Soil ecosystem engineering by the train millipede Parafontaria laminata in a Japanese larch forest

ARTICLE IN PRESS

Soil Biology & Biochemistry 38 (2006) 1840–1850 www.elsevier.com/locate/soilbio

Soil ecosystem engineering by the train millipede Parafontaria laminata in a Japanese larch forest Ayu Toyota,1, Nobuhiro Kaneko, Masamichi T. Ito Soil Ecology Research Group, Graduate School of Environment and Information Sciences, Yokohama National University, Yokohama, 79-7 Tokiwadai, Yokohama 240-8501, Japan Received 15 July 2005; received in revised form 21 November 2005; accepted 6 December 2005 Available online 2 March 2006

Abstract Periodic swarming by adult train millipedes Parafontaria laminata (Attems, 1909) occurs in central Japan on an 8-year cycle, and the emergence of new adults is highly predictable. Millipede biomass reaches a maximum and feeding habits change upon the emergence of adults. Larvae are geophagous while adults feed on both litter and soil. We hypothesized that the shift in the developmental stages of P. laminata influenced the carbon dynamics in the soil and conducted a field mesocosm experiment in a larch plantation forest over 2 years (1999 and 2000) using three developmental stages: sixth- and seventh-instar larvae and adults. By experimentally manipulating millipede density at four levels, we obtained the following results: larvae were geophagous, while adults consumed both litter and soil (mixedfeeding) and consequently showed stronger density effects on litter decomposition rates than larvae; adult activities in the high-density treatment increased soil microbial biomass but not at low adult densities or at the larval stages; and adults increased the carbon accumulation in soil layers especially at high densities due to their mixed-feeding on litter and soil. We determined that due to synchronized postembryonic development with high densities and changes in feeding habits, the train millipede periodically sequestered carbon in this forest. r 2006 Elsevier Ltd. All rights reserved. Keywords: Diplopoda; Feeding shift; Litter decomposition; Soil carbon sequestration; Soil microbial biomass; Soil organic matter

1. Introduction Most terrestrial net primary production ultimately enters the soil subsystem (Cebrian, 1999) where decomposition processes occur. Global soil carbon storage in the upper 2 m of mineral soil is estimated to be about three times that of atmospheric carbon (Ko¨rner, 2000), and thus soil is the largest carbon reservoir in terrestrial ecosystems. Soil carbon influx and efflux is a major concern in current atmospheric carbon enrichment due to anthropogenic activities (e.g., Wang and Hsieh, 2002). Carbon accumulation in forest soils is a long-term process (e.g., Kaye et al.,

Corresponding author. Tel.: +81 45 339 4358; fax: +81 45 339 4379.

E-mail address: [email protected] (A. Toyota). Present address: Japan Wildlife Research Center, 3-10-10 Shitaya, Taito-ku, Tokyo 110-8676, Japan 1

0038-0717/$ - see front matter r 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.soilbio.2005.12.015

2003; Vitousek, 2004). Turnover times of soil organic matter are estimated 300–175 000 years (Torn et al., 1997). Organic matter on the forest floor forms part of the labile carbon pool (Green et al., 1993; Currie, 1999), which is transferred into the soil or atmosphere by the activity of soil microbes and fauna (Petersen and Luxton, 1982). Studies of various ecosystems have indicated that soil invertebrates mediate about 15% (Anderson, 1995) and 7–13% (Schro¨ter et al., 2003) of the carbon turnover. The contribution of soil invertebrates to decomposition depends on their feeding habits. Earthworm feeding types have been divided into litter-feeding (humus formers) and soil-feeding (humus eaters), although gut-content analyses have revealed that litter feeders actually eat litter and soil (Judas, 1992; Uchida et al., 2004). Litter feeders increase the surface area of organic matter by fragmentation of litter, whereas soil feeders mix organic matter with mineral soil (Lavelle et al., 1997) and thus promote organo-mineral

ARTICLE IN PRESS A. Toyota et al. / Soil Biology & Biochemistry 38 (2006) 1840–1850

mixtures as soil aggregates. Wolters (2000) reviewed invertebrate effects on the stabilization of soil organic matter, and simulations of carbon dynamics in earthworm casts have predicted long-term stabilization of carbon in agricultural soils (Lavelle et al., 1997; Wolters, 2000; Lavelle and Spain, 2001). Geophagous animals such as earthworms and termites mold the soil, thereby enhancing the accessibility of organic matter to microorganisms; therefore, they are good examples of ecosystem engineers (Jones et al., 1994; Lavelle et al., 1997; Wardle, 2002). Most diplopods are considered to be litter transformers (Lavelle et al., 1997). Although Bocock and Heath (1967) observed geophagy in Glomeris marginata, Glomeris spp. have been studied primarily in their role as litter feeders, e.g., decomposition of their feces (Tajovsky et al., 1992) and their effect on microbial activity (Anderson and Bignell, 1980; Ineson and Anderson, 1985; Maraun and Scheu, 1996). Recently, Carcamo et al. (2000) examined the feeding preference of adult millipedes Harpaphe haydeniana haydeniana Wood (Polydesmida: Xystodesmidae) on litter. The train millipede Parafontaria laminata (Attems, 1909), another xystodesmid, is the most abundant species of soil macrofauna in the foothills of Mt. Yatsugatake, Japan (Yoshida, 1987; Ito et al., 2001; Hashimoto et al., 2004). Because of its wide distribution and often high population density, this species can have major effects on decomposition processes occurring on the forest floor. Adult P. laminata have been observed to feed on litter (Niijima, 1984). At our study site, we confirmed that adults feed on litter and soil (Hashimoto et al., 2004). The adult population biomass was 28.6 g m–2 dry weight in September 2000 (Hashimoto et al., 2004). This abundance and the occurrence of litter-feeding under natural conditions provide a good opportunity to investigate the functions of train millipedes on litter decomposition processes. The life cycle of this species, from egg to adult stages, lasts 8 years. Molting is observed during summer and is physiologically controlled by cold temperatures in winter (Fujiyama, 1996). Adult P. laminata swarm on the forest floor for mating, and the population consists of only a single age cohort (Ito et al., 2001); therefore, swarming occurs in 8-year intervals (Niijima and Shinohara, 1988). We expected the shift in the developmental stages of P. laminata to influence the carbon dynamics in the soil. To test the hypothesis, we conducted experiments from 1999 to 2000 using three developmental stages: sixth- and seventhinstar larvae and adults. Because larval density is highly heterogeneous (Toyota and Kaneko, 2004) and the effects of P. laminata may vary along a density gradient, we manipulated millipede density at four levels: zero (control), half (low), average (intermediate), and about twice (high) that of the average field density using a field mesocosm method. The objectives of this field experiment were to compare the effects of different developmental stages and to determine density effects of the train millipede on carbon dynamics and soil microbial biomass.

