Document Type

Thesis - Open Access

Award Date

2006

Degree Name

Master of Science (MS)

Department / School

Geography

Abstract

High spatial resolution Landsat Thematic Mapper (TM) and Enhanced Thematic Mapper Plus (ETM +) bands 4, 5, and 7 coupled with pre-existing Moderate Resolution Imaging Spectroradiometer (MODIS) 250m resolution maps as reference information were employed in an attempt to catalog anthropogenic impacts on the Congo Ba in's humid tropical forest. The utility of Landsat multi-spectral data depends both on the extent to which surface properties can be estimated from the radiometric measurements and on the ability to observe the surface through the atmosphere. Cloud cover is a major impediment to optical remote sensing of humid tropical regions. Accordingly, applications may require cloud-free portions from various scene dates for each vegetation map or for each of the times that bound change assessment. The variously dated scenes or cloud-free scene portions that might constitute an image mosaic will differ in atmospheric conditions, phenological state, sun-target-sensor geometry, and sensor calibration. These differences cause the relationships between land-cover classes and pixel brightness values to vary across space over a mosaic period (the time period spanning the cloud-free scenes or scene parts that compose an image mosaic). Consequently, radiometric normalization of the scenes to be used for the change analysis is indispensable. This thesis investigates the applicability of MOD IS 250m resolution tree cover map in performing relative radiometric standardization of Landsat scenes used for the Congo Basin's forest change mapping. The relative radiometric normalization protocol discussed in this thesis uses the pre-existing MODIS 250m resolution tree cover map coupled with Landsat thermal band threshold and mean value bias adjustment to radiometrically standardize the Landsat TM and ETM+ bands 4, 5, and 7. Under this normalization protocol, it was assumed that the evergreen humid tropical forest is a seasonal or pseudo-invariant or constant reflector across the scenes. The MO DIS tree cover map and the thermal band threshold bitmap were intersected using the logical "and" operator to flag or identify the normalization target, the evergreen forest pixels in the Landsat imagery. Using the normalization target (the evergreen forest pixels) as a mask, a histogram was generated for each band to obtain the mean value. The mean value of each band under the forest mask was then adjusted to an arbitrary constant. A distribution of ±2 standard deviations was used to normalize the Landsat TM and ETM + scenes to a common radiometric scale. Scatter plots generated under the forest mask for the normalized and the non-normalized images demonstrated that the relative radiometric normalization protocol using the dark target (the evergreen tropical forest) reduced the radiometric variability among the TM and ETM + bands. The implementation of the protocol has a consequence of shifting the scatter plot to lie on a one-to-one line. Besides scatter plots to evaluate the performance of the normalization procedure, a statistical spectral separability measure, Bhattacha1.1Ja distance, was used to quantitatively appraise the impact of the dark target on feature selection among the raw digital numbers (DN), the at-satellite top of atmosphere reflectance (TOA), the normalized DN and the normalized TOA images. The spectral separability assessment was conducted using the MODIS 250m resolution forest and nonforest cover maps as spectral signatures. It was determined that the use of the evergreen humid tropical forest as a normalization target improved the ability to extract information than would have been the case by just using the raw DN and the TOA images.

Library of Congress Subject Headings

Remote-sensing images -- Data processing

Forests and forestry -- Congo (Democratic Republic) -- Remote sensing

Forest canopies -- Congo (Democratic Republic)

Landsat satellites

Format

application/pdf

Number of Pages

81

Publisher

South Dakota State University

Share

COinS