The study was carried out in the Sharavathi River Basin (13°43´24 - 14°11´57N to 74°40´58 - 75°18´34E) located in the Central Western Ghats, Karnataka, India (Fig. 1). The river basin covers a total area of 1991.43 km² with the Western region receiving ~2500 mm of annual rainfall and the eastern region ~900 mm (ShivalingaMurthy 2008). The vegetation, similar to elsewhere in the Western Ghats, is a mosaic of evergreen forests, moist- and drydeciduous forests and scrub jungles interspersed with plantations of Acacia sp.

The study area provides habitat for 93 species of ants representing 36 genera (Narendra et al. in review). Of these we here focus on the nesting site characteristics of 13 species that represent the major functional groups in this region (Narendra et al. in review): Generalised Myrmicinae (represented by Myrmicaria brunnea Saunders, Crematogaster sp. 1, Pheidole sp. 2), Tropical Climate Specialists (Cataulacus taprobanae Smith, Oecophylla smaragdina (Fabr.)), Specialist Predators (Harpegnathos saltator Jerdon, Leptogenys processionalis (Jerdon), Pachycondyla rufipes (Jerdon)), Opportunists (Anoplolepis gracilipes (Smith), Paratrechina longicornis (Latr.), Technomyrmex albipes (Smith), Hot Climate Specialists (Meranoplus bicolor (Guérin-Méneville)) and Cryptic Species (Pheidologeton diversus (Jerdon)). Species designation to functional groups was carried out following Brown’s (2000) detailed distribution, biology and ecology of world ant genera.

We sampled along three 20 km transects along South, East and West sides of the reservoir and one 4 km transact along the North side of the reservoir (Fig. 1). The transect towards the North was short as it was close to the reservoir. At 4 km intervals along each transect we established three sampling plots, each measuring 30 x 30 m. These plots were along a mini-transect at 200 m intervals and set perpendicular to the main transact. This resulted in a total of 60 sampling plots, distributed across five forest types: scrub jungles, acacia plantations, dry deciduous forests, moist deciduous forests and evergreen forests.

Between 2000 and 2002 we located nest sites of the 13 focal ant species in each sampling plot. A systematic visual sampling was carried out at each plot during 09:00-11:00 h and 15:00-17:00 h which involved checking under tree bark, rotting logs and leaf litter. To increase our chances of locating the nest we set up terrestrial and arboreal bait traps and followed either the ant trail or individual foragers that were returning to the nest with food. Baits consisting of 70% honey, tuna fish and fried coconut were provided as both terrestrial and arboreal baits. Terrestrial baits were placed on the ground and the arboreal baits were tied to a tree at a height of two metres from the ground. The bait traps were laid at 07:00 h and retrieved at 17:00 h. The baits were checked once every 30 minutes and ants that had visited the bait were recorded and their nests were located. Presence or absence of nests of each of the 13 species was determined by a one-time sampling at each of the 60 plots. Ants collected from the two methods were sorted, cleaned in saltwater solution, preserved in 70% ethyl alcohol and identified using keys provided by Bingham (1903) and Bolton (1994). Scientific names are based on the current nomenclature (Bolton 1995) and were cross verified with the online ant database (Agosti & Johnson 2005). Specimens have been deposited at the Insect Museum, Centre for Ecological Sciences, Indian Institute of Science.

We used a single cloud-free multispectral satellite image acquired on 5 March 1999 (Path 97 – Row 63) captured by the Linear Imaging Self-scanning Sensor (LISS) onboard the Indian Remote sensing Satellite IRS-1D. Data were purchased from the National Remote Sensing Agency, Hyderabad, India. The image covers the entire study area. Bands 2 and 3 (VIS: 0.62 – 0.68; NIR: 0.77 – 0.86) were extracted and geo-referenced by means of ground control points, established during fieldwork using a GPS receiver. Both bands feature a spatial resolution of 23.5 m. Our use of the satellite imagery acquired before the sampling period is justifiable, since land cover in this region had not changed significantly between these two dates. Evidence for this comes from our analysis of 2002 satellite imagery for this region (unpublished results).

From these two bands we calculated the NDVI following the established formula :

NDVI = (NIR-RED)/(NIR+RED)

(Lillesand et al. 2004). Using NDVI as a surrogate for vegetation status in this region is justifiable since the terrain in this region is very hilly and NDVI is one of the vegetation indices that minimises topographic effects in vegetated areas (Lillesand et al. 2004). Its value ranges from –1 to +1; negative or near-zero values indicate nonvegetated areas (e.g. soil, water), while positive values represent vegetated areas.

We identified five major habitats, without using remote sensing data, to sample ants. Our first question was hence to assess the correlation between LISS-derived NDVI values at the visited ant nest sites and the five habitat types. Our motivation behind this was to validate the remotely sensed NDVI data as a descriptor of the predefined habitat types, before using it in support of our interpretation of the relationship between ant nest site choice and habitat type. To evaluate this correlation, we extracted the NDVI values from all nest sites using GIS software, grouped all ant nest sites according to their designated habitat type, and carried out an analysis of variance (ANOVA) of the corresponding NDVI values followed by a post-hoc Tukey test to assess whether differences in NDVI were significant between these groups.

Our second question was to find out whether the ant functional groups to which the 13 species belong establish their nests at locations characterised by different NDVI values. To test this we grouped ant nest sites according to their designated functional group (Brown 2000; also see Andersen 1995, 2000). We then carried out an analysis of variance (ANOVA) of the associated NDVI values, followed by a post-hoc Tukey test to assess whether differences in NDVI were significant between these groups.

Our third question was to test whether differences in NDVI values associated with ant nesting locations were distinct at the species level. To test this we carried out the same analysis as for the functional groups using nest site NDVI values grouped at species level.

Our fourth question was established in response to the results we obtained from the three analyses above. Most ant nests were found to be associated with NDVI values that corresponded well with the expected habitat type for the species (see results section for details). For one of the 13 species, Pachycondyla rufipes, we obtained nest site NDVI values that were surprisingly low and did not match the NDVI range of the habitat type in which the species occurred. We tested this against the NDVI values observed at all other nestsites.

To consolidate our interpretation of this result (see discussion section for details), we conducted an assessment of the robustness of P. rufipes prevalence in this NDVI range using independent data obtained from additional fieldwork. We randomly selected 25 locations from previously un-sampled regions with NDVI values in the same range in which we had found P. rufipes nests before (0.015 – 0.1779) and conducted a visual all-out search for the nests of this species at these sites. Due to topographical limitations only 17 of the 25 plots were visited. Next, we calculated the respective prevalence value for our original dataset, using only those locations, which featured NDVI values within the same range (0.015 – 0.1779). We then compared both values to assess whether the difference in prevalence was significant. Non-parametric tests were used when data was not normally distributed.