Available online at www.sciencedirect.

com

Agriculture, Ecosystems and Environment 127 (2008) 3136 www.elsevier.com/locate/agee

Role of native bees and natural habitats in eggplant (Solanum melongena) pollination in Kenya
Barbara Gemmill-Herren a,*, Alfred O. Ochieng b
b

Food and Agriculture Organization of the United Nations, Viale delle Terme di Caracalla, Rome 00153 Italy University of Nairobi, Department of Botany and International Centre of Insect Physiology and Ecology, P.O. Box 30772, Nairobi, Kenya Received 25 March 2007; received in revised form 31 January 2008; accepted 4 February 2008 Available online 5 May 2008

Po st -P rin tv
provisioning of pollination services in farming systems in Australia, California, Costa Rica, and Indonesia (Blanche et al., 2006; Greenleaf and Kremen, 2007; Klein et al., 2003; Kremen et al., 2002; Ricketts, 2004). In the Nguruman farming area of southwestern Kenya, horticultural crops including eggplant (S. melongena L.) are being grown intensively on plots recently cleared from riverine umbrella Acacia (Acacia tortilis (Forsk.) Hayne) forests. Eggplant bears owers with abundant pollen that can only be dispersed through small orices in the anthers. Eggplant owers are hermaphroditic, and are capable of some self-pollination (Free, 1993), but the process of expelling pollen from these owers onto female ower parts requires shaking either by wind or by the action of insect visitors. Honeybee (A. mellifera) visitation has been shown to signicantly increase fruit weight in eggplant (Levin, 1989). Some bee species can expel and efciently harvest pollen from owers with poricidal anthers by vibrating their bodies while in contact with the stamen, effecting a process termed buzz pollination (Buchmann, 1983). Honeybees

Abstract

The pollination requirements of eggplant (Solanum melongena) were investigated. One variety of eggplant exhibited a signicantly reduced seed set in absence of pollinators, and two varieties signicantly increased seed numbers when pollen deposition was enhanced. Two solitary bee species, Xylocopa caffra and Macronomia rupes, were identied as effective pollinators of the crop. The visitation rates of these pollinators to eggplant owers declined signicantly with distance from the wild habitat. The importance of wild plants as alternative forage source for pollinators was assessed by a survey of the owering plants in different habitats surrounding eggplant elds. While ruderal farm weeds provide much of these resources, the Acacia tortilis riverine forest experienced high visitation rates in one critical period of the dry season. The spatio-temporal foraging behaviour of eggplant pollinators highlights the role of the agricultural matrix in conserving ecosystem services. Interspersion of wild habitats with cultivated land promoted increased pollination services. # 2008 Food and Agriculture Organization of the United Nations. Published by Elsevier B.V. All rights reserved.
Keywords: Eggplant; Pollination; Ecosystem services; Wild habitat; Agroecosystems; Landscape ecology

si er

1. Introduction Because many horticultural crops are dependent on animal pollinators for optimal yields (Klein et al., 2007), concern has mounted about possible declines in pollinator abundance and diversity caused by anthropogenic disturbances, including habitat modication and fragmentation, and the consequent impacts on pollination services for agriculture. The domesticated honeybee, Apis mellifera L. has been utilized to provide managed pollination systems, but for many crops, honeybees are either not effective or are suboptimal pollinators (Westerkamp, 1991) and pollination services provided by wild biodiversity may be of key importance (Klein et al., 2007; Kremen et al., 2007). Recent ecosystem-level approaches in macademia, watermelon, tomato and coffee cropping systems have documented the role of wild bees and landscape conguration in the
* Corresponding author. Tel.: +39 06 57056838. E-mail address: Barbara.herren@fao.org (B. Gemmill-Herren).

