AGRICULTURAL AND FOOD SCIENCE Agricultural and Food Science (2022) 31: 241–259 241 https://doi.org/10.23986/afsci.115968 Testing the aerobic stability of silage-based mixed rations using a multiple indicator approach Siriwan D. Martens1, Annette Jilg2, Mariana Schneider3, Barbara Misthilger3, Susanne Ohl4 and Olaf Steinhöfel1 1Saxon State Office for Environment, Agriculture and Geology, Department of Animal Husbandry, Am Park 3, 04886 Köllitsch, Germany 2Agricultural Centre for cattle production, grassland management, dairy food, wildlife and fisheries Baden-Württemberg (LAZBW), Atzenberger Weg 99, 88326 Aulendorf, Germany 3Bavarian State Research Center for Agriculture, Institute for Animal Nutrition and Feed Management, Prof.-Dürrwaechter-Platz 3, 85586 Poing, Germany 4Chamber of Agriculture Schleswig-Holstein, Research and Training Centre Futterkamp, Gutshof, 24327 Blekendorf, Germany e-mail: siriwan.martens@smekul.sachsen.de Total Mixed Rations (TMR) including silages are commonly fed to cattle in many parts of the world, but they tend to deteriorate on exposure to air, especially in the warm season. The aim of the study was to develop an easy to im- plement protocol in order to compare the aerobic stability of different feed mixtures for lactating dairy cattle and to test the potential of TMR stabilizing products to delay spoilage. The experimental conditions were standardized to 25 °C ambient, and the sample dry matter was adjusted to 400 g kg-1 to challenge shelf life. Temperature rise in an insulated vessel, which is frequently used when testing silages, was shown to be one promising indicator of spoil- age activity. Furthermore, determination of pH, scores for visual occurrence of yeasts and moulds and condensation from 0–4 were successfully applied at the end of the 72 h period. A dosage of 4.5 l propionic acid t-1 proved suit- able as a positive control when compared to 1.5 and 3.0 l t-1. The commercial products tested at the recommend- ed dose had a similar efficacy. The stabilizing effect depended principally on the original hygienic condition of the ration. Visual evaluation plus the recording of pH, in addition to continuous temperature measurement facilitated this appraisal. Early signs of spoilage, in particular, can only be assessed by visual appraisal. Thus, an evaluation of a combination of indicators of aerobic stability is recommended. Key words: TMR, aerobic deterioration, spoilage indicators, temperature rise, pH change, protocol Introduction Silages and total or partial mixed rations (TMR, PMR), which include ensiled forages, grains or by-products, are commonly fed to cattle on many farms around the world. These feeds are aerobically instable, i.e. they tend to deteriorate upon exposure to air. Spoilage of feed results in losses of dry matter (DM) and reduction in nutritional value and hygienic quality, which leads to depressed feed intake and ultimately impaired animal performance (Whitlock et al. 2000, Gerlach et al. 2014, Borreani et al. 2018). The investigation of the phenomenon of aerobic deterioration of silages increased in the 1970s, although in 1964, Beck and Gross (1964) had already recognized that yeasts might play an important role in the process. Many yeast species are able to oxidize lactic acid, which increases the pH of silage and leads to further changes brought about by other aerobic spoilage organisms (Mid- delhoven and Franzen 1986, Pahlow et al. 2003). Later, Spoelstra et al. (1988) confirmed the role of Acetobacter spp. especially in maize silages. Different factors accelerate the microbial spoilage process. These include elevated temperature, humidity, expo- sure time and microbial load (Seppälä et al. 2013, Borreani et al. 2018). Despite increasing knowledge, there are still many gaps in the understanding of e.g. microbial interactions and apparently contradictory observations, es- pecially in unconventional feeds such as chicory and potato by-products (Martens 2006, Avila and Carvalho 2020). A temperature rise of 2 or 3 K above ambient is internationally recognized as a reliable indicator for aerobic insta- bility of silages (O’Kiely 1993, Cherney and Cherney 2003). However, indicators of microbial spoilage do not always match observed temperature rises. Varying observations of no or very low temperature increase, despite other signs of spoilage, are summarized in Table A1 (Appendix). Therefore, not only temperature but other indicators of aerobic respiration should be monitored, for example, pH increase, visible fungal infestation and CO2 develop- ment where feasible (Shan et al. 2021a). Received 7 April 2022 / Accepted 24 November 2022 The Scientific Agricultural Society of Finland ©This is an open access article under the CC BY 4.012 S.D. Martens et al. 242 Currently, various products are available on the European market that aim to stabilize the moist mixed rations, which might be of interest in farm practice, especially when storing TMR in the warm season. Apart from the chem- ical composition of the offered products, their efficacy might also depend on other factors related to the feed and the environment, such as hygiene, humidity and temperature. The aim of the collaborative study of four Institutes of Applied Agricultural Research was to develop a simple scheme to challenge the aerobic stability of mixed rations on a laboratory scale, in order to test the efficacy of different TMR stabilizing agents and their dosage. For this purpose, in addition to temperature, other potential indicators that could easily be carried out without special laboratory equipment were recorded and evaluated. Initially, the minimum dosage of propionic acid was determined to serve as a positive control in subsequent trials. Materials and methods Pre-trials served to standardize the experimental conditions in terms of duration of the test, ambient tempera- ture, the sample moisture range, observed spoilage indicators (Martens and Steinhöfel 2019), and the method is presented here. Fresh TMR for lactating dairy cows yielding ≥ 30 kg milk per cow d-1 were obtained from the experimental farm stations of the four institutes involved in the study. All of the rations contained 600–650 g silage kg-1 DM (maize + grass) and included cereal grains (barley, wheat or rye), rapeseed meal and minerals as common feedstuffs in var- ying proportions (Tables 1 & 2). The DM content was determined at 105 °C in a drying oven over night. Based on this information, another fresh TMR portion was taken for the aerobic stability test on the following day. Hence, it was possible to calculate the volume of water needed to achieve the target DM of 350–400 g kg-1 in order to challenge aerobic deterioration (Rinne et al. 2018, Martens and Steinhöfel 2019). Table 1. Total Mixed Ration composition in Trials 1 and 2 (% of dry matter) Trial 1 Trial 2 