Research in Action South African Journal of Science 104, July/August 2008 247

closely synchronized with the distinct Dasineura rubiformis (Diptera: flowering pulse exhibited by mearnsii during spring (September/Octo- Cecidomyiidae), a new biological ber). Adult D. rubiformis live only a few days and females require open flowers for control agent for oviposition, so this synchrony is essential. The eggs are laid within the perianth tube in South Africa of the flower and, on hatching, the larvae start feeding on the surface of the ovary, F.A.C. Impsona,b*, C.A. Kleinjanb, J.H. Hoffmannb and J.A. Posta at the same time inducing gall-formation and preventing pod set by affected flow- ers. The flowers of A. mearnsii occur in cecidomyiid midge, Dasineura rubifor- programme against invasive Australian globular flower-heads, each with about 45 mis, is the most recent addition to the Acacia species in South Africa has been flowers. Afflicted flower-heads generally Asuite of biological control agents that fraught with conflict, particularly due to produce a small, tightly packed cluster of have been deployed in South Africa against the economic importance (for tannin up to 36 galls [10.5 ± 1.0 s.e., n =83at invasive Australian Acacia species. This 14–18 is associated with Acacia mearnsii (black and paper pulp) of black wattle. As a Stellenbosch in July 2007 (unpublished wattle), which is extremely invasive, but also compromise, the choice of biological data)] instead of pods. Each gall within an important agro-forestry species, in South control agents has been restricted to those the cluster contains 1–5 chambers, and Africa. It induces development of galls in the that reduce the reproductive capacity but generally a single develops within flowers of A. mearnsii, thereby preventing not the useful attributes of the plants.6,18,19 each chamber. Third-instar larvae emerge pod development and reducing the reproduc- The concerns about biological control from the galls during winter (June/July) tive capacity of the plants. The useful attributes potentially disrupting commercial supplies and drop to the soil where they pupate in of this economically important plant species of seed were resolved by demonstrating should not be affected by the introduction of silken cocoons. D. rubiformis. The midge is established in the that the prospective agents can be effec- vicinity of Stellenbosch, where it is increas- tively controlled with conventional insec- Host range ing in abundance. Studies have been initiated ticides,3,20 some of which are routinely Surveys in Australia of 147 native Acacia to (i) evaluate the performance of the midge; applied in wattle plantations.21–23 species, including 27 within the section (ii) confirm that galling does not cause a reduc- The first agent to be released against Botrycephalae, along with three intro- tion in vegetative growth of A. mearnsii; and A. mearnsii was a seed-feeding weevil, duced African species and one intro- (iii) determine the potential effectiveness of Melanterius maculatus, which is now estab- duced American species,3 showed that, D. rubiformis as a biological control agent lished at a number of sites across the besides A. mearnsii, D. rubiformis may be of A. mearnsii. All indications are that the 6 insect has the potential to become an excellent country. Although M. maculatus can associated with the following Botry- seed-reducing biological control agent of cause substantial levels of seed reduction cephalae species in eastern Australia: Aca- A. mearnsii. of A. mearnsii, considerable quantities cia parramattensis, A. irrorata, A. deanei, of seed are still produced annually (F. A. leucoclada and A. constablei. The struc- Background Impson, unpublished data). The addition ture of galls on these five species resem- A diverse array of gall midges (Cecido- of complementary biological control bles that of D. rubiformis on A. mearnsii myiidae) is associated with the Australian agents to further reduce reproductive and DNA sequences of larvae from the in their native habitat.1 In the late output of A. mearnsii was therefore desir- first four species matched those of D. rubi- 1990s, attention was focused on them in a able. formis. Larvae from the fifth species were drive to find additional biological control Eight gall midge species are associated not sequenced. To confirm these putative agents for use against invasive Australian with the reproductive structures of black identifications, adults need to be reared Acacia species in South Africa with empha- wattle in Australia. Of these, D. rubiformis from the galls and examined.1 In Western sis on black wattle (Acacia mearnsii).2,3 was considered to be the most suitable Australia, where both