Reduced pesticide toxicity and increased woody vegetation cover account for enhanced native bird densities in organic orchards

Authors

  • Catriona J. MacLeod,

    Corresponding author
    1. Landcare Research, Private Bag 1930, Dunedin 9054, New Zealand
      Correspondence author. E-mail: macleodc@landcareresearch.co.nz
    Search for more papers by this author
  • Grant Blackwell,

    1. Agriculture Research Group on Sustainability, Centre for Study of Agriculture, Food, Energy and Environment, University of Otago, PO Box 56, Dunedin 9054, New Zealand
    Search for more papers by this author
  • Jayson Benge

    1. Agriculture Research Group on Sustainability, The Agribusiness Group, c/o ZESPRI International Limited, PO Box 4043, Mt Maunganui 3149, New Zealand
    Search for more papers by this author

Correspondence author. E-mail: macleodc@landcareresearch.co.nz

Summary

1. Organic farming is often promoted as a solution for counteracting the adverse impacts of agricultural intensification on biodiversity. However, it is unclear whether the biodiversity benefits derived from organic farming require an adoption of organic farming in its entirety (i.e. a systems-level approach) or whether the benefits derived are because of just a small subset of the associated management practices.

2. Using bird survey data collected from kiwifruit orchards in New Zealand, we assessed whether orchards managed under an organic system support higher bird densities than those under integrated management systems. To determine whether biodiversity gains might also be achieved on non-organic orchards, we tested whether variation among kiwifruit orchards in the amount of (non-crop) woody vegetation cover, density of shelterbelts and toxicity of pesticide applications are better predictors of bird densities than management systems.

3. Composite measures of breeding season densities of all native species and the subset of native insectivores were higher on organic orchards than integrated management orchards. Densities of introduced bird species were comparable among management systems.

4. Pesticide use and habitat composition variables were better predictors of native bird densities than management system, with native bird densities negatively associated with pesticide toxicity ranking and/or positively associated with woody vegetation cover.

5.Synthesis and applications. A complete conversion to an organic system may not be required to improve biodiversity in agroecosystems. Instead, the transfer of specific land management practices known to benefit biodiversity in organic systems has the potential to enhance biodiversity in other more intensively managed systems (e.g. integrated management). This may be a path towards attaining biodiversity benefits at a larger scale, because such changes may be more straightforward than conversion to an organic system.

Introduction

Loss of biodiversity and ecosystem services (e.g. nutrient recycling, hydrological processes and carbon sequestration) in agricultural landscapes in response to changes in land management practices, particularly intensification, is a major concern at both global and national scales (Matson et al. 1997; Altieri 1999; Foley et al. 2005; Butler, Vickery & Norris 2007; Power 2010). In Europe and North America, for example, significant declines in common and widespread farmland birds have been linked to agricultural intensification (Krebs et al. 1999; Chamberlain et al. 2000; Donald, Green & Heath 2001), with similar trends observed for plants and invertebrates (Wilson et al. 1999; Benton et al. 2002). However, robust empirical information for assessing the effectiveness of different governance strategies in supporting ecosystem services, resource sustainability and biodiversity in farmland is limited (Kenward et al. 2011). For example, surprisingly, little is known about how farmland biodiversity declines are linked to land-use intensity (Kleijn et al. 2009).

Organic farming systems are often promoted as a potential solution to reverse the impact of agricultural intensification, as these systems tend to support higher species abundance and/or diversity for a range of taxa than conventional systems (Bengtsson, Ahnström & Weibell 2005; Fuller et al. 2005; Hole et al. 2005; Gomiero, Pimental & Paoletti 2011), and have less negative environmental impacts (Reganold et al. 2001; Lotter 2003). However, in ecological terms, it is unclear whether an organic systems-level approach is required to enhance biodiversity on farms or if implementing one or more key management practices associated with organic systems could also benefit biodiversity in other systems (Hole et al. 2005; Gomiero, Pimental & Paoletti 2011). This is a fundamental gap in our understanding, because the cost of implementing management actions for biodiversity benefit in agricultural landscapes would be greatly reduced (and hence such actions more likely to be adopted) if only one or two key changes, as opposed to whole system changes, were required for benefits to be realised. Yet, research aiming to move beyond identifying patterns, to identifying key ecological processes and farm management practices impacting biodiversity in the agricultural landscape, is often hampered by poor study design and/or difficulties obtaining relevant, comparable and detailed management information from study farms (Hole et al. 2005).

