A Promising Agent for Nodding Thistle Is Found, Studied and Introduced

Agriculture and Agri-Food Canada (AAFC) in 1962 contracted CABI to survey all European thistle species to obtain field host range information on the insects on both the target and other introduced thistles (Zwölfer, 1965). The most promising for nodding thistle was the seed-head weevil, Rhinocyllus conicus (Curculionidae: Coleoptera) (Fig. 8.2). Although competitively inferior to three other insects, high fecundity and wide egg distribution (dispersion coefficient 0.7) result in attack of up to 98% of the heads. The next best was the seed-head gall fly, Urophora solstitialis (Tephritidae: Diptera) , which attacks up to 41% with a clumped distribution (dispersion coefficient 9.5) (Zwölfer, 1973).

Fig. 8.2. Rhinocyllus conicus adult.

CABI was then contracted to conduct host-specificity tests on a population of R. conicus at Mulhouse in the French Rhine valley. Single plant confinement showed the weevil nibbled on many plants; but normal feeding and egg development was confined to three closely related genera: Carduus, Cirsium and Silybum . The oviposition range was similar, although eggs were laid on dummy heads of cotton attached to Carduus leaves. Larvae from transferred eggs did not penetrate into Silybum heads although it is a field host (Zwölfer and Harris, 1984). Field studies found the weevil consisted of populations with host preferences that were overridden by confinement or high weevil density (Zwölfer and Preiss, 1983). The Canadian and US review committees approved release in time for the 1968 season. Subsequently, the insect was released throughout North America, as well as in other countries.

In Canada, both nodding thistle flowering and weevil oviposition are synchronized to a short early summer period. The result was seed reduction of about 50%. Thistle stand perpetuation depended on the germinating seeds covering the ground with rosettes before grass encroached. Rosette densities of around 100/m 2 then self-thinned to 20/m 2 flowering plants. With less seed, grass gradually returned, preventing seedling establishment. Two Saskatchewan sites with 92/m 2 and 179 seed/m 2 in 1969 declined to 1.5/m 2 in 1981 and 2.5 seed/ m 2 in 1982. The thistle now survives in pasture breaks, such as ground squirrel diggings. It is now rarely an agricultural problem but is still a nuisance in disturbed urban sites. The weevil is poorly synchronized with plumeless thistle, which has a later and longer flowering season; seed declined by only 15% and stands did not collapse. Similarly, nodding thistle was not controlled by introduction of the weevil in New Zealand and New South Wales, Australia, which have a long flowering season.

Releases of a Weevil against Plumeless Thistles Not a Success in Canada

A rosette weevil investigated as Trichcosirocalus horridus (Curculionidae: Coleoptera) by Virginia Polytechnic Institute and State University was released against plumeless thistles in 1975 in Virginia with stock from Italy and in Canada with stock from Germany. This practice was a bad one, as recent studies found T. horridus to be a complex of three species. Fortunately, those from Carduus all seem to be Trichcosirocalus mortadelo (D. Briese, 2001, personal communication). It has been ineffective in eastern Canada, where oviposition is restricted to about 3 weeks in early spring, but a small plumeless thistle infestation collapsed in British Columbia. The weevil has two generations in Virginia, and plumeless thistle in mixed stands collapsed in 7–13 years, about twice the time taken to control the nodding thistle (Kok and Mays, 1991). Stock from Canada sent to New Zealand, and subsequently Australia, controlled nodding thistle with a 72% seed reduction. This increased to 81% where R. conicus and a seed-head fly, Urophora solstitialis , are also present (Woodburn, 1997).

Another Agent Found, Studied and Released against Plumeless Thistle

AAFC contracted CABI to test the gall fly U. solstitialis , as in Germany it attacks 60–75% of plumeless thistle heads. Stock from Austrian nodding thistle was released in 1991. In a mixed thistle stand in Ontario, both the fly and the weevil preferred nodding thistle. Attack of plumeless thistle heads in the autumn was 5% by U. solstitialis and 6% by R. conicus . By 1998, both nodding thistle and R. conicus had disappeared and U. solstitialis increased to attack 58% of the heads. The impact on seedproduction should be greater than an equivalent attack by R. conicus , as the gall is a powerful metabolic sink, which sequesters resources from all parts of the plant.

A Rust is Also Introduced

The USDA screened a rust disease of nodding thistle, Puccinia cardorum (Uredinales: Fungus), and obtained permission in 1997 for a limited release in Virginia. Limited releases are a fiction as by 1998 it had spread to California and in 1999 to Nevada. Releasing it in Canada would offer little gain, but it will arrive anyway.

