Journal of Insect Conservation

, Volume 16, Issue 1, pp 13–23

The importance of viticultural landscape features and ecosystem service enhancement for native butterflies in New Zealand vineyards

Authors

    • Bio-Protection Research CentreLincoln University
  • Steve D. Wratten
    • Bio-Protection Research CentreLincoln University
Original Paper

DOI: 10.1007/s10841-011-9390-y

Cite this article as:
Gillespie, M. & Wratten, S.D. J Insect Conserv (2012) 16: 13. doi:10.1007/s10841-011-9390-y

Abstract

The fragmentation of habitats in intensively managed farming landscapes is often considered to be partly responsible for butterfly population decline in Europe and the USA. Although relatively little is known about New Zealand butterfly ecology, agricultural landscapes in lowland New Zealand are managed similarly to those in Europe and ecosystem services (ES) in these landscapes are generally at a low level. In the northern hemisphere, attempts are being made to address the problem through agri-environment schemes, but such farmer compensation is not available in New Zealand. Instead, landowner- and research-led initiatives are currently the only potential approaches. One such project in the Canterbury province, New Zealand, is the Greening Waipara project. This aims to return native plants to viticultural landscapes and enhance ES, and while research has sought to quantify economic benefits of the project, there has been no work to establish if the plantings are improving or are likely to improve non-target invertebrate biodiversity, for example arthropods that are not biocontrol agents. In the first study of its kind in New Zealand, butterfly surveys were conducted in vineyards and linear mixed modelling techniques were used to identify the most important vegetation and structural features which may influence butterfly distribution. While the native planting areas were not important for butterflies, remnant patches of native vegetation in unproductive areas were vital for sedentary species. These results are discussed in relation to the conservation of native species in New Zealand vineyards and in the context of conservation in and around farmland in general.

Keywords

ConservationHost plantsLycaena salustiusNectarVegetationZizina oxleyi

Introduction

In some countries such as the USA, the UK and Japan, the decline of butterfly populations has been linked with changes in land use and the fragmentation of habitats in intensively managed farming landscapes (Pyle 1976; Erhardt 1985; Sibatani 1992; Kleijn and Snoeijing 1997; Thomas et al. 2001). However, the study of butterflies and agriculture has demonstrated the importance to butterflies of farmland features such as field margins (Dover 1994; Feber et al. 2007), hedgerows (Dover and Sparks 2000) and remnant areas of native vegetation (Shepherd and Debinski 2005; Schmitt et al. 2008; Franzen and Nilsson 2008). In addition, to complement the monitoring data of programs such as the UK Butterfly Monitoring Scheme, recent farm-scale studies have sought to assess the influence of other biotic and abiotic factors influencing butterfly abundance and distribution in different landscapes (Dover 1996; Clausen et al. 2001; Pywell et al. 2004). An understanding of individual species’ requirements and the identification of influential features of the landscape is vital to butterfly conservation on farmland, particularly for some of the rarer and specialist species (Dennis 2004; Pywell et al. 2004). Much of this research can also inform the improvement of agri-environment schemes (AES) (Kleijn and Sutherland 2003; Kuussaari et al. 2007; Franzen and Nilsson 2008), in widespread attempts to encourage farmers to enhance the provision of ecosystem services (ES; the goods and services provided to mankind by nature) in agricultural landscapes, including the aesthetic ES of butterfly conservation.

In New Zealand, lowland agricultural areas are less diverse than those in Europe (Tscharntke et al. 2007), with less than 1% of native vegetation remaining (Meurk 2008) and biodiversity is likely to be at a similar or lower level. Despite this, farmer compensation schemes such as AES do not currently exist. Instead, landowner- and research-led initiatives are the only potential approaches to enhancing ecosystem services and biodiversity on farmland. An example of this is a government funded project called ‘Greening Waipara’ (http://www.bioprotection.org.nz/greening-waipara) led by Lincoln University and Landcare Research (http://www.landcareresearch.co.nz/) in the viticultural region of Waipara, North Canterbury. The project is encouraging and assisting participating vineyard owners to replant some of the once common native plant species in and around the properties. Prior to anthropogenic disturbance, Waipara was originally covered by tall native forests of totara (Podocarpus totara G. Benn. (Podocarpaceae)), matai (Prumnopitys taxifolia Banks & Sol. (Prumnopityaceae) and kowhai (Sophora spp. L. (Fabaceae)), and shorter kanuka (Kunzea ericoides A. Rich. (Myrtaceae))-kowhai woodland (Meurk 2008). Butterflies would have taken advantage of the opening of the vegetation and halting of succession by browsing flightless birds such as the now extinct moa (Struthioniformes: Dinornithidae), and of the ridges, scarps, coasts and young terraces covered with matagouri (Discaria toumatou Raoul (Rhamnaceae)), native grasses and divariacting shrubs. Polynesian settlers then introduced fire and land mammals from c. 1,000 years B.P. (Davidson 1984) causing greater opening of the vegetation, which may have benefitted many of the butterfly species as grasslands and shrublands became more prevalent (McGlone 1989). European settlement, widespread in the 1840 and 1850s, brought extensive land clearing through grazing and mixed farming with exotic and more competitive plant species. Later, forestry, horticulture and viticulture completed the transformation of the landscape to that seen today. Commercial viticulture in particular was introduced in Waipara in the 1980s and expanded to around 80 vineyards, so that viticulture covers 1442.8 ha of the Waipara valley (New Zealand Winegrowers 2009). Today, native vegetation in Waipara is restricted to pockets of matagouri and associated shrubs and grasses growing on marginal and unproductive land such as steep slopes and rocky ground, although there is an extensive area of native vegetation on the limestone pavement tops of Mount Cass (523 m.a.s.l).