Component Inst.1 Inst.2 Inst.3 Inst.4 Inst.1 Inst.2 Inst.3 Inst.4 Maize silage 34.4 24.2 31.2 29.6 34.4 28.5 27.6 29.6 Grass silage 26.0 38.5 34.0 33.2 26.0 11.5 30.6 33.2 Grass-clover silage 0 0 0 12.3 0 0 0 12.3 Crimped maize 0 0 0 7.09 0 5.11 0 0 Grain maize/sorghum 11.9 4.28 0 1.09 0 0 0 7.09 Barley/Wheat/Rye 3.17 7.13 0 2.18 11.9 7.74 4.51 1.09 Barley/Wheat straw 1.51 0 2.54 3.42 3.17 12.0 13.57 2.18 Lucerne hay 0 2.14 0 0 1.51 0 6.36 3.42 Pelleted pressed beet pulp 5.81 4.04 0 0.446 0 3.85 0 0 Rapeseed meal 6.34 7.13 12.6 7.48 5.81 3.91 0 0.446 Sunflower seed meal 7.84 0 0 0 6.34 9.67 11.34 7.48 Soybeans 0 0 0 2.33 7.84 0 0 0 Field peas & beans 0 0 0 0 0 0 0 2.33 Cattle salt 0 0 0.192 0.050 0 0 4.51 0 Glycerin 0 0.950 0 0 0 0.043 0.19 0.050 Mixture of cereal grains and minerals + trace elements 2.41 11.6 18.9 0.842 0 2.20 0 0 Urea 0.399 0 0.240 0 2.41 0.400 1.09 0.842 Vinasse from sugar beets 0.310 0 0.240 0 0.399 0 0.24 0 Inst. = institute Agricultural and Food Science (2022) 31: 241–259 243 Trial 1: Dose-response study with propionic acid Trial 1 was conducted in late October-early November 2020 (i.e. autumn) at the four localities in Northern (1), Eastern (1) and Southern (2) parts of Germany. Four different levels of propionic acid (n = 6): 0, 1.5, 3.0, 4.5 ml kg-1 original TMR were applied to determine the concentration, that would be reasonably certain to assure aero- bic stability and serve as a positive control. A fifth level (6.0 ml kg-1) was introduced in a second experimental run in Institute 2. They are referred to as Pr0.0, Pr1.5, Pr3.0, Pr4.5 and Pr6.0. Manufacturers of products using the same agent currently recommend 1–3 l t-1 to farmers. When the TMR had to be remoistened, the propionic acid (Art. No. 6026, Carl Roth GmbH + Co. KG, Karlsruhe, Germany) was mixed in the respective amount of tap water necessary to achieve at the most 400 g DM kg-1. This liquid was then added to the weighed amount of TMR and mixed thoroughly. When the original TMR already had the target DM (Institute 4), the propionic acid (undiluted) was sprayed on evenly with a pump sprayer to assure a homogeneous distribution. Institute 2 ran a second experiment with a slightly lower moisture content because of high instability observed in its first run, which had not allowed differentiation between the treatments. Trial 2: TMR stabilizing products and both positive and negative control Trial 2 took place between the end of November 2020 and the beginning of January 2021 at the four Institutes. In addition to the negative and positive control (no additive and 4.5 ml propionic acid kg-1 TMR; Pr0.0 and Pr4.5) there were four other treatments (n = 4): potassium sorbate (> 99 %) (0.4 g kg-1, powder, to be dissolved) (STAB1), Table 2. Chemical (g kg-1 DM) composition and microbial counts (log cfu g-1 fresh matter) of the Total Mixed Rations in Trials 1 and 2 Trial 1 Trial 2 Parameter Inst. 1 Inst. 2(1) Inst. 2(2) Inst. 3 Inst. 4 Inst. 1 Inst. 2 Inst. 3 Inst. 4 Original DM (g kg-1) 360 506 507 367 399 438 504 439 382 DM after remoistening (g kg-1) 361 359 390 347 399 393 369 362 363 Crude ash 71.0 66.0 71.0 75.0 83.0 72.0 60.0 74 70.0 Crude protein 146 154 166 141 134 157 157 164 148 Crude fibre 176 176 163 173 n.a. 174 151 187 209 Ether extract 29.0 30.0 34.0 31.0 n.a. 33.0 39.0 30.0 40 aNDFom 361 344 328 n.a. n.a. n.a. n.a. n.a. n.a. ADFom 194 204 196 n.a. n.a. n.a. n.a. n.a. 246 Sugar (water soluble) 57.0 62.0 61.0 59.0 63.0 53.0 20.0 40.0 26.0 Starch 242 218 215 236 158 238 249 204 212 Metabolizable Energy (MJ) 11.0 11.1 11.5 11.1 10.29 11.3 11.8 11.0 10.6 Net Energy Lactation (MJ) 6.70 6.70 7.10 6.80 6.16 6.90 7.27 6.70 6.40 pH 4.19 4.57 4.73 4.27 4.33 4.35 4.18 4.05 4.60 Lactic acid 41.0 62.6 62.9 24.3 42.0 44.7 n.a. 68.7 34.0 Acetic acid 13.1 7.12 6.99 7.65 9.00 14.9 n.a. 19.0 12.0 Propionic acid 0.19 0.13 0.13 n.d. n.d. <0.30 n.a. n.d. n.d. Butyric acid n.d. 0.46 n.d. 0.273 n.d. <0.55 n.a. 0.40 n.d. Ethanol 4.26 1.69 1.33 6.83 2.00 5.60 n.a. 18 2.00 1,2-Propanediol 5.30 0.76 0.93 3.01 n.a. 4.1 n.a. 5.90 n.a. Undissociated VFA 10.4 4.67 3.65 5.96 6.52 10.7 16.2 7.03 Undis. VFA/(sugar + LA) 0.11 0.04 0.03 0.07 0.06 0.11 0.15 0.12 NH3-N (g kg-1 N) 39.0 38.0 30.0 n.a. n.a. n.a. n.a. n.a. 1.00 Moulds <2.00 2.40 <3.00 <2.00 2.00 3.48 3.65 <1.0 2.30 Yeasts 6.20 8.37 6.42 4.81 3.90 4.67 6.20 3.00 4.83 Inst. = Institute; DM = dry matter; ± standard deviation; aNDFom = amylase treated neutral detergent fibre exclusive of residual ash; ADFom = acid detergent fibre exclusive of residual ash; (1) = first run; (2) = second run; Undissociated VFA = sum of undissociated acetic, propionic, butyric and valeric acids; Undis. VFA/(sugar + LA) = ratio of sum of undissociated VFA to sugar + lactic acid as suggested by Gomes et al. (2021); n.a. = not analyzed; n.d. = not detectable S.D. Martens et al. 244 a combination of propionic (~38 %) and formic acid (~34 %) (3.5 ml kg-1, liquid) (STAB2), a combination of sodium benzoate (300 g kg-1) and diacetate (20 g kg-1) (3.0 ml kg-1, liquid) (STAB3), and a combination of sodium formate and potassium sorbate (concentrations not disclosed by the manufacturer) (2.0 g kg-1, granulate) (STAB4). The additives were applied in the same proportions as in the first trial, i.e. mixed with the tap water used for remois- tening. The only exception was the granulate (STAB4), which was spread evenly on the TMR before remoistening, according to the manufacturer’s instructions. The parameters, which were determined by the treatments before the aerobic stability test were: DM, pH, yeast and mould numbers, crude ash, crude protein, ether extract, neutral detergent fibre treated with an amylase and exclusive of residual ash (aNDFom), acid detergent fibre exclusive of residual ash (ADFom), starch, water soluble carbohydrates, lactic, acetic, butyric and propionic acid, NH3-N of total N, ethanol (VD- LUFA 1976, VDLUFA 2012). The undissociated form of each volatile fatty acid in the TMR was calculated as (Henderson-Haselbalch equation) and multiplied by the respective acid concentration. The treated material was weighed into containers based on the model of System Völkenrode in replicates (around 250 g fresh matter (FM) each), and the weight recorded (Honig 1990). A sample container consisted of 20 cm lengths of a polyvinyl chloride (PVC) drainage pipe (PVC-KG-pipe DN110, Ø 11 cm), and closed with PVC caps (DN110) at the bottom and the top. A hole of Ø 10 mm was drilled in the centre of each cap to allow air to circulate. A layer of cotton gauze was placed on the bottom of the container to avoid losses by trickling. A diagram of this system is shown in Figure A1 (Appendix). When about one third of the container was filled a temperature data logger (TG 4080, Gemini Data Loggers Ltd, Chichester, UK), wrapped in a disposable polyethylene bag, was placed centri- cally in