Diversity and turnover in moth assemblages in rainforests on a remote oceanic island

The study was conducted in near-pristine lowland rainforest remnants on the eastern and southern slopes of La Réunion (21°S 55.5°E), where average annual rainfall is between 3 and 5m (Dupont 2016). Thébaud et al. (2009) provide a succinct introduction to the biology of the island. Numerically dominant indigenous species in these forests include Gaertnera vaginata (Rubiaceae), Molinaea alternifolia (Sapindaceae), Hancea integrifolia (Euphorbiaceae), Phyllanthus phillyreifolius (Phyllanthaceae) and Nuxia verticillata (Stilbaceae) (see Strasberg et al. 2005 for a detailed description of the flora). Since human settlement (from about 1640, Cheke & Hume 2008), the island has been invaded by a range of invasive non-indigenous woody plants such as Psidium cattleianum, Syzygium jambos Alston (Myrtaceae) or vines such as Rubus alceifolius (Rosaceae), Hiptage benghalensis (Malpighiaceae) (Kueffer & Lavergne 2004). Our study sites, however, are still mainly dominated by indigenous species and maintain what is, presumably, their original vegetation structure (Macdonald et al. 1991; Strasberg et al. 2005).

The invertebrate fauna of La Réunion, as elsewhere, is perhaps the least well known of its metazoan fauna (Legros et al. 2019). Nevertheless, the Lepidoptera, Coleoptera, Odonata and Ephemeroptera have been well worked and modern identification guides exist (Gomy et al. 2016; Martiré and Rochat 2008; Martiré 2010).

We sampled at ten plots of 0.25 ha (50 × 50m) previously established by DS and colleagues in the lowland rainforest band at altitudes between 393 and 792 m asl along the south-eastern and eastern coasts of the island (Fig. 1; Emerson et al. 2017). Plot locations were chosen based on accessible forest habitat that was deemed to be relatively undisturbed in terms of human impact and invasive plant species.

Within each 50 × 50m study plot all trees > 10 cm dbh were measured and identified to species (Borges et al. 2018). Twenty 5 × 5m subplots were additionally established, along two 50 m long transects positioned along the western border of the plot (10 subplots) and the eastern border of the plot (10 subplots) in each study plot, in order to sample small trees, saplings and shrubs which are all important host plants for moths. Within each of these subplots every woody stem > 1cm dbh was measured and identified to species. The vegetation data in the 50 × 50m plots and 5 × 5m subplots were pooled together. Reference material was collected and is deposited in the herbarium of the Université de La Réunion.

On two occasions (April 2018 – the end of the ‘wet’ season; and November 2018 – the end of the ‘dry’ season) the moth fauna of all ten vegetation plots was surveyed using Pennsylvania-style light traps (Frost 1957; Kitching et al. 2005). Each trap comprises a vertically mounted actinic light tube driven by a 12v battery. Three vertically mounted transparent vanes surround the light and a bucket beneath them accumulates the catch. A small block of Dichlorvo-impregnated resin among egg-box fragments in the bucket both killed the moths caught and ensured they were in reasonably good condition for assessment. A rain-guard above the trap and a battery container beneath the trap adapted them well for use in wet forests. At each site on each occasion we aimed to sample a minimum of 400 individual moths and continued sampling until this goal was achieved. We ran two to four traps per plot each night until the sample size was achieved, with traps placed at least 20m apart. We avoided sampling in the three nights on either side of a full moon whenever possible. Trap catches were processed daily, sorted to morphospecies and each ‘species’ counted. All moths with a forewing-length greater than 5mm were sorted. We used this cut-off for feasibility of identification, as many of the micromoths that were excluded would require dissection to be correctly identified. This meant that a small number of largely Gracillariidae were not included. A reference collection of all morphospecies was created with (where possible) several representative individuals of each species. Identifications were finalized and confirmed using reference texts (Guillermet 2009a, b, c, 2011, 2016; Martiré and Rochat 2008) under the supervision of JR. The reference collection is deposited in the Réunion Island collection held by the Université de La Réunion (PVBMT).

Total stems varied from 2511 in the Bras Laurent plot to 749 in the Mare Longue plot (Table 1). This pattern was not reflected in the total basal area measurements which were at a maximum of 14.58m2 in the Mare Longue site and a minimum of 2.61 m2 at Bras Laurent. Total species richness varied from 55 species at Basse Vallee to just 31 at Piton Nelson. Dominant species (in terms of basal area) varied across sites although Nuxia verticillata (Stilbaceae) ranked first in seven of ten sites. All species that were either dominant or one of the two highest ranked subdominants were all indigenous species (ten of twelve being Mascarene endemics) except at the Bras Laurent site where the strawberry guava, Psidium cattleianum (Myrtaceae), was dominant, although with a relatively very low total basal area (Table 1).