The emphasis of the project is on providing multiple ecosystem services (Fiedler et al. 2008), such as weed and disease suppression through planting low-growing plants beneath vines, natural enemy population enhancement via flowering plants and cultural services from educational biodiversity trails close to wineries and other native plantings outside vine blocks. While research surrounding the plantings has focussed on tangible economic benefits (Berndt et al. 2006; Tompkins 2009), there has been no work to evaluate the impact on general biodiversity not specifically targeted by the plantings. In addition, the study of butterflies in New Zealand is undeveloped, especially compared with Europe (Gibbs 1980). Although there are no endangered species (Patrick and Dugdale 2000), this may be because of poor population records. Populations on farmland are especially likely to face threats similar to their European counterparts. For example, although the New Zealand wine industry promotes limited use of broad-spectrum pesticides (Scarratt 2005), the use of land for viticulture is one of the most intensive forms of agriculture in terms of monocultures and the fragmentation resulting from field re-alignment in the conversion from arable or pastoral farming (Schmitt et al. 2008). While ecosystem services in such extreme monocultures are at a low level (Fiedler et al. 2008), there are opportunities to enhance them and to halt or reverse the assumed decline in native butterfly populations.

The Greening Waipara project provides an ideal opportunity to study both the butterfly fauna of New Zealand vineyards and its requirements, and the extent to which ecosystem service enhancement is improving or has potential to improve the biodiversity of the viticultural landscape. This study explores the factors that influence butterfly diversity and abundance within one region of New Zealand vineyards. It consists of the first known survey of butterfly species on farmland of any type in New Zealand and the first attempt at associating farm-scale features with the New Zealand butterfly fauna using linear mixed modelling.

Methods

Study sites

All measurements were carried out on and around six vineyards in the Waipara Valley, North Canterbury (Fig. 1). According to Gibbs (1980), this region is well within the national distribution of ten of the 23 butterfly species known to reside in or visit New Zealand.
https://static-content.springer.com/image/art%3A10.1007%2Fs10841-011-9390-y/MediaObjects/10841_2011_9390_Fig1_HTML.gif
Fig. 1

Location of the Waipara Valley within New Zealand (top) and the vineyards used for butterfly surveys in the Waipara valley (bottom)

The term “vineyards” in this study includes all parts of a property of which the predominant product is wine grapes. Therefore, despite comprising largely monocultural areas, the vineyard sites included a wide range of vegetation types not directly used in wine production such as pasture, wide field margins for turning machinery, gravel tracks, river beds and ‘Greening Waipara’ planting areas. In addition, each site was chosen due to the presence of areas of remnant native vegetation (hereafter termed “remnants”), either on the property itself or on the neighbouring land. These remnants typically consisted of stands of matagouri, Muehlenbeckia spp. Meisn. (Polygonaceae), NZ bindweed (Calystegia tuguriorum G. Forst. (Convolvulaceae)), as well as European Gorse (Ulex europaeus L. (Fabaceae)) and Broom (Cytisus scoparius L. (Fabaceae)) and a range of exotic herbs and grass species. The remnants were also considered marginal and relatively unproductive land due to shallow soils or impractical topography, although many of them were grazed occasionally by stock. The vineyards were spread across the valley so that no study site was closer than 1.5 km to its nearest neighbour. The sites all underwent typical vineyard management operations, although there were differences between properties in mowing and spraying regimes for example, which have not been accounted for.

Butterfly recording

Butterflies were counted along fixed transect routes following the standard methodology used by the UK Butterfly Monitoring Scheme (UKBMS) described by Pollard et al. (1986) and Pollard and Yates (1993). Transects were established at each site so that they passed through the different vegetation types represented on or adjacent to the properties. Homogeneous sections of each transect were identified on the basis of topography and vegetation type and data were pooled for each section. Transects had between 9 and 14 sections, with a total of 66 sections across the six sites. The total length of transects was 14,065 m.

