Introduction

Generalist predators such as spiders and ground beetles have received more attention than specialist predators or parasitoids as biological control agents in annual cropping systems (Symondson et al. 2002; Welch et al. 2012). They have various characteristics that determine the success of pest suppression. For example, they can persist on alternative prey before pest herbivores colonize a crop field after an agronomic disturbance. This habit promotes a lying in wait strategy for pest suppression, resulting in early season pest control (Welch et al. 2012). There is considerable agreement on the potential of spiders as biocontrol agents (Riechert 1998; Sunderland and Samu 2000; Wise 1993). However, their trophic cascade effects on vegetable yield have been studied by only a few research groups (Halaj and Wise 2002; Snyder and Wise 1999, 2001; Vichitbandha and Wise 2002).

In Japan, Kiritani and colleagues have demonstrated the impact of the wolf spider, Pardosa pseudoannulata (Bösenberg and Strand), on the populations of the green rice leafhopper, Nephotettix cincticeps (Uhler), in paddy fields (Kiritani and Kakiya 1975; Kiritani et al. 1972; Sasaba and Kiritani 1974). They also found that linyphiid spiders disrupted the larval aggregation of Spodoptera litura (F.) in a taro, Colocasia esculenta (L.) Schott, field (Nakasuji et al. 1973). Since then, these studies have been followed up by several laboratory feeding trials (Ozaki et al. 1977; Yamada and Yamaguchi 1985; Yasudomi et al. 1993) and an immunological study (Nemoto 1986). These laboratory studies have shown that the wolf spiders (Lycosidae) are potential biological control agents of Brassica insect pests, particularly the diamondback moth, Plutella xylostella (L.).

In general, wolf spiders are abundant worldwide in Brassica crop fields and considered to be potential predators of P. xylostella (Furlong et al. 2013). However, the influence of spiders on the populations of Brassica insect pests and vegetable yield has so far received less attention (Furlong et al. 2013; Liu et al. 2014). Many studies have focused on laboratory feeding trials and monitoring of spiders (Miranda et al. 2011; Senior and Healey 2011). Only a few studies have investigated the impact of spiders on vegetable yield or marketability (Furlong et al. 2004).

In this study, we manipulated the density of the wolf spiders in enclosed cabbage plots to investigate their impact on the pest populations and crop yield. For this purpose, we chose the wolf spider, Pardosa astrigera L. Koch, a common inhabitant of crop fields in Japan (Miyashita 1969; Suzuki and Okuma 1975). Straw mulch was applied to the soil surface of the experimental plots to provide a refuge for spiders, thereby enhancing pest control.

Materials and methods

Study site

The experiments were conducted in a cabbage field (11.5 × 24.5 m) at the National Research Institute of Vegetables, Ornamental Plants and Tea (currently National Agriculture and Food Research Organization (NARO), Institute of Vegetable and Tea Science), Ano, Mie Prefecture, Japan, in 1995. The field was surrounded by fallow land and vegetable fields planted with cabbage and tomato.

Experimental design

Field studies were conducted in the spring and autumn of 1995. In both experiments, nine plots (2 × 2 m) were established in the field shortly after it was plowed, and planting beds (1 m wide) were formed. Of the nine plots, six were enclosed by transparent acrylic resin fences (45 cm tall). The fences were buried to a depth of 15 cm, leaving 30 cm above the ground. The tops of the enclosures were left open to allow the entry of winged herbivores. However, each of the four top corners of the enclosures was covered with a square board (5 × 5 cm) of the same material as the fence. This was to prevent spiders crawling up the corners and escaping from the enclosures.

Three of the six enclosures were chosen for augmentation with P. astrigera (spider-addition plots). The remaining three enclosures served as controls, with the wolf spiders removed (spider-removal plots). In the spider-removal plots, P. astrigera was removed using two pitfall traps (11 cm in diameter) throughout the experiment as mentioned below. The remaining three unfenced plots (open reference plots) were naturally colonized by herbivores and ground-dwelling predators. These three treatments, each replicated three times, were arranged in a randomized block design.

Cultivation and management

In the spring experiment, after the fences had been erected in mid-April, cabbage seedlings (Brassica oleracea L. variety capitata ‘Kinkei 201’) at the four- to six-leaf stage were transplanted on 21 April 1995. In the autumn experiment, after the fences had been erected in late September, cabbage seedlings (‘YR-Nodoka’) at the four- to six-leaf stage were transplanted on 26 September 1995. Each plot contained two rows of planting beds with two rows of plants for each bed. The spacing between plants was 30 cm and between rows was about 50 cm (i.e., 24 plants/plot).

