Host and parasitoid rearing
Our laboratory colony of CLB was initiated using various life stages of the beetle that were collected from coconut groves near Sanya, on Hainan Island in southern China. The colony was maintained in a growth chamber at 25–30°C, 70–80% RH, an a 12:12 (L:D) photoperiod, following Lu et al. (2008). Coconut spear leaflets were cut into 5–7 cm long pieces and placed in plastic rearing boxes (20 cm in length, 13 cm in width, and 7 cm in depth). Each box was equipped with a ventilated lid. Adults of CLB were introduced into these containers for oviposition and larval rearing. The foliage in the containers also facilitated the pupation process, allowing for the whole life cycle of the beetle to be completed within the rearing container.
Both parasitoids, A. hispinarum and T. brontispae, were acquired from commercial insectaries and were reared under the same physical conditions as our CLB colony until they were used in laboratory or field experiments. Both parasitoids were reared in transparent plastic containers (50 × 50 × 50 cm) with mesh tops for facilitation. During rearing, the parasitoid-to-host ratio was held at ca 1 parasitoid adult to 5 CLB larvae or pupae. Parasitoids were supplied with hosts and fed diluted honey. When needed for experiments, adult parasitoids were collected from their rearing containers using a small aspirator.
Exp. #1. Parasitoid performance and host impact in mixed species releases: a laboratory study
To evaluate the effects of interspecific competition between A. hispinarum and T. brontispae on parasitoid performance, we examined three different treatments, each in a separate container with 50 CLB larvae and 50 CLB pupae. Treatments were as follows: A: 100 A. hispinarum; B: 100 T. brontispae; and C: 50 of each parasitoid species. After 25 days, we evaluated the longevity of the initially introduced adult parasitoids (parental generation) having isolated them before the emergence of F1 adults. This separation was achieved by transferring each parasitized CLB larva or pupa into individual, labeled containers before emergence of F1 adults. We then monitored and recorded the lifespan of the parental parasitoids until their natural death. To measure the effects of mixed-parasitoid versus single-parasitoid releases on host parasitism levels and host survival to the adult stage, we also recorded the total number of parasitized CLB individuals (larvae or pupae) and the number of CLB individuals that escaped parasitism and become adults. Each treatment was replicated 10 times.
Exp. #2. Effects of parasitoid release strategies under laboratory conditions
To assess the impacts of various release strategies, we examined three factors: (1) the ratio between the two parasitoid species, (2) the parasitoid-to-host ratio, and (3) the frequency of releases. For all treatments, the CLB test population comprised 50 larvae and 50 pupae, totaling 100 individuals. Run one of this experiment (a 25-day trial) tested the effects of various ratios of the two parasitoid species, evaluating seven ratios (1:1, 2:1, 3:1, 4:1, 1:2, 1:3, and 1:4), each with a total count of 100 parasitoids (total of both species). Run two, again a 25 days trial, tested eight parasitoid-to-host ratios (2:1, 4:1, 6:1, 8:1, 10:1, 12:1, 14:1, 16:1) for each parasitoid species, each with a total count of 100 parasitoids (total of both species). Run 3 tested the effects of seven release frequencies (1, 2, 3, 4, 5, 6, or 7 releases, distributed over a 200-day trial), with all releases made at 30 days intervals and terminated when the desired number of releases had been made. In this run, we used the most effective ratio of the two parasitoids and the most effect ratio of parasitoid to hosts, as found in runs 1 and 2 of this experiment.
In all three runs of this experiment, the outcomes that were counted to assess the results were the number of parasitized CLB individuals and the number of coconut leaf beetles that were not parasitized and became adults. For Run 1 and Run 2, each with a duration of 25 days, the number of parasitized CLB individuals and surviving CLBs was measured at the conclusion of the trial. In Run 3, which extended over a period of 200 days with parasitoid releases at 30-day intervals, we conducted a single assessment of parasitized CLB individuals at the end of the trial. Each of the above three runs was replicated 10 times.
Exp. #3. Assessment of parasitoid releases in coconut tree plantations
The field trials were conducted in Haina, China, an area notably affected by CLB infestations. Six distinct coconut groves in Hainan, each 20–30 hectares in size, were selected for the field tests (Grove A: 109.64° E, 18.35° N; Grove B: 110.55° E, 19.14° N; Grove C: 110.50° E, 19.80° N; Grove D: 109.21° E, 18.44° N; Grove E: 110.22° E, 19.32° N; Grove F: 110.40° E, 19.74° N). Groves A, B, and C were used in year one of the trial and groves D, E. and F in the second year.
On average, each hectare had about 150 coconut trees, resulting in a substantial tree population for study. These groves were strategically positioned, with over 30 km between groves. Trees in groves were of similar size, averaging between 15 to 30 meters in height and 8 to 10 years of age. All sites showed pronounced CLB damage. Pest density, measured by our research team through systematic sampling and direct observation in early spring of the study year, was high at all sites, exceeding 50 beetles per tree. Over the years, these groves have been maintained with minimal intervention, avoiding the use of chemical pesticides and allowing trees to grow naturally. In this field trial, work in the first year of this two years trial was done in groves A, B, and C, and in year two, we use groves D, E, and F. In each grove, we estimated the overall population density of CLB based on randomly selected samples. Our focus was on heart leaves, the primary habitat for CLB larvae and pupae. For sampling, in each grove we selected ten trees that were well spread out over the grove. For each selected tree, our sample method was to climb the coconut tree, use a saw to carefully cut down the heart leaves, and then take them to the laboratory where we counted the number of CLB larvae and pupae present within these leaves. Finally, we estimated the total population of CLB larvae and pupae within the entire coconut grove by multiplying the number of trees in the grove, by the average numbers per tree in the sampled trees.
