Do host-plant requirements and mortality from soil cultivation determine the distribution of graminivorous sawflies on farmland?

Authors


Dr A.M. Barker, The Game Conservancy Trust, Fordingbridge, Hants SP6 1EF, UK. Fax: 01425 655848; E-mail: abarker@game-conservancy.org.uk

Abstract

1. There are many arthropod species on farmland that are not economic pests but which have an important role in the wider ecology of agricultural ecosystems. Few of these have been studied in detail. This study investigated the factors underlying the distribution of one such group, graminivorous sawflies (Hymenoptera: Tenthredinidae).

2. In a farm survey, sawfly larvae of the genera Dolerus and Pachynematus were over six times more abundant in fields of sown perennial rye-grass Lolium perenne L. than in fields of spring barley, winter wheat or long-established grazing pastures.

3. In rearing trials the growth rates and survival of larvae of the two genera were high on L. perenne and low on two cereals, spring barley and winter wheat. Differences were sufficiently marked to have had a strong influence on the distribution of larvae between crops in the field.

4. In an experimental study, nearly twice as many adults emerged from uncultivated as from cultivated ground, suggesting that disturbance of the overwintering prepupae and pupae in the soil caused about 50% extra mortality. This mortality will have reinforced the effects of host-plant requirements, suppressing sawfly numbers in annually cultivated cereal fields whilst numbers built up in less-frequently cultivated grass fields.

5. Sawfly abundance is known to have declined on farmland. Our results suggest that changes in patterns of crop rotation have been the main cause of this decline. Farming practice in Britain has moved away from mixed rotations alternating cereals with undersown grass fields. Modern intensive farms grow monocultures of cereal crops or of heavily grazed grass, both poor habitats for sawflies.

6. Increasing insecticide use in cereals has probably been only a minor influence on population abundance of sawflies as wheat and barley fields are not high-quality habitats.

7. Provision of undisturbed grassy habitats would promote the conservation of graminivorous sawflies on farmland.

Introduction

Declines of several bird species have been attributed directly to shortages of insect food, particularly during the early growth and development of chicks (Potts 1986; Potts & Aebischer 1991; Aebischer & Ward 1997). Increasing use of insecticides (with directly toxic effects on insects) and herbicides (which eliminate their host-plants from the arable landscape) are thought to be a major cause of the shortage of insects (Campbell et al. 1997). However, other elements of the intensification of farming have also been important. These include processes increasing the area under annual cultivation, such as greater cropping of marginal land, and the decrease in the rotational practice of undersowing spring cereals with grass (Potts 1997; Sotherton 1998). There is a continuing need to identify which of these factors have contributed to the decline of key insect groups in bird diets; a priori assumptions that pesticides are solely responsible will not aid in developing scientifically-based strategies for the conservation of farmland wildlife.

The larvae of sawflies have been identified as an important food source in farmland habitats for the chicks of several bird species known to be in decline. These include the grey partridge Perdix perdix (Green 1984; Potts 1986), corn bunting Emberiza cirlus (Aebischer & Ward 1997), reed bunting Emberiza schoeniclus (Badulin 1981) and skylark Alauda arvensis (Poulsen, Sotherton & Aebischer 1998). The two most common sawfly genera in these habitats are the graminivorous Dolerus and Pachynematus. Dolerus species are univoltine with spring-flying adults, a larval period running from June to July and then a long overwintering phase in an earthen cell in the soil. Pachynematus sawflies may have two or three generations from spring to early autumn (March–September) but also overwinter as pupae in cocoons in the soil. Both are recorded as sporadic minor pests of cereals in continental Europe (Faber 1970; Wetzel & Freier 1980; Heyer & Wetzel 1988; Miczulski & Lipinska 1988), America (Kamm 1975) and Asia (Xu & Chen 1991). However, a long-term monitoring study has shown graminivorous sawfly numbers to be low and in decline in cereal fields in southern Britain (Aebischer 1991). Aebischer (1990) related this to reductions in the practice of undersowing spring cereal crops to establish a ley after harvest, leading to increases in the area under annual cultivation. Sawflies are likely to be at risk from cultivation while overwintering as pupae in the soil (Potts & Vickerman 1974). Deep ploughing has been used as a cultural control method for Dolerus species in China (Anonymous 1992), suggesting that soil disturbance causes significant mortality. Additionally, Aebischer (1990) explored the potential effects of insecticide use on sawflies, which have been shown to have little resistance to broad-spectrum insecticides (Sotherton 1990). He estimated that populations could take as long as 7 years to recover from a single application of summer insecticide.

However, all these studies only considered sources of direct mortality, ignoring a potentially important indirect factor – the availability of suitable host plants. The behavioural responses of ovipositing females in selecting hosts can determine the subsequent abundance of larvae, but are often ignored in studies of insect population dynamics as they are difficult to measure directly (Preszler & Price 1988). Barker & Maczka (1996) showed experimentally that winter wheat was a much less suitable host for three graminivorous sawfly species than selected grass species, and was avoided by ovipositing females. Declines in the availability of suitable grasses through loss of mixed arable/grass rotations (Potts 1997) and in the availability of grass weeds (Vickerman 1974) might therefore be a hidden factor in the reported decline of sawflies in Britain.

The aims of this study were to discover the field distribution of sawflies between grass and cereal crops and to evaluate the importance of host-plant requirements and soil cultivation in determining this distribution. Results of a survey of sawfly larval abundance in different crops are compared with measurements of larval performance on different hosts and with estimates of the mortality caused by soil cultivation during the overwintering period.