Here, we address these design issues using an established model system of environmental performance and farm practice monitoring across organic and integrated management systems in New Zealand’s kiwifruit (Actinidia spp.) production sector, to test whether biodiversity benefits associated with organic systems can indeed be ascribed to specific land management practices and habitat composition on kiwifruit orchards as opposed to the system as a whole. Recent research has shown that, relative to kiwifruit orchards under integrated management, organic orchards can provide enhanced environmental outcomes, with the latter supporting higher plant and invertebrate biodiversity and soil quality (Moller et al. 2007; Carey, Benge & Haynes 2009; Todd et al. 2011). Here, we focus on birds, which are often selected as an indicator of sustainable land management in other countries (Gregory et al. 2005). We first tested whether organic orchards (with the Green or Hayward kiwifruit variety Actinidia deliciosa) do indeed support higher densities of common native (n = 7) and introduced (n = 13) bird species than orchards managed using an integrated management system either with the same kiwifruit variety or a different variety (Gold or Hort 16A Actinidia chinensis). We then investigated whether any observed differences among the management systems can be accounted for by differences in habitat composition and/or pesticide use, or whether there are other elements of the organic management system having beneficial influences on this element of biodiversity.

Destruction of non-crop habitats and intensive use of agrochemicals are considered by many as the two aspects of intensive agricultural management that pose the most significant threat to biodiversity (Campbell et al. 1997; Benton et al. 2002; Morris et al. 2005; Gomiero, Pimental & Paoletti 2011). We predict that woody vegetation cover will be a positive predictor of bird density, through the provision of shelter, nesting sites and food resources (as has been shown for other systems; Haslem & Bennett 2008; Fischer et al. 2010; Hanspach et al. 2011), while pesticide use will be a negative predictor of bird density because of its impacts on non-target invertebrate food resources (Campbell et al. 1997; Benton et al. 2002; Morris et al. 2005). We also predict that native bird species, which tend to have small and fragmented populations in farmland areas, will be more sensitive to land-use management practices than introduced species that are generally widespread and abundant (MacLeod et al. 2009, 2012). Finally, we predict that insectivores, especially native ones, will be more abundant on orchards that are managed under organic systems (Genghini, Gellini & Gustin 2006) or use less toxic pesticides (Benton et al. 2002). If all observed management system effects are accounted for by just habitat composition and/or pesticide use, this study will demonstrate that the application of specific management practices could potentially achieve the biodiversity benefits associated with organic systems not only in this model system but also in other agricultural systems, being particularly relevant to intensively managed orchards in the Europe and North America (Reganold et al. 2001; Genghini, Gellini & Gustin 2006).

Materials and methods

Study Sites

Bird densities were estimated for 36 kiwifruit orchards in New Zealand. The orchards (1·39–24·9 ha) were located within 12 clusters, with all except one cluster on the North Island (MacLeod et al. 2012). Each cluster had three orchards: one integrated management ‘Green’, one integrated management ‘Gold’ and one organically managed Green ‘Organic’. All properties within a given cluster were no more than 3 km apart, but most were within 1 km of each other. Within each cluster, the three properties were matched, as closely as possible, according to location and topography (Table 1) as well as climate and soil type (Carey, Benge & Haynes 2009). However, to avoid the risk of excluding the very differences that may be responsible for creating observed variability in biodiversity among the different management systems (Unwin et al. 1995), orchards were not matched according to any other land-use characteristics or management practices (Table 1).

