Full text
| |
| Author(s): |
Katia Maria Paschoaletto Micchi de Barros Ferraz
Total Authors: 1
|
| Document type: | Doctoral Thesis |
| Press: | Piracicaba. , gráficos, ilustrações, tabelas. |
| Institution: | Universidade de São Paulo (USP). Escola Superior de Agricultura Luiz de Queiroz (ESALA/BC) |
| Defense date: | 2004-09-23 |
| Advisor: | Luciano Martins Verdade |
| Field of knowledge: | Biological Sciences - Ecology |
| Indexed in: | Banco de Dados Bibliográficos da USP-DEDALUS |
| Location: | Universidade de São Paulo. Biblioteca Central da Escola Superior de Agricultura Luiz de Queiroz; t636.9323; F381d 83014 |
| Abstract | |
The Piracicaba river basin, like the whole southeastern Brazil, has been suffered landscape alterations that certainly influence distribution and abundance of vertebrates. Apparently, the capybara is one of the species that has been influenced by this process, since large groups can be observed in anthropogenic habitats, possibly due to the great availability of food, open areas, and the local extinction of large predators. The main goal of this study was to develop a predictive model of capybara spatial distribution in relation to the landscape of Piracicaba river basin in the Sao Paulo, Brazil. The present study had three steps: 1) Distribution modeling: SPIP model (weighted-Iayers overlay) and GARP model (genetic algorithm for rule-set prediction), both assessed by Geographic Information System (GIS); 2) Aerial videography: Characterization of physical environment and study sites location assumed as adequate to the species; and, 3) Terrestrial surveys: capybara distribution estimated by presence/absence Index of individuais and/or tracks in the study sites. 89 presence points and 66 absence points were used to calibrate and validate the models. The unclassified Landsat TM image, classified Landsat TM image, land uselland cover, digital elevation model, aspect, slope, curvature of terrain and water distance gradient were the environmental variables used to generate the models. The relative frequency of capybara presence was 57.42%, and the animais were observed at only 8.38% of the sites. Capybaras were associated mainly to the agricultural habitats, with lower slopes, nearby the stream network, and with strong human presence. 100% of presence was accurately predicted by the SPIP model, with 79.77% in areas with higher probability of occurrence. The area predicted by the SPIP model represented the 99% of the total basin area. The 79.96% of the predicted area had medium-high probability of occurrence with 67.53% in the agricultural areas. The predictive variables indicated by the GARP model to explain the capybaras spatial distribution were the unclassified Landsat TM image, digital elevation model, curvature of terrain, land uselland cover, and soil type. 44.04% of the total area had medium-high probability of capybaras occurrence, but 23.93% of the higher probability was sugar cane and 12.25% was pasture. Predictions of presence were highly significant (p < 0.001); however, predictions of absence were only marginally accurate. The inclusion of presence points in the GARP model calibration improved its performance, explaining the low type II error probability, and, consequently, the high accuracy (97%). The presence/absence Index was efficient for the modeling processo GARP was the most accurate model. However, it should be validated for other areas with different landscape attributes where the species is not as abundant or widespread. Predictive models of wildlife spatial distribution can be helpful for the decision-making process in management actions. (AU) | |