Methods

Survey of the occurrence of larvae in cereal and grass crops

A survey of larvae in grasses and cereals on farmland in southern England was carried out over a 4-year period using sweep-net sampling. In each year, four farms were selected that each had the following crops: winter wheat Triticum aestivum L., spring barley Hordeum vulgare L., permanent grazing pasture and herbage seed, which is a crop of perennial rye-grass (L. perenne) grown to provide seed for silage or ley mixtures. None of the fields used in the survey had received summer insecticide applications prior to sampling. Sampling dates were in mid June – early July (1 July 1993, 14–16 June 1994, 19–22 June 1995, 24–26 June 1996), which is the time of peak larval numbers, especially for the univoltine Dolerus species. Changes in cropping patterns between years dictated that although three farms (Farms A, B and C) were used in all years, the fourth changed in 3 of the 4 years (Farm D in 1993, Farm E in 1994, Farm F in 1995 and 1996). In 1993, three fields of each crop were sampled on each farm; this was increased to four fields of each crop in all subsequent sampling except on Farm E, where only three fields of each were available. This meant that overall 59 fields of each type were sampled (except spring barley, where a shortage of available fields on one farm in one year reduced the number to 57).

In 1993, three sub-samples each of 50 sweep-net strokes were taken in each field; in the following years, this was increased to six sub-samples in an attempt to increase the catch per field. Each set of three sub-samples in 1993 was taken in a transect line from one field edge starting inside the crop headland and travelling into the field (spanning about 12–150 m in from the edge). Crop headlands were not sampled in order to avoid potential edge effects caused by factors such as aspect and boundary type, and the possible confounding presence of grass weeds in some cereal headlands. From 1994 to 1996, pairs of transects were taken about 10 m apart to obtain the six sub-samples. The sweep-nets used had a diameter of 0·45 m and the average sweep length was ≈ 1 m of net movement through the crop. To maximize between-crop comparability, all samples on a single farm were collected either by one person or by the same pair of people each taking half of the sub-samples in each field.

Sampling in 1993, 1994 and 1996 was carried out during the day. In 1995, the sampling was carried out in the evening (18.00–22.00 h British Summer Time) as there was some evidence suggesting that at this time sawfly larvae might be more active in crops and so more available to the sweep-net (A.M. Barker, unpublished data). However, the total catch did not increase markedly in 1995 so this was not repeated. The patterns of catches and of species caught between crops were not affected by this change (crop × year effects were not significant in analyses) so all years have been included in the overall analysis.

All sub-samples were placed in separate plastic bags, kept cool after collection and then frozen. They were subsequently sorted and all sub-samples were preserved in alcohol in separate tubes. Sawfly larvae were identified to genus by comparison with reared reference samples bred from identified adults collected in southern England. The Dolerus spp. larvae were further identified to species, using a key developed using reared larvae (Barker 1998).

Analysis of larval abundance in relation to crop type

Analysis of catch sizes was carried out separately for Dolerus spp. larvae and Pachynematus spp. larvae. These two genera comprised 98% of the catch; Selandria serva (F.) larvae and a few larvae of species that feed on broad-leaved weeds made up the remainder. Data for each field were scaled to the number of larvae caught per 50 sweeps. Catch sizes were often low (zero in many fields) so that the data were highly skewed. The data were log (n + 1)-transformed before carrying out a three-way analysis of variance incorporating factors for year, farm and crop with the catch for each individual field as the basic data unit.

Sawfly densities may have been influenced by the cropping regime on the farm in previous years. Factors were therefore coded to allow testing of the effects of prior cropping in the overall model. Initial analysis using a factor coding for five different prior crops (herbage seed, winter wheat, spring barley, permanent pasture or other crops such as peas) found that, of these crops, only the presence of herbage seed made a major difference to subsequent densities. This factor was therefore simplified to one coding: prior herbage seed vs. all other prior crops. Analysis of variance was used to test whether previous cropping made a significant extra contribution to explaining the variation in larval abundance. The proportional difference in mean catch between fields which were previously cropped as herbage seed and those which were not was given by the exponential of the estimated regression coefficient from the anova model.

During the survey, records were kept of visual estimates of grass height on a simple scale (short: ≈ 0–5 cm, medium: ≈ 5–10 cm, long > 10 cm) in permanent pasture fields. The relationship between height and the presence or absence of sawfly larvae in samples from these fields was investigated with a log-likelihood test of association.

Analysis of catch composition in relation to crops

Compositional analysis (Aitchison 1986; Aebischer, Robertson & Kenward 1993) was used to test for differences in catch composition between crops. Where either or both Dolerus or Pachynematus larvae were caught in a field, log-ratios were calculated from the abundance of each genus in that field [i.e. log (number of Pachynematus larvae/number of Dolerus larvae)]. Similarly, for all fields in which one or more Dolerus larvae were caught and identified, log-ratios were calculated for the relative numbers of each species in each field. The log-ratio transformation was used to overcome the problem of linear dependence of proportional data and to allow statistical analysis (Aitchison 1986). For the generic analysis, with only one log-ratio to consider, an anova was used to test for the effect of crop after adjusting for farm and year effects. The equivalent multivariate statistic, Wilk's Lambda, was calculated for the effect of crop on Dolerus species composition, assuming a multivariate normal distribution. Following Aebischer, Robertson & Kenward (1993), values of zero were replaced before calculation of log-ratios by a small number which was an order of magnitude less than the smallest non-zero proportion in the data. In the generic analysis, for example, this proportion was 5% or 0·05, giving a replacement value of 0·005 for zeroes.