Table 1.   A summary of location, topographic, management and land-use data for kiwifruit orchard properties managed under integrated systems (Gold and Green) and organic systems (Organic); n = 12 per system
VariableGoldGreenOrganic
MeanMinMaxMeanMinMaxMeanMinMax
Latitude (ºS)−37·8−41·1−35·2−37·8−41·1−35·2−37·8−41·1−35·2
Longitude (ºE)175·7173176·5175·7173176·5175·7173176·5
Elevation (m)89521378·4517577·55185
Orchard age (years)25·7113428·2195428·22435
Property size (ha)7·91·423·26·4311·69·5424·9
Kiwifruit canopy (ha)3·91·17·33·62·26·13·81·910·1
Green canopy (ha)1·705·13·62·26·13·81·910·1
Gold canopy (ha)2·20·76000·5000
Kiwifruit canopy (%)55·730·483·158·644·284·849·69·582·2
Other fruit trees (%)6·6037·72·20122·9026·9
Woody vegetation (%)5·8045·32·8019·614·6063·4
Shelterbelt density (m ha−1)25311050724114235621096359

The orchards studied were established 11–54 years ago, with organic orchards fully certified for 10–24 years (by Bio-Gro New Zealand, which complies with the IFOAM certification; Table 1). Gold orchards were established in the 1990s and approximately half of those also grow the Green kiwifruit variety. Some properties also include orchards of other species (Citrus spp., avocado Persea americana and tamarillo Solanum betaceum), with a mean of 54% kiwifruit cover (range: 9–84%; Table 1). Both crop and non-crop habitats were mapped in ArcMap software (ESRI, Redlands, California, USA), using GPS locations recorded during field surveys. Two types of non-crop woody vegetation were present on the study orchards (Table 1): (i) shelterbelts, which consisted of a tree line (1–2 m wide; primarily Japanese cedar Cryptomeria japonica; she oak Casuarina spp. and willow Salix spp.; Moller et al. 2007), used to subdivide the kiwifruit canopy into small blocks and (ii) ‘woody vegetation’ patches of indigenous forest or exotic plantations (the latter were only found on two orchards). ‘Woody vegetation’ cover was the proportion of the property land area covered by forest or plantations (excluding shelterbelts). The density of shelterbelts (m ha−1) was the total shelterbelt length divided by the property area excluding the woody vegetation cover. Note that the amount of woody vegetation cover, which was inversely related to shelterbelt density (Rs = −0·37, P = 0·024), also varied among the different management systems (Table 1).

Annual records of the number and type of pesticide applications to each kiwifruit block within each orchard for three growing seasons (2007–2008, 2008–2009 and 2009–2010) were collated from the ZESPRI® Spray Diary database (see Appendix S1, Supporting information for list of sprays considered). Pesticides were ranked from least to most toxic (scale: 1–4; Appendix S1, Supporting information) based on the risk those chemicals pose to invertebrate taxa; non-hazardous chemicals were ranked as zero. New Zealand’s system for classifying hazardous substances (ERMA 2008), which is based on Organisation for Economic Co-operation Development test guidelines, assesses the oral and contact toxicity of substances to terrestrial invertebrates (using median lethal concentration, LC50; milligrams of substance in diet per invertebrate; very ecotoxic: <2 μg per invertebrate; ecotoxic: 2 ≤ LD50 < 11 μg per invertebrate; harmful: 11 ≤ LD50 < 25 μg per invertebrate; non-hazardous: >25 μg per invertebrate). For each property, the toxicity score was the sum of the individual spray applications per year, weighted according to their toxicity rankings and the proportion of the property area sprayed, and averaged across the 3-year period. Invertebrate toxicity scores were also indicative of risk to other taxa, as they were highly correlated with matched data for vertebrate (Spearman’s rank correlation coefficient: Rs = 0·97, P < 0·001) and soil taxa (Rs = 0·87, P < 0·001; Appendix S1, Supporting information; Jayson Benge, unpublished data).

