Trophic niche but not abundance of Collembola and Oribatida changes with drought and farming system

Study site

The study was performed in 2017 in the DOK trial, an agricultural long-term field experiment established in 1978 comparing different organic and conventional farming systems. The DOK trial is located in Therwil, Switzerland, at 300 m above sea level on a Haplic Luvisol on deep deposits of alluvial loess (Fließbach et al., 2007). The mean annual temperature over the last five years was 10.5 °C and the mean annual precipitation was 842 mm (Krause et al., 2020). For this study we used winter wheat fields with soybean as the previous crop. The experimental fields were organized in four blocks each comprising a conventionally and an organically managed field (factor farming system, CONMIN and BIODYN systems of the DOK trial, respectively). Conventionally managed fields received mineral fertilizer (40–60 kg N/ha in March, April and May), herbicides (0.1 l/ha of Husar OD, Bayer, Zollikofen, Switzerland, and 1 l/ha of Mondera, Switzerland, once in March), insecticides (0.1 l/ha of Audienz; Omya, Oftringen, Switzerland, in May) and fungicides (1.5 l/ha Pronto Plus in April and 1 l/ha AviatorXpro and Miros FL in May; all Bayer) as well as plant growth regulators (1.5 l/ha Cycocel extra; Omya, in March). Organically managed fields received only organic fertilizers (farmyard manure, compost and slurry), biodynamic preparations and mechanical weed control (Krause et al., 2020; Kundel et al., 2020). All fields were ploughed up to a depth of 20 cm and seedbed preparation was done with a tooth harrow to a depth of 10 cm. In both systems 415 grains/m² were sown. All fields followed the same 7-year crop rotation with soybean as the preceding crop. On each field one drought treatment and one control plot were established (factor drought). We simulated drought by using experimental rainout-shelters that excluded 65% of the precipitation (for details on the shelter construction see Kundel et al., 2018). On the control plots, we established a similar shelter construction with the difference, that it did not reduce precipitation entering the plot. Thereby, we accounted for possible side effects caused by the roof construction itself (Kundel et al., 2018).

Sampling

Samples were taken in May, eight weeks after the establishment of the experiment, with soil cores of 5 and 20 cm diameter to a depth of 10 cm in the center of the plots (n = 16). Microarthropods with high densities, in our case the Collembola Mesaphorura sp. and Oribatida, were taken from the five cm cores. For the other Collembola, i.e., Isotoma caerulea, Isotomurus maculatus and Orchesella villosa, abundances in the small cores were too low to obtain enough material for stable isotope measurements, so we took them from the large soil cores that were initially taken to extract macrofauna. Soil animals were extracted from intact soil cores by gradually increasing the temperature from 25 to 55 °C over ten days (Macfadyen, 1961; Kempson, Lloyd & Ghelardi, 1963), collected into a glycol-water solution (1:1) in canisters underneath the soil coresand and stored in 70% ethanol. We first sorted the extracted animals to order level under a stereomicroscope (Stemi 2000; Zeiss). Thereafter, Collembola and Oribatida were identified to species or genus level under the microscope (Axioplan; Zeiss). As slide-mounting medium we used 70% ethanol, because other commonly used solutions like lactic acid may change stable isotope compositions. We used keys by Hopkin (2007), Fjellberg (1998); Fjellberg (2007) and Weigmann (2006).