each tube. The loggers were programmed to record the temperature at half-hourly intervals. The filled con- tainers were then closed and placed in a polystyrene cylinder (EPS25, 6 cm wall thickness) which provided temper- ature insulation. Polystyrene covers (EPS25, 6 cm thick) were placed on the top and bottom of the cylinders. They had a V-shaped notch (5 mm deep, 1 cm wide) passing straight through the middle of the cover to permit air flow. The samples were stored at 25 (± 1) °C for 72 h. When taking them out of the polystyrene cylinder for evaluation, all PVC lids were firstly removed to evaluate the possible loss of condensed water, from both the cap and the inner walls, and the actual volume ranked between 0 and 4 (Table 3; Appendix Fig. A2). The vessels were then weighed with their contents, but without the lids. The complete samples were examined for visible signs of yeast and mould growth (Table 4; Appendix Figs. A3 & A4). DM and pH were also determined. FM and DM losses after the aerobic stability test (AST) were calculated as follows: FM loss (%) = 100 – net weight after AST/net weight before AST × 100, DM loss (%) = 100 – (net weight after AST × DM [%] after AST / 100)/(net weight before AST × DM [%] before AST / 100) × 100. Table 3. Humidity score after the aerobic stability test, with immediate evaluation after opening the lids (see also Appendix Fig. A2) Points Observation 0 No condensation 1 Light condensation at container wall 2 Wall continuously moist 3 Large drops at the wall (but water does not merge to a pool in the lid) 4 Wall and lid very wet Table 4. Visual evaluation of yeast and mould growth (see also Appendix Figs. A3 & A4) (Pahlow 1997, personal communication, DLG TestService GmbH 2018) Points Yeasts Moulds 0.0 None visible None visible 0.5 Traces of yeasts A very small area of mould 1.0 Yeasts ~10% ------- 1.5 More yeasts Some small mould agglomerations 2.0 Yeasts continuously present ------- 2.5 Yeasts continuously present More mould agglomerations 3.0 Heavy presence of yeasts Mould in every part 4.0 Completely deteriorated Completely deteriorated 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢. = 1 1 + 10(𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝−𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝) Agricultural and Food Science (2022) 31: 241–259 245 A demonstration of the whole procedure can be viewed in the following video: https://lsnq.de/tmraerobicstabilitytest Statistical analyses Both one and two-factorial designs were used. Variance analysis using the univariate and multivariate procedures was performed for both Trials, followed posthoc by a Tukey test. For Trial 1, the treatments were first evaluated within institute (per test run) in order to see in more detail how the different parameters would react depending on the underlying ration (Figs. 2–5): Yi = μ + CONCi + εi where i = 1, 2, …, 4 (0, 1.5, 3.0 or 4.5 ml propionic acid kg-1 TMR) and ε = error. Furthermore, they were evaluated across all test runs (Table 5): Yi = μ + CONCi + RUNj + CONC x RUNij + εij where i = 1, 2, …, 4 (0, 1.5, 3.0 or 4.5 ml propionic acid kg-1 TMR), j = 1, 2, …, 5 (test runs) and ε = error. For Trial 2, the treatments were evaluated across all institutes (Table 6): Yi = μ + STABi + INSTj + STAB x INSTij + εij where i = 1, 2, …, 6 (stabilizers), j = 1, 2, …, 4 (Institutes) and ε = error. In some cases, the temperature did not increase by 2 K above ambient within the limited time span of the AST. In those cases, it was decided to add 0.25 d (i.e. 6 h) to the maximum evaluated time. This was done to be able to include those samples in the statistical evaluation e.g. variance analysis and the post-hoc test, and to considering the practical significance on-farm of stocking ready mixed feeds. The software IBM® SPSS® Statistics (Version 19, SPSS, Inc., IBM Company©) was used. In Trial 2 the slope of linear regression lines (extended to axes) of the tem- perature curves was determined (SigmaPlot 12.5, Systat Software, Inc.). Results Trial 1 There was a highly significant effect of the test runs at the different Institutes on the evaluated parameters. The same was true for the treatment effect except for pH. In addition, the interaction of both factors significantly in- fluenced all parameters but FM losses and maximum temperature difference (Table 5). Pr0.0 represents Total Mixed Ration without additives, Pr1.5, 3.0, 4.5 with 1.5, 3.0 and 4.5 l propionic acid t-1 Fresh Matter; different superscript letters in the same row refer to significant differences among treatments (p < 0.05, Tukey test). TD, temperature difference to ambient; (0–4) refers to the score given in Table 3 and 4; SEM, stand- ard error of the mean. Tr Treatment, Run Test run. In three out of five test runs the temperature of the control Pr0.0 rose steeply within the first 24 h (Fig. 1). S.D. Martens et al. 246 *Excluding Institute 1 as this parameter was not determined here. Fig. 1. Temperature profiles for samples treated with increasing concentrations of propionic acid (0, 1.5, 3, 4.5, 6.0) ml kg-1 FM) within 72 h of aerobic stability testing. (a) Institute 1; (b) Institute 2(1); (c) Institute 2(2); (d) Institute 3 (samples not insulated); (e) Institute 4 Table 5. Shelf life of differently treated Total Mixed Ration stored for 72 h under aerobic conditions in Trial 1. Results of test runs at four different Institutes (n = 6 per Institute). p-value Measurements Pr0.0 Pr1.5 Pr3.0 Pr4.5 SEM Treatment Run Tr x Run h until ≥2 K TD 26.6d 32.3c 39.9b 47.7a 0.30 <0.001 <0.001 <0.001 h until ≥3 K TD 28.7d 36.3c 43.3b 49.4a 0.37 <0.001 <0.001 <0.001 Max. TD (K) 11.6a 10.9a 10.5a 9.16b 0.154 0.001 <0.001 0.075 h until max. TD 43.9c 52.6b 58.8a 57.4ab 0.71 <0.001 <0.001 <0.001 Final pH 5.79 5.75 5.67 5.61 0.053 0.563 <0.001 <0.001 Humidity (0–4) 3.55a 3.39a 2.57b 2.04c 0.046 <0.001 <0.001 <0.001 Yeasts (0–4) 2.81a 2.13b 1.71c 0.87d 0.041 <0.001 <0.001 <0.001 Moulds (0–4) 0.484a 0.161b 0.089b 0.093b 0.016 0.001 <0.001 <0.001 FM losses (%) 3.40a 3.06ab 2.70bc 2.42c 0.059 <0.001 <0.001 0.459 DM losses (%) 6.65a 7.12a 5.21ab 4.42b 0.265 0.001 <0.001* 0.004* a) e) d)c) b) Agricultural and Food Science (2022) 31: 241–259 247 Notably, in two of those (a, b), the temperature rise of the acid treatments was only slightly delayed, and there was almost no difference between Pr1.5 and Pr3.0. They had an initial total yeast count of 6.0 (a), 8.4 (b) and 6.4 (c) log10 cfu g-1 FM (Table 2). In the case of no thermal insulation (d) there was only a slight increase, which start- ed at the end of the first day, followed by a flattening of the curve. The initial yeast count was 4.8 log10 cfu g-1 FM. A test run (e) had shown the latest time for temperature to start to increase i.e. after 40 h, and there was almost no heating in Pr4.5 within 72 h. At the start of the experiment it had 