The study was conducted under GLP conditions. The test field was a meadow near Homberg (Ohm) at the FNU research centre Neu-Ulrichstein, Hesse, Germany. Since 2010 it was used as meadow without the use of mineral fertilisers and pesticides. For the characterization of the field the soil particle size distribution, organic matter content of the soil, soil pH, total N-content and water holding capacity of the A-horizon was determined from samples taken at 0–30 cm depth once on 18 September 2018 in every quadrant of the experimental field using a Puerkhauer sampler. On each sampling date, vegetation height and coverage of the plots for each treatment group was recorded. On the dates of soil core sampling, soil moisture was determined using a Theta-Probe (HH2 meter, Delta-T Devices, Cambridge, UK). Continuous weather data (air temperature, precipitation and soil temperature) during the study period were obtained from the DWD (Deutscher Wetter Dienst) weather station (World Meteorological Organization number 10537) about 500 m away from the test field. Climatic conditions (air temperature, air humidity, soil temperature and on date of application wind velocity) on the experimental field were recorded with suitable instruments (ALMEMO type 2290-8 or 2590; AHLBORN Mess-und Regelungstechnik GmbH, 83,602 Holzkirchen, Germany) on the date of application and each sampling date.
Test design and test item/reference item application
The test was designed according to ISO 11268-3 [5] with five treatments and four replicates each (20 plots in total) as a randomized block design. The plot size was 10 × 12 m (120 m2) with a distance of at least 3 m between adjacent plots and a plot margin (unused) of at least 10 m. Four plots were left untreated as unfertilized control. Four plots were treated with 200 kg Perlka®/ha and 400 kg Perlka®/ha, respectively. The urea fertiliser Piagran®46 (SKW Piesteritz, total nitrogen content approx. 46.5% (nominal)) was applied to four plots at the same total nitrogen rate as provided by the high Perlka® application rate, corresponding to 172.9 kg Piagran®/ha. As reference item Agriclor® (480 g chlorpyrifos/L) was used at a rate of 1.449 L product/ha (corresponding to 0.72 kg. chlorpyrifos/ha) in 400 L water/ha. The reference item was applied once with the first application of the mineral fertilisers. The first application of Perlka® (batch-no.: SWSE-18-068; content of a.i.: 45.0% calcium cyanamide (analysed)), Piagran® and the reference item was on 28 September 2018. The second application of Perlka® (batch-no.: SWSE-19-017; content of a.i.: 44.4% calcium cyanamide (analysed)) and Piagran® was on 2 April 2019.
The mineral fertilisers were applied using an accurate fertiliser spreader for granules (Hege 80 Parzellenstreuer, System Weihenstephan, Model 422437 with a working width of 1.50 m and 10 gape pipes, Hans-Ulrich Hege Saatzuchtmaschinen, 74638 Waldenburg, Germany). For the application of the reference item a movable plot sprayer for field application; type PSG 4 FE (Fa. Schachtner Gerätetechnik, 71640 Ludwigsburg, Germany) was used, with an extension tube including 5 spraying nozzles (Lechler IDK 120–04; distance between nozzles: 50 cm) operating with compressed air and a boom width of 250 cm and a distance to soil of approximately 50 cm. Since Perlka® and Piagran® are formulated as granules, verification of the application rate cannot be carried out in the same way as for plant protection products, where overspray of soil samples or filter material in Petri dishes is the recommended method for quantifying the applied amount of test item [3]. Therefore, a method according to DIN 13739-1 [12] was used to verify the applied quantity and distribution homogeneity. Five collecting trays (37.5 × 50 cm) were placed on each Perlka® and control fertiliser plot. The trays were passed once during the application. Weight of the granules in the trays was determined. The amount of granules applied per hectare was calculated on the basis of the surface area of the collection trays and the amount of fertiliser collected.
After the first application, the test field was irrigated to achieve at least 10 mm precipitation. After the second application, sufficient natural rainfall occurred. Due to a summer characterized by high temperatures and long periods of drought, irrigation was carried out before the start of the study to achieve appropriate sampling conditions (11 mm/m2 on 30 August, 31 August, 3 September, 4 September and 13 September 2018, respectively, in total 55 mm/m2). In preparation of the autumn samplings in 2019 the following irrigation schedule was implemented: 3 September 2019 7 L/m2, 4 September 2019 8 L/m2, 10 September 2019 8 L/m2, 11 September 2019 9.6 L/m2, 12 September 2019 10 L/m2.
