Tri‐trophic interactions among Fopius arisanus, Tephritid species and host plants suggest apparent competition

Abstract When several polyphagous herbivore species share a parasitoid, the tri‐trophic interaction networks can be difficult to predict. In addition to direct effects, the parasitoid may influence the herbivore community by mediating indirect interactions among hosts. The plant species can also modulate the parasitoid preference for a specific host. One of the indirect effects is apparent competition, a negative interaction between individuals as a result of the action of shared natural enemies. Here, we focus on the interactions between the parasitoid Fopius arisanus (Braconidae) and two generalist fruit fly pests: Bactrocera dorsalis and Bactrocera zonata (Tephritidae). This parasitoid was introduced into La Réunion in 2003 to control populations of B. zonata and can also interact with B. dorsalis since its invasion in 2017. Our main objective is to characterize the tri‐trophic interactions between F. arisanus, fruit fly and host plant species. We developed a long‐term field database of fruit collected before and after the parasitoid introduction and after the B. dorsalis invasion in order to compare parasitism rate and fruit fly infestation for the different periods. In laboratory assays, we investigated how the combination of fruit fly species and fruit can influence the preference of F. arisanus. In the field, before the invasion of B. dorsalis, the parasitism rate of F. arisanus was low and had a little impact on the fruit fly infestation rate. After the B. dorsalis invasion, we observed an increase in parasitism rate from 5% to 17%. A bioassay showed that females of F. arisanus could discriminate between eggs of different fruit fly and host plant species. The host plant species preference changed in relation to the fruit fly species inoculated. Field observations and laboratory experiments suggest the possible existence of apparent competition between B. dorsalis and B. zonata via F. arisanus.


