Spatial Distribution of the Anecic Species of Earthworms Dendrobaena nassonovi nassonovi (Oligochaeta: Lumbricidae) in the Forest Belt of the Northwestern Caucasus

Abstract

The mountain forests of the Northwestern Caucasus represent unique refugia for the native biodiversity of flora and fauna. Endemic species are also preserved among soil invertebrates, including the group of earthworms, which are important ecosystem engineers. This study assesses the spatial distribution of the endemic anecic species of earthworms Dendrobaena nassonovi nassonovi Kulagin, 1889 in deciduous, coniferous–deciduous, small-leaved, and conifer forests of the Northwestern Caucasus (a total of 1028 geographical points were surveyed, of which the species was found in 185 points) based on our own field data by modeling the current potential areas using the Maxent software. The D. n. nassonovi potential area maps show a high probability of the species inhabiting mid- and high-mountain forests of the Northwestern Caucasus and being confined to mixed coniferous–deciduous and beech forests. The optimum soil and climatic parameters, as well as a lack of large-scale business operations in the mid- and high-mountain forests, make it possible for these ecosystems to remain suitable refugia, in particular for the endemic anecic species of Caucasus earthworms D. n. nassonovi.

1. Introduction

Earthworms are ecosystem engineers and the most important detritivores in forest soils. Earthworms play a key role in soil organic matter transformations and nutrient dynamics at different spatial and temporal scales. In terrestrial ecosystems, soil biodiversity promotes multiple functions simultaneously, within which, ecosystem engineers are the main drivers. Earthworms, as ecosystem engineers, create biogenic structures (aggregates, burrows) that may serve as habitats for other species than themselves [1,2]. Therefore, knowledge of the distribution of earthworms in forest ecosystems can serve as a scientific basis for understanding the distribution of other species and groups of soil invertebrates.

Epigeic, endogeic, and anecic ecological types [3], or morpho-ecological groups [4], of earthworms differ in their soil habitats, their functional role in the mineralization of organic residues [5,6], and their impact on soil water regimes [7]. Epigeic and endogeic species groups are usually much more common under current conditions than anecic earthworms, which are more often confined only to well-developed, rich soils [4,8]. At the same time, anecic earthworms in natural forest ecosystems are usually confined to the best-preserved old-growth forests that have not recently experienced large-scale wildfires, logging, or other types of impacts that would violate the forest ecosystem integrity [9].

Anecic earthworms differ from other species groups because of their much larger size (the biomass of an anecic species is much higher than that of an epigeic one), longer life expectancy of up to 9 years compared with the average of 1–2 years in other groups [10], and the fact that they build permanent slightly branched vertical passages to a depth of up to 1 m, which can be clearly shown today by using X-ray tomography methods [7]. Anecic earthworm passages persist significantly longer than the lifespan of an earthworm [11], which, under certain conditions, allows researchers to conduct paleoreconstructions of these animals’ habitats [12]. Anecic earthworms go deep into the ground during periods of adverse conditions (droughts, negative temperatures) and enter diapause. Anecic earthworms exhibit behavioral plasticity and can behave like epigeic earthworms in forest soils at the optimum soil moisture and temperature [8]. Anecic earthworms feed on plant litter and are capable of carrying organic substances rich in carbon and nitrogen in their coprolites to great depths because of vertical migration [13].

The cosmopolite species Lumbricus terrestris L. is the most studied among the anecic earthworms in Russia [14]. A detailed study on the distribution of the Crimean–Caucasian endemic species of earthworms Dendrobaena nassonovi nassonovi Kulagin, 1889 (synonyms: Dendrobaena mariupolienis Wyssotzky, 1898; Allolobophora crassa Michaelsen, 1900; Eophila crassa Michaelsen, 1900) was conducted by E.Sh. Kvavadze [15], a renowned researcher of Lumbricidae fauna in the Caucasus in the 20th century. Later, the data on the distribution of this species were generalized and clarified by T.S. Perel, the greatest taxonomist of earthworms in Russia [16,17,18]. Today, the data on the distribution of the species in the Caucasus, mainly in forest ecosystems, may be found in articles by I.B. Rapoport [19,20,21,22] and A.P. Geraskina [9]. These studies offer a synonym for the species, Dendrobaena mariupolienis, as the one most commonly used in references prior to an update of the checklist of valid species and families of earthworms [23]. In general, D. n. nassonovi is the only representative of the anecic species group in forests of the Northwestern Caucasus [9]. In this article, we supplement information on the distribution of this species with previously unpublished quantitative data obtained by us during field expeditions.