1841

2. Materials and methods 2.1. Study site This study was conducted in the foothills of Mt. Yatsugateke in Yamanashi Prefecture, central Japan (1400 m a.s.l., 35154’35N, 135123’29E), which has a cool temperate climate. Mean annual precipitation and air temperature recorded at Oizumi, near our study site, were 1100 mm yr–1 and 10.6 1C, respectively (Japan Weather Association 1998). The vegetation is plantation forest of Japanese larch (Larix kaempferi [Lamb.] Sargent) planted in 1961; the average diameter and height of L. kaempferi were 19.4 cm and 14.6 m, respectively. The shrub layer was sparse and was dominated by Quercus crispula Blume, Prunus incisa Thunb., Ligustrum tschonoskii Decaisne, Symplocos coreana Ohwi, and Rhododendron obtusum Planch. Dwarf bamboo (Sasa nipponica Makino) dominated the herb layer with L. kaempferi at about 50 cm in height. The soil type was an Andosol (FAO et al., 1998), with a color classification of 7.5YR1/2 and a welldeveloped aggregate structure. The organic layer on the forest floor was mainly composed of larch leaves mixed with larch twigs, and stalks and leaves of bamboo, and had an average thickness of 1.5–3 cm. 2.2. Experimental design: field soil mesocosms Soil mesocosms were established on the forest floor from 17 May to 5 November 1999 (174 d) and from 24 April to 15 October 2000 (174 d). Each mesocosm was made of a PVC cylinder with an inner diameter of 15.5 cm and a height of 30 cm. The O-layer and soil (depth 0–30 cm) were collected separately from the forest floor in the two successive years. Stones, twigs (45 mm diameter), roots (42 mm diameter), and macrofauna were removed by hand from the O-layer and soil, which were separately homogenized before placement into the mesocosm. Wet soil, equivalent to 779 and 708 g dry weight, was distributed in each mesocosm in 1999 and 2000, respectively. Homogenized fresh O-layer, equivalent to 22 g dry weight, was placed on top of the soil in 1999, and fresh Oi (5.6 g dry weight) and Oe (8.4 g dry weight) were deposited in 2000. The amounts of O-layer were determined based on the observed average accumulation on the forest floor (Hashimoto et al., 2004) and additional samplings. We separated the O-layer into two different decay stages in 2000, since we expected a different feeding preference by adult millipedes. However, this was not expected in 1999 because the juveniles are geophagous; therefore, we used homogenized O-layers. The population at the study site consisted of sixth-instar larvae in autumn 1998, which became seventh-instar larvae in autumn 1999. Adults emerged in September 2000, and swarming on the forest floor was observed from September to October (Hashimoto et al., 2004). We manipulated the millipede density into four levels: Control,

ARTICLE IN PRESS A. Toyota et al. / Soil Biology & Biochemistry 38 (2006) 1840–1850

1842

Low-, Intermediate-, and High-density treatments (refer hereafter as Control, L, I, and H). The density of train millipedes was measured on the forest floor in 25
25-cm quadrats up to a depth of 30 cm in October 1998 for sixthinstar larvae (n ¼ 3 quadrats) and in April 2000 for seventh-instar larvae (n ¼ 6). The density of sixth-instar larvae ranged from 160 to 1088 m–2 in October 1998. At another site about 3 km east of our study site, the density of sixth-instar larvae was 106–2156 m–2 in June 1999 (Toyota and Kaneko, 2004). Therefore, in May 1999, we set the four experimental densities at zero (Control), 530 (L), 1060 (I), and 2121 (H) individuals m–2 (Table 1). In April 2000, the density of seventh-instar larvae ranged from 144 to 720 m–2 with a mean7SE of 4697240 individuals m–2 at this site; therefore, the four experimental levels were set at zero (Control), 212 (L), 424 (I), and 636 (H) m–2. Since the sexes of the millipede were identifiable after the seventh instar, equal numbers of male and female seventh-instar larvae were introduced in 2000. The sex ratio was not controlled for sixth-instar larvae. The biomass of the introduced train millipedes was estimated by measuring individual weights. The mesocosms were placed on the forest floor within an area of 900 m2 in a randomized block design of six replicates (24 mesocosms in total). The upper end of each mesocosm was covered with a mesh lid to allow free rainfall and gas exchange. Glass beads and fiberglass were placed on the bottom of each mesocosm to ensure adequate drainage. 2.3. Measurements At the start of the field incubation, subsamples of soil and O-layer from the mesocosms were collected for analysis of soil moisture and carbon concentration. All mesocosms were destructively sampled on 5 November 1999 and 15 October 2000. The O-layers remaining on the surfaces of the mesocosms were collected and weighed. The litter weight loss upon ignition (600 1C for 6 h) was measured only in 2000 to evaluate the amount of contamination by mineral soil in the O-layer. Soil in the mesocosms was separated into two layers of 0–5 and 5–10 cm (1999) and 0–3 and 5–8 cm (2000). In 2000, at the

