masses were used to calculate the DM losses of the silage during the fermentation process. Two balances were used to balance the barrel (range 0–35,100 g, readability of 0.10 g; BBK 422-35 LA, Mettler Toledo, Germany) and the glass jar weights (range 0–2410 g, readability of 0.01 g, linearity ± 0.05 g; KB 2400-2N, Kern & Sohn GmbH, Balingen, Germany). Crude ash, sugar, starch, crude fibre, crude protein, utilisable crude protein at the duodenum, and metabolizable energy concentrations were analysed according to the German Handbook of Agricultural Research and Analytic Methods [46].

Organic acids, alcohols and esters, pH, and water soluble carbohydrates (WSC) were analysed in aqueous silage extracts after mixing 50 g of frozen silage material with toluene (1 mL) and distilled water (300 mL) [47]. Subsequently, various analyses were performed after filtration (MN 615 filter paper; Machery-Nagel, Düren, Germany) and microfiltration (0.45 µm pore size, Minisart RC, Sartorius, Göttingen, Germany) on the following day. Lactic acid (LA) was detected using high-performance liquid chromatography (refractive index detection; LC-20 AB, Shimadzu Deutschland, Duisburg, Germany; [48]). Volatile organic acids, alcohols, and ethyl esters (including EA) were determined using gas chromatography (GC) with a free fatty acid phase column (Permabond FFAP 0.25 µm, Macherey–Nagel, Düren, Germany) or an optima wax column (Macherey–Nagel, Düren, Germany), respectively, and a flame ionisation detector (GC-2010, Shimadzu, Deutschland, Duisburg, Germany) [47]. The detection limit for all parameters with the free fatty acid phase column was 0.01% of FM, and that with the optima wax column was 0.001% of FM. WSC were analysed by the anthrone method [49] using a continuous flow analyser (Scan++ , Skalar Analytical, Breda, The Netherlands). The pH was analysed potentiometrically using a calibrated pH electrode. The DM concentration was corrected based on Weißbach and Strubelt [50].

The microbial analysis procedure was conducted by two different laboratories. The first analysed the fresh material samples according to the methods of VDLUFA [46] for aerobic and mesophilic bacteria, moulds, Dematiaceae, and yeasts (methods 28.1.2 and 28.1.3). In the first step, 20 g of silage was suspended in 180 mL of solution (pH 7.0, 0.58 g L^–1 NaH₂PO₄, 2.5 g L^–1 Na₂HPO₄, 4.0 g L^–1 NaCl, 1.0 g L^–1 peptone, and 0.3 mL L^–1 Tween 80) and treated with a paddle blender. From this solution, subsequent dilutions were prepared in phosphate buffer (pH 7.0, 0.58 g L^–1 NaH₂PO₄ × 2 H₂O, 2.5 g L^–1 Na₂HPO₄ × 2 H₂O, and 4.0 g L^–1 NaCl), and the appropriate dilutions were used for microbial analysis. For bacteria, tryptose/TTC agar (pH 7.3, 20.0 g tryptose, 1.0 g L^–1 glucose, 5.0 g L^–1 NaCL, 15.0 g L^–1 agar, and 10 mg L^–1 2,3,5-triphenyltetrazolium chloride) was used, and the plates were incubated for 2 days at 30 °C. For fungi, rose-bengal chloramphenicol agar supplemented with Tergitol (pH 7.2, 5.0 g peptone, 10.0 g L^–1 glucose, 1.0 g L^–1 K₂PO₄, 0.5 g L^–1 MgSO₄, 0.05 mg L^–1 Rose-Bengal, 15.5 g L^–1 agar, 0.1 ml L^–1 Tergitol, and 20 mg L^–1 chlortetracyclin-HCl) was incubated for 3 days at 25 °C. For enumeration of mesophilic LAB (method 28.3.3, [51]), pour-plates of de Man, Rogosa, and Sharpe agar (68.2 g L^–1; type 1.10660, Merck, Darmstadt, Germany)—to provide micro-aerophilic conditions—with an overlay were prepared from the dilutions used for determination of aerobic, mesophilic bacteria, and fungi and incubated at 30 °C for 5 days. The second laboratory analysed the samples collected on ensiling days 2, 14, 30, and 135. For this purpose, 30 g of silage was suspended and homogenised in ¼-strength Ringer solution (0.05 g L^–1 NaHCO₃, 0.06 g L^–1 CaCl₂, 0.105 g L^–1 KCl, 2.25 g L^–1 NaCl; Merck, Darmstadt, Germany). This suspension was used for the analysis of total bacterial counts on plate-count agar (pH 7.0, 1.0 g L^–1 glucose, 2.5 g L^–1 yeast extract, 5.0 g L^–1 enzymatic digest of casein, 15 g L^–1 agar; Merck, Darmstadt, Germany) after 2 days of incubation at 30 °C; LAB counts on de Man, Rogosa, and Sharpe agar (Merck, Darmstadt, Germany) after 3 days of incubation at 30 °C under anaerobic conditions; and yeasts and moulds on yeast extract glucose chloramphenicol agar (pH 6.6, 0.1 g L^–1 chloramphenicol, 5.0 g L^–1 yeast extract, 14.9 g L^–1 agar, and 20.0 g L^–1 glucose; Merck, Darmstadt, Germany) after 3 days of incubation at 25 °C.

Furthermore, silage quality in Experiment A2 was assessed using the V-Score according to the procedure described by the Society of Utilisation of Self Supplied Feeds (2009) [52] and applied by Tian et al. [53]. V-Scores of Y > 80 were considered favourable, 80 ≥ Y ≥ 60 average and 60 > Y bad silage quality [54].

In Experiment A1, silage quality was assessed using the silage scoring system of the German Agricultural Society [55, 56] after opening the barrels. This methodology helps to assess the quality of various silage parameters, such as smell, structure, colour, and mould. A more detailed qualitative observation supplemented this scoring system to assess mould and yeast contamination by two trained persons. For this purpose, a scale from 0 (very good) to 5 (very bad) was used, which included the following values: 0.0 = no mould/yeast spots; 0.5 = occasional mould/yeast spots, approx. < 5% of the surface; 1.0 = occasional mould/yeast spots, approx. 5% of the surface; 2 = small mould/yeast nests, approx. 15% of the silo face; 2.5 = small mould/yeast nests, approx. 20% of the silo face; and 3.5 = multiplied mould/yeast nests, approx. 30% of the silo face.

Measurement of silage emissions

During the anaerobic storage period, measurements of silage emissions occurred regularly in Experiment A1 (Fig. 1), hereafter referred to as measurement time points.

Each silo barrel had four hose connectors (Sect. “Forage material, silage treatments, and ensiling management”). Two connectors (at the level of the floor space) were closed during the ensiling process to ensure anaerobic conditions.

One of the connectors in the cover was attached to a short hose (all hoses in the experimental set-up were made of polytetrafluoroethylene unless otherwise stated), to which a ball valve and a rubber septum were mounted. The ball valve was only opened for gas sampling. For this purpose, a laboratory syringe (50 mL volume) was inserted into the rubber septum, and 50 mL of gas was removed and pumped back into the barrel twice. This procedure was used to ensure homogeneous mixing of the gas in the barrel headspace. A double needle was subsequently inserted, and four vacuumed glass vials (20 mL volume each) were filled one after the other. This procedure was performed for all the barrels in the anaerobic storage phase, i.e. for 12 barrels between ensiling days 0 and 30 and 6 barrels between ensiling days 30 and 135, respectively. A similar methodology was used by Schmithausen et al. [2].

The other connector in the cover was connected to a gas sampling bag (nominal volume 25 L), which collected the gases formed by the silage during the ensiling process [30]. For the bag connection, polyurethane and polytetrafluoroethylene hoses were fitted using connectors. The system of a barrel and a gas bag can be regarded as a zero-pressure system, as inflation of the bags captured the formed gas. After gas sampling at ensiling hour 36, each gas bag was exchanged for an empty, new gas bag due to the high gas formation of the silage material.

After gas sampling, the gas bags were carefully clamped in a calliper with flat pads until the bags were under tension to a certain degree. The filling volume of the bags could be measured using the deflection of the calliper and a preliminary calibrated scale. Afterwards, all hose connectors were checked for gas tightness, and the bags were stored until the subsequent measurements. To ensure the comparability of this procedure, all the measurements were taken by only two trained persons. Furthermore, this procedure was carried out indoors to avoid any change in the gas volume due to temperature (Sect. “Ambient air and silage temperatures”). After 36 h of ensiling, the first and second gas bag volumes were added to determine the total gas formation quantity per silo.