Where species composition was significantly different between crops, the mean log-ratios of species-pairs within crops were predicted using the statistical model and the mean proportional compositions calculated to show which genera or species were relatively more abundant in each crop.

Larval performance

Larvae were reared from eggs laid by adult female sawflies collected from farm sites in Hampshire, Dorset and Scotland. In 1995 and 1996, females were caged with potted L. perenne plants and in 1997, as part of an oviposition experiment, with a choice of winter wheat Triticum aestivum L., spring barley Hordeum vulgare L., L. perenne and Festuca rubra L. plants. Choice of sawfly species was determined by the availability of adult females. Species used were: in 1995, Dolerus haematodes Schrank, D. puncticollis Thomson, and D. niger L.; in 1996, D. aeneus Hartig and Pachynematus extensicornis Norton; and in 1997, D. gonager Fab., D. aeneus, D. haematodes and D. puncticollis. Adults used were provided with cotton wool soaked in a dilute sugar/water solution as food. All eggs laid were monitored and when hatching began larval collection was carried out daily. Most larvae collected were newly hatched, but the occasional larvae not discovered until they had reached the second or third instar were also included in the experiment.

Larvae were reared individually following the techniques given in Barker & Maczka (1996). Larvae collected on each date from each host were allocated randomly to one of four host plants: F. rubra (commercial variety, cultivar not known), L. perenne cv. Talgo (1995/96) or Macho (1997), spring barley cv. Chariot, or winter wheat cv. Hunter (1995/96) or Consort (1997). Plants were obtained from the field a few weeks before use to ensure their growth stages were realistic; the changes in cultivars used were due to changes in availability over the 3-year period.

Larval fresh weight was measured for each individual to the nearest 0·1 mg every 1–3 days. Larval survival was also monitored. Survival was analysed for each sawfly species in each year by using a log-likelihood (G) test to compare the relative numbers of larvae that survived to the prepupal stage on each grass or cereal species. As in Barker & Maczka (1996), the instantaneous growth rate was estimated for each individual larva from the day of its collection until it finished feeding or died. This is the slope of a linear relationship between the logarithm of larval weight against time (in days). This technique makes use of all data available on each individual's rate of growth. It also avoids the potential problems associated with the assumptions of linearity implicit in the use of indices such as relative growth rate (RGR) (Raubenheimer & Simpson 1992). The calculated growth rates were analysed using an anova to test for the effect of host plant.

Individuals for which there were less than three appropriate observations (equating to survival for less than 5 days) were excluded from the analysis. A few larvae of most sawfly species in each year survived without weight gain (i.e. zero growth) for 5 days or more before dying. This was apparently not host-specific so these were also excluded.

Effects of cultivation on overwintering survival

Spring emergence of adult sawflies was measured in three rotational set-aside fields on a Scottish farm in 1996, in two set-aside fields on a farm in southern England in 1996 and three fields (a set-aside, a winter wheat and a grass ley) on the same English farm in 1997. Emergence traps were placed on paired blocks of cultivated and uncultivated land in each field. Cultivated areas were ploughed and harrowed in spring (February/early March) in 1996 and early winter (November) in 1997; cultivation disturbed the soil to a depth of about 25 cm. The selected blocks were 110 × 1·5 m-strips in all fields except the ley, where two 30 × 25-m rectangular blocks were used so that stock could be fenced out. Fifteen traps were evenly spaced along each strip and placed in a 5 ×3 grid on each rectangular block.

Emergence traps consisted of 1-m2 wooden boxes with spiked corners and black 1·5-mm mesh netting lids; they contained yellow water traps (210 × 100 × 60-mm aluminium trays sprayed inside with yellow paint) filled with a dilute solution of detergent to collect emerging insects. The spikes on the boxes were pushed firmly into the ground and the base of each box was sealed on both sides with loose earth. Traps were checked twice weekly, taking care to look for live sawflies as well as those in the water traps; all adult sawflies were collected and identified. Trapping was carried out from late March to early June, when emergence finished.

The numbers of Dolerus spp., Pachynematus spp. and total sawflies caught in traps on cultivated and uncultivated land were calculated for each field. Log-linear modelling assuming a Poisson distribution was used to test for differences in the catches between treatments, using paired within-field cultivated/ uncultivated replicates. For groups where the effect of cultivation was significant, the log-linear model was used to estimate the mean proportional difference in catches between treatments.

The estimated difference in Dolerus catches between cultivated and uncultivated areas of fields (a value of 2·35) was used to weight the Dolerus catch in the field survey. For example, a catch of 1 larva per 50 sweeps in a winter wheat, spring barley or autumn-sown herbage seed field was scaled up to 2·35 larvae per 50 sweeps. This estimated the sawfly density in each field that might have been expected had none of the fields been cultivated. A three-way anova incorporating factors for year, farm and crop was carried out on the log(n + 1)-transformed weighted data to see if larval abundance was still dependent on crop type after allowing for the effect of cultivation.