Transects were walked every 2 weeks between October 2008 and April 2009, covering the flight periods of all common butterfly species, with a total of 13 visits per transect. The route was walked at a steady pace counting all butterflies encountered up to 2.5 m either side and in front of the observer. Where possible, the butterflies were recorded to species and sex. Where the species was uncertain, the most common species was assumed. In the case of the blue butterfly species, the Southern blue (Zizina oxleyi C & R Felder (Lycaenidae)) and the Common blue (Zizina labradus Godart (Lycaenidae)) are difficult to distinguish in the field. However, in a separate molecular and morphological study of 45 blue butterflies sampled from around Waipara, all specimens were classed as Z. oxleyi (Gillespie 2010). All blue butterflies encountered were therefore assumed to be Z. oxleyi.

Environmental variables

Biotic and abiotic parameters were recorded in January 2009 and are listed in Table 1. The choice of variables followed those of Dover (1996), Clausen et al. (2001) and Pywell et al. (2004) where applicable. The first variable, vegetation type, categorised each section into one of seven possible types based on predominant vegetation and the principal vineyard operation:
Table 1

Environmental variables recorded for each transect section

Variable

Description and scale

Vegetation type

Vinerows, margin, river, planting, pasture, track, remnant; as described in text

Insolation

Estimated from aspect 0 = S, 2 = SE or SW, 4 = W or E, 6 = NE or NW, 8 = N

Shelter

Estimated on 0–8 scale following Dover (1996); number of directions from which section is sheltered from wind

Slope

Presence of slope: 0 = no slope, 1 = slight slope, 2 = steep slope.

Shade

Degree of shade over section: 0 = no shade, 1 = under 50% shaded, 2 = over 50% shaded

Grazing

0 = ungrazed, 1 = grazed by rabbits only, 2 = grazed by rabbits and occasionally by stock (sheep or cattle), 3 = heavily grazed

Mowing

0 = no mowing, 1 = occasional mowing, 2 = frequent mowing

Grass height

Mean height of grass (N = 20 per section)

Vegetation height

Mean height of ground (non woody) vegetation (N = 20 per section)

Native Cover

% cover of native plant species estimated for length of section

Tree Cover

% cover of trees estimated for length of section

Shrub cover

% cover of all shrub species for length of section,

Herb cover

% cover of all herb species for length of section

Grass cover

% cover of grasses for length of section

Bare Ground cover

% cover of bare ground for length of section

Nectar

Summed nectar abundance score per 100 m of transect section

Nectar richness

Maximum number of flowering species per section over summer

Useable nectar score

Summed nectar abundance score per 100 m of transect section of those species used by adult butterflies

  • Vinerows—grape vines through which the transects passed via a single row. Ground vegetation consisted mainly of exotic grasses dominated by perennial ryegrass, bare ground and other agricultural weed species in the Asteraceae and Leguminosae.

  • Margin—consisted of ground vegetation similar to the above, but lacked the cover of the vines, and often included additional flowering perennials at fencelines e.g., members of the Malvaceae and Boraginaceae.

  • River—sections passing along the gravel bed of the Waipara River, predominantly vegetated by gorse, European broom, Muehlenbeckia axillaris Hook f. (Polygonaceae) and occasional agricultural weed species.

  • Planting—the Greening Waipara native planting sites on the vineyards. On the sites studied, these were small (100–200 m2) areas located close to buildings or situated away from main operations. They were planted with a range of native shrub and grass species and were usually bordered by vegetation similar to that of margins.

  • Pasture—areas kept for stock grazing, with typical vegetation of exotic grasses, members of the Leguminosae, Asteraceae and other common agricultural weed species.

  • Track—any gravel roads used by farm vehicles and which were often bordered by or infested with common agricultural weed species.

  • Remnants—as defined in the Study sites section.

The remaining variables in Table 1 were recorded for each transect section in the 5 m strip of land in which butterflies were observed while walking the transects. Also, the vegetation composition of each section was estimated using quadrats. Each transect section over 100 m was divided into 10 subsections in which a circular quadrat of 1.5 m diameter was placed randomly. Where sections were less than 100 m in length, subsections were allocated every 10 m. Within each quadrat, the percentage cover of each plant species was estimated visually. For simplicity in data analysis, these observations were subsequently grouped by family (except for grasses which were grouped together), and additionally into native and non-native species. The abundance of nectar sources was estimated using the five point DAFOR scale (sensu Clausen et al. 2001) (5 = D—dominant, 4 = A—abundant, 3 = F—frequent, 2 = O—occasional, 1 = R—rare) and this was recorded during most of the butterfly transect sessions. The number of each species in flower was visually estimated in blocks of 20 inflorescences. A particular flowering species was considered dominant if more than 10 blocks were counted, abundant if 5–10 blocks were counted, frequent if 2–5 were counted, occasional with 1–2, and rare if counts did not constitute a block of 20. Nectar source abundance categories were summed for each section for an overall nectar abundance score, and the total species richness across the summer was also calculated. In addition, a nectar abundance score was calculated to include only those nectar sources that butterflies are known to use or were observed using.