Identifying environmental predictors of species distribution and understanding their distribution is possible using machine learning techniques. Currently, there are various methods or algorithms for modeling species distribution models (SDMs), such as Classification Tree Analysis (CTA), Random Forest (RF), Multivariate Adaptive Regression Spline (MARS), Generalized Linear Models (GLMs), Generalized Boosted Regression Models (GBMs), Maximum Entropy (Maxent), and others. Each of them has its own strengths and weaknesses [24]. Among the “presence-only” models, one of the most widely used models is the Maximum Entropy method, implemented in the Maxent program. The Maxent model is based on machine learning algorithms and has been validated in numerous studies as an effective algorithm for predicting species distribution, including for the study of earthworms [24,25,26,27,28,29].

We believe that the large field material we have accumulated, as well as the use of the Maxent modeling method, will allow us to answer the following scientific questions: what is the current distribution of the anecic species D. n. nassonovi in different forest types of the Northwestern Caucasus and what predictors determine this distribution?

2. Materials and Methods

2.1. Study Area

The materials were collected during expedition routes in the Northwestern Caucasus (Krasnodar Krai, Republic of Adygea, and the Karachay-Cherkess Republic, Russian Federation) in the spring–summer seasons of 2014 through 2019. The research was conducted in the forest belt at altitudes of 42 m to 1965 m a.s.l. The climate in the area is temperate and humid: the average annual temperature is 7.7–11.1 °C above zero; the average temperature in January is 4–5 °C below zero and in July and August about 15 °C above zero; the annual amount of precipitation varies from 500 mm to 3000 mm. Along the routes from the lower to the upper forest border, black alder forests, small-leaved forests (birch, aspen, birch and aspen, hornbeam forests), hornbeam–beech forests (forests with predominant Carpinus betulus L. and Fagus orientalis Lipsky; the stands may also consist of Acer platanoides L., Fraxinus excelsior L., and Cerasus avium (L.) Moench with no coniferous tree species), coniferous–deciduous forests (the stands most often consist of Fagus orientalis Lipsky, Quercus robur L., Acer platanoides L., Fraxinus excelsior L., Abies nordmanniana (Stev.) Spach, and Picea orientalis (L.) Link and less often Cerasus avium (L.) Moench, Tilia begoniifolia Stev., and Taxus baccata L.), and coniferous forests (spruce, fir, fir–spruce, and pine forest) were examined (

Table 1. Main forest types in the Northwestern Caucasus.

Forest Types	Number of Geographical Locations	Number of Soil Samples	Plant Litter Thickness, cm	Soil Moisture, %	Soil pH
Black alder forests (large-fern and tall-grass)	34	102	2–8	45 ± 3	6.0 ± 0.5
Small-leaved forests (small-grass, multi-grass)	305	915	1–3	35 ± 2	5.8 ± 0.6
Hornbeam and beech forests (wood–fern, tall-grass, small-grass, fescue)	280	840	1–2	30 ± 5	5.6 ± 0.2
Mixed coniferous–deciduous forests (small-grass, multi-grass, dead-cover)	281	843	2–4	25 ± 5	5.5 ± 0.1
Coniferous forests (dead-cover, small-grass, green-moss)	281	843	2–4	25 ± 5	5.5 ± 0.1

). The soils are mostly brown earth or alfisols; in terms of granulometric composition, they are most often medium- and light-loamy (in black alder forests, they are heavy-loamy with signs of gleysol) [30].

