Nymphalidae sample composition
Within the Nymphalidae as a whole, the samples totalled 1518 individuals (Table S1). Some species that were difficult to identify precisely were grouped at the genus level, for example, Actinote sp., Pteronymia sp., Tegosa sp., Junonia sp., Blepolenis sp., Opsiphanes sp., Yphthimoides sp., Hermeuptychia sp., and Catonephele sp.. Besides that, Marpesia chiron (1 individual in conventional plantation), Marpesia petreus (1 individual in native and 1 individual in agroforestry) and Marpesia zerynthia (1 individual in conventional and 1 individual in agroforestry plantation) were grouped into Marpesia sp. because of undetermined phylogenetic relationships. The final data set comprised 89 Nymphalidae taxa (genus or species) (Table S1; Fig. S2).
The species with the most even distribution among the treatments share a remarkable tiger colour pattern. Belonging to a Müllerian mimicry system, H. ethilla narcaea, Mechanitis lysimnia and Placidina eurynassa, were considered potential candidates for population analysis as they were widely spread among the treatments. Other species within the same tiger mimicry ring, Eueides isabella, Lycorea halia, and Consul fabrius, were also recorded, but were more restricted in number and between treatments. While Eueides isabella, and Lycorea halia are not documented in the native forest areas, Consul fabrius was not found in the conventional plantations (Table S1). H. ethilla narcaea was selected for the population genetic analysis, as reference genomes were not available for the other genera at the time.
Population structure
LEA analysis (Frichot and Francois 2018) suggested there was only one ancestral population within the genomic data (Fig. S3a), indicating that H. ethilla narcaea shows very little genetic population structure. The DAPC analysis (Jombart et al. 2010), which maximises the differences among genetic groups, suggested there were six genomic clusters in the dataset (Fig. 3a, Fig. S3b). The most common clusters were one, four, and six, while the least common clusters were two, three, and five, with two and five being the most genetically distinct (Fig. 3; Fig. S4; Table S4).
The rarest and most distinct genetic clusters detected by DAPC (two, three, and five) occurred only in native and agroforestry sites (Fig. 3b; Fig. S4; Table S4). Importantly, the individuals belonging to clusters two and three were collected in agroforestry and native forest plots belonging to different, spatially separated sampling blocks (cluster two is present in blocks Ja and Lu, while cluster three is present in blocks Ad and Lu, Fig. 3c), suggesting that their absence from conventional plantations is not due to the limited spatial distribution of these genotypes. On the other hand, cluster five was only found in the most distant sampling block (Ze), suggesting this may be spatially restricted, but was still absent from the conventional plantation plot in this block (Table S4).
Even amidst a continuous genetic landscape, as indicated by our LEA analysis, our study underscores the profound impact of treatment on the likelihood of affiliation with distinct genetic groups (Fig. 3b). Generalised linear mixed models showed good fit (Fig. S5), and confirmed the significance of the interaction between native forest habitat and genetic composition, as shown in Table 1, (p = 0.025), suggesting that membership probability to the different genetic groups differed significantly from chance in the native forest. The native forest had the highest membership probability of genetic groups two, three and five, but the lowest probability of genetic group six. Specifically, rarer genetic variants exhibit a pronounced preference for native habitats (genetic group two, three and five, Fig. S6), contrasting with more ubiquitous genetic clusters that span across various treatment conditions (Fig. 3).
Table 1
Assessing the impact of treatment on the probability of membership to distinct genetic groups in H. ethilla narcaea using a Generalized Linear Mixed Model. Corresponding Standard Errors (SE) and Significance (*) are provided for each parameter, based on a total six genetic groups across three treatments.
Parameter | Estimate | SE | p - value |
Conventional farm | -1.568 | 0.799 | 0.0497* |
Native area | 2.394 | 0.983 | 0.0148** |
Agroforestry farm | 0.093 | 0.979 | 0.3402 |
Genetic group | 0.025 | 0.180 | 0.1749 |
Native: Genetic group | -0.512 | 0.229 | 0.0256** |
Agroforestry: Genetic group | -0.245 | 0.232 | 0.2922 |
Zero-inflation model | -1.609 | 0.632 | 0. 0109** |
Furthermore, our analysis reveals compelling disparities in the estimated log odds between native and conventional agricultural settings (2.394 and − 1.558, respectively, Fig. S6), with both being significantly different from the null expectation (p = 0.0497 and p = 0.015). This discrepancy suggests an elevated likelihood of encountering distinct genetic groups within native habitats, juxtaposed against a diminished probability in conventional farming landscapes. Conversely, the log odds largely conform to the null expectation within the agroforestry plantations (p = 0.3402, Fig. S6), indicative of an even distribution of genotypes therein (varying from 30–40% of membership between all genetic groups).
Treatment effects on dispersal and diversity
We observed differences in the number of individuals among the treatments (F2,24 = 8.41, p < 0.001, Fig. 4a). The mean, standard deviation and total number of individuals was largest in agroforestry area (82.78 ± 35.53, 745 individuals), as compared to conventional area (51.22 ± 21.32, 461 individuals), and native forest (34.67 ± 14.18, 312 individuals). The number of species also varied between treatments (F2,24 = 8.97, p = 0.001, Fig. 4b), with both banana plantations, agroforestry (21.33 ± 2.24) and conventional (19.11 ± 5.55), hosting more species than native forest (12.78 ± 4.84). The highest number (31 species) was recorded in a conventional area (Ti block), while the smallest numbers (5 and 6 species) were observed in native areas (To and Di blocks, respectively).
