Bottom-up control of water hyacinth weevil populations: do the plants regulate the insects?

Authors


*Correspondence author. E-mail: ted.center@ars.usda.gov

Summary

1. Nutrient availability varies widely in aquatic systems and is likely to profoundly affect the outcomes of plant–herbivore interactions in aquatic environments. This has implications in programmes aimed at the biological control of water weeds. We hypothesized that nutrient flux seasonally affects ovarian development of two specialist weevil species, Neochetina eichhorniae and Neochetina bruchi, thereby influencing population growth on the floating plant Eichhornia crassipes.

2. We also hypothesized that the weevils differed in their sensitivities to nutritional quality and interacted differently depending on host quality.

3. To test these hypotheses, we cultured E. crassipes plants under 20 fertilizer regimens then introduced reproductively dormant, over-wintering female weevils as single or mixed species populations. They were later retrieved and dissected to ascertain ovarian status. F1 progeny was extracted biweekly to compare fecundity with population growth.

4. Ovaries regained functionality and the reproductive proportion of the population increased with fertilizer rate as did the fecundity of reproductive individuals. F1 progeny increased exponentially from 7 to over 300 weevils tank−1 in the highest fertilizer treatment.

5. Mixed species treatments produced more offspring under high fertilizer rates than single species treatments (236 ± 32 mixed vs. 155 ± 25 N. bruchi or 170 ± 33 N. eichhorniae pure). Neochetina eichhorniae was more productive throughout than N. bruchi, but N. bruchi performed nearly as well in high fertilizer treatments.

6. Ratios of the two species remained consistent whether in single or mixed species populations suggesting a lack of interspecific interference. However, N. eichhorniae seemed more adaptable to a wider range of plant quality and more tolerant of lower nutritional circumstances.

7.Synthesis and applications. The coupling of reproductive status of the parental generation with the population response of the F1 generation indicates that plant quality drives population growth of both weevil species, demonstrating bottom-up regulation. This largely explains the variable performance of these weevils as biological control agents. These results emphasize the importance of considering bottom-up regulation in evaluating host range trials, predicting efficacy and in post-release evaluations. In addition, they suggest that partial herbicide treatments to accomplish phased removal of infestations could enhance the quality of the remaining plants, thereby better integrating biological control into current E. crassipes management systems.

Introduction

The acquisition of adequate nourishment from plant tissue constitutes a major impediment to phytophagy (Strong, Lawton & Southwood 1984). Plant-feeding insects require dietary nitrogen in the form of protein for growth and reproduction, which they derive from tissue composed primarily of carbohydrate (Strong, Lawton & Southwood 1984). The amount of nitrogen in this tissue can be strongly influenced by nutrient flux in the habitat where the host plant occurs (Mattson 1980). This is especially true in aquatic systems, where nutrient availability is seasonally variable or occurs in pulses (Room & Gill 1985; Room & Thomas 1986; Taylor 1989; Carignan, Neiff & Planas 1994; Xie et al. 2004). Specialized plant-feeders on aquatic plants must therefore adjust behaviourally or physiologically to often unpredictable alterations in host quality.

Biological control utilizes specialized insect herbivores whose diets are usually limited to a single host species (Harley & Forno 1992; Center, Dray & Frank 1997; Littlefield & Buckingham 2004). These insects, being unable to switch to alternative hosts should the primary host fail to meet nutritional needs, are at the mercy of environmentally induced host quality differences (Room & Thomas 1985; White 1993). In the aquatic environment, these differences can occur seasonally or result from stochastic events such as drought, flooding or pollution from fertilizer runoff. Thus, we reasoned that bottom-up regulation, i.e. constraints caused by limitations in resource quality (Gratton & Denno 2003; Walker & Jones 2003; Richmond et al. 2004), should influence the population dynamics of two biological control agents, the tropical, semi-aquatic weevils Neochetina eichhorniae Warner and N. bruchi Hustache, on their host plant Eichhornia crassipes (Mart.) Solms (water hyacinth).