2.2. Data Collection

A total of 1028 geographic locations were surveyed with at least 200 m distance between the locations. For earthworm records, 3–4 soil samples were collected in each location (single sample size: 25 × 25 cm², depth: 30 cm). Earthworms in deadwood were recorded as well [9]. Soil and deadwood were manually parsed directly in the field. Soil moisture and soil acidity were identified for each sample using a portable pH 300 field indicator (Table 1). D. n. nassonovi was found in 185 geographical locations (Figure 1). Earthworms were fixed in 96% ethanol solution. Sampling and fixation of earthworms in the field were carried out. The biomass of earthworms was determined by weighing individuals with a full intestine.

2.3. Identifying of Earthworms

The species composition was identified in the laboratory using the Russian Cadastre, the earthworm taxonomic key [17], and the supplement to Caucasus fauna [18]; the diagnoses were refined in accordance with the key of earthworms of Turkey [31,32]. Additional details of the areas are provided in accordance with the checklist of the earthworm fauna of Turkey [33] and the combined checklist of earthworms of the Northeastern Mediterranean region [34]. This study presents materials concerning a single species, the Crimean–Caucasian endemic earthworm Dendrobaena nassonovi nassonovi, which belongs to the anecic form.

2.4. Statistical Analysis of Field Data

The non-parametric Kruskal–Wallis test was used when comparing the field data samples to identify significant differences (p < 0.05).

2.5. Environmental Predictors

The study used WorldClim 2 resources [35]: 19 standard annually averaged bioclimatic variables at a 30-arcsecond spatial resolution. Bioclimatic variables were as follows: bio1—average annual temperature; bio2—average monthly temperature; bio3—isothermality (bio2/bio7) × 100); bio4—temperature seasonality (standard deviation × 100); bio5—maximal temperature in the warm months; bio6—minimal temperature in the cold months; bio7—annual amplitude temperature (b5–b6); bio8—average temperature in the most humid quarter; bio9—average temperature in the driest quarter; bio10—average temperature in the warmest quarter; bio11—average temperature in the coldest quarter; bio12—amount of precipitation per year; bio13—amount of precipitation in the most humid month; bio14—amount of precipitation in the driest month; bio15—precipitation seasonality (variation coefficient); bio16—amount of precipitation in the most humid quarter; bio17—amount of precipitation in the driest quarter; bio18—amount of precipitation in the warmest quarter; bio19—amount of precipitation in the coldest quarter.

2.6. Model Building via Maxent and Model Evaluation

We used Maximum Entropy modeling (Maxent 3.4.4: Columbia University, New York, NY, USA) based on a machine learning technique [36]. A total of 185 species observation points were used in the model. To create the Maxent model, full sets of applicable bioclimatic variables and the following parameters were used: characteristics—automatic; output format—cloglog [36]; regularization multiplier = 1. To reduce the risk of misidentification, an upper threshold was chosen to define habitats with a high degree of sustainability [37]. Accordingly, our study used a fixed difference threshold of ≥0.6 (60%–100% probability of detection) for optimal habitats, consistent with a low false-positive rate model [38]. A potentially suitable habitat was achieved using a fixed threshold of >0.4 (probability of presence >40%) [39]. In addition to generating probability distribution maps, Maxent identified environmental predictors for the distribution object. Convenient options for assessing predictors of innovation are relative percentage contributions to the construction of models (percentage contribution, %).

One of the most common estimates of model quality based on the Maxent method is AUC—the area under the spectral characteristic of the receiver (the area under the receiver operating characteristic curve). AUC measures patterns of specificity and sensitivity in distinguishing points from background (random) points [40]; i.e., it shows the number of correctly classified presences and absences [41]. A model is considered good when AUC = 0.7–0.8 and excellent when AUC = 0.8–0.9 and higher [36].

Maps of species distribution were produced with ArcGIS 10.6.1 software package (Environmental Systems Research Institute, Redlands, CA, USA).