4.2 log10 cfu yeasts g-1 FM. Whilst the maximum temperature rise was not different among treatments during the rapidly accelerating runs (Institutes 1 & 2), lowest temperature difference was observed for Pr4.5 in the two slower runs (Fig. 2a), which was reflected in the evaluation across the test runs (Table 5). However, time to achieve the maximum temperature was more delayed at Institutes 2, 3 and 4 (Fig. 2b). Time to achieve 2 or 3 K above ambient was clearly differentiated between the treatments (Fig. 2c & d, Table 5), and was most clearly demonstrated in the slowly increasing test runs. In all of the test runs there was an increase in pH (Fig. 3). However, only in Institute 1 and 4 was there a statistically significant effect of treatment, which contrasted at both the institutes. Where the rapid temperature increase had taken place (Institute 1), the highest pH was found for the highest acid addition Pr4.5. In contrast, at Institute 4, the pH was highest in the untreated control where there was only slow heating. This contrast led to the insignifi- cant effect of the treatment on the final pH when the statistical analysis was applied across the test runs, in con- trast to the interaction with the test run (Table 5). Fig. 2. (a) Maximum temperature difference; (b) hours until maximum temperature difference; (c) hours until 2 K difference above ambient; (d) hours until 3 K difference above ambient in the different treatments at the different institutions. The figures in the legend represent the applied dosage of propionic acid in L t-1 FM TMR. Error bars represent the standard deviation (SD). The p-value indicate the significance of the dosage effect within each test run. a) b) c) d) S.D. Martens et al. 248 The humidity was very high in the test runs with a rapid and steep temperature rise (Institutes 1 and 2) and only modest and small rises with the high acid treatment Pr4.5 (Fig. 4a, Table 5). Signs of yeast growth were more varied across the institutes; the highest acid treatment usually showing the least yeast growth (Fig. 4b, Table 5). Mould growth was only observed at Institute 1 (Table 5). Fig. 3. Initial and final pH values for the different treatments at different Institutes. The figures in the legend represent the applied dose of propionic acid in l t-1 FM Total Mixed Ration. Error bars represent the standard deviation (SD). The p-value indicates the significance of the dosage effect within each test run. Fig. 4. (a) Score of condensed water (Table 3) at the end of the aerobic stability test at the different Institutes. (b) Score of visible yeast growth (Table 4). The figures in the legend represent the applied dosage of propionic acid in l t-1 FM Total Mixed Ration. Error bars represent the standard deviation (SD). The p-value indicates the significance of the dosage effect within each test run. Fig. 5. (a) Fresh matter losses; (b) dry matter losses in the different treatments at the different institutions. The figures in the legend represent the applied dosage of propionic acid in l t-1 FM Total Mixed Ration. Error bars represent the standard deviation (SD). The p-value indicates the significance of the dosage effect within each test run. (n.a. = not analyzed) a) b) a) b) Agricultural and Food Science (2022) 31: 241–259 249 While FM losses were consistently higher in the control (Fig. 5a) the DM losses were less consistent (Fig. 5b, Table 5). Trial 2 As results from Trial 1 pointed to a reasonably safe effect of Pr4.5 on aerobic stability it was chosen as positive control for Trial 2. Here, the treatment had a significant effect on all parameters just as the interaction between treatment and Institute (Table 5). The same applied for the factor Institute per se except for time until maximum temperature difference. In contrast to the first trial, the temperature only started to rise on the second day irrespective of location (Figs. 6a & b). The average slope of the linear regression line is presented in Table 5. The TMR, which showed a steep increase in temperature in the control (Fig. 6b) had an initial yeast count of 6.2 log10 cfu g-1 FM while the one with more gradual heating (Fig. 6a) had a count of 4.7 log10 cfu g-1 FM. The one with a similar yeast count (4.8 log10 cfu g-1 FM) started to increase in temperature later, but rose sharply (Fig. 6d). In one of the test runs, the tempera- ture did not increase at all during the 72 h period (Fig. 6c), and yeasts counts were 3.2 log10 cfu g-1 FM (Table 2). The time until the temperature reached 2 or 3 K ≥ ambient was shortest for the negative control, on average for the four test runs (Table 6). The maximum temperature difference discriminated the treatments only in two out of four cases when evaluating per Institute (Fig. 7a). However, there was a highly significant effect of treatment and Institute in the overall evaluation (Table 6). When 2 K difference was used as the criterion, more samples could be evaluated properly as their temperature increased within the test time (Fig. 7c and d). Final pH, humidity, yeast score and losses were highest on average in the negative control (Fig. 9). However, in the post hoc test across in- stitutes, only Pr4.5 and STAB4 were significantly less humid, and in STAB4 also less yeasts appeared (Table 6). Al- though no temperature increase was apparent for Institute 3 (Fig. 6c) first signs of yeast growth were observed there (Fig. 8c). Moulds appeared only in one out of the four test runs and in two of the treatments, the negative control and STAB3 (Fig. 8d, Table 6). Fig. 6. Temperature development during a 72 h aerobic stability test in the second trial; (a) Institute 1, (b) Institute 2, (c) Institute 3, (d) Institute 4 a) b) c) d) S.D. Martens et al. 250 Fig. 7. (a) Maximum temperature difference to ambient; (b) hours until maximum temperature difference; (c) hours until 2 K ≥ ambient; (d) hours until 3 K ≥ ambient (Inst1–4 = Institute 1–4) Table 6. Shelf life of differently treated Total Mixed Ration stored for 72 h under aerobic conditions in Trial 2. Results of test runs at four different institutes (n = 4 per Institute). p-value Items Pr0.0 Pr4.5 STAB1 STAB2 STAB3 STAB4 SEM Treatment Inst. Tr x Inst. h until ≥2 K TD 46.6b 67.7a 66.2a 68.0a 64.3a 70.0a 1.63 <0.001 <0.001 <0.001 h until ≥3 K TD 48.3b 70.6a 69.6a 69.6a 66.2a 70.7a 1.51 <0.001 <0.001 <0.001 Max. TD [K] 10.5a 3.66b 4.78ab 4.71ab 5.65ab 3.57b 0.666 <0.001 <0.001 <0.001 h until max. TD 59.5b 67.2a 66.8ab 67.6a 68.7a 66.7ab 0.73 <0.001 0.958 <0.001 Final pH 5.82a 4.55b 4.63b 4.75b 4.72b 4.69b 0.067 <0.001 <0.001 <0.001 Humidity (0–4) 3.00a 1.12b 1.47ab 1.41ab 1.69ab 1.18b 0.161 <0.001 <0.001 <0.001 Yeasts (0–4) 2.16a 0.94ab 1.06ab 0.91ab 1.41ab 0.68b 0.124 <0.001 <0.001 <0.001 Moulds (0–4) 0.16 0.00 0.00 0.00 0.09 0.00 0.0180 0.004 <0.001 <0.001 FM losses (%) 2.75a 1.20b 1.46b 1.33b 1.49b 