0167-8809/$ see front matter # 2008 Food and Agriculture Organization of the United Nations. Published by Elsevier B.V. All rights reserved. doi:10.1016/j.agee.2008.02.002

on

32

B. Gemmill-Herren, A.O. Ochieng / Agriculture, Ecosystems and Environment 127 (2008) 3136

are not able to effectively buzz pollinate, so it is likely that wild insect visitors that are capable of handling owers in this way could be even more effective in accessing pollen from poricidal anthers and effectuating pollination (Kearns and Inouye, 1997). Moreover, eggplant owers produce no oral nectar and bees that pollinate eggplant may be expected to depend on other sources of nectar nearby (Free, 1993). To operationalise the concept of ecosystem services in a practical sense, it is important to link research ndings to the land-use decisions of the farming community. The effects of pollination by insect vectors on two commonly grown eggplant cultivars were studied, within a research framework that asked:  Do wild bees act as effective eggplant pollinators?  Does the level of pollination service vary over the farm/ forest mosaic landscape?  What types of habitats within this farm/forest mosaic provide critical alternative resources to eggplant crop pollinators?  Are there not only spatial, but also temporal differences in when such habitats increase or decrease in importance to crop pollinators?

subsequent studies, seed numbers are used as a metric for production values. Four treatments were applied to both eggplant varieties: (1) autonomous self-pollination (owers bagged to exclude both wind and pollinators during the owering period) (2) wind-pollination (caged to exclude any pollinators but allow wind-borne pollen) (3) open pollination (unrestricted access to pollinators on owers that were marked but not bagged or caged); (4) hand-pollination with pollen from other plants. A split plot design was used to test differences in pollination vectors due to plant cultivars. An area of eggplant crop within the Nguruman cultivation area, at 50 m distance from eld/forest edge, was divided into two blocks, one for each cultivar. Each block was divided into four plots of 20 m2. The four treatments were randomly assigned to plots within each block. Treatments were applied to three owers on each of ten plants per plot (n = 228). ANOVA (Proc GLM, SAS) was performed to compare the effects of each treatment on mean number of seeds set by owers of the two eggplant varieties. Pollinators were observed from 6:30 a.m. to 12:30 p.m., after which time the number of visiting insects dropped to near zero. Visitation rates of pollinators were observed by recording individual ower visitors and the number of owers visited to the two crop varieties in 1-m2 plots for standard 10 min observation periods, standardized by recording the number of owers per plot, throughout the day during the owering period of the crop and at varying distances from eld/forest edge. Environmental factors, including temperature, humidity, lux, cloud cover and windspeed were recorded for each observation period. Differences in visitation rates over time were determined using a non-parametric one-way ANOVA (proc npar1way; SAS) statistical test. The pollination performance of the most common pollinators was investigated by using a direct measure of pollinator reproductive success (following the denitions of Dafni et al., submitted for publication), during the peak eggplant growing season from March to June 1999. Young ower buds of the two common varieties (Black Beauty and Early Long Purple) were bagged a day before they opened to exclude pollinators. Since the receptivity of the stigma is thought to be greatest on the second day after opening (see Free, 1993) on this day the bags were removed and the owers allowed only a single visit by a pollinator, noting its species. The owers were bagged again to prevent subsequent visits. Fruits were harvested upon maturation and the seeds counted. A metric of pollinator performance, or effectiveness (PE) was calculated using Spears (1983) formula, which assesses the plant response to a single pollinator visit as a measure of its contribution to plant reproductive success. This technique assumes that owers with unrestrained visitation will average greater yields than a single visit during the life of the ower.

2. Materials and methods

The study was conducted in the horticultural area of Nguruman in southwestern Kenya, from October 1998 to July 1999; a second phase, to assess impacts on pollination by distance from wild habitat, was carried out from April to June 2003. The farms were located on an arid Maasai group ranch, where riverine A. tortilis forests occur at the base of an escarpment, in an agricultural area recently cleared from forest. Farmers cultivate small elds that are contiguous, forming a large cultivation area. From a small river descending off the escarpment, a system of irrigation provides water to grow vegetables year round. Farming is mixed but primarily eggplant is grown for an export market that has recently been developed, with eggplant crops covering 83% of the land area within the main agricultural area. Throughout the agricultural area, management practices within eggplant elds were very similar, because the exporting companies specify the timing and quantities of inputs such as planting material and pesticides. Pesticides were applied almost weekly during the growing season, primarily Lambda cyhalothrin and Chlorpyrifos, but no other agricultural chemicals such as fertilisers. Site characteristics were also homogenous, as elds were recently cleared from Acacia riverine forest and did not have long periods of differential management. The relationship between fruit weight and seed set was determined by weighing fresh fruit and counting seeds of eggplant fruit from two commonly grown varieties, Early Long Purple (n = 47) and Black Beauty (n = 50). In all