For indigenous species the “Intermédiaire” plot has the maximum number of 1408 stems whereas Mare Longue has just 749 stems. When we examine total basal area however, Mare Longue, with a preponderance of large trees, shows the maximum measure of 14.58m2 compared with the minimum value of 1.86m2 at Bras Laurent. Total indigenous species richness varied from 53 species at Basse Vallée to just 29 species at Piton Nelson. Non-indigenous species were absent from the Mare Longue plots but, in terms of numbers of stems > 1 cm DBH totally dominated the Bras Laurent site (with 1888 of 2511 stems). In terms of total basal area, however, non-indigenous plants contributed no more than 7.2% of total basal area except at Bras Laurent (28.7%).

A total of 12,974 moths were sampled over the two seasons representing 239 morphospecies (Table 2). Of the 239 morphospecies, 57 of 191 (29.8%) in the April samples were unique to that sampling period: in November 105 of the 211 (49.8%) sampled species were unique to that period. There were clear differences in assemblage structure between the two sampling periods (Permanova test with month as a fixed factor: r² = 0.55, p < 0.001, Fig. 2.). A repeat of this analysis after removal of ‘rare’ species (i.e. defined as those for which n < 28 following Novotny et al. 2007a, b) removed this seasonal differences. We conclude, accordingly, that the clear difference between the seasons is due to species richness and identity differences, not relative abundances. Across all sites in both seasons the Pyraloidea, that is: the sister families Crambidae and Pyralidae, dominated the samples numerically (Table 2.). In many instances this was due to the superabundance of the scopariine crambid, Scoparia resinodes. The second most abundant higher taxon was the Geometridae in April and the Erebidae in November, although on both occasions the diverse Erebidae were marginally more speciose in our samples.

We calculated rarefaction curves in terms of species richness, inverse Shannon and inverse Simpson index. This shows that plant sampling (Figure S1) encountered virtually all species present. As in most snapshot studies of this kind (where the baseline fauna is known to be very large), however, species richness of moths on both occasions showed that further richness remained unencountered (Figures S2, S3). Values of both diversity indices, nevertheless, rapidly approached asymptotes on all sampling occasions. In terms of similarity/dissimilarity analyses, therefore, we used the Chao-Sørensen measure which takes into account both species richness and relative abundance. We found strong beta turnover for plants and moths, supporting our first hypothesis that there is beta turnover with distance in the plant and moth assemblages (Table 3., Fig. 3.).

For the plants we compared the turnover/distance relationship for endemic (that is: La Réunion specialists), indigenous (that is: plants occurring naturally on La Réunion but not endemic to the island) and non-indigenous species (that is: plant known to have been introduced by human agency to the island)(Table 3). The strength of the relationship in declining order was endemic >  > indigenous < non-indigenous, with that involving endemic plants only being much stronger than the other two. For the moths on each sampling occasion we subdivided the data initially into La Réunion endemics, Mascarene endemics and non-endemic species (Table 3). For the April data, significant turnover relationships were identified for all three data subsets, but the relationship was strongest in the case of the La Réunion endemic species (r = 0.58, p = 0.001) compared with the Mascarene endemics (r = 0.42, p = 0.008) and the non-endemics (r = 0.24, p = 0.02). The relationships between moth turnover and distance for both La Réunion endemic and non-endemic moths showed a very similar slope although that for non-endemics showed much wider scatter as inter-site distances increase (Fig. 4). For the Mascarene endemic moths there was a much steeper relationship between turnover and distance with a wide scatter of points around the relationship line. For the November data relationships were far less clear and none are formally significant (ie p < 0.05). Turnover for La Réunion endemics showed a stronger trend with inter-site distance than was evident for either of the other two subsets. Splitting plants into endemics vs indigenous species gave a much stronger turnover for the former. Splitting moths into endemics and non-endemics also showed stronger turnover relationships for the former vs the latter, robustly for the April data, weakly and non-significantly for the November data. We predicted the turnover of endemic plant and moth species should be stronger than for the entire species assemblage. This was strongly the case for plants. However, for the April moths there was virtually no difference between the two relationships and for the November moths, turnover of the entire moth assemblage was stronger than that for endemics alone – contrary to our hypothesis.

In order to test our third hypothesis, we compared turnover/distance relationships between small and large moths (for definition see Methods). For the April data both ‘large’ and ‘small’ moths showed significant declines in similarity with distance. The relationship for ‘smaller’ was stronger (p = 0.003) than for the group classified as larger (p = 0.009) but in both cases was strong. As with other comparisons the relationships for the November data were much weaker and were non-significant for both size classes.