Cabbage yield was measured at the end of experiments. In the spring experiment, cabbages remaining in the enclosures (14 plants) after the last count of herbivores (28 June) were cut off at ground level in mid-July 1995. Heads were then individually weighed, retaining their outer leaves. In the autumn experiment, only cabbage heads were cut off from 20 plants in mid-January 1996 and weighed individually. Cabbage plants from which heads were cut off were left in the field until the collection of P. astrigera in mid-April 1996.

No herbicides, insecticides, or fungicides were applied to the experimental field or to adjacent habitats during each experiment. Weeds were, however, occasionally removed outside the enclosures (including the open reference plots) by hoeing early in the season, but allowed to grow later in the season. In contrast, weeds in the enclosures were allowed to grow naturally through most of the season, but cut as needed to a height below that of the cabbage or the fences to prevent spiders from escaping from the enclosures by means of creeping weeds.

Application of straw mulch

To provide spiders with a refuge (Halaj et al. 2000; Riechert and Bishop 1990), the spider-addition and spider-removal plots were covered with rice straw mulch (1 kg/plot in dry weight) after cabbages had been transplanted. The straw was cut into halves and evenly placed on the soil surface of the enclosures, forming a thin layer through which the soil surface was seen. The open-reference plots received no rice straw.

Manipulations of spider densities

Pardosa astrigera used in these experiments is a common inhabitant of the Japanese crop fields and shows two peaks of activity in the Tokyo area of Japan (Miyashita 1969). However, in our institute, it showed two or more peaks of occurrence from early March through early September, with immatures relatively active from late September through November (Suenaga and Hamamura, unpublished data). P. astrigera overwinters as a last instar and acquires resistance to hunger but can develop when food is provided (Miyashita 1968, 1969). The mean body lengths are 7–10 mm for female adults and 6–8 mm for male adults (Chikuni 1989).

In the spring experiment, P. astrigera was released into spider-addition plots 13, 15, 27, and 29 days after planting. The numbers of spiders released on these dates were 32, 18, 11, and 39 individuals per plot for a total of 100 spiders (25 individuals/m2). They were collected from the immediate area of the experimental field on the day of introduction. The stage structure of the introduced spiders was the same for all the spider-addition plots: 57 % unsexed immature, 21 % last-instar males, 14 % adult males, and 8 % adult females without egg sacs.

In the autumn experiment, immatures of P. astrigera were introduced into the enclosures at a density of 193–211 spiders per plot (about 50 spiders/m2) on the day of planting. The immatures were all 5 mm or smaller in body length; about half of these (53 %) measured <3 mm. These spiders had been collected from the immediate area of the field in late September; they were individually kept in ventilated glass vials (18 × 60 mm) together with small moistened cotton balls for 5 days at 15 °C before release.

Monitoring of spiders and other predators

In the spring experiment, two pitfall traps without preservatives were installed diagonally in each plot 12 days after planting (early May), with a gap of 140 cm between traps. The traps were white plastic cups 11.5 cm deep with openings of 11.5 cm and with their rims buried flush with the soil surface. Three drainage holes (1.6 cm diameter) were opened on the bottom of the cups and covered with a polyethylene screen (1.2-mm mesh openings). These traps were operated for two consecutive days per week in the spider-addition plots to check whether the addition of spiders was effective. Trapped P. astrigera were counted to investigate the seasonal abundance and activity (activity density, Luff 1987), and then released back into the plots. In contrast, traps in the spider-removal plots were operated throughout the experiment to remove trapped P. astrigera. Other predator species, such as carabid beetles and other spider species, captured in the spider-addition and spider-removal plots were recorded and then released back into the enclosures. At the end of the spring experiment (mid-July 1995), the wolf spiders were collected from the enclosures and counted by sex and development stage. P. astrigera was also monitored in the open-reference plots and field boundary (about 3 m wide), each using 10 pitfall traps to investigate its natural activity density. These traps were continuously operated during the experiment, and captured P. astrigera were counted every 1–3 days and then released back to their sites of capture.

In the autumn experiment, the pitfall traps were installed in early October, as in the spring experiment, and captured P. astrigera and other predators were treated as in the spring. However, the numbers of P. astrigera and other predators captured were not systematically recorded. In mid-April 1996, spiders were collected from the enclosures and counted by sex and development stage.

Monitoring of herbivores

Herbivorous insects were counted on 20 plants per plot early in the season but on only 10 plants late in the season in both experiments. This is because counting insects on fully grown cabbage plants late in the season is time consuming and labor intensive.