end of experiment, adults disturbed the soil layer due to their vertical mobility. To avoid a mixture between upper and lower soil layers within a mesocosm, soil was separated into 0–3-cm and 5–8-cm layers in 2000. All surviving millipedes were collected and weighed individually after defecation for 1–2 d. The gut contents of millipedes were also analyzed for carbon concentration (n ¼ 3 in 1999, n ¼ 8 in 2000). The total carbon concentrations of the litter, soil, and millipedes were determined using a Sumigraph NC-95A total C and N analyzer (Shimadzu Corp., Kyoto, Japan). Litter, soil, millipedes, and gut contents were dried for 72 h at 65 1C, and subsamples were collected for measurements of water content (dried for 48 h at 105 1C). Soil microbial biomass carbon was determined using the fumigation–extraction method (Vance et al., 1987). Fresh soil samples equivalent to 8 g dry weight were fumigated with alcohol-free chloroform for 72 h at 20 1C in the dark. Non-fumigated and fumigated soil samples were extracted in 80 ml 10% KCl for 60 min using a horizontal shaker at 160 rev min–1 and then filtered. Microbial carbon was calculated as the difference between extractable carbon of fumigated and non-fumigated samples using a conversion factor of 2.04 (Inubushi, 1997). The microbial biomass carbon from the 1999 experiment was measured by UV absorbance at 280 nm (Nukan et al., 1998) using a Hitachi U-1100 spectrophotometer (Hitachi Instruments Inc., Tokyo, Japan). UV absorbance of the filtrates was recorded immediately after filtration. In the 2000 experiment, the organic carbon of the filtrates was determined using a Shimadzu TOC-5000A analyzer. Microbial carbon was simultaneously measured using UV absorbance and TOC analyzer in 24 samples in 1999. Since the two values were linearly correlated (r ¼ 0.80), microbial carbon for the 1999 measurement was calibrated based on the absorbance value. To determine the carbon dynamics in the mesocosms, the carbon budget accounted for by the net carbon (NET C) change during the experiments was calculated using the equation NETC ¼ DO  layer C þ DSoil C þ DMillipede C, where DO-layer ¼ (O-layer carbon at the end of the experiment  initial litter carbon), DSoil ¼ (soil carbon

Table 1 Number of introduced train millipedes, sixth- and seventh-instar larvae per mesocosm and density and total biomass in the mesocosm experiments in 1999 and 2000 Year:

1999

2000

Developmental stage:

Sixth-instar larvae

Seventh-instar larvae

Treatment

Number

Density (m2)

Biomass (g dry m2)

Number

Density (m2)

Biomass (g dry m2)

Control Low Intermediate High

0 10 20 40

0 530 1060 2121

0 3.4 6.8 13.8

0 4 8 12

0 212 424 636

0 4.8 9.5 15

ARTICLE IN PRESS A. Toyota et al. / Soil Biology & Biochemistry 38 (2006) 1840–1850

1843

Table 2 Individual weights, mortality, and carbon concentrations of train millipedes at the end of the field mesocosm experiments (mean7SE) Year (stage)

Treatment

Individual weight (dry mg, Mean7se)

Mortality (%)

Total carbon (%)

1999 (Sixth-seventh instar larvae)

L I H

18.572.63a 17.673.16a 17.572.82a

48.377.92a 54.0716.76a 43.878.84a

25.870.33a 24.870.14a 24.171.50a

(27) (42) (75)

Male 2000 (Seventh-instar larvae–adults)

L I H

(3) (3) (3)

Female a

43.673.16 44.172.69a 38.771.06a

(6) (8) (18)

65.274.90b 60.674.20b 49.971.61a

(4) (8) (15)

58.3710.54a 77.5711.46a 51.4717.54a

25.570.56ab 26.070.49b 23.970.35a

(10) (11) (13)

Figures in parentheses show the number of individuals for which weights and carbon concentrations were measured. Individual weights, mortality, and chemical properties marked with the same letter were not significantly different among treatments in each year at the 5% level (Scheffe´’s test).

at the end of experiment  initial soil carbon), and DMillipede C ¼ (final millipede carbon  initial millipede carbon). 2.4. Statistical analysis

instar larvae in 1999; however, similar amounts of total biomass were introduced in both years (Table 1). The millipedes became seventh-instar larvae and adults at the final samplings in 1999 and 2000, respectively. Average individual body weights increased by 2.7–2.9 times in the period from spring (sixth instar) to autumn (seventh instar) in 1999. The average weight of seventh-instar larvae did not differ among treatments at the end of the experiment in 1999. In 2000, the weights of adult males were about 70% of adult females, and adult females weighed significantly less in H than in L or I, whereas male weights did not differ among treatments (Table 2). During the experiment, male body weight increased twofold; in contrast, female body weight increased 2.5 times in L and I, and two times in H. Mortality rates did not differ among treatments in both years (Table 2). Body carbon content was not significantly different among treatments in 1999, whereas the carbon content of adults was significantly lower in H than in I.

Initial individual weight, carbon concentration, and mortality of train millipedes among the density treatments were compared using one-way analysis of variance (ANOVA) followed by Scheffe´’s test. The carbon content of millipede gut contents and soil and O-layer carbon concentrations were also compared among developmental stages using one-way ANOVA followed by Scheffe´’s test to identify means that were significantly different at Po0.05. Carbon concentrations in soil and litter were compared using one-way ANOVA followed by a Tukey–Kramer test to assess differences among millipede density treatments. The amounts of litter remaining among density treatments and developmental stages were compared by two-way ANOVA followed by a Tukey–Kramer test. To test for equality of variances and normality, Bartlett’s and Kolmogorov–Smirnov tests were performed prior to ANOVA, respectively. If necessary, log transformations were performed to standardize variances and improve normality. Slopes of the linear regressions for the remaining litter against millipede density were used to represent litter decomposition rates. Significant differences between the slope coefficients of linear regressions were determined using an F-test. Litter consumption per individual millipede at each developmental stage was regressed against millipede density using best-fit parameters in an exponential regression model. All analyses were performed using Stat View ver 5.0J (SAS Institute Inc., Cary, NC, USA).

