Study site
This study was conducted at the Abisko Scientific Research Station (68°21′17.0′′ N; 18°48′54′′ E) about 200 km north of the Arctic Circle. The station is located within the sporadic permafrost zone, with a mean annual temperature around 0 °C and mean annual precipitation of 335 mm (1981–2010, Abisko Scientific Research Station, 380 m.a.s.l).
Common garden experiment
We established a mesocosm (N = 48) experiment in the experimental garden of the Abisko Scientific Research Station. An overview of the experimental design is shown in Fig. S1 and a timeline for all activities is shown in Table S1. The mesocosm experiment consisted of natural tundra vegetation–soil monoliths from two vegetation types (meadow and heath) installed during the autumn of 2013. Monoliths for the mesocosms were collected in the Kärkevagge valley, about 20 km northwest of Abisko, Sweden, (at around 68°24′36′′ N; 18°19′11′′ E) at about 700 m elevation. Here, we selected representative patches of both heath and meadow vegetation. The heath was dominated by dwarf shrubs (Empetrum hermaphroditum, V. myrtillus, V. vitis-idaea), with a considerable amount of graminoids (Carex bigelowii, Deschampsia flexuosa, Festuca ovina) and bryophytes (Pleurozium schreberi, Hylocomium splendens). The meadow was dominated by forbs (Alchemilla glomerulans, Bistorta vivipara, S. alpina) and graminoids (C. bigelowii, D. flexuosa, F. ovina). Soils were pooled by vegetation type and homogenized to ensure a similar depth of soil under each vegetation patch regardless of their original soil depth. Mesocosms were constructed using polypropylene plastic boxes (50 × 39 × 30 cm, with 10 drainage holes with a diameter about 1 cm at the bottom) with a layer of weed cloth, a fine layer (c. 1 cm) of sand, and the mixed heath or meadow soil (intact vegetation including organic sods of 8 cm, placed on top of an Ah mineral soil horizon of 22 cm). The thickness of the O horizon are typical to that of local soils prior to earthworms invasions27. Average soil organic matter content for the upper 10 cm for the heath and meadow mesocosm were 12.1 ± 0.8% (mean ± std err) and 11.4 ± 1.3%, respectively.
In the common garden, eight wooden raised beds (300 cm × 160 cm, 40 cm high) were prepared, insulated with 5 cm thick styrofoam and filled with sand. The mesocosms were then installed into the raised beds at a randomly determined location. Either five or seven mesocosms were installed per raised bed and these mesocosms presented both heath and meadow vegetation in each bed at approx. equal numbers. The soil surface of the mesocosm was levelled with the surrounding sandy beds to maintain realistic soil temperature fluctuations. The mesocosms were allowed to recover from disturbance between the summer of 2013 and spring 2017 prior to the onset of the experiment. During the experiment soil temperature (−1 cm) and volumetric soil moisture content (down to −10 cm) were recorded at an hourly interval using data loggers (Em50 ECH2O) equipped with temperature (EC-5) and soil moisture (ECT) sensors (Decagon Systems, Washington state, US). See Supplementary Fig. 1 for location of loggers.
Earthworm treatments
Earthworms were first introduced to the mesocosms on 9th of June 2017 at relevant densities. The earthworm species’ used (Aporrectodea sp. and Lumbricus sp.) are observed to radiate out from anthropogenic sources in the surrounding arctic environment with four known invasion sites situated within 12 km from our mesocosm experiment27. All earthworms used in this experiment were collected manually in forest adjacent to farmlands within the municipality of Umeå (63°50′17.7′′ N; 20°18′46′′ E), Sweden. In each mesocosm, we added a total earthworm fresh weight of 24.2 g ± 0.3 (mean ± SE) of endogeic Aporrectodea sp. (Aporrectodea trapezoids, Aporrectodea tuberculata, Aporrectodea rosea, 16–17 individuals per mesocosm) and epi-endogeic Lumbricus rubellus (27–29 individuals per mesocosm), corresponding to densities of 87 and 140 individuals m−2, respectively. As a comparison, measured earthworm densities dominated by Aporrectodea sp. frequently exceed 200 individuals m−2 in nearby (<12 km) arctic earthworm invasion gradients27. One day prior to worm addition, all mesocosms were watered with 10 L each (corresponding to 51 mm precipitation), to prevent desiccation of earthworms directly after release. To prevent earthworms from migrating into mesocosms designated as earthworm-free controls, we assigned whole raised beds as either earthworm addition or controls resulting in four raised beds with and four without earthworms.