| INTRODUC TI ON
In the context of human-induced changes with unintentional (invasion) and voluntary (biological control) alien species introductions, new interactions between species have become frequent and can impact the ecological networks. Studying the ecological mechanisms underlying novel species interactions is a significant challenge to understanding fluctuation in population and community assemblage, such as species colonization and range expansion (Strauss et al., 2006;Wang et al., 2013). However, the ecological outcomes of species interactions can only be fully understood after considering the multi-trophic approaches in which the species are embedded, i.e. beyond the simple pairwise interactions, the emergent features of interactions visible at least at a tri-trophic should also be considered (Fortuna et al., 2012;Harvey et al., 2003;Perović et al., 2018;Price et al., 1980;Singh, 2003). Understanding multi-trophic interactions are fundamental in the context of biological control and pest invasions (Schulz et al., 2019;Tylianakis & Binzer, 2014). For example, the fluctuation of pest herbivore populations can be mediated by resource availability and presences of natural enemies (parasitoids, predators, or pathogens). In return, plants can affect how natural enemies impact herbivore populations (Abdala-Roberts et al., 2019;Price et al., 1980). However, the tri-trophic interaction networks (parasitoid -herbivores -host plants) can be complex and difficult to predict. In addition to the direct negative effect of parasitism, the parasitoid may influence the host species' community structure by mediating negative or positive indirect interactions among hosts (Abrams et al., 1996;Chaneton & Bonsall, 2000;van Veen et al., 2006). Apparent competition refers to an indirect negative interaction between individuals due to the action of shared natural enemies (Bonsall & Hassell, 1997;Holt & Bonsall, 2017;van Veen et al., 2006). Apparent competition can occur when the presence of one prey species increases predator density, thus increasing predation on other species (Density-dependent indirect effects, Holt & Lawton, 1993;Long et al., 2012). Moreover, apparent competition can occur when the presence of one prey species induces changes in predator traits or behavior, which alter the interaction of the predator with other prey species (trait-mediated indirect interactions, Werner & Peacor, 2003;Banerji & Morin, 2014). One mechanism underlying these effects is predator or parasitoid selectivity. If the two host species are not equivalent or if the parasitoid has a host preference, the preferred prey species is likely to become extinct (Chailleux et al., 2014;Chaneton & Bonsall, 2000;van Veen et al., 2006). In addition, the plant species can modulate the parasitoid preference for a specific host when herbivore hosts are polyphagous (Traine et al., 2021). Although biological control is founded on the concept of trophic interactions, the impact of indirect effects due to parasitoids is largely unexplored.
One example of complex interactions is found between the parasitoid Fopuis arisanus (Sonan, 1932) (Hymenoptera: Braconidae) and the two tephritid species: Bactrocera dorsalis (Hendel, 1912) and B. zonata (Saunders, 1841) (Diptera: Tepritidae). These three species currently coexist in several parts of the world. F. arisanus was introduced in many countries for tephritid biological control , and these two Bactrocera species are major invasive pest species both present in Sudan, Pakistan, Mauritius, and La Réunion (Abro, 2020;Mahmoud, Abdellah, et al., 2020;Sookar et al., 2021). Furthermore, their distribution overlap could increase if we consider climate change and their potential future distribution area, which has been modeled by several authors (De Villiers et al., 2015;Ni et al., 2012).
However, the dominant species may vary from region to region.
B. zonata is the dominant species in Sudan , while Bactrocera dorsalis is the dominant species in La Réunion and Mauritius Sookar et al., 2021). The outcome of the competition is modulated by factors such as climatic tolerance. Indirect effects linked to parasitoids could also influence the interactions between these two species.
In La Réunion, F. arisanus was released between 2003 and 2005.
The primary purpose of its introduction was to control B. zonata detected on the island for the first time in 2000, but also two Ceratitis species with economic impact, Ceratitis quilicii De Meyer, Mwatawala and Virgilio, 2016 and Ceratitis capitata (Wiedemann, 1824) (White et al., 2000). However, after the invasion of B. dorsalis on the island in 2017, the ability of the well-established F. arisanus populations to parasitism again its ancestral host was uncertain. With these multiple unintentional (invasion) and voluntary (biological control) species introductions, La Réunion (France) represents a particular area to study how new interactions can impact ecological networks and tri-trophic interactions. We explored these questions using a longterm field database of fruit collected before and after the parasitoid introduction and after the B. dorsalis invasion (from 1991 to 2009 and 2018 to 2019). In addition, laboratory experiments were carried out to study the tripartite interactions between host plant, fruit fly species and F. arisanus in La Réunion (France). First, we analyzed the change in the infestation and parasitism rate since the introduction of F. arisanus in 2003. We supposed that the introduction of F. arisanus reduced the infestation rate of B. zonata and Ceratitis species.
After the B. dorsalis invasion, we hypothesize that indirect interactions among the two main hosts (Bactrocera species) via the parasitoid could exist. Secondly, in laboratory experiments, we analyzed interactions between Tephritidae and F. arisanus and how the host plant influenced Tephritidae/parasitoid interactions. It was proven that F. arisanus could discriminate and choose between fruit-fly species eggs for oviposition (Ayelo et al., 2017;Bautista & Harris, 1996;Mohamed et al., 2010;Rousse et al., 2006), and we supposed a preference for Bactrocera species in comparison to Ceratitis species.
However, the preference between B. zonata and B. dorsalis was more challenging to predict. While B. dorsalis is the ancestral parasitoids' host, Fopius arisanus interacted with B. zonata for 14 years in La Réunion . From a tri-trophic viewpoint, we also supposed that the host plant could modulate fruit fly preferences of the parasitoid. Finally, we discussed how field samplings and experimental results suggest an apparent competition between these species.

| Fopius arisanus and historical data of releases
Fopius arisanus is an egg-larval parasitoid species regularly used for the biological control of Tephritidae. The species is native to the Indo-Malayan region. It is a solitary koinobiont endoparasitoid that attacks the eggs of fruit fly species and emerged from the puparium (Rousse, 2007). It was used as a biological control for the first time in Hawaii in 1946. Then, it was introduced from Hawaii to many other parts of the world, including Africa and the Indian Ocean, to control tephritid pests Purcell, 1998;Rousse et al., 2005). Fopius arisanus can attack numerous fruit fly species, but it predominantly attacks Bactrocera species (Mohamed et al., 2010;Rousse et al., 2006;Zenil et al., 2004). In the introduction regions, this generalist species was regularly exposed to several hosts that coexist, for example, F. arisanus control B. dorsalis, Bactrocera kirki (Froggatt, 1911), and Bactrocera tryoni (Froggatt, 1897) in French Polynesia (Vargas et al., , 2012. In La Réunion, F. arisanus can attack Bactrocera dorsalis, Bactrocera zonata, and Ceratitis species (Rousse et al., 2006 (Rousse et al., 2006). Approximately 74,800 individuals were released in different parts of the island (Table 1; Quilici et al., 2005).

