The Isotopic Signature of the “Arthropod Rain” in a Temperate Forest: Wingless Are From Crowns, Winged Are From Soil

Invertebrate phyto-, sapro-, and microbophages, as well as predators and parasitoids, densely populate tree crowns. Eventually, all crown inhabitants fall from the trees and become a food source for litter-dwelling predators, scavengers, and saprophages. However, the functional signicance of the arthropod rain, i.e., the ux of invertebrates falling from the crowns, remains unexplored. We collected arthropod rain in a temperate mixed forest throughout the growing season. The δ 13 С and δ 15 N values of the arthropods (730 samples in total) were compared to a large reference dataset of the isotopic composition of soil animals from temperate forests. The most numerous taxa in the arthropod rain were collembolans and mites. The most diverse orders were Diptera (18 families) and Coleoptera (29 families), which formed the major portion of the winged specimens. The total ranges of δ 13 С and δ 15 N values of individual animals forming arthropod rain reached 14 ‰ and 26 ‰ , respectively. Nevertheless, invertebrates forming arthropod rain were on average depleted in 13 C and 15 N by 1.6 ‰ and 2.7 ‰ , respectively, compared to the soil-dwelling animals. This difference can be used to detect the contribution of the arthropod rain to detrital food webs. Low average δ 13 С and δ 15 N values of the arthropod rain were primarily driven by the presence of microphytophages, represented mainly by Collembola and Psocoptera, and macrophytophages, mainly aphids, caterpillars, and heteropterans. Furthermore, wingless arthropods were depleted in heavy isotopes relative to winged specimens. Among wingless invertebrates, predators and parasitoids differed signicantly in δ 15 N values from phytophages and microbi/saprophages. In contrast, there was no consistent difference in δ values between saprophages and predators among winged insects, all of them being similar in the isotopic composition to soil-dwelling invertebrates. This result suggests that winged insects in the arthropod rain represented a random assemblage of specimens originating in different biotopes, but most were tightly linked to soil food webs. Overall, our data suggest that invertebrates falling from the crown space and ying arthropods originating from the soil are an important channel connecting food webs in tree crowns and in the soil.