Other orchard practices likely to differ between management systems, but not considered in our analysis, included fertiliser composition and application rates (Carey, Benge & Haynes 2009), vine pruning and ring barking (which alter the canopy cover and structure to enhance fruit quality and yields; Miller et al. 2001), grass understorey mowing and shelterbelt pruning regimes (Moller et al. 2007).

Bird Surveys

Distance sampling surveys were undertaken along transects within orchard production areas in the breeding season (November–January). Each orchard was surveyed three times over a 6-year period (2004–2009; MacLeod et al. 2012). A complete survey of the production areas was carried out for most study sites but, where that was not feasible, the kiwifruit blocks of the same classification as the orchard (i.e. Green, Gold or Organic) were prioritised over other kiwifruit blocks and orchard types (e.g. avocado, citrus). Transects (which started at a random point within 50 m of a property’s boundary corners) were 50 m apart and usually ran parallel to the vines. All transects on a single property were surveyed between 09·00 and 16·00 h (avoiding the peak calling periods at dawn and dusk to minimise bias associated with time-of-day effects; Dawson & Bull 1975). Heavy rain and strong wind conditions were avoided. Observers were rotated between properties and management systems to control for potential bias, with properties within each cluster usually surveyed concurrently. For each independent sighting of an individual bird or group of birds, the observer recorded the first detection cue (seen or heard), their number and location (MacLeod et al. 2012). The perpendicular distance of the bird from the transect line was calculated using exact distance and angle information in 2004/2005 and distance bands (metres, 0–5; 6–15; 16–25; 26–50; 51–00; >100) in subsequent surveys. No distance measures were recorded when the observer was uncertain about the location of the bird(s), and individuals that were only observed flying overhead (and not associated with any specific habitat feature at the study site) were excluded from the analyses. Average wind speed (km h−1) was recorded using a Kestral 4000 portable weather meter (Nielsen-Kellerman, Boothway, PA, USA).

Estimating Bird Densities

We used distance sampling software to estimate bird densities (Buckland et al. 2001; Distance Version 6.0 Beta 1; Thomas et al. 2006) by modelling decline in detectability of individuals or clusters of individuals with distance from the transect line. A global detection function was fitted to bird observation data from all surveys, including covariates of detectability where appropriate (wind speed, whether a bird was first seen or heard, observer identity or management system; see MacLeod et al. 2012). To reduce the risk of generating unrealistically large flock sizes and high variance density estimates, we (i) log-transformed flock sizes prior to fitting a regression model for observed flock size and distance from the transect; and (ii) only estimated flock size using regression models that met a significance P-value threshold of >0·15, otherwise the mean was applied (Buckland et al. 2001). Density estimates were then extracted for individual farms from the best-fitting model, using Distance’s post-stratification features.

Total densities of the following subsets of species were calculated: (i) all introduced species (house sparrow Passer domesticus, chaffinch Fringilla coleobs, goldfinch Carduelis carduelis, greenfinch C. chloris, redpoll C. flammea, yellowhammer Emberiza citrinella, skylark Alauda arvensis, myna Acridotheres tristis, starling Sturnus vulgaris, blackbird Turdus merula, song thrush T. philomelos, pheasant Phasianus colchicus, California quail Callipepla californica); (ii) introduced insectivores (blackbird, song thrush, myna, starling, chaffinch); (iii) introduced granivores (house sparrow, goldfinch, greenfinch, redpoll, yellowhammer, skylark); (iv) all native species (fantail Rhipidura fuliginosa, grey warbler Gerygone igata, welcome swallow Hirundo tahitica, silvereye Zosterops lateralis, tui Prosthemadera novaeseelandiae, pukeko Porphyrio porphyrio, kingfisher Halcyon sancta); (v) native insectivores (fantail, grey warbler, welcome swallow); (vi) native nectar-feeders (silvereye, tui). (Note: while other species were detected on the orchards, they were excluded from this analysis because there were too few sightings to estimate density.)

