APICULTURE AND SOCIAL INSEcrS Pollination of Greenhouse Tomatoes by the North American Bombus vosnesenskii (: )

M. H. DOGTEROM, J. A. MATIEONI,l AND R. C. PLOWRIGHT2

J. Econ. EntomoL 91(1): 71-75 (1998) ABSTRACT The effectiveness of pollination of tomatoes by the bumble bee Bombus vos- nl'senskii Radoszkowski in greenhouses was determined between May and August 1991 by

measuring fruit size and seed content. Bumble bee pollination was compared with no pollination, Downloaded from https://academic.oup.com/jee/article/91/1/71/2216810 by guest on 29 September 2021 manual pollination, and manual plus bumble bee pollination. Bumble bee-pollinated flowers produced larger fruit than non-bumble bee-pollinated flowers. Fruit shape was not affected by bumble bee pollination. Our results show that B. vosnesenskii is an effective pollinator of tomatoes in greenhouses.

KEY WORDS Bombus vosnesenskii, pollination, tomatoes, greenhouse, fruit size, seed content

GREENHOUSE TOMATOES, Lyeopersieon esell/entum Greene), largely for pollinating greenhouse toma- Miller, require supplemental pollination for fruit set toes. It has not been reported whether additional (McGre~or 1976, Review by Picken 1984, Free bumble bee species in western North America can 1992) and are usually pollinated by mechanical vi- effectively pollinate greenhouse tomatoes. One po- bration (manual pollination), which is labor inten- tential bumble bee pollinator is B. vosnesenskii Ra- sive and thus expensive. In Europe, laboratory- doszkowski, which occurs from southern British Co- reared colonies of Bombus terrestris L. have been in lumbia, Canada, down to southern California, USA tomato ~reenhouses since 1987 and have subse- (Stephen 1957). quently replaced manual pollination (van Raven- Our objective was to evaluate the western North stijn and Nederpel1988, van Ravenstijn 1989, van America B. vosnesenskii as a pollinator of ~reen- Hecmcrt et al. 1990). Pollination by B. terrestris house tomatoes at 2 levels of foraging activity. We resulted in si~nificantly heavier fruit (Banda and also examined if B. vosnesenskii selected previously Paxton 1991, Ravestijn and Sande 1991) when com- pollinated flowers (by manual pollination). pared with manual pollination, although fruit was of similar wei~ht in another study (Kevan et al. 1991). Materials and Methods In North America, Agriculture Canada and the United States Department of Agriculture restrict Two separate experiments were conducted in a 2 the importation of European bumble bee species, greenhouse (5,600 m ) in Surrey, BC, Canada. The and restrict the movement of bumble bee species greenhouse was rectangular, glass-Venlo, with liq- within North America to north-south movements uid-feed systems (Anonymous 1996), and ambient within east or western provinces or states. Thus, lighting. Daily temperatures were maintained be- several native bumble bee species have been inves- tween 18 and 22°C. Density of 'Dombito' tomato ti~ated for suitability as pollinators of greenhouse plants ('Beefsteak') was =2.5 plants per square tomatoes because of restrictions to their movements meter. Trusses were pruned to 3- 4 fruit per truss and secondarily because not all bumble bee species and plants were maintained according to standard visit tomatoes (Bombus hortorum L. in Pinchinat et commercial practices. Plants were 5 mo old in mid- al. 1979; Brewer and Denna 1980). Researchers in- April at the start of the experiment. Fruit was picked vesti~atin~ bumble bee pollination in greenhouses during July and August. in Canada (Plowright and Laverty 1987) reported Test plants were selected in 4 rows that were that 2 bumble bee species (B. bimaeu/atus Cresson spread through the greenhouse, and location within and B. impatiens Cresson) effectively pollinated rows was selected at random. Each treatment rep- greenhouse tomatoes (Kevan et al. 1991). Since that licate was represented by 3 plants adjacent to each time, commercial suppliers produce bumble bees other. Treatment replicates were placed in the same year-round in both eastern North America (B. im- order for all plots to minimize error in treatments patiens) and western North America (B. oeeidentalis given to each plant. Manual pollination was com- pleted between 1000 and 1200 hours 3 times per week. I Kwantlen University Colle!\e, P.O. Box 9030, Surrey, BC, Can- uda V3W 2M8. Bumble bee colonies (modified after Plowright " 482 Montee de la Source, Cantley, PQ, Canada J8V 3H9. and Jay 1966) were evenly spaced along the center