insect and host Acacia mearnsii is one of South Africa’s candidate for use in the biological control plant are naturalized, D. rubiformis was most widespread and problematic invasive programme against black wattle in South found only on A. mearnsii.1 Of the species plants.4–7 Its presence in natural forests, Africa. This decision was based partly on listed above, only A. mearnsii currently grasslands and water courses continues observations that D. rubiformis substantially occurs in South Africa. to threaten ecosystems in terms of loss of reduces pod production of A. mearnsii in Host-specificity tests, conducted at the biodiversity, reduced water supplies, in- Western Australia, where both the midge Plant Protection Research Institute in creased soil erosion and by exacerbating and host plant have become naturalized Stellenbosch between 1999 and 2001, the intensity of fires.8–12 The success of after being introduced from eastern confirmed the restricted host range of the plant as an invasive weed and its Australia.3 In this article we summarize D. rubiformis. During these studies flow- persistence in the landscape is largely progress with D. rubiformis in South Africa ering branches of 10 Australian Acacia attributable to the annual production of and discuss the issues that need to be species [including four species from the enormous seed crops which accumulate addressed before the insect is exploited section Botrycephalae, namely, A. mearn- and persist in the soil for many years.6,13 fully as a biological control agent of sii, A. dealbata, A. baileyana and A. decur- Historically, the biological control A. mearnsii. rens, 13 African Acacia species, an Ameri- can acacia, and an additional four species aPlant Protection Research Institute, Private Bag X5017, Stellenbosch 7600, South Africa. Life cycle of D. rubiformis of test plant (Paraserianthes lophantha, bZoology Department, University of Cape Town, Private The life cycle of D. rubiformis has been Mimosaceae; Cydonia oblonga and Prunus Bag, Rondebosch 7701, South Africa. described by Adair.3 It is a univoltine armeniaca, both Rosaceae; and Vitis vini- *Author for correspondence. E-mail: [email protected] species in which adult emergence is fera, Vitaceae)] were exposed to cohorts 248 South African Journal of Science 104, July/August 2008 Research in Action

of D. rubiformis adults and monitored for Africa. Qualitative assessments have Weeds, ed. N.R. Spencer, pp. 605–614. Bozeman, Montana. subsequent development of galls. Under shown that levels of galling are increasing 3. Adair R.J. (2004). Seed-reducing Cecidomyiidae as the test conditions, gall induction occurred (unpublished data). In 2006, galls were potential biological control agents for invasive Austra- only on A. mearnsii.3 detectable over a range that extended lian wattles in South Africa, particularly Acacia for about 800 m. In 2007 this range had mearnsii and A. cyclops. Ph.D. thesis, University Host plant growth of Cape Town, South Africa. increased sixfold, up to 5000 m. In Western 4. Stirton C.H. (ed.) (1978). Plant Invaders, Beautiful One of the concerns when using biolog- Australia D. rubiformis has demonstrated but Dangerous. Department of Nature and Envi- ical control agents that reduce seed pro- an ability to locate disparate populations ronmental Conservation of the Cape Provincial duction of invasive acacias by forming Administration, Cape Town. of A. mearnsii, including trees that occur in 5. Versfeld D.B., le Maitre D.C. and Chapman R.A. 3 galls is that the galling may also suppress isolation. The midge will probably show (1998). Alien invading plants and water resources in the growth rate of the host plants. A case the same dispersal ability and eventually South Africa: a preliminary assessment. WRC Report in point is that of the gall wasp, Trichilogaster become established widely in South No. TT99/98, CSIR No. ENV/S-C 97154. Pretoria. acaciaelongifoliae, which is an extremely 6. Dennill G.B., Donnelly D., Stewart K. and Impson Africa. At present no efforts are being F.A.C. (1999). Insect agents used for the biological effective biological control agent of A. made to distribute the manually control of Australian Acacia species and Paraserian- 18,24 longifolia in South Africa. Although until the completion of studies, to confirm thes lophantha (Willd.) Nielsen (Fabaceae) in South most T. acaciaelongifoliae galls are induced Africa. African Entomology, Memoir No. 1, 45–54. that D. rubiformis is no threat to the integ- 7. Henderson L. (2001). Alien Weeds and Invasive in flower buds, the wasps also stunt vege- rity of the wattle industry. Plants. A complete guide to declared weeds and tative growth of A. longifolia because: (i) These studies have begun and include invaders in South Africa. Agricultural Research Council, Pretoria. galls induced in vegetative meristematic monitoring of the extent and intensity of tissues destroy the growth points which 8. Macdonald I.A.W. and Jarman M.L. (eds) (1984). galling on A. mearnsii along with rates of Invasive alien organisms in the terrestrial ecosystems of would give rise to new stems; and (ii) galls spread and impact on pod production. the fynbos biome, South Africa. S. Afr. Natl Sci. Prog. suppress growth indirectly because their Physiological studies are being under- Rep. 85. CSIR, Pretoria. biomass is routinely much higher than 9. VersfeldD.B. and van Wilgen B.W.