Proximity to rainforest enhances pollination and fruit set in orchards

Summary

• 1
  Tropical rainforests are potential reservoirs of insects that could enhance crop pollination, but only a few instances of the provision of such services by tropical rainforest insects have been reported. Our field study aimed to determine the relative importance of such insects to the pollination of macadamia Macadamia integrifolia and longan Dimocarpus longan crops on the Atherton Tableland, north Queensland, Australia.
• 2
  We quantified initial fruit set, a measure of pollination success, in treatments designed to assess the relative importance of the possible modes of pollination. The treatments were applied in orchards that varied in distance from rainforest, in order to compare the effects of the contrasting pools of available pollen vectors. We also recorded the insect species present and estimated the number of visits each made to flowers in crops near and far from rainforest.
• 3
  For both crops there was an interaction between pollination treatment and distance from rainforest. Maximum fruit set was only achieved when pollen vectors had access to flowers and orchards were close to rainforest. Exclusion of pollinators near rainforest reduced initial fruit set to a greater extent than exclusion of pollinators far from rainforest.
• 4
  We confirmed that pollen transfer in macadamia is by autogamous self-pollination and by pollen vectors, but our design did not distinguish among pollen vectors. The only abundant insects in macadamia orchards were honeybees Apis mellifera. There were more honeybee visits to macadamia flowers in orchards near rainforest than far from rainforest, but we detected no relationship between honeybee visits and initial macadamia fruit set in our sample of observations on a per raceme basis. More detailed studies are needed to identify the pollen vector responsible for enhanced pollination of macadamia near rainforest.
• 5
  We established for the first time that pollen transfer in longan is by a combination of autogamous self-pollination, wind and bees. Longan flowers were visited by stingless bees and honeybees but only stingless bees had a positive relationship with initial longan fruit set and higher visitation rates near rainforest than far from rainforest. This suggests that enhanced pollination in longan near rainforest resulted primarily from a more abundant supply of stingless bees from the rainforest.
• 6
  Synthesis and applications. By demonstrating that tropical rainforest can act as a reservoir of pollen vectors that benefit crops, our study highlights the existence of a largely unrecognized resource available to agriculture. At the same time our results make a significant contribution to the growing database of studies that underscore the importance of tropical rainforest conservation. Policy and management aimed at sustainable use of this resource would satisfy the goals both of agriculturalists, to improve crop yields, and conservationists, to conserve tropical rainforest.

Introduction

Many tropical rainforest plants are insect pollinated (Bawa 1990; Sands & House 1990; Momose et al. 1998; Liow, Sodhi & Elmqvest 2001) but the role of tropical rainforest as a reservoir of insect pollinators that could be used in agricultural crops has rarely been established. Exceptions are the demonstration that rainforest fragments are sources of pollinating bees for coffee Coffea spp. flowers in Indonesia (Klein, Steffan-Dewenter & Tscharntke 2003a,b,c) and Costa Rica (Ricketts 2004) and of pollinating beetles for atemoya Annona squamosa L. × Annona cherimola Mill. hybrids flowers in north Queensland, Australia (Blanche & Cunningham 2005). While pollination services stemming from tropical rainforest have the potential to provide economic benefits to growers (Klein et al. 2002), they also add to the reasons for conserving and restoring such forests (Ricketts et al. 2004).

To help expand knowledge of the relative importance of tropical rainforest in providing pollination services for crops, our research aimed to assess the identity, source and contribution to pollination of the pollen vectors of two subtropical tree crops growing in a diverse agricultural landscape, with scattered patches of rainforest and savanna remnants, in north Queensland, Australia. These crops were macadamia Macadamia integrifolia Maiden and Betche, a native of the subtropical rainforests of south-eastern Australia, and longan Dimocarpus longan Lour., a native of the mountain chain from Myanmar (Burma) to southern China and possibly also south-western India and Sri Lanka (Wong 2000).