3. Results

3.1. Confinement of D. n. nassonovi to Various Forest Types (Field Data Results)

The largest average abundance (14.1 ± 6.0 ind.*m⁻²) and biomass (Figure 2) of the anecic species D. n. nassonovi was were found in mixed coniferous–deciduous forests. These are mainly beech and fir forests that can be admixed with Acer platanoides L., Quercus robur L., Picea orientalis (L.) Link, Tilia begoniifolia Stev., Fraxinus excelsior L., Cerasus avium (L.) Moench, Carpinus orientalis Mill., Pyrus caucasica Fed., and (rarely) Taxus baccata L. Sometimes there is hazel in the undergrowth. The forests are located in automorphic landscape positions and less often on slopes with a height range of 487 m to 1379 m a.s.l. The soils are well developed, thick, and of the alfisol or brown earth type with no traces of wash. The forests often have traces of selection logging and no traces of fires. There is deadwood at the late decomposition stages in these forests. In these forests, anecic earthworms were encountered in an average of 35% of soil samples. In forests with a predominance of deciduous tree species—aside from Fagus orientalis Lipsky, with Acer platanoides L., Fraxinus excelsior L., and Tilia begoniifolia Stev. with a small-grass or multi-grass ground cover—the occurrence of D. n. nassonovi in samples reached 45%, the abundance reached 19.1 ± 5.2 ind.*m⁻², and the biomass was 33.0 ± 3.2 g*m⁻². In fir and beech dead-cover forests, the occurrence of D. n. nassonovi in samples was 25%, the abundance was 9.5 ± 4.8 ind.*m⁻², and the biomass was 18.0 ± 2.6 g*m⁻². Often, earthworms of this species are found in the plant litter where they feed. It is in the plant litter that large mature individuals are often found. Juvenile earthworms are more often found in deeper organogenic horizons. In very rare cases, mature and juvenile individuals have occasionally been found in Tilia begoniifolia Stev., Fraxinus excelsior L., Acer platanoides L., and Fagus orientalis Lipsky deadwood at late decomposition stages or under the deadwood. The species was not found in the deadwood of Abies nordmanniana (Stev.) Spach or Picea orientalis (L.) Link.

In hornbeam and beech forests, the average abundance of D. n. nassonovi was 1.5 times (9.3 ± 4.0 ind.*m⁻²) lower and the biomass (Figure 2) was two times lower than in coniferous–deciduous forests. These forests are equally characteristic of both automorphic and transit landscape positions in a range of altitudes from 200 to 1500 m a.s.l. The soils are usually medium-thick and of the brown earth type. They have no traces of fires; sometimes, there are traces of selective logging. Deadwood is present, but it is often at the initial decomposition stages, so anecic earthworms were not found in it. The occurrence of D. n. nassonovi in soil samples was 20%–25%. Earthworms, both mature and juvenile, are most often found in upper organogenic horizons. Hornbeam and beech small-grass and tall-grass forests showed a significantly higher abundance (12.7 ± 2.1 ind.*m⁻²) and biomass (16.5 ± 4.8 ind.*m⁻²) of D. n. nassonovi (Kruskal–Wallis test, p < 0.05) than hornbeam and beech wood–fern and fescue forests with 6.1 ± 2.0 ind.*m⁻² and 8.3 ± 4.0 g*m⁻², respectively.

In small-leaved forests of Carpinus betulus L., Populus tremula L., and Betula pubescens Ehrh., occasionally mixed with Quercus robur L. and Pinus sylvestris L., the abundance was 6.5 ± 2.5 ind.*m⁻² and the biomass was 7.0 ± 2.5 g*m⁻² (Figure 2). The forests mainly occupy automorphic landscape positions in an altitude range from 340 to 750 m a.s.l. The soils are thick and of the brown earth type. They have no traces of fires; sometimes, there are traces of selective logging. There is little deadwood. Only deadwood of Betula pubescens Ehrh. or Populus tremula L. at the initial decomposition stages is rarely present. The occurrence of D. n. nassonovi in soil samples was at best 15%. Earthworms, both mature and juvenile, are most often found in upper organogenic horizons. There were no statistically significant differences in the abundance, biomass, or occurrence of D. n. nassonovi in samples between small-grass and multi-grass small-leaved forests.