Data Analysis

We used linear mixed models (Bates & Maechler 2010) to test the relative importance of management system, toxicity of pesticide applications, shelterbelt density and woody vegetation cover on the property, as predictors of bird densities (Tables 2 & 3). Separate models were built for each subset of species, with the total density of birds per hectare for each property as the response variable, cluster identity as a random effect and a normal error distribution specified. The same set of 14 candidate models was fitted for each subset of species (see Results), using maximum likelihood and Akaike’s Information Criteria corrected for small sample sizes (AICc) to identify which model or set of models best described the data (Burnham & Anderson 2001). Where significant management system effects were detected, pair-wise comparisons were conducted using Tukey tests (Hothorn, Bretz & Westfall 2008). For models where organic management system effects were detected, we tested for a significant correlation between the predicted residuals and the time since conversion to organics. The partial effects of each predictor in the best-fit models were plotted using Markov chain Monte Carlo credible intervals following Baayen (2007). The software package R was used for all statistical analyses (R Development Core Team. 2011).

Table 2.   Candidate models fitted to introduced bird density data for all species and two subsets of species (insectivores and granivores). Best-fit models identified by the model selection (highlighted in bold, based on a ΔAICc threshold value of 2). Predictor variables were kiwifruit ‘management’ system (three-level factor: Gold, Green and Organic), ‘pesticide’ toxicity levels, the density of ‘shelterbelts’ and the amount of (non-crop) ‘woody’ vegetation cover (excluding shelterbelts)
ModelHypothesiskAll introduced speciesIntroduced insectivoresIntroduced granivores
AICcΔAICcWeightAICcΔAICcWeightAICcΔAICcWeight
m1Null3229·900·257210·300·337198·81·30·154
m2Management5232·32·40·078214·94·60·034200·530·066
m3Woody4232·42·50·074212·72·40·1012013·50·051
m4Shelterbelts42300·10·245211·81·50·159197·500·294
m5Pesticide4232·22·30·0812121·70·144201·33·80·044
m6Pesticide + Woody5234·950·021214·54·20·041203·76·20·013
m7Pesticide + Shelterbelts5232·52·60·07213·43·10·071200·22·70·076
m8Pesticide + Management6233·940·035213·83·50·058202·85·30·021
m9Woody + Management6235·15·20·019217·87·50·008203·45·90·015
m10Shelterbelts + Management6232·52·60·07216·66·30·014198·71·20·161
m11Pesticide + Woody + Management7236·970·008216·96·60·012205·98·40·004
m12Shelterbelts + Woody + Management72355·10·02219·79·40·003200·93·40·054
m13Pesticide + Shelterbelts + Management7235·35·40·017216·76·40·014201·74·20·036
m14Pesticide + Shelterbelts + Woody + Management82388·10·0042209·70·003204·26·70·010
Table 3.   Candidate models fitted to native bird density data for all species and two subsets of species (insectivores and nectar-feeders). Best-fit models identified by the model selection (highlighted in bold, based on a ΔAICc threshold value of 2). Predictor variables were kiwifruit ‘management’ system (three-level factor: Gold, Green and Organic), ‘pesticide’ toxicity levels, the density of ‘shelterbelts’ and the amount of (non-crop) ‘woody’ vegetation cover (excluding shelterbelts)
ModelHypothesiskAll native speciesNative insectivoresNative nectar-feeders
AICcΔAICcWeightAICcΔAICcWeightAICcΔAICcWeight
m1Null3110·25·70·02135·53·60·0447620·082
m2Management5107·63·10·07531·900·26877·83·80·034
m3Woody4107·83·30·06837·45·50·0177400·224
m4Shelterbelts4111·87·30·00937·55·60·01678·54·50·024
m5Pesticide4105·71·20·195320·10·25574·70·70·158
m6Pesticide + Woody5104·500·35533·920·0997400·224
m7Pesticide + Shelterbelts5107·93·40·06535·13·20·05477·43·40·041
m8Pesticide + Management6108·64·10·04634·62·70·0777·33·30·043
m9Woody + Management6108·43·90·0534·72·80·06678·14·10·029
m10Shelterbelts + Management6110·560·01834·82·90·06380·56·50·009
m11Pesticide + Woody + Management71083·50·06237·65·70·01675·71·70·096
m12Shelterbelts + Woody + Management7111·46·90·01137·85·90·01480·46·40·009
m13Pesticide + Shelterbelts + Management71116·50·01437·75·80·01580·36·30·01
m14Pesticide + Shelterbelts + Woody + Management8111·26·70·01240·990·0037950·018