In contrast to our population structure analysis, we did not find any evidence for genetic differences between treatment types using PERMANOVA to assess genetic distance (F2, 62 = 1.02, R2 = 0.03, p = 0.43) or using dispersion to assess genetic diversity (F2,62 = 0.28, p = 0.75, Fig. 5a), although there is a visible trend towards reduced dispersion (diversity) in the conventional plantations. These results also support that dispersal between sites is high, leading to a lack of strong differentiation between them.
Similarly, at the phylogenetic levels, there was no effect of production regime, be it conventional and agroforestry banana plantations or native forest, using PERMANOVA, on the Nymphalidae phylogenetic community distances among sites (F2,24 = 1.13, R2 = 0.08, p = 0.23) (Fig. S5B), or variability among sites, using dispersion (F2,24 = 1.22, p = 0.31, Fig. 5b). However, the species composition of the Nymphalidae community yielded different results, with a significant difference in the compositional distance between treatment types in the PERMANOVA analysis (F2,24 = 2.44, R2 = 0.17, p = 0.002) (Fig. 5c), and a significant effect of plantation type on diversity in the dispersion analysis (F2,24 = 5.74, p = 0.01), with the native forest cluster being separated from the banana plantation clusters (Fig. 5c). We further assessed the effect of treatment on diversity by Rao's quadratic entropy. In H. ethilla narcaea populations it was marginally non-significant (F2,18 = 0.55, R2 = -0.05, p = 0.059; Fig. 6a; Table S5) but did follow the trend identified in the population analysis of reduced diversity in the conventional plantation populations and higher diversity in the native forest. Similarly, Nymphalidae phylogenetic diversity (F2,24 = 2.80, R2 = 0.12, p = 0.08; Fig. 6b; Table S5) and species diversity (F2,24 = 0.38, R2 = 0.05, p = 0.68; Fig. 6c; Table S5) did not exhibit significant differences between treatments, but sites appear to be more similar in diversity in conventional plantations compared with agroforestry and native forest.
In the assessment of clustering by treatment, the mean nearest taxon distance (MNTD) between H. ethilla narcaea individuals in the native forest was found to be lower than expected by chance (z = -1.95, p = 0.04, Table 2, Fig. S7a). Consequently, individuals in the native forest tended to be over-represented in certain clusters at the tips, potentially a result of the rarest genetic groups found in this area (Fig. S6a). However, these individuals were otherwise distributed randomly across the entire tree, as indicated by the mean pairwise distance (MPD) result (Table 2). This pattern was not observed in either of the plantations, where MNTD did not significantly differ from the null distribution (Table 2). Clustering at the tips could also result from the sampling of genetically related individuals, possibly indicating breeding within the native forest but not in either of the plantations.
Table 2
Standardised effect size of mean pairwise distance (MPD) and mean nearest taxon distance (MNTD) within each treatment in H. ethilla narcaea populations and Nymphalidae communities compared to a null model. The corresponding number of individuals or species in each case are represented as n. Standardised effect size in MPD.n and MNTD.n considering distance weighted by species abundance. The bold values represent a p-value < 0.05 when compared to a null distribution.
| | | | Treatment | |
| | | Conventional | Native | Agroforestry |
H. ethilla narcaea | | | | | |
| | n | 19 | 21 | 25 |
| MPD | z | -0.24 | -0.17 | -0.75 |
| | p | 0.35 | 0.37 | 0.19 |
| MNTD | z | -0.25 | -1.95 | -0.14 |
| | p | 0.34 | 0.04 | 0.41 |
Nymphalidae | | | | | |
| | n | 61 | 50 | 64 |
| MPD | z | -2.98 | -0.12 | -1.29 |
| | p | 0.01 | 0.51 | 0.10 |
| MPD.n | z | -0.58 | -2.43 | -0.33 |
| | p | 0.23 | 0.03 | 0.25 |
| MNTD | z | 2.36 | 0.11 | -0.96 |
| | p | 1.00 | 0.59 | 0.14 |
| MNTD.n | z | 0.44 | -0.51 | -0.32 |
| | p | 0.71 | 0.31 | 0.40 |
In terms of community composition, Nymphalidae species demonstrated more phylogenetic clustering in the conventional plantation than expected by chance, indicated by lower MPD values in the conventional plantation when species abundance was not considered (z = -2.98, p = 0.01, Table 2). This suggests that the conventional plantation may provide specific conditions or resources that favour the prevalence of particular phylogenetic groups or exclude others. Conversely, when considering the abundance of each species (MPD.n), greater clustering was observed in the native forest (z = -2.43, p = 0.03, Table 2, Fig. S6b). All native sites presented a similar phylogenetic composition with the exception of Ja and Ad sites (Fig. S7b). The MPD.n result implies that besides the phylogenetic proximity found between native sites, the abundance of the species is crucial for clustering the native sites. Similar grouping was found using species composition (Fig. S7c), which is consistent with our dispersion analysis. In the agroforestry system, a neutral process was identified, considering both MPD and MNTD, where no clustering or overdispersion occurred. The lack of a clear pattern in this case may suggest a broader distribution of species within the agroforestry system, perhaps representing an intermediate habitat type where the conventional plantation and forest species are both found. This is reminiscent of the genetic results, which also suggested a more even distribution of genotypes within the agroforestry plantation.