| Field collection
To study interactions among fruit fly and parasitoid species, we performed field campaigns on the entire island of La Réunion. La Réunion is located in the southern Indian Ocean (55°30′E; 21°10′S), around 700 km off the coast of Madagascar. It is a volcanic island that rises to an altitude of 3100 m. Its topography is rugged and has a humid tropical climate, with a dry season from May to October and a wet season from November to April.
Sampling was regularly performed between 2000 and 2003, just after the B. zonata invasion, between 2004 and 2009 (except 2008), during and after the release of F. arisanus (Duyck et al., 2008) and between 2018 and 2019 after the B. dorsalis invasion . The same data collection method was used throughout the different sampling periods. We collected ripe fruit samples on the ground or on trees from different plant species (cultivated, ornamental or wild) all over the island. Whenever possible, we sampled 15 fruits for each plant species found per location and date. In total, we collected more than 33,500 individual pieces of fruit from 112 potential host plant species.
In the laboratory, the fruit samples were individually weighed, placed in plastic boxes with sand as pupation substrate, and covered with a fine-mesh cloth. We put fruit samples in a maturation room (25°C ± 2°C and 70 ± 20% humidity) until pupation. Fruit samples were regularly inspected for 3 weeks, and the sand was sifted to look for pupae. Pupae were kept in a climatic room in plastic boxes until their emergence, when they were taxonomically identified to species level. We identified fruit flies and parasitoids (Appendix S1) using morphological criteria (Virgilio et al., 2014;Wharton & Yoder, 2021). Identification was performed at emergence. Fruit could be infested by several fruit flies and it was impossible to determine which fruit fly species was parasitized.
We recorded the number of emerging individuals for each fruit fly species or parasitoid according to fruit (species and weight), site and date (of collection). We calculated (i) the fruit fly infestation rate as the number of emerged flies per kg of collected fruit and (ii) the parasitism rate as the number of parasitoids on the number of emerged imago (flies and parasitoids). Following other studies on parasitism of fruit flies (Aluja et al., 1990;Dieng et al., 2020;García-Medel et al., 2007;Ovruski et al., 2004), we calculated the parasitism rate (PR) of Fopius arisanus for each host plant species separately with the formula:  (Duyck & Quilici, 2002). Fruit fly eggs used for bioassays were collected from routine rearing cages (housing a few thousand females), into which we placed a perforated plastic ball containing a small piece of fruit (guava, lime, mango, or papaya) to stimulate egg laying inside this oviposition device. Eggs were never rinsed and were manipulated with a fine wet paintbrush.
Parasitoids and flies were reared in a 45 × 45 × 45 cm plastic screened cage at 25 ± 2°C, 70 ± 20% RH, with a 12 L:12D photoperiod. The adults were given free access to water and food consisting of sugar and enzymatic protein hydrolysate.

| Fruits
We chose host plant species according to the infestation rates observed in the field in La Réunion for the target tephritid species Delatte, 2021). We used ripe fruit with no pesticide treatment. We protected guava and mango with fine-mesh nylon bags at the unripe stage to avoid infestation by wild fruit flies. We collected unripe papaya and kept it in the laboratory at room temperature until the ripe stage. We visually checked the absence of stings on the limes. To provide a standardized oviposition substrate, fruit samples were cut into small pieces of about 9 cm 2 with two slits of 5 mm deep to slip in the eggs of fruit flies.

| General protocol
We tested whether the oviposition choice of F. arisanus was influenced by the host plant and fruit fly species. Using a fine wet paintbrush, we gently deposited 50 <4 h old fruit fly eggs in each slot (100 eggs per fruit). Fruit samples were spaced approximately 10 cm apart and exposed to naïve and mated parasitoid females (4-15 days old) for 24 h in 30 × 30 × 30 cm cages with natural light. At the end of the experiment, we rinsed fruit samples with water and sieved eggs on a piece of thin netting. We dechorionated the eggs using the same protocol as Rousse et al. (2006). Eggs were immersed for 60 s in a 2.6% NaClO solution and then rinsed with water. They were deposited onto a microscope slide with mineral oil and observed under a binocular microscope at 100× magnification. The proportion of parasitized eggs was calculated as the number of parasitized eggs over the total number of counted eggs.