1.15b 0.068 <0.001 <0.001 <0.001 DM losses (%) 5.50a 1.25b 0.75b 0.85b 0.22b -0.63b 0.329 <0.001 <0.001 <0.001 Slope m 0.266 0.120 0.140 0.125 0.154 0.111 0.0350 Pr0.0 = Total Mixed Ration without additives; Pr4.5 = with 4.5 l propionic acid t-1 FM, STAB1–4 = the treatments with commercial TMR stabilizing products; different superscript letters in the same row refer to significant differences among treatments (p< 0.05, Tukey test). TD = temperature difference to ambient; (0–4) = the score given in Table 3 and 4; m = slope of linear regression line of temperature curve (total n = 3, Institute 3 was left out because no heating occurred); SEM = standard error of the mean; Inst. = Institute, Tr = treatment. a) b) c) d) Agricultural and Food Science (2022) 31: 241–259 251 Overall when evaluating all four test runs together, the time until 2 or 3 K temperature difference, pH and losses demonstrated the effectiveness of the tested products compared to the untreated control (Table 6). Fig. 8. (a) Final pH; (b) Humidity score; (c) Yeast score; (d) Mould score after visual evaluation (Inst1–4 = Institute 1–4) Fig. 9. (a) Fresh matter losses in %; (b) Dry matter losses in % (Inst1–4 = Institute 1–4) a) b) c) d) a) a) b) S.D. Martens et al. 252 Discussion Temperature increase and pH changes In general, undissociated volatile fatty acids (VFA) may exhibit fungistatic effects, and thus, Gomes et al. (2021) found that the higher the proportion of those acids in grass silage compared to the sum of soluble carbohydrates and lactic acid, the more stable the pH was on exposure to air. However, in an untreated TMR for high yielding dairy cows the amount of VFA is negligible compared to concentration of the soluble carbohydrates (Table 2). Thus, other factors may determine more clearly the course of aerobic deterioration in this kind of feed. In terms of temperature, the results presented for Trial 1 showed higher increases occurring earlier than for Trial 2. This was reflected by the yeast counts, which could be a result of both a new batch of silage from another silo (at least at Institute 1) and/or of the oncoming cold season. In Trial 1 at Institute 3 the samples were not insulated, which probably explains why the temperature curve flat- tened early, as heat diffused to the environment. For this reason the use of a thermal insulation is of paramount importance, as Honig (1990) had suggested when working with small sample sizes. It was more helpful to evaluate the time taken for the temperature to rise ≥ 2.0 K above ambient than 3.0 K, especially in Trial 2 when samples were more stable, as it allowed at least a numerical differentiation within the specified time when looking at a particular test run. Often, this single point in time is determined when running an aerobic stability test, but it is recommended to also map the temperature development. Mathematically this can be done either by linear regression equations to show the gradient or by more sophisticated models such as the Gompertz function (Zeyner et al. 2018). In the presented case, the slope of the curve showed the changes simultaneously, at least for the speed and extent of temperature increase. The results presented for Trial 1 also demonstrated that TMR stabilizing products are more effective when the orig- inal stability of the feed was moderate, i.e. a yeast load of < 105 cfu g-1 FM. In that situation a clear dose response is more probable. This is in accordance with the findings of Rinne et al. (2018) and Seppälä (2020). Increasing the dosage of fungistatic agents within reasonable economic constraints is unlikely to stop spoilage of feeds with a low initial hygienic quality. The final pH allowed a reasonable differentiation of treatments overall, and in Institute 4 in the first trial and again in Trial 2 at Institutes 4 and 1. Measuring the pH after the stability test only allows a reasonable interpretation when the samples are not yet completely spoiled, as was the case in the first trial in Institutes 1 and 2(1). In this case, the results appear to be in accordance with the visual evaluation of humidity and yeasts as shown in the second trial. Otherwise, the highest level of propionic acid treatment can lead to the highest pH, as was the case for Institute 1 in Trial 1. Thus, increase in pH and visual yeast occurrence at the end of the test can tell whether spoilage has taken place, but not when it started. In an experiment with tropical grasses, pH clearly rose before temperature in several treatments (Gomes et al. 2021). It would be desirable to monitor pH over time to best make use of this parameter, which suggests that oxidation of lactic acid has taken place (Middelhoven and Franzen 1986, Pahlow et al. 2003, Martens 2006). Shan et al. (2021b) used a special pH electrode to monitor the fermentation process in a mini-bioreactor. Other researchers tried to develop a wireless pH sensor for application to feeds (Huang et al. 2012, Marsh et al. 2020). However, none of these approaches yet seems suitable or available for routine applica- tions such as the current aerobic stability test. Visual assessment of condensed water and fungal infestation Humidity is a good indicator of ongoing oxidative processes. In order to minimize the subjective influence of visual inspection, different data loggers for relative humidity were tested in several preceding trials. Most of them failed in the special environment, which is why it was decided to use a visual inspection. However, in analogy with pH, its evaluation can only help to differentiate treatments when samples are not yet completely spoiled, as was the case with samples from Institute 1 in Trial 1. Visual evaluation of signs of yeast growth allowed an early detection of spoiling processes in the case of Institute 3 in the second trial. Although no temperature increase was detected during the 72 h of the aerobic stability test, some yeast points were scored for all treatments at opening. This demonstrated that heating is a subordinate indicator in the early stage of yeast development. On the other hand, when oxidation is at an advanced stage, Agricultural and Food Science (2022) 31: 241–259 253 yeasts cannot be detected easily by visual inspection. Either the sample is too wet to identify single yeast points or they are overgrown with mould, as was observed in pre-trials. This is the reason why other authors attempted to develop another method of observing fungal growth using a transparent covering, which unfortunately lacked insulation (Franco et al. 2018, Stefanski et al. 2018). In our trials, moulds hardly ever occurred, as they usually appear after the yeast infestation (Pahlow et al. 2003). However, moulds were observed in two out of 6 treatments in Institute 2, Trial 2. This phenomenon might be a starting point for further product development work, which