0022-0493/98/0071-0075$02.00/0 © 1998 Entomological Society of America 72 JOURNAL OF ECONOMIC ENTOMOLOGY Vol. 91, no. 1 working aisle within each greenhouse and supplied bees visited between 80 and 100% of the flowers with diluted with water (2:1, vol:vol) every examined for bruising; a low level of foraging ac- 2-3 d. Four colonies of B.vosnesenskii (7 colonies per tivity occurred between 17 June and 3 July (LOW hectare) were introduced on 19 April 1991, and foraging activity) when bumble bees visited be- experimental pollination occurred from 8 May to 3 tween 50 and 80% of the flowers examined for bruis- July. Bumble bee foraging activity was assessed on ing. The latter period coincided with a decline in 7 and 17 June at 1300 hours by counting incoming nest size, number of workers in the colony, and the and exiting bumble bees. Intensity of foraging ac- presence of nonforaging sexual reproductive fe- tivity was 16 bees per 5 min per colony (mean of 4 males and males. colonies) of which 0.81 bees were incoming Experiment 2. A 2nd test was conducted to de- foragers. At midday on 17 June, the average bumble termine if there was an effect by the presence of bee colony population estimate was 115 workers manually pollinated flowers on adjacent plants pol- Downloaded from https://academic.oup.com/jee/article/91/1/71/2216810 by guest on 29 September 2021 (range, 100-150). This colony population estimate linated by bumble bees. Bumble bee-pollinated test excludes foragers working on the crop. The majority plants were located adjacent to either manual or of bumble bee foraging occurred between 1000 and bumble bee pollinated plants. Thus, yield (fruit 1500 hours. weight) and seed count of bumble bee-pollinated The level of bumble bee foraging activity on flow- plants were compared when these plants were ad- ers was determined by the presence of brown, jacent to manually pollinated plants and bumble bruised spots on the anther cones (Bin and Sorressi bee-pollinated plants. In addition, no-pollination 1973) that were damaged by bees' tarsal scratch test plants were compared when plants were located marks. Bumble bee foraging activity was monitored adjacent to either manual- or bumble bee-polli- (Ravestijn and Sande 1991) 3 times per week by nated plants. examining these marks on the anther cones of 100 Tomatoes were harvested from each truss when senescing flowers. Flowers were picked randomly fruit color was "orange-coloured" stage or darker along 5 rows (20 flowers per row) that were spread (Anonymous 1988). Ripening fruit were picked ev- through the greenhouse. Individual flowers in the ery 2-3 d. All fruit were frozen at collection time, bumble bee treatments were not monitored for vis- individually examined for defects, weighed, and itation by bumble bees. We assumed that all flowers measured with calipers (at a later date). A round- in the bumble bee treatments were visited at least ness index was calculated by dividing maximum by once by bumble bee(s). It is possible that some minimum diameter. Each fruit was prepared for flowers in the bumble bee pollinated group may not seed counting by thawing and removing the major- have been visited by any bumble bees. However, ity of the pulp. The remaining pulp was pushed this is unlikely because bumble bee foraging activity through a No. 20 sieve (850 /L) and the remaining was monitored before flowers were thinned from 10 seeds were counted. to 12 flowers per truss down to 4 to 6 flowers per Tomato measurements were averaged by plant truss. and then by plot, and means of plot were analyzed Bumble bees were prevented from pollinating 2 by 2-way analysis of variance (ANOVA) using the non-bee treatments (manual pollination and no- general linear model (GLM) procedureofSAS (SAS supplemental pollination) by enclosing the flower- Institute 1988). Duncan multiple comparison test ing portion of the individual plants with a polyester separated the means when significant differences mesh bag (100 cm long by 75 cm wide; 3 mm holes). were indicated by GLM procedure. Chi-square was The bagging effect was not addressed in this study, used to compare pollination treatments for each although it was minimized by removing the bag from tomato grade. Correlations and regressions were flowers as soon as fruit had set. When all flowers in calculated using the PROC CORR and PROC REG a truss were set, the fruit-truss plus adjacent leaves procedures, respectively (SAS Institute 1988). Re- were moved outside the bag to allow for normal gression slopes were compared using the general growth and fruit development. No more than 4 linear model (GLM procedure, SAS Institute 1988). leaves were left inside the bag at anyone time. Tukey studentized range test was used to separate The effects of pollination by B. vosnesenskii were differences between regression slopes of treat- determined by measuring its impact on the fruit size, ments. weight, and seed count of Dombito. Tomatoes from all treatments were graded by Experiment 1. Tomato fruit yields and seed num- weight into 6 commercial grades; from small to max- bers were determined for 6-8 trusses in the follow- imum-extra-Iarge (BC Greenhouse Vegetable Re- ing 4 treatments: (1) bumble bee pollination, (2) no search Council). supplemental pollination (bagged), (3) manual pol- lination with a mechanical vibrator (bagged), and Results (4) bumble bee plus manual pollination. Fruit weight and seed number were analyzed for Experiment 1. Pollination treatments had a sig- 2 periods based on an assessment of bee foraging nificant impact on fruit weight (F = 12.41; df = 3,15; activity by the percentage of bruised flowers. A high P < F = 0.0001), seed count per fruit (F = 34.12; df = level of foraging activity occurred between 8 May 3,15; P < F = 0.0001), fruit height (F = 16.82; df = and 10 June (HIGH foraging activity) when bumble 3,15; P < F = 0.0001) and maximum fruit diameter Ft'bruary 1998 DOGTEROM ET AL.: GREENHOUSE TOMATO POLLINATION BY BUMBLE BEES 73