| Fruit fly species
To test parasitoid choice according to fruit fly species, we exposed eight F. arisanus females to eggs of different combinations of two fruit fly species (B. dorsalis/B. zonata; B. dorsalis/C. quilicii or B. zonata/C. quilicii). We arranged two pieces of guava, one with 100 eggs of one species and the other with 100 eggs of the second species. Each cage constituted a replicate (n = 8 for each species combination). We had four experimental blocks in which each combination was tested simultaneously (3 species combination × 2 F. arisanus colonies). We also conducted no-choice tests following the same protocol but using the same species on both pieces of guava (n = 5).

| Host plant species
To test parasitoid choice regarding host plant species, we exposed 16 F. arisanus females to eggs (100 eggs per fruit) deposited on a piece of guava, lime, mango, and papaya, simultaneously. This experiment was carried out with eggs from the three fruit fly species.

| Statistical analyses
All analyses were conducted in R (R Development Core Team, 2021), and data are presented as mean ± standard error. When we used Generalized Linear Mixed Models (GLMM), we always checked the homoscedasticity, normality, and independence of residuals graphically.

| Field collections
We compared the infestation rate of B. zonata and C. quilicii (not enough data for doing any statistical analysis for C. capitata using   Similarly, we performed a GLMM to test the influence of host plant species on the proportion of eggs parasitized by F. arisanus.

| Experimental test
In this case, the proportion of parasitized eggs was the response variable; we tested the influence of host plant species, fruit fly species, and the colony of F. arisanus (fixed factors). The interactions between fruit fly species and host plant species were also tested.
We added cages as a random factor.

| Fruit fly species
We did not observe a significant difference in the proportion of parasitized eggs between the colony of F. arisanus reared on B. dorsalis eggs, and the colony reared on B. zonata eggs during choice experiments ( 2 1 = .041, p = .839). In no-choice tests, proportions of parasitized eggs were 0.15 ± 0.07 for B. dorsalis eggs, 0.19 ± 0.09 for B. zonata eggs, and 0.04 ± 0.03 for C. quilicii eggs and were significantly higher for B. zonata eggs than for C. quilicii eggs (z value = 3.639, p < .001, Figure 3).

| Host plant species
We did not observe a significant difference in the proportion of par-  Table 3    However, the parasitism rate was highly variable according to the host plant species and location. In our results, this parasitoid was absent from 32 plant species infested by B. dorsalis or other generalist species, while the infestation rate reached 41 ± 17% for Cananga odorata. According to , in the plant species most infested by B. dorsalis, the parasitism rate by F. arisanus was 17 ± 3% for M. indica, 37 ± 2% for T. catappa, 16 ± 2% for S. jambos, 19 ± 1% for P. cattleianum, and 24 ± 2% for P. guajava. These values are low compared to parasitism rates observed in Hawaii and French Polynesia (Bess et al., 1961;van den Bosch & Haramoto, 1951;Vargas et al., 1993Vargas et al., , 2007Vargas et al., , 2012 where parasitism rates of P. cattleianum, P. guajava, and T. catappa were included between 41% and 73%. The global parasitism rate observed in our

F I G U R E 4 The proportion of parasitized eggs (mean ± SE) by
Fopius arisanus according to host plant species on which eggs were deposited and fruit fly species. Different letters indicate a significant difference in parasitism rate among host plant species for each fruit fly species. study (17%) is more similar to values recorded in Africa, where this parasitoid was introduced from Hawaii, and where the average parasitism rate varied according to studies from 1.7% in Mozambique to 14% in Senegal (Cugala et al., 2016;Gnanvossou et al., 2016;Ndiaye et al., 2015). The discrepancies in parasitism efficacy observed between the islands in the Pacific Ocean and Africa (including the Indian Ocean islands) could be linked to several factors. However, the host plants (very similar exotic species are found in these countries), and climatic conditions (the introduced areas cover a wide range of climatic conditions), do not appear to be the main explanatory factors for these differences. Other factors may be involved.
First, when the F. arisanus population was initially introduced, only a few individuals were used. Consequently, the effective population size was small. This increased the effects of inbreeding and genetic drift, leading to a greater loss of genetic diversity and potentially affecting population fitness (Zaviezo et al., 2018). Another hypothesis could be that not all species of Tephritidae are suitable hosts for the parasitoid; and if eggs are laid in some non-host species, it could be a dead-end host for F. arisanus (Rousse et al., 2006). In Africa, in areas where it was recently introduced, a very different and broad community of Tephritidae species is found, which could also explain its reduced efficacy.