In contrast to invasive earthworms populations currently thriving in the local sub-arctic environment27, the earthworms in the mesocosm were unable to migrate below a soil depth of 30 cm. Given that no behavior based strategies could be adopted to cope with the low winter temperatures (down to −15 °C in the mesocosm) and soil frost in the 0.3 m deep mesocosms (soil frost extended down to 1 m during the first winter according to data from Abisko research station, measured <100 m from the experiment), eradication of the earthworm population during winter was considered unavoidable. Therefore, addition of earthworms was repeated in the following season (2018) to assure presence of living individuals. Here, we added earthworms to each mesocosm, corresponding to a fresh weight of 11.7 ± SE 0.3 g on 16 June 2018, to compensate for winter mortalities.
Addition of 15N labeled litter
To assess eventual impact of plant uptake of N from the litter layer, we introduced 15N labeled coarsely ground plant material (12 g) into half of the mesocosms on 19 June 2017 to isotopically spike the litter layer, while the remaining mesocosms received similarly ground but unlabeled litter (12 g). The litter added was a mixture of plant litter from meadow and heath vegetation with very similar species composition as the one used in this study43. The added litter had an atom % 15N of 0.36 ± 0.00 (unlabeled control, average ± SE) and 0.73 ± 0.00 (labeled litter), and a δ15N of −4.89 ± 0.31 (unlabeled control) and 984.44 ± 12.11 (labeled litter). Besides the 15N composition, there was no major difference in the N content between the two litter types (unlabeled: 0.97 ± 10% and labeled 0.88 ± 0.01%).
Plant–root-simulator probes
We used PRSTM soil probes (Western Ag Innovations Inc., Saskatoon, Canada), commercially manufactured ion-exchange membranes, to assess availability of inorganic forms of nitrogen (NO3–N and NH4–N, as well as Ca, Mg, K, P, Fe, and Mn). We installed two pairs in each mesocosm which were pooled before extraction. PRS-probes were buried from 16 June 2017 to 3 October 2017 at a depth of 4–10 cm.
Soil sampling
We collected soil samples to analyze the effects of earthworms on the microbial community composition on the 18 August 2017. From each mesocosm, a composite soil sample was taken by coring to 10 cm depth (four cores, corer diametre 1 cm), including the litter layer but removing the mineral soil horizon and pooling these samples together per mesocosm. In the laboratory, plant roots and litter were manually removed from each composite sample, a sub-sample was analyzed for soil moisture (105 °C, 18 h) and organic matter content (475 °C, 4 h) and the remaining soil sample was frozen (−20 °C) within four days of sampling and consequently freeze dried for PLFA analyses, conducted at the Biogeochemical Analyses Laboratory at the Swedish University of Agricultural Science, Umeå. PFLA analyses are reported in nmol/g dry soil.
Plant foliar nutrient content analyses
We collected green foliage from two to three plant species in each mesocosm at the end of the growing season (21 August 2017). From the meadow mesocosms, we collected the leaves of the two forbs that were present in the highest fraction of the meadow mesocosms; i.e., S. alpina and B. vivipara. From the heath mesocosms, we collected the leaves and stems of two dwarf shrub species, which were found in the highest fraction of the heath mesocosms, i.e., V. vitis-idaea and V. myrtillus. From both vegetation types, we collected the leaves of a jointly common graminoid; F. ovina. Thereby, our leaf samples at each vegetation type represented plant species constituting on average 60% of the vegetation cover. Since present in both vegetation types, F. ovina allowed comparison of plant N acquisition between heath and meadow without the interference caused by interspecific variation. Within each mesocosm, we collected ≥ten leaves of randomly distributed plants with fully expanded green leaves. All collected material was immediately air-dried ≥22 °C. Leaves were sorted from stems (V. myrtillus and V. vitis-idaea), dried (approximately 48 h, 60 °C) and milled into a fine powder (Bertin precellys 24). The ground materials were subsequently analyzed for total N and δ15N (Isotope ratio mass spectrometer DeltaV, Thermo Fisher Scientific, Bremen, Germany; and Elemental analyzer Flash EA 2000, Thermo Fisher Scientific, Bremen, Germany).
