Effects of disturbance and landscape structure on the biodiversity value of remnant tropical forest corridors

The study will be based on community-wide quantitative inventories of plants (trees and woody lianas >5 cm in diameter at breast height), diurnal birds, and non-flying mammals at 30 spatially independent remnant riparian corridors and 10 riparian sites embedded within large tracts of continuous forest. As many as 10 plots will be sampled at each site, but this will unavoidably depend on corridor patch size. All sites will be pre-selected using a georeferenced 2006 Landsat image (positional error <10m) within a 3,500 km2 landscape polygon including both the fragmented matrix and continuous forests. To control for differences in source populations of area-sensitive species, all corridors will be connected to forest patches of at least 1,000 ha, but sampling site location will explicitly take into account distance to these patches. During a previous study, we developed a collaborative network of 154 landowners within a 80-km radius of Alta Floresta, which will greatly facilitate access to survey sites and survey logistics. Patterns of species diversity and species composition will be quantified and then compared with previous surveys on the same taxa in variable-sized forest patches in the same region (Michalski & Peres 2005, in press, Peres & Michalski 2006, Lees & Peres in press). Bird and mammal surveys will include diurnal line-transect censuses, surveys of mammal tracks, camera trapping, armadillo burrow counts, limited-radius point-counts. Floristic inventories will be based on 0.25-ha (10 x 250m) plots. To reconstruct changes in landscape structure in the last decade, all sites will be GPS-located and then related to a biannual time series of Landsat images (1996-2006, pixel size = 15 m in recent images). Environmental habitat variables obtained in site will include tree density, forest basal area, tree species composition, canopy cover (using digital hemispherical photographs), understorey density, and vertical stratification of foliage. Plant identification will be conducted by a collaborating botanist, aided by a reference collection of plant vouchers that is already available at the local UNEMAT herbarium. In addition, vertical corridor profiles along sections of 50 m in length will be assessed using digital photos taken from neighbouring pasture sites (two photos per plot along both sides of the corridor). Forest profile photos will be thresholded and processed using a dedicated software that we have compiled. The resulting patch metrics (i.e. corridor height, area of vertical profiles) will then be related to the forest habitat quality of the same sites using pixel-scale multi-spectral data from recent Landsat images, which provides an aerial view of the horizontal plane of the same corridors. Most corridor sites are part of the working countryside and have been subjected to a recurrent history of different forms of anthropogenic forest disturbance. Willing local informants from neighbouring households (large cattle ranches and smallholders) will thus be interviewed to recover the history of structural and non-structural disturbance at each site, including selective logging, wildfires, cattle grazing, and game hunting. Corridor disturbance history will be further verified using an image-processing technique that relies on a gap fraction index (the Normalized Difference Fraction Index, NDFI) which combines the green and non-photosynthetic vegetation, soil and shade fractions derived from SMA (spectral mixture analysis) to enhance the detection of anthropogenic canopy disturbance (Souza et al. 2005). Data analysis of community-wide properties of each assemblage, and the patterns of both occupancy and abundance of single species will be carried out primarily using generalized linear modelling and logistic regressions taking into account patch and landscape variables, as well as the history of disturbance within a site.