Eichhornia crassipes is a nitrophilous floating aquatic species (Imaoka & Teranishi 1988; Reddy, Agami & Tucker 1989) native to South America, but widely naturalized in tropical and subtropical regions. Like many aquatic species (Wheeler, Van & Center 1998) the nutritional quality of the plant tissue is poor. Tissue N concentrations during autumn in the southeastern USA vary extensively ranging from 1·8% to 5·0% of dry mass (Center et al. 1999). The leaf tissue consists of 89–94% water (Penfound & Earle 1948) so fresh leaf tissue is typically <0·5% N. The plant tissue becomes more succulent as fertility of the growth medium increases and more fibrous as it decreases (Tucker & DeBusk 1981a; Tucker 1981). Thus, a plant feeder ingests more water at the one extreme and more carbohydrates at the other. The plant grows rapidly in response to high fertility becoming taller with more structural tissue as density increases (Center & Spencer 1981), further diluting nutrients within a large bulk of emergent biomass (Tucker & DeBusk 1981b). As a result, even though the total leaf nitrogen may not change, tissue concentrations may decrease as the leaf area expands (Center & Van 1989).

Neochetina eichhorniae and N. bruchi feed and reproduce only on E. crassipes (DeLoach 1975, 1976; DeLoach & Cordo 1976). Eggs are inserted in leaf tissue and larvae mine the thick leaf petioles (DeLoach & Cordo 1976; Center 1994). Reproduction is reduced in the southeastern USA during late autumn when the ovaries of most females degenerate (Grodowitz, Center & Freedman 1997). However, N content of the plant tissue increases as plant growth slows, and peaks as the plants recover from winter (Tucker & DeBusk 1983; Center 1994). The increased N supply during this period (e.g. Center & Spencer 1981) may be an important event governing the dynamics of the insect populations. This could be a crucial point during which bottom-up regulation of the weevil populations occurs and it may determine population levels attained during the ensuing growing season. It is unclear whether the individuals present at this time have overwintered as adults and have regained reproductive capacity or have emerged from overwintered pupae, but bottom-up regulation during this period may explain the inconsistent levels of biological control produced (Spencer & Ksander 2004).

Grodowitz et al. (1997) reported that Neochetina spp. females with functioning ovaries comprised <10% of December populations, but this proportion rapidly increased during late winter into spring. We suspected that a portion of this increase represented females from the previous season whose ovaries regained functionality as nutrition improved, although this has never been demonstrated. We therefore hypothesized that non-reproductive females would become reproductive, and that this transformation would be triggered by some threshold level of leaf tissue N. We also hypothesized, based on past observations (Center & Dray 1992; Center et al. 1999), that plant quality differences would affect reproduction of the two species and that they would respond differently – with N. bruchi being more responsive to plant quality (Heard & Winterton 2000). To address these hypotheses we collected post-reproductive female weevils during autumn and placed them onto E. crassipes cultivated under a range of fertilizer regimens. Individuals retrieved during late winter were dissected to assess reproductive status. Comparison of the F1 generation in pure or mixed colonies of the two species allowed us to test for bottom-up regulation of the weevil populations and, thus, the relationship between reproductive potential and population growth.

Materials and methods

Eichhornia crassipes plants collected locally were propagated in concrete mesocosm tanks (0·8 m wide × 2·2 m long × 0·65 m deep, water depth 0·5 m, vol. 0·88 m3). The stock tanks, filled with water from the source described by Van & Steward (1986), were fertilized with 150 g Scotts® Osmocote Plus 15-9-12 N:P:K, Southern 8- to 9-month formulation (Scotts Professional, Marysville, Ohio, USA) and 18 g Miller® Iron Chelate DP 10% Fe (Miller Chemical & Fertilizer, Hanover, Pennsylvania, USA).

Eichhornia crassipes rosettes, selected to be of similar form and stature, were cleaned of debris, attached litter and developing offshoots. Thirty rosettes were randomly distributed to each of 60 tanks during 28–30 July 2004. A 1-m tall cage (7·0 × 5·5 strands cm−1 mesh screen) was placed over each tank. These cultures were thinned to 15 plants tank−1.