The amount of litter remaining and millipede densities showed a significant negative relationship (Fig. 1a, b). The amount of litter remaining was affected by developmental stages (two-way ANOVA, F1,40 ¼ 143.24, Po0.01) and density treatments (F3,40 ¼ 29.51, Po0.01), although their interaction was not significant (F1,40 ¼ 1.02, P ¼ 0.39). The slope coefficients of linear regressions differed between the developmental stages (F ¼ 4.86, Po0.05). The litter consumption per individual was constant in 1999 (Fig. 1c), whereas it was significantly reduced with increasing density in 2000 (Fig. 1d).

3. Results

3.3. Carbon concentration in the litter, soil, and gut contents

3.1. Influence of population densities on millipede growth

Larvae significantly reduced the total carbon concentration of the O-layer in 1999 (Table 3). In 2000, the amount of P. laminata fecal pellets mixed in the O-layers in L, I, and H were 1.17, 2.07, and 1.19 g per mesocosm,

The numbers of introduced seventh-instar larvae per mesocosm in 2000 were lower than the numbers of sixth-

3.2. Effects on litter decomposition along gradients of developmental stages and density

ARTICLE IN PRESS A. Toyota et al. / Soil Biology & Biochemistry 38 (2006) 1840–1850

1844

1.5

Amount of litter remaining (kg m-2)

1.5

(a)

(b)

y = -0.0002x + 1.02 r2

r 2 = 0.76**

= 0.48 *

1

1

0.5

0.5

0

0 0

500

1000

1500

0

2000

Millipede density (individuals m-2)

500

1000

1500

2000

Millipede density (individuals m-2) 10

10 Amount of litter consumed per individual (g)

y = -0.001x + 0.68

(c)

(d) y = 0.543e-0.0249x

8

y = 3.967e-0.2493x

8

r 2 = 0.08 NS

r 2 = 0.75 **

6

6

4

4

2

2

0

0 0

10

20

30

Number of larvae per mesocosm

40

0

10

20

30

40

Number of adults per mesocosm

Fig. 1. Relationship between millipede density and amount of litter remaining in the mesocosms at the end of the experiment in 1999 for seventh-instar larvae (a), and in 2000 for adults (b). Relationship between the number of millipedes and amount of litter consumed per individual at the end of the experiment in 1999 (c), and 2000 (d). Regression lines (a, b), best-fit parameters for an exponential decay model (d), coefficients of determination, and significance levels (*Po0.05, **Po0.01, NS ¼ not significant) are shown.

respectively, and showed no significant difference among treatments. The soil used in this study was humus-rich and contained 22–23% carbon at the beginning of the experiment. The total soil carbon concentration in 1999 did not differ between the mesocosms with and without millipedes (Table 3). In 2000, the carbon concentration in the upper soil layer in H was significantly higher than without millipedes (Control) and in L, but not in the deeper soil layers. Gut carbon concentrations of adults were significantly higher than that of the seventh-instar larvae and intermediate between the values of soil and litter; whereas, the gut carbon content of seventh-instar larvae was similar to soil carbon (Fig. 2). 3.4. Effects on soil microbial biomass carbon In 1999, the average soil microbial biomass carbon contents in the upper and deeper layers of each treatment were in the order of Control4L4I4H (Fig. 3a, b), although these differences were not significant. In contrast, in 2000, the order of the average soil microbial biomass carbon was ControloLoIoH (Fig. 3c, d). Although the differences in the upper layer were again not significant, the I and H mesocosms were about 2.9 times higher than the

Control and L in the deeper layer (Po0.01; Fig. 3d). Simultaneously, the Control and L treatments showed almost the same biomass between their upper and deeper layers, whereas in I and H, biomass was about three times higher in the deeper layers than in the upper layers. Soil microbial biomass carbon comprised 0.5–0.8% of the soil carbon in the mesocosms, except in the deeper soil layers of I and H, where it comprised 1.9% in 2000. 3.5. Effects on the carbon budget The carbon dynamics of each density treatment and developmental stage are shown in Fig. 4. Carbon in the millipede body was negligible compared to that in the Olayer and soil. In 2000, DO-layer of all treatments with millipedes was significantly lower than without millipedes (Control), and that in H was significantly lower than in I or L. Even at the larval stage, I and H were significantly reduced compared to the Control. Interestingly, the density effects on DSoil carbon were nonlinear in 2000; DSoil carbon was significantly higher in H than in the Control and L, whereas DSoil carbon in 1999 did not significantly differ among the treatments. NET C in 1999 did not significantly differ among the treatments, whereas in 2000

ARTICLE IN PRESS A. Toyota et al. / Soil Biology & Biochemistry 38 (2006) 1840–1850

1845

Table 3 Total carbon concentrations of the organic layer and soil at the beginning and end of the field mesocosm experiments (%, mean7se, n ¼ 6) Year (stage)

Layer

Initial

Layer

Treatment Control (C)

1999 (Sixth-seventh instar larvae)

2000 (Seventh-instar larvae-adults)

O-layer Soil

49.0271.187 23.2970.205

Oi

45.5672.552

Oe Soil

43.8670.234 22.0170527

Low (L)

O-layer Soil 0–5 cm Soil 5–10 cm O-layer*

b

42.0270.734 21.0770.304a 21.0270.296a 45.2570.338a

Soil 0–3 cm Soil 5–8 cm

21.5970.214a 21.3370.294a

Intermediate (I) ab

High (H)

40.5570.947 21.4770.226a 20.8770.348a 44.1370.423a

ab

39.0870.588 21.4770.207a 20.6270.303a 44.1870.520a

37.6271.426a 21.3470.314a 20.4070.709a 42.4271.367a

21.5270.239a 21.1870.238a

23.4071.174ab 21.5870.332a

24.4070.505b 22.0070.222a

Within rows, total carbon concentrations in the organic and soil layer marked with the same letter were not significantly different among treatments at the 5% level (Tukey-Kramer test).* The values for the O-layer in 2000 were on an ash-free basis.