could include further indicators for fungal growth such as monoclonal antibodies (Le Cocq et al. 2020). Losses FM losses can be used, but with some reservations, as not all respired H2O will have evaporated from the ves- sel. Part of it stays within the sample, another part has condensed on the wall and in the cap; this is why the lat- ter was not included in the weighing. Such FM loss will only reflect the loss of organic matter to a certain degree and when considering the determination of the DM losses this error will multiply. Water from the wall, which has been included in the FM weight, is now multiplied with the DM concentration values of the sample. This is one reason for an apparent “gain” in DM as calculated for the second run in Institute 2, Trial 1. That is why Knicky and Spörndly (2015) introduced a correction factor of 1.44 to take the water into consideration as a respiratory loss. Another reason is the inherent error in the method. Oven drying at 100, 103 or 105 °C has become a standard for DM determination of forages in many laboratories because of ease of handling and reduced risk for health and environment compared to alternative methods. Protocols using toluene or gas chromatograph to determine wa- ter content for example or freeze drying have been compared (Minson and Lancaster 1963, Aerts et al. 1974, Hui- da et al. 1986, Alomar et al. 1999). An overview on methods for determining forage moisture content is given by Cherney and Cherney (2003) in their chapter on silage quality assessment. All these studies concluded that loss of volatile compounds has to be considered when oven drying. Thus, correction factors depending on analyzed vola- tile organic acid concentration, pH and/or ammonia have been suggested for different types of silages (Weissbach and Kuhla 1995, Porter and Murray 2001, Weissbach and Strubelt 2008). However, a correction factor for silages that have undergone aerobic spoilage has not been published, as usually the disappearance of volatile organic acids in spoiled samples is not documented analytically, and was not the subject of this investigation. Thus, in the presented case no correction factors were applied before or after the test. DM values of silages will increase by correction. Thus, the simple gravimetrical determination of mass losses is a weak tool. Alternative indicators and evaluation options Another way of monitoring aerobic deterioration processes is by the continuous measurement of CO2. Honig who invented the System Völkenrode for example measured CO2 concentrations, which he then correlated with temperature measurements with the intention of replacing the more elaborate CO2 measurement with the latter (Honig 1990). He then calculated DM losses using the respiration equation C6H12O6 + O2 → H2O + CO2. However, it is not documented how he verified his assumption. Firstly, only a minor part of silage carbohydrates consists of glucose and many intermediate steps would have to be considered. Secondly, different respiratory pathways exist with different organisms such as alternative oxidase in many fungi, and with differing extents of heat production (Joseph-Horne et al. 2001). Thirdly, carbon dioxide escaping from silage has different sources which need to be distinguished. One is the efflux of gas, which has accumulated in the pores during anaerobic storage, the other is production caused by microbial activity (Shan et al. 2019). Thus, Shan et al. (2021a) attempted to identify the pro- portion of aerobic microbial respiration by determining O2, CO2 and temperature in different layers in triticale silage. Mapping the aerobic stability indicators presented of different feed materials in a normalized (i.e. minimum val- ue becomes 0, maximum possible value becomes 1) radar plot can help to understand the type of microbial be- haviour inherent in the feedstuff. An example where TMR was compared with pressed beet pulp silages can be found in the supplementary material (Appendix Fig. A5). These samples were differentiated upon visible appear- ance of signs of yeasts and/or moulds. While moulds alone or in combination with yeasts were predominant in the spoilage process of pressed beet pulp silages, moulds did not occur solely in TMR. In that case, the combina- tion of yeasts and moulds led to the highest temperature peaks and pH increases. The rapid detection of biogenic amines and pathogens might also help to indicate spoilage going beyond the de- scribed and usual parameters. Respective sensors are being developed in the food sector (Park et al. 2015, Schaude et al. 2017, Müller and Schmid 2019). Such additional information might increase the practical benefit for farmers in making management decisions in the future. S.D. Martens et al. 254 Conclusions The extended protocol described allows the evaluation of TMR stabilizing products in a specified time if the feed is of moderate hygienic quality. Highly contaminated rations will deteriorate without delay whilst hygienic rations will remain stable for a reasonably long time. This also means that the use of an antimicrobial agent cannot re- place good agricultural practice in feed conservation. In the evaluated feed materials temperature increase is usually a good indicator for spoilage processes. It is rec- ommended that the whole temperature curve is used rather than merely a single point in time. However, early signs of respiratory activity can only be detected by visual evaluation of fungal growth. Thus, it should always ac- company the appraisal of aerobic stability and equally, the determination of pH before and after the test. Parameters such as maximum temperature difference, time to maximum temperature difference, FM losses and evaluation of condensation give an additional clue to the extent of microbial spoilage that has occurred. They cannot be evaluated in isolation. Acknowledgements We kindly thank the DLG (German Agricultural Society) Commission for Silage Additives for encouraging this in- vestigation. Thanks go to Hannah Jilg for the digital drawing of Fig. A1. We are very grateful to Dr. Roger Merry for the intermediate and Dr. David Davies for the final language editing. References Aerts, J.V., De Brabander, D.L., Cottyn, B.G., Buysse, F.X. & Moermans, R.J. 1974. Comparison of methods for dry matter de- termination of high moisture roughages and faeces. Journal of the Science of Food and Agriculture 25: 619–627. https://doi. org/10.1002/jsfa.2740250604 Alomar, D., Fuchslocher, R. & Stockebrand, S. 1999. Effects of oven- or freeze-drying on chemical composition and NIR spectra of pasture silage. Animal Feed Science and Technology 80: 309–319. https://doi.org/10.1016/S0377-8401(99)00053-X Avila, C.L.S. & Carvalho, B.F. 2020. Silage fermentation - updates focusing on the performance of micro-organisms. Journal of Ap- plied Microbiology 128: 966–984. https://doi.org/10.1111/jam.14450 Beck, Th. & Gross, F. 1964. Ursachen der unterschiedlichen Haltbarkeit von Gärfutter. Das wirtschaftseigene Futter 10: 298–312. (in German). Borreani, G., Tabacco, E., Schmidt, R.J., Holmes, B.J. & Muck, R.E. 2018. Silage review: Factors affecting dry matter and quality losses in silages. Journal of Dairy Science 101: 3952–3979. https://doi.org/10.3168/jds.2017-13837 Cherney, J.H. & Cherney, D.J.R. 2003. Assessing Silage Quality. In: Buxton, D.R., Muck, R.E.