Results

For introduced species, there was weak support for the models that included shelterbelt density for all three composite measures explored and pesticide toxicity for introduced insectivores (Table 2). Introduced bird densities increased with shelterbelt density, and introduced insectivore densities declined with increasing pesticide toxicity. However, little confidence can be placed in these relationships as the null model (m1) was included in the subset of best-fit models for all three composite measures investigated.

In contrast, for native bird species density, the model including management system alone (m2) provided a better fit to two of the three composite measures explored: all native species and the subset of native insectivores (Table 3). In both cases, lower bird densities were detected on Green integrated management orchards relative to Green Organic ones, with similar (but non-significant for all native species; P < 0·1) trends of lower density also observed for Gold orchards (Fig. 1). There was no evidence of a ‘time since conversion’ effect for either of these native species groups on green organic orchards (Rs = 0·5, P >0·05).

Figure 1.

 Mean (±SE) bird densities in relation to three different management systems (Gold, Green and Organic) in kiwifruit orchards for subsets of introduced and native species. The letters ‘a’ and ‘b’ indicate pairs of management systems that were significantly different from each other.

Relative to the pesticide toxicity and woody vegetation cover variables, however, management system was a poor predictor of all native bird species density and was excluded from all best-fit models (determined by AICc values and Akaike weights for the full set of candidate models; Table 3). Parameter estimates (from the model with the greatest Akaike weights) show that densities were negatively associated with pesticide toxicity for all native species together (Fig. 2). The same pattern was observed for the subsets of native insectivores and nectar-feeders (Table 3; Fig. 2), although the subsets of best-fit models for these two density measures (based on Akaike weights) did also include management system (m2 and m11, respectively; Table 3). Furthermore, all native bird species density and the subset of nectar-feeders were also positively associated with woody vegetation cover (Table 3; Fig. 2).

Figure 2.

 Partial effects of each predictor (solid lines) in the best-fit models for native bird densities (Table 3), with Markov chain Monte Carlo confidence intervals (dashed lines), showing the relationship between densities and pesticide toxicity levels and the amount of vegetation cover. Points show observed data from each orchard management system (triangles = Gold; squares = Green; circles = Organic).

Discussion

Enhanced Biodiversity in Organic Systems

Our study provides evidence that the benefits of organic farming practices extend beyond arable and mixed farming systems (Hole et al. 2005) to horticultural systems (Reganold et al. 2001; Genghini, Gellini & Gustin 2006). Higher densities of the ‘all native species’ composite measure were detected on Green kiwifruit orchards managed under organic systems, relative to integrated ones (Table 3; Fig. 1), lending further support to the hypothesis that organic farming systems can sustain enhanced biodiversity compared to more intensive ones (Bengtsson, Ahnström & Weibell 2005; Fuller et al. 2005; Hole et al. 2005; Genghini, Gellini & Gustin 2006). Although we found no evidence of a time-lag in the response of birds to any benefits generated by a switch from conventional to organic farming, it is important to note that we did not have the statistical power to detect short-term time-lags as all the organic conversions occurred at least 10 years prior to our study. Given that our results align with previous demonstrations of improved soil quality and enhanced populations of other taxa for kiwifruit orchards managed under organic farming systems (Moller et al. 2007; Carey, Benge & Haynes 2009; Todd et al. 2011), native bird densities may be a good indicator of sustainable land management practices within this agroecosystem.