The gas samples in the vials were used to analyse the composition of the gas mixture. One of the vials was used to analyse the greenhouse gas concentrations using a gas chromatograph (electron capture detector and a flame ionisation detector; model 8610C, SRI Instruments, Torrance, CA, USA). Due to the high CO₂ concentrations in the barrel headspace, a diluted sample (diluted 1:101 with room air) was analysed (detection limit of 50.00 ppm). The initial CO₂ concentration was calculated using the CO₂ concentration in the room air. Subsequently, the undiluted gas sample was used to analyse the concentrations of CH₄ (detection limit of 0.08 ppm) and N₂O (detection limit of 0.01 ppm). The subsequent values considered the amount of gas taken for CO₂ analysis. This procedure of Experiment A1 is similar to the methodology used in the previous trials [2, 35].

Another vial was used to analyse the concentrations of VOCs. For this purpose, the sample air was diluted with room air (dilution 1:153) and then analysed using infrared photoacoustic spectroscopy (PAS; Multi-Gas Analyser INNOVA 1312; LumaSense Technologies SA, Ballerup, Denmark). Cross and water compensation were turned on for measurement [21, 35]. The initial concentration was calculated again based on the results of the room air analysis. The accuracy of the Multi-Gas Analyser INNOVA 1312 was 3% of the gas concentration, and the detection limits, based on the calibration chart of the manufacturer, were 0.1 ppm for ethanol and 0.02 ppm for EA. Furthermore, the analyser was used to measure the CO₂ concentration. These data are not shown but should be considered in terms of cross-compensation (Sect. “Examination of the methodological procedure”).

Each gas bag was sampled after removing it from the barrel.

The remaining two vials per barrel and measurement time point were stored in the laboratory for use in case of erroneous measurements. In Experiment A1, 2280 vials were filled; 1681 gas samples were analysed using GC, and 926 using PAS.

Calculation of gas emissions during anaerobic storage

Due to the accumulation of gases formed in the zero-pressure system (barrel plus gas bag), cumulative gas emissions were calculated for each measurement time point in Experiment A1.

For this purpose, homogeneous gas dispersion within the total gas space of each zero-pressure system was assumed. The total gas space was calculated using Eq. 1

$${V}_{\text{gas}}= {V}_{\text{bag}}+{V}_{\text{headspace}}+{V}_{\text{gas pores}}+{V}_{\text{floorspace}},$$

(1)

where ${V}_{\text{gas}}$ is the volume of the total gas space [L]; ${V}_{\text{bag}}$ is the volume of the gas bag [L]; ${V}_{\text{headspace}}$ is the volume of the headspace [L] (3 L volume); ${V}_{\text{gas pores}}$ is the volume of the gas pores in the packed silage material [L] (Eqs. 2 and 3); and ${V}_{\text{floorspace}}$ is the volume of the floorspace [L] (3 L volume).

The volume of the gas pores was calculated using Eqs. 2 and 3 [57]

$${V}_{\text{gas pores}}= {V}_{\text{silage}}*{\text{porosity}}_{\text{silage}}$$

(2)

$${V}_{\text{gas pores}}= {V}_{\text{silage}}*\left(1.733*\text{DM}-0.256*\text{density}+39.778\right),$$

(3)

where ${V}_{\text{gas pores}}$ is the volume of the gas pores in the packed silage material [L]; ${\text{porosity}}_{\text{silage}}$ is the ratio of gas pores in the volume of packed silage [%]; ${V}_{\text{silage}}$ is the volume of the packed silage material in the barrel [L] (28.8 L volume); DM is the dry matter concentration of the silage material [%] (425 g_DM kg_FM^–1 during silo packing); and density is the packing density [kg_DM m^–3] (150.6 kg_DM m^–3). At the time of silo closure, the porosity was 74.90% ± 0.02%, ${V}_{\text{gas pores}}$ 21.57 ± 0.01 L and ${V}_{\text{gas}}$ 27.57 ± 0.01 L for the twelve barrels.

The cumulative gas emission quantities were calculated for each measurement time point i in Experiment A1 using Eq. 4

$${M}_{\text{gas},\text{ i}}= {V}_{\text{gas},\text{ i}}*{c}_{\text{gas},\text{ i}},$$

(4)

where, ${M}_{\text{gas},\text{ i}}$ is the cumulative gas emission mass [g]; ${V}_{\text{gas},\text{ i}}$ is the total gas space at this measurement time point i [L] (Eqs. 1 to 3); and ${c}_{\text{gas}, i}$ is the gas concentration in the gas bag at measuring time point i [g_gas L^–1].

Equation 4 was also used for calculating the gas emission masses in the various gas bags. After the gas bags of all the zero-pressure systems were changed at ensiling hour 36, the gases in the system (barrel plus second gas bag) were added to the gases in the first bags.

The maximum gas emissions per silo in Experiment A1 were the gas emission masses at that measurement time point when the gas formation quantity—i.e. the sum of the gas volume of the gas bag one and two per silo—reached its maximum.

Data processing and statistics

The following conversion ratios were used: 1 ppm CO₂ = 1.83 (mg CO₂) m^–3; 1 ppm CH₄ = 0.67 (mg CH₄) m^–3; and 1 ppm N₂O = 1.83 (mg N₂O) m^–3. In Experiment A1, the gas quantities formed per silage mass are given for the DM mass at the time of ensiling (day 0), according to Bueno et al. [29]. The following GWP were applied according to the fourth assessment report of the Intergovernmental Panel on Climate Change (IPCC) [58]: CO₂ = 1, CH₄ = 25, and N₂O = 298. The CO₂ equivalent (CO₂eq) emissions derived in this trial considered only climate-relevant gases, i.e. CH₄ and N₂O; GHG emissions also included CO₂. DM losses are indicated by negative values, and DM gains are indicated by positive values.

In Experiment A2, the chemical compositions and microbial counts of the fresh and ensiled materials were compared using one-way analysis of variance (ANOVA). In Experiment A1, the gas formation quantities, the gas concentrations, and the subsequently calculated gas emission masses were compared using mixed ANOVA, with subsequent one-way ANOVAs for each measurement analysis interval. These intervals differ from the measurement time points described above (Sects. “Measurement of silage emissions” and “Calculation of gas emissions during anaerobic storage”). Multiple measurement time points were combined into one measurement analysis interval to ensure a sufficient sample size for the ANOVA (Table S1). For each one-way ANOVA, if homogeneity of variance was given, Tukey’s-HSD test was used for post hoc significance comparison; if not, a Welch-ANOVA was followed by a Games–Howell post hoc test. Linear correlations were analysed using Spearman correlation. In all the analyses, p < 0.05 was considered to indicate statistical significance.

Microsoft Office Excel 2019 (Microsoft Corporation, Redmond, Washington, DC, USA) was used for descriptive data analysis. IBM SPSS 26.0 (International Business Machines Corporation Armonk, New York, NY, USA) was used for statistical analysis.

Results

Ambient air and silage temperatures

The maize was harvested at an ambient air temperature of 12.2 ± 0.4 °C. At the time of silo closing, the silage temperatures were 13.9 ± 0.1 °C for CON, 13.1 ± 0.1 °C for BIO, and 13.4 ± 0.1 °C for CHE barrels in Experiment A1. After 2 h, the barrels and glass jars were stored indoors (19.0 ± 1.4 °C). CON barrels reached the temperature level of ambient air after 1.22 ± 0.02 d, BIO after 1.18 ± 0.01 d, and CHE after 1.23 ± 0.02 d. Afterwards, the temperatures of the BIO barrels were greater than those of the CON (+ 0.3 ± 0.1 K) and CHE (+ 0.3 ± 0.1 K) barrels until ensiling hour 36. The silage temperatures subsequently remained at steady levels in Experiment A1: 18.7 ± 0.9 °C for the CON barrels, 18.7 ± 1.0 °C for the BIO barrels, and 18.7 ± 1.0 °C for the CHE barrels.