Statistical analysis

While the transect method developed by Pollard and Yates (1993) remains the most effective way to collect population data on butterflies, it was designed for monitoring purposes rather than to analyse habitat feature associations. In regression analyses therefore, the data suffer from problems of temporal and spatial pseudoreplication due to repeated measurements and the close proximity of transect sections, respectively. This leads to the inflation of the degrees of freedom and violation of the independence of errors assumption (Crawley 2007). In the present study, temporal pseudoreplication was removed by summing all butterfly counts of individual species and of total butterflies over the summer for each transect section.

To address spatial pseudoreplication, a mixed effect modelling approach was chosen to analyse the data, as this allows the incorporation of a random factor for site to control for spatial non-independence arising from grouped observations (Venables and Ripley 2002). Initially, the summed data were analysed in R 2.9.2 (R Core Development Team 2009) using generalised linear mixed models with penalised quasi-likelihood (GLMM PQL) using a Poisson error distribution. However, for species richness the model was inadequate due to non-constancy of variance. The butterfly abundance and individual butterfly species abundance data were found to follow a negative binomial distribution, and fitting this using the glmm.ADMB library lead to convergence problems. It was therefore decided to transform the response variable data and fit linear mixed effects models.

To normalise the residuals, the number of butterfly species (species richness) and the total butterfly individuals were transformed by first dividing observations by the log length of the section, and then taking the log of these standardised values. Total abundance of Z. oxleyi was transformed in a similar way, except that the formula was log(n + 1/log(section length)) to account for the presence of zeros. It was not possible to transform data of other species satisfactorily due to the large number of zero counts, so data for these were not analysed further. However, these numbers were included in the data for species richness and total butterfly abundance.

The first analysis was a separate linear mixed model for each of the three response variables (species richness, total butterfly abundance, abundance of Z. oxleyi) with vegetation type as the explanatory variable and site as random factor. Three further models were then parameterised for each response variable using the remaining environmental (explanatory) variables in Table 1. Initially, these variables were tested for collinearity using the Pearson product moment correlation (r). If two variables had a value of r = 0.7 or above, one of them was removed from the analyses. The remaining variables were then added to the analyses and backwards stepwise selection was used to eliminate those that were not significant at the P = 0.01 level. The variables that showed a high level of significance were then considered to be those that best explained variation in the response variables (Zuur et al. 2009).

The 0.01 cut-off criterion was used as opposed to the traditional 0.05 level because regression models may still suffer from multi-collinearity between explanatory variables despite initial testing. The data may also violate the assumption of independence on a finer scale than site due to the varying proximity of sections to one another (Dover 1996). A strict selection criterion can reduce this possibility.

A final set of linear mixed models were performed using individual flower species as explanatory variables and site as random factor. Backwards stepwise selection was used in the same way as above, until only significant terms remained (the flower species that were significant in explaining the variation in response variables). All analyses were performed using R 2.9.2 (R Core Development Team 2009).

Results

Butterfly assemblage in Waipara vineyards

A total of 2988 butterfly individuals were recorded on 13 visits to each of the six vineyards. The butterfly fauna was limited to a maximum of five species on any one site at any one recording session, a maximum of seven species over the summer on any one site, and a total of eight species over the summer (Table 2). The majority of individuals belonged to the Lycaenidae (87%); 74% were Zizina oxleyi, which was prevalent in most places due to the ubiquity of its host plants, members of the Fabaceae. The New Zealand lycaenids, like those in other countries, are relatively sedentary species, and were almost always observed close to host plants. The other four species recorded are all mobile and as the low number of observations suggest, were most often seen on the wing passing through the sites. With the exception of the pest, the cabbage white (Pieris rapae L. (Pieridae)), it cannot be stated with any confidence whether any of these mobile species were breeding at the study sites as none of their host plants was encountered. The site with the highest mean species number was Waipara West (2.1 ± 0.5) and the lowest was Fancrest (0.9 ± 0.1). The highest mean abundance of butterflies per 100 m was observed at Greystone (36.5 ± 13.2) and the lowest was at Dickson vineyard (11.2 ± 3.4).
Table 2