Tobit' 1. (:omparison of l1It"on ± SE weifiljht, seed count, fruit "eight, and rOWldnes8 per plant for 4 treatments; no pollination, numuul l,ollil1otiol1, bumbl •• b••••1101Ii11otiol1,mId hl1mble hees plus monuol pollinotion

Bumble bees Yidd llnd No pollination Manual Bumble bees plus manual quality (,t = 12)" (n = 11) (n = 12) (n = 12) Wd~ht. ~ 149.9:!: 6.8a 159.1 :!: 7.0a 188.4:!: 4.5b 181.0:!: 6.ob S""d (~ount) 165.2 :!:8.4a 213.1 :!: 11.9b 277.8:!: 1O.1c 279.5:!: 9.2c H,·i~ht. mm 55.5:!: 0.7a 56.7 :!:0.7a 60.6:!: O.4b 61.0:!: LOb ~1t\xdium. mm 70.1 :!: 1.2a 71.6 :!: 1.2ab 76.0:!: 0.7c 74.3:!: 0.8cb Roundn •.ss" 1.056 :!: .005a 1.059 :!: .0050 1.054 :!: .0040 1.050 :': .004a

M•.ans in th •. same row followed by different letters are significantly different at the P < 0.05 level as determined by 2-way ANOVA, follow"d by tl•••Dun~an multiple comparison test. Downloaded from https://academic.oup.com/jee/article/91/1/71/2216810 by guest on 29 September 2021 " Numb"r in brackets indicates the number of plants per treatment. I, Fruit roundn,'ss is measured by the ratio of maximum diameter to minimum diameter.