As mentioned above, the study was conducted over two years, with the first phase running from April 2021 to May 2022 in groves A, B, and C. The second phase continued from April 2022 to May 2023 in a different set of groves, labeled D, E, and F. Each grove was a treatment: the groves A and D received releases of the parasitoid A. hispinarum. Groves B and E received releases of T. brontispae only, and grove C and F received releases of both A. hispinarum and T. brontispae in a 3:1 ratio. (Note: the total number of parasitoids released in each treatment was the same.) In all treatments, five releases were made in total, being done at monthly intervals. Each release used a parasitoid: host ratio of 10:1, with hosts counted as the number of CLBs larvae or pupae. The field release strategy for parasitoids was determined based on the optimal release approach derived from indoor experimental results.
Insect density and parasitism were estimated by field sampling in each grove once per month. In our study, we randomly selected ten trees and collected samples of heart leaves, counting both surviving and parasitized CLB larvae and pupae. To determine parasitism, we identified parasitized individuals by observing external signs of parasitism, such as mummification. Additionally, for larvae and pupae that did not show external signs of mummification were dissected to look for the presence of parasitoid eggs or young larvae, ensuring a comprehensive assessment of parasitism. The data collection process lasted a total of 390 days summed over the two years. This schedule was strategically designed to maximize the impact of the parasitoids on the CLB population, while also allowing for thorough data collection on parasitism rates and host density variations over time.
Statistical analysis
We used two generalized linear models (GLMs) with Poisson distributions to compare differences in adult longevity and emergence rates for A. hispinarum or T. brontispae under both individual and mixed releases, in each model, the explanatory variable was the type of release, while the response variables were either adult parasitoid longevity or the number of adult parasitoids that emergenced. Furthermore, we used Kruskal-Wallis (K-W) tests to compare the number of parasitized CLB or surviving CLB between single-species versus two-species releases of the two parasitoids.
To analyze the impact of different release strategies on the number of hosts parasitized and the number that survived to become CLB adults, we constructed six GLMs with negative binomial distributions. In each model, the explanatory variable was the ratio of the two parasitoids, the parasitoid to host ratio, or the number of releases, while the response variables were the number of (1) CLB larvae or pupae that were parasitized or (2) that survived to become CLB adults.
To analyze the reduction of CLB populations and the corresponding parasitism rates after parasitoid releases, we considered data on parasitism and host density spanning two years. The parasitism rate was determined as the proportion of parasitized CLBs to the total CLB larvae and pupae population, encompassing all stages identified as parasitized in the collected samples. The CLB reduction rate was calculated as follows:
$$\text{C}\text{L}\text{B} \text{r}\text{e}\text{d}\text{u}\text{c}\text{t}\text{i}\text{o}\text{n} \text{r}\text{a}\text{t}\text{e} \left(\text{%}\right) = \frac{\text{C}\text{L}\text{B} \text{c}\text{o}\text{u}\text{n}\text{t} \text{b}\text{e}\text{f}\text{o}\text{r}\text{e} \text{t}\text{r}\text{e}\text{a}\text{t}\text{m}\text{e}\text{n}\text{t} - \text{C}\text{L}\text{B} \text{c}\text{o}\text{u}\text{n}\text{t} \text{a}\text{f}\text{t}\text{e}\text{r} \text{t}\text{r}\text{e}\text{a}\text{t}\text{m}\text{e}\text{n}\text{t}}{\text{C}\text{L}\text{B} \text{c}\text{o}\text{u}\text{n}\text{t} \text{b}\text{e}\text{f}\text{o}\text{r}\text{e} \text{t}\text{r}\text{e}\text{a}\text{t}\text{m}\text{e}\text{n}\text{t}}$$
The two study years of our field trial were analyzed independently. To do so, we constructed two Generalized Linear Mixed Models (GLMMs) utilizing a β-distribution function for each year. In analyzing the differences in CLB reduction rates among the different release methods, the GLMM was structured with the release method as the fixed effect, the CLB reduction rate as the dependent variable, and the time post-release as a random factor. Similarly, for parasitism rates, the model was constructed with the release method as the fixed effect, the parasitism rate as the dependent variable, and time post-release as the random factor.
Statistical analyses were conducted using R Studio (version 3.6.3), with key packages including 'lme4' for linear mixed-effects models (Bates et al. 2015), 'emmeans' for post-hoc tests (Lenth et al. 2018), 'ggeffects' for estimated marginal means (Daniel 2018), and 'DHARMa' for diagnostics of hierarchical models (Hartig 2016). We assessed data for normality through the Shapiro-Wilk test, opting for nonparametric methods when necessary. For model comparisons, we implemented a pairwise approach, adjusting P values using a method analogous to Tukey’s test. Model diagnostics were performed by visually inspecting residuals against fitted values, confirming the absence of discernible patterns indicative of model misfit.