50

c

Carbon concentration (%)

40 b 30 a

a

20

10

0 Soil

7th instar larvae in 1999

Adults in 2000

O layer

Fig. 2. Total carbon concentrations (mean7SE) in the gut contents of adults (n ¼ 10) and seventh-instar larvae (n ¼ 3) in mesocosms. Average initial values of soil and O-layer carbon concentrations for both years are shown. Means with same letters were not significantly different according to Scheffe´’s test.

it was significantly greater in H than in L. In 2000, NET C was enhanced in I and H compared to the Control at 94 and 391 g m–2, respectively, whereas that in L decreased by 178 g m–2. 4. Discussion The mesocosm method has been successfully used to measure faunal effects or the interactions between soil animals and microflora on various soil processes (e.g., Anderson and Ineson, 1982; Kampichler and Kandeler, 1995; Verhoef, 1996). The average adult biomass in 2000 was 28.6 g dry weight m–2 at our study site (Hashimoto et al., 2004), which was ten times higher than that of reported

total soil macrofauna biomass in temperate coniferous forests (Petersen and Luxton, 1982). In this study site, the average field densities were 448 individuals m–2 and 311 m–2 in September 1999 and August 2000, respectively (Hashimoto et al., 2004). The final densities in each treatment were 274, 488, and 1192 m–2 in October 1999 for L, I, and H, respectively, whereas they were 88, 95, and 309 m–2, respectively, in October 2000. Therefore, we assume that the density dynamics in treatment I are valid because the change in natural density corresponded to the manipulated density in 1999. Even the highest density used in treatment H in 2000 was observed under normal field conditions due to high variability of the natural density. The developmental phenology and survival rate of millipedes in the field (Toyota and Kaneko, 2004) were similar to those inside our mesocosms. In addition, the life history of train millipedes in our mesocosms was also similar to other field populations at Yanagisawa Pass located about 45 km southeast of our study site, where the adult density was 96 m–2 (Niijima, 1984). Therefore, our mesocosms provided suitable conditions for train millipede growth. However, the individual weight of adult females in H was significantly lower than in L or I. It is not clear whether female weight reduction occurs at high densities in the field because density should be normalized by dispersal.

4.1. Larval geophagy Geophagy by millipedes is not common (Hopkin and Read, 1992). Our gut content analysis indicated that in 1999, seventh-instar larvae ingested soil particles. Therefore, the larvae did not directly consume litter. In contrast, the amount of remaining litter showed a negative correlation with larval density in 1999. The carbon concentration in the O-layer tended to be lower in the higher density treatment at the end of the experiment in 1999. We observed larval feces in the O-layer, which will likely change the nutritional condition of this layer. The mixing effects of fecal pellets in the O-layer may have triggered

ARTICLE IN PRESS A. Toyota et al. / Soil Biology & Biochemistry 38 (2006) 1840–1850

1846

500 (a)

(b)

Soil microbial biomass C (mg 100g dry soil-1)

400 300 a 200

a

a

a

a a

a

a

100 0 500 (c)

(d)

b

b

400 300 200

a

a

a

L

I

a

a

a

C

L

100 0 C

H

I

H

Treatment Fig. 3. Soil microbial biomass carbon in field mesocosms. Each datum represents the mean7SE (n ¼ 6). Soil depths are (a) 0–5 cm and (b) 5–10 cm in 1999, and (c) 0–3 cm and (d) 5–8 cm in 2000. C, Control; L, Low density; I, Intermediate density; H, High density. Means with same letters were not significantly different according to a Tukey–Kramer test (Po0.05).

enhanced activity of microbes and other soil mesofauna, leading to higher rates of litter decomposition. Larval feces in the O-layer would affect the activity of other organisms, whereas in the soil layer, larvae did not change the soil microbial biomass. Grazing by larvae may suppress soil microbial biomass because the gut enzymes of arthropods have the potential to digest microbes (Bignell, 1984). Yeast and some fungus families were considerably reduced by gut passage of millipedes (Byzov et al., 1998). Conversely, larval feces should supply available carbon as energy for microbes. The labile carbon from feces to soil that is useful for soil microorganisms would be assimilated by larval coprophagy in the soil. Therefore, the negative effects of larval grazing and the positive effects of larval labile carbon release on soil microbial biomass can be balanced. Overall, despite the decline in O-layer mass by indirect larval effects, NET C in the mesocosms did not change with larval density because the carbon pool was higher in the soil than in the O-layer in the mesocosms. 4.2. Feeding shift in adults Adult P. laminata can directly exploit the O-layer as a food source (Niijima, 1984; Hashimoto et al., 2004). The

assimilation rates of the litter-feeding millipedes G. marginata and H. haydeniana are only 6% litter (David and Gillon, 2002) and less than 10% litter (Carcamo et al., 2000), respectively. Similarly, the assimilation efficiency of adult P. laminata would be expected to be low. Therefore, the feces of litter transformers such as diplopods or isopods contain undigested dead plant tissues (Tajovsky et al., 1992; Lavelle et al., 1997). The adults supply fragmented organic matter as feces in the soil layer due to their vertical movement. Nutrient-rich feces in the soil layer are a favorable resource for microbes. Substrate carbon is correlated very closely with microbial biomass carbon (Wardle, 1992), and thus, increasing the soil carbon concentration through the addition of adult feces is expected to lead to an increase in microbial biomass. Although soil carbon significantly increased in the upper layer, soil microbial biomass carbon significantly increased in the deeper layers in 2000. In the I and H treatments, biomass C showed a threefold increase over that of the Control and L in the deeper soil layers. Bioturbation in the deeper soil layers due to adult vertical mobility was observed more with increasing adult density. Nevertheless, fecal production was higher with increasing density, and the amount of fecal pellets in the O-layer did

ARTICLE IN PRESS A. Toyota et al. / Soil Biology & Biochemistry 38 (2006) 1840–1850

1847

1999 (sixth - seventh instar larvae) 2000 (seventh instar larvae - adults) 0.1

Millipede

0.05 0 0 -2 -4 -6 -8 -10

A B

a

B

ab

C

O Layer

b

b

Carbon (g mesocosm-1)

