The results of this study show that in the Western Ghats, some ant species and even some functional groups establish their nests at locations that can be clearly distinguished using LISS-derived NDVI. However, the discriminative success of NDVI was limited to ant species that nest in areas with low NDVI, i.e. in relatively sparse vegetation. The reason for the failure of NDVI to distinguish ant nesting locations at the species (and functional group) level when they are located in dense vegetation is probably that in these cases, ant species do share similar habitat in terms of greenness and / or biomass: H. saltator and L. processionalis are specialist predators on other arthropods and known to inhabit deciduous and evergreen forests; O. smaragdina and C. taprobanae are arboreal ants that are dominant in humid tropical regions, and P. diversus is a cryptic species that nests and forages within soil and leaf litter (Figs. 3, 4). To distinguish nesting sites of such species it might therefore be necessary to use additional variables.

A potential limitation of this study is that our interpretation of the relationship between ant nest sites and habitat type was based on nest-site NDVI only. Although we complemented this with a correlation analysis of NDVI and the pre-defined habitat types, a direct assessment of speciesspecific nest site frequency vs. habitat type is missing. Regardless of this shortcoming, our purely NDVI-based analysis results match the documented habitat preferences of species inhabiting areas with less dense vegetation quite well: invasive species such as A. gracilipes and P. longicornis along with T. albipes were abundant in scrub jungles and acacia plantations that have low NDVI (Fig. 4). All three species are Opportunists that exhibit unspecialised food and niche requirements, are poorly competitive, and are dominant in disturbed habitats (Andersen 1995). The seed harvesting ant M. bicolor was abundant in regions with low NDVI such as scrub jungles. On the other hand, a particular wide range of NDVI niches were occupied by the ubiquitous species of the myrmicine community (Crematogaster sp. 1, Pheidole sp. 2 and M. brunnea) that do not have highly specific niche requirements (Fig. 4).

A surprising finding was that P. rufipes, a specialist predator on termites (Narendra & Kumar 2006), built nests at sites with low NDVI despite the fact that this species was collected only from deciduous and evergreen forests (Fig. 4). Similar observations on the niche occupied by this ant species have been reported (Narendra & Kumar 2006; Narendra et al. in review). NDVI at the collection sites of P. rufipes was similar to that in the scrub jungles (Fig. 3), a habitat from which this species was never collected. And although our validation fieldwork confirmed the initial observation that both nesting sites and foragers of P. rufipes coincided with canopy gaps of deciduous and evergreen forests, the question remains: why is P. rufipes found in canopy gaps?

Individually foraging ants rely on direct-ional information gathered either from celestial cues (Wehner 2001) or from landmarks present in the foraging environment (Fukushi 2001; Narendra 2007). Solitary foraging ants are well known for their ability to return to the nest by matching the previously seen views (Wehner & Räber 1979; Narendra et al. 2007). In fact a congeneric species of P. rufipes, Pachycondyla tarsata (Fabr.) (previously known as Paltothyreus tarsatus) uses the contrast available in the canopies (Hölldobler 1980) to match its previously acquired image to its current images to return to the nest. In the landmarkrich habitats of P. rufipes it is quite unlikely for individual trees to act as a beacon and utilisation of celestial cues may be hindered. It is perhaps because of this that P. rufipes colonises canopy gaps, a micro-niche that would enable the ants to forage using information derived from both canopies and sky.

The example of P. rufipes demonstrates the potential of high-resolution remotely sensed NDVI data in delineating preferred nesting sites for species whose habitat preferences are clearly different from those of other ant species in the study area, in terms of both density and geometry of the local vegetation. It does not however shed much light on the potential of NDVI as a predictor variable to model ant species distributions, as this would require validating results for each species at random locations from anywhere in the study area and not only from within the high-probability range. Also, if NDVI was to be tested as predictor variable for presence of P. rufipes nests, we would recommend to not only employ the absolute value but also a derived variable that quantifies the difference in NDVI between neighbouring pixels, i.e. taking into account the “canopy gap” as an argument for habitat suitability. We emphasise that our study did not intend to validate the pre-defined habitat categories using remotely sensed NDVI; instead, we merely assessed the correlative strength between these two at the selected nest sites. It would be interesting however, to use the LISS-derived NDVI image to establish habitat types for the whole study area (e.g. by means of a supervised classification) and then explore the relationship between ant nest site locations and habitat type at both the species and the functional group level.

We conclude that LISS-derived NDVI has considerable value in deriving the nest site locations of some ant functional groups and even at species level, especially regarding ant species that belong to the functional groups Hot Climate Specialists and Opportunists. These results are encouraging for decision-makers dealing with invasive species, which are often opportunistic. Officers in the land-use and conservation sector who need to monitor the effects of intensifying human land-use and climate change stand to benefit as well, since many ant species are considered reliable indicators for ecosystem change