Fertilizer enclosed in screen packets suspended within the root zone of the plants was supplied at 20 treatment levels. Each level of fertilizer was applied to three tanks at rates of 30–600 g tank−1 in 30 g increments on 10 August 2004. Iron (18 g Miller® 10% Iron Chelate DP) was added initially and to all tanks whenever deficiency symptoms appeared in any of the tanks.

Plant tissue analyses

Three plants were withdrawn from each tank on 10 December 2004, a few days prior to introducing the weevils, to determine tissue N levels, thus leaving 12 plants per tank. They were cleaned as before, separated into submerged and emergent components, weighed and dried at 50 °C to constant weight. The emergent portion was ground in a Willey mill to pass a 40-mesh screen. Samples were analysed with a C-H-N analyzer (Perkin-Elmer® Series II CHNS/O Analyzer Model 2400; Perkin-Elmer Corporation, Waltham, Massachusetts, USA) and compared against tomato leaf standards to determine % N and C:N ratio. Percentage N was converted to a fresh weight basis using the dry to fresh weight ratio.

Weevils and weevil treatments

Post-reproductive overwintering female weevils were collected locally. Dissection of 31 N. eichhorniae and 24 N. bruchi revealed that all possessed degenerated ovaries and sperm-filled spermathecae. The pronota of each weevil was marked using a fine-tipped paint pen. Limited availability of N. bruchi necessitated making releases in three installments (15 December 2004 to 21 January 2005) to attain infestations of 12 weevils tank−1 (1 female/plant). One tank of the three representing each of the 20 fertilizer levels received N. eichhorniae females only, one received N. bruchi only and a third received an equal mix of the two species.

Marked females were retrieved during 14–17 February 2005 by submersing the plants and collecting the weevils as they surfaced. They were held on ice while awaiting dissection. The reproductive status of each was assessed as per Grodowitz et al. (1997). Females with healthy, functioning ovaries were considered ‘reproductive’. In these cases, the proximal follicles contained fully formed oocytes followed distally by a sequence of progressively less mature oocytes. The total numbers of follicles in all four ovarioles were counted.

Egg load was based on ovulated eggs present in the oviducts. Excessive holding without an opportunity to oviposit can cause egg retention, so we were careful to ensure that holding times were distributed evenly across treatments.

The extraction procedure was repeated to capture the F1 adults. Extractions were carried out biweekly when the F1 generation first appeared (16 March 2005), and continuing until 20 June 2005 when the F2 generation began to emerge. These were removed and tallied by species and gender.

Resource quantity

It was necessary to non-destructively estimate the amount of plant biomass available to support the F1 populations in order to partition the effects of resource quantity vs. quality on weevil abundance. We reasoned that the above-water volume of plant material should estimate resource availability. Plants were therefore gently pushed to one end to produce full coverage over a portion of each tank; this length was multiplied by tank width to determine area then multiplied by modal canopy height (measured at several points), to determine emergent plant volume. This was performed twice, once after emergence of the F1 generation (23–27 May 2005), and 7 weeks later (12 July 2005) when the plants were harvested. The later data established an allometric relationship between emergent biomass and plant volume used to estimate biomass for the earlier date. Plants were harvested and cleaned of debris and attached litter. The living material was separated into emergent and submerged components, weighed fresh, then dried and reweighed.

Data analyses

Descriptive statistics and standard data analyses (anova, t-tests, etc.) were performed using sigmastat (v. 3., Systat 2006) and sas (v. 9.1; SAS Institute 2004). Data were tested for normality and homogeneity of variance and then transformed as needed prior to analysis. Average values are presented as arithmetic means with standard errors. Graphical analyses were carried out using sigmaplot (v. 11, Systat 2008). Lines were fitted to data using the SigmaPlot regression utility.