(1986). Impacts taken to confirm that compensatory of woody aliens on ecosystem properties. In The that of normal seed pod loads (that is, that photosynthesis occurs and to determine Ecology and Control of Biological Invasions in South 24,25 would occur on ungalled trees). the extent of photosynthesis by the galls Africa, eds I.A.W.Macdonald, F.J.Kruger and A.A. Conversely, two other introduced gall- Ferrar, pp. 239–246. Oxford University Press, themselves. A comparison of the vegeta- Cape Town. forming insects on Australian acacias in tive growth of A. mearnsii, relative to 10. Macdonald I.A.W.(1991). Conservation implications South Africa have had little or no indirect intensity of galling, is being undertaken of the invasion of southern Africa by alien organisms. Ph.D. thesis, University of Cape Town, South effect on the vegetative growth of their to confirm that growth is not suppressed. host plants.26 Physiological studies have Africa. Additional studies will determine how 11. Cowling R.M. and Hilton-Taylor C. (1994). shown that compensatory photosynthesis D. rubiformis (i) interacts with the intro- Patterns of plant diversity and endemism in in adjacent phyllodes offsets the carbon duced seed weevil, M. maculatus; and (ii) southern Africa: an overview. In Botanical Diver- demands placed on the host plant by galls sity of Southern Africa, ed. B.J. Huntley. Strelitzia 1, might be influenced by acquired para- 31–52. of Trichilogaster signiventris on Acacia 12. De Wit M.P., Crookes D.J. and van Wilgen B.W. 27 sitoids and predators in South Africa. pycnantha and by galls of Dasineura dielsi Cecidomyiid midges have a poor reputa- (2001). Conflicts of interest in environmental management: estimating the costs and benefits of on Acacia cyclops (C. Moseley, University tion as biological control agents because of Cape Town, unpublished data). The a tree invasion. Biol. Invas. 3, 167–178. they are prone to attack when 13. Pieterse P.J. and Boucher C. (1997). Is burning a compensation is possible because the introduced into new areas.28–30 In Western standing population of invasive legumes a viable control method? Effects of a wildfire on an Acacia galls induced by these two species do not Australia, however, D. rubiformis has be- disrupt plant tissue functioning.27 The mearnsii population. S. Afr. For. J. 180, 15–21. come abundant and effective in suppress- 14. Annecke D.P. (1975). Biological control of Austra- studies also demonstrated that photo- ing pod production despite the acquisition lian Acacia species. Official communication ref. synthetic activity of the galls themselves of .3 13/3/417/6 dated 26 November, 1975 from director, may contribute substantially to the car- Plant Protection Research Institute, to director, bon budgets of the galls,27 thus further off- Wattle Research Institute. Conclusion 15. Stubbings J.A. (1977). A case against controlling setting the carbon demands of the galls. Based on available evidence, D. rubiformis introduced acacias. In Proc. Second National Weeds Dasineura rubiformis is expected to re- Conference of South Africa, pp. 89–107, Stellenbosch. will be restricted to A. mearnsii in South semble D. dielsi most closely and not cause A.A. Balkema, Cape Town. Africa with the capacity to reduce pod 16. Anon. (1978). Biological control of Australian any reduction in the growth rates of the production substantially while not reduc- acacias. Wattle Growers News 62, 10–12 host plant because: (i) D. rubiformis lays its 17. Anon. (1987). Editorial. The history of the conflict ing the vigour of the plants. All indications eggs exclusively in flowers of A. mearnsii, of interest regarding the biological control of alien are that the midges will be compatible and consequently galls never develop in acacias in South Africa. South African Institute of with the commercial exploitation of Ecologists. Bulletin 6(3), 1–9, eds B.W. van Wilgen vegetative meristematic tissue; and (ii) the wattle. In combination, it is anticipated and B. McKenzie. biomass of each D. rubiformis gall is much 18. Dennill G.B. and Donnelly D. (1991). Biological that M. maculatus and D. rubiformis will less than a seed pod [gall mean dry bio- control of Acacia longifolia and related weed make a beneficial contribution to curbing species (Fabaceae) in South Africa. Agric. Ecosyst. mass = 5.8 ± 0.3 mg (n = 50), seed pod the invasiveness of one of South Africa’s Environ. 