15 10

B

Soil

AB

5 0 -5

A

A

-10 -15

a

a

a

a

-20 10 B

NET 0 AB

AB

-10

A -20

a

a

a

a

-30 Control

Low

Intermediate

High

Treatment Fig. 4. Carbon budgets during the experiment. NET C ¼ (DO-layer+DSoil+DMillipede; see text for details). Values of the control treatments in both years are shown as broken and dotted lines for 1999 and 2000, respectively. Means with same letters were not significantly different according to a Tukey–Kramer test (Po0.05).

not differ among treatments. The addition of glucose enhances microbial growth in a nutrient-limited soil (e.g., Scheu, 1990). Thus, labile carbon supplied by adult feces and bioturbation in the deeper soil layers where the nutrient supply is poor could cause immediate increases in soil microorganisms, whereas in the upper layer, grazing of soil by adult millipedes may equilibrate microbial activity. Scheu et al. (2002) showed that epigeic litterfeeding earthworms increase labile carbon for soil microorganisms, whereas grazing by endogeic soil-feeding earthworms reduces microbial biomass due to competition for labile carbon in the soil. Despite the increase in soil microbial biomass with increased adult density, the estimated net carbon loss and soil carbon loss in H was significantly lower than in L. The carbon input from the O-layer to the soil contributed to soil carbon accumulation with increasing adult density. Therefore, decomposability of soil carbon could be strongly controlled by the characteristics of adult fecal pellets. Millipede feces usually consist only of litter and are

relatively crumbly (Lavelle et al., 1997); however, the feces of train millipedes were compact. The gut contents of adults indicated that adults ingested litter together with mineral soil. Mixed-feeding on litter and soil by adults has also been found under laboratory conditions (Hashimoto et al., 2004). Mixed-feeding on litter and soil is not common among diplopods (Hopkin and Read 1992), although adult millipedes Alloporus uncinatus in Zimbabwe similarly ingest mineral soil with litter (Dangerfield, 1993). Importantly, geophagy has been suggested to promote carbon stabilization through interactions of soil organic matter with mineral particles in the excrement (Wolters, 2000). For example, enchytraeid casts were coated with clay minerals that provide protection for organic matter particles (Marinissen and Didden, 1997). Thus, the reduced soil carbon loss rate would result from adults feeding on mixed litter and soil. Differences in soil carbon loss rate in relation to adult density could be associated with the proportion of litter

ARTICLE IN PRESS 1848

A. Toyota et al. / Soil Biology & Biochemistry 38 (2006) 1840–1850

and soil in the feces. Our results showed that litter consumption per individual adult increased with increasing adult density, corresponding to an earlier food-preference experiment (Hashimoto et al., 2004). The mass of ingested soil compared to ingested litter per adult increased with the increment of adult density in a laboratory experiment in which adults were provided adequate quantities of both litter and soil (Hashimoto et al., 2004). Overall, an increase in the consumption rate of soil compared to litter at high densities led to more compact feces and the formation of recalcitrant organic compounds mixed with mineral particles in the soil. Soil invertebrates have generally been considered to enhance soil organic matter decomposition (e.g., Anderson and Ineson, 1982), the effects may depend on the duration of observation. The endogeic earthworm Millsonia anomala accelerates the mineralization of soil organic matter during digestion over the short term (several hours), whereas over the long term (420 d), incubations have shown that the annual decomposition rate is three times lower in the casts of M. anomala than in the surrounding soil (Martin, 1991). Moreover, soil carbon mineralization rates in agricultural ecosystems were reduced by earthworms after about 15 years in a 30-year simulation (Brown et al., 2000). In our study, the carbon accumulation at high densities of adults occurred in 174 d; however, more long-term monitoring of carbon dynamics is needed. 4.3. Interaction between developmental stages and density gradients The effect of earthworms on soil processes suggests that function varies among species (e.g., Je´gou et al., 1998; Brown et al., 2000; Scheu et al., 2002; Uchida et al., 2004). However, little is known about a species’ changing effects on decomposition processes according to its developmental stage and density. Our experiment indicated that a temporal shift in diet with advanced developmental stages in the train millipede caused significant changes in O-layer accumulation and soil carbon dynamics. The O-layer and soil in 2000 at this study site had been affected by larval activity in the previous year. The initial quantities of litter in the mesocosms in 1999 were much higher than in 2000 to reflect the field conditions. Our results demonstrated that the mass loss of carbon from the O-layer in all treatments in 1999 was much greater than in 2000. Increases in the carbon loss in the O-layer with increasing density were similar between both years. Therefore, carbon input from the O-layer was not the only reason for soil carbon accumulation in 2000, when millipedes were at the adult stage. Moreover, carbon losses from the O-layer and soil in the Control mesocosms in 2000 were markedly less than in 1999. These differences may have been caused by an increase in carbon recalcitrance in the soil through the change in physical structure caused by larval soil-feeding in 1999.