In black alder forests, most often with Carpinus betulus L. and less often with Abies nordmanniana (Stev.) Spach, Acer platanoides L., Quercus robur L., and Fagus orientalis Lipsky, the abundance and biomass of D. n. nassonovi were significantly lower (Kruskal–Wallis test, p < 0.05) than in small-leaved forests, amounting to 4.5 ± 2.0 ind.*m⁻² and 2.5 ± 0.5 g*m⁻², respectively. These forests often occupy accumulative landscape positions and, less often, transit positions on flat slopes located at an altitude of 42 to 1150 m a.s.l. The soils are thick and of the brown earth type, with well-developed plant litter horizons. These forests have no traces of fires, but there are frequent traces of selection logging. There is little deadwood, the most present being hornbeam deadwood at decomposition stages 1–2. The occurrence of D. n. nassonovi in soil samples was 15% in tall-grass black alder forests and no more than 6% in large-fern black alder forests. The tall-grass black alder forests showed an abundance and biomass of 6.5 ± 2.2 ind.*m⁻² and 3.5 ± 0.8 g*m⁻², respectively, whereas large-fern black alder forests showed an abundance and biomass of 2.5 ± 2.0 ind.*m⁻² and 1.5 ± 0.5 g*m⁻². Earthworms, both mature and juvenile, are most often found in the plant litter horizon.

Since there were no differences found in the abundance or biomass of D. n. nassonovi in light coniferous (pine) and dark coniferous (fir–spruce, spruce, fir) forests, they were combined in this study. Anecic earthworms are very rare in these forests (at best in 5% of soil samples), and the abundance (0.5 ± 0.5 ind.*m⁻²) and biomass values were minimal (Figure 2) among the forests surveyed. These forests most often occupy transit landscape positions at an altitude of 1100–1965 m a.s.l. The soils are thin-layered and of the brown earth type, with outings of rocks and stones. Pine forests often have traces of fires on tree trunks and in the soil. The deadwood of Pinus sylvestris L. and Picea orientalis (L.) Link is present at the initial decomposition stages, but no anecic earthworms were found in it. No differences between small-grass, green-moss, or dead-cover coniferous forests for this species were found.

3.2. Modeling of Current Potential Areas of Earthworms

The results of modeling the range of the anecic species D. n. nassonovi show that the potential area covers mainly mid- and high-mountain forests (probability level > 60%) (Figure 3), which corresponds to the best-preserved old-growth coniferous–deciduous and beech forests.

An analysis of the contribution of bioclimatic parameters shows that the amount of precipitation in the driest month (43.1%) is the most significant parameter for this species. The annual mean temperature (18.5%) and the precipitation of the warmest quarter (11.8%) are also significant. The Jackknife test also allows precipitation to be added in the wettest quarter of the year. This means that precipitation levels during the warm season are especially important for Dendrobaena nassonovi nassonovi. Other bioclimatic parameters related to temperature and humidity did not show significant contributions to the modeling of the potential area for this species.

The Maxent model performance was almost perfect (especially taking into account a few samples) because the AUC value for 25 replicates equals 0.977 (Figure 4).

4. Discussion

The biota of the Caucasus developed during the Cenozoic, and the formation of its current soil fauna, as well as the entire fauna, began in the Oligocene. In the Paleogene, most of the Lesser Caucasus was under the Tethys Sea, but numerous islands merged in its place, and by the end of the Oligocene, the land associated with the Asia Minor plate was formed. In the Paleocene, there was a mountainous island in the area of the Greater Caucasus, covered with moisture-loving vegetation of a tropical type. The islands were likely inhabited by the earthworm species that gave rise to the Lumbricidae fauna of the Greater Caucasus. The insular isolation of the Greater Caucasus is associated with the occurrence of Caucasian lumbricids of the genus Dendrobaena, including the anecic species Dendrobaena mariupolienis [15] or, according to contemporary taxonomy, D. n. nassonovi.