| Host plant preference
We demonstrated the capacities of F. arisanus to discriminate fruit substrate for oviposition. For example, eggs deposited in lime (C. aurantifolia) were neglected in favor of other host plants.
Citrus species have been widely recognized as poor hosts for fruit flies because of the chemical resistance in the peel (Greany et al., 1983;Papachristos & Papadopoulos, 2009;Ruiz et al., 2014).
On the contrary, F. arisanus preferred guava and mango, hosts of high nutritional quality for polyphagous fruit fly species (Hafsi et al., 2016). Host selection by parasitoids seems to match the preference-performance hypothesis. This hypothesis describes how the female selects the oviposition site to optimize the development of its progeny (Gripenberg et al., 2010). This trend was observed in parasitoids, including F. arisanus (Ayelo et al., 2017;Bautista & Harris, 1996), but it is less common in generalist species (Gripenberg et al., 2010;Monticelli et al., 2019). Moreover, the preference for a host plant varied according to the species of eggs deposited. In the no-choice (tephritid host) experiment, F. arisanus preferred to lay eggs in the guava and mango when it was infested by C. quilicii eggs, the papaya and mango when it was infested by B. dorsalis eggs, and the papaya when it was infested by B. zonata eggs. Fopius arisanus adapted its preferences for the oviposition site according to the fruit fly species present. The preferenceperformance hypothesis was not always confirmed. For example, F. arisanus preferred to lay eggs in papaya when B. zonata infested the fruit, whereas Hafsi et al. (2016) have shown that survivorship of B. zonata was very low on papaya. Fopius arisanus is classified as a generalist parasitoid, reported to be able to develop on over 80 host plant species from diverse families and on at least 35 host fly species belonging to Tephritidae (Gnanvossou et al., 2016;Nanga Nanga et al., 2019;Rousse et al., 2005). It has been suggested that the strength of the preference-performance relationships depends on the specificity of the diet (Gripenberg et al., 2010).
In generalist species, insect behavior can be constrained by their ability to recognize specific cues of a fruit fly, host plant species, and a combination of the two.
Preferences of F. arisanus in the laboratory were consistent with field observations. We observed a higher parasitism rate on C. papaya and P. guajava (24 ± 2% for both), than on M. indica (17 ± 3%), and the parasitism rate was zero for Citrus species (except Citrus tangerina). While most studies focused on some highly parasitized species (mango, guava, tropical almond), we collected cultivated, ornamental, and wild host plant species. Some of these host plants had a significant infestation rate but a lower or null parasitism rate. For example, we found a parasitism rate of 2% for Diospyros kaki, Ziziphus mauritiana and 0% for Musa sp., Prunus sp., and Pyrus sp. It is essential to consider these species because they may represent a refuge for fruit flies. The 'refuge theory' proposed by Hawkins et al. (1993) predicts that if hosts occupy a large niche, parasitoids may fail to sufficiently reduce the host population's density for effective biological control. We were able to highlight refuge plants for B. dorsalis, C. capitata, and C. quilicii, but not B. zonata and C. catoirii (see the network shown in Figure 2). The absence of parasitism in some host plant species could result from the combination of sampling effort and the spatio-temporal variations of the parasitism rate. Parasitoid populations can fluctuate as a function of climatic factors, host plant availability, and fruit fly density. Parasitoids can be attracted to highly infested patches or avoid already parasitized hosts (Aguiar-Menezes & Menezes, 2001;Kitthawee, 2000). Models have shown that the spatio-temporal heterogeneity in parasitism rate and the presence of host refuges can stabilize parasitoid-host interactions (Briggs & Hoopes, 2004;Holt & Hassell, 1993). Nevertheless, empirical studies are required to understand the different parameters influencing parasitism rates in fruit fly parasitoids.