Stable isotope analysis

The four most abundant Collembola taxa (I. caerulea, I. maculatus, O. villosa and Mesaphorura sp.) and the two most abundant oribatid mite species (Scheloribates laevigatus and Tectocepheus sarekensis) were chosen for stable isotope analysis. To achieve at least 10 µg of animal dry weight per sample we used 1–14 individuals per sample. To have at least three values for every species x drought x farming system combination we included pseudoreplicates in plots with many individuals (Table 1). Animals were weighed into tin capsules and dried at 60 °C for 24 h. Wheat from every plot was dried, milled and weighed into tin capsules (ca. 1 mg per sample). Stable isotope analysis of animals was done with a coupled setup of an elemental analyzer (Eurovector, Milano, Italy) and a mass spectrometer (Delta Vplus, Thermo Fisher Scientific, Bremen, Germany) adjusted for small sample sizes (Langel & Dyckmans, 2014). Stable isotope analysis of wheat was done with another set of elemental analyzer and mass spectrometer (Flash 2000 elemental analyser coupled to a DELTA Plus XP continuous-flow IRMS via a ConFlo IV interface, Thermo Fisher Scientific, Bremen, Germany). Variations in stable isotope ratios including baseline correction were expressed using the delta notation with ΔX = (R_SAMPLE/R_STANDARD)/R_STANDARD ×1000 with X representing the target isotope (¹³C, ¹⁵N), and R_SAMPLE and R_STANDARD the ratios of the heavy to the light isotope (¹³C/¹²C, ¹⁵N/¹⁴N) of the sample and the standard, respectively. As standard for ¹³C PeeDee Belemnite and for ¹⁵N atmospheric air was used (Coplen et al., 2002). Acetanilide was used for internal calibration.

Statistical analyses

All statistical analyses were done in R version 4.0.2 (R Development Core Team, 2020).

We calculated mean abundances for each species. Abundance data were analyzed with linear mixed effects models (LMMs) for individual species with farming system and drought as fixed factors, and field as random factor using the package nlme (Pinheiro et al., 2021).

Stable isotope data were baseline corrected using wheat stable isotope values of the respective plot and analyzed with a LMM with farming system and drought as fixed factors, and plot as random factor to account for differences in sample size. Because the interaction species × drought as well as species × farming system was significant (Table 2), we ran individual LMMs for each species to detect species-specific effects of drought and farming system. In these models we again included drought and farming system and their interaction as fixed factors, and plot as random factor.

The size of the isotopic niches of each species in the two farming systems and in the two drought treatments was calculated and visualized with the R package SIBER (Jackson et al., 2011). Standard ellipse areas with a correction for small sample sizes (SEAc) based on maximum likelihood were estimated and used to visualize isotopic niches of all species in the two farming systems and drought treatments. To compare isotopic niche widths between farming systems and drought treatments within species, Bayesian multivariate normal distributions were fitted to the two levels of the factor farming system and drought, with prior settings of length, number and iterations of sampling chains, and distribution parameters as recommended by Jackson (2019). Based on these probability distributions Bayesian standard ellipse areas were calculated and plotted using the function siberDensityPlot() including 50%, 75% and 95% credible intervals. For statistical comparison of isotopic niche sizes of the farming systems and the drought treatments for individual species, we compared probability distributions from the Bayesian standard ellipses with 95% credible intervals.

Soil characteristics

Water holding capacity, pH and total carbon were higher in organically compared to conventionally managed fields (Table S1). Total carbon at our study site is equivalent to total organic carbon, because the soil is free of carbonates. Soil water content was decreased by experimental drought by 4.23% and was generally higher in organically compared to conventionally managed fields (Fig. S1; see Meyer et al., 2021).

Abundance

Based on their mean abundance the six mesofauna taxa could be separated into two groups of high and low abundance with abundances of the former being 23 to 73 times higher than that of the latter. Highly abundant taxa included S. laevigatus, T. sarekensis and Mesaphorura sp. (overall average of 7648 ± 1528, 7392 ± 1286 and 5312 ± 1734 ind m⁻², respectively; mean ± SE). Species with low abundances included I. caerulea, I. maculatus and O. villosa (106.8 ± 32.7, 105.0 ± 29.3 and 227.5 ± 49.8 ind m⁻², resepectively). Generally, abundances of individual species did not change significantly with drought treatment or farming system (Table 3, Fig. S2).

Isotope values

Mean stable isotope values were significantly different between species, spanning over two δ units for ¹³C and over four δ units for ¹⁵N (Fig. 1, Table 2). The Δ¹³C values of the two oribatid mite species were three to four δ units higher than those of the three Collembola species. Mean Δ¹⁵N values spanned over four δ units with the values of S. laevigatus exceeding those of the other species by three to four δ units.