Macadamia flowers in Hawaii are reported to be mainly pollinated by honeybees Apis mellifera L. (Urata 1954; Shigeura, Lee & Silva 1970) but in south-eastern Australia stingless bees, native Trigona spp., are also thought to be major pollinators (Heard 1994). Small bees, like Trigona spp., are potentially more effective pollinators of macadamia than honey bees (Heard 1994) and macadamia flowers are not thought to be wind pollinated (Heard 1993). In other areas of Australia, outside the natural range of Macadamia integrifolia, the lycid beetle Metriorrhynchus rhipidius (MacLeay), the flower wasp Campsomeris tasmaniensis Saussure and the halictid bee Homalictus sp. are capable of contributing to macadamia pollination (Vithanage & Ironside 1986). Longan flowers are thought to be pollinated by Apis and Trigona bee spp. and possibly also by hoverflies (Syrphidae) (Boonithee et al. 1991) but the relationship between bee visitation and longan pollination has not been verified.

Materials and methods

sites and rationale

The study was carried out on the Atherton Tableland in north Queensland, Australia. The tropical rainforests of the Wet Tropics World Heritage Area lie to the east but extensive clearing since the mid-1800s means that the study area is a mosaic of mainly agricultural land interspersed with remnants of rainforest and eucalypt savanna vegetation (Frawley 1983; Collins 1994).

For each crop, we selected orchards near and far from rainforest to establish a contrast in the pools of available flower visitors. Our design was based on the hypothesis that the visitation of crops by insects associated with rainforest vegetation would be higher near the rainforest than farther away. We used three macadamia and two longan orchards near rainforest vegetation (≤ 0·5 km away) and two macadamia and four longan orchards far from rainforest vegetation (4–30 km away). The number of replicate sites was constrained by the number of suitable orchards available.

treatments

In a randomized complete-block split-plot design (Steel & Torrie 1980), pollinator exclusions were applied in a balanced way to each tree and trees were in orchards nested within distance from rainforest. At each macadamia orchard we applied the same three treatments to eight randomly selected trees. On each tree, three clusters, each of one to 10 immature racemes (inflorescences in which stalked flowers are spaced along an axis), were tagged and assigned one of the following three treatments: (i) a control treatment, in which the racemes were left open, allowing all modes of pollination; (ii) a bagged treatment, in which the racemes were completely enclosed in a fine voile bag supported by a plastic frame, allowing only autogamous self-pollination; (iii) a frame-only treatment, in which the racemes were enclosed only by a plastic frame, allowing all modes of pollination but taking possible effects of the frame into account.

At each longan orchard we applied the same three treatments to single panicles (compound racemes, i.e. branched clusters of flowers in which the branches are racemes) with unopened flower buds on four trees. Because the pollination of longan flowers has not been studied in detail, we also applied two additional treatments, one to separate the effects of pollination by small insects from that of large insects, and another to assess pollination by wind. The additional longan treatments comprised: (iv) a coarse mesh treatment, in which the panicle was completely enclosed in a 5 × 5-mm mesh bag supported by a plastic frame, excluding large insects like honeybees but allowing pollination by small insects like stingless bees and by wind and autogamous self-pollination; (v) a fine mesh treatment, in which the panicle was completely enclosed in a 1 × 1-mm mesh bag supported by a plastic frame, excluding all insects large enough to carry pollen but allowing pollination by wind and autogamous self-pollination.

Treatments were left in place until all flowers had senesced and immature fruits were observed (4–6 weeks). The racemes and panicles were then removed from the trees, placed in individual plastic bags, and taken to the laboratory where the number of attached developing fruits was counted. Macadamia yields were standardized as number of developing fruits per raceme, because the number of flowers available for fruit production increases with the number of racemes. Initial fruit set is a reliable measure of fertilization in macadamia (Urata 1954). We standardized our yield estimates for longan by measuring the length of stem available for flower production in each panicle and expressing the results as number of fruit initiated per centimetre of panicle.