Introduction
Numerous feedback mechanisms link the above-and belowground components of ecosystems (Scheu 2001; Wardle et al. 2004; Van Der Putten et al. 2009). For instance, aboveground consumers of higher trophic levels receive energy subsidies from detrital food webs, which allows for a larger population of aboveground invertebrate generalist predators (Miyashita et al. 2003; Moore et al. 2004;Haraguchi et al. 2013). This phenomenon subsequently impacts plant productivity by controlling the populations of phytophages (Halaj and Wise 2002). On the other hand, detrital food webs in soil generally based on dead plant tissues can be subsidized by other sources of nutrients and energy from above ground. In forest ecosystems, soil predators such as ants often forage in the canopy (Zakharov 2012;Iakovlev et al. 2017), but the "arthropod rain," i.e., various animals and their derivatives falling from the crowns (Chan et al. 2008), may be a more important link connecting soil food webs and the canopy layer. However, only a few studies have thus far addressed the contribution of above-ground animals (likely belonging to grazing food webs) to belowground detrital food webs, although above-ground prey providing a subsidy to soil predators is likely to be a widespread phenomenon (Pringle and Fox-Dobbs 2008).
Tree crowns are populated by phytophages and sapro-and microbophages, as well as predators and parasitoids. Eventually, all types of crown inhabitants fall from the trees and become a food source for litter-dwelling predators, scavengers, and saprophages. Our rst arthropod rain study in a temperate forest (Rozanova et al. 2019) showed that the rain was taxonomically diverse: it comprised approximately 15 orders and more than 60 families of invertebrates. Some of these animals are slowmoving and defenseless and can be easily captured by soil predators. Moreover, approximately 28% of the total mass of arthropod rain is formed by dead animals and exuviae; this fraction can be consumed by saprophages such as collembolans. A few studies using stable isotope labeling of tree crowns con rmed these suggestions: invertebrate predators and saprophages inhabiting the upper layer of litter obtained the crowns' isotopic label more often than the inhabitants of lower litter layers and soil (Goncharov et al. 2016;Potapov et al. 2016b).
Theoretically, the contribution of the arthropod rain into the diet of soil-dwelling predators and saprophages can be estimated using the intrinsic difference in the isotopic composition of "green," i.e., grazing, and "brown," i.e., detrital, food chains (Hyodo et al. 2010). It is well known that members of detrital food webs are 2-5‰ enriched in 13  Furthermore, microarthropods inhabiting the canopy are likely microphytophages feeding on algae, lichens, and mosses (Potapov et al. 2018 (Potapov et al. 2016a). This depletion should result in the relatively low δ 15 N values in the arthropod rain. In contrast, most soil animals are strongly enriched in 15 N compared to phytophages due to the accumulation of heavy nitrogen in microbial biomass at basal levels of detrital food webs (Potapov et al. 2019). Overall, due to the prevalence of macrophytophages and microphytophages in the arthropod rain, it can be expected to be depleted by 2-3‰ in 13 С and 15 N content relative to the animals belonging to detrital food webs in soil.
Although a considerable difference in resource base (and consequently in the isotopic composition) of the crown fauna and soil-dwelling species is expected, some members of the arthropod rain would, in fact, belong to detrital food webs. First of all, these are winged insects that could spend their early life stages or feed in the soil (Wallwork 1970). Second, besides winged insects, actively moving (climbing) specimens of wingless macrofauna (mostly predators like harvestmen, spiders, ants) move freely between the litter and the tree canopy connecting belowground and aboveground food webs (Scheu 2001;Zakharov 2012). Finally, detritus and small-scale detrital food webs can be quite abundant in the canopy, supporting a relatively rich fauna of typical microbivores and detritivores such as oribatid mites or collembolans (Behan-Pelletier and Winchester 1998; Yoshida and Hijii 2011).
Thus, the study of the isotopic composition of the arthropod rain would contribute to elucidating trophic relationships of its constituent invertebrates. Furthermore, it would allow us to assess the possibility of evaluating the contribution of arthropod rain to the nutrition of soil invertebrates and clarify the nature of the "detrital shift" in soil food webs.
In this study, we estimated the stable isotope composition of invertebrates forming arthropod rain in a temperate forest. Our main goal was to compare δ 15 N and δ 13 C values of the arthropod rain with those of soil-and litter-dwelling invertebrates. We hypothesized that (1) invertebrates forming arthropod rain are depleted in 13 C and 15 N compared to the soil-dwelling animals. We further proposed that (2) winged insects in the arthropod rain have on average higher δ 13 C and δ 15 N values than wingless invertebrates because the former are likely to have tight trophic connections with soil and detrital food webs.

Study site and sampling
The study was conducted in two forested plots near Malinky Biological Station (Moscow region, Russia, 55°27'42" N, 37°11'10" E). The rst plot was located in a mixed forest with spruce (Picea abies L.) and lime (Tilia cordata Mill.) forming the upper canopy. The second plot was situated nearby in a ca. 50-yearold pure P. abies plantation. The two forest plots represent typical forest types in the central part of European Russia.
The arthropod sampling was conducted using six custom-made traps in each of the two plots. The traps were open for 24 hours once every two weeks (± three days) throughout the growing season from May to October 2017 (12 samplings in total). Further details are given in Rozanova et al. (2019). Collected arthropods were preserved in 75% ethanol and subsequently identi ed to the order or family level. In addition to arthropods, fresh leaf litter and pollen were collected from the traps. Other sampled substrates included soil (upper 5 cm) and mixed leaf litter from the soil surface. Lichens, bark, tree branches, green leaves, and needles were sampled at different heights of the canopy trees (from 1 to 20 m) in July 2017. Isotopic composition and the number of replications for these substrates are given in Table S1. Five samples of soil and litter (25 x 25 cm, 10 cm deep) were taken in each study plot, and soil macrofauna was extracted by hand-sorting. These soil-dwelling animals were subsequently used for validating the reference dataset (see below).