The Δ¹³C but not Δ¹⁵N values differed significantly among the studied mesofauna species between the drought treatments (Table 2), with this pattern being driven by a significant reduction in the Δ¹³C values of S. laevigatus under drought; Δ¹³C values of the other species were not significantly affected by drought (Fig. 2, Table 3). By contrast, both Δ¹³C and Δ¹⁵N values of mesofauna species varied significantly with farming system (significant species × farming system interaction; Table 2). In organically managed fields the Δ¹³C value of T. sarekensis and the Δ¹⁵N values of I. caerulea and O. villosa significantly exceeded those in conventionally managed fields (Fig. 3, Table 3).

Drought significantly reduced the isotopic niche width of S. laevigatus (P = 0.016), I. caerulea (P = 0.003) and I. maculatus (P = 0.032) (Fig. 4), with isotopic niches of S. laevigatus partly overlapping between the two drought treatments, whereas in I. caerulea and I. maculatus they overlapped in full (Fig. 5). Further, the isotopic niche space of I. caerulea and I. maculatus was significantly smaller in organically compared to conventionally managed fields, with isotopic niches of I. caerulea partly overlapping between the two farming systems, whereas those of I. maculatus overlapped in full (Fig. 6).

Trophic positions

Overall, stable isotope values of the studied microarthropods spanned two δ units in ¹³C and four δ units in ¹⁵N, indicating the utilization of different C resources and the representation of at least two trophic levels, assuming an enrichment of about 3 δ units per trophic level (Post, 2002). Based on the Δ¹³C and Δ¹⁵N values of the individual species, the studied taxa can be separated into three groups, the three Collembola species, the Oribatida species T. sarekensis and the Oribatida species S. laevigatus.

The three Collembola species I. caerulea, I. maculatus and O. villosa had Δ¹⁵N values close to zero, indicating they are closely linked to wheat plants and suggesting that they live as primary decomposers that are little enriched in ¹⁵N (−0.05‰, −0.12‰, 1.23‰, respectively). Earlier studies also found large epi- and hemiedaphic Collembola species, such as the ones we studied, to predominantly feed on plant-derived resources in both agroecosystems and forests (Pollierer et al., 2009; Birkhofer et al., 2016; Potapov et al., 2016). Ngosong et al. (2009) further found plant rather than fungal resources to be incorporated by Collembola in agricultural systems, and results of the study of Li et al. (2020) suggest that root-derived carbon is a major resource.

Isotope values of ¹³C of both Oribatida species exceeded those of the three Collembola species by one to two δ units, indicating that both are linked to resources enriched in ¹³C. However, their Δ¹⁵N values indicated that they occupy different trophic levels with T. sarekensis living as primary decomposer and S. laevigatus as secondary decomposer or predator, similar to what has been previously suggested (Schneider et al., 2004; Haynert et al., 2017). The average δ¹³C value of T. sarekensis being 4.55‰ higher than plant litter indicates that T. sarekensis is linked to older carbon resources, probably soil organic matter in deeper soil layers (Potapov, Tiunov & Scheu, 2019). The average δ¹⁵N value of S. laevigatus being 3.77‰ higher than that of plant litter indicates a mixed diet consisting of mainly microorganisms, but in part also microbial feeders such as nematodes.