Composition of the silage

The DM losses in Experiment A1 and the chemical compositions, the V-Score, and microbial counts of the fresh and ensiled materials of the silage treatments in Experiment A2 are shown in Tables 1 and 2. In Experiment A1, all the silage barrels exhibited decreasing FM weights throughout the fermentation process. These results were as follows: for d2, CON − 0.33% ± 0.01%, BIO − 0.36% ± 0.01%, and CHE − 0.29% ± 0.01%; for d14, − 0.51% ± 0.02%, − 0.55% ± 0.02%, and − 0.46% ± 0.02%; for d30, − 0.56% ± 0.03%, − 0.66% ± 0.02%, and − 0.54% ± 0.02%; and for d135, − 0.74% ± 0.01%, − 1.27% ± 0.04%, and − 0.71% ± 0.03%, respectively. Considering the DM of the silage material (based on the DM in Experiment A2), CON barrels indicated the most considerable DM losses between ensiling days 0 and 14, followed by BIO and CHE (p < 0.05). However, the DM losses were not linear for all the treatments over time (Table 1). CON and CHE presented DM mass increases between ensiling days 14 and 30, and subsequent losses occurred between days 30 and 135. BIO showed a constant increase in DM after ensiling day 14. Based on the silo fresh mass decrease, this DM mass increase resulted from rising DM concentrations (Sect. “Gas quantities decrease and DM mass increases”). In Experiment A2, the glass jars indicated similar FM and DM mass losses throughout the fermentation process.

Table 1 Chemical composition and energy concentration of fresh maize and ensiled material for the silage treatments in Experiment A2 (unless otherwise stated)

Table 2 Chemical composition, the V-Score, and microbial counts of fresh maize and ensiled material for the silage treatments in Experiment A2

In Experiment A2, VOC concentrations in the silage material differed between the treatments. At ensiling day 30, the ethanol concentrations did not vary, but BIO had higher EA concentrations than CON and CHE did (p < 0.05). On ensiling day 135, BIO had higher ethanol, propanol, 1,2-propanediol, 2-butanol, and EA but lower ethyl lactate concentrations than CON (p < 0.05). In the BIO treatment, a decrease in LA and an increase in AA concentrations were consistent with an increase in EA and a decrease in ethyl lactate concentrations.

All silage treatments are characterised by a quick decrease in pH within the first 2 ensiling days in Experiment A2; BIO increases after ensiling day 30. This aligns with a decrease in LA, sugar, and WSC concentrations, and an increase in AA concentrations. All the silages were free of butyric acid (not shown), and only BIO had small amounts of propionic acid at ensiling day 135.

In Experiment A2, all the treatments had higher LAB counts on ensiling day 2 than on day 0. This difference aligns with the findings of LA formation and a decrease in pH during this period. Subsequently, the LAB counts decreased continuously until ensiling day 30 in all the treatments. BIO had significantly greater counts since ensiling day 30; the counts at ensiling day 135 were similar to those at ensiling day 2, while CON and CHE showed noticeable decreases. Yeast counts decreased during anaerobic storage in all the treatments. In detail, the decline is fastest in the CHE treatment and slowest in the CON treatment. The high yeast counts in the CON treatment at ensiling day 135 are due to an outlier (possibly a tiny oxygen leakage in one of the silage glass jars), proven by the high standard deviation. However, since the other parameters of this sample did not show outliers, this sample was not excluded.

The V-Score showed consistently high values of ≥ 90 indicating favourable silage quality. The lowest values were indicated by the BIO treatment based on the ongoing AA production.

In Experiment A1, the inspection of the silage faces in the silage barrels on the days of opening showed a consistently high silage quality (Tables S2 and S3). The BIO barrels indicated a slight smell of alcohol on ensiling day 30, even though the BIO ethanol concentrations in the material (Experiment A2) were slightly below the level of CON. At ensiling day 30, BIO and CHE barrels showed a pungent odour of AA, which increased in the BIO treatment until ensiling day 135. In addition, slight mould growth and a recurring yeast formation were evident in individual barrels. At ensiling day 30, 5–20% of CONs, 15% of BIOs, and 15% of CHE’s silage faces were covered with yeast spots; at day 135, 15–30%, < 5–5%, and 5–15%, respectively (Tables S2 and S3).

Formation of greenhouse gases

In Experiment A1, the courses of the gas volumes and the corresponding cumulative gas quantities within the barrels’ head space and gas bags, respectively, are visualised in Fig. 3. The central part of the gas formation occurred within the first ensiling days. The gas bags’ volume increased after ensiling hour 8. Until this time, barrels formed 11% ± 1% of the total CO₂ gas quantities. After 2 ensiling days, 78% ± 4% of the CO₂ was generated, and after 5 days, 93% ± 4% was generated. CON indicated increasing cumulative gas quantities until ensiling day 19.5 ± 6.4, BIO until ensiling day 60.0 ± 49.5, and CHE until ensiling day 14.0 ± 0.0 (Fig. 3, S1 and Table 3). CHE had a slower increase in gas volume (for ensiling days 0–14; see Fig. S1), with significantly lower volumes since ensiling day 0.5 (p < 0.05; see Table S4 for selected data). CON and BIO differ significantly (p < 0.05) for ensiling days 30–70. However, the large standard deviation in the BIO treatment since ensiling day 70 affects the statistical analysis. The central part of the gas formed is CO₂, which shows a strong positive correlation between the gas and CO₂ quantities (Fig. S2). Most of the CO₂ generated during the main fermentation phase (ensiling days 0.3–4.0). During this phase, CO₂ concentrations reach 1.20 × 10⁶ mg m^–3 (≈ 6.6 × 10⁵ ppm), with a subsequent regressive decrease. After a certain point, i.e. between ensiling days 50 and 80, depending on the treatment, the CO₂ quantities within the total gas space remain at a (nearly) constant level.

Fig. 3

Cumulative gas, CO₂, CH₄, and N₂O quantities within the zero-pressure systems during the ensiling process. Treatments: treatment containing no additive (CON), treatment containing biological additive (BIO), and treatment containing chemical additive (CHE). Error bars indicate the standard deviation of all treatments' measurement time point values within each analysis interval

Table 3 Maximum cumulative gas quantity collected within the gas bags and cumulative GHG and VOC quantities at these time points

The CH₄ concentrations in the headspace reached 4.69–5.13 mg m^–3 (≅ 7.04–7.69 ppm) between ensiling hours 132 and 144. The N₂O concentrations peaked at 49.9–66.2 mg m^–3 (≅ 27.4–36.4 ppm) at ensiling hour 44. CHE had lower CH₄ quantities than CON did between ensiling days 20 and 135 (p < 0.05). The combination of lower gas formation and CH₄ concentrations leads to less CH₄ emissions in the CHE treatment. A noticeable standard deviation of the BIO values reduces the statistical accuracy. For N₂O, CHE had significantly greater quantities than CON and BIO since ensiling day 1.5 (p < 0.05).

Several barrels show a constant decrease in gas quantities in the gas bags after 14–25 ensiling days (Fig. 3, S1 and Table 3; exceptions barrels BIO3 and BIO4). However, focusing on the individual gases, CO₂ quantities peaked between ensiling days 4.0 and 7.5, and an average reduction of − 25.3% occurred until ensiling day 30 (− 44.3% until ensiling day 135). CH₄ quantities peaked between ensiling days 5 and 12 and decreased by − 31.0% and − 49.0%, respectively; N₂O quantities peaked between ensiling days 1.5 and 2.7 and decreased by − 51.2% and -76.3%, respectively.

All the barrels exhibited DM loss and gas formation until ensiling day 14 (Fig. S3). These parameters show strong correlations. Subsequently, the gas quantities decreased, and the DM masses of the silos increased (+ 0.88% for CON, + 0.99% for BIO, and + 1.64% for CHE; Table 1) between ensiling days 14–30 compared to ensiling day 14 (Fig. S4). Statistical correlations were partly given. However, the net losses were still negative. Between days 30 and 135, CON and CHE exhibited DM losses and a decrease in gas quantities, while BIO exhibited nearly constant gas quantities and an increase in DM mass.

When gas bags were exchanged at ensiling hour 36, GHG concentrations were greater in the headspace than in the first gas bags; on ensiling days 30 and 135, the opposite was true for the second gas bags.

Formation of ethanol and ethyl acetate

The cumulative ethanol and EA quantities during the anaerobic storage period are displayed in Fig. 4 and S5. Ethanol quantities rapidly increased within the first 5 ensiling days, and the gas quantity peaked at ensiling day 4.7 ± 0.4. After 2 ensiling days, 55% ± 8% ethanol was formed, and after 5 days, 94% ± 6% ethanol was produced. Subsequently, constant quantities were observed for CON and CHE silos after a certain point during the anaerobic storage period. Ethanol gas quantities decreased by − 38% to − 47% of the maximum quantities until ensiling day 30 and − 30% to − 62% until ensiling day 135. BIO showed a constant increase after ensiling day 60 due to the increasing quantities in the BIO4 barrel (Table 3). CHE showed the lowest cumulative ethanol formation quantities throughout the anaerobic storage period (ensiling days 0.5 to 135, p < 0.05) compared to CON and BIO; between the latter two, CON values were significantly lower than BIO values after ensiling day 70 (p < 0.05).

