Information about ecological and geographical distribution of species is essential to understand spatial patterns of biodiversity and to chalk out robust conservation strategies (Rushton et al. 2004; Graham et al. 2006). However, for most species, the number and spatial density of confirmed occurrences is very low. Hence ecologists use species distribution models in an attempt to provide predictions of distribution of species over large areas by relating presence/ absence or density of flora and fauna to environmental predictors (Elith et al. 2006). These species distribution models are typically built using the ecological niche concept (Hutchinson 1957). An ecological niche of a particular species is defined by long-term stable constraints on the potential of its geographic distribution such as habitat suitability. Ecological niche modeling 52 Remote detection of ant nesting locations techniques quantify and exploit such constraints (Peterson 2003) and comprehensive reviews of the available species distribution modeling approaches have been provided by Guisan & Zimmermann (2000) and Elith et al. (2006). Aggregating results of several species distribution models can improve our understanding of the relationship between environmental parameters and species richness (MacNally & Fleishman 2004). Other potential application areas include research into the ecological and geographical differentiation of distribution of congeneric species (Cicero 2004; Graham et al. 2004), and investigations of the invasive potential of non-native species (Peterson 2003; Goolsby 2004; birds: Peterson & Vieglais 2001; Fuller et al. 2007; plants: Panetta & Dodd 1987; insects: Eyre et al. 2004; Roura-Pascual et al. 2006). Regardless of the intended application area and the choice of modeling algorithm model reliability is determined to a large degree by the quality of the input data. The present study intends to augment our knowledge in this field by exploring the potential of the Normalised Difference Vegetation Index (NDVI) data, derived from the Linear Imaging Self-scanning Sensor (LISS) on the Indian Remote sensing Satellite IRS-1D, to distinguish between habitat types used by a variety of ant species to build their nests. This analysis is performed at (a) the level of functional groups and (b) species level. Furthermore, we use these remotely sensed data to investigate the prevalence of Pachycondyla rufipes nest sites within the area identified as suitable in terms of NDVI, which supports our interpretation of this species’ peculiar choice of nesting site.

The rationale for investigating the potential of NDVI as a discriminatory variable for ant nest site allocation is the assumption that ants consider vegetation characteristics when establishing nests, and NDVI is one of the commonly used vegetation indices. NDVI is calculated from the reflectance values in the red and near infrared electromagnetic spectrum. NDVI thus quantifies chlorophyll activity of plants by relating the absorption of light at wavelengths of around 0.6 – 0.7 μm (red) to the reflection of light at wavelengths of around 0.7 – 0.9 μm (near infra-red). Vegetation with high chlorophyll activity is characterized by large NDVI values, which may indicate high greenness, high biomass or both. After taking into account subvegetation ground reflectance properties and some other additional factors, a basic delineation of vegetative from non-vegetative land cover features may also be achieved (Lillesand et al. 2004).

The use of LISS-derived NDVI data allows us to exploit two major advantages of remotely sensed data. First is the fast capture and consistent representation of NDVI values across a fairly large area, especially when temporal variability is taken into account, e.g. by conducting a time series analysis. Second, available remote sensing data continues to increase in spatial resolution, which is an essential prerequisite for building models at a scale appropriate for the studied organism. LISSderived NDVI products feature an acceptable fine spatial resolution for the purposes of this study.

NDVI has been used as predictor variable in several studies: to determine the extent of vegetation cover (Narendra 2000), to monitor productivity and health of an ecosystem (Zhang et al. 1997; Ikeda et al. 1999), to determine outbreaks of insects (Leckie et al. 1992), to map crop conditions of agricultural fields (Brewster et al. 1999) and to characterise the structure of forest canopies (Gamon et al. 1995; Wannebo & Rosenzweig 2003). NDVI has previously been used in determining the ecological habitat requirements of an invasive ant species, and its correlation with ant presence was found to be quite consistent (Roura-Pascual et al. 2006). However, in the above study NDVI was used in conjunction with another vegetation index (EVI) and topographical variables, i.e. the individual potential of NDVI was not assessed. In this study, we analyse the relation between LISS-derived NDVI alone and the location of nesting sites of different ant species.