Characterizing larval breeding sites: physical characteristics
PCA analysis summarizing the 11 physical variables in La Lopé showed that the four larval breeding site groups (two habitats × Ae. aegypti present/absent) overlap extensively in the plane described by the first two principal components, which together account for 38% of the total variance (Figure 2a). However, forest and peridomestic sites appeared to differ slightly. In support of that, MRPP tests found a significant multivariate difference between habitats when including bothAe. aegypti present and absent sites (p = 0.001). Such between-habitat different was less significant when examining onlyAe. aegypti present sites (p = 0.316), possibly due to the small sample size (only five samples in the forest Ae. aegypti present group). Ae. aegypti present and absent sites did not differ significantly (all sites regardless of habitats: p = 0.311, within forest sites: p = 1, within peridomestic sites: p = 1). The axis that differentiated forest and peridomestic larval breeding sites roughly aligned with the vectors of a few environmental variables, including ambient temperature and humidity, canopy coverage, container opening height, and water pH (Figure 2a). These variables also generally had high variable important ranks in the random forest analysis (Figure S2a), reflecting their difference between habitats.
In Rabai (Figure 2b, PCA summarizing 16 physical variables), similarly, forest and peridomestic sites were modestly but significantly different in their physical characteristics (MRPP, p = 0.001). The differentiation between forest and domestic larval breeding sites was more apparent, while peridomestic and domestic sites shared similar conditions.Ae. aegypti present and absent sites in the forest did not show a strong difference. Consistent with PCA, MRPP found significant multivariate differences in most comparisons (p = 0.001), except between forest sites with Ae. aegypti present vs. absent (p = 0.157) and between domestic vs. peridomestic sites (p = 0.192 for all sites and p = 0.147 for Ae. aegypti present sites only). Forest and domestic breeding sites were separated primarily on the first PC, explained by container size (e.g., diameter, circumference, etc.), water volume, and water pH (Figure 2b). Domestic sites tended to be larger, contained more water and had higher pH (i.e., more alkaline). These variables were also identified as most influential in distinguishing habitats in the random forest analysis. In addition, canopy coverage, a measure of shading, also showed a high rank in variable importance in comparisons between habitats (Figure S2b). This is not surprising as domestic containers were under roof and peridomestic containers were exposed outdoor, while forest tree holes were partially shaded by the canopy.