insect observations

During peak flowering time, 1 day was spent at each orchard monitoring flower-visiting insects on four trees. The four macadamia trees selected for observation comprised an evenly spaced subset of the eight experimental trees. The longan trees used for observation were the same as those used in the pollination exclusion experiment. Days chosen had similar weather conditions (i.e. fine and sunny with little wind) and were within 10 days of each other for each crop. An observer spent 10 min tree⁻¹ h⁻¹ between 08:00 and 17:00 recording the number of insect visits, and the identity of each insect visitor, to mature, open flowers on a known number of macadamia racemes (6–20) or a single longan panicle, respectively. Representative specimens of each insect species visiting flowers were collected and later identified in the laboratory. These specimens were then deposited with the Australian National Insect Collection, Canberra, Australia.

analyses

The effect of pollinator exclusion on initial fruit set was tested using restricted estimates maximum likelihood variance analysis (REML) (Genstat 2002), with treatment as a fixed effect and tree nested within site (orchard) as a random effect. To establish post-hoc the source of differences, we compared means (predicted from the REML model) with least significant difference (LSD) at α= 0·05.

A difference in initial fruit set in orchards near rainforest compared with orchards far from rainforest, if the result of a difference in pollination rates, would be expressed as a significant interaction between distance and pollinator treatment. It would be difficult to interpret such an interaction if there were more than two pollinator treatments, so we reduced the data sets for these analyses to only one pollinator access treatment and one pollinator exclusion treatment. These treatments were selected after considering the results of the analysis of treatment effects. We tested for a distance by pollination interaction using a REML analysis, with distance and pollination treatment as fixed effects (with a two-way interaction) and tree nested within orchard, nested within distance, as a random effect. REML analyses were favoured because they perform better than analysis of variance when the design is unbalanced, as was the case here with uneven numbers of orchards in the near and far categories. We found that the distribution of residuals from the models was approximately normal if data were transformed as log₁₀(x + 1).

The total visits of dominant insect species, per raceme or panicle, during the 9-h observation period, were summed for each tree. The mean of the four trees at each orchard was used as the measure of insect visitation at that orchard. Nested one-way analysis of variance was used to determine the effect of distance from tropical rainforest on insect visitation. Linear regression analysis was used to explore the relationships between insect visitation and initial fruit set in the open treatments. All data were transformed as log₁₀(x + 1).

Results

macadamia

Mean initial fruit set per raceme in macadamia exceeded 12 when racemes were open to all pollinators but was less than two when pollinators were excluded (Fig. 1a), leading to a significant effect of pollinator exclusion (d.f. = 2, Wald = 95·8, P < 0·001). Initial fruit set in the frame-only control was not significantly different from that in the open treatment.

Figure 1. Effect of pollination treatments on fruit set for (a) macadamia and (b) longan. Data presented are untransformed means ± SE but data were transformed as log₁₀(x + 1) for REML analyses. Means with different letters above the bar were significantly different from one another (α = 0·05) using the LSD test on predicted treatment means from REML analyses.

When examining the distance by pollination treatment interaction we chose to be conservative and use the data from the frame-only control, rather than from the open treatment, for comparison with the bagged treatment. There was a significant interaction between pollination treatment and distance of the orchard from rainforest (Table 1). The effect of pollinator exclusion in orchards near rainforest was greater than in orchards far from rainforest (Fig. 2a).

Table 1.  REML analysis of variance of fruit set for macadamia (fruit per raceme, n= 77) and longan (fruit per cm of panicle, n= 48)

Crop	Effect	d.f.	Wald statistic	P
Macadamia	Distance	1	0·01	0·925
	Treatment	1	60·18	< 0·001
	Treatment × distance	1	4·67	0·031
Longan	Distance	1	2·51	0·113
	Treatment	1	22·85	< 0·001
	Treatment × distance	1	3·83	0·050

Figure 2. Interaction plots showing the effect of distance from rainforest and pollinator exclusion on fruit set for (a) macadamia and (b) longan. There was a significant interaction between distance and pollination treatment in both crops (Table 1). Data presented were back-transformed from the means and SE of the log₁₀(x + 1)-transformed data.