Two-way analyses of variance with interaction were used to determine whether weevil species (N. bruchi or N. eichhorniae) or fertilizer level (as pooled groupings, see Fig. 1) explained variation in response variables (ovarian follicles and ovulated eggs per female). The makeup of the weevil population (as pure or mixed species), initially included, was found to be non-significant and dropped. Analyses were performed using the General Linear Model procedure (proc glm) available in SAS (Littell, Stroup & Freund 2002). Means were separated using the Ryan–Einot–Gabriel–Welsch multiple range test. Data were examined for underlying assumptions and found to be within tolerances described by Zar (1999). All analyses were based on means per tank to avoid pseudoreplication (Hurlbert 1984).

Figure 1.

 Nitrogen concentrations of emergent tissue of Eichhornia crassipes plants grown under varying fertilizer levels. The top graph shows N as a percentage of dry weight (inline image ± SE); the bottom graph as a percentage of fresh weight. The inset depicts the relationship between tissue N concentrations and % dry weight. The ellipses in the bottom graph show the groupings used to pool data into low, medium, or high fertilizer categories. Values represent means and standard errors. Fertilizer groupings are coded as low (white), medium (grey) and high (black).

Loglinear hierarchical categorical data analyses were used to analyse frequency data consisting of total counts of marked female weevils retrieved from the tanks as well as counts of those with functioning ovaries. Frequency data for each fertilizer level were arranged as contingency tables with cells containing counts of weevils classified according to fertilizer treatment (20 levels), weevil species or population makeup. Many cells contained too few observations so the tables were collapsed to combine the fertilizer treatments into low, medium or high groupings based on fresh weight tissue-N concentrations (see Fig. 1). These data were analysed using the sas v. 9.1 catmod procedure employing hierarchical loglinear analyses for multi-way contingency tables of categorical data in which the primary interest is the presence of interactions (Sokal & Rohlf 1995; Gotelli & Ellison 2004). All four-way interactions were analysed first which included ovary status (functional vs. degenerated), fertilizer levels (high, medium or low), weevil species (N. bruchi or N. eichhorniae) and population makeup. Four-way interactions that failed tests of significance (Maximum Likelihood Estimates and Wald chi-square) were deleted and all three-way interactions were compared in a similar manner. This iterative process continued so as to maximize the value of the joint multinomial likelihood function.

Average cumulative numbers of weevils produced in the F1 generation were compared among three pooled fertilizer groupings, between species and between population makeup types within species, using heterogeneity G-tests that produce anova-like tables (Sokal & Rohlf 1995) to partition main effects from interactions.

Results

Plant quality

Tissue N levels induced in the plants ranged from about 1·5% to 4·4%. Tissue-N concentrations increased with increasing fertilizer until reaching maximal values at fertilizer levels above 400 g tank−1 (Fig. 1, top). The succulence of the tissue (the inverse of % dry weight) increased (Fig. 1, inset) as tissue-N concentrations increased.

Based on fresh weight (Fig. 1, bottom) data segregated into three significantly different groups (one-way anova, F = 73·0, < 0·001) with fertilizer levels of <210 g tank−1 producing similar low concentrations (0·19 ± 0·01%), 210–330 g tank−1 producing similar intermediate concentrations (0·25 ± 0·02%) and levels >330 g tank−1 producing similar high levels (0·31 ± 0·02%).

Plant abundance

The E. crassipes plants grew in size and number during the latter part of the study so increased resource quantity potentially affected weevil population growth. Fertilizer level (X, g tank−1) proved to be a predictor of plant volume (inline image, m3 tank−1) as: inline image; r=0·894, < 0·001. Plant volume (X), in turn, proved to be a reasonable predictor of emergent biomass (inline image, g tank−1) as: inline image; r= 0·95, <0·001.

Status of the released weevils upon recovery

The numbers of weevils recovered were independent of weevil species, population makeup or fertilizer levels (Fig. 2, top; Table 1) based on the maximum likelihood analysis of variance (Proc catmod; SAS Institute Inc. 2004). No four- or three-way interactions were significant so they were excluded. Population makeup had no effect so only 2 two-way interactions remained: reproductive status by fertilizer treatment and reproductive status by weevil species. The fertilizer interaction was much more important but about 10% more N. bruchi than N. eichhorniae possessed functioning ovaries across all fertilizer levels (Fig. 2, bottom).