Processes for Enhancing Biodiversity Beyond Organics

Three broad processes strongly associated with organic farming are expected to be generally beneficial to farmland biodiversity (Hole et al. 2005): (i) Prohibition or reduction in the use of chemical pesticides and inorganic fertilisers is likely to have a positive impact through the removal of both direct and indirect negative effects on plants, invertebrates and vertebrates; (ii) Sympathetic management of non-crop habitats and field margins can enhance diversity and abundance of plants, invertebrates, birds and mammals; and (iii) Preservation of mixed farming through the provision of greater habitat heterogeneity at a variety of temporal and spatial scales within the landscape. While we did not address the third prediction, our analyses support the first two: native bird densities on kiwifruit orchards were negatively associated with the toxicity of pesticide applications and positively associated with the amount of woody vegetation cover (Table 3; Fig. 2). Furthermore, as predicted, densities of both native and introduced insectivorous species were negatively associated with pesticide toxicity, suggesting insectivorous species are indeed most susceptible to pesticide impacts. Native species, which are primarily aerial or canopy-feeders (McLean 1989), are presumably attracted to orchards that use less toxic pesticide applications because their kiwifruit canopies support enhanced populations of invertebrate populations (Todd et al. 2011). Similarly, as the woody vegetation cover considered in our analysis was primarily native bush or forest, this non-crop habitat probably provided native species with important resources such as nesting sites, shelter and food resources (e.g. nectar and invertebrates). However, controlled experimental manipulations are still required to confirm a causal link between reduced pesticide toxicity and/or increased woody vegetation cover and enhanced populations of native bird species and other taxa on orchards.

Addressing the primary aim of this study, our results show that a complete conversion to an organic system may not be required to improve biodiversity in kiwifruit orchards; the greater woody vegetation cover and reduced pesticide toxicity associated with organic systems alone accounted for all the bird biodiversity benefits observed under this management system. Furthermore, that these practices were stronger predictors of native bird densities than management system alone (Table 3) shows that reducing the toxicity of pesticide applications and increasing the amount of woody vegetation should also benefit biodiversity on orchards in integrated management systems. This finding has important implications for biodiversity conservation at the wider agricultural landscape scale in New Zealand, as integrated management orchards make up a large proportion of the total land area managed for kiwifruit production compared to organic systems (Gomiero, Pimental & Paoletti 2011).

Study Design Considerations

Relative to earlier studies comparing organic and conventional systems (Hole et al. 2005), our study design had two key strengths. First, detailed and comparable information about land-use and management practices were recorded in parallel with the bird survey data for three farm management systems, allowing us to move beyond measuring differences in biodiversity among different management systems to identifying specific management practices that benefit biodiversity irrespective of the overarching management system. Secondly, we avoided three methodological issues that are often overlooked but may result in erroneous conclusions being drawn by: (i) controlling for any confounding effects of landscape composition orchards by deliberately selecting 12 clusters of three orchards (each managed under a different system) and including cluster identity as a random effect in our regression analyses; (ii) avoiding the risk of excluding the very differences that create variability in biodiversity in the first place by not attempting to match orchards within the different management systems according to specific practices or land-use characteristics; and (iii) accounting for stochastic variability by collecting bird and pesticide data over three breeding seasons during a 6-year period. That we were able to demonstrate the likely transferability of management practices that benefit biodiversity between management systems indicates that this approach (and perhaps even the specific results) would be a powerful way in which changing management in other agricultural systems (particularly intensively managed orchards) could likewise benefit biodiversity.