Po

st

-P

rin

tv

si er

on

B. Gemmill-Herren, A.O. Ochieng / Agriculture, Ecosystems and Environment 127 (2008) 3136

33

To test for differences in visitation rates by pollinators due to eld distance from the eld/forest edge, ve blocks of the Black Beauty cultivar were divided into three subblocks. The rst of the sub-blocks was located within 20 m of eld edge and wild habitat. The second sub-block was located between 20 and 50 m of the eld/forest interface, and the third was located from 50 to 100 m from eld edge. Within each sub-block, detailed observations of insects visiting eggplant owers were made in standardized 1-m2 quadrats, distributed randomly within blocks. Each observation lasted 10 min, and was made between 06:30 a.m. and 12:30 p.m., with observations times distributed to correspond to the frequency of visitation over the day (see Fig. 2). Number of individuals of each species visiting crop owers and numbers of owers visited by each individual were counted. Number of owers per observation plot was also recorded. Since insect visitors showed no preference for one variety over another, the investigation of the effects of distance from natural habitat on visitation rates focused on only one crop variety (Early Long Purple). An ANOVA (Procedure GLM, SAS) was performed to test if there was a signicant difference in the number and type of foraging pollinators at differing distances from eld/forest interface and between elds, standardized by number of owers per m 2. The land cover within and surrounding agricultural elds was characterised from georectied aerial photographs obtained by the Kenya Trypanosomiasis Research Institute in 1998, and classied into agricultural elds, grassy sward, thick riverine Acacia forest and dry Acacia savanna. Vegetative composition of the cover classes adjoining the agricultural elds, and along open farm pathways was determined by means of three 50-m line intercept transect measurements in taken in January and May, 1999, replicated twice and averaged together. From January to May of 1999, insects visiting owers in these three types of wild habitat adjacent to eggplant elds were observed regularly (on average, 2 days a month) over standardized 1-m2 plots, for 10 min. Monthly from January to May of 1999, the presence of eggplant pollinating species in vegetation communities (as characterised by Agnew et al., 2000) was investigated. A 5km long and 2-m wide transect through the eggplant elds (all originating as clearings in the surrounding riverine forest) was established, and extending beyond these to adjacent remaining riverine Acacia forest, dry Acacia woodland and grassy swards. The transect was sampled once a month over a 5-month period, except in 1 month (April) when logistics prevented the gathering of data. It was surveyed at a constant pace to record both the alternative forage plants and the pollinators in their relative frequencies in the different vegetation communities, the method being based on that of Silveira and Godinez (1996). Given the constant pace, time spent in a particular habitat was taken as the percentage cover of that habitat, over the 5 km 2 m transect area. Comparison was

made of the total number of pollinators observed on each host plant and in each habitat type, in relation to the percentage cover of habitat type. The University of Nairobi Herbarium identied plant species. Pollinator species were collected and sent to the Plant Protection Research Institute in Pretoria, South Africa and the Institut Royal des Sciences Naturelles de Belgique in Belgium for identication, where needed. Voucher specimens have been deposited at the Department of Invertebrate Zoology, National Museums of Kenya.

3. Results Number of seeds per fruit showed a signicant positive correlation with fruit weight (r2 = 0.23, p < 0.05); seed set has thus been used as the metric of pollination effects in the subsequent investigations. Dependence on pollinating vectors varied between crop varieties (Fig. 1). Seed set was highest for both varieties under those treatments that reect the role of animal vectors of pollination. The Early Long Purple variety had signicantly higher numbers of seed when open or hand pollinated than when selfed or exposed to wind pollination;

Po

st

-P

rin tv
Fig. 1. Pollination treatment effects on seed counts of two eggplant varieties (BB Black Beauty, ELP Early Long Purple; letters denote signicant differences between treatments). Fig. 2. Visitation rates of common oral visitors over time (letters denote signicant differences between treatments).