Aboveground plant sampling
In 27 July 2017, a vegetation and litter survey was conducted by the point intercept method50 in all mesocosms. We used two 50-cm wide rows of ten vertical pins at every 10 cm and counted the total number of times each species was intercepted by 20 pins. The total number of hits was normalized to hits per 100 pins in each mesocosm. This data were further used to calculate the abundance of plant functional groups (graminoids, forbs, evergreen, and deciduous dwarf-shrubs, mosses, and lichens) at the mesocosm level. For each mesocosm, we also measured NDVI (Normalized Difference Vegetation Index) from 0.5 m above each mesocosm using a hand-held pole and two channel sensors (SKR 1800D/SS2, SKL925 logger, SpectroSense2, Sky Instruments, Llandrindon Wells, Wales UK). This index measures the differences between near-infrared light (which vegetation strongly reflects) and red light (which vegetation absorbs), where higher values are indicative of higher photosynthetic activity for the plant community (used as a measure of vegetation greenness).
As an additional proxy for plant growth, we measured the length of the highest individuals and floral shoot numbers of F. ovina, and D. flexuosa (species that occurred in most of the mesocosms) in late autumn of 2018.
Below ground plant production
Total fine root length and fine root growth were measured on 13 June, 7 August, and 25 September 2017 using minirhizotrons (Bartz Technology Corporation, Carpinteria, CA, USA), a nondestructive method for observing temporal development of fine roots. These are transparent tubes with an inner diameter of 5 cm, which were placed horizontally at a depth of 10 cm into the mesocosms during their initial installation.
Nitrogen in earthworm products
To assess mineralization and nitrification rates in earthworm casts in comparison to litter and humus we conducted a water leaching experiment. All material was collected in situ from the earthworm invasion gradient at Jiebren, situated <10 km from the mesocosm experiment, described in detail elsewhere27. Fresh excreted casts were gathered from collected earthworms (Aporrectodea sp.) excreting cast directly into vials, while aged earthworm casts deposited during the season on top of litter was collected manually from the soil surface of forb and dwarf shrub vegetation communities growing under sparsely growing sub-alpine birch forest. Litter and humus (sieved through a 2 mm mesh) from meadow and heath vegetation communities was obtained from uninvaded soil behind a stream functioning as a migration barrier27. Sampled material was homogenized and subsectioned. One sub-sample was shaken in Milli Q water (sample v-water v, 1:5) for 2 h, centrifuged at 4000 rpm and the supernatant filtered through a 0.2 µm filter to remove bacteria. Sample was directly frozen after filtering and analyzed for inorganic N using a segmented flow analyzer (QuAAyro39, Seal Analytica) and total N was measured using a Formacs HT-I (Skalar) with a mounted nitrogen detector (ND 25). Organic N was calculated by subtracting the inorganic N from the total N.
Statistics
The effect of earthworms, vegetation type and 15N labeling of litter (labeled and unlabeled) on plant N concentrations, plant δ15N, soil nutrient availability, abundance of bacteria, fungi, litter, and plant functional types and height of graminoids was tested with a linear mixed effect model using the lme command from the nlme package51 within the statistical environment R (R Core Team 2018). Earthworm treatment, vegetation type, and 15N labeling was treated as fixed categorical factors and block was treated as a categorical random factor to account for the spatial autocorrelation within blocks. Heteroscedasticity was tested by inspecting residuals and data was log transformed in a few cases to achieve homoscedasticity. The number of floral shoots was tested using the glmmPQL function in the statistical package MASS (Venables & Ripley 2002) within the statistical environment R (R Core Team 2018).
Source: Ecology - nature.com
