Ten years after the release of the water hyacinth mirid Eccritotarsus catarinensis in South Africa: what have we learnt?

J.A. Coetzee,1,2 M.P. Hill1 and M.J. Byrne2

Summary Water hyacinth, Eichhornia crassipes (Mart.) Solms, is the worst aquatic weed in South Africa, and biological control offers the most sustainable control option. The mirid, Eccritotarsus catarinensis (Carvalho) (: ), was released against water hyacinth in South Africa in 1996 and shown to be damaging to the plant and host-specific within the Pontederiacae. Feeding, oviposi- tion and nymphal development were recorded on pickerelweed, Pontederia cordata L., an important aquatic plant in North America but a potential weed in South Africa. The release of this agent allowed us to test in the field that pickerelweed was not part of the mirid’s realized host range. The agent subsequently established at 15 sites around South Africa, including those where climatic modeling had indicated that it would not due to low winter temperatures, calling into question the usefulness of climate-matching techniques in the absence of microclimate and behavioural data. Hypertrophic nutrient conditions also reduced the effectiveness of E. catarinensis due to rapid proliferation of the plant, but the mirid reduced both the vigour and competitive ability of water hyacinth in mesotrophic and eutrophic water. E. catarinensis is emerging as an effective agent in areas of medium to low- nutrient status with a warm climate and should be considered for release in other areas of the world, particularly Africa, where few Pontederiaceae occur. This programme shows the value of considering fundamental vs realized host ranges but suggests that more data are needed to increase confidence in climate compatibility predictions.

Keywords: host specificity, realized host range, climate-matching, post-release evalua- tion, agent impact.

Introduction and released in South Africa in 1996 as a new natural enemy of water hyacinth (Hill et al., 1999). Water hyacinth, Eichhornia crassipes (Mart.) Solms, Hill et al. (1999) found that E. catarinensis had po- remains the world’s worst aquatic weed, even though tential as a control agent of water hyacinth in South up to seven biological control agents have been released Africa due to its host specificity within the Pontederia- against it in at least 30 countries (Julien and Griffiths, ceae and because it has long-lived, mobile adults that 1998). The effects of these agents are spatially and tem- are obviously damaging to the plant – the four nymphal porally variable such that water hyacinth still causes instars and the adults feed gregariously, resulting in problems in many regions, including South Africa (Hill chlorosis and ultimately death of the leaves (Hill et al., and Olckers, 2000). One of the newer agents against 1999). Since 1996, the mirid has been released at least water hyacinth is the mirid, Eccritotarsus catarinensis 18 sites in South Africa (Hill et al., 1999) and has es- (Carvalho) (Hemiptera: Miridae), which was screened tablished at 15. In this paper, we review the results of the last 10 years of research since the mirid’s release. We have conducted 1 Rhodes University, Department of Zoology and Entomology, P.O. Box a range of laboratory and field experiments to (1) further 94, Grahamstown, 6140, South Africa. evaluate the realized host range of the mirid, (2) deter- 2 University of the Witwatersrand, School of , Plant and Environ- mine its thermal physiology and potential distribution mental Sciences, Private Bag X3, Wits 2050, South Africa. Corresponding author: J.A. Coetzee . in South Africa and (3) assess the impact it is likely to © CAB International 2008 have on water hyacinth. Based on these findings, we

512 Ten years after the release of the water hyacinth mirid Eccritotarsus catarinensis in South Africa were able to predict what might occur in the field once tions are subject to frost and winter dieback. Biologi- it was released, and over time, we have been able to cal control in these areas is not as successful as that in assess these predictions. frost-free areas because, in colder areas, regrowth of water hyacinth occurs during spring, whereas the Host specificity populations only reach significant levels during - mid summer (Hill and Cilliers, 1999). This lag period may Before the release of E. catarinensis, host-specificity allow the plant populations to increase unchecked and trials demonstrated that pickerelweed, Pontederia cor- could be responsible for the variable results achieved ­data L., an important, native, littoral plant of waterways by water hyacinth biological control agents in these in the United States, may be at risk because feeding, regions. oviposition and nymphal development were recorded When biological control agents are released into a on it in the laboratory (Hill et al., 1999). This did not new country, they should ideally be species or strains prevent the release of the mirid in South Africa, as from a climatically matched area (Williamson, 1996). pickerelweed is neither indigenous nor economically We therefore investigated various aspects of the mirid’s important. thermal physiology to determine whether it might be Hill et al. (2000) predicted that the results of the limited by cold winter temperatures in South Africa. laboratory host-specificity tests were indicative of an We determined the critical thermal minimum (CTMin, artificially expanded host range, and despite feeding on a point short of death where locomotory impairment non-target pickerelweed under laboratory conditions, occurs, but from which recovery is possible) of E. ca- ° the mirid would have minimal, if any, non-target effects tarinensis to be 1.2 C and the lower lethal limit (LT50, on this species in the field, where natural host-selection the temperature at which 50% of the population dies) cues would prevail. Since E. catarinensis had already to be -3.5°C (Coetzee et al., 2007a). Neither of these been released in South Africa, we were presented with limits is particularly low, and they might prevent the an ideal opportunity to test its realized host range in the mirid from establishing in areas that receive consider- field. We first attempted to force the mirid to establish able winter frost. on pickerelweed plants in the absence of water hya- Another method available for climate matching cinth by sleeving them onto leaves to prevent their ini- is degree-day modeling, which uses temperature and tial dispersal (Coetzee et al., 2003). The mirids fed on time to predict the number of generations that an in- the leaves and produced offspring within the sleeves. sect can complete at a given locality. We calculated that Once the sleeves were removed after 5 weeks, the pick- the mirid’s thermal constant K was 342-degree days, erelweed was monitored for establishment of the mirid. above a developmental threshold t of 10.2°C (Coetzee Under field conditions, E. catarinensis did not sustain et al., 2007a). These values were then used to calculate a population on pickerelweed in the absence of water accumulated degree days according to the methods of hyacinth (Coetzee et al., 2003). Campbell et al. (1974) for 128 South African localities,