Mean butterfly abundance and species richness/100 m of transect recorded at each of the 6 study sites (n = number of sections after removal of those where no butterflies were found)

Species

 

Riverside sites

Hillside sites

 
  

Dickson

n = 9

Dunstaffnage

n = 9

Waipara West

n = 12

Fancrest

n = 8

Greystone

n = 8

Mountford

n = 10

All

n = 56

Lycaenidae

 Zizina oxleyi

Southern blue

6.8

4.8

9.5

27.0

30.5

18.9

97.5

 Lycaena salustius

Common copper

2.8

2.1

3.3

0.7

0.8

2.2

12.0

 Lycaena feredeyi

Glade copper

0.1

0.1

0.1

0.0

0.0

0.1

0.4

 Lycaena boldenarum

Boulder copper

0.0

3.5

1.1

0.0

0.0

0.0

4.6

Nymphalidae

 Vanessa gonerilla

Red admiral

0.0

0.0

0.0

0.1

0.0

0.1

0.2

 Vanessa itea

Yellow admiral

0.1

0.0

0.1

0.2

0.1

0.1

0.6

 Danaus plexippus

Monarch

0.0

0.0

0.1

0.0

0.1

0.0

0.2

Pieridae

 Pieris rapae

Cabbage white

1.4

2.7

3.4

0.8

5.1

2.8

16.3

 

Total butterflies

11.2

13.2

17.4

28.8

36.5

24.0

131.1

 

Species richness

1.7

1.3

2.1

0.9

1.2

1.4

1.5

Across all sites, two out of three of the most abundant lycaenids showed clear fidelity to certain vegetation types (Fig. 2). Lycaena salustius Fabricus (Lycaenidae) was found almost exclusively in the remnants where its larval host plants were abundant. Similarly, L. boldenarum White (Lycaenidae) was restricted to the section types where its larval host was present. In contrast, Z. oxleyi was relatively well represented in all vegetation types. This distribution of species also illustrates the relative importance of the different vegetation types. In particular, while the remnants were most important for butterflies, the plantings were least important.
https://static-content.springer.com/image/art%3A10.1007%2Fs10841-011-9390-y/MediaObjects/10841_2011_9390_Fig2_HTML.gif
Fig. 2

Mean number of adult butterflies of three of the lycaenid species recorded in surveys on different vegetation types, with 1 ± SE. Dark grey bars: Z. Oxleyi; light grey bars: L. boldenarum; white bars: L. salustius

Host plant species were encountered for all four of the lycaenids and for P. rapae. For Z. oxleyi, white clover (Trifolium repens L. (Fabaceae)) was the most abundant host plant species. Muehlenbeckia complexa A. Cunn (Polygonaceae) was the most abundant host for both the L. salustius and L. feredeyi Bates (Lycaeanidae), although M. australis G. Forst (Polygonaceae) and M. astonii Petrie (Polygonaceae) were present in smaller numbers; M. astonii was a frequently used species in the Greening Waipara plantings. Muehlenbeckia axillaris, the host plant of the L. boldenarum was found either abundantly in river beds or was irregularly seen in some of the remnants. Host plants for P. rapae included hedge mustard (Sisymbrium officinale L. (Brassicaceae)) and brassica crops in nearby arable farmland.

The most common nectar sources used by butterflies, and observed being used by NZ butterflies could all be categorised as agricultural weeds. Species common at most sites were: dandelion (Taraxacum officinale Weber (Asteraceae), mean total abundance score per section 7.8 ± 0.7), T. repens (4.8 ± 0.8), creeping thistle (Cirsium arvense L. (Asteraceae), 2.9 ± 0.5), and common yarrow (Achillea millefolium L. (Asteraceae), 1.9 ± 0.6). However, the most abundant flower species unused on certain sites were Viper’s bugloss (Echium vulgare L. (Boraginaceae), 4.9 ± 0.8), large flowered mallow (Malva alcea L. (Malvaceae), 4.3 ± 0.8), small flowered mallow (Malva parviflora L. (Malvaceae), 2.3 ± 0.5) and dog rose (Rosa canina L. (Rosaceae), 1.1 ± 0.3). Native flowers consisted only of C. tuguriorum (0.2 ± 0.1) which was not used, and Muehlenbeckia species (M. axillaris: 0.4 ± 0.3, M. complexa: 1.4 ± 0.5) which were used occasionally.