Figure 2.

 Loglinear hierarchical analysis for the total number of female weevils recovered (cross-hatched bars) from the six low-, five medium- and nine high-fertilizer treatments and the portion with functioning ovaries (solid bars, top graph). The bottom graph shows the number of reproductive females as a percentage of the total.

Table 1.   Maximum Likelihood Analysis of Variance (sas catmod procedure) testing the effects of fertilizer level (high, medium or low), weevil species (Neochetina bruchi vs. N. eichhorniae) and colony type (single vs. mixed species) on the number of marked female weevils recovered from mesocosm tanks
Sourced.f.Chi-square> Chi-square
Fertilizer (F)20·190·9112
Species (S)10·430·5104
F × S20·050·9744
Colony type (C)10·000·9908
F × C20·030·9856
S × C10·030·8536
Likelihood ratio20·030·9847

No weevils from the lowest fertilizer treatments (30 g tank−1) were reproductive when recovered. Weevil species did not interact with fertilizer treatments (F = 0·04, = 0·9560) so the interaction term was dropped and the data reanalysed. The overall anova explained a significant amount of the total variation (F = 11·3, < 0·0001, 3 and 36 d.f.). Follicle counts in functioning ovaries were influenced primarily by fertilizer level and to a lesser degree by weevil species (Fig. 3a,b).

Figure 3.

 Analyses of variance comparing fecundity in terms of follicular development (a) between species, Neochetina bruchi (NB) and Neochetina eichhorniae (NE); (b) among pooled fertilizer treatments and (c) comparing egg load among fertilizer treatments. Bars represent means with standard errors. Those labelled with the same letter did not differ at = 0·05.

None of the female weevils in the lowest fertilizer treatment (30 g tank−1) contained eggs when recovered. Otherwise, egg load increased with fertilizer level (Fig. 3c), which was the only significant treatment effect being independent of weevil species or population makeup. There was no species difference (F = 0·10, = 0·1453) and no interaction effect (F = 0·20, = 0·7936) (Fig. 2b), so these components were deleted and the data reanalysed for fertilizer effect alone (Fig. 3b).

Weevil populations

Weevil populations increased as fertilizer level increased (Fig. 4), but plant volume also increased. The increases in plant abundance, having occurred after the parental females had been removed, would not have influenced oviposition but may have affected larval survival. Resource quality, accounted for 67% of the variability in weevil populations (F = 119·48, < 0·001). However, quality and quantity (i.e. plant volume) were correlated, confounding factors (r = 0·801, < 0·001). A multiple regression analysis including these two non-orthogonal factors had greater predictive power, by about 10%, than the model using quality alone (F = 52·15, < 0·0001). Quality had a substantial influence (fertilizer level type III F = 5·86, = 0·019) on weevil abundance after adjusting for quantity, but the reverse was not true (plant volume type III F = 0·55, = 0·461). The two factors could not be completely parsed, however, and the interaction was significant (type III F = 5·63, = 0·021). This suggests that the plant quality effect on the weevil population changed depending on the amount of plant material available, so resource availability may have had some post-reproductive influence, possibly affecting larval survival.

Figure 4.

 The average number of weevils (both species and all culture types combined) and the amount of plant material produced per tank compared with fertilizer levels.