Fig. 4

Cumulative ethanol and ethyl acetate quantities within the zero-pressure systems during the ensiling process. Treatments: treatment containing no additive (CON), treatment containing biological additive (BIO), and treatment containing chemical additive (CHE). Error bars indicate the standard deviation of all treatments' measurement time point values within each analysis interval

For EA, after the gas bags were exchanged at ensiling hour 36, all the treatments exhibited a spontaneous increase in the calculated quantities (Fig. S5). Subsequently, CHE increased until ensiling days 5 to 6; CON and BIO increased until ensiling days 12 and 30. After this, the EA quantities remained steady or showed small decreases (Fig. 4). At ensiling day 30, CON and CHE exhibited losses of − 17% to − 23% compared to the peak quantities. The decline changed to − 30% to − 34% at ensiling day 135. However, the BIO3 and BIO4 barrels exhibited a substantial increase in EA formation during the ongoing storage period (Table 3). These barrels formed only 23%–61% of the final quantity of EA gas before ensiling day 30. CHE indicates significantly lower EA gas quantities than CON and BIO at ensiling day 2 and between ensiling days 4 and 100 (p < 0.05). BIO had greater quantities than CON did between ensiling days 3 and 100 (p < 0.05).

The cumulative ethanol and EA gas quantities exhibited a strong linear correlation for all the treatments (Fig. S6).

The relationships between the ethanol and EA concentrations in the silage material and the cumulative gas quantities are shown in Fig. 5. For ethanol, the CON material exhibited a concentration increase until ensiling day 30 and a subsequent decrease (Table 2). BIO and CHE levels continuously increased. However, the gas quantity peak during the first days cannot be explained by a material concentration peak but aligns with the ethanol formation in the silage material. For EA, concentrations within the silage material increase until ensiling day 30, including local minima on day 14 for BIO and CHE; after this, all treatments indicate a decrease, which is in line with decreasing gas quantities for CON and CHE. Higher concentrations of EA in BIO material beginning on ensiling day 30 led to increased quantities of EA gas. However, higher material concentrations provide a limited explanation for the differences between BIO3 and BIO4.

Fig. 5

Ethanol and ethyl acetate concentrations in the silage material and cumulative gas quantities in the zero-pressure systems. Treatments: treatment containing no additive (CON), treatment containing biological additive (BIO), and treatment containing chemical additive (CHE)

Ethanol concentrations within the material correlated positively with EA concentrations in CON (r_S = 0.655, p < 0.05), and tended to in BIO (r_S = 0.545, p = 0.08). For BIO, the ethanol concentrations in the material (r_S = 0.609, p < 0.05) and the gas quantities (r_S = 0.691, p < 0.05) correlate with the quantity of EA gas. The treatment CHE showed no significant correlations.

Emissions of climate- and environment-relevant gases

Table 3 shows the maximum emission quantities of climate- and environment-relevant gases. CHE indicates significantly lower gas formation quantities than CON and BIO but the highest N₂O and CO₂eq emission quantities (p < 0.05). Furthermore, CHE had lower CO₂ and ethanol emission quantities than CON did (p < 0.05). EA emission quantities show no differences. BIO indicates lower CH₄, N₂O, and CO₂eq emission quantities than CON (p < 0.05).

Discussion

Composition of the silage

Overall, the DM of the fresh material was greater than the target value of 30%–40% [59, 60]. However, the silage quality meets the requirements for modern maize silage [59] and the V-scores indicate excellent qualities. DM affects microbial activity and subsequent gas formation. With high DM, microbial metabolism seems to be less active: fewer acid quantities are generated [61], gas formation is delayed [2], and less gas is formed [38]. The pH values determined via laboratory analysis are in the recommended range of 3.7–4.0, the latest since ensiling day 14 [60]. A continuous increase in LA is an indicator of homofermentative LAB activity, and an increase in AA is an indicator for heterofermentative LAB activity [6, 18, 62]. The LA concentrations for BIO on ensiling day 135 are below the target range (30–60 g kg_DM^–1); the opposite applies for AA concentrations (10–30 g kg_DM^–1) [60]. However, this pattern results from the use of biological inoculants: Llb. buchneri leads to the degradation of lactic acid to acetic acid, 1,2-propanediol, ethanol, and CO₂ [18, 62, 63]. The presence of small amounts of propionic acid at ensiling day 135 in BIO aligns with the activity of Llb. buchneri [60, 64].

The DM losses showed initial losses between ensiling days 0 and 14 and subsequent increases in DM mass up to day 30 for CON and CHE or day 135 for BIO (Sect. “Gas quantities decrease and DM mass increases”). The DM losses in the silage barrels are on the levels reported for commercial maize silos [10, 65]. However, as stated by Ostertag et al. [66], farm-scale silos exhibited greater DM losses than laboratory silos. Higher farm-scale losses seem reasonable due to prolonged oxygen supply [34, 41], lower gas tightness of the silos, and possible air leakages within silo sealing [14]. Furthermore, 67%–80% of the maximum DM losses (ensiling day 14) were detected within the first 2 ensiling days. The high DM and low packing density of silage lead to high porosity and high oxygen quantities in the silos at the moment of silo closure. This can promote the activity of plant material and aerobic microbes, e.g. enterobacteria or yeasts [1, 7]. Sun et al. [67] reported a negative correlation between packing density and Enterobacter abundance. Moreover, some articles have reported correlations between DM concentration and packing density and subsequent DM losses [7, 9, 68, 69]. For instance, Borreani et al. [7] report the findings of Holmes [70]: DM loss [%] = 29.1 – 0.058 × DM density [kg_DM m^–3]. The highest losses occurred with increasing DM concentration and low packing density [71]. These findings align with the noticeable GHG and VOC formation observed during the first days of ensiling (Sects. “Formation and emissions of greenhouse gases” and “Formation and emissions of ethanol and ethyl acetate”). In principle, higher DM leads to lower gas formation in the silage if the porosity in laboratory silos is consistently low [38]. Compared to silage with low DM, the fermentation intensity is reduced with dry silage, as the acid tolerance of the microbes is lower. In practical silos, however, compacting the dry particles is usually associated with greater effort; higher porosities and increased residual oxygen supply are the result. This was modelled in the laboratory silos here with high porosity. Therefore, the regression equations presented should be able to estimate the quantities of gas and CO₂ emitted by farm-scale silos during the main fermentation phase of maize silage with similar DM.

Formation and emissions of greenhouse gases

All three treatments indicate a rapid increase in GHG quantities formed and emitted into gas bags, which aligns with former reports stating a regressive course of accumulative CO₂ emission quantities [4, 5, 13, 38, 72, 73]. Emitted gas quantities vary between plant species [4] and are affected by SA use [5, 72]. In detail, SA did not affect the quantity of gas formed by maize silage during the first 10 ensiling days [5]. However, during the ongoing activity of heterofermentative LAB, additional gas was generated. The initial course of gas formation correlates with DM losses. The initial gas formation in the first phase of the ensiling process [1] is based on the aerobic activity of the plant material and microorganisms [2, 74]. The barrel silos indicated noticeable yeast counts on the silage face at ensiling day 30. This could result from high initial oxygen availability and low yeast inhibition due to lower CO₂ concentrations in zero-pressure systems [75].

Several studies have measured gas concentrations within silos [2, 34, 76, 77]. The CO₂ concentrations presented here (approximately 66%) support the data from Peterson et al. [76] and Wang and Burris [34], who reported CO₂ concentrations of 75–85% after 48 h of ensiling in commercial farm-scale maize silos. The values seem comparable, considering the gas tightness of the zero-pressure set-up and the quantities of N₂ remaining in the system.

In addition to plant and aerobic microorganism activity, LAB are considered to be tolerant to aerobic conditions and are active in the first phase of ensiling [7, 78,79,80]. For instance, Lpb. plantarum can metabolise glucose first to lactate and subsequently to acetate under aerobic circumstances resulting in CO₂ formation [78, 81, 82]. However, based on the abundance in the epiphytic microbial community [1, 83], proteobacteria and particularly enterobacteria seem to be the most important microorganisms for oxygen depletion, CO₂ formation, and pH decrease in this phase [80, 84]. Other microorganisms such as yeasts contribute to CO₂ formation during aerobic respiration.