Stable isotope analysis
All materials were dried at 50°C for at least 72 h. Identi ed animals were weighed (dry wt.) individually or in batches of several conspeci cs using a Mettler Toledo MX5 microbalance with 2 µg accuracy. For the isotope analysis of macrofauna, legs and/or head capsules of large arthropods were used (Tsurikov et al. 2015). Small animals were analyzed alone or as a group of several individuals from the same taxonomic group (minimum sample weight was approximately 50 µg). Soil and plant materials were ground to powder using an MM200 ball mill (Retsch, Germany). Stable isotope analyses were performed using a Flash 1112 Elemental Analyzer (Thermo Fisher, USA) and a Thermo Delta V Plus isotope ratio mass spectrometer in the Joint Usage Center "Instrumental Methods in Ecology" at the A.N. Severtsov Institute of Ecology and Evolution, Russian Academy of Sciences, Moscow. The carbon and nitrogen isotopic compositions were measured as deviations from the international standards (Vienna Pee Dee belemnite and atmospheric N 2 , respectively) and expressed in conventional δ values (‰): δX = [(R sample /R standard ) -1] * 1000, where X is the element of interest (carbon or nitrogen), and R is the molar ratio of its heavy and light isotopes. The standard deviations of δ 15 N and δ 13 C values in laboratory standards were < 0.15‰.
Preservation of sampled animals in 75% ethanol could affect their isotopic composition. In particular, the δ 13 С values could slightly increase due to the washing out of 13 С-depleted lipids. However, the effect of ethanol preservation is typically small and does not exceed 1‰ (Ventura and Jeppesen 2009; Krab et al. 2012). We, therefore, did not apply any correction related to the preservation of samples.