However, our analyses do not account for the possibility that the organic orchards in our study were managed by people already predisposed to environmentally friendly farming practices anyway (Lotter 2003), or the land had previously been managed less intensively and so was easier to convert successfully to organic (Hole et al. 2005). Indeed, this is one possible reason why no ‘time since conversion’ effects were detected in our study. A better understanding of how orchard managers relate to their agricultural land is needed to facilitate change in management practices to reduce adverse impacts on biodiversity and the environment (Hunt 2010).

Management Implications

Intensification is an ongoing and accelerating threat for biodiversity in New Zealand’s agricultural landscape (MacLeod & Moller 2006; Moller et al. 2008). However, the nature and extent of its impact is unknown because there is little biodiversity and environmental monitoring in farmland areas. Our study highlights that native bird species are likely to be suitable indicators for monitoring the impact of changes in land management practices within kiwifruit orchards. Although only relatively small differences between bird densities were associated with the different management systems in our data, the alignment of these differences with previously observed benefits of organics in terms of higher plant and invertebrate biodiversity and soil quality means that they are likely to reflect large differences at the total biodiversity scale. In addition, the lack of association, or weak relationships only, between introduced bird densities and management system, pesticide toxicity, shelterbelt density and woody vegetation composition suggests that these species’ populations must be limited by other factors in this agroecosystem (e.g. winter food resources, exposure to extreme weather events or natural enemy regulation; MacLeod & Till 2007; MacLeod et al. 2009, 2010). Thus, the management actions for improving native species identified here appear unlikely to increase the abundance of non-native species, including pest birds linked to both agricultural (MacLeod et al. 2008, 2011) and disease impacts (Sturrock &  Tompkins 2008; Tompkins et al. 2011).

On a global scale, the agricultural industry is coming under increasing pressure from local and international consumers, communities and governments to produce sustainable food products (Golden et al. 2010). Maintaining food safety standards and minimising adverse impacts on the environment are two key concerns that food producers need to address to gain and maintain access to high-value food markets in particular (Campbell et al. 2011). While the uptake of organic farming practices has increased at the global scale in recent years, this sector still targets and accesses a relatively small and marginal niche market. Development of other ‘green’ audit systems that embrace some of the principles of organic farming (e.g. food safety and environmental sustainability) but are less stringent (e.g. ‘integrated systems’) has allowed retailers to supply larger volumes of food products into the ‘green’ market (Campbell et al. 2011).

While there is growing support to show that organic systems can enhance biodiversity on farms, it appears likely that integrated management systems can also contribute to biodiversity conservation. The regulations that organic and integrated farms need to meet to secure market access, however, do not necessarily have a strong ecological basis, with biodiversity benefits often being assumed rather demonstrated (Hole et al. 2005; Walker 2005). Our study demonstrates that reducing the frequency and toxicity of pesticide applications within one horticultural sector not only addressed consumers’ concerns about adverse health impacts of spray residue on fruit in the international market (Walker 2005), but probably also alleviated adverse impacts on biodiversity. That the biodiversity benefits associated with organics can be accounted for by such relatively straightforward management changes offers a way forward for other agricultural sectors to make similar advances.

Acknowledgements

Funding was provided by the Ministry for Science and Innovation (Contract Number AGRB0301) and the Certified Organic Kiwifruit Producers Association, ZESPRI Innovation Company, University of Otago and Landcare Research. Data were collected by C. Bragg, D. Clarke, F. Weller, F. Buzzi, G. Coleman, J. Lurling, J. Lach, L. Mckinnel, L. Penniket, N. Percey, S. Rate, S. McGuire, T. Maegli and T. Dearlove. We thank F. Weller for completing the distance analysis and H. Moller, D. Tompkins, D. Raffaelli, the editor and an anonymous reviewer for their constructive reviews.

Ancillary