Unfortunately, gas sampling and analysis could not measure the oxygen concentrations within the silos’ headspaces. Therefore, it is unclear when absolute anaerobic conditions were present. Some of the O₂ within the barrel may already be respired between filling and sealing the barrel. The availability of oxygen differs for the various layers and positions within silage due to metabolic respiration and limited oxygen diffusion [2, 20, 85, 86]. In former trials, the aerobic phase in maize silos lasted only for several hours, e.g. 1.4–3.0 h in laboratory silos and 6.0 h in farm-scale silos [34, 41, 80]. The rapid decrease of pH until ensiling day 2 confirmed anaerobic conditions before this point. It is assumed that anaerobic conditions apply the latest at ensiling hour 8, based on the literature review, and that the gas bag volume starts to increase. The respiration of O₂ to CO₂ considers the constant gas volume of both gases.

After this point, enterobacteria, yeasts, LAB, and propionic acid bacteria can form CO₂ during anaerobic fermentation [8, 87]. Approximately 87%–93% of the total CO₂ is generated after ensiling hour 8, i.e. in the anaerobic phase. The formation of LA—in combination with decreased water activity, microbial substrate concurrence, and additional factors—inhibits microbial activity when the pH tolerance of the bacteria is undercut. Currently, it is unclear which ratios of total CO₂ production can be assigned to which phyla or genera of microorganisms. Sun et al. [80] described a high abundance of enterobacteria in the first hours after silo closure. The increase in 2,3-butanediol quantities formed within the first 2 ensiling days indicates the metabolism of enterobacteria during neutral fermentation [8, 79, 87]. In this phase, anaerobic fermentation by enterobacteria leads to the synthesis of AA or CO₂, among other products_, and DM losses of -17% occur [79]. For this purpose, LA is a potential substrate [88] that affects the slope of the pH decrease. Vigne [84] reported that enterobacterial metabolism is the major gas producer in this phase. Upon further metabolism under acidic conditions (down to pH ≥ 4.5) [1], glucose is metabolised to lactate, acetate, ethanol and CO₂ [79]. However, the abundance of enterobacteria decreased with decreasing pH to > 5.5 [80]. Yeasts ferment glucose to ethanol and CO₂, leading to DM losses of − 49% [7, 79]. Therefore, the formation of 1 mol of ethanol by yeasts is associated with greater DM losses (− 24.5%) than is the formation of ethanol by enterobacteria (− 17%) or heterofermentative LAB (− 17%). The ratio of ethanol production during the first 2 ensiling days of maximum ethanol quantities (65–68%) was less than the initial CO₂ and gas quantity formation (71–79%). Schmidt et al. [13] measured unrestrained CO₂ production in the first 2 ensiling days after treatment with natamycin. This food and feed additive should inhibit yeast spoilage [29] but not enterobacterial activity. Thus, in this study, a significant portion of the initial gas formation seemed to be metabolised by microorganisms such as enterobacteria or LAB rather than by yeasts. Enterobacteria counts were not quantified, but the literature reports significant counts and activity in the initial days [8]. A more detailed analysis of the microbiological community and fermentation products, such as formic acid, 1-butanol, and hydrogen gas, could provide additional information concerning CO₂ formation.

After the final pH values were reached, most of the CO₂ formed could be assigned to obligate heterofermentative LAB, such as Llb. buchneri. These bacteria convert LA to 1,2-propanediol, ethanol, AA, and CO₂ [6, 18]. This anaerobic fermentation, which also occurred in the late phase of the storage period, explains the increase in gas quantities (up to ensiling day 60 or 135) measured in barrels BIO3 and BIO4 in combination with the increase in 1,2-propanediol, ethanol, and AA concentrations. Thus, the application of heterofermentative LAB using biological SA can lead to increased CO₂ formation. However, the cumulative CO₂ quantities did not differ between CON and BIO despite increasing gas bag volumes in BIO3 and BIO4. This aspect may be attributed to the short fermentation period of 30 days for some barrels and to CO₂ degradation or fixation pathways (Sect. “Gas quantities decrease and DM mass increases”). To the best of the author's knowledge, scientific research on gas formation during the fermentation process of maize silage treated with chemical SA is lacking.

After 80 ensiling days, CON (425 g_DM kg_FM) formed ≈ 7.4 (L gas) kg_DM^–1. A preliminary trial by our working group (unpublished data) determined cumulative gas quantities of 10.0 L kg_DM^–1 for untreated maize silage (400 g_DM kg_FM) after 83 days of storage. Daniel et al. [5] reported gas quantities of ≈ 13.0 L kg_DM^–1 for untreated maize silage after 83 days (380 g_DM kg_FM). Therefore, gas quantity formation seems to increase with decreasing DM. These findings align with the observations of Gomes et al. [38], who reported decreased gas formation with increased wilting intensity in oat silage. Nevertheless, while high DM seem to decrease CO₂ emissions during anaerobic storage, higher emissions are possible during the aerobic phases after silo closure or in the feed-out phase (Part B) [7]. Therefore, managing the DM seems to be a compromise but is important for minimising DM losses and gas formation.

Although other gases remain in the zero-pressure system, CH₄ concentrations (maximum of 6.8–8.0 ppm) exceed the values in grass and lucerne silage (maximum of 4.6–5.8 ppm) [2]. CH₄ concentrations are on the level of Schmidt et al. [13], who measured 7 ppm methane in maize silo emissions on ensiling days 5 and 15. After oxygen depletion in the whole gas volume or in some areas of silage [2, 20, 85, 86], enterobacteria can provide free hydrogen (H₂) during formate degradation in the first hours of the (partially) anaerobic phase [1, 2]. This H₂ can be used partly by archaea for methanogenesis. This process can stop if the pH is ≤ 6.8, a typical threshold for common archaea in silage-based methanogenesis in biogas plants [89]. However, the highest CH₄ concentrations were measured at ensiling days 5.5–6.0, and laboratory analysis indicated that the pH was ≤ 5.9 beginning on day 0. A literature review indicated that methanogenesis occurs in acidic environments by acetoclastic methanogens, using acetate as a substrate, or via hydrogenotrophic methanogenesis by specific archaea, using H₂ as an electron donor [90]. A decrease in the pH to 3.8 promotes the latter pathway based on the pH tolerance of various microorganisms, e.g. Methanobacterium and Methanothermobacter [90,91,92,93]. However, to the author's knowledge, little is known about the presence of these microorganisms in maize silage. In particular, the sensitivity of various microorganisms to changes in pH [90, 94] requires additional research. Other biochemical pathways may also be involved [2]. Modified gas analysis, which involves measuring H₂ concentrations within the silos, could supply additional information regarding methanogenesis and enterobacterial activity.

The N₂O concentrations (maximum of 24–38 ppm) are less than the maximum values mentioned by Schmithausen et al. [2] (686–1118 ppm), Zhao et al. [33] (1806–1836 ppm), and Wang and Burris [34] (10,000–43,500 ppm). On the other hand, Schmidt et al. [13] reported concentrations of 1 ppb in the cumulative emissions of maize silage. The variety of values can be partly explained by the varying experimental set-ups, sampling points, or technical limitations [33]. N₂O formation seems to be part of denitrification [2, 95, 96]. Franco [97] reported that silage treated with potassium sorbate emitted greater amounts of nitrogen oxides (NO_x) during the initial fermentation period. The same applies to N₂O in the CHE treatment. Franco [97] assumed that potassium sorbate has an inhibitory effect on denitrificating microorganisms and limits the degradation of nitrate and nitrite to N₂. However, all three treatments showed no signs of extended proteolytic breakdown, which could lead to additional N₂O formation.

The LAB inoculation in BIO was able to reduce organic acids in CHE increased CO₂eq emission quantities. Thus, the use of SA affects the formation of climate-relevant emissions, but the relevance of these differences has to be contextualised (Part B). All treatments show higher N₂O emission quantities than CH₄. This, in combination with the different GWPs, shows that the climate relevance of N₂O emissions is more significant than that of CH₄. Any mitigation measures should therefore start with a reduction in N₂O emissions. To minimise DM losses and CO₂ formation, the aim should be to achieve rapid oxygen exclusion through low porosity and rapid covering of the silo as well as a rapid pH reduction to minimise the activity of yeasts and enterobacteria.