We also conducted choice tests in the field by placing using the equation: pickerelweed plants among water hyacinth plants in a (T + T ) { max min – t heavily infested river that had a large, well-established K = S { 2 population of mirids. Monitoring of the pickerelweed plants showed that, although feeding damage was evi- The mean annual degree days accumulated for each dent, it was far less than on water hyacinth and was location was then calculated, which predicted the num- indicative of spillover feeding damage because of the ber of generations that E. catarinensis could complete high mirid population levels (Hill et al., 2000). at each locality. The number of generations that the Therefore, the prediction of Hill et al. (1999) that, mirid could complete during the winter months (April under restricted laboratory conditions, pickerelweed to August) was also calculated. The results of both was a more suitable host for E. catarinensis than under of these calculations indicated that fewer generations field conditions, was correct. were possible at high altitudes, as was expected (Coe- tzee et al., 2007a). Thermal physiology From these data, we predicted that low winter tem- peratures will limit the establishment of E. catarinensis In South Africa, where at least five biological control in the field. This explained the failure of mirid estab- agents have been released, water hyacinth control is not lishment at Delta Park in Johannesburg, a high-altitude as successful as that in tropical areas (Hill and Olckers, site, where it has been released at least three times in 2000), and it was assumed that low winter temperatures early summer, established, but then has not persisted play a crucial role in the successful control of water through the winter as a consequence of heavy frosts. hyacinth in South Africa. Many of the worst water hya- However, our prediction could not explain why the mi- cinth infestations in South Africa occur at high-altitude rid did establish at another high-altitude site on the Vaal sites that are typified by cold winters (Hill and Olckers, River, which experiences similar climatic conditions as 2000). At these high elevations, water hyacinth infesta- Delta Park. It is likely that the effects of microclimates