Linear regression analysis: vegetation type

The first linear mixed model analysis of the response variables modelled against vegetation type gave a highly significant effect of type for species richness (F = 5.03, df = 6, 49, P < 0.001, Deviance explained = 46.7%). There was also a significant effect for total butterfly abundance (F = 2.98, df = 6, 49, P = 0.015) which was slightly outside the P = 0.01 criteria used in this study. Subsequent examination of the effects of the different section types suggests that the remnants had significantly higher species richness and butterfly abundance than the other types apart from the river sections (Fig. 3). However, there were only three river transect sections. Multiple comparisons were therefore not conducted on these data due to the unbalanced nature of the dataset. The total amount of each vegetation type in the transects is shown in Table 3.
https://static-content.springer.com/image/art%3A10.1007%2Fs10841-011-9390-y/MediaObjects/10841_2011_9390_Fig3_HTML.gif
Fig. 3

a Mean butterfly species richness per log-length of section (m) plotted by vegetation type with 1 ± SE; b mean butterfly individuals per log-length of section (m) plotted by vegetation type with 1 ± SE

Table 3

The total length in metres of each vegetation type included in the transect surveys and percentage of total transect length

Vegetation type

Total length (m)

Percentage of total transect length

Vinerows

1,395

10.0

Margin

2,250

16.0

River

895

6.3

Planting

390

2.8

Pasture

3,420

24.3

Track

2,360

16.8

Remnant

3,355

23.9

Multiple regression analysis: environmental variables

In the multiple regression analysis using environmental variables (Table 4) the only significant explanatory variable for species richness was percentage cover of native plant species. The total abundance of butterfly individuals was explained by the cover of native plants and the abundance of useable nectar sources. Variation in the abundance of Zizina oxleyi was explained by the abundance of useable nectar sources plants and the percentage cover of legumes.
Table 4

Significant environmental variables in the linear mixed effects models of species richness, total butterfly abundance and abundance of Zizina oxleyi, after backwards stepwise selection

Model

Estimate

t

P value

Species richness

 Intercept

−1.00

  

 Native cover

0.02

4.06

<0.001

 Deviance explained 31.8%

   

Total individuals

 Intercept

0.27

  

 Native cover

0.03

2.83

0.007

 Useable nectar score

0.04

6.20

<0.001

 Deviance explained 46.1%

   

Zizina oxleyi

 Intercept

−0.15

  

 Fabaceae cover

0.04

2.90

0.005

 Useable nectar score

0.03

4.15

0.0001

 Deviance explained 39.4%

   

Mutilple regression analysis: nectar abundance

The linear mixed model analyses of the response variables modelled by the abundance of individual flowering species are shown in Table 5. The most important flower species explaining butterfly species richness were C. arvense and T. repens. Cirsium arvense was also important for abundance of individuals, together with T. officianale. For the model of Z. oxleyi abundance, the resulting residuals suffered from non-constancy of variance. This was addressed by transforming the explanatory variable (log(n + 1)) and resulted in a significant effect of the abundance of T. repens flowers on Z. oxleyi abundance.
Table 5

Results of linear multiple regression analysis of nectar sources with species richness, total individuals and Z. oxleyi individuals after backwards stepwise selection

Model

Estimate

F

P value

Species richness

   

 Intercept

−0.91

  

 Cirsium arvense

0.06

4.56

<0.001

 Trifolium repens

−0.03

−3.03

0.004

 Deviance explained 39.1%

   

Total individuals

   

 Intercept

1.91

  

 Cirsium arvense

0.13

4.03

<0.001

 Taraxacum officianle

0.10

3.70

<0.001

 Deviance explained 42.1%

   

Zizina oxleyi

   

 Intercept

−3.13

  

 Trifolium repens (log(n + 1))

0.52

3.41

0.001

 Deviance explained 22.6%

  

Discussion

Butterfly assemblage in Waipara vineyards

The number of butterfly species recorded visiting or residing in Waipara vineyards may seem low compared with European agricultural areas. However, the New Zealand butterfly fauna comprises an extremely small number of species compared with most countries in the northern hemisphere. The reasons for this are unknown, although they may be linked to the relative isolation of the archipelago and the changes that took place during the Gondwana separation (Gibbs 1980). Of the 23 butterfly species known to reside in or visit New Zealand, the distribution of ten includes the Waipara region. Therefore, the recording of eight species is actually quite high for this region.