Farming system

Our second hypothesis was partly supported by the significantly higher isotope values of T. sarekensis, I. caerulea and O. villosa in the organic compared to the conventional farming system. However, we did not find differences for the other taxa, which is in line with earlier studies comparing different agricultural systems (Haubert et al., 2009; Birkhofer et al., 2011; Lagerlöf, Maribie & John, 2017). The higher isotope values of T. sarekensis, I. caerulea and O. villosa in the organic farming system likely are related to the higher soil organic carbon content that was found in this system. The higher Δ¹³C values of T. sarekensis in organically compared to conventionally managed fields indicate that they more intensively feed on old carbon resources in the organic system, which is richer in soil organic matter due to long-term input of farmyard manure and compost (Mäder et al., 2002). In I. caerulea and O. villosa Δ¹⁵N values were higher in organically compared to conventionally managed fields, pointing to a higher proportion of microorganisms in their diet in the organic system. In fact, previous studies conducted in the same long-term experiment as the present study found higher microbial biomass in the organically than the conventionally managed fields indicating a higher availability of microbes as food resource (Esperschütz et al., 2007; Fließbach et al., 2007). This is likely to be a consequence of higher soil water content and soil organic carbon in the organically managed fields. This effect might further be enhanced by higher δ¹⁵N values of the organic fertilizer (farmyard manure) compared to the inorganic fertilizer in the conventional system (Birkhofer et al., 2011). Higher stable isotope values in the organic system may additionally be caused by stable isotope enrichment of soil organic matter due to stronger internal nutrient cycling (Vervaet et al., 2002; Hobbie & Ouimette, 2009).

Besides the comparison of mean stable isotope values, additional information on trophic shifts of species can be obtained by comparing trophic niche width and trophic niche space (Behan-Pelletier, 1999; Bearhop et al., 2004). Our hypothesis on changes in trophic niche width with farming system was based on the assumption that in less disturbed habitats consumers would have a greater range of potentially available food resources, from which species could select according to their preferences. By contrast, in more severely disturbed systems, preferred resources may not be available, forcing consumers to feed on a wider range of resources resulting in broader trophic niches. Our data support this hypothesis only partly for the two farming systems and, interestingly, in some species showed the opposite pattern for the drought treatment (see below). Variations in Δ¹³C values in I. caerulea and I. maculatus were small in the organic farming system indicating a diet consisting of fresh litter or root exudates, whereas in the conventional farming system diets varied more widely. This suggests the utilization of a wider range of resources including old litter and microorganisms resulting in increased δ¹³C values (Potapov, Tiunov & Scheu, 2019) or algae resulting in decreased δ¹³C values (Tozer et al., 2005). In the conventional system, the amount of litter input is low and limited to plant residues from the crop plant, i.e., mainly roots, whereas in the organic system plant residues in the organic fertilizer provide additional food resources. Further, the amount of rhizodeposits in the conventional system is likely to be lower than in the organic system, thereby providing fewer resources to the belowground food web (Jones et al., 2001; Li et al., 2016; Wang, Chapman & Yao, 2016). The lower availability of preferred food resources in the conventional compared to the organic farming system may force soil invertebrates to broaden their trophic niche. Further, in the conventional farming system more algae may be present due to the scarcity of weeds (Meyer et al., 2021) providing additional food resources that are not equally available in organic farming systems.

Drought

Drought significantly decreased soil moisture in both farming systems. However, contradicting our second hypothesis, drought did not affect stable isotope values of most taxa and there was no significant interaction with farming system. Only S. laevigatus had lower Δ¹³C and constantly high Δ¹⁵N values in the drought treatment indicating prey switching. Assuming that S. laevigatus, at least in part, feeds on nematodes, this might represent a switch from microbial-feeding to plant-feeding nematodes, due to microbial-feeding nematodes being heavily stressed under dry conditions due to reduced microbial activity (Kundel et al., 2020). In contrast to conventional farming and contrasting our second hypothesis, drought decreased the trophic niche width in some species (S. laevigatus, I. caerulea and I. maculatus). For S. laevigatus this was caused by lower Δ¹³C values in the drought treatment, probably due to prey switching (see above). In the two Collembola species the decreased trophic niche width was due to decreased variation in Δ¹³C values, but not lower mean Δ¹³C values, indicating more restricted consumption of plant-derived resources rather than algae and microorganisms. Accessibility of algae and microorganisms is likely to decrease at low soil moisture, whereas the availability of (higher) plant-derived resources may be less affected. In fact, plant-related variables, including root biomass, shoot biomass and grain yield, did not differ between the drought treatments in this experiment (Kundel et al., 2020). For I. caerulea, additionally, the smaller variation in Δ¹⁵N values, but no changes in mean Δ¹⁵N values, in the drought treatment supports the conclusion of narrower trophic niches due to more pronounced feeding on plant material.