& Harrison, J.H. (eds.). Silage Science and Technology. Madison, Wisconsin, USA. p. 141–198. https://doi.org/10.2134/agronmonogr42.c4 DLG TestService GmbH 2018. DLG Testing Guidelines for the award and use of the DLG Quality Mark for ensiling agents. Frankfurt am Main. http://www.dlg.org/siliermittel.html Franco, M., Jalava, T., Kahala, M., Järvenpää, E., Lehto, M. & Rinne, M. 2018. Preservatives can improve aerobic stability of potato by-products. In: Udén, P., Eriksson, T., Spörndly, R., Rustas, B.O. & Liljeholm, N. (eds.). Proceedings of the 9th Nordic Feed Science Conference, Uppsala, Sweden 12–13 June 2018. p. 143–148. Gerlach, K., Ross, F., Weiss, K., Büscher, W. & Südekum, K.H. 2014. Aerobic exposure of grass silages and its impact on dry mat- ter intake and preference by goats. Small Ruminant Research 117: 131–141. https://doi.org/10.1016/j.smallrumres.2013.12.033 Gomes, A.L.M., Auerbach, H.U., Lazzari, G., Moraes, A., Nussio, L.G., Jobim, C.C. & Daniel, J.L.P. 2021. Sodium nitrite-based addi- tives improve the conservation and the nutritive value of guinea grass silage. Animal Feed Science and Technology 279: 115033. https://doi.org/10.1016/j.anifeedsci.2021.115033 Honig, H. 1990. Evaluation of aerobic stability. In: Proceedings of the EUROBAC Conference, 12–16 August 1986, Uppsala. p. 76–82. Huang, W., Deb, S., Seo, Y., Rao, S., Chiao, M. & Chiao, J.C. 2012. A Passive Radio-Frequency pH-Sensing Tag for Wireless Food- Quality Monitoring. IEEE Sensors Journal 12: 487–495. https://doi.org/10.1109/JSEN.2011.2107738 Huida, L., Väätäinen, H. & Lampila, M. 1986. Comparison of dry matter contents in grass silages as determined by oven drying and gas chromatographic water analysis. Annales Agriculturae Fennia 25: 215–230. Joseph-Horne, T., Hollomon, D.W. & Wood, P.M. 2001. Fungal respiration: a fusion of standard and alternative components. Bio- chimica et Biophysica Acta (BBA) - Bioenergetics 1504: 179–195. https://doi.org/10.1016/S0005-2728(00)00251-6 Knicky, M. & Spörndly, R. 2015. Short communication: Use of a mixture of sodium nitrite, sodium benzoate, and potassium sorb- ate in aerobically challenged silages. Journal of Dairy Science 98: 5729−5734. https://doi.org/10.3168/jds.2015-9332 Le Cocq, K., Brown, B., Hodgson, C.J., McFadzean, J., Horrocks, C.A., Lee, M.R.F. & Davies, D.R. 2020. Application of monoclo- nal antibodies in quantifying fungal growth dynamics during aerobic spoilage of silage. Microbial Biotechnology 13: 1054−1065. https://doi.org/10.1111/1751-7915.13552 Agricultural and Food Science (2022) 31: 241–259 255 Marsh, P., Manjakkal, L., Yang, X., Huerta, M., Le, T., Thiel, L., Chiao, C., Cao, H. & Dahiya, R. 2020. Flexible Iridium Oxide Based pH Sensor Integrated With Inductively Coupled Wireless Transmission System for Wearable Applications. IEEE Sensors Journal 20: 5130–5138. https://doi.org/10.1109/JSEN.2020.2970926 Martens, S. 2006. An In-Vitro Study of Aerobic Changes in Silages: Effects of microbial activities and impact factors. Doctoral the- sis, University of Rostock. Tönning, Lübeck, Marburg, Der Andere Verlag. 151 p. https://doi.org/10.18453/rosdok_id00000271 Martens, S.D. & Steinhöfel, O. 2019. Development of an evaluation scheme for aerobic deterioration of total mixed rations. In: Jambor, V.& Mala, S. (eds.). Proceedings of the 18th International Symposium Forage Conservation, 13–16th August, 2019, Brno, Czech Republic. p. 102–103. Middelhoven, W.J. & Franzen, M.M. 1986. The Yeast Flora of Ensiled Whole-crop Maize. Journal of the Science of Food and Agri- culture 37: 855–861. https://doi.org/10.1002/jsfa.2740370906 Minson, D.J. & Lancaster, R.J. 1963. The effect of oven temperature on the error in estimating the dry matter content of silage. New Zealand Journal of Agricultural Research 6: 140–146. https://doi.org/10.1080/00288233.1963.10419331 Müller, P. & Schmid, M. 2019. Intelligent Packaging in the Food Sector: A Brief Overview. Foods 8. https://doi.org/10.3390/ foods8010016 O’Kiely, P. 1993. Influence of a Partially Neutralised Blend of Aliphatic Organic Acids on Fermentation, Effluent Production and Aerobic Stability of Autumn Grass Silage. Irish Journal of Agricultural and Food Research 32: 13–26. Pahlow, G., Muck, R.E., Friehuis, F., Elferink, S.J.W.H. & Spoelstra, S.F. 2003. Microbiology of Ensiling. In: Buxton, D.R., Muck, R.E.& Harrison, J.H. (eds.). Silage Science and Technology. American Society of Agronomy, Inc., Crop Science Society of America, Inc., Soil Science Society of America, Inc. Madison, Wisconsin, USA. p. 31–93. https://doi.org/10.2134/agronmonogr42.c2 Park, Y.W., Kim, S.M., Lee, J.Y. & Jang, W. 2015. Application of biosensors in smart packaging. Molecular & Cellular Toxicology 11: 277–285. https://doi.org/10.1007/s13273-015-0027-1 Porter, M.G. & Murray, R.S. 2001. The volatility of components of grass silage on oven drying and the inter-relationship between dry-matter content estimated by different analytical methods. Grass and Forage Science 56: 405–411. https://doi.org/10.1046/ j.1365-2494.2001.00292.x Rinne, M., Franco, M., Kuoppala, K., Seppälä, A. & Jalava, T. 2018. Response to total mixed ration stabilizers depends on feed quality. In: Gerlach, K. & Südekum, K.-H. (eds.). Proceedings of the XVIII International Silage Conference, Bonn, Germany, 24–26 July 2018. p. 538–539. Schaude, C., Meindl, C., Fröhlich, E., Attard, J. & Mohr, G.J. 2017. Developing a sensor layer for the optical detection of amines during food spoilage. Talanta 170: 481–487. https://doi.org/10.1016/j.talanta.2017.04.029 Seppälä, A. 2020. Controlling aerobic stability of silage based total mixed rations. Natural resources and bioeconomy studies 45. Natural Resources Institute Finland (Luke), Helsinki. 