Formation and emissions of ethanol and ethyl acetate

The course of ethanol concentrations within the material aligns with values stated by Weiß et al. [20], who reported a rapid increase within the first two days of the ensiling period and a subsequent slow increase until ensiling day 30. Weiß et al. [98] support these data, stating that high proportions of maximum ethanol concentrations are formed in the early fermentation phase. The formation of ethanol starts in silos with temporal–spatial variation considering oxygen depletion [20] and pH [32]. Weiß et al. [20] reported a very similar pattern in yeast counts and ethanol concentrations in promptly or delayed sealed maize silage and a strong linear correlation between DM losses and ethanol concentrations. However, the role of enterobacteria in ethanol fermentation should also be considered. The initial DM losses and gas formation, possibly unrestrained by antifungal SA (Sect. “Formation and emissions of greenhouse gases”), led to the assumption that the production of ethanol was also based on the metabolism of enterobacteria. However, further research is necessary to clarify this phenomenon in various silages.

Concerning the continuous increase in BIO, Weiß et al. [47] measured heterogeneous ethanol concentrations in farm-scale silos, but the concentrations decreased with the addition of heterofermentative LAB. These findings align with those of Arriola et al. [6]. The laboratory-scale trial of Hafner et al. [19] measured higher ethanol and EA concentrations in heterofermentative LAB-treated maize silage than in CON silage (p < 0.01). The following assumptions for ethanol formation apply: (a) Ethanol can be produced in a pathway parallel to the degradation of LA to 1,2-propanediol or AA by Llb. buchneri [18]. Therefore, increasing ethanol quantities are an indicator of the activity of heterofermentative LAB [18, 19, 32]. In addition to the formation of AA, heterofermentative LAB can produce ethanol during the anaerobic fermentation of glucose [79]. As shown, the counts of LAB and AA and ethanol concentrations are greater in BIO; (b) yeasts are a typical producer of ethanol in silage production [16, 20]. The increasing AA concentrations and decreasing yeast counts in the glass jars exclude this formation pathway in the BIO treatment. Furthermore, the excellent silage face scoring of the BIO barrels excludes a difference between the glass jars and the silo barrels. The formation of ethanol by yeasts may be from higher priority in practical silos than in laboratory silos which are characterised by rapid compaction and sealing. Based on the literature review, assumption a) seems the most reasonable in this trial. Thus, heterofermentative LAB may be beneficial for ethanol reduction in practical silos by suppressing yeast metabolism but contradictory in laboratory silos.

The only article considering the measurements of ethanol gas formation during silage-related, broth-based anaerobic fermentation was by Shan et al. [32]. Their results indicate rapid ethanol production within the first 20–30 h with reduced gas formation quantities for the treatment supplemented with Llb. buchneri and Lpb. plantarum compared to that of Llb. buchneri alone. However, the trial presented here indicated no difference between CON and BIO during the first days of the ensiling process. Ethanol gas formation stopped at ensiling day 4.67, at which point no further increase in ethanol gas quantity occurred. The ethanol concentrations in the silage material further increased. At present, why gas formation stopped is unclear, but pH may be a factor. In the trials of Shan et al. [32], ethanol gas formation stopped rapidly at pH 3.7–4.0.

Furthermore, previous research has shown that the gas quantities in silos are based on the VOC concentrations of the material considering Henry’s law [16, 31, 32]. Moreover, ethanol gas formation may depend on ambient and silage temperatures during anaerobic fermentation, aligning with emission patterns in the feed-out phase [21, 36]. This may be true, but correlations between ethanol and EA concentrations and between ethanol and gas quantities are small or nonexistent. Therefore, measurements of ethanol or ethyl acetate concentrations within the gas phase of farm-scale silos may provide only limited information for assessing material concentrations. Consequently, this procedure would not be an improvement over a standard laboratory analysis.

Hafner et al. [31] established a calculation model for VOC emissions during the fermentation process and calculated losses of 0.2%–1.0% of the present ethanol. For this purpose, they assumed uniform ethanol concentrations for the ensiling process, which should be constant at the final level. Based on the results presented here, the ratios of treatments CON, BIO, and CHE were 0.2%, 0.1%, and 0.1%, respectively. The percentages for EA are 1.1% for CON, 0.8% for BIO, and 1.1% for CHE. The experimental data are lower than the calculated data, which aligns with the difference between assumed continuous ethanol concentrations and the actual pattern measured in the material [16].

The course of EA concentrations is, in principle, in line with the previous literature showing an increase in the first 16 [98] to 30 ensiling days [20], followed by a subsequent decrease. However, local minima at ensiling day 14 have not yet been reported.

The formation of EA by yeasts was detected under various metabolic and processing conditions [99, 100]. In silage, EA can be produced through several biochemical pathways as previously described by Weiß et al. [20]. EA formation within the first days after silo closure can be mainly attributed to the enzymatic yeast activity of esterase, hemiacetal dehydrogenase, and alcohol acetyltransferase. Thus, EA may be formed by dehydration of acetate and ethanol, from acetyl-CoA and ethanol or from reduced hemiacetals like ethanol and aldehyde, respectively [20, 101]. In detail, the metabolism of some yeast genera shows a substantial increase in ethanol and especially EA formation shortly after oxygen depletion [102]. This could explain the rapid increase in material concentrations and gaseous emissions of EA during the initial phase of the ensiling process. Unfortunately, changing the gas bags at ensiling hour 36 could have affected the gas concentration measurements. In addition, the enzymatic activity of esterase in Acetobacter sp. [103] could also be of minor relevance in this first period.

Due to microbial analysis and silage face scoring, the activity of yeasts can be neglected during the late increase in EA in the BIO treatment. At this point, other possible explanations could be the metabolism of LAB. Previous research has focused on the activity of the ferulate-esterase formed by Llb. buchneri to affect silage digestibility by ruminants [104]. Furthermore, naturally occurring LAB on plants cause EA formation [105]. Therefore, LAB can synthesise esters and probably also EA in silage. Thus, the increase in the EA concentration in BIO may be attributed to the activity of (heterofermentative) LAB. This aligns with the increased EA concentrations in Llb. buchneri-treated silage treatments observed by Gomes et al. [38]. Moreover, combining lactic acid and acetic acid bacteria can enhance EA formation [106]. However, during the ongoing fermentation phase, enzyme activity seems to decrease with increasing concentrations of acids and decreasing pH [20, 107]. Thus, further research to determine a more precise bacterial species and strain specification is required concerning EA synthesis.

According to the authors' information, the formation of gaseous EA emissions during the anaerobic fermentation of silage has not yet been reported. Therefore, this study presents novel information concerning potential VOC emissions from silage production.

Overall, the VOC concentrations in all silage treatment materials were lower than those in the former literature [20, 39, 47, 98, 108, 109]. Reviewing these studies leads to the assumption that ethanol (2.6–33.6 g kg_DM) and EA (0.02–1.60 g kg_DM) concentrations vary widely and EA concentrations decrease with increasing DM in maize silages. However, the literature data show high variation in trial set-up, plant variety, and storage length and conditions. Weiß et al. [110, 111] reported that VOC concentrations increase with strict anaerobic conditions, lower ambient air storage temperatures, and high packing density. For legume silage, VOC concentrations decrease with increasing DM [111]. Therefore, low CO₂ concentrations within zero-pressure barrels, constant indoor storage temperatures, low packing density, and high DM could lead to reduced ethanol and subsequent EA formation compared to farm-scale silos. Whether practice silos with higher DM also form less EA cannot be generalised. The multifactorial influences, such as the achieved compaction, O₂ supply, and resulting microbial activity, can have an impact.

Examination of the methodological procedure

The packing density of 150.60 ± 0.08 kg_DM m^–3 was significantly below the theoretical recommendations for farm-scale silos of approximately 346 kg_DM m^–3 for fresh material with 42.5% DM [112, 113]. However, these theoretical target values cannot be achieved in practice silos, where a density of around 260–300 kg_DM m^–3 should be aimed for. The calculated porosity of the material was, therefore, very high. The trial set-up involved the combination of high DM and low packing density to provoke heating of the silage during the aerobic storage period (Article Part B) according to the German procedure for assessing the ASTA of silages [114]. With this, DM losses seem to be high compared to former laboratory-scale studies but at the level of well-managed commercial silos [7, 10, 65, 115,116,117]. Therefore, the correlation between DM losses and CO₂ emissions could also apply to farm-scale silos, since the basics of metabolic processes differ only slightly between systems [14]. Nevertheless, diverse microbiota [75] should be considered, and farm-scale silos with similar properties (high DM and low packing density) are likely to suffer greater DM losses. This aspect is especially relevant for increased oxygen exposure due to small leakages or during the aerobic feed-out phase (Part B).