Figure 5 shows cumulative numbers of weevils extracted from single vs. mixed species populations. In all cases, N. eichhorniae lagged behind N. bruchi. The summed totals of the two species produced in mixed populations exceeded either pure population in high fertilizer treatments (G = 16·90, < 0·001 and = 10·78, < 0·001 for N. bruchi and N. eichhorniae respectively), whereas under low fertilizer conditions mixed populations were more productive than N. bruchi (G = 5·23, = 0·023) but not N. eichhorniae (G = 0·33, = 0·521) alone. Productivity of N. bruchi differed by fertilizer level (G = 11·70, = 0·004), with mixed populations more productive than pure populations at high fertilizer levels but less productive at intermediate levels. Neochetina bruchi was unproductive (only seven individuals total) at low fertilizer levels regardless of population composition. Also, N. eichhorniae out produced N. bruchi regardless of fertilizer level in both pure (244 vs. 204 individuals; G = 3·57, = 0·062) and mixed (173 vs. 126 individuals; G = 7·42, = 0·007) populations. The ratio of the two species remained constant regardless of whether in mixed or pure populations (F = 0·87, = 0·358). However, ratios were affected by fertilizer level wherein N. eichhorniae was more highly represented at low fertilizer levels than N. bruchi. This is shown in Fig. 6 by the significant difference of the experimental slopes from a hypothetical 1:1 ratio (F = 26·54, < 0·0001). Neochetina eichhorniae was more productive in mixed colonies than pure colonies (G = 21·10, < 0·001) at all fertilizer levels, considering that only half as many were released (i.e. counts in mixed populations were doubled for comparison with pure populations).

Figure 5.

 The number of F1 weevils from each population from Eichhornia crassipes plants cultivated in high, medium or low fertilizer levels. Half as many of either species were released in the mixed populations, so counts should be doubled for comparison with the pure cultures. Also note the difference in Y-axis scaling for each fertilizer level.

Figure 6.

 The relationship between the numbers of Neochetina bruchi and Neochetina eichhorniae produced per tank was independent of whether from mixed (circles, solid line) or pure (triangles, dashed line) populations. Half the number of either species was released in the mixed parental populations so counts were doubled for comparison with the pure populations. Fertilizer levels are coded as low (white), medium (grey) and high (black). Deviations from equal representation of the two species (dotted line) were greater and favoured N. eichhorniae at low fertilizer levels.

Discussion

Akbay, Howell & Wooten (1991) presented a temperature-based simulation model for N. eichhorniae and N. bruchi populations using field data from Florida sites. The simulation fit the actual field data for N. eichhorniae adults fairly well, but not data for adult N. bruchi or for larval populations of either species. We contend that the poor performance of the model was due to the lack of consideration for bottom-up effects. Temperature is not the primary force driving populations of these tropical species and host quality must be taken into consideration, especially for N. bruchi.

Our findings accord well with field observations in which the variable reproductive status of Neochetina populations corresponded with nutritional quality of plants resident at 54 Florida sites during autumn (Center et al. 1999). They are also consistent with the findings of Heard & Winterton (2000) who found that plants grown at higher nutrient concentrations were superior hosts for N. bruchi. The weevils must have the capacity to produce populations of sufficient size to damage and suppress E. crassipes populations to be effective biological control agents; but population growth depends on reproductive ability. The nutritional value of the plant tissue profoundly affected this ability, with N. bruchi more sensitive to plant quality than N. eichhorniae. Potential for population growth, reflected as the percentage of reproductive females in the population as well as by individual fecundity, was dependent upon plant quality but the two species exhibited relatively minor differences in fecundity. Neochetina bruchi populations responded more quickly to good quality, whereas N. eichhorniae thrived across a wider range of nutritional conditions. The dissimilar performance of the two species in terms of population development may be attributable to plant growth and density-dependent larval survival (DeLoach & Cordo 1976; Wilson, Rees & Ajuonu 2006) rather than fecundity inasmuch as differences in egg load and follicular development were slight. We have observed N. bruchi larvae feeding in axillary buds (Center 1994). The requirement for this highly nutritious tissue may explain their greater sensitivity to plant quality. The combined feeding of the weevils may promote lateral bud formation, which would provide additional resources for N. bruchi in mixed populations. This would explain why mixed populations produced greater numbers than either pure population in high nutrients.