In the spring experiment, the numerically dominant species in the experimental field were the diamondback moth P. xylostella, the common cabbage worm Pieris rapae crucivora Boisduval, the beet semi-looper Autographa nigrisigna (Walker), and the green peach aphid Myzus persicae (Sulzer). These herbivores were counted on eight dates at intervals of 6–14 days, between late April and late June 1995. On the last survey date (66 days after planting), plants were cut off at ground level, and herbivores were counted on the heads and outer leaves. In the autumn experiment, the four species mentioned above and the cabbage armyworm Mamestra brassicae (L.) were investigated. They were counted on five dates at intervals of about 10 days, except at the last interval of 21 days, between early October and late November 1995. Larvae and pupae were counted together for each of the lepidopterous species, and adults and nymphs were counted together for the aphid species.

Statistical analysis

The data of the open reference plots were excluded from the statistical analyses because several conditions, such as weed management and enclosure effects, differed from those used in the enclosed plots. Thus, we tested only the differences between the spider-addition and spider-removal treatments.

We used repeated-measures ANOVA (Keppel 1991; von Ende 2001) to analyze the impact of spider predation on pest densities. In this design, we treated the spider density (treatment) as the between-subjects factor and the date of count (time) as the within-subjects factor. Though the experimental plots were arranged in a randomized block design, replications were simply treated as subjects in the analysis. Each subject was repeatedly measured for the herbivore density over time. To correct possible violation of the sphericity assumption and to make the analysis more conservative, the Greenhouse–Geisser adjustment was used; the degrees of freedom of the critical F-statistic were multiplied by the Greenhouse–Geisser ε (0 < ε < 1). Significant interactions between treatment and time showed the effect of spider predation on herbivore density over time.

Statistical analyses were performed individually for each season and each pest species. Mean pest densities per plant of each replicate plot (x) were √(x + 0.05) transformed to stabilize variance. The constant 0.05 was used instead of 0.5 (Yamamura 1999) because the discrete unit of the mean pest density per plant was 0.1 in the case of sample size 10. The smaller sample size (n = 10) was used instead of a larger one (n = 20) for the calculation of the constant (Yamamura, personal communication). We did not use multivariate analysis of variance (MANOVA) because of the low number of subjects (three) within each treatment and the high number of levels of time (five and eight) (von Ende 2001). The fresh weight of cabbage plants was analyzed for the difference between spider-addition and spider-removal treatments by a t-test. The average cabbage head weight (x) of each replicate plot was log10(x) transformed before analysis to meet the assumptions of ANOVA. These statistical tests were performed using the Statistica software (StatSoft Inc 1999).

Results

Impact of spider predation on pest density

Pardosa astrigera significantly suppressed the numbers of P. xylostella and M. persicae in the spring and autumn experiments, respectively (Fig. 1; Table 1). Significant interactions between treatment and time suggested the impact of the spider predation on the pest numbers (Table 1). The density-growth curves of P. xylostella diverged late in the season; the herbivore density in the spider-addition plots was suppressed to nearly half the value of the spider-removal plots (e.g., 63 vs 120 individuals/plant in late June, Fig. 1a). The abundance of M. persicae in the spider-addition plots was strongly suppressed and kept at low densities throughout the autumn experiment (e.g., 33 vs 816 individuals/plant in late November, Fig. 1b).

Table 1 Significance test for the impacts of P. astrigera on pest densities evaluated by the treatment (spider density) × time interactions in the repeated-measures ANOVA
Full size table
Fig. 1
figure 1

Impact of P. astrigera on cabbage pest densities in (a) the spring and (b) the autumn experiments. Each symbol represents mean number (±SD) of pest insects in the spider-addition, spider-removal, and open reference plots. Where not shown, error bars are smaller than the symbols. Arrows in the figures indicate the dates when P. astrigera was introduced

Full size image

Pardosa astrigera had no significant impact on the abundance of other lepidopterous species (Table 1). The densities of P. rapae crucivora and A. nigrisigna were not significantly different between the treatments in both experiments (Fig. 1; Table 1). In the autumn experiments, M. brassicae density was also not affected by the spider predation (Table 1). These three lepidopterous species were larger than P. xylostella and less abundant, with density being less than seven insects per plant in both experiments.

Accidental differences in the number of alate colonizers of M. persicae may have affected the density increase. However, the numbers of alate aphids per plant were not significantly different between treatments; the mean total numbers of alatae found early in the season (6, 16, and 26 October) were 0.33 ± 0.32 and 0.22 ± 0.10 per plant for spider-addition and spider-removal treatments, respectively (treatment, df = 1, 4, F = 0.23, p = 0.65; treatment × time, ε = 0.998, df = 2, 8, F = 0.52, p = 0.61).

