Companion planting with French marigolds protects tomato plants from glasshouse whiteflies through the emission of airborne limonene

Abstract

Horticulturalists and gardeners in temperate regions often claim that planting marigolds next to tomato plants protects the tomatoes from the glasshouse whitefly (Trialeurodes vaporariorum Westwood). If shown to hold true, this technique could be used in larger-scale tomato production, protecting the crop and helping to introduce greater plant diversity into these agro-ecosystems. Here we present two large-scale glasshouse trials corresponding to the two main ways growers are likely to use marigolds to control whiteflies. In the first, marigolds are grown next to tomato throughout the growing period and we quantify whitefly population growth from the seedling stage over a 48 day infestation period. Here we show that association with marigolds significantly slows whitefly population development. Introducing additional whitefly-attractive 'pull' plants around the perimeter of plots has little effect, but reducing the proportion of marigolds and introducing other non-hosts of whiteflies (basil, nasturtium and Chinese cabbage) also reduces whitefly populations on tomato. The second experiment assesses the efficacy of marigolds when used as an 'emergency' measure. Here we allow whitefly populations to build to a high density on unprotected tomatoes then introduce marigolds and assess whitefly population over a further period. Following laboratory work showing limonene to be a major chemical component of French marigolds and a negative behaviour response of whiteflies to this compound, limonene dispensers are added as an additional treatment to this experiment. "Emergency" marigold companion planting yielded minimal reductions in whitefly performance, but the use of limonene dispensers was more effective. Our work indicates that companion planting short vine tomatoes with French marigolds throughout the growing season will slow development of whitefly populations. Introducing marigolds to unprotected tomatoes after significant whitefly build-up will be less effective. The use of limonene dispensers placed near to tomato plants also shows promise. It is argued that this work supports the possibility of the development of a mixture of tomato companion plants that infer 'associational resistance' against many major invertebrate pests of tomato. Such a mixture, if comprising edible or ornamental plants, would be economically viable, would reduce the need for additional chemical and biological control, and, if used outdoors, would generate plant-diverse agro-ecosystems that are better able to harbour invertebrate wildlife.

Fig 1. Population development of the whitefly, T. vaporariorum on tomato in the glasshouse, with all whitefly life stages (adult, eggs and nymphs) contributing to the average number of whitefly/leaf (n = 8).

Day 0 is 10th August 2016 and the ‘push-pull’ was started 12 days after the “push” experiment. Whitefly abundance data from Fig 1A and 1B was (log +1) transformed and analysed over time with repeated measures ANOVA’s and Bonferroni corrected post-hoc comparisons, in which the p values were multiplied by the total number of comparisons (n = 3). Significant observations between treatments at individual sampling points are annotated onto the graphs with asterisks, “*” indicates a p value of <0.05, “**” p <0.001. All other p values were non-significant at α = 0.05. Fig 1A Shows the “push” experiment in which repellent plants such as marigold (LD) and marigold plus non-hosts (HD) are intercropped amongst tomato plants. There was a significant effect of the treatment (rm ANOVA F _(2,126) = 18.85, p < 0.001) and treatment x time (rm ANOVA F _(10,126) = 2.14, p = 0.025) across the sampling period. Fig 1B Shows the ‘push-pull’ experiment which is the same as the ‘push’ experiment but additionally a single (LD) and several (HD) host plant species surround the repellent hosts and tomato. There was a significant effect of treatment (F _(2,105) = 3.73, p = 0.027) but no significant effect of treatment x time (F _(8,105) = 1.49, p = 0.166) following repeated measures ANOVA’s. Post-hoc comparisons from Fig 1A showed there were significantly less whiteflies in LD and HD plots at day 34 (LD, t = 2.89, df = 126, p = 0.013. HD, t = 3.61, df = 126, p = 0.001), day 43 (LD, t = 3.59, df = 126, p = 0.001. HD, t = 3.92, df = 126, p < 0.001) and day 48 (LD, t = 3.66, df = 126, p = 0.001. HD, t = 3.29, df = 126, p = 0.003). For Fig 1B, there were significantly less whiteflies on the HD treatment at day 43 (t = 3.018, df = 105, p = 0.009) but not on the LD treatment (t = 2.20, df = 105, p = 0.088). Fig 1C shows the data in Fig 1A and 1B expressed as effect size (relative to control). The Cliff’s d measure is used as this is suitable for non-normal data of the type observed in experiments. Values of 1 or -1 (the sign shows the direction of effects relative to control) indicate complete non-overlap between the groups under consideration and a value of 0 indicates complete overlap. Ninety five percent confidence intervals have been calculated and are available in the supporting information (S1 Dataset), but to aid visualisation they have been removed from the figure.