si er

on

34

B. Gemmill-Herren, A.O. Ochieng / Agriculture, Ecosystems and Environment 127 (2008) 3136

Table 1 Visitation rates of oral visits to eggplant owers at varying distances from forest/eld edge Species Mean number of ower visits per m2, over 10 min visitation periods Near sites All species Xylocopa caffra Macronomia rupes Amegilla calens Xylocopa avorufa Pseudapis sp. Xylocopa albiceps 0.214 0.33a,b 0.131 0.29d 0.029 0.08e,f 0.020 0.06h,i 0.001 0.008 0.007 0.06 0.006 0.04 Middle sites 0.160 0.25a,c 0.092 0.20d 0.018 0.07e,g 0.012 0.03h,j 0.002 0.018 0.004 0.03 0.010 0.09 Far sites 0.105 0.15b,c 0.081 0.14 0.004 0.01f,g 0.004 0.01i,j 0.003 0.015 0.002 0.01 0.00 0.01 Signcance level a, P < 0.01; b, P < 0.001; c, P < 0.01 d, P < 0.01 e, P < 0.01; f, P < 0.001; g, P < 0.05 h, P < 0.05; i, P < 0.001; j, P < 0.05 n.s. n.s. n.s.

the Black Beauty variety responded with signicantly higher seed numbers when hand pollinated, above self, wind or open pollination. Nine bee species were seen visiting the eggplant blossoms: Amegilla calens Lepeletier, A. nubica Lepeletier,

A. mellifera L., Crocisaspida sp., M. rupes Smith, Pseudapis sp., X. caffra Vachal, X. albiceps Fabricius, and X. avorufa/inconstans Smith. Only A. calens visited owers before 7:30 a.m. in the morning. Flower visitation rates peaked from 8.30 to

Table 2 Vegetative cover in wild habitat types adjacent to eggplant crop elds, % of total pollinator visits in the habitat type to particular plant species, and number of species of species visiting wild plant species that are also crop pollinators

Po

st

Open pathside

Total number of pollinator visits observed = 901 Commicarpus helenae (J. A. Schultes) Meikle Justicia ava Vahl, Duospermum kilimandscharium (Lindau) Dayton Cajanus cajan (L.) Millsp. Solanum incanum L. Hibiscus palmatus Forsk. Barleria eranthemoides R.Br. Bare ground Thick Acacia tortilis forest (overstory) Total number of pollinator visits observed = 8 Acacia tortilis (Forsk.) Hayne Thick Acacia tortilis forest (understory) Total number of pollinator visits observed = 239 Tephrosia villosa (L.) Pers. Lippia javanica (Burm.f.) Spreng Grewis bicolor Juss. Cordia sinensis Lam. Vigna vexillata (L.) A.Rich. Abutilon hirtum Lam. Solanum incanum L. Acryanthes aspera L. Bare ground Grassy sward Total number of pollinator visits observed = 109 Leucas massaiensis Oliv. Solanum incanum L. Ruellia patula Jacq. Cynodon dactylon (L.) Pers. Bothriocline tomentosa (Oliv. & Hiern) N.E.Br Bare ground

% Cover

% of total pollinator visits 38 34 16 5 3 3 1 0

No. of species of crop pollinators visiting this wild species 4 7 6 3 2 1 0 0

-P
28.3 9.3 9.2 4.6 17.3 6.2 7.8 17.3 % Cover 45.00 % Cover 3.8 4.5 9.7 5.3 31.4 25.3 2.2 13.8 4 % Cover 7.7 4.5 7 56.7 7.3 0

rin tv
% of total pollinator visits 100 % of total pollinator visits 36 19 17 15 6 3 3 2 0 % of total pollinator visits 78 16 6 0 0 0