37, 115–135. mean dry biomass = 292 ± 27 mg (n = 19. Impson F.A.C.and Moran V.C.(2004). Thirty years most troublesome alien plant species. 40)]. The number of galls that are pro- of exploration for and selection of a succession of duced may exceed the number of pods Melanterius weevil species for biological control of 1. Kolesik P., Adair R.J. and Eick G. (2005). Nine new invasive Australian acacias in South Africa: that would normally be generated, but species of Dasineura (Diptera: Cecidomyiidae) should we have done anything differently? In this would need to be about 50-fold to from flowers of Australian Acacia (Mimosaceae). Proc. XI International Symposium on Biological require equivalent resources. Syst. Entomol. 30, 454–479. Control of Weeds, eds J.M. Cullen, D.T. Briese, D.J. 2. Adair R.J., Kolesik P. and Neser S. (2000). Austra- Kriticos, W.M. Lonsdale, L. Morin and J.K. Scott, Evaluation studies lian seed-preventing gall midges (Diptera: Cecido- pp. 127–134. CSIRO Entomology, Canberra, myiidae) as potential biological control agents for Australia. Dasineura rubiformis is currently estab- invasive Acacia spp. in South Africa. In Proc. X 20. Donnelly D., Calitz F.J. and van Aarde I.M.R. lished in the Stellenbosch region of South International Symposium on Biological Control of (1992). Insecticidal control of Melanterius servu- Research in Action South African Journal of Science 104, July/August 2008 249

lus (Coleoptera: Curculionidae), a potential bio- analysis of the SAWGU Survey, 1953–1997. ICFR Africa. Biol. Control 25, 64–73. control agent of Paraserianthes lophantha (Legumi- Bulletin Series 4/99. 27. Dorchin N., Cramer M.D. and Hoffmann J.H. nosae), in commercial seed orchards of black wat- 24. Dennill G.B. (1988). Why a gall former can be a (2006). Photosynthesis and sink activity of tle, Acacia mearnsii (Leguminosae). Bull. Ent. Res. good biocontrol agent – the gall wasp Trichilo- wasp-induced galls in Acacia pycnantha. Ecology 82, 197–202. gaster acaciaelongifoliae and the weed Acacia longi- 87(7), 1781–1791. 21. Govender P.and Atkinson P.R. (eds) (1997). Wattle folia. Ecol. Entomol. 13, 1–9. 28. Harris P. (1991). Classical biocontrol of weeds: its Pests and Diseases. Institute for Commercial 25. Dennill G.B. (1985). The effect of the gall wasp definitions, selection of effective agents, and Forestry Research and the South African Wattle Trichilogaster acaciaelongifoliae (Hymenoptera: administrative-political problems. Can. Entomol. Growers Union,Pietermaritzburg, South Africa. Pteromalidae) on reproductive potential and 123, 827–849. 22. Nel A., Krause M., Ramautar N. and van Zyl K. vegetative growth of the weed Acacia longifolia. 29. Goeden R.D. and Louda S.M. (1976). Biotic inter- (1999). A Guide for the Control of Plant Pests, 38th Agric. Ecosyst. Environ. 14, 53–61. ference with insects imported for weed control. edn. National Department of Agriculture, Pretoria. 26. Hoffmann J.H., Impson F.A.C., Moran V.C. and Ann. Rev. Entomol. 21, 325–342. 23. Atkinson P.R. (1999). On the population dynamics Donnelly D. (2002). Trichilogaster gall wasps 30. Harris P.and Shorthouse J.D. (1996). Effectiveness of the wattle bagworm, Chaliopsis () (Pteromalidae) and biological control of invasive of gall inducers in weed biological control. Can. junodi (Heylarts) (: Psychidae): an golden wattle trees (Acacia pycnantha) in South Entomol. 128, 1021–1055. cotton-growing regions in Kenya. Data Prediction of cotton yield in Kenya volumes from a given cotton region corre- lated with the amount of cotton grown in that region. This ensured that the data *‡ * Josphat Igadwa Mwasiagi , Xiu Bao Huang and in hand were statistically representative Xin Hou Wang* of the local industry. There are three cotton-growing regions in the country: OTTON YIELD IS ONE OF THE INDICATORS influence cotton yield are government eastern/central (e/c), coastal (c) and west- for describing agricultural efficiency policies, crop husbandry methods and ern (w). Their respective percentages of Cfrom different resource management cotton price.1–3 Crop husbandry methods cotton lint production for the eight years methods in the cotton-growing industry. such as land preparation, the type and Selected cotton-growing cost factors were from 1996 to 2003 were 54%, 15% and used to design an artificial neural network quantity of