| Parasitoid-Tephritidae interaction
This study also shows how females of F. arisanus can discriminate between eggs of different fruit fly species. We have demonstrated that the preference for the host plant species varies depending on the fruit fly species infesting the fruit. Our original findings reveal that when the parasitoid had the choice between B. dorsalis and B. zonata eggs, it had a preference for the latter.
However, our study disregarded these marking pheromones because we moved eggs from the artificial support to the piece of fruit.
Thus, only compounds present on the eggs can influence the observed behavior. Rousse et al. (2007) demonstrated that females of F. arisanus respond to kairomones emanating from the egg masses of Tephritidae, which could explain this behavior.
In choice and no-choice experiments, F. arisanus preferred eggs of Bactrocera species to eggs of C. quilicii. This result was consistent with previous studies (Ayelo et al., 2017;Bautista & Harris, 1996;Mohamed et al., 2010;Rousse et al., 2006). It shows that F. arisanus can discriminate between fruit fly species. In this situation, the parasitoid preference is in line with performance. F. arisanus has a much higher survival rate when it parasitizes B. zonata (75.7%), than when it parasitizes C. quilicii (22.0%, Rousse et al., 2006). This could result from the long co-evolution of these species. In its region of origin (Indomalayan region), as well as in regions of introduction (Hawaii), F. arisanus is found to parasitize Bactrocera species (Ramadan et al., 1992

| Indirect interactions
In our results, many parameters suggest that indirect interactions could exist between B. zonata and B. dorsalis via F. arisanus.
First, both species were suitable hosts for F. arisanus (Harris & Bautista, 2001;Rousse et al., 2006) and share the same ecological niche in La Réunion . Moreover, we observed a greater abundance of F. arisanus and a decrease in B. zonata infestation rate and the adult population just after the B. dorsalis invasion. This could be due to apparent competition, a mechanism that is mediated by density, whereby the greater abundance of one host allows an increase in parasitoid abundance and then has a negative impact on a second host species. In addition, although not tested here, trait-mediated indirect interactions could add up to densitymediated interactions if B. dorsalis induces changes in F. arisanus traits (morphological or behavioral) that could alter its interactions with B. zonata. Other studies show that field observation suggested an indirect effect even during the biological invasion (Chaneton & Bonsall, 2000). For example, (Settle & Wilson, 1990)  Furthermore, the preference of F. arisanus for B. zonata could influence indirect interactions between the two Bactrocera species, with a shift towards B. zonata. If the natural enemy has a feeding preference for one type of prey, the interactions between the host species could be asymmetric, i.e. one prey species can have a negative effect on another prey species, while the reciprocal effect is near zero (i.e. amensalism). This situation is common (Brassil & Abrams, 2004;Chaneton & Bonsall, 2000) and could contribute to the significant decrease of the B. zonata population observed in La Réunion, following the B. dorsalis invasion .
In La Réunion, B. zonata populations almost disappeared only 2 years after B. dorsalis was first detected. In 2022, no B. zonata was caught in traps installed around the island (Appendix S2). This observation could result from both direct and indirect competition between the two fruit fly species. Despite all the cases of invasion in fruit fly species, competitive exclusion is very rare. In fruit flies, the only case of exclusion was reported for C. catoirii in Mauritius because of pressure from successive invasions of different species over the years (Duyck et al., 2004(Duyck et al., , 2022. Although populations may be sufficiently abundant during biological invasions to cause interspecific competition (Duyck et al., 2022), many authors suggest that direct competition is not the determinant mechanism for phytophagous communities (Kaplan & Denno, 2007), which includes fruit flies (Clarke, 2016). On the contrary, more and more articles show that indirect interactions are common, such as apparent competition, which structures insect communities and produces similar patterns to those found when there is competition for resources (Bird et al., 2019;Frost et al., 2016;Morris et al., 2005;van Veen et al., 2006).
To conclude, with field sampling and experimental bioassays, our study suggests that direct and indirect interactions could significantly modulate the population of species in a tripartite network, even leading to the disappearance of a resident species. However, other experimental studies are necessary to confirm the part of indirect interactions in the network (Chaneton & Bonsall, 2000). In the context of invasion and biological control, understanding the outcomes of these multilevel interactions is necessary to predict the outcome of population control strategies.

ACK N OWLED G M ENTS
This study was funded by CIRAD, the "Conseil Régional de La Réunion" and the European Agricultural Fund for Rural Development (EAFRD). The authors acknowledge the Plant Protection Platform (3P, IBISA), where all experiments were conducted. We would also like to thank Jim Payet, Serge Glénac, Antoine Franck, Christophe Simiand, and Patrick Turpin for collecting field data over the years.
This research was conducted within the framework of the UMT BAT: 'Biocontrole en Agriculture Tropicale'.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data are available on CIRAD Dataverse https://doi.org/10.18167/ DVN1/NYZ2NR (https://datav erse.cirad.fr/).