Data analysis
The stable isotope composition of arthropods and litter collected at two experimental plots did not differ. Furthermore, we did not detect signi cant changes in the isotopic composition of animals during the growing season (data not shown). All materials collected were therefore analyzed together. Local δ 13 C and δ 15 Table  S1): ΔX normalized = δX arthropod -δX litter Dry-mass weighted mean δ 13 С and δ 15 N values of the arthropod rain were calculated using the following equations: i.e., mass proportion of the individual samples in a total ux, m i − dry mass of i-th of n samples from a group of the arthropod rain and δX i -isotope signatures (δ 13 С or δ 15 N) of the individual sample.
The trophic position of individual taxonomic groups of arthropods was derived based on the known morphological and behavioral traits (Striganova 1980) and isotopic studies (Potapov et al. 2019). Some groups of arthropods that have fallen into our traps can move between the forest canopy and the litter layer. These are winged animals (the imago stage of various insects) and actively moving wingless predators, such as ants, harvestmen, and spiders. We, therefore, separated winged animals and predators from other arthropods.
We compared the isotopic composition of the arthropod rain (original data, n = 730) to that of soil invertebrates using the results of 23 studies performed in temperate forests compiled in Potapov et al. contains averaged values rather than individual measurements, it fully re ects data on the isotopic composition of soil and litter-dwelling mesofauna and macrofauna in the vicinity of the Malinky Biological Station (Fig. S1). In addition, we compared the isotopic composition of high-rank taxonomic groups that were well represented (more than 30 measurements) in the original and published datasets (Diptera, Coleoptera, Collembola, and Araneae).
The isotopic composition of arthropod rain and soil-dwelling invertebrates was compared using standard ellipses limiting the 95% con dence interval (Jackson et al. 2011 in Psocoptera to 18.7‰ in imago Figitidae, Hymenoptera), respectively. The δ 13 С and δ 15 N values of excrements (frass) and exuviae were within the range of arthropod δ values (Fig. S2).
The previously reported ranges of mean litter-normalized δ 13 C and δ 15 N values of soil invertebrates in temperate forests (n = 1300, Potapov et al. 2019) were similar to those of the arthropod rain. In the isotopic bi-plot ( Fig. 1) nearly all soil invertebrates are found within the convex hull formed by the arthropod rain. Standard ellipses of the arthropod rain and soil invertebrates largely overlapped, but the arthropod rain ellipse was shifted towards lower δ 13 C and δ 15 N values. Therefore, mean δ 13 C and δ 15 N values of soil invertebrates (4.0 ± 0.1‰ and 4.2 ± 0.1‰, respectively) were signi cantly higher than those of the arthropod rain (2.4 ± 0.1‰ and 1.5 ± 0.1‰, respectively) (Mann-Whitney test, P < 0.05). Due to the contribution of large specimens of spiders and especially harvestmen, weighted mean δ 13 C and δ 15 N values of the total arthropod rain were somewhat higher (2.7‰ and 2.4‰, respectively) ( Table 1). Table 1 Isotopic composition (litter-normalized Δ 13  As a rule, individual taxonomic groups forming arthropod rain had δ 13 C and δ 15 N values lower than those of animals collected from soil and litter. Signi cant differences both in δ 13 C and δ 15 N values between specimens originating from the arthropod rain (original data) and the soil (published data) were found in Diptera, Collembola, and Araneae (Fig. 2). Coleoptera from the arthropod rain were signi cantly depleted in 15 N but enriched in 13 C compared to soil-dwelling animals.
Wingless and winged invertebrates within the arthropod rain differed considerably in isotopic composition. Wingless arthropods were depleted in heavy isotopes relative to winged specimens ( Table 1). The differences in both δ 13 C and δ 15 N values were signi cant. Among wingless invertebrates, predators and parasitoids differed signi cantly from phytophages and microbi/saprophages (Fig. 4A). In contrast, there was no difference in δ values between saprophages and predators in winged insects, all of them being similar in the isotopic composition to soil-dwelling invertebrates (Fig. 4B). . As a note of caution, it should be stressed that sizes of the standard ellipses re ecting "isotopic space" of soil animals and arthropod rain (Fig. 1) cannot be compared directly, as they were based on species means and individual measurements, respectively. Nevertheless, their centroids can be accurately compared. This accuracy is con rmed by a close similarity in the isotopic signatures of soil macro-and mesofauna collected in this study and those represented in the reference dataset (Fig. S1).