Seasonal abundance of P. astrigera and other predators

Manipulation of the spider density was effective during the spring experiment. The introduction of P. astrigera into the enclosures successfully raised its activity density compared with the ambient level early in the season (Fig. 2). However, the activity density declined to a level nearly equal to that of the open reference plots late in the season. No P. astrigera were captured in the spider-removal plots during the spring experiment, suggesting that the early erection of fences effectively prevented spiders from entering the enclosures.

Fig. 2
figure 2

Seasonal changes in the activity density of P. astrigera in the spider-addition plots, open reference plots and field boundary surrounding the experimental field in the spring experiment. Each symbol represents 2-day catch by two traps in each plot or site. Error bars (SD) were only assigned to the data obtained from the spider-addition plots

Full size image

Other generalist predators recorded in the spider-removal plots during the spring experiment were mainly carabid beetles. They consisted primarily of the genera Amara, Bembidion, Harpalus and Chlaenius. These carabid beetles were trapped from early May to early July, with the total numbers of the beetles caught ranging from 13 to 27 per plot. Chlaenius micans (F.), a voracious predator of P. xylostella (Suenaga and Hamamura 1998), was trapped from late June to early July (1–6 beetles/plot in total). The larvae of C. micans, which are more voracious feeders (Suenaga and Hamamura 1998), were not found on cabbage plants during the spring experiment. As for spiders, only three individuals of Linyphiidae and one individual of Lycosidae were captured during the experiment. No carabid beetles or other spider species were captured from the spider-addition plots during the spring experiment, except that an individual linyphiid spider and two individual carabid beetles were captured from three replicate plots in mid-June.

Crop yield

Despite the suppression of P. xylostella and M. persicae abundance, the cabbage yield did not significantly increase (Fig. 3). In particular, in the autumn experiment, despite the marked suppression of aphid density, the yield was not significantly increased (Figs. 1b, 3). The overall crop yield was higher in the spring experiment than in the autumn experiments, and it was attributed to the difference in the harvesting method; the crop yield was measured for both heads and outer leaves in the spring experiment but for heads alone in the autumn experiment.

Fig. 3
figure 3

Mean weight (±SE) of cabbage heads of each treatment for the spring and autumn experiments. Differences in mean weight were tested between the spider-addition and the spider-removal treatments. The overall difference in the mean weight between the spring and autumn experiments reflects the difference in harvesting techniques, i.e., heads with outer leaves and only heads were harvested in the spring and autumn experiments, respectively

Full size image

Collection of P. astrigera

The number of P. astrigera collected at the end of experiments increased in the spring experiment but decreased in the autumn experiment compared with the numbers initially introduced (Fig. 4). In the spring experiment, the number of adult females and immatures increased by 23 %, with about 70 % of the adult females carrying egg sacs or hatchlings. No P. astrigera was collected from the spider-removal plots. In the autumn experiment, P. astrigera decreased by 62 %, with the majority (89 %) of them being immatures (>3 mm in body length). The remainder were in the adult stage, with 71 % of the female adults (15 of 21 in total of the three plots) carrying egg sacs. Only one female was collected from the spider-removal plots. The fact that almost no wolf spiders were found in the spider-removal enclosures suggests that manipulation of the spider density was effective during each experiment.

Fig. 4
figure 4

Mean numbers of P. astrigera introduced to and collected from the spider-addition enclosures. Error bars represent standard deviations. No error bars were assigned to initial totals because they were exactly (spring) or nearly (autumn) the same for all replication plots

Full size image

Discussion

Our results showed that P. astrigera suppresses the populations of certain pest species if it occurs abundantly early in the crop season and remains stable throughout the season. Riechert and Bishop (1990) demonstrated that pest density and leaf damage were reduced in broccoli plots where the abundance of the wolf spiders was increased up to 10 times the natural density by applying grass hay mulch. Other studies have also shown that spiders regulate prey populations if spider density is enhanced early in the crop growing season (Riechert 1998; Sunderland and Samu 2000; Welch et al. 2012). In the current study, the activity density of P. astrigera was actually higher in the spider-addition plots than in the ambient levels of the open reference plots early in the season (Fig. 2).