The set-up of barrels and gas bags is a variation of previous trials in silage [2, 30] and differs from the approaches of Brazilian researchers [4, 26, 38]. The importance of short measurement intervals has been discussed previously [2]. Barrel silos are established for laboratory-scale investigations and, at the same time, have larger masses than glass jars. This fact was utilised to achieve larger emission quantities. As a result, these corresponded to the measuring ranges of gas measurement technology, especially during the aerobic emission measurement periods (Part B). The addition of gas bags enables the quantification of GHG and VOC emissions. If no bags had been used, the gas quantities would have been released through the lid seal, and it would only have been possible to determine gas concentrations in the headspace and not emission quantities [2]. However, the storage of all gases, especially residual N₂ and formed CO₂, could affect the emission behaviour or gas fixation of silage due to varying volatilisation and diffusion rates based on Fick’s and Henry’s laws [15, 41]. Furthermore, little is known about the dispersion of gases in silos. The first gas bags had lower gas concentrations than did the headspace; the second gas bags had lower gas concentrations. A significant portion of the residual N₂ was transferred into the first gas bag due to the formation of CO₂ in the silage material. Thus, further trials should use larger gas bags (≥ 15 L kg_DM^–1) to avoid methodological implications. Nevertheless, whether gases are subject to stratification in barrels and gas bags is unclear. Therefore, some variation between the calculated and actual gas quantities can apply. Furthermore, the storage of formed gases within a zero-pressure silo system differs from that of commercial silos with outwards-directed mass flow. Thus, gas dynamics could vary between this trial and farm-scale silos.

The silage treatments showed an increase in DM mass at certain stages of the fermentation process. The weights of the barrel and glass jar silos were measured using scales with adequate measuring ranges. The method of DM analysis and correction was frequently used [2, 20, 39, 47, 50, 98] and is considered the best practice. Nevertheless, the DM at ensiling day 14 showed a local minimum. Therefore, the increase in DM could be based on methodological errors, but a literature review suggested that gas fixation may be possible (Sect. “Gas quantities decrease and DM mass increases”). However, further studies are needed to verify these hypotheses in this new field of research [42].

In this trial, barrel silos were employed in Experiment A1 and glass jar silos in A2. The variation in silo geometry and storage conditions may influence the fermentation process and silage characteristics. For instance, glass jar silos were packed to provide the same silage porosity as the barrel silos, but lacked a head or floor space. Consequently, the residual oxygen supply within each glass jar was likely to be smaller affecting the microbial activity in this phase. However, this compromise was unavoidable due to the overarching trial set-up. The silage material in the barrels (Experiment A1) was to be stored untouched. This also meant that the quality of the silage in experiment A1 could only be qualitatively assessed once. A laboratory analysis of the material would have been desirable, but was not possible. However, the specific silo characteristics were needed to conduct emission measurements during fermentation and feed-out (Part B). Furthermore, the trial was conducted in a way to minimise differences in external effects such as ambient air temperature.

For statistical analysis, single measurement time points were merged into measurement intervals. This approach can lead to minor deviations from actual and stated time points for gas formation. However, the differences in gas formation patterns among the various silages were obvious. Furthermore, minor differences in the methodological procedure used by the two laboratories for microbial analysis may have affected the values. However, the course of microbial counts seems reasonable based on changes in chemical composition and gas formation.

The use of these measurement technologies limits the number of silos used during subsequent aerobic emission measurement periods (Article Part B). More precisely, the Multipoint Sampler and Doser (INNOVA 1303, LumaSense Technologies SA, Ballerup, Denmark) used in the aerobic emissions measurement periods for the PAS technology has 6 measuring points, which limited the number of barrels (n = 2; Article Part B). Due to the extensive manual sampling and dilution, the experimental design was limited to a maximum of 2 aerobic measurement periods—i.e. a total of 12 barrels. This affects the fermentation trial and restricts the statistical significance of the results shown. In particular, the significant deviation of the BIO3 and BIO4 barrels highlights the weakness of the small sample size. Nevertheless, new essential findings in silage research can be presented due to the extensive and detailed execution of the experiments and subsequent literature review. This case study should be considered an additional step in silage emission research and should be supplemented by further studies.

Gas quantities decrease and DM mass increases

CON and CHE showed a substantial increase in DM and a decrease in gas quantities between ensiling days 14 and 30, and BIO between days 14 and 135. On this basis, the following hypotheses were derived to explain the measured phenomena.

When setting up the experiment, care was taken to use barrel, hose, and gas bag materials impermeable to relevant gases. In addition, all the hose connectors were checked regularly. Knicky et al. [30] and Schmithausen et al. [2] used similar approaches in silage research. Other trial set-ups [26, 29] used silicone hoses known to be CO₂ permeable, leading to losses.

Long-term gas losses cannot be excluded in this trial. The decreasing quantities of gases in the zero-pressure system could result from tiny leakages, e.g. inside the seal of hose connectors. Similar loss rates for all silos and treatments are assumed if these losses apply, but it is unclear if they vary for the individual gases. For example, a greater decrease in N₂O could result from a greater permeability of this gas through the materials of the set-up. Furthermore, gas loss via permeability through the material could also be applied to other laboratory or commercial silos [7]. However, the constant gas and CO₂ quantities in the late phase of the storage period make leakage unlikely. Rather, the regressive course of CO₂ quantities strengthens the suspicion of controlled—and, from a particular point, time-saturated—gas fixation in contrast to uncontrolled, linear gas escape due to leakage [42].

Although an increase in DM during extended anaerobic storage periods is not mandatory, this phenomenon has been reported previously [6, 14, 118]. Savoie et al. [118] assumed errors in methodology and analysis (Sect. “Examination of the methodological procedure”) or in the formation of effluent. The effluent was not detectable in this trial due to the set-up but seemed unlikely due to high DM [119].

These decreasing gas volumes contradict the findings of Daniel et al. [4, 5], who described degressive courses. The Brazilian silos indicate positive pressure, and gases were released after pressure measurements. The set-up used here is based on the zero-pressure principle. These differences could depend on the various solubilities of CO₂ in the liquid phase, considering the varying DM. Furthermore, the concentrations in the gas space of the silos affect the solubility and volatilisation of substances, but this methodology was necessary to quantify the decrease in gas quantities. Furthermore, BIO had a pH value of 4.3, whereas it was 3.8 for the treatment ‘ComboHigh' (Llb. buchneri and Lpb. plantarum) in Daniel et al. [5]. This could increase the solubility of CO₂ in the form of carbonic acid in the liquid phase [120].

Other studies have reported negative pressure in silos after the completion of the main fermentation phase [2, 26, 44]. Krueger et al. [14] discussed the variance between the expected and calculated CO₂ emissions of maize silage. Furthermore, CO₂ absorption by maize silage exceeded the CO₂ solution potential in the liquid phase [26, 42, 43]. In those studies, this phenomenon can be attributed to the CO₂ permeability of the silicone hoses used in the trial. However, the decrease in the CO₂ quantity also exceeded the solution potential in the trial presented here.

Gas fixation in silage was described in recent reviews [42, 84], but these findings may be affected by erroneous methodology. Currently, two pathways of CO₂ fixation in silage are a matter of conjecture. First, Brazilian researchers assumed microbiological CO₂ reduction to acetate by acetogenic bacteria (Wood–Ljungdahl pathway) [26]. In this pathway, CO_2, in combination with ‘free’ protons and electrons, is the substrate for the formation of acetate and water [121]. Several species are known for this reductive acetyl-coenzyme A pathway [121]. However, to the authors’ knowledge, only one species (Sporomusa ovata) has been detected in silage [122, 123]. Additional species could be present in the epiphytical microbial community. The application of this pathway could explain the differences in AA concentrations between laboratory-scale and farm-scale silages [124, 125]. The improved gastightness of laboratory-scale silos could lead to higher quantities of CO₂ remaining inside for additional acetate provision. These findings align with those of Vigne et al. [45] reported in Schmidt and Vigne [42]. In particular, this could explain the increase in AA concentrations during the late phase of the ensiling process, while LAB counts remain at similar levels [125]. For commercial silos, the improved gastightness of tower silos could affect this pathway compared to bunker silos.