(F == 8.96; df == 3, 15; P < F = 0.0002), but no impact Experiment 2. There were significant differences on fruit roundness (F = 0.09; df = 3, 15; P > F = between bumble bee-pollinated and no-pollination 0.4517). treatments for fruit weight (F = 27.79; df = 3,9; P < Flowers pollinated by B. vosnesenskii produced F = 0.0001) and seed count (F = 50.97; df = 3,9; P < lar~er fruit than manually pollinated flowers (Table F = 0.0001). However, there were no significant 1) as evidenced by significant increases (P < 0.05) differences between bumble bee-pollinated £lowers in fruit weight, height, maximum diameter, and seed when test plants were located adjacent to either count. Mean fruit weight for bumble bee pollination bumble bee or manually pollinated plants (P > 0.05) was 18.4% higher than for manual pollination. Over- all, bumble bee treatments (bumble bees and bumble bees plus manual) resulted in significantly 300 greater (P < 0.05) fruit weight, seed count, fruit 'HIGH' 'LOW Foraging FOlllglng diameter, and fruit height compared with non-bee 250 Activity Activity treatments (manual and no-pollination). In con- b trast, fruit roundness did not differ between treat- § 200 b ments (P > 0.05). .... • b X The significant differences obtained for seed C) 150 number remained consistently higher for bumble bee pollination than manual pollination during both ~ 100 HIGH and LOW levels of bumble bee foraging ac- tivity (Fig. 1). Similarly, fruit weight was greater for 50 bumble bee pollination than manual pollination during both HIGH and LOW forager activity, al- 0 400 though at HIGH foraging activity this difference was 'HIGH' 'LOW not significant. 350 ForagIng FOlllglng Bumble bee treatment also had a higher fre- Activity c c Activity 0: 300 b quency of larger commercial grades of fruit than in W b the manual-pollination treatment. The frequency of a:I 250 b :::iii the extra-large grade was 30% greater in the bumble ::::l 200 bee-pollinated fruit than in the manual-pollinated Z • C fruit (Table 2). Manual pollination produced W 150 W smaller grade fruit (large, medium, small, and extra- U) 100 small) in higher numbers than in bumble bee pol- lination. 50 The largest correlation coefficient was between weight and maximum diameter (1" = 0.925); this was followed by weight and minimum diameter (1" = 0.886), weight and seed number (1" = 0.749), and weight and height (1" = 0.651) (P < 0.001). The regression analyses of weight and seed between the Fig. 1. Comparison of fruit weight (grams) and seed 4 treatments are significantly different (F = 11.10; (number) (=SE) during 'LOW' (50-80%) and 'HIGH' df = 3, 15; P < F = 0.0022). Bumble bee treatments (80-100%) levels of bumble bee foraging activity (exper- were similar to each other (bumble bee and bumble iment 1) for 4 treatments: no-pollination, manual (polli- bee plus manual), but different from non-bee treat- nation), bumble bee (pollination), and bumble bee plus ments (no-pollination and manual) (P > 0.001, manual (pollination). Bars labeled with different letters Tukey studentized range test). are significantlydifferent (P < 0.05). 74 JOURNAL OF ECONOMIC ENTOMOLOGY Vol. 91, no. 1

Table 2. Frequency of 6 commercinl grades of tomatoes produced by flowers pollinated by manunl or bumble bee pollination

Weight range, Manual pollination Bumble bee pollination Commercial X' grade name(l gm (n = 188) (II = 225) Extra-small "=35 < 55 0 0 Small "=55<80 3.2 0.4 0.14 NS Medium 2: 80 < 110 7.4 1.8 0.06 NS Large 2: 110 < 160 43.4 17.9 15.290 Extra-large 2: 160 < 265 41.3 75.8 24.520 Maximum-extra large "=265 4.8 4.2 0.04 NS

Chi-square test: difference between treatments for each grade and the level of significance P < 0.01 (0) and P > 0.05 (NS, not significant). a Commercial grades from the British Columbia Greenhouse Vegetable Research Council. Downloaded from https://academic.oup.com/jee/article/91/1/71/2216810 by guest on 29 September 2021