si er

No. of species of crop pollinating bees visiting this wild species 2

on
2 3 2 1 0 0 3 0 0 5 2 1 0 0 0

No. of species of crop pollinating bees visiting this wild species

No. of species of crop pollinating bees visiting this wild species

B. Gemmill-Herren, A.O. Ochieng / Agriculture, Ecosystems and Environment 127 (2008) 3136

35

10:30 a.m., at an average of 1.98 visits per standardized 10 min observation periods, and decreased thereafter. X. caffra visited owers signicantly more frequently than all other visitors in most time periods, with distinct peaks from 8:30 to 10:30 a.m., and 11:30 to 12:30 a.m. (Fig. 2). A. mellifera visited owers rarely, even though some local farmers kept managed bee colonies. Visitation rates by species did not differ signicantly between crop varieties. Of the nine bee species observed visiting the eggplant blossoms, only two species- X. caffra and M. rupes- were frequent enough to be used in the pollination effectiveness (PE) tests. Seed set due to one visit from each of the two most common pollinators was not signicantly different between the two pollinating species, but the pollinator effectiveness of one visit to the Black Beauty cultivar was signicantly higher than the effect of one visit to the Early Long Purple variety. There were signicant differences in visitation rates with distance from the eld/forest interface (Table 1). Floral visitation rates for all species decreased signicantly with distance from the foresteld edge. For X. caffra, the most abundant oral visitor, visitation rates dropped signicantly between forest edge elds and midelds. For M. rupes and A. calens, the next most abundant species, visitation rates dropped most signicantly between near elds and elds far from the forest edge. For other minor visitors no spatial differences were evident. There were no signicant differences between ve block locations. From the visitation patterns to wild plant species in different habitats near to the elds, it is evident that the ruderal weed community alongside farm paths provides an important alternate forage resource to the wild bees servicing the crops (Table 2). Within the open pathside vegetation community, Justica ava Vahl and Duospermum kilimandscharium (Lindau) Dayton were particularly sought by a large number of crop pollinating bee species. The grassy sward habitat also hosted one species that was highly attractive to crop pollinators, Leucas massaiensis Oliv. The thick (riverine) A. tortilis forest provided less attractive forage plants, although the A. tortilis overstory is an important resource when it owers at the start of the rainy season.

In all months, pollinators foraged more on the farmland than in any other habitat, as represented in Fig. 3. Bee abundance in the farm area was highest in terms of both species composition and species richness due to the constant occurrence of alternative forage plants species persisting as ruderal weeds in farm margins. In other habitats, the alternative forage plants owered over more restricted periods. However, in the month of February (which is at a peak of the dry season in this region), bees also exhibited a selective preference for the forest habitat. Many of the favoured plant species occurring along farm paths and edges dry out during the dry season, but nd a more favourable habitat under the shade of the forest trees.

4. Discussion and conclusion Pollinating insects may signicantly increase seed numbers in eggplant. As fruit weight corresponds to seed number, the scope for signicantly increasing yields of eggplant crops by managing and enhancing insect visitation rates and the pollination services they provide is indicated from these eld trials. In the Nguruman area of Kenya, at least two species of solitary bees were found to be effective pollinators of eggplants. Visitation rates of the all bees taken as a group were signicantly higher closer to the foresteld edge, and dropped off markedly at distances from 50 to 100 m from the Acacia forests. Within the farm landscape, the favoured alternative forage species of pollinating bees were three indigenous plant species. These resources are abundant most of the year as ruderal weeds along paths in the farmland area, but there was a detectable shift in dependence to resources occurring in the forests toward the end of the dry season. With crops such as eggplant that are not optimally pollinated by managed (honeybee) pollinators and must depend on other oral resources for nectar, the availability of alternate resources in the non-farmed habitat adjacent to farm elds is clearly important. The present investigation showed that important alternative oral resources occurred abundantly in farmland as ruderal weeds, for most of the growing season. The health and diversity of the weed community is probably the most important landscape feature for alternative food plants for crop pollinators, in general in this system. Looking at this data alone, one might suggest that the A. tortilis riverine forest, which is being cleared to bring more land into cultivation, is not an important resource from the standpoint of pollination. However, in one month pollinators preferentially chose the riverine woodland, in addition to the weeds in the farm landscape. This shift in resource use occurred at the end of the dry season, when oral resources are scarce. Flowers of the Solanaceae family including eggplant are well known for not being highly attractive to pollinators,

Fig. 3. Ratio of the number of bees in each habitat in relation to the percentage cover of this habitat represented in the transect.