A specific threshold level of tissue-N was not apparent inasmuch as the reproductive response seemed to vary directly with fertilizer rate, but none of the females from the lowest fertilizer treatment (30 g tank−1) were reproductive when retrieved. This corresponds to tissue-N levels of about 1·7% which could be construed as a minimum threshold. Another way of viewing this is in terms of the tissue levels needed to achieve replacement of the parental populations; 24 weevils species−1 tank−1 in pure populations or 12 weevils species−1 tank−1 in mixed populations, assuming 1 : 1 sex ratios. Minimal tissue N concentrations needed for positive population growth would then be about 3·0% for N. bruchi (270–300 g tank−1) but only about 2·1% (120–150 g tank−1) for N. eichhorniae.

Understanding these dynamics in the field requires knowledge of nutrient flux in the aquatic environment and the associated changes that occur in the nutritional status of E. crassipes stands. However, data pertaining to this relationship are scant. Tucker & DeBusk (1981b, 1983) demonstrated that N content of E. crassipes plants cultivated in a constant nutrient supply was elevated during winter and spring in south Florida, attained maximal values during February (3·7% N), and was minimal during summer and autumn (2·1–2·2% N). In contrast, plant growth rates were lowest during winter and maximal during summer. Boyd & Blackburn (1970) showed that N concentrations ranged from 3·8% during spring to about 2·5% during summer, but they did not sample during the winter months. These seasonal maximal and minimal N levels correspond to our high and low quality plants respectively. Center (1994) compared seasonal tissue N levels between sites with low quality or high quality plants and showed a similar seasonal pattern in both but with consistently lower N concentrations at the low quality site. Egg populations tracked N levels, with maximal numbers in March and April. Grodowitz et al. (1997) found that in southern Texas the percentage of weevils with functioning ovaries was low during late autumn and early winter then increased dramatically during February and remained high until summer. The number of ovarian follicles was minimal during winter then tripled by February and attained maximal levels during spring before rapidly decreasing. The accumulation of these observations, when coupled with the present study, suggests that the reproductive biology of N. eichhorniae and N. bruchi is governed from the bottom up by variation in plant quality. The possibly reinforcing role of abiotic factors, such as photoperiod or temperature, is unknown but nutritive quality of the plant tissue is clearly a primary factor.

The present study along with published data on the seasonal variability in plant nutrient status allows us to envision a conceptual model of the Neochetina–Eichhornia interaction. The plants produce large amounts of biomass during the summer months causing competition for space and nutrients, thereby depleting nitrogen from the water column. The leaves form elongate petioles to compete for light, which requires additional structure, causing the C:N ratio to increase as the leaf nitrogen becomes dispersed within more biomass. Neochetina populations, unable to derive adequate nutrition, respond by losing reproductive capacity as ovaries degenerate. Plant growth slows as solar flux wanes and the senescing summer leaves are replaced by progressively smaller leaves (Center & Spencer 1981). All emergent tissue can be killed in areas subjected to frost, which further opens the canopy and reduces competition for light. Decomposition of this dead tissue slows during winter but accelerates as temperatures rise; the decomposing tissue thus nutrifying the water at the end of winter. Submerged roots are less affected by cold than emergent shoots so the root : shoot ratio increases (Center & Spencer 1981) allowing rapid absorption of the released nitrogen thus producing higher tissue-N levels and initiating new growth. The weevils synchronize with this burst of plant growth and respond with renewed reproductive activity; but N. bruchi responds more quickly than N. eichhorniae and appears earlier. As plants grow and competition increases the tissue-N concentrations decrease and weevil reproduction slows. Due to their greater sensitivity to N-levels, N. bruchi populations are reduced more than N. eichhorniae and may disappear altogether during the summer at nutrient-poor sites.

This conceptual model is overly simplified pertaining to dense stands where nutrient cycles are not overly influenced by external inputs. Aquatic habitats, however, are heterogeneous and dynamic and many stochastic factors change nutrient availability, which can be extremely unstable often occurring in pulses (Room & Gill 1985; Room & Thomas 1986; Carignan, Neiff & Planas 1994). These rapid changes can induce rapid changes in the nutritional value of aquatic plant tissue (Taylor 1989) and the associated Neochetina populations, so the pattern of ovarian maturation is unlikely to be consistent across habitats. Nonetheless, N-limitation (and thus bottom-up regulation) should be apparent in aquatic systems, especially within populations of specialist herbivores that feed on floating aquatic plants (Room & Thomas 1985).