(Fig. 2). Similarly, no significant differences were Kevan et al. (1991) found no significant difference found between no-pollination treated flowers when in a similar study. test plants were located adjacent to either bumble Regression slopes were different between bee- or manually pollinated plants (P> 0.05, Dun- bumble bee and non-bumble bee treatments, sug- can multiple range test; not graphed) gesting there is a different relationship between fruit weight and seed number for different pollina- tion treatments. Perhaps pollination quality is dif- Discussion ferent when pollination is completed by bumble These experiments indicate that B. vosnesenskii bees versus manual pollination. Manual pollination provides a greater yield of Beefsteak tomatoes (va- is conducted according to a schedule of 3 times per riety Dombito) than does manual-pollination or no- week, whereas bumble bees may visit flowers at an pollination under greenhouse conditions. Fruit optimal time (for fertilization) and perhaps visit weight, height, minimum and maximum diameter, flowers more than once. Thus, although differences grade, and seed count were higher for bumble bee- between regression equations are significantly dif- pollinated flowers than for non-bumble bee-polli- ferent, these differences did not appear to consti- nated flowers (manual and no pollination), al- tute practical differences between treatments. though fruit quality, as indicated by roundness in- The high correlation coefficient (1" = 0.925) for dices, did not significantly differ between fruit weight and maximum fruit diameter indicates treatments. Larger fruit size resulting from bumble that weight is an excellent predictor of maximum bee pollination of flowers was reported in other diameter. studies (Sande 1989, Banda and Paxton 1991, Rav- Our results were not adversely affected by locat- estijn and Sande 1991, Kevan et al. 1991). Banda and ing treatments adjacent to manually pollinated Paxton (1991) and Ravestijn and Sande (1991) re- treatments, Fruit size and seed count of bumble bee- ported that bumble bees produced heavier fruit pollinated fruit was similar when test plants were than when manual pollination was used, although located adjacent to manually pollinated plants or adjacent to bumble bee-pollinated plants. Bumble

300 bees, then, equally pollinate plants that are adjacent to manually pollinated or not manually pollinated. Perhaps bumble bees cannot identify manually pol- 250 linated flowers that have had some of their pollen removed. These results could be important for fu- 200 ture experiments where bumble bee and non-bee co r:: treatments may be located adjacent to each other. :: 150 :IE The effect of bagging plants was not addressed in this study. It is possible that bagging decreased the 100 amount of light that reached the developing fruit, thus inhibiting fruit development. However, the 50 length of time that the top part of each tomato plant remained inside the bag was kept to a minimum by Weight continually moving the bags when flowers were set. Bumble bee pollinated plants adjacantto: The level of flower bruising can be used to mon- o MANUALLY POLUNATED PLANTS e:aBUMBLE BEE! POWNAT'ED PLANTS itor £lower visitation by bumble bees (Ravestijn and Sande 1991) and could be used to indicate when Fig, 2. Comparison of means (weight, height, mini- additional bumble bee colonies are required in the mum,and maximumdiameter, and seed count) (::':SE)for fruit produced from bumble bee-pollinated plants adja- greenhouse. In this study we monitored £lower vis- cent to either manually pollinated plants or bumble bee- itation by bumble bees on each £lower-truss before pollinated plants. There were no significant differences thinning down from 10-12 to 4-6 £lowers per truss. between means (P> 0.05). An 80% level of bruised £lowers would very likely be February 1998 DOGTEROM ET AL.: GREENHOUSE TOMATO POLLINATION BY BUMBLE BEES 75