In the second pathway, CO₂ may be chemically fixed in the liquid phase during the ensiling process [41], e.g. in the form of solved CO₂, bicarbonate (HCO₃⁻), or carbonic acid (H₂CO₃). At pH 4, no bicarbonate should actually occur, and the most common form should be protonated carbonic acid (H₃CO₃⁺), which is in equilibrium with CO₂. Furthermore, former titration trials indicated no presence of carbonates or bicarbonates in the press juice of silage samples [126, 127]. Nevertheless, microbial activity could affect the equilibrium. This includes transmembrane CO₂ transport by LAB using aquaglyceroporin [128]. The transport of CO₂ into the cytosol and its hydration to H₂CO₃ and bicarbonates are assumed to regulate the pH of the cell [14]. Furthermore, LAB can convert CO₂ to H₂CO₃ and HCO₃⁻ through carbonic anhydrase activity [129]. Subsequently, bicarbonate can be used for the formation of pyrimidines and arginines [129, 130]. This pathway can be auxotrophic at low CO₂ concentrations in one-third of Lpb. plantarum-related strains [131]. Other LAB exhibit phosphoenolpyruvate carboxylase activity and upregulated bioenergetic metabolism in a CO₂-enriched atmosphere [132]. The relevance of this pathway has yet to be proven for silage, and the possible impacts of chemical or physical pathways, e.g. buffer systems or pH changes, have yet to be assessed.

CH₄ quantities substantially decreased between ensiling days 5 and 30. Little is known about the underlying reasons for this pattern, but two assumptions apply: (a) methane leakage or relocation within the zero-pressure system, or (b) special acid-tolerant, anaerobic, methane-oxidising archaea metabolise methane to CO₂ in ‘reverse methanogenesis’ [133, 134]. Concerning a), low-density CH₄ could have been shifted earlier and in larger parts into the gas bags. After formation, the low density leads to an accumulation in the headspace of the barrel from where it is subsequently driven into the bag. This would influence the calculation of the gas quantities in the zero-pressure system, as this is based on the gas concentration in the headspace (Sect. “Examination of the methodological procedure”). At this point, it is unknown whether the pathways of assumption b) are relevant for anaerobic fermentation of silage.

After the initial peaks, N₂O exhibited a greater decrease in quantity than did CO₂ and CH₄ for all the treatments. This pattern could be based on (a) N₂O gas relocation or leakage at higher rates than CO₂ and CH₄, (b) denitrification of N₂O to N₂ in the silos, or (c) solubility of N₂O in the liquid phase of silage material [95, 135]. Pathway a) was already discussed. Pathway b) is also possible, since N₂O gas concentrations peak at ensiling hour 44 at pH values approximately or slightly above 4.6–4.7, and denitrification was shown to possibly occur under these acidic conditions [136]. Theoretically, denitrification leads to N₂ formation. However, in a former trial by Wang and Burris [34], a decrease in N₂O concentrations did not lead to detectable increases in N₂ concentration. However, denitrification of the small quantities of N₂O in the present study did not impact N₂ or other gas concentrations. Concerning pathway c), N₂O dissolves in the liquid phase spontaneously and in parallel with the increase in gas concentration. The increase in N₂O emissions during aerobic emission measurements (Article Part B) underlines this fact. Thus, a mix of pathways b) and c) seems most reasonable.

Little is known about ethanol degradation during the ongoing ensiling process. Some LAB and acetic acid bacteria are reported to metabolise ethanol to organic acids, e.g. AA [137, 138]. To the authors’ knowledge, no such metabolic pathways have been detected in silage. The same applies to EA degradation, considering the lower rates of decrease.

Krueger et al. [14] mentioned that silage could serve as a sink and thus absorb CO₂. Schmidt and Vigne [42] expressed similar hopes, stating the open question: ‘Can silage absorb more carbon than what it emits during fermentation?’. The results presented here clearly show that although silage can absorb gases during anaerobic storage, the net balance of the fermentation process is still negative. Consequently, reducing avoidable losses in the early phases of ensiling seems more of a matter.

Implications for ensiling management and ensiling research

Ongoing climate change can make optimal harvesting and silage management difficult for farmers worldwide. Sudden rainfall can lead to short wilting periods [2]. On the other hand, draughts increase temperatures and solar radiation, leading to high forage DM. Improving ensiling management—i.e. by shortening chopping lengths, increasing packing density, and rapidly closing the silo—or applying SA can minimise losses during anaerobic storage. Nevertheless, other types of forage conservation, e.g. hay making, are also related to CO₂ emissions and should be compared to the silage process chain.

Further research should include O₂ analysis at narrow intervals during the first days of the ensiling process to determine the start of anaerobic conditions. For this purpose, gas sampling at different positions within the silos and more detailed specifications for microbiological analysis are recommended in future studies. With this, further conclusions can be drawn regarding the temporal–spatial variation in microbial activity. Calculating the molar masses in the zero-pressure system could also help.

To the authors' knowledge, this research article is the first to quantify the formation and emission of gaseous ethanol and EA during the ensiling process of (maize) silage. Currently, at least 46 different VOCs have been identified in silage [16]. However, ethanol can be used as an indicator substance for the VOC formation pattern since approximately 80% of silage VOC emissions are alcohol [15, 16, 139], with ethanol as the primary contributor (> 70% of alcohol emissions). Furthermore, ethanol is essential for forming esters, e.g. ethyl lactate or EA [20]. EA is reported to have antibacterial and antifungal properties. This phenomenon is primarily important for industrial processes and products, indicating the inhibitory effects of EA on yeast growth or specific bacteria [22, 23]. Thus, higher concentrations within BIO barrels could be a factor in suppressing adverse microorganisms and improving the ASTA of the BIO treatment.

VOC can lead to ground-level ozone formation and affect air pollution and climate change [15, 17]. However, the impacts are difficult to quantify, but emissions are relevant for ecological systems and can lead to substantial DM losses. This process is especially important for transforming photosynthetically bonded CO₂ to carbohydrates and subsequent emissions of methane, considering the higher GWP or high-energy and environment-relevant VOC.

Measurements of GHG and VOC gas concentrations within silos can provide important information for assessing microbial activity and ensiling quality. For instance, a previous study shows that increasing CH₄ concentrations in the silo can be used as an indicator for clostridial activity and quality decrease [2]. Furthermore, the formation of CO₂ or VOC gases may be used to determine the efficiency of microbial metabolism in silage fermentation [32]. Therefore, this approach can complement the preliminary methodology used in silage research. If applied in reverse, easily assessed parameters such as DM losses could be used to derive accurate models to calculate silage emissions and effects on the CF of silage production (Part B). VOC concentrations in the material have been part of the silage evaluation criteria for a long time, e.g. in the methods used in this trial. However, it is still unclear whether VOC gases in silos can be used to assess fermentation quality. Data show that correlations between concentrations in the material and the quantities in the gas phase are limited.

Nevertheless, even if the basic principles of anaerobic fermentation in laboratory and practical silos are similar, the transferability of the test results to the field requires special attention. The larger dimensions, extended oxygen provision during silo filling and compaction, the greater risk of air infiltration through small leakages, and the fundamental difference in gas exhaustion compared to gas storage in zero-pressure gas bags can influence gas dynamics of practical silos. Thus, future tests should also be carried out on commercial silos. The integration of experiments conducted at both scales could potentially yield a synergistic enhancement in the quality of the findings. Furthermore, parameters of ensiling management, e.g. cutting length, packing density, time period until silo sealing, or external factors, e.g. ambient temperature, may be relevant factors.

Conclusions

This study effectively measured the gaseous emissions of GHG, ethanol, and ethyl acetate during the fermentation process of maize silage. These findings aid in the evaluation of the distinct gas formation patterns of laboratory-scale silos during the various stages of the ensiling process. Notably, CO₂ formation is linked to DM loss under aerobic conditions and before the pH drops to 3.9–4.3. The BIO treatment, which was supplemented with Llb. buchneri and Lpb. plantarum, exhibited prolonged CO₂ and ethyl acetate formation. For the first time, records of ethanol and ethyl acetate gas formation during the fermentation process were made, validating earlier emission models. However, a few relationships were found between the material's VOC levels and gaseous emissions. BIO emitted less methane and nitrous oxide than the untreated control (CON). The CHE treatment, supplemented with sodium benzoate and potassium sorbate, emitted more nitrous oxide than CON (p < 0.05). CHE had the highest climate-relevant CO₂eq emission quantities, BIO the lowest (p < 0.05). After a certain point, the data indicate decreasing gas quantities, particularly CO₂. This suggests a shift from formation to fixation, which coincides with an increase in DM. The results of these experiments are worthy of further research. Nevertheless, gas formation clearly exceeds gas fixation during anaerobic storage of silage.. Further research should assess parameters that may influence silage gas dynamics, e.g. plant species, forage DM and maturity, epiphytic microbial counts, silo type, including laboratory- and farm-scale silos, management decisions such as chopping length or compaction effort, and malfermentation or external factors such as ambient temperature.