From an applied standpoint, the important aspect of this relationship is the ability of the weevils to control the plant. Obviously, the stimulus provided by increased nitrogen can cause weevils populations to increase. However, it also stimulates plant growth that can compensate for damage caused by the insects. Examination requires interruption of this feedback loop, i.e. by studying the effect of the plant on the insect and then independently studying the effect of the insect on the plant. We have accomplished the first part of this in the present study. Research is continuing on the second part.

Recommendations

Eichhornia crassipes remains a major world-wide problem, especially in poor countries which cannot afford perpetual herbicidal control programmes, despite the broad dissemination of biological control agents throughout its adventive range. Nonetheless, biological control has produced relief in many areas, especially in the Old World tropics. The unpredictability of success has led many to question the efficacy of N. bruchi and N. eichhorniae. Sceptics frequently promote alternative explanations for observed declines (Williams, Duthie & Hecky 2005; Williams, Hecky & Duthie 2007), even in cases where remarkable control has resulted (e.g. Lake Victoria; Wilson et al. 2007). Our data suggest that uniform results should not be expected inasmuch as the potential for weevil population growth will vary widely depending on host plant quality. The nature of this variation is not yet clear; plants already under nutrient stress may succumb to low levels of herbivory more readily, or plants growing in high nutrients may be better able to survive intense herbivory. Furthermore, herbivory can alter plant quality (Heard & Winterton 2000) which could impinge upon the reproductive ability of the herbivores. Nonetheless, assessment of the nutritional value of the plant can improve predictability in biological control programmes.

Recently, there has been a call for increased pre-release evaluation of the efficacy of weed biological control agents, so that ineffective agents are not introduced into the environment (McClay & Balciunas 2005). The fear is that such agents could increase in abundance causing adverse food web effects without reducing the weed populations. Such pre-release studies should be approached with caution as variation in food quality can affect conclusions in profound ways. Erroneous predictions could cause rejection of effective agents capable of alleviating devastating environmental effects of invasive plants. They could also lead to acceptance of ineffective agents if the testing scheme, utilizing high quality plants, did not reflect natural conditions. For example, DeLoach & Cordo (1976) predicted that N. bruchi would be more effective in Florida than N. eichhorniae because, in Argentina, it seemed more tolerant of cooler temperatures, became reproductive earlier in spring, had a greater intrinsic rate of increase, was generally more abundant at field sites, and destroyed laboratory colonies of E. crassipes more quickly than N. eichhorniae (DeLoach & Cordo 1976). Contrary to this prediction, N. eichhorniae predominates in Florida (Center & Dray 1992; Center et al. 1999).

Our results also have implications for integrated management schemes and suggest that improved nutritional quality of the host plant could lead to more effective biological control. Earlier studies (Center et al. 1999) indicated that plants in areas subjected to herbicidal control tend to be higher quality, presumably due to reduced competition and release of nutrients from decomposing, herbicide-treated plants. While applications of nitrogen-rich fertilizer would not be advisable, as was carried out to improve biological control of Salvinia molesta Mitchell in New Guinea (Room & Thomas 1985), partial herbicide treatment to accomplish removal of infestations gradually or in phases could enhance the quality of the remaining plants (Thomas & Room 1986) and provide harborage for insects escaping from treated areas (Haag, Glenn & Jordan 1988).

Acknowledgements

We thank the following for many useful and constructive comments on early drafts of the manuscript: Rieks van Klinken, Martin Hill, Julie Coetzee, Phil Tipping, Ray Carruthers and Mike Grodowitz. We also thank Judy Myers and Peter McEvoy for sharing fruitful ideas and stimulating discussion. We gratefully appreciate the efforts of Willey Durden who oversaw and coordinated technical aspects of the experiments described herein. Mention of trade names or commercial products in this publication is solely for the purpose of providing specific information and does not imply recommendation or endorsement by the U.S. Department of Agriculture.

Ancillary