Results

Crop survey: larval abundance

In total, 900 larvae were caught during the 4-year survey, with 529 Dolerus spp. larvae, 353 Pachynematus spp. larvae and 18 larvae of other genera. The abundance of larvae of both common genera varied between crop types (Dolerus: F3,170 = 36·88, P < 0·001, Pachynematus: F3,170 = 92·12, P < 0·001; Fig. 1). However, the relative magnitude of the differences varied between years and farms, with significant interactions for crop × farm (F15, 170 = 2·19, P < 0·01) and crop × farm × year (F21, 170 = 4·66, P < 0·001) for Dolerus and for crop × farm (F15, 170 = 3·90, P < 0·001) for Pachynematus. These interactions may have been partly an artefact of the low numbers caught. Mean catches per field ranged from 0 to 3·06 larvae per 50 sweeps (Table 1), and no larvae were caught at all in 47% of the sampled fields, exaggerating the influence of small differences in relative magnitude. Examination of the individual catches (Table 1) indicated that there was a general pattern with the highest catches in herbage seed fields on most farms in most years. Only four of the 32 year/farm combinations for the two genera were exceptions to this pattern. (These were Farm E in 1994 where there were marginally more Dolerus larvae in permanent pasture, Farm B in 1993 and C in 1994 where more Dolerus larvae were caught in winter wheat than in other crops, and Farm D in 1993 where only two Pachynematus larvae were caught in total.) Overall, six times as many larvae of both genera were caught on average in herbage seed fields as in the other crops (Fig. 1). Larvae were caught in 95% of the herbage seed fields in comparison to 42% of the spring barley fields, 37% of the winter wheat fields and 40% of the permanent pasture fields.

Figure 1.

Mean catch of Dolerus (Üwc4,50Ý▪Üwc4Ý) and Pachynematus (Üwc4,25Ý▪Üwc4Ý) larvae per field (expressed as catch per 50 net sweeps) in four crops in a 4-year farm survey, with 95% confidence intervals (error bars). Values have been back-transformed from log (n + 1)-transformed data. A Tukey HSD test on the effect of crop on catch size showed the mean catch of both genera per field of herbage seed to be significantly higher than that in the other three habitats (P < 0·001). HS = herbage seed (Lolium perenne), SB = spring barley, PP = permanent pasture, WW = winter wheat.

Table 1.  Mean and standard error of catch size (number of larvae per 50 sweep samples) per field in each of four crops for individual farms from 1993 to 1996, with number of fields sampled (n). Values have been back-transformed from the log (n + 1)-transformed data on which all analyses were performed. Bold text shows the crop with the highest larval density on each farm in each year (a) Dolerus spp. larval catches 1993–96
Herbage seedSpring barleyPermanent pastureWinter wheat
CropnMeanSEMMeanSEMMeanSEMMeanSEM
Farm A
199330·980·10 000·10 0·10 0·10 0·10
199443·060·240·32 0·09 0·59 0·120·12 0·07
199540·730·26 000·33 0·33 00
199640·290·03 00000·08 0·05
Farm B
199330·10 0·10 0·19 0·19 000·440·08
199440·760·08 0·74 0·21 0·04 0·04 0·510·12
199542·890·33 00000·04 0·04
199640·750·27 0·20 0·07 000·16 0·16
Farm C
199332·730·50 0·19 0·190000
199440·39 0·130·04 0·04 0·21 0·21 1·360·47
199540·310·12 0·04 0·04 0·04 0·0400
199640·500·21 0·04 0·04 000·20 0·09
Others
1993 Farm D30·490·23 000·26 0·26 00
1994 Farm E30·26 0·26 0·31 0·14 0·350·19 00
1995 Farm F40·580·44 0·04 0·04 0·30 0·1400
1996 Farm F40·200·07 0·04 0·04 0·04 0·04 0·08 0·05
(b) Pachynematus larval catches 1993–96
Herbage seedSpring barleyPermanent pastureWinter wheat
CropnMeanSEMMeanSEMMeanSEMMeanSEM
Farm A
199330·440·08 000·10 0·10 00
199440·800·11 000·26 0·20 00
199540·580·14 000·04 0·04 00
199641·040·24 000000
Farm B
199331·230·34 0·33 0·33 0·10 0·10 00
199441·970·21 0·27 0·11 0·04 0·04 0·04 0·04
199542·330·21 000000
199640·620·14 000000
Farm C
199330·640·12 0·10 0·10 0·190·19 00
199440·450·06 0·22 0·13 000·07 0·07
199540·41 0·16 000000
199640·27 0·11 0·04 0·04 000·04 0·04
Others
1993 Farm D30·100000·10000
1994 Farm E30·410·41 0·05 0·05 0·10 0·10 00
1995 Farm F41·090·30 0·16 0·06 0·24 0·09 00
1996 Farm F40·430·14 0·07 0·07 0·07 0·07 0·04 0·04

The importance of a field's history in determining current numbers was highlighted by analysis of previous cropping patterns. The presence or absence of herbage seed as the previous crop in a field was a significant factor affecting the mean catch of both Dolerus larvae (F1,169 = 18·46, P < 0·001) and Pachynematus larvae (F1,169 = 4·86, P < 0·05). Calculations from the estimated regression coefficients showed that the mean catch in a field was 21% higher for Dolerus larvae and 5% higher for Pachynematus larvae if the field had been cropped as herbage seed in the previous year. Fields of herbage seed are often left for 2 years, with the rye-grass re-growing after the first harvest. This might have magnified the mean catches in herbage seed fields in relation to those in other fields, because more of these had been in this crop in the previous year (49% compared to 2% of spring barley and 21% of winter wheat fields). However, even with the effect of the previous crop accounted for in the model the crop type sampled still significantly affected the mean catch (Dolerus: F3,169 = 18·97, Pachynematus: F3,169 = 66·16, both P < 0·001, but still with significant two- and three-way interactions).