Of the mobile species recorded in this study, only the introduced pest P. rapae occurred in large numbers. The remaining three mobile species encountered, the two natives Vanessa itea Fabricius (Nymphalidae) and V. gonerilla Fabricius (Nymphalidae), and the introduced Danaus plexippus L. (Nymphalidae), were observed feeding from non-native nectar sources, but could not be considered to be supported by the vineyards because of their extremely low numbers and an absence of larval host plants. The less mobile butterflies consisted of four native species of the Lycaenidae. Only the southern blue, Z. oxleyi, the most abundant butterfly overall by far (74%), occurred in all vegetation types, which can be explained by the ubiquity of the species’ larval host plants, members of the Fabaceae. This is supported by the importance of Fabaceae cover in the model of Z. oxleyi abundance. The three other lycaenids were restricted to the remnants and river beds that contained their Muehlenbeckia spp. host plants. For example, L. salustius and L. feredeyi were restricted almost entirely to the remnants of native vegetation on or adjacent to vineyard properties. When L. salustius was recorded in other vegetation types, it was observed nectaring, suggesting that it may search for food sources away from the host plant patches. This is supported by Gibbs (1980) who observed the butterfly nectaring up to 250 m from the nearest larval host plant. Lycaena boldenarum was recorded only in association with M. axillaris, the sole host plant of this species, which was most often found on the gravelly river banks adjacent to Dickson, Dunstaffnage and Waipara West vineyards, although some records of both plant and butterfly came from the remnants at the same sites.

Vegetation associations of butterflies in Waipara vineyards

Few butterflies were associated with the plantings of the Greening Waipara project. Zizina oxleyi was the most abundant in this vegetation type, but this is likely to be due to the presence of legumes in the vicinity. While some M. astonii plants were included in most of the plantings, these sites were isolated from native patches, with apparently insufficient resources to attract L. salustius. While more mobile species are not constrained by larval host plants in their distribution on farmland (Pywell et al. 2004), a general lack of suitable nectar sources and host plants in plantings and elsewhere is likely to explain at least in part the paucity of these species on vineyards.

More species and individuals were recorded in the remnants than in any other vegetation type. The remnants were the main location for native shrubs, which was an important factor in both species richness and abundance models. The presence of Muehlenbeckia spp. and associated Lycaena spp. are likely to have inflated both summary response variables (species richness and total butterfly abundance) for such a small fauna. However, aside from providing the main source of larval host plants in vineyards, the importance of the remnants and native shrubs to the New Zealand Lycaenidae is unsurprising given the structural heterogeneity they provide. Although resource requirements are not known in detail for these butterfly species, it is likely that, in Waipara vineyards, remnant stands of native shrub vegetation are also the best sources of utility resources, i.e., non-consumable aspects of the vegetation that are also important components of a butterfly habitat under a resource-based definition (Dennis 2010). For example, remote from much of the disturbance of intensive agricultural operations such as machinery use and chemical application, which are detrimental to butterflies (Marini et al. 2008; Franzen and Nilsson 2008), remnants are likely to provide shelter in otherwise open farmland (Dover 1996; Dover et al. 2000; Pywell et al. 2004), corridors to facilitate movement between resource patches (Haddad 1999) and structural resources for roosting, perching, basking and mate location (Dover and Fry 2001; Dennis 2004; Hardy and Dennis 2007, Dennis 2010). Fallow, marginal or unproductive lands are also important refuges of specialist butterflies in urban areas (New 2007), European vineyards (Schmitt et al. 2008) and other types of farmland (Clausen et al. 2001; Summerville et al. 2005; Ockinger et al. 2006; Kuussaari et al. 2007).

The remnants may even represent the closest approximation to ‘natural habitat’ for New Zealand butterflies in lowland agricultural settings. Prior to Polynesian settlement, endemic butterflies such as L. salustius are likely to have persisted close to their host plants in seral shrubland communities maintained by browsing flightless birds such as moa. Fires started by Polynesians for land clearance may have similarly created or maintained such vegetation structure following the extinction of these birds, before the degradation that accompanied European settlement (McGlone 1989).

The abundance of useable nectar sources was also important in this work in explaining variation in total butterfly abundance and the abundance of Z. oxleyi. In agricultural landscapes, where perennial flowering plants are removed directly or indirectly through the use of artificial fertilisation (Boatman 1992), nectar sources can be the limiting factor for many butterfly species. Nectar has been correlated with butterfly abundance and species richness in a number of similar studies (Dover 1996; Clausen et al. 2001; Pywell et al. 2004; Nelson and Wydoski 2008). In this study, C. arvense, Trifolium repens and Taraxacum officinale were important species. The relationship between the abundance of these exotic plants and the abundance of butterflies suggests that native New Zealand butterfly species have adapted to introduced plants in agricultural landscapes. However, New Zealand farmers are unlikely to be receptive to the idea of encouraging such ‘weed’ species in the name of butterfly conservation, and the improvement of native flower provision is recommended.