Po

st

-P

rin

tv

si er

on

36

B. Gemmill-Herren, A.O. Ochieng / Agriculture, Ecosystems and Environment 127 (2008) 3136 Buchmann, S.L., 1983. Buzz pollination in angiosperms. In: Jones, C.E., Little, R.J. (Eds.), Handbook of Experimental Pollination Biology. Van Nostrand Reinhold Company, New York, pp. 63113. rgens, A., Neeman, G., Newstrom-Lloyd, L., Potts, S.G. Dafni, A., Ju Towards a standardized framework for comparing pollinator performance: effectiveness and efciency (submitted for publication). Free, J.B., 1993. Insect Pollination of Crops, second ed. Academic Press, London, UK. Greenleaf, S.S., Kremen, C., 2007. Wild bees enhance honey bees pollination of hybrid sunower. Proc. Natl. Acad. Sci. USA 103 (37), 13890 13895. Kearns, C.A., Inouye, D.W., 1997. Pollinators, owering plants, and conservation biology. BioScience 47 (5), 297307. Klein, A.-M., Steffan-Dewenter, I., Tscharntke, T., 2003. Fruit set of highland coffee increases with the diversity of pollinating bees. Proc. R. Soc. Lond. B Biol. Sci. 270, 955961. ` re, B.E., Cane, J.H., Steffan-Dewenter, I., CunningKlein, A.-M., Vaissie ham, S.A., Kremen, C., Tscharntke, T., 2007. Importance of pollinators in changing landscapes for world crops. Proc. R. Soc. Lond. B Biol. Sci. 274, 303313. Kremen, C., Williams, N.M., Aizen, M.A., Gemmill-Herren, B., LeBuhn, G., Minckley, R., Packer, L., Potts, S.G., Roulston, T., Steffan-Dewenter, I., Vazquez, D.P., Winfree, R., Adams, L., Crone, E.E., Greenleaf, S.S., Keitt, T.H., Klein, A.-M., Regetz, J., Ricketts, T.H., 2007. Pollination and other ecosystem services produced by mobile organisms: a conceptual framework for the effects of land-use change. Ecol. Lett. 10, 299314. Kremen, C., Williams, N.M., Thorp, R.W., 2002. Crop pollination from native bees at risk from agricultural intensication. Proc. Natl. Acad. Sci. USA 99 (26), 1681216816. Levin, M.D., 1989. Honey bee pollination of the eggplant (Solanum melongena L.). The 31st International Agricultural Congress, August 1987, Warsaw, Poland. Apimondia Publishing House, Bucharest, pp. 344348. Ricketts, T.H., 2004. Tropical forest fragments enhance pollinator activity in nearby coffee crops. Conserv. Biol. 18, 12621271. Silveira, F.A., Godinez, L.M., 1996. Systematic Survey of Local Bee Faunas. Melissa, 9. Spears Jr., E.E., 1983. A direct measure of pollinator effectiveness. Oecologia (Berlin) 57, 196199. Westerkamp, C., 1991. Honeybees are poor pollinators why? Plant Syst. Evol. 177 (1-2), 7175.

since their oral resources are usually limited to pollen. This study highlights the use of alternative resource, both close to and distant from farm elds, by pollinators. In conclusion, it has been possible to document an important pollination ecosystem service provided by wild habitat to horticultural development in a smallholder farming system in Kenya. Conserving this service will require sensitive development, and the conservation of wild habitat in and among and around the crop elds. Improved understanding of how pollination services both vary with respect to, and can be manipulated through, the spatial and temporal distribution of resources surrounding the agricultural elds sites could lead to additional practical management guidelines for building sustainability into agricultural systems as they intensify.

Acknowledgements The authors thank the Conservation, Food and Health Foundation of the United States, and the Commercial Insects Department of the International Centre of Insect Physiology and Ecology, in Kenya, for support for this research; the eggplant farmers of the Nguruman community, particularly Mr. Lentai Musa for kindly accommodating our research on their farms; and two anonymous reviewers for their many helpful comments.

Po

st

-P

rin

References

Agnew, A.D.Q., Mwendia, C.M., Oloo, G., Roderick, S., Stevenson, P., 2000. Landscape monitoring of semi-arid rangelands in the Kenyan Rift Valley. Afr. J. Ecol. 38, 277285. Blanche, K.R., Ludwig, J.A., Cunningham, S.A., 2006. Proximity to rainforest enhances pollination and fruit set in orchards. J. Appl. Ecol. 43, 11821187.

tv

si er

on