adequate to ensure that all tomatoes left to mature Bin, F., and G. P. Sorressi. 1973. Insetti pronubi e produ- and ripen were visited by bumble bees. In experi- zione di seme ibrido di pomodoro. Genet. Agrar. 27(1): ment I, the drop below 80%flower bruises appeared 35-74. to have coincided with a drop in colony population Brewer J. W., and D. W. Denna. 1980. Observations on and indicated the need for an additional colony to pollinators of Lycopersicon peruvianuln and L. esculentuln in relation to the production of F, hybrid be placed into the greenhouse. The new colony tomato seed. J. App!. Entomo!. 90: 314-319. should offset the drop-off in the number of foraging Free, J. B. 1992. Insect pollination of crops. Academic, bumble bees and increase the flower bruising back New York. to >80%, which in turn would result in adequate Heemert, C. van, A. de Ruijter, J. van den Eijnde, and J. pollination levels. van der Steen. 1990. Year round production of bumble Seed count may be the most accurate method for bee colonies for crop pollination. Bee World 71 (2); determining levels of pollination because fruit 54-56. weight, but not seed count (Picken 1984), is influ- Kevan, P. G., W. A. Straver, M. Offer, and T. W. Laverty. Downloaded from https://academic.oup.com/jee/article/91/1/71/2216810 by guest on 29 September 2021 enced by environmental conditions such as plant 1991. Pollination of greenhouse tomatoes by bumble resources. Manual pollination significantly in- bees in Ontario. Proc. Entomo!. Soc. Onto 122; 1.5-17. McGregor, S. E. 1976. Insect pollination of cultivated creased seed count over the no-pollination treat- crop plants. U.S. Dep. Agri., Agric. Res. Servo Handb. ment. A further increase in seeds resulted from 496; 357-361. bumble bee pollination, indicating that bumble bees Picken, A.J.F. 1984. A review of pollination and fruit set arr better pollinators than the manual pollination in the tomato (Lycopersicon esculentum Mill.). J. Hor- technique. tic. Sci. 59(1}; 1-13. Pinchinat, B., M. Bilinski, and A. Ruszkowski. 1979. Pos- sibilities of applying bumble bees as pollen vectors in Acknowledgments tomato FJ hybrid seed production. Proceedings IVth International Symposium on Pollination. Md. Agric. Our sincere thanks to grower D. Ryall for the use of his Exp. Stn. Spec. Misc. Pub!. 1; 73-90. greenhouse, and to his staff for their assistance. Also thanks to S. Chittaranjan and M. Balasundar of Simon Fraser Plowright, R. C., and S. C. Jay. 1966. Rearing bumble bee colonies in captivity. J. Apic. Res. 5(3): 155-16.5. University and S.Berezowskyj for their technical help, and a special thanks to B. Mauza for his encouragement Plowright, R. C., and T. M. Laverty. 1987. Bumble becs and crop pollination in Ontario. Proc. Entomol. Soc. throughout the project. Thanks also go to K. Naumann and B. Mauza for their review of an earlier draft of the manu- Ont. U8; 155-160. script, R. Currie for his advice on statistics and method- Ravenstijn, W. van. 1989. Hommels ook in ronde tomaat ology, and R. Balshaw for statistical consultation. In ad- goede vervangers van trillen. Groenten en Fruit dition, I thank the editor and 2 anonymous reviewers for 45(20); 42-43. their comments. This project was funded in part by West- Ravenstijn, W. van, and L.S.R. Nederpe!. 1988. Trostrill- gro Sales, Incorporated, Delta, BC, Canada; British Co- ers in Belgie aan de kant: Hommels doen het werk. lumbia Greenhouse Vegetable Research Council; and Groenten en Fruit 43(32}; 38-41. I.R.A.P. (National Research Council Canada, Industrial Ravenstijn, W. van, and J. van del' Sande. 1991. Use of Research Assistance Program). bumble bees for the pollination of glasshouse tomatoes. Sixth International Symposium on Pollination. Acta Horticult. 288: 204-212. References Cited Sande, J. van del'. 1989. Hommels goed alternatief voor trostrillen vleestomaat. Groenten Fruit 45(20): 40-41. Anonymous. 1988. Tomatoes. International standardiza- SAS Institute. 1988. SAS/STAT user's guide, release 6.03 tion of fruit and vegetables. Organization for Economic ed. SAS Institute, Cary, NC. Co-operation and Development, Paris, France. Stephen, W. P. 1957. Bumble bees of western America 1996. Greenhouse vegetable production guide for com- (Hymenoptera: Apoidea). Technical Bulletin 40. Ag- mercial growers, 1996/97 ed., pp. 20-41. J. Portree ricultural Experiment Station, Oregon State College, [ed.], Province of British Columbia, Ministry of Agri- Corvallis. culture, Fisheries and Food, Victoria, BC, Canada. Banda, H. J., and R. J. Paxton. 1991. Pollination of green- house tomatoes by bees. Sixth International Sympo- Received for publication 30 December 1996; accepted 6 sium on pollination. Acta Horticult. 288; 194-198. October 1997.