| Remote sensing is a technology that has been used for many years to generate land cover maps. These maps provide insight as to the landscape, and features that are on the ground. One way in which this is useful is through the visualization of forest cover types. The forests of New England have been notoriously difficult to map, due to their high complexity and fine-scale heterogeneity. In order to be able to better map these features, the newest satellite imagery available may be the best technology to use. Landsat 8 is the newest satellite created by a team of scientists and engineers from the United States Geological Survey and the National Aeronautics and Space Administration, and was launched in February of 2013.;The Landsat 8 satellite sensor is considered an improvement over previous Landsat sensors, as it has three additional bands: (1) a coastal/ aerosol band, band 1, that senses light in deep blue, (2) a cirrus band, band 9, that provides detection of wispy clouds that may interfere with analysis, and (3) a Quality Assessment band whose bits contain information regarding conditions that may affect the quality and applicability of certain image pixels. In addition to these added bands, the data generated by Landsat 8 are delivered at an increased radiometric resolution compared with previous Landsat sensors, increasing the dynamic range of the data the sensor can retrieve.;In order to investigate the satellite sensor data, a novel approach to classifying Landsat 8 imagery was used. Object-Based Image Analysis was employed, along with the random forest machine learning classifier, to segment and classify the land cover of the Coastal Watershed of southeastern New Hampshire. In order to account strictly for band improvements, supervised classification using the maximum likelihood classifier was completed, on imagery created: (1) using all of the original bands provided by Landsat 8, and (2) an image created using Landsat 8 bands that were only available on the previous Landsat sensor (Landsat 7). Once classification had been performed, traditional and area-based accuracy assessments were implemented. Comparison measures were also calculated (i.e. Kappa, Z test statistic). The results from this study indicate that, while using Landsat 8 imagery is useful, the additional spectral bands provided in the Landsat 8 Operational Land Imager (OLI) and Thermal Infrared Sensor (TIRS) do not provide an improvement in vegetation classification accuracy in this study. |