Anecic earthworms, as ecosystem engineers, play several critical roles in forest soils that are not duplicated by other organisms, including bioturbation and creating vertical systems of burrows and cavities. A vertical branched system of passages contributes to the flow of air, water, and dissolved substances into deep soil horizons [7,42]. Earthworms use these passages for several years, so any soil disturbances like wildfires, plowing, overgrazing farm animals, and the use of equipment in logging result in disruptions of local earthworm habitats and high mortality in juvenile individuals because earthworm cocoons are deposited in the upper soil horizons that are most vulnerable to external influences. Our results show that the Crimean–Caucasian endemic species of earthworms D. n. nassonovi is highly confined to the mountain–forest belt of the Northwestern Caucasus, which is confirmed by both our own field data and data collected by other authors [19,20]. A lack of large-scale wildfires, clear logging, plowing, and overgrazing [43] contributes to the preservation of this endemic species in the mid- and high-mountain forests of the Caucasus. On the contrary, the impact of all said factors on the plain land most likely limits the distribution of the species; according to the literature data, other representatives of anecic species, such as the well-known cosmopolitans L. terrestris and Aporrectodea longa, can only be found there occasionally [19,44].

According to our field data, as well as the modeling results, D. n. nassonovi is highly confined to coniferous–deciduous forests, especially forests with predominant deciduous tree species (maple, linden, ash) with an admixture of conifers (fir). It is likely that the optimum trophic and topical qualities of plant litter are developed in these forests, which is important for anecic earthworms not only as a trophic resource but also as a horizon of temporary habitat. Despite the high secondary metabolite content in beech litter, this species is also confined to old-growth beech forests. It has been shown that large- and medium-sized soil decomposers, such as Oligochaeta and Isopoda, may have a stronger tolerance to some chemical defense substances and stronger adaptability to different environments [45]. In addition, species conservation is facilitated thanks to the deadwood of deciduous tree species at late stages of decomposition in these forests, which earthworms can also temporarily inhabit. A decrease in the abundance and biomass of the anecic species was found in the following order of forests: mixed coniferous–deciduous—hornbeam and beech—small-leaved—black alder—coniferous (both light and dark coniferous). This decrease may be associated with both the deterioration of the trophic and topical properties of plant litter and soil moisture dynamics. Both low soil moisture in coniferous forests (especially in pine forests) and waterlogging in black alder forests (Table 1) may serve as an unfavorable factor for anecic species, which prefer moderately moist, well-drained soils [15]. At the same time, a positive correlation between the biomass of the anecic species D. n. nassonovi and the litter thickness (R² = 0.89) was revealed, which indirectly confirms the role of this species in the regulation of litter stocks. A negative link with soil acidity was found (R² = −0.72). Earthworms prefer less acidic soils, and calcium accumulation in their calciferous glands is also known to reduce the acidity of processed organic matter [46].

Among the bioclimatic parameters, the amount of precipitation in the driest month is the most significant, which confirms the high dependence of the species on moisture. The contribution of temperature was less significant, probably because anecic earthworms, thanks to a system of deep vertical passages, can avoid critical temperature deviations in both the winter and summer seasons (anecic earthworms are known to experience summer and winter diapauses).

Consequences and Limitations of the Maxent Method

The Maxent modeling method was used for a deeper analysis of our field data, making it possible to analyze field observations in more detail and identify the most favorable forest types and the main environmental predictors of the distribution of the endemic earthworm species D. n. nassonovi. Although Maxent is still a relatively new method that has not yet been fully developed, Maxent modeling has frequently outperformed a number of other approaches that rely on presence-only data [47]. Because it relies only on presence data, it lacks many of the complications associated with presence–absence analytical methods [27,36]. However, the limitations of the method must be respected, and improvements are still required in steps such as developing a general method for establishing threshold levels, developing a methodology for selecting the best approximating model, and establishing a protocol for assessing habitat selection based on the repeated sampling of individuals [24,47]. The limitations of the model must be taken into account when interpreting the results obtained.

5. Conclusions

The results of modeling the potential range of D. n. nassonovi show a high probability of this species inhabiting mid- and high-mountain forests of the Northwestern Caucasus and confining itself to mixed coniferous–deciduous and beech forests. The field data also show the high-scale biotopic confinement of the anecic earthworm species D. n. nassonovi to coniferous–deciduous and deciduous forests because of its biological features. Moreover, given the absence of large-scale anthropogenic activities in mid- and high-mountain forests that could violate natural habitats, said areas remain suitable refugia for the endemic anecic species of earthworms of the Caucasus. The protection of endemic species among ecosystem engineers in particular can provide a scientific basis for the conservation of these forests.