The only abundant flower visitor in macadamia orchards was the honeybee. Significantly more honey bees were seen visiting macadamia flowers near rainforest (mean ± SE visits per raceme, 11·4 ± 1·2) than far from rainforest (mean ± SE visits per raceme, 3·2 ± 0·4; F_1,3 = 62·3, P= 0·004) but there was no significant relationship between initial fruit set in the open treatment and observed honeybee visits on a per raceme basis (linear regression r² = 0·22, P= 0·43).

longan

Longan fruit set was significantly reduced by pollinator exclusions (d.f. = 4, Wald = 26·7, P < 0·001; Fig. 1b) but the frame did not significantly inhibit initial fruit set. Initial fruit set in the coarse mesh treatment was not significantly different from the frame-only treatment, but the fine mesh and bagged treatments were both significantly lower. Initial fruit set in the bagged treatment was significantly lower than that in the fine mesh treatment.

Observations revealed that honeybees were able to crawl through the holes in the coarse mesh bags meant to exclude them, thus we were unable to separate the contributions of small and large insects to pollen transfer. Initial fruit set in the bagged treatment, a measure of autogamous self-pollination, was 7% of the open treatment. The contribution of wind (fine mesh minus bagged treatment) to pollen transfer was 31% and that of insects (frame-only minus fine mesh treatment) 62%.

We chose to use the frame-only control and the fine mesh treatments to examine the distance by pollination treatment interaction because this comparison was conservative and would separate the input of pollinating insects from pollination by wind. There was a significant interaction between pollination treatment and distance of the orchard from rainforest (Table 1). There was a greater effect of pollinator exclusion in orchards near rainforest than in those far from rainforest (Fig. 2b).

Eight species of bee were observed visiting longan flowers but only honeybees and stingless bees were abundant. The number of honeybee visits to longan flowers at orchards near rainforest (mean ± SE visits per panicle, 177·4 ± 85·9) was not significantly different from the number observed at orchards far from rainforest (mean ± SE visits per panicle, 208·3 ± 77·8) but the number of stingless bee visits to longan flowers was significantly higher at orchards near rainforest (mean ± SE visits per panicle, 539·9 ± 514·4) than at orchards far from rainforest (mean ± SE visits per panicle, 3·0 ± 1·9; F_1,4 = 8·8, P= 0·042). There was also a significant positive relationship between initial fruit set in the open treatment and number of stingless bee visits (linear regression a= 0·05, b= 0·03, r² = 0·71, P= 0·035).

Discussion

Our results demonstrate that pollination rates in both macadamia and longan, as represented by initial fruit set, are enhanced substantially by proximity to tropical rainforest. Our study is the first to document that autogamous self-pollination, wind and insects can all contribute to longan pollination but that insects are the main pollen vectors in longan orchards. The positive relationship between the number of stingless bee visits to longan flowers and initial fruit set suggests that stingless bees are the major pollinating insects of longan flowers. The higher visitation of stingless bees to longan flowers near rainforest suggests that rainforest is a source of these bees.

Our findings are less clear-cut for macadamia. Even though we confirmed that macadamia flowers benefit from pollen transfer by vectors, and demonstrated that macadamia pollination by vectors was higher near rainforest, our design did not distinguish between insect and wind pollination. There were more honeybee visits to macadamia flowers in orchards near rainforest, which supports a honey bee-mediated advantage for orchards near rainforest, but there was no relationship between honeybee visitation and fruit set on a per raceme basis. This lack of agreement between the two analyses could be the result of the effect of pollen vectors that we did not observe (and the honeybee pattern is therefore weak) or (if honeybees are important) to a poor congruence between honeybee visits estimated at the site level and fruit set measured at the raceme level. The short period of the total flowering time that the racemes were observed could be responsible for either possibility.