We confirmed that the train millipede not only provides litter resources to soil microbes directly but also modifies the physical states of the litter and soil. Such manipulations are not common among millipedes. Ecosystem engineers are organisms that directly or indirectly modulate the availability of resources to other species by causing physical state changes (Jones et al., 1994). Physical artifacts such as soil aggregates and channels continue to function in the absence of the living organism that created them and act cumulatively on soil processes (Anderson, 1995). The soil profile of a typical grassland is composed of a high proportion of fecal pellets (Barois et al., 1998; Ciarkowska and Niemyska-Lukaszuk, 2002). At our study site, the surface soil also mainly consisted of aggregates of train millipede feces. These feces can persist, resulting in a longer period of soil carbon accumulation. This prediction is supported by data of high soil carbon concentrations at this study site. Carbon concentrations exceeded 20%, which was significantly higher than that in a low-density area near our study site (ca. 12%; Toyota, unpublished data). Periodic swarming has occurred for at least 80 years because serious traffic disturbances attributable to swarming adults have been recorded near the study site since 1920 (Niijima and Shinohara, 1988). Further investigations should be conducted to establish whether soil organic matter is accumulated by larvae under field conditions, although the effects of larvae from the egg to sixth-instar stages on soil carbon would not be very substantial due to their relatively low biomass. Because the repeated synchronized occurrence of adults in the same field at high densities could lead to a gradual enrichment of carbon in the soil, the train millipede acts as an ecosystem engineer in this forest. 5. Conclusions The developmental stage shifts and density gradients of train millipedes influenced the soil as follows: the litter decomposition rates were positively correlated with density at multiple developmental stages, although the correlation was stronger at the adult stage than at the larval stages; the activities of adults at high densities increased the soil microbial biomass, although not at low densities or at the larval stages; and adults increased carbon accumulation in the soil layer especially at high densities due to mixedfeeding on litter and soil. Mixed-feeding on soil and litter, which is not common among millipedes, can strongly mediate soil carbon stabilization. Acknowledgments We thank Dr. K. Niijima for valuable advice on the biology of train millipedes and Dr. M. Hashimoto and the members of the Soil Ecology Research Group, Yokohama National University, for assisting with the field study. Provision of the study site by Yamanashi Prefecture is gratefully acknowledged. We thank Prof. R. T. Koide and

ARTICLE IN PRESS A. Toyota et al. / Soil Biology & Biochemistry 38 (2006) 1840–1850

Dr. R. J. Blakemore for comments on earlier drafts of this manuscript.

References Anderson, J.M., 1995. Soil organisms as engineers: microsite modulation of macroscale processes. In: Jones, C.G., Lawton, J.H. (Eds.), Linking Species and Ecosystems. Chapman & Hall, London, pp. 94–106. Anderson, J.M., Bignell, D.E., 1980. Bacteria in the food, gut contents and feaces of the litter-feeding millipede Glomeris marginata (Villers). Soil Biology & Biochemistry 12, 251–254. Anderson, J.M., Ineson, P., 1982. Soil microcosm system and its application to measurements of respiration and nutrient leaching. Soil Biology & Biochemistry 14, 415–416. Barois, I., Dubroeuq, D., Rojas, P., Lavelle, P., 1998. Andosol-forming process linked with soil fauna under the perennial grass Mulhembergia macroura. Geoderma 86, 241–260. Bignell, D.E., 1984. The arthropod gut as an environment for microorganisms. In: Anderson, J.M., Rayner, A.D.M., Walton, D.W.H. (Eds.), Invertebrete- Microbial Interactions. Cambridge University Press, Cambridge, pp. 205–228. Bocock, K.L., Heath, J., 1967. Feeding activity of the millipede Glomeris marginata (Villers) in relation to its vertical distribution in the soil. In: Graff, O., Satchell, J.E. (Eds.), Progress in Soil Biology. Friedr. Vieweg & Sohn GmbH, Braunschweig, pp. 233–240. Brown, G.C., Barois, I., Lavelle, P., 2000. Regulation of soil organic matter dynamics and microbial activity in the drilosphere and the role of interactions with other edaphic functional domains. European Journal of Soil Biology 36, 177–198. Byzov, B.A., Kurakov, A.V., Tretyakova, E.B., Thanh, V.N., Luu, N.D.T., Rabinovich, Y.M., 1998. Principles of the digestion of microorganisms in the gut of soil millipedes: specificity and possible mechanisms. Applied Soil Ecology 9, 145–151. Carcamo, H.A., Abe, T.A., Prescott, C.E., Holl, F.B., Chanway, C.P., 2000. Influence of millipedes on litter decomposition, N mineralization, and microbial communities in a coastal forest in British Columbia, Canada. Canadian Journal of Forest Research 30, 817–826. Cebrian, J., 1999. Patterns in the fate of production in plant communities. The American Naturalist 154, 449–468. Ciarkowska, K., Niemyska-Lukaszuk, J., 2002. Microstructure of humus horizons of gypsic soils from the Niecka Nidzianska area (South Poland). Geoderma 106, 319–329. Currie, W.S., 1999. The responsive C and N biogeochemistry of the temperate forest floor. Trends in Ecology and Evolution 14, 316–320. Dangerfield, J.M., 1993. Ingestion of mineral soil/litter mixtures and faecal pellet production in the southern African millipede, Alloporus uncinatus (Attems). Pedobiologia 37, 159–166. David, J.F., Gillon, D., 2002. Annual feeding rate of the millipede Glomeris marginata on holm oak (Quercus ilex) leaf litter under Mediterranean conditions. Pedobiologia 46, 42–52. FAO, ISSS, ISRIC, 1998. World reference bases for soil resources, Rome, pp. 109. Fujiyama, S., 1996. Annual thermoperiod regulating an eight-year lifecycle of a periodical diplopod, Parafontaria laminata armigera Verhoeff (Diplopoda). Pedobiologia 40, 541–547. Green, R.N., Trownbridge, R.L., Klinka, K., 1993. Towards a taxonomic classification of humus forms. Forest Science Monograph 29, 1–48. Hashimoto, M., Kaneko, N., Ito, M.T., Toyota, A., 2004. Exploitation of litter and soil by the train millipede Parafontaria laminata (Diplopoda: Xystomidae) in larch plantation in Japan. Pedobiologia 48, 71–81. Hopkin, S.P., Read, H.J., 1992. The Biology of Millipedes. Oxford University Press, New York, pp. 43–58. Ineson, P.J., Anderson, J.M., 1985. Aerobically isolated bacteria associated with gut and faeces of the litter feeding macroarthropods Oniscus asellus abd Glomeris marginata. Soil Biology & Biochemistry 17, 843–849.