While the findings of the regression analyses can be interpreted with apparently simple explanations, butterfly habitat provision is complex and the results should be considered along with the potential drawbacks of the methodology. In particular, the limitations of linear regression using count data include the possibility of lost information during the pooling of data across the summer and through data transformation (Zuur et al. 2009). The data may also suffer from multi-collinearity and a lack of independence (Kivinen et al. 2006) at finer scales than at the site level and the use of multiple explanatory variables can result in the identification of significant values for terms by chance or due to correlation between variables (Dover 1996; Clausen et al. 2001). Every effort has been made to minimise collinearity in the present study and the parsimonious 0.01 significance level has been used in stepwise selection to reduce the likelihood of these possibilities. In future studies, the use of transect sections of a standard length (e.g. 100 m) and spatially remote from each other depending on the mobility of the butterfly fauna studied, as used by Kivinen et al. (2006) and Kuussaari et al. (2007) for example, are likely to reduce these problems further.

Conservation implications and the importance of enhanced ecosystem services

In Waipara vineyards, the absence of farmland features traditionally associated with butterflies, such as hedgerows and florally rich field margins, is reflected in the low numbers of mobile and specialist butterfly species. Such a degraded landscape may represent the end product of intensification of agriculture with only the most successful generalists being supported (P. rapae, Z. oxleyi). Furthermore without farmer incentive schemes such as AES, conservation relies on the enthusiasm of landowners and research extension activities. The extent and nature of funding of research-led projects is also important. The scenario presented here may illustrate the prospects for European agriculture should AES schemes fail to be effective or renewed.

While the Greening Waipara plantings were the least important vegetation type because of their recent history, low numbers of host plants and isolation from larger patches of native vegetation, there is potential to link these areas to the remnant vegetation. These latter sites represent important patches of resources for butterflies in Waipara vineyards; for the less mobile species, they are isolated patches and are likely to form the core locations of their resources. For the more mobile butterflies, they may provide important sources of nectar (e.g., C. arvense) that are absent from other areas, although host plants for these mobile species were not found. Thus, remnants have much potential to form the focus of future conservation efforts on farmland and, while butterflies were not the target of the native plantings, such areas could be developed with some form of connectivity to the larger remnants and more attention being paid to planting of appropriate butterfly host plants and nectar sources. For example, climbing shrubs such as M. complexa and other shrubs rich in nectar sources such as Hebe spp. (Gillespie 2010) would make ideal hedge plants, and may further enhance multiple ecosystem services (Fiedler et al. 2008) including those associated with beneficial insects. Furthermore, additional research should focus on effective management strategies for increasing the abundance of the mobile species and attracting those species not recorded here but known to occur nearby. The tussock butterflies (Argyrophenga spp. Doubleday (Nymphalidae)) in particular are a poorly studied group that utilise tussock grasses included in the Geening Waipara plantings. These butterflies occur in the nearby Mount Cass area (M. Gillespie, pers. obs.) and warrant investigation in relation to the prospects of their colonisation of viticultural and other farm landscapes.

Many other workers have emphasised that providing habitats for a number of butterfly species does not involve only the improvement of larval and adult food supply, but should also aim at creating general structural heterogeneity, providing resources that can be used for different aspects of adult behaviour and life cycle stages (Clausen et al. 2001; Pywell et al. 2004). This emphasis can also be extended to the whole farm, or the so-called matrix (the area between habitat patches as defined largely by host plants (Dennis and Hardy 2007)), which may make conservation more successful at larger scales. However, in vineyards, the challenge remains in convincing landowners to commit to conservation measures and to coordinate these activities spatially and temporally. In Europe, the use of AES to address this challenge has had variable success (Kleijn et al. 2006) and schemes are more likely to have positive outcomes when landowners in the surrounding landscape adopt similar practices (Rundlof et al. 2008). Such financial compensation is not available to New Zealand vineyard owners and managers, and may not always be available in Europe. On vineyards, landowners are unlikely to be able to incorporate many diversification ideas into working practice, due to the semi-permanent nature of the crop and the dedication of resources to the production of a high-value product, wine. Instead, incentives in the form of evidence of the economic benefits of habitat management techniques are required. To this end, subsequent research should be focussed on linking the conservation of butterflies with beneficial outcomes (i.e., multiple ecosystem service enhancements) and the aesthetic and economic effects on tourism and marketing. Such work would be of benefit to global agricultural biodiversity in encouraging the adoption of AES or alternative evidence-based projects.

Acknowledgments

This work could not have been completed without the fieldwork assistance of Mariska Anderson and Emma Thomas. Thanks also to Brian Patrick, George Gibbs, Tim New and Nick Sotherton for valuable advice and comments on the work as it progressed, and to the vineyard owners who allowed us access to their properties. Mark Gillespie was funded by a New Zealand Educated International Doctoral Research Scholarship (NZIDRS) and by the Bio-Protection Research Centre of Lincoln University and the New Zealand Foundation for Research Science and Technology (FRST; LINX 0303).

Copyright information

© Springer Science+Business Media B.V. 2011