The absence of stingless bees in macadamia orchards during our observations was at first surprising given their apparent importance in macadamia orchards in south-east Queensland (Heard 1994) and their occurrence during pre-study surveys. However, we later discovered that the orchards were sprayed with chemicals to control pests in the period between the pilot surveys and our observational study. Honeybees may have been more resistant than the stinglessbees to the pesticides used. This may explain why honeybees were the only pollinating insects remaining when we carried out our observations. We do not know whether stingless bee numbers recovered in time to pollinate the flowers in our study.

The amount of wind-blown pollen reaching receptive macadamia flower stigmas in our study may have been greater than anticipated. Wind may be a particularly important agent of cross-pollination on the Atherton Tableland, where strong south-easterly winds prevail from July to September, when macadamias are flowering. To our knowledge no-one has tried to separate the contribution of wind-blown macadamia pollen from pollen transfer by other agents, although this has been attempted for other crop species such as oilseed rape or canola Brassica napus L. (Hayter & Cresswell 2006). Urata (1954) suggested that the role of wind-blown pollen in macadamia pollination should not be discounted but most experiments carried out since then have not taken wind into account. Bags with mesh smaller than the diameter of macadamia pollen, used to exclude insects, also exclude wind-blown pollen, but the combined effects of excluding both have been attributed solely to the exclusion of insects (Vithanage & Ironside 1986).

Our observations did not allow for the possibility that nocturnal insects or vertebrates play a part in the pollination of these crops, but it has already been established that night-flying insects (and by default bats) have a negligible effect on the pollination of macadamia flowers (Vithanage & Ironside 1986; Heard 1993). The role of these pollinators has not been investigated for longan but as pollination of longan flowers is reported to be most effective between 08:00 and 14:00 (Wong 2000) it is unlikely that nocturnal pollinators contribute greatly to longan pollination.

Birds may make a substantial contribution to macadamia pollination (Heard & Exley 1994) but their role has yet to be quantified. In the macadamia orchards we surveyed, birds were only observed visiting flowers during the first hour of observations. The few birds observed targeted flowers at the top of the canopy several metres above the flowers used for our study. The hole size of the mesh used in the coarse mesh treatment applied to longan flowers was too small for vertebrate pollinators like birds to pass through. The fact that their exclusion by the coarse mesh treatment did not significantly reduce pollination rates, compared with the frame-only treatment, suggests that vertebrate pollinators did not contribute to longan pollination.

It seems clear that rainforest acts as a reservoir for stingless bees that enhance longan pollination, and it is likely that honeybees are also harboured by rainforest and that these can be important pollinators of macadamia. More detailed studies that separate the contributions of potential pollen vectors are required before the vectors responsible for the enhancement of pollination in macadamia orchards near rainforest can be determined. The importance of these pollinators might vary between years such that they are of even greater significance when other modes of pollen transfer are not adequate (Waser et al. 1996; Kremen, Williams & Thorpe 2002).

Our findings have important ramifications for crop management and rainforest conservation. Rainforest is commonly perceived as a source of pest insects that damage crops but the role of rainforest as a source of beneficial insects is rarely recognized (Blanche et al. 2002). Our study and others (Klein, Steffan-Dewenter & Tscharntke 2003a,b,c; Ricketts 2004; Blanche & Cunningham 2005) underscore the benefits of rainforests. This new perspective draws attention to the advantages to be gained from policy and management practices that allow rainforest vegetation to remain near crops, or be planted within crops, to improve visitation by beneficial insects. Such strategies have the potential to improve crop yields while conserving tropical rainforest.

Acknowledgements

We thank the macadamia and longan growers who allowed us to carry out our field experiments and observations in their orchards. We are also grateful for field and laboratory assistance provided by Robert Bauer and comments on the manuscript by Ingolf Steffan-Dewenter. The study was part of a CSIRO Ecosystems Services Project. It was funded jointly by the Myer Foundation and the Cooperative Research Centre for Tropical Rainforest Ecology and Management.