Introduction
Leafcutter ants, genera Atta and Acromyrmex, are considered important ecosystem engineers in the Neotropics (Fowler et al 1989; Hölldobler and Wilson 1990). They are also of great relevance as key herbivores during successional processes (Vasconcelos and Cherrett 1997; Vasconcelos 1997) and as pest species in planted forest such as agroforestry and restoration programs, and agricultural land (Della-Lucia 2003). Their nests are heterogeneously distributed (Jaffe and Vilela 1989; Lapointe et al. 1998; Schoereder 1998; Meyer et al. 2009). For example, Atta cephalotes (L., 1758), Atta laevigata (F. Smith, 1858) and Atta colombica (Guérin-Méneville, 1844) nests in rainforests are more frequently found in vegetation gaps or near edges (Jaffe and Vilela 1989; Farji-Brener and Illes 2000; Peñaloza and Farji-Brener 2003; Wirth et al. 2003; Urbas et al. 2007; Wirth et al. 2007; Meyer et al. 2009; Silva et al. 2009), which are characterized by pioneer plant species (Turner 2004).
The palatable forage hypothesis argues that leafcutter ants are favored by these pioneer plant species due to their lower chemical and higher nutritional content (Farji-Brener 2001). Many rainforest Atta species also actively create gaps by removing all vegetation on and around the nest (Gonçalves 1967). Gap creation, soil perturbation (Alvarado et al. 1981; Moutinho et al. 2003; Verchot et al. 2003) and deposition of organic waste and seeds on the soil surface (Haines 1978; Sternberg et al. 2007; Hudson et al. 2009), are the most important ingredients of the underlying mechanisms that make these Atta colonies key modifiers of rainforest vegetation structure and composition (Garrettson 1998; Hull-Sanders and Howard 2003; Peñaloza and Farji-Brener 2003; Farji-Brener 2005; Meyer 2008; Corrêa et al. 2010).
Contrary to site selection of most Atta species, Atta sexdens (Linneaus, 1758) nests in the Colombian Amazon were observed only in habitat types with high canopy cover (van Gils et al. 2010, 2011). This study aimed to evaluate whether nests within these habitat types are located under relatively higher (as seems to be suggested by van Gils et al. 2010, 2011) or lower (as is the case for A. cephalotes, A. laevigata and A. colombica) canopy cover. This study contributes to our knowledge of the impact of each leafcutter ant species on different, and especially human-disturbed, ecosystems (Wirth et al. 2003; Côrrea et al. 2010). A. sexdens has rarely been studied for its impact on the vegetation surrounding the nest.
Materials and Methods
Study area. This study was carried out in the community of Palmeras (3°48′40.9"S 70°17′53.3"W) and on the private farm Versalles (3°49′51.2"S 70°14'26.9″W), both located along the Colombian Amazon River. Mean annual rainfall in the study area is 3200 mm. There is a distinct dry season from June to September and a wet season from October to May. Average relative humidity is approximately 86% (Riaño 2003). The study area is part of the hilly sedimentary Amazon plain, which dates from the Tertiary (Riaño 2003). The soil is of marine and continental origin and mostly classified as Typic Hapludult (Ultisol) (Brady and Weil 2002). Natural vegetation cover is tropical rainforest, with Lecythi-daceae, Myristicaceae, Violaceae, Moraceae and Fabaceae as abundant plant families (IGAC 1997). Since the beginning of last century, human activity has also created patches of secondary forest, agricultural plots and housing sites in the area (Riaño 2003; Pinilla 2004). Three habitat types were included in this study: OF (old rainforest), AF (secondary rainforest - abandoned agricultural plots) and GF (secondary rainforest - abandoned grassland). In each habitat type, the location of active A. sexdens nests was determined directly for the year 2008. The locations were determined indirectly for 2006, 1999 and 1992 and they were mapped on satellite images (as described in van Gils et al. 2011). Only the 2006 and 1992 satellite images were considered sufficiently free of cloud formation to be used in the current study.
Canopy cover determination. In 2008, canopy cover was determined with a Robert E. Lemmon densiometer (Model A - Forestry Suppliers) (Englund et al. 2000). Measurements were carried out at each active A. sexdens nests in OF (N = 26), AF (N = 6) and GF (N = 24) (van Gils et al. 2011), and at an equivalent number of random sites in the same habitat types. Four measurements were taken towards the north, east, south and west on the most central nest mound or at each random site. The same procedure was repeated 15 m to the north, east, south and west. This way, a canopy area of approximately 45 x 45 m was covered. Twenty measurements were taken at each nest or random site. The average was multiplied by 1.04 to establish the average percent canopy cover.
To get a relative indication of vegetation cover on active A. sexdens nests and random sites without nests in 2006, ASTER satellite image AST_L1B_003_0429200615070_vnir. img was visualized in gray scale stretched mode of band ASTER_Band3N in the GIS software program ArcMap (ESRI Inc. 2006). For 1992, LANDSAT satellite image TM004_063_19922808 (ID: 015-784) was visualized in gray scale stretched mode of band Layer_5 in the same program. On these two images, all active nests of the respective years were projected. ASTER and LANDSAT satellite images visualized in gray scale stretched mode consist of pixels assigned with values according to light reflection. Pixel values, therefore, do not represent actual canopy cover percentages, but give a relative indication: Low canopy cover reflects more light and it causes brighter grey pixels with higher values. In contrast, high canopy cover reflects less light and causes darker grey pixels with lower values. A LANDSAT image pixel measures 45 x 45 m and an ASTER image pixel 15 x 15 m. On the satellite images, it was aimed to determine pixel values of a canopy area similar to that covered by the densiometer method (i.e. approximately 45 x 45 m).