Soil cultivation is one possible cause of sawfly mortality, and pupae overwintering in second-year herbage seed fields will have had a year's respite from soil disturbance. However, many of the first-year herbage seed fields were sown, like cereal crops, directly into a cultivated seed-bed. Catches of sawfly larvae in these fields were still at least four times higher than those in cereal fields [Dolerus: direct-sown first year herbage seed (19 samples): mean 0·625, SEM 0·125, wheat (59 samples): mean 0·165, SEM 0·041, spring barley (57 samples): mean 0·124, SEM 0·028; Pachynematus: herbage seed: mean 0·485, SEM 0·084, wheat: mean 0·013, SEM 0·007, spring barley: mean 0·074, SEM 0·022, values back-transformed from log (n + 1)-transformed data]. This would suggest that differences between crops were not wholly due to their cultivation regimes.

In permanent pasture fields, the proportion of fields in which larvae were caught was affected by the length of the grass (G2 = 9·864, P < 0·01). There were catches in 80% of the fields with long grass, but only 35% of the medium length and 25% of the short grass fields. Catches of Dolerus in long grass fields were quite high (back-transformed mean = 0·578, SEM= 0·118, n = 10) and comparable to those in herbage seed fields (Fig. 1), but Pachynematus catches were not (mean = 0·135, SEM= 0·054, n = 10).

Crop survey: catch composition

Overall, the proportions of Dolerus and Pachynematus larvae in the catch varied significantly between crops (F3,121 = 10·20, P < 0·001; no significant interactions between crop type and year or farm). The average proportion of Pachynematus larvae in the catch was 58% in herbage seed fields, 34% in permanent pasture, 17% in spring barley, and only 1% in wheat.

Nearly all (96%) of the 526 Dolerus larvae caught were from four species: D. haematodes, D. gonager, D. puncticollis and D. nigratus Müller. The relative proportions of these four species differed between crops (crop effect in compositional analysis: Wilk's Λ= 0·687,

image

 = 39·04, P < 0·001; no significant interactions between crop type and year or farm). As shown in Fig. 2, D. haematodes averaged nearly 80% of the larvae caught in the herbage seed fields and 91% of those caught in spring barley. Permanent pasture had a higher proportion of D. puncticollis, at about 50% percentages of D. haematodes, D. gonager and D. puncticollis were similar in wheat with a smaller percentage of D. nigratus.

Figure 2.

Composition of the mean catch in four different crops in a 4-year farm survey. HS = herbage seed (L. perenne), SB = spring barley, PP = permanent pasture, WW = winter wheat. Lengths of bars represent the proportions of the catch divided between the four Dolerus species: D. haematodes (Üwc4,50Ý▪Üwc4Ý), D. gonager (Üwc4,80Ý▪Üwc4Ý), D. puncticollis (Üwc4,15Ý▪Üwc4Ý), and D. nigratus (Üwc4,30Ý▪Üwc4Ý)

Larval performance

Larval performance varied significantly between plant species for most of the Dolerus species tested. For D. haematodes and D. niger, larval survival (Table 2) was high on L. perenne and spring barley, and growth rates (Table 3) were significantly higher on L. perenne than all other host plants. Most larvae died before pupation when reared on winter wheat. Survival of D. puncticollis and D. gonager larvae was significantly higher on both grasses than on cereals, with growth rates highest on L. perenne. P. extensicornis larval growth rates were 40% higher on L. perenne than on the other three species and its survival was also best on this species. D. aeneus suffered high rearing mortality, so the results were less conclusive; in one out of the two years studied (1996) larvae grew significantly faster on grass than cereal species but the difference between the grasses and spring barley was not significant in 1997.

Table 2.  Percentage survival to the prepupal phase for six sawfly species on two cereals (winter wheat and spring barley) and two grasses (perennial ryegrass and red fescue). Significance of relative survival on different larval diets is given by log-likelihood (G) tests
Plant species
Sawfly
species
Year T. aestivumH. vulgareL. perenneF. rubraG (d.f.)P
Dolerus1995n3124222664·51  < 0·001
haematodes % survival0%71%91%42%(3) 
 1997n
% survival
13
8%
14
64%
12
75%
15
27%
17·48
(3)
 < 0·001
Dolerus niger1995n
% survival
9
22%
10
70%
10
70%
8
12·5%
11·05
(3)
 < 0·05
Dolerus1995nNot tested2621Not tested23·28  < 0·001
puncticollis % survival 38%95% (1) 
 1997n
% survival
9
11%
11
27%
9
55%
10
70%
9·05
(3)
 < 0·05
Dolerus gonager1997n
% survival
16
56%
18
50%
18
94%
19
89%
15·09
(3)
 < 0·01
Dolerus aeneus1996n
% survival
11
0%
12
0%
12
25%
11
9%
6·98
(3)
NS
 1997n
% survival
9
0%
12
0%
11
0%
12
25%
2·43
(3)
NS
Pachynematus1996n3735383438·22  < 0·001
extensicornis % survival14%23%55%0%(3) 
Table 3.  Mean growth rates (mg mg–1 day–1) of the larvae of six sawfly species on four host plants, with standard errors (SEM) Growth rates were calculated for larvae that survived for more than 5 days; (*) indicates cases where insufficient larvae survived to include in the analyses. Results of anova on the effect of plant species on growth rate are given. Significant differences between host plants for each sawfly species, calculated using a Tukey HSD test, are indicated by different letters after the means. [In one case (+), the overall anova was significant but no pairwise differences were quite significant.] Maximum growth rates for each species have been highlighted
Plant species
Sawfly
species
Year T. aestivumH. vulgareL. perenneF. rubraF (d.f.)P
Dolerus1995Mean0·064 a 0·069 a 0·092 b0·064 a27·67 (3,80) < 0·001
haematodes SEM0·005
(n = 13)
0·002
(n = 25)
0·002
(n = 26)
0·002
(n = 20)
  