fertilizers used, and weed and 31%. Corresponding records of data model to predict cotton yield in Kenya. This pest control methods, have been reported collected from the eastern/central, coastal neural network model was able to predict to have a primary influence on cotton and western regions were 39, 11 and 22, cotton yield with a satisfactory performance quality and yield.4–6 respectively. The DAOs normally collect error of 0.204 kg/ha and a regression correla- Cotton yield can be predicted using cost factor data, crop yields and prices tion coefficient between network output and satellite forecasting.7,8 This method makes that were paid for all the crops grown in a actual yield of 0.945. use of the latest technologies in surveil- given district. We identified a total of ten Introduction lance and data collection, and is set to cost factors within the archived cotton- Cotton can be grown in all the eight replace traditional means of crop predic- growing data. These cost factors are provinces of Kenya and especially in the tion. While prediction of cotton yield using ploughing (plo), harrowing (hr), seed cost semi-arid regions where few other com- satellite forecasting is gaining popularity, (sd), fertilizer cost (ft), planting (pla), mercial crops are viable. The Kenyan cot- especially in the main cotton-producing weeding (we), pesticide cost (pes), harvest- ton-processing industry has an installed countries, there are other, less-affluent ing (ha), transportation to the ginnery (tr), capacity of over 120 000 bales of cotton lint countries where the cotton-growing in- as well as other miscellaneous expenses per year, which is able to meet half of the dustry struggles to survive, and a cheaper (ot). These factors, together with the related national demand for cotton products. yet effective forecasting method is still crop yields, form the main core of our Kenya has consistently produced less needed. Such a method will serve as an data, from which we correlated produc- than 30 000 bales of cotton lint annually interim measure for a struggling cotton- tion costs with crop yields using our since 1990, so the country is a net importer growing industry to rationalize itself and ANN yield prediction model. The aim in of the commodity, causing adverse strain achieve a requisite level of competitive- designing the model was to create a on its foreign exchange reserves. ness. predictive tool for cotton yield before Crop yield is one of the indicators for Kenya, being mainly an agricultural harvest-time. Seven cost factors—plough- describing agricultural efficiency arising economy, has a well-established ministry ing, harrowing, seeds, fertilizers, planting, from different resource management of agriculture, spreading its operations weeding and pesticides—were selected, methods. The need to monitor cotton over the entire country. The district being data that a DAO collects from the yield in Kenya is essential, given that agricultural officers (DAOs), at district farmers prior to harvest. The input and cotton-growing and yields have under- level, are responsible for collection of output vectors were 72×7and72×1, gone an unprecedented decline in the last specific data pertaining to all agricultural respectively. A back-propagation feed- ten years. Cotton yield forecasts, a few crops grown in their area. These data are forward ANN, with one hidden layer was weeks before harvest time, can be of archived at district, provincial and national used, and the number of neurons in the strategic importance for advance trade levels. We have undertaken a study of the hidden layer varied from 2 to 20 in steps of planning, in order to pre-empt national data for cotton growing from 1996 to 2003. two. The performance of the ANN was cotton lint demand, and rationalize cotton The cost factors that influence cotton assessed using root mean squared error lint ginning and marketing. An effective growth, together with corresponding (RMSE),9 defined in Equation (1), where N cotton yield model can also be useful for yields have been used to predict cotton is the number of input–output pairs, y is analysis of factors that affect cotton agri- yield by means of an artificial neural the cotton yield and y$ is the ANN output culture in a region. The main factors that network (ANN) model. (predicted cotton yield). RMSE has the *Department of Textile Engineering, Donghua University, same units as the factor being predicted. Shanghai, China. Materials and methods ‡Author for correspondence.Address for correspondence: Department of Textile Engineering, Moi University, P.O. Cotton-growing data were accessed Box 3908, Eldoret, Kenya. E-mail: [email protected] from district agricultural officers in all the