On the LANDSAT image of 1992, the values were determined for only those pixels on which more than 50% of each nest symbol was located (N = 34 in OF; N = 13 in AF; N = 34 in GF - van Gils et al. 2011). In the same image, the values of an equal number of randomly located pixels in each habitat type were also determined. For the ASTER image of 2006, the values were determined of the pixel on which more than 50% of each nest symbol was located (N = 59 in OF; N = 24 in AF; N = 35 in GF - van Gils et al. 2011), plus the values of all eight surrounding pixels. This way, a canopy area of 45 x 45 m was covered. Then, the average of these nine pixels was calculated to obtain one value per nest. In the same image, the average values of an equal number of randomly located groups of such nine pixels in each habitat type were also determined.
Statistical analysis. Canopy cover percentages of 2008 and average pixel values of 2006 and 1992 were compared per habitat type and for all habitat types together with Kolmogorov-Smirnov tests carried out in statistics software program SPSS 15.0.1 (SPSS Inc. 2006).
Results
In 2008 the percent canopy cover in OF, GF and all habitat types together was higher at A. sexdens nests than at random sites; only in AF the difference was not significant (Table 1). In 2006 and 1992, nest pixel values were significantly lower in OF, GF and all habitat types together (Table 1). In AF, nest pixel values showed the same trend, but differences were not significant.
Discussion
Contrary to the nests of other common rainforest Atta species, this study showed that A. sexdens nests were consistently located under higher than average canopy cover in old rainforest (OF), abandoned grassland secondary forest (GF) and in general. The lack of a clear pattern and significant differences in abandoned agricultural plot secondary forest (AF) may be due to a relatively large variability in canopy cover in this habitat or to the low sample number. It seems unlikely that the higher canopy cover at nests was caused only by non-removal of vegetation from A. sexdens nest mounds (Gonçalves 1967). Although A. sexdens founding queens seem to prefer open spaces for nest establishment (Vasconcelos 1990), there may be some advantages to being located at more shaded micro-sites: Nests, young ones in particular, may be less conspicuous to natural enemies, and micro-climatic conditions may be more favorable (van Gils 2011). Atta worker ants frequently travel as far as 200 m to reach suitable forage (Eidmann 1932; Cherrett 1968). Therefore, high canopy cover micro-sites should not limit the access of ants to palatable pioneer plants.
Nests of other common rainforest leafcutter ant species, such as A. colombica and A. cephalotes, receive significantly more light due to the gaps they create overhead by removing vegetation from nest mounds and surroundings up to canopy level (Garrettson et al. 1998; Farji-Brener and Illes 2000; Hull-Sanders and Howard 2003; Wirth et al. 2003; Farji-Brener 2005; Meyer 2008; Côrrea et al. 2010). Partly due to this activity, both species strongly affect the surrounding vegetation structure and composition (Garrettson et al. 1998; Hull-Sanders and Howard 2003; Wirth et al. 2003; Farji-Brener 2005; Meyer 2008; Côrrea et al. 2010). However, colonization patterns and fine plant root growth of plants growing near nest mounds differ between nests of these two leafcutter ant species due to the above vs. underground location of their organic waste dumps (Farji-Brener and Medina 2000). Another study on A. sexdens nests in cerrado vegetation (Schoereder and Howse 1998) found that there they had no observable effect at all on vegetation structure. These studies clearly illustrate how the overall impact of each Atta species on the surrounding rainforest ecosystem is defined differently according to their particular behavior and nest-building characteristics (Fowler et al. 1989; Fowler and Claver 1991; Farji-Brener 2005; Côrrea et al. 2010). Contrary to A. colombica and A. cephalotes nests, those of A. sexdens were located under higher than average canopy cover. This implies that less light reaches the forest floor at nest locations, even after colony death. Furthermore, A. sexdens does not accumulate organic waste or seeds on the soil surface but in deep underground chambers and does not remove all vegetation from the nest mounds and surroundings (Eidmann 1932; Gonçalves 1967).
Conclusion
Increased light and nutrient availability and a bare forest floor are the principal ingredients for vegetation diversification and growth, especially for shade-intolerant plant species (Côrrea et al. 2010). However, none of these are attributes of A. sexdens nests, during or after colony demise. It seems that A. sexdens nests in human-disturbed rainforests do not provide a suitable niche at all for light-demanding, nutrient-limited, and small-seeded litter-gap demanding plant species (Côrrea et al. 2010). Therefore, general seedling germination and plant growth may even be reduced on and around these nests. The influence of A. sexdens on vegetation structure and composition is likely to differ from the influence of other rainforest Atta species. Further studies should elucidate the extent of A. sexdens engineering role in rainforest and other ecosystems.