 1997Mean
SEM
0·088
0·010
(n = 3)
0·087
0·005
(n = 10)
0·094
0·010
(n = 12)
0·064
0·020
(n = 4)
1·07 (3,25) NS
Dolerus niger1995Mean
SEM
0·046 a
0·007
(n = 4)
0·071 b
0·006
(n = 7)
0·096 c
0·004
(n = 6)
*
*
17·50 (2,14) < 0·001
Dolerus1995MeanNot tested0·066 a 0·079 bNot tested6·73 (1,36) < 0·05
puncticollis SEM0·004
(n = 17)
0·002
(n = 21)
    
 1997Mean
SEM
*
*
0·099 a
0·005
(n = 6)
0·117 b
0·003
(n = 7)
0·095 a
0·005
(n = 8)
6·53 (2,18) < 0·001
Dolerus gonager1997Mean
SEM
0·084 ab
0·004
(n = 12)
0·082 a
0·005
(n = 11)
0·105 c
0·004
(n = 19)
0·099 bc
0·003
(n = 16)
7·32 (3,54) < 0·001
Dolerus aeneus1996Mean
SEM
0·053
0·011
(n = 4)
0·041
0·015
(n = 3)
0·074
0·006
(n = 6)
0·083
0·004
(n = 3)
3·93 (3,12) < 0·05 + 
 1997Mean
SEM
*0·086
0·005
(n = 3)
0·094
0·007
(n = 6)
0·088
0·003
(n = 7)
0·60 (2,13) NS
Pachynematus1996Mean0·090 a 0·096 a 0·132 b0·087 a 4·40 (3,53) < 0·01
extensicornis SEM0·007
(n = 8)
0·011
(n = 15)
0·009
(n = 26)
0·014
(n = 8)
  

In general, percentage survival was highest on L. perenne for five out of six species, as were larval growth rates. On the second grass species, F. rubra, results were more variable; this was a good or adequate host for some sawfly species, although generally supporting lower growth rates than L. perenne, but survival of other species such as D. haematodes was poor. Survival and growth on winter wheat were generally poor; spring barley was also unsuitable for some species, but for D. haematodes and D. niger it appeared to be a viable host, with about 70% of larvae reared on it surviving to pupation.

Effects of cultivation on overwintering survival

Most of the sawflies caught in the emergence traps were Dolerus or Pachynematus species, with a few individuals of the genera Tenthredopsis and Nematus. Significantly more adult sawflies emerged from uncultivated than cultivated land (Table 4, F1,7 = 12·71, P < 0·01) with an average of 1·91× as many in each field. On average 2·35× as many Dolerus adults per field emerged from uncultivated as from cultivated land (F1,7 = 12·84, P < 0·01). Pachynematus adults were caught in only four of the eight fields used, and the results for this species were inconclusive; the catch on uncultivated land was not significantly different from that on cultivated land (F1,7 = 1·53, P > 0·05) but in the one field where adults of the genus were caught in number (Table 4), there were over twice as many caught in the traps on uncultivated ground. Overall, the results suggested that soil cultivation caused a 50% mortality of overwintering sawfly prepupae and pupae over and above normal winter losses.

Table 4.  Numbers of Dolerus and Pachynematus adults and total number of adult sawflies emerging from cultivated and uncultivated ground in five fields in Sussex and three in Scotland
DolerusPachynematusTotal sawflies
 CultivatedUncultivatedCultivatedUncultivatedCultivatedUncultivated
Sussex 1996821331125
Sussex 1996712301113
Scotland 1996240027
Scotland 1996030013
Scotland 1996030003
Sussex 1997313374
Sussex 1997298191128
Sussex 1997110011
Total235417254484

When the Dolerus catches from the field survey were weighted to allow for this mortality from cultivation there was still a significant difference in the catch between crops (F3,170 = 26·83, P < 0·001). As with the unweighted data set, significant crop × farm (F15, 170 = 2·61, P < 0·01) and crop × farm ×year interactions (F21, 170 = 4·51, P < 0·001) made the results difficult to interpret fully. The unweighted mean catches of Dolerus larvae in herbage seed were about six times those from winter wheat, spring barley and permanent pasture (Fig. 1). Correcting for mortality from cultivation reduced the difference in catch sizes between herbage seed and cereals to about three or four times [herbage seed: mean 0·942, SEM 0·084, winter wheat: mean 0·314, SEM 0·065, spring barley: mean 0·247, SEM 0·051, mean and SEM values calculated using log(n + 1)-transformed data and back-transformed]. However, it accentuated the difference between all the planted crops and the uncultivated permanent pasture fields (mean 0·130, SEM 0·033). The results suggest that about one-third of the differences between herbage seed and cereal crops were due to the differences in cultivation regimes between these crops.