Discussion
Consistent with our rst hypothesis, invertebrates forming arthropod rain were on average depleted in 13 C and 15 N compared to the soil-dwelling animals. Preservation of the arthropod rain invertebrates in 75% alcohol could not affect this conclusion since the expected change in 13 С content due to leaching of lipids (Post et al. 2007) would increase, rather than decrease, the δ 13 С value of the arthropod rain.
The difference in the isotopic composition of soil animals and arthropod rain was mainly driven by the presence in the latter of a signi cant proportion of macro-and microphytophages with relatively low δ 13 (Fig. 3), consistent with the difference in isotopic signatures of their basic trophic resources: non-vascular plants, such as algae and lichens, and fresh leaves, respectively. Indeed, the difference in Δ 15 N values between micro-and macrophytophages roughly corresponded to the difference between crown lichens and fresh leaf litter (Table S1). These data corroborate previous reports on the importance of non-vascular plants in forest food webs (Potapov et al. 2018).
Microphytophages depleted in 15 N were represented mainly by Psocoptera and Collembola (Table S1). Psocoptera grazing on epiphytes are typical components of crown fauna (e.g., Southwood et al. 2005), while Collembola are usually regarded as typical soil animals feeding predominantly on fungi.
Nevertheless, feeding of collembolans on 15 N-depleted lower plants has been repeatedly noted. According to Potapov et al. (2016), at least 20% of collembolan species in temperate forest soils are depleted in 15 N relative to litter, suggesting they are trophically linked to non-vascular plants, predominantly algae (Chahartaghi et al. 2005;Potapov et al. 2018). Thus, even in the soil, there are many phycophagous collembolans, but in the crowns, microphytophagy is apparently more widespread, as suggested by signi cantly lower δ 15 N and δ 13 С values in the Collembola from the arthropod rain than in soil-dwelling Collembola (Fig. 2).
Among other groups of arthropods well represented in both datasets, Diptera and Araneae were depleted in 13 С compared to soil-dwelling animals. This observation further con rms that the "detrital shift," i.e., enrichment of detrital food webs with 13 С due to interactions with saprotrophic microorganisms (see Potapov et al. 2019 and references therein) can be traced both in micro-and macroarthropods and also at higher trophic levels. Coleopterans did not follow this pattern (Fig. 2) likely because they were represented mainly by winged imagoes (Table S1).
Dead stems and branches, bark crevices, suspended litter and soil support a substantial amount of detritus in the crown space, which in turn harbors rich fauna of detritophagous arthropods (Winchester et al. 1999;Grove 2002). Thus, the detrital shift can be expected and was observed in the canopy food webs (e.g., Erdmann et al. 2007). Nevertheless, the isotopic signature of non-winged specimens, which presumably fed in the crowns, suggests that the effect of the detrital shift in crown fauna was considerably less pronounced than in the soil food webs (Fig. 4А). Furthermore, soil-dwelling taxa associated with mineral soil that are the most enriched in 13  On the other hand, a large range of δ 13 С values in macrophytophages (ca. 8‰, Fig. 3B) can be related to the "canopy effect," i.e., a gradient in the concentration of 13 C in green leaves growing at different heights (Brooks et al. 1997). Therefore, phytophages that consumed green parts of vascular plants at different canopy heights could differ greatly in isotopic carbon composition.
As suggested by our second hypothesis, decreased δ 13 С and δ 15 N values were typical of wingless arthropods, while winged insects collected in the traps hardly differed in the isotopic composition from soil animals ( Table 1, Fig. 4B). Another important feature of winged insects was the lack of difference between predators and phytophages or microbi/saprophages, while in the wingless arthropods, this difference was pronounced (Fig. 4). This observation con rms that winged insects collected in the traps represented a random assemblage of specimens originating in different biotopes or local ecosystems. Nevertheless, isotopic signatures of the winged insects suggest that they mostly originated from the soil. This localization is especially true for Diptera and Coleoptera (Table S1) that often have litter-dwelling larvae (Wallwork 1970

Conclusions
Overall, our data suggest that invertebrates falling from the crown space and ying arthropods originating from the soil are an important channel connecting food webs in the crown and the soil. Due to the large contribution of micro-and macrophytophages, the fraction of the arthropod rain consisting of wingless specimens differs considerably in δ 13 C and δ 15 N values from soil invertebrates belonging to detrital food webs. This difference can be considered in further studies of the contribution of crown arthropods to the nutrition of soil predators in forest ecosystems.
(gray triangles and solid line, published data, points represent the mean value of species in different ecosystems, n = 1300). Standard ellipses show the 95% con dence interval.

Figure 2
Stable isotope composition of arthropods from soil and litter (Potapov et al. 2019, gray) and the arthropod rain (original data, green). Litter-normalized Δ13C and Δ15N are shown as violin plots (mirrored kernel density estimation) with medians and 2.5-97.5% percentiles. Only groups represented by more than 30 data records are shown, and the number of samples is given in brackets. The difference within pairs was signi cant in all cases (Mann-Whitney test, P < 0.05).

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. RozanovaSMTheisotopicsignature.docx