Our first analysis shows that LISS-derived NDVI values mirror the general trend in supposedly increasing biomass from scrub jungle to evergreen forest (P < 0.001; F_4,53= 6.5322, ANOVA; Fig. 2). Evergreen forests habitats had NDVI values that differed significantly from all habitats except moist deciduous forests. NDVI values between scrub jungles and moist deciduous forests were also significantly different.

Nest site NDVI at the level of functional groups was significantly different (P < 0.001, F_5,106=14.1029, ANOVA; Fig. 3). Hot Climate Specialists and Opportunists occupied a narrow range of NDVI niche, while Generalised Myrmicinae occupied the broadest range. Posthoc> tests revealed that NDVI at the nest sites of Tropical Climate Specialists, Cryptic Species and Specialist Predators did not differ and were comparatively high (Fig. 3). NDVI at nest sites of Opportunists were significantly different from nest sites of Cryptic Species, Specialist Predators and Tropical Climate Specialists but were similar to Generalised Myrmicinae. NDVI at nest sites of Hot Climate Specialists were comparatively low and were found to be similar to the niche occupied by Opportunists and Generalised Myrmicinae. Our results suggest that in terms of NDVI the Opportunists and Hot Climate Specialists group occupy a very narrow niche within the broad habitat range of Generalised Myrmicinae (Fig. 3).

Analysis of nest site NDVI at the species level revealed large differences (P < 0.001, F_12,118=9.2476, ANOVA; Fig. 4). Nesting sites of C. taprobanae, O. smaragdina, H. saltator, L. processionalis and P. diversus were found in areas with a high mean and an intermediate variance in NDVI. Nesting sites of Crematogaster sp. 1, Pheidole sp. 2 and M. brunnea exhibited an intermediate mean and a particular high variance in NDVI. Nesting sites of A. gracilipes, P. longicornis, T. albipes and M. bicolor were found in areas with a low mean in NDVI and these species displayed the lowest variance in NDVI at their nest sites.

The above findings reflect fairly well the habitat preferences of the studied ant species – except for one of the specialist predators, Pachycondyla rufipes (Fig. 5), whose nest sites showed surprisingly low NDVI values (0.09 ± .069; mean ± SD; n=11) and differed significantly (P < 0.05; t = – 2.046; t-test) from the ensemble NDVI of all other nest sites examined (0.201 ± 0.106; mean ± SD; n = 49). The nest site NDVI analysis suggests that P. rufipes prefers to nest in scrub jungle – a completely unexpected result, since we collected P. rufipes only from deciduous and evergreen forests, but never from scrub jungles or acacia plantations. In the search for an explanation we realized that previously we had observed the nesting locations of P. rufipes to be in dense vegetation patches that had large gaps in the canopy. We wondered if canopy breaks in dense forests were indeed nesting sites of P. rufipes and if these locations could be identified using NDVI. Our subsequent analysis showed that prevalence of P. rufipes in the NDVI range in which the species was initially observed (0.015 – 0.1779), was not significantly different between our validation and initial dataset (P = 0.6223, U=28.0; Mann-Whitney test). In fact, we found P. rufipes nests in all of the 17 new locations, which strongly supports our hypothesis of P. rufipes’ preference for canopy gaps in nest site selection. Our subsequent analysis showed that prevalence of P. rufipes in the NDVI range in which the species was initially observed (0.015 – 0.1779), was not significantly different between our validation and initial dataset (P = 0.6223, U=28.0; Mann-Whitney test). In fact, we found P. rufipes nests in all of the 17 new locations, which strongly supports our hypothesis of P. rufipes’ preference for canopy gaps in nest site selection.