513 XII International Symposium on Biological Control of Weeds provided by an abundance of overhanging vegetation sink as the plants started to die. Eventually, most of played a major role in providing thermal refuges for the water hyacinth sank, and the majority of the im- E. catarinensis at the Vaal River site, thereby allowing poundment remains clear, with fringing populations of it to persist through the winter. Furthermore, the mir- water hyacinth. Based on examination of the plants in id’s behaviour might allow it to escape extreme tem- the field, the mirid is having an impact on the plants, as peratures in the canopy of the plant by moving down the leaves are brown and clearly chlorotic. towards the crown in cold weather. Therefore, the as- sumption that standard meteorological data can be used Discussion and conclusion to represent the conditions actually experienced by the in the field is a generalization (McClay and In the 10 years that this agent has been established in Hughes, 1995; van Klinken, et al., 2003), highlighting South Africa, it has been shown to have a wider distri- the fact that we cannot ignore the effects of microcli- bution than was first predicted (Coetzee et al., 2007a), mates and behaviour on establishment patterns of the and while prediction of the insect’s thermal resilience mirid in South Africa. was reasonably accurate, the thermal buffering of mi- croclimates considerably underestimated the eventual Impact distribution. Inability to include local microclimates is an inevitable failing of models, which use climate en- Another factor contributing to the variable results velopes as a basis for predicting distributions (Sutherst, achieved by water hyacinth biological control agents 2003), while presence of the weed is not necessarily in South Africa is the effect of eutrophication of bod- a predictor of climatic suitability for the agent (van ies of water (Hill and Olckers, 2000). Water hyacinth Klinken et al., 2003). proliferation is usually closely linked to increases in Laboratory host-specificity trials can produce- am eutrophication in these systems (Hill, 1999), and as a biguous results, as the differences between the fun- result, the effect of feeding by biological control agents damental/physiological and realized/ecological host is often insufficient to retard water hyacinth growth ranges are difficult to determine (van Klinken, 2000). (Hill and Cilliers, 1999). Undertaking field host-range studies in the region of A previous study investigated the effects of herbiv- origin of the weed can often resolve these ambigui- ory by the mirid on water hyacinth grown at high, me- ties, but in most cases, only plants common to both dium and low nitrogen (N) and phosphorus (P) nutrient countries can be tested (Olckers et al., 2002). Pickerel- concentrations under laboratory conditions (Coetzee weed’s presence in South Africa allowed validation of et al., 2007b). The results showed that water-nutrient laboratory host-specificity results in the field, including concentration affected plant growth parameters of wa- assessment of the mirid’s suitability for release in the ter hyacinth significantly more than did herbivory by United States. Hill et al. (1999) predicted that at best the mirid. At high-nutrient concentrations, leaf and pickerelweed would be an inferior host in comparison daughter plant production were more than double than to water hyacinth. Field trials confirmed that the mirid at low-nutrient concentrations, while stems were twice would not establish on pickerelweed in the absence of as long at high-nutrient concentrations compared to water hyacinth, but where the two grow sympatrically, low concentrations. Chlorophyll content was also twice some spillover feeding is expected (Hill et al., 2000; as high at high-nutrient concentrations as at low con- Coetzee et al., 2003). However, this spillover feeding centrations. Although herbivory by E. catarinensis did on pickerelweed has not been quantified and possibly not have as great an effect on water hyacinth vigour should be, before the agent is considered for release in as nutrient concentration, it significantly reduced the the United States. In this case, the laboratory results length of second petioles, chlorophyll content of water overestimated the potential impact on the non-target hyacinth leaves and the production of daughter plants species. (Coetzee et al., 2007b). These results are important E. catarinensis was expected to contribute to the because water hyacinth populations increase rapidly control of water hyacinth (Hill et al., 1999), but its by vegetative reproduction through the production of overall impact was predicted to be subtle in comparison daughter plants (Edwards and Musil, 1975), so any re- to the Neochetina spp. weevils that are the mainstay of duction in daughter plant production will have negative water hyacinth biological control (Coetzee et al., 2005) consequences for the rate of spread of water hyacinth. and likely to be negligible in highly eutrophic condi- From these results, we predicted that the mirid would tions (Coetzee et al., 2007b). Five to six years after re- have the greatest impact on water hyacinth infestations lease, mirid populations were generally low and their under mesotrophic conditions. At Clairwood quarry in impact slight. However, in the past 2 to 3 years, several KwaZulu-Natal, a eutrophic site, E. catarinensis has outbreaks of the mirid have been seen, resulting in wa- had a major impact on the infestation and is responsible ter hyacinth mats collapsing, even at eutrophic sites. It for clearing the weed. Initially, large, brown, circular is uncertain if these high populations of the mirid will patches appeared in the mat, which gradually began to persist, and this aspect warrants further study.

