Files
Abstract
With the increased capabilities of satellite sensors in recent years, there has been acute interest among researchers to utilize high-resolution satellite data in projects requiring land cover mapping. In Chapter 1 of this thesis, I investigated the advantages and limitations of three land cover classification approaches: supervised/maximum likelihood classifier, unsupervised/ clustering classifier and hierarchical/rule-based knowledge classifier (HRB). I found that the two traditional approaches (supervised/unsupervised) overall performed poorly, but that each of the classification methods had notable strengths when classifying a high-resolution image. In Chapter 2, I compared the use of different data sources in bird presence/absence habitat models. I compared the performance of models that used high-resolution (<4 m) satellite data to models that used a coarser resolution (~30 m) data. I found that the results depend upon the species modeled, the scale of the features with which that species is associated and whether or not that data type can capture those features adequately.