60 p. Seppälä, A., Heikkilä, T., Mäki, M., Miettinen, H. & Rinne, M. 2013. Controlling aerobic stability of grass silage-based total mixed rations. Animal Feed Science and Technology 179: 54–60. https://doi.org/10.1016/j.anifeedsci.2012.11.011 Shan, G., Buescher, W., Maack, C., Zhou, H., Grantz, D.A., Lipski, A., Acir, I.H. & Sun, Y. 2019. An automatic smart measurement system with signal decomposition to partition dual-source CO2 flux from maize silage. Sensors and Actuators B: Chemical 300: 127053. https://doi.org/10.1016/j.snb.2019.127053 Shan, G., Maack, C., Buescher, W., Glenz, G., Milimonka, A., Deeken, H., Grantz, D.A., Wang, Y. & Sun, Y. 2021a. Multi-sensor meas- urement of O2, CO2 and reheating in triticale silage: An extended approach from aerobic stability to aerobic microbial respiration. Biosystems Engineering 207: 1–11. https://doi.org/10.1016/j.biosystemseng.2021.04.004 Shan, G., Rosner, V., Milimonka, A., Buescher W., Lipski, A., Maack, C., Berchtold, W., Wang, Y., Grantz, D.A. & Sun, Y. 2021b. A Multi-Sensor Mini-Bioreactor to Preselect Silage Inoculants by Tracking Metabolic Activity in situ During Fermentation. Frontiers in Microbiology 12: 2048. https://doi.org/10.3389/fmicb.2021.673795 Spoelstra, S.F., Courtin, M.G. & Beers, J.A.C.V. 1988. Acetic acid bacteria can initiate aerobic deterioration of whole crop maize si- lage. The Journal of Agricultural Science 111: 127–132. https://doi.org/10.1017/S0021859600082915 Stefanski, T., Franco, M., Kautto, O., Jalava, T., Winquist, E. & Rinne, M.2018. Grass silage for biorefinery - Separation efficiency and aerobic stability of silage and solid fraction. In: Udén, P., Eriksson, T., Spörndly, R., Rustas, B.O. & Liljeholm, N. (eds.). Proceed- ings of the 9th Nordic Feed Science Conference, Uppsala, Sweden, 12-13 June 2018. p. 153–158. VDLUFA 1976. VDLUFA Methodenbuch Band III Die chemische Analys e von Futtermitteln. 3rd ed. VDLUFA-Verlag, Darmstadt. 2190 p. (in German). VDLUFA 2012. Mikrobiologische Verfahren. In: VDLUFA Methodenbuch Volume III Die chemische Untersuchung von Futtermitteln (1976). Supplement 8 as of 2012. VDLUFA-Verlag. Darmstadt. (in German). Weissbach, F. & Kuhla, S. 1995. Substance losses in determining the dry matter content of silage and green fodder: arising errors and possibilities of correction. Übersichten zur Tierernährung 23: 189–214. Weissbach, F. & Strubelt, C. 2008. Die Korrektur des Trockensubstanzgehaltes von Zuckerrübensilagen als Substrat für Biogasanla- gen. LANDTECHNIK - Agricultural Engineering 63: 354–355. (in German). https://doi.org/10.15150/lt.2008.879 Whitlock, L.A., Wistuba, T., Siefers, M.K., Brent, B.E., Bolsen, K.K. & Pope, R.V. 2000. Effect of level of surface-spoiled silage on the nutritive value of corn silage-based rations. Kansas Agricultural Experiment Station Research Reports 0: 22–24. https://doi. org/10.4148/2378-5977.1785 Zeyner, A., S. Hoedtke, S., Martens, S.D., Steinhöfel, O. & Wensch-Dorendorf, M. 2018. Extension of the biological relevance of the Rostock Fermentation Test by curve fitting and interpretation. In: Gerlach, K.& Südekum, K.-H. (eds.). Proceedings of the XVIII International Silage Conference, Bonn, Germany, 24–26 July 2018. p. 28–29. Agricultural and Food Science (2022) 31: 241–259 256 Appendix Table A1. Examples where no or very slight temperature increase (< 3 K above ambient) occurred during the aerobic stability test (AST), and which however showed visible signs of deterioration after the test (values in brackets mean ranges) Feed material DM (g kg-1) Ambient (°C) AST (h) Max.TD (K) Yeasts (0–4) Molds (0–4) pH before pH after AST n Year observed Grass silage 428 20.6 215 2.3 3.0 3.5 4.70 8.37 1 2020 [1] Grass silage, 2nd cut 380 24.7 310 0.7 (0–1.5) 0 2.3 (1.5–3.0) 4.19 4.36 (4.20V–4.63) 6 2020 [2] Grass silage, 1st cut 344 22.6 336 0.4 (0–1.5) 0.9 (0.5–1.5) 0.4 (0–2.5) 4.00 4.02 (3.94–4.08) 7 2019 [2] Grass silage, 1st cut (L. multiflorum) 352 24.8 333 1.4 (0.6–2.8) 0.7 (0.5–1.0) 0 3.98 4.09 (3.94–4.16) 5 2019 [2] Lucerne silage (M. sativa) 465 23.7 499 0.5 (0.0–0.9) 0.8 (0.5–1.0) 2.7 (2.5–3.0) 5.09 5.58 (5.11–6.50) 3 2020 [2] Lucerne silage 470 23.7 499 0 0 1.5 4.54 4.64 (4.60–4.66) 3 2020 [2] Lucerne silage 387 24.1 331 0 0 1.5 (0.5–2.5) 4.44 4.44 3 2020 [2] Pressed beet pulp silage 230 17.3 158 0.8 (02.85) 0.5 (0–1.5) 0.8 (0–2.5) 3.67 (3.53–3.85) 3.77 (3.55–4.32) 15 2017 [3] Chicory silage (C. intybus) 144 23.9 192 1.5 (0–2.8) 0 2.1 (1.5–3.0) 4.29 (4.13–4.62) 4.46 (4.23–5.01) 7 2016 [3] Maize silage 370 19 192 2.3 0 2.0 3.80 n.a. 6 2019 [4] Guinea grass silage 226 25 150 < 2.0 4.79 5.29 4 2021 [5] Guinea grass silage 234 25 144 < 2.0 4.44 4.94 4 2021 [5] Fresh potato by-products 153 20 96 0 0 4.0 5.31 n.a. 2018 [6] Note: Max. TD = maximum temperature difference; (0–4) points: see Table 3 and Figs. 3 and 4. References: (1) Jilg, A. (Agricultural Centre for cattle production, grassland management, dairy food, wildlife and fisheries Baden-Württemberg (LAZBW), Aulendorf, Germany). Personal communication, 2021. (2) Ohl, S. (Chamber of Agriculture Schleswig-Holstein, Research and Training Centre Futterkamp, Gutshof, Blekendorf, Germany). Personal communication, 2021. (3) Martens, S.D. (Saxon State Office for Environment, Agriculture and Geology, Department of Animal Husbandry, 04886 Köllitsch, Germany). Personal communication, 2021. (4) Sun, Y. & Maack, C. (Department of Agricultural Engineering, The University of Bonn, Bonn, Germany). Personal communication, 2021. (5) Franco, M., Jalava, T., Kahala, M., Järvenpää, E., Lehto, M. & Rinne, M. 2018. Preservatives can improve aerobic stability of potato by-products. 9th Nordic Feed Science Conference, Uppsala, Sweden, 12-13 June 2018. Rapport 298: 143–148. Uppsala, Swedish University of Agricultural Sciences, Department of Animal Nutrition and Management. (6) Gomes, A.L. M., Auerbach, H.U., Lazzari, G., Moraes, A., Nussio, L.G., Jobim, C.C. & Daniel, J.L.P. 2021. Sodium Nitrite-Based Additives Improve the Conservation and the Nutritive Value of Guinea Grass Silage. Animal Feed Science and Technology 279: 115033. https://doi.org/10.1016/j. anifeedsci.2021.115033 Agricultural and Food Science (2022) 31: 241–259 257 Appendix Fig. A1. Schematic of the general procedure from preparing and evaluating the aerobic stability test of fermented feeds Fig. A2. Humidity score (0–4) from visual evaluation after the aerobic stability test Agricultural and Food Science (2022) 31: 241–259 258 Appendix Fig. A3. Exemplary yeast score (0–4) from visual evaluation after the aerobic stability test. Fig. A4. Exemplary mold score (0–4) from visual evaluation after the aerobic stability test. Agricultural and Food Science (2022) 31: 241–259 259 Appendix Fig. A5. Example of normalized radar plots with aerobic stability indicators sorted by the occurrence of no yeasts + molds, only yeasts, only molds and yeasts + molds in (a) pressed beet pulp, (b) Total Mixed Rations. Testing the aerobic stability of silage-based mixed rations usinga multiple indicator approach Introduction Materials and methods Trial 1: Dose-response study with propionic acid Trial 2: TMR stabilizing products and both positive and negative control Statistical analyses Results Trial 1 Trial 2 Conclusions Discussion Temperature increase and pH changes Visual assessment of condensed water and fungal infestation Losses Alternative indicators and evaluation options Acknowledgements References Appendix Table A1. Fig. A1. Fig. A2. Fig. A3. Fig. A4. Suppl. Fig. A5.