1849

Inubushi, K., 1997. Method of microbial biomass C. In: Society for the study of soil microbes (Eds.), Methods of Experiment with Soil Microbe. Yoken-Do Press, Tokyo, pp.173–183 (in Japanese). Ito, M.T., Mima, J., Yoshida, T., Hashimoto, M., Toyota, A., Kaneko, N., Tomita, D., Uchida, K., 2001. Distribution of YatsugatakeKirigamine population of a train millipede (Parafontaria laminata armigera Verhoeff, 1936) in the outbreak year 2000. Edaphologia 68, 39–42 (in Japanese with English summary). Japan Weather Association, 1998. Weather Yearbook 1998. Ministry of Finance Japan, Tokyo (in Japanese). Je´gou, D., Cluzeau, D., Balesdent, J., Tre´hen, P., 1998. Effect of four ecological categories of earthworms on carbon transfer in soil. Applied Soil Ecology 9, 249–255. Jones, C.G., Lawton, J.H., Shachak, M., 1994. Organisms as ecosystem engineers. Oikos 69, 373–386. Judas, M., 1992. Gut content analysis of earthworms (Lumbricidae) in a beechwood. Soil Biology & Biochemistry 24, 1413–1417. Kampichler, C., Kandeler, E., 1995. Microbial-faunal interactions in soils. In: Schinner, F., Ohlinger, R., Kandeler, E., Margesin, R. (Eds.), Methods in Soil Biology. Springer, Berlin, pp. 367–372. Kaye, J.P., Binkley, D., Rhoades, C., 2003. Stable soil nitrogen accumulation and flexible organic matter stoichiometry during primary floodplain succession. Biogeochemistry 63, 1–22. Ko¨rner, C., 2000. Biosphere responses to CO2 enrichment. Ecological Applications 10, 1590–1619. Lavelle, P., Spain, A.V., 2001. Soil Ecology. Kluwer Academic Publishers, Netherlands. Lavelle, P., Bignell, D., Lepage, M., Wolters, V., Roger, P., Ineson, P., Heal, O.W., Dhillion, S., 1997. Soil function in a changing world: the role of invertebrate ecosystem engineers. European Journal of Soil Biology 33, 159–193. Maraun, M., Scheu, S., 1996. Changes in microbial biomass, respiration and nutrient status of beech (Fagus sylvatica) leaf litter processed by millipedes (Glomeris marginata). Oecologia 107, 131–140. Marinissen, J.C.Y., Didden, W.A.M., 1997. Influence of the Enchytraeid worm Buchholzia appendiculata on aggregate formation and organic matter decomposition. Soil Biology & Biochemistry 29, 387–390. Martin, A., 1991. Short-and long-term effects of the endogeic earthworm Millsonia anomala (Omodeo) (Megascolecidae, Oligochaeta) of tropical savannas, on soil organic matter. Biology and Fertility of Soils 11, 234–238. Niijima, K., 1984. The outbreak of the train millipede. Japanese Journal of Forest Environment 26, 25–32 (in Japanese). Niijima, K., Shinohara, K., 1988. Outbreaks of the Parafontaria laminata group (Diplopoda: Xystodesmidae). Japanese Journal of Ecology 38, 257–268 (in Japanese with English summary). Nukan, N., Morgan, M.A., Herlihy, M., 1998. Ultraviolet absorbance (280 nm) of compounds released from soil during chloroform fumigation as an estimate of the microbial biomass. Soil Biology & Biochemistry 30, 1599–1603. Petersen, H., Luxton, M., 1982. A comparative analysis of soil fauna populations and their role in decomposition processes. Oikos 39, 288–388. Scheu, S., 1990. Changes in microbial nutrient status during secondary succession and its modification by earthworms. Oecologia 84, 351–358. Scheu, S., Schlitt, N., Tiunov, A.V., Newington, J., Jones, T.H., 2002. Effects of the presence and community composition of earthworms on microbial community functioning. Oecologia 133, 254–260. Schro¨ter, D., Wolters, V., De Ruiter, P.C., 2003. C and N minerlisation in the decomposer food webs of a European forest transect. Oikos 102, 294–308. Tajovsky, K.S., Santtruckova, H., Hanel, L., Balik, V., Lukesova, A., 1992. Decomposition of faceal pellets of the millipede Glomeris hexasticha (Diplopoda) in forest soil. Pedobiologia 36, 146–158. Torn, M.S., Trumbore, S.E., Chadwick, O.A., Vitousek, P.M., Hendricks, D.M., 1997. Mineral control of soil organic carbon storage and turnover. Nature 389, 170–173.

ARTICLE IN PRESS 1850

A. Toyota et al. / Soil Biology & Biochemistry 38 (2006) 1840–1850

Toyota, A., Kaneko, N., 2004. Relation between larval train millipede density and soil microbial biomass under two different forests. Edaphologia 74, 15–25 (in Japanese with English summary). Uchida, T., Kaneko, N., Ito, M.T., Futagami, K., Sasaki, T., Sugimoto, A., 2004. Analysis of the feeding ecology of earthworms (Megascolecidae) in Japanese forests using gut content fractionation and d15N and d13C stable isotope natural abundances. Applied Soil Ecology 27, 153–163. Vance, E.D., Brookes, P.C., Jenkinson, D.S., 1987. An extracted method for measuring soil microbial biomass C. Soil Biology & Biochemistry 19, 686–696. Verhoef, H.A., 1996. The role of soil microcosms in the study of ecosystem processes. Ecology 77, 685–690. Vitousek, P., 2004. Nutrient Cycling and Limitation. Hawai’i as a Model System. Princeton University Press, Princeton.

Wang, Y., Hsieh, Y.P., 2002. Uncertainties and novel prospects in the study of the soil carbon dynamics. Chemosphere 49, 791–804. Wardle, D.A., 1992. A comparative assessment of factors which influence microbial biomass carbon and nitrogen levels in soil. Biological Reviews 67, 321–358. Wardle, .A., 2002. Communities and ecosystems, Linking the aboveground and belowground components. Princeton University Press, Princeton. Wolters, V., 2000. Invertebrate control of soil organic matter stability. Biology and Fertility of Soils 31, 1–19. Yoshida, T., 1987. Biology of periodical Diplopoda, Parafontaria laminata armigera Verhoeff. II Distribution range of the Diplopodea broken out in Nagano Pref., Japan. Research of Sugadaira 8, 83–87 (in Japanese).