Fig 2

The percentage of whiteflies on tomato (n = 200) when given the choice between four un-accompanied “Elegance” tomato seedlings or four “Elegance” seedlings intercropped with either flowering marigold plants (Fig 2A) or limonene dispensers (Fig 2B). Whitefly eggs were also recorded for each replicate (Fig 2C and 2D). Values were transformed to percentages and the Pearson’s chi-squared test was used to test differences in settling and oviposition behaviour. Significant differences are annotated onto the graph for each treatment with the following format; “*” indicates a p value of <0.05, “**” p <0.001. Comparisons between individual replicates and controls for all four graphs (Fig 2A, 2B, 2C and 2D) are as follows; df = 1 for all reps, Fig 2A, rep 1 X ² = 21.26, p = <0.000; rep 2 X ² = 3.43, p = 0.064; rep 3 X ² = 17.01, p = <0.000; rep 4 X ² = 11.17, p = 0.001; rep 5 X ² = 5.95, p = 0.015; rep 6 X ² = 15.72, p = <0.000. Fig 2B, rep 1 X ² = 2.92, p = 0.087; rep 2 X ² = 8.33, p = 0.004; rep 3 X ² = 0.985, p = 0.321; rep 4 X ² = 8.33, p = 0.004; rep 5 X ² = 5.32, p = 0.021; rep 6 X ² = 0.501, p = 0.479. Fig 2C, rep 1 X ² = 13.82, p = <0.000; rep 2 X ² = 0.323, p = 0.570; rep 3 X ² = 12.65, p = <0.000; rep 4 X ² = 1.65, p = 0.198; rep 5 X ² = 2.05, p = 0.152; rep 6 X ² = 9.51, p = 0.002. Fig 2D, rep 1 X ² = 2.31, p = 0.128; rep 2 X ² = 0.731, p = 0.398; rep 3 X ² = 6.876, p = 0.009; rep 4 X ² = 2.97, p = 0.084; rep 5 X ² = 0.02, p = 0.887; rep 6 X ² = 2.49, p = 0.114. On average over the 6 replicates, 21.93% of whiteflies settled on tomato intercropped with marigold and 36.12% whiteflies settled on tomato intercropped with limonene dispensers. 31.44% of eggs were laid on marigold intercropped tomato and 37.24% were laid on limonene intercropped tomato. Average distribution (n = 6) of whiteflies in limonene and marigold treatments were compared to ascertain whether marigold control was significantly more effective at repelling whitefly, however this was not the case (X ² = 2.016, p = 0.156, df = 1).

Fig 3. Whitefly population development in the second glasshouse trial with marigolds and limonene “emergency” intercropping treatments to control an established whitefly population (n = 8).

Treatments comprised intercropping eight heavily whitefly-infested tomato plants with eight further tomato plants (control), eight French marigold plants (marigold treatment), or eight limonene dispensers (limonene treatment). The first data point in Fig 3A, 3B and 3C represents whitefly abundance (adults/eggs/nymphs) on the focal tomato plants immediately before treatment plants were introduced. ‘*’ indicates a p value of <0.05, ‘**’ indicates a p value of <0.001, NS = not significant. Fig 3A displays the average number of adult whitefly settled on each focal tomato plant in the three treatments over the course of the experiment (n = 8). No significant effect on settling adult whiteflies was observed between treatment (rm ANOVA F _(2,105) = 1.71, p = 0.185) and treatment x time (rm ANOVA F _(8.105) = 0.896, p = 0.522) across the sampling period. Fig 3B displays the average number of eggs laid on each focal tomato plant over the course of the experiment (n = 8). No significant effect on whitefly egg abundance was observed between treatment (rm ANOVA F _(2,105) = 0.477, p = 0.625) and treatment x time (rm ANOVA F _(8,105) = 0.351, p = 0.943) across the sampling period. Fig 3C displays the average number of whitefly nymphs of all stages counted on each focal tomato plant over the experiment (n = 8). There was no significant effect on treatment (rm ANOVA F _(2,105) = 2.72, p = 0.070) but there was a significant effect on treatment x time (rm ANOVA F _(8,105) = 2.62, p = 0.011). The limonene treatment had significantly less nymphs than the control on day 29 (t = 4.18, df = 105, p < 0.001) and neared significance at day 22 (t = 2.17, df = 105, p = 0.095). The limonene treatment also had significantly less whiteflies than the marigold treatment at day 29 (t = -2.61, df = 105, p = 0.030). Ninety five percent confidence intervals have been calculated and are available in the supporting information (S2 Dataset), but to aid visualisation they have been removed from the figure.

Fig 4. Plant development characteristics at the end of the 2017 assay into the efficacy of marigolds and limonene as an emergency treatment for the control of established whitefly populations (n = 8).

Fig 4A displays the median aboveground tissue weight of each focal tomato plant, excluding tomatoes, at the end of the 29 day experiment (n = 8). Tomato plants from the marigold treatment were near-significantly lighter than tomato plants from the control (t = -1.86, df = 14, p = 0.085) and significantly lighter than tomato plants from the limonene treatment (t = -3.3, df = 14, p = 0.005) according to t-tests. Tomato plants from the limonene treatment were significantly heavier than those from the control according to a t-test (t = 2.15, df = 14, p = 0.05). Fig 4B displays the median weight of all tomatoes from each focal tomato plant at the end of the 29 day experiment (n = 8). The total tomato weight from plants in the marigold treatment was near-significantly lower than plants in the control (t = -1.76, df = 14, p = 0.100) and significantly lower than the total tomato weight of plants from the limonene treatment (t = -2.8, df = 14, p = 0.014). There was no significant difference in tomato weight between the control and limonene treatment.