Discussion

Crop types and the distribution of larvae

The results of this survey demonstrated a clear six-fold difference in the field between the average abundance of Dolerus and Pachynematus larvae in fields of uncut, ungrazed L. perenne and their average abundance in three other crops, spring barley, winter wheat and long-established grazed pasture. Both genera were found in all crops but Pachynematus larvae were proportionately less common in cereals, especially winter wheat, where they made up on average only 1% of the larval catch. It is difficult to relate these differences between crops directly to published observations, as other studies have not directly compared numbers in grasses and cereals. Haris (1994a,b) found L. perenne to support large numbers of both sawfly genera, although numbers were even higher in a number of other planted grass species. He suggested that various grass species rather than neighbouring wheat fields were the primary hosts of Dolerus species but that wheat was a good host plant for some species (Haris 1995). Studies of sawfly larvae in cereal crops have tended to find numbers higher in spring barley than in winter wheat (Wetzel & Freier 1980; Miczulski & Lipinska 1988) but the statistical significance of the difference was not assessed. This difference was not apparent on the farms in this survey, where numbers were low in both cereals.

Sawfly abundance was low even in some apparently suitable fields on some farms in some years (Table 1). Potts & Vickerman (1974) found that sawfly distribution tended to be clumped, with little dispersal from emergence sites. Such weak dispersal may limit the number of suitable fields that are colonized across the arable landscape, where such sites may be widely spread and will change from year to year with crop rotation. Over time, such poor ‘tracking’ of the changing environment could lead to falling abundance over whole farms even when grass crops are grown.

Whilst sampling efficiency was unlikely to be identical between all crop types, it would have been similar in wheat, barley and herbage seed, where the sweep net passes through about the top 30 cm of the leafy part of the crop. Samples were all taken at the same time of year and at this time all these crops had reached their maximum heights and were flowering. Efficiency in permanent pasture would have been more variable as height varied with grazing intensity from < 5 cm to 30 cm, so results for this crop may not be as consistent an estimate of larval abundance. In short grass the sweep net will sample the full depth of the vegetation but will tend to be in contact over a shorter distance.

Host-plant requirements

Laboratory trials identified distinct differences in larval performance between different hosts. Pachynematus extensicornis and five Dolerus species all grew and survived better on L. perenne than on modern cultivars of winter wheat and spring barley. L. perenne also appeared to be a better host than a second grass species, F. rubra, for most of the species tested. Winter wheat has been shown to be a less suitable host than L. perenne for three sawfly species and than F. rubra for two of these (Barker & Maczka 1996). The current study showed that both winter wheat and spring barley are poor hosts for a wide range of graminivorous sawfly species. Measured differences in host quality were clearly of sufficient magnitude to affect the survival of field populations of sawflies. For example, survival rates on L. perenne and winter wheat differed by as much as 91% (for D. haematodes in 1995, Table 2) and growth rates differed between these two hosts by up to 47% (P. extensicornis, Table 3).

Rearing trials may not fully represent the costs and risks to larvae of feeding on particular hosts in the field where, for example, differential risks from natural enemies may counteract direct host-plant effects (Bernays & Graham 1988). Nevertheless, the differences in larval performance between hosts matched the observed field distribution of larvae, which were concentrated in planted L. perenne rather than in winter wheat or spring barley. Variations in host-plant preferences between sawfly species could also be related to their field distributions. For example, over 90% of the larvae caught in spring barley were D. haematodes, which survived well on this host (Table 2). Dolerus gonager larvae, which survived better than other Dolerus species on winter wheat, formed a greater proportion of the catch in this crop. These results support the hypothesis of a relationship between host-plant requirements and sawfly field distribution.

High densities of sawfly larvae have been recorded sporadically in spring barley and winter wheat on farms in eastern Europe and Asia (e.g. Freier & Wetzel 1984; Miczulski & Lipinska 1988; Xu & Chen 1991). Freier & Wetzel (1984) recorded a maximum density of 22 larvae per 50 sweeps, compared with an average total catch of only 0·25 larvae per 50 sweeps in cereals in this study. Given the high mortality of larvae on cereals in laboratory trials such densities are surprising. The explanation may lie in regional differences in the characteristics of cereal cultivars, which can vary in their suitability for sawfly larvae (Heyer & Wetzel 1988), or possibly in differences in the phenology of cereal crops relative to adult flight times. There is some evidence that wheat growth stage affects colonization by Dolerus adult females (Haris 1995). Cereals are most likely to be suitable hosts during the limited period when there are lots of young growing leaves which will be most suitable for sawfly oviposition (Price, Craig & Roininen 1995; Barker & Maczka 1996). Sawfly density in cereal crops may also be higher where grass weeds are available as a food source. Vickerman (1974) found a significant reduction in sawfly numbers when grass weeds, primarily Poa trivialis L., were sprayed out in a winter cereal crop. Fields in this survey received standard agricultural herbicide inputs to control grass and broad-leaved weeds – generally an autumn and a spring herbicide treatment on winter wheat, spring herbicide on spring barley and autumn herbicide for cereal volunteers and broad-leaved weeds on herbage seed. The highest density of larvae in winter wheat in the study occurred in two fields in Farm B in 1994 (Table 1) which were observed to have high densities of grass weeds, primarily Poa trivialis L. (A.M. Barker, personal observation).