514 Ten years after the release of the water hyacinth mirid Eccritotarsus catarinensis in South Africa

Since its first introduction to South Africa in 1992, cal Control of Weeds in South Africa (1990–1998). Afri- there have been ten scientific publications on the biol- can Entomology Memoir 1, 103–112. ogy, host specificity and impact of E. catarinensis on Hill, M.P. and Olckers, T. (2000) Biological control initiatives water hyacinth. Several of the predictions made about against water hyacinth in South Africa: constraining fac- the agent proved to be accurate, while some were un- tors, success and new courses of action. In: Julien, M.H., Hill, M.P., Center, T.D. and Jianqing, D. (eds) Biologi- derestimates and others overestimates. Despite this cal and Integrated Control of Waterhyacinth, Eichhornia close examination, we are still unclear of the interac- crassipes. Proceedings of the 2nd Meeting of the Global tion this agent will have with the other agents released Working Group for the Biological and Integrated Control against water hyacinth in South Africa (e.g. Ajuonu et of Waterhyacinth. Beijing, China, 9–12 October 2000. al., 2007) and its long-term impact on water hyacinth Australian Centre for International Agricultural Research, populations. Selection of a good agent retains the ele- Canberra, Australia, pp. 33–38. ments of art, even as we improve the science (Hoelmer Hill, M.P., Cilliers, C.J. and Neser, S. (1999) Life history and and Kirk, 2005). laboratory host range of Eccritotarsus catarinensis (Car- valho) (: Miridae), a new potential natural enemy released on water hyacinth (Eichhornia crassipes References (Mart.) Solms-Laub.) (Pontederiaceae) in South Africa. Biological Control 14, 127–133. Ajuonu, O., Byrne, M., Hill, M., Neuenschwander, P. and Hill, M.P., Center, T.D., Stanley, J., Cordo, H.A., Coetzee, Korie, S. (2007) Survival of the mirid Eccritotarsus ca- J.A. and Byrne, M.J. (2000) The performance of the wa- tarinensis as influenced by Neochetina eichhorniae and ter hyacinth mirid, Eccritotarsus catarinensis, on water Neochetina bruchi feeding scars on leaves of water hya- hyacinth and pickerel weed: a comparison of laboratory cinth Eichhornia crassipes. BioControl 52, 193–200. and field results. In: Spencer, N.R. (ed)Proceedings of the Campbell, A., Frazer, B.D., Gilbert, N., Gutierrez, A.P. and Xth International Symposium on the Biological Control MacKauer, M. (1974) Temperature requirements of some of Weeds. Bozeman, MT, USA, 4–14 July 1999, Montana aphids and their parasites. Journal of Applied Ecology 11, State University, Bozeman, MT, USA, pp. 357–366. 431–438. Hoelmer, K.A. and Kirk, A.A. (2005) Selecting Coetzee, J.A., Byrne, M.J. and Hill, M.P. (2003) Failure of biological control agents against arthropod pests: can the Eccritotarsus catarinensis, a biological control agent of science be improved to decrease the risk of releasing inef- waterhyacinth, to persist on pickerelweed, a non-target fective agents? Biological Control 34, 255–264. host in South Africa, after forced establishment. Biologi- Julien, M.H. and Griffiths, M.W. (1998) Biological Control cal Control 28, 229–236. of Weeds. A World Catalogue of Agents and their Target Coetzee, J.A., Center, T.D., Byrne, M.J. and Hill, M.P. (2005) Weeds. Fourth Edition. CABI Publishing, Wallingford, The impact of Eccritotarsus catarinensis, a sap-feeding UK, 223 pp. mirid biocontrol agent, on the competitive performance of McClay, A.S. and Hughes, R.B. (1995) Effects of tempera- waterhyacinth. Biological Control 32, 90–96. ture on developmental rate, distribution, and establish- Coetzee, J.A., Byrne, M.J. and Hill, M.P. (2007a) Predicting ment of Calophasia lunula (Lepidoptera: Noctuidae), a the distribution of Eccritotarsus catarinensis, a natural biocontrol agent for Toadflax (Linaria spp.). Biological enemy released on water hyacinth in South Africa. Ento- Control 5, 368–377. mologia Experimentalis et Applicata 125, 237–247. Olckers T., Medal, J.C. and Gandolfo, D.E. (2002) Insect her- Coetzee, J.A., Byrne, M.J. and Hill, M.P. (2007b) Impact bivores associated with species of Solanum (Solanaceae) of nutrients and herbivory by Eccritotarsus catarinensis in northeastern Argentina and southeastern Paraguay, with on the biological control of water hyacinth, Eichhornia reference to biological control of weeds in South Africa crassipes. Aquatic Botany 86, 179–186. and the United States of America. Florida Entomologist Edwards, D. and Musil, C.J. (1975) Eichhornia crassipes in 85, 254–260. South Africa – a general review. Journal of the Limnologi- Sutherst, R. (2003) Prediction of species geographical ranges. cal Society of Southern Africa 1, 23–27. Journal of Biogeography 30, 1–12. Hill, M.P. (1999) What level of host specificity can we expect van Klinken, R.D. (2000) Host specificity testing: why we and what are we prepared to accept from new natural en- do it and how we can do it better. In: Spencer, N.R. (eds) emies for water hyacinth? The case of Eccritotarsus ca- Proceedings of the Xth International Symposium on the tarinensis in South Africa. In: Hill, M.P., Julien, M.H. and Biological Control of Weeds. Bozeman, MT, USA, 4–14 Center, T.D. (eds) Proceedings of the First IOBC Global July 1999, Montana State University, Bozeman, MT, Working Group Meeting for the Biological and Integrated USA, pp. 54–68. Control of Water Hyacinth, 16–19 November 1998, van Klinken, R.D., Fichera, G. and Cordo, H. (2003) Target- Zimbabwe, pp. 62–66. ing biological control across diverse landscapes: the re- Hill, M.P. and Cilliers, C.J. (1999) A review of the arthropod lease, establishment and early success of two insects on natural enemies, and factors that influence their efficacy, mesquite (Prosopis spp.) in Australian rangelands. Bio- in the biological control of water hyacinth, Eichhornia logical Control 26, 8–20. crassipes (Mart.) Solms-Laubach (Pontederiaceae), in Williamson, M. (1996) Biological Invasions. Chapman and South Africa. In: Olckers, T. and Hill, M.P. (eds) Biologi- Hall, London.

515