The increased spider activity density in the spider-addition plots may be attributed in part to the application of straw mulch. In particular, early application of straw mulch before the weeds covered the plots would have provided a refuge for spiders. Indeed, we observed P. astrigera gathering along the base of the enclosure walls before the straw mulch was added. They, however, became dispersed throughout the enclosures after the mulch was applied. These observations suggest that the application of straw mulch would have reduced the rates of contact between the conspecific cannibals and, in addition, it would have increased the time spent searching for prey items. The complex structure provided, for example, by straw mulch serves as a refuge from cannibalism for spiders (Halaj et al. 2000; Langellotto and Denno 2006; Rickers and Scheu 2005). The increased activity density could also be ascribed to the small enclosed area in which spiders moved about repeatedly. The decreased activity density late in the season may be attributed to the inhibition of spiders’ movement by weeds.

Despite the increase in the spider activity density within the enclosures, P. astriger failed to consistently suppress pest density (Table 1; Fig. 1). For example, P. astrigera suppressed the numbers of P. xylostella and M. persicae only in either the spring or the autumn experiment. Both species were suppressed in the season when they were abundant but not in the season when they were less abundant (Fig. 1). This may be ascribed to the switching of prey species by spiders from a less to a more abundant one (Halaj and Wise 2002; Welch et al. 2012). However, it was unknown whether such switching occurred in our experiments. Other examples of the wolf spiders failing to control herbivore density were the cases of P. rapae crucivora, A. nigrisigna, and M. brassicae. A possible reason for the failure of the wolf spiders in controlling these lepidopteran species may be that their full-grown larvae (about 30 mm or larger in body length, Okada 2003) are larger than P. astrigera. Indeed, spiders generally prefer prey smaller than their own size (Nentwig 1987).

The rapid divergence in the growth curves of M. persicae was unexpected (Fig. 1b). Predation of the aphids by P. astrigera was indeed not observed during the autumn experiment. We assumed that the divergence in growth curves may be attributed to the possible chance difference in the number of alate colonizers; however, alate numbers were not different between treatments. Thus, the divergence in the aphid growth curves may be ascribed to the manipulation of the spider density. As described above, the immatures of P. astrigera (<5 mm in body length) would have preferentially consumed aphids smaller than their own body size (Nentwig 1987). Aphid densities are suppressed early in the crop growth by spiders (Birkhofer et al. 2008; De Roincé et al. 2013; Schmidt et al. 2004), in spite of their low food quality (Toft 2005).

Despite the suppression of pest density, P. astrigera failed to increase the cabbage yield (Fig. 3). Previous studies also failed to reveal measurable effects of spider predation on crop yield. Researchers have ascribed the absence of cascading effects (indirect effects of predators on plants via herbivores) to herbivore density that is too low to affect crop yield (Carter and Rypstra 1995; Halaj and Wise, 2002; Vichitbandha and Wise 2002). In the spring experiment, the density of P. xylostella was indeed below an economic injury level of five larvae per plant (Morishita 1998) in the pre-heading stage, which is vulnerable to herbivore attack (Ayalew 2006; Morishita 1998) (Fig. 1a). Myzus persicae probably also had little influence on the cabbage yield because they were less abundant early in the season (Fig. 1b).

The impact of wolf spiders on herbivore populations and crop yield has often been measured in the presence of carabid beetles (Halaj and Wise 2002; Lang 2003; Snyder and Wise 1999, 2001; Vichitbandha and Wise 2002) or mirid bugs (Finke and Denno 2004). Some of these studies demonstrated that carabid beetles also have an impact on herbivores (Lang 2003; Snyder and Wise 2001). In the present study, carabids were also trapped in the enclosures in the spring experiments. However, most of them were species of Amara, Bembidion, and Harpalus, which are less voracious and poor plant climbers (Suenaga and Hamamura 1998), and thus would have had little influence on pest densities. Adults of C. micans, voracious ground beetles (Suenaga and Hamamura 1998), were also captured in spring; however, it was late in the season that they were trapped, when the investigation was almost completed. Indeed, C. micans has been reported to occur from late June to late September (Suenaga and Hamamura 2001), consequently suggesting that C. micans did not occur in the autumn experiment. Therefore, carabid beetles probably contributed little to pest suppression in both experiments.

In conclusion, our results suggest that P. astrigera is an effective biocontrol agent, provided that it abundantly occurs early in the crop season. However, P. astrigera alone would not be able to suppress herbivore assemblages, suggesting that diverse predator assemblages are necessary for the control of diverse herbivore communities. For example, carabid beetles of the genus Chlaenius, another beneficial predator of insect pests (Suenaga and Hamamura 1998, 2001), may suppress lepidopterans other than P. xylostella. To enhance the diversity and abundance of natural enemies, habitat manipulation within and around the fields would be an effective strategy (Ditner et al. 2013). We need further investigations to elucidate the impact of wolf spiders on the crop yield and quality.