In the field, host-plant quality could potentially affect sawfly distribution through two different mechanisms. As well as the direct effects of host plants affecting larval survival, female oviposition choices between hosts may determine where larvae occur. Females of some graminivorous species have been shown to lay eggs in the most suitable larval host plant species when a choice is available (Barker & Maczka 1996). However, on the farm scale females will often encounter monocultures rather than mixed stands of grasses and cereals. Oviposition behaviour in such non-choice situations is unknown for Dolerus and Pachynematus species. Preszler & Price (1988) and Craig, Itami & Price (1989) showed that willow-galling sawflies Euura lasiolepis would withhold eggs rather than lay into poor quality hosts. If this behaviour extends to the graminivorous species, egg distribution will be concentrated into fields containing suitable hosts.

Mortality from post-harvest cultivation and other factors

There were several factors in addition to host-plant quality that might have been influencing differences in sawfly densities between crops in this study: post-harvest cultivation, grazing pressure and pesticide use. Soil cultivation as a source of mortality has been stressed by previous authors (Potts & Vickerman 1974; Aebischer 1990) who highlighted the concentration of sawfly numbers around grass leys undersown into spring cereal fields and left to grow after harvest without disturbing the soil. Vickerman (1978) found fewer sawflies in emergence traps on ploughed than unploughed land but numbers caught were very low and the results not significant. This study has been able to quantify the effect of soil cultivation on overwintering sawfly prepupae and pupae for the first time, showing that it halved the numbers emerging successfully as adult sawflies. Analysis of data weighted by this mortality factor estimated that about 30% of the disparity in catches of Dolerus larvae between crop types was due to differences in cultivation regimes.

Pasture fields, with no soil cultivation and grasses that were likely to be suitable host-plants, might have been expected to have been an ideal habitat for sawfly larvae. However, catches in pasture fields were low except in those that had not recently been grazed. It is possible that to some extent this difference was due to variations in efficiency of sweeping in this habitat due to variable grass length, but it is likely that grazing renders permanent pasture fields a poor habitat. Grazing will disturb the host plant during the sawflies’ breeding period, causing direct egg mortality and removing the leaves at the top of growing shoots which are the preferred oviposition sites (Barker & Maczka 1996).

Farms used in this study were not organic. They used standard fungicide applications on all three sown crops, and often applied an autumn aphicide to winter wheat; there is evidence that both pesticide types can depress subsequent summer sawfly populations (Ewald & Aebischer 1998). No summer insecticides were sprayed onto any fields before sampling took place, so differences between crops were not due to direct mortality from insecticide use. However, some winter wheat fields were sprayed with insecticide after the survey in two of the four years. Sawfly populations can be decimated by insecticide use and take many years to recover (Aebischer 1990), so in theory these applications might have had serious long-term effects on local sawfly abundance. However, there were relatively few larvae present in the sprayed fields, with higher numbers in neighbouring unsprayed herbage seed fields. Numbers on the farm overall were therefore not likely to be much reduced.

The effects of changes in farming practice

The results of this study emphasize how changes in farming practice may influence insect populations in ways that may not be obvious. For example, a reduction in the availability of suitable grass hosts on arable farms will have had negative effects on sawfly abundance, as will increases in the proportion of land cultivated after harvest or increased grazing pressure on grassland. Traditionally, mixed crop rotations provided grass undersown into spring cereal crops and left undisturbed until the following year. There has been a widespread shift in Britain away from such rotations into either arable farming, with a high proportion of cereal crops and no grassland, or stock farming, with heavily grazed pasture (Potts 1997).

Increases in the use of pesticides in farming have been cited as the cause of the declines of many farmland insect species (Campbell et al. 1997; Potts 1997). Increased usage of herbicides to control grass weeds may have reduced sawfly abundance by removing grasses from cereal fields (Vickerman 1974). More widespread summer insecticide use on cereals, however, is unlikely to have been the prime cause of sawfly decline on farmland. The proportion of the sawfly population affected by summer insecticide applications will be high only on farms where no undisturbed grassland habitats are available, and these populations will be already suppressed by soil disturbance and lack of quality host plants. Moreby et al. (1994) found higher sawfly densities on organic than conventional farmland, but in their study summer insecticide use and grass weed availability were the same in both systems. Contrasts between systems in crop rotations, choice of cereal varieties, winter insecticide use, and treatment of non-cropped land were all considered as possible reasons for the differences in density.

A reduction of insecticide input to cereal crops as part of a conservation strategy to restore sawfly populations on farms would therefore achieve relatively little. Provision of suitable host plants, preferably on land that will not be cultivated annually, would be more effective. Selectively sprayed conservation headlands might help by increasing grass weed availability, but they would subsequently be cultivated. Experimental trials have not shown increases in sawfly numbers on conservation headlands (Rands 1985; Moreby 1997). Better results might be obtained by reviving the practice of undersowing spring cereals with ley or grass seed crops or by planting semipermanent grass strips around field margins. Both these forms of land management are currently being encouraged in Britain under some options of the Arable Stewardship and Countryside Stewardship Schemes run by The Ministry of Agriculture, Fisheries and Food (Anonymous 1998a,b).

Acknowledgements

The authors would like to thank J. P. Hughes for carrying out the Scottish part of the cultivation experiment and allowing us to use his data, N.W. Sotherton and three anonymous referees for comments on the manuscript and N.J. Aebischer for statistical advice and additional comments. Many thanks to the farmers of the six farms used in the survey for allowing us to use their land and to Mr C. W. Passmore for his help in setting up the cultivation experiment on his farm. This research was supported by grants from The Esmeé Fairbairn Charitable Trust awarded to the Habitat Research Trust and from The Leverhulme Trust.

Received 22 January 1998; revision received 1 February 1999

Ancillary