Date of Award

Fall 12-1-2003

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Geography, Geology, and Anthropology

First Advisor

Paul Mausel

Second Advisor

Robert Larson

Third Advisor

Brian Ceh

Abstract

There has been extensive use of remote sensing to study secondary succession of moist tropical forests in the Amazon Basin. However, scientists have not totally understood the dynamic processes of secondary succession of Amazon forests. A lack of consensus remains about what is the best approach for biomass estimation in the Amazon Basin using remotely sensed data. It is not clear which spectral wavelength bands have the best relationship with biomass values and what are the specific relationships between biomass and TM satellite radiance. The strategies in this research, based on analysis of detailed ground-based data and satellite TM image data, employ integrated GIS and remote sensing technology to identify the dynamic processes of secondary succession at a subregional scale in the Altamira area. A non-linear regression biomass-spectral model was successfully developed and implemented based on detailed field data and using 1991 TM data. Residual analysis showed that the non-linear model had good representation of population based on the sampled data. The model was also applied to estimate biomass values of secondary succession using 1985 and 1988 TM spectral data informed by the regression model. Estimated biomass values were classified into several distinct clusters using cluster analysis technology for spatial and temporal analysis. Spatial and temporal patterns of secondary succession were processed using GIS overlay analysis and feature classification technology based on the estimated classified biomass in the 1985-1991 TM time series used. GIS Feature classification is important in this study because it makes the spatial and temporal analysis based on vegetation charactesistics possible. GIS overlay technology is an efficient way in spatial and temporal analysis. IV Research results indicate that there are significant non-linear relationships between satellite TM spectral response and biomass values in the Altamira moist mature forest areas, especially in middle-infrared bands (MIR). Both MIR band 5 and band 7 had an inverse relationship with estimated biomass at their lower and higher spectral response. The TM middle-infrared band 5 was found to have the best relationship with biomass in the study area. It is found that the Cubic model with Band 5 as independent variable had the highest R squared values: 0.876. This research indicates that the optimal number of distinct secondary succession stages was three in number based primarily on the biomass characteristics of vegetation stand parameters. The regression model performs best for biomass estimation when spectral data are in the range of the model independent data. This research also found that extending the model developed to 1985 and 1988 to estimate biomass using 1991 TM modeled results as a base was unsuccessful because vastly different weather conditions existed in the months prior to 1985 and 1988 TM data acquisition compared to 1991. This condition affected overall vegetation vigor and spectral responses that could not be standardized using an image-based DOS atmospheric correction approach. Multi temporal modeling of biomass through signature extension of one date of remotely sensed data to another wi11 require additional modeling that incorporates weather condition impacts prior to data acquisition.

Download

Share

COinS