Sampling and identification of Collembola
Collembola were sampled by collecting 6 soil cores per plot with a depth of 5 cm using a core sampler with 5 cm diameter followed by heat extraction with a modified MacFadyen extraction apparatus according to ISO 23611-2 [13]. Soil core samples were taken on 11 sampling dates: 26 September 2018; 2 October 2018; 12 October 2018; 26 October 2018; 1 April 2019; 5 April 2019; 16 April 2019; 30 April 2019; 24 June 2019; 23 September 2019 and 15 October 2019. The specimens were identified at species level and the number per soil core was determined.
In addition, four funnel pitfall traps according to Barber and Melber [14, 15] were installed in the central area of each plot with at a distance of 2 m from each other. The traps were opened for 4 days at each sampling and were equipped with a preservative (50% ethylene glycol, 50% water) in order to conserve the captured organisms. The samples were taken on 11 occasions: 25 September 2018; 5 October 2018; 12 October 2018; 26 October 2018; 1 April 2019; 8 April 2019; 16 April 2019; 30 April 2019; 24 June 2019; 23 September 2019 and 15 October 2019. After sampling the pitfall trap specimen were washed with tap water to remove soil and plant material; washing water was poured through sieves with a mesh size of 150 µm and the samples were fixed with 70% ethanol until the taxonomic evaluation was performed. The specimens were identified at species level and the number of individuals per trap for each species was determined.
Taxonomic identification was carried out using a dissecting microscope and/or transmitted light microscope and the following keys [16,17,18,19,20,21,22,23,24,25,26].
Data evaluation
For the statistical evaluation the results of both sampling types were considered individually.
To test the significance of the differences between the mean values of controls and treatments for each taxon and sampling date, the multiple t-test by Williams [27, 28] was used, which provides the NOER (No Observed Effect Rate) at the population level (α = 0.05, one-sided). For regulatory purposes direct effects are in the focus, expecting monotonic concentration-responses of the test item. Therefore, Williams test was selected for the evaluation. Furthermore, one-sided tests were selected to increase the power of the test to detect direct effects with a monotonous dose–response. Direction of the test was determined from comparison of the means of controls and highest treatment level. If for example the mean abundance at the highest test concentration was lower than the mean of the control, the test was conducted for a decrease of abundance, and vice versa.
The minimal detectable difference (MDD) at the NOER in accordance with Brock et al. [29] was also calculated as an indication of the statistical power. For NOER-calculation abundance data of the taxa were log transformed [y’ = ln (ay + 1)] with a = 2/min(x) before analysis in order to better approximate normality and homoscedasticity (homogeneity of variances) requirements, whereby min(x) was the smallest value of the data set, which was greater than zero [30]. NOER-calculations were done with the program Community Analysis (CA) 4.3.14 [31].
As some statistically significant differences could be determined between the control and the fertiliser control, the statistical evaluation of the effects of the Perlka® application rates were done separately in comparison with the untreated control and the fertiliser control.
Number of species, Shannon Index and evenness were used to describe the diversity of the community. Shannon Index, a diversity measure depending on species richness and frequency distribution of the individuals of the species, was calculated using the following formula:
$$H_{{\text{S}}} = - \sum {p_{j} \ln (p_{j} )} ,$$
with HS = Shannon Index, pj = relative abundance of species j.
Evenness was calculated as follows: E = HS/HSmax = HS/ln (n), with E = evenness, n = number of species.
To determine differences between the means in controls and treatment the multiple t-test by Williams [27, 28] of the program Community Analysis (CA. 4.3.14) was used.
In order to evaluate the validity of the study, the comparison of control and reference item was done separately. Abundance of Collembola was tested for normality (Shapiro–Wilk test) [32] and homogeneity of variance (Levene’s test) [33, 34] using ToxRat® Professional (ToxRat® Solutions GmbH, 52477 Alsdorf, Germany, Version 3.3.0). Depending on the results, Student’s t-test, Welch t-test or Mann–Whitney U-test was selected to compare control and reference item. These tests were calculated one-sided smaller (α = 0.05). The comparison of the untreated control and fertiliser control was evaluated in the same way, but the respective tests were calculated two-sided (α = 0.05). According to de Jong et al. [3], only taxa with a mean abundance of > 2.5 individuals per sample in the control were considered relevant for the statistical evaluation.