Lessons Learned

The normal end of a biocontrol project is that a single agent species controls the weed in a habitat, in this case represented by different climates. However, R. conicus has attacked native Cirsium spp. (Strong, 1977) beyond the expectation indicated by the tests of the Mulhouse population. Mulhouse stock did not establish on Cirsium pycnocephalus or Silybum marianum in California, but Italian collections from these thistles (Goeden and Ricker, 1977, 1985), as well as nodding thistle, have been established in the USA. Subsequent European allozyme studies showed that R. conicus separates into two groups with a genetic distance of 0.073 (Klein and Seitz, 1994). This is in the invertebrate subspecies range with weevils from S. marianum being in the taxon formerly known as Rhinocyllus oblongatus . There were also geographical differences within the two groups. Thus, genetically diverse stock has been established, but tests done only on those from Mulhouse.

Mulhouse stock in Saskatchewan is uncommon on Cirsium spp. unless with nodding thistle. It was not released in Alberta, where the wish was eradication with herbicides. The weevil in S. Alberta, apparently immigrated from Montana, which received both Mulhouse and central Italian stock, is presently common on the native Cirsium undulatum as well as the introduced Canada thistle. Both thistles remain abundant, and coincidentally with the weevil arrival, the rust disease, Puccinia punctiformis, became common on Canada thistle. Regulators no longer provide permits for the release of untested populations, and testing of new agents for thistle biocontrol in North America has ceased as researchers perceive that they will be rejected regardless of specificity.

The Ecological Risks and Benefits

Ecologically, the project has been beneficial, as range sites with pure thistle have returned to a species complex. A similar change following biocontrol has been documented for leafy spurge ( Euphorbia esula ) (Mico and Shay, 2002). Biocontrol reduced spurge canopy cover from 59% to 6%, increased plant diversity by 16 vascular plants, six of them natives, with native diversity increasing after 6 years. Spurge was displacing the northern prairie skink, the western spiderwort in its restricted Manitoba enclaves and the western prairie fringe orchid on the Sheyenne National Grasslands of North Dakota. Thus, both doing and not doing biocontrol may impact rare species. A US legal requirement that threatened or endangered natives should not be harmed is a sword over the reviewers’ heads, and their safest course is to deny release even though the weed itself is a greater ecological threat. Preferably, reviewers will weigh the risks and select the least ecologically damaging course, as required in Australia. Plant acceptance by an agent is not always detrimental, as in spite of attack by the beetle Aphthona nigriscutis (Chrysomelidae: Coleoptera), the scattered native spurge, E. robusta , has increased following leafy spurge decline.

The native thistle receiving most concern is Cirsium canescens , which has a flowering phenology similar to that of nodding thistle (Louda, 1988). It is widespread in the central Great Plains, where it forms dense stands in disturbed areas, despite a 76–91% destruction of seeds by native insects and pathogens and death of 75–87% of the seedlings, partly from cattle grazing and trampling (Lamp and McCarty, 1981). Most of R. conicus impact must be at the expense of the existing agents and is not added destruction. I am more concerned for Cirsium pitcheri , which is a threatened species confined to shoreline dunes along the western Great Lakes. It has the same enzyme loci as C. canescens , but a depauperate subset of alleles (Loveless and Hamrick, 1988). Possibly the greatest danger to C. pitcheri is hybridization with C. canescens.

Prediction of biocontrol risk relies on host range tests. North America still uses the no-choice test to demonstrate that the insect larva cannot mature on desirable species. The test worked when concern was limited to cropplants, but present concern for native relatives of the weed mean that 85% of the time the test results are broader than the field host range. Many of the remaining 15% may be species complexes, as in T. horridus , and need DNA analysis. The poor predictability arises because the tests are done on larvae. The most mobile insect stage is usually the adult, so it is up to the female to optimize her progeny survival by choosing appropriate plants. This trait is firmly held, with selection rewarding those getting it right. Externally feeding larvae can distinguish the host from interlaced vegetation, but not necessarily from close host relatives. New Zealand and Australia have changed their tests, but it is difficult in North America with two countries and many agencies whose representatives often have little insect ecology (see ‘no-choice tests’ and ‘host specificity’ on the website). Preferably Carduus biocontrol would have started with the release of U. solstitialis , but funds limit testing of several agents to select the most host specific.

This example shows that weed biocontrol with insects can have considerable ecological as well as economic benefits. Ecologically, there is no free lunch, as there will be changes, even if only those arising from replacement of the weed. Weed biocontrol is done by government in the public interest, which is more easily determined in economic than ecological terms. Indeed, it is not clear how to balance an endangered species threatened by the weed versus another by a biocontrol agent. Species of insects attacking weeds are not well known and may well be a species complex. Thus, it is essential to only release a tested population. Testing must involve the adult insect, which is usually responsible for field host selection. These are some of the technical problems, but there is also a political element, with different people having different interests and different fears.