Across boreal forests and resource rich areas, human-induced change is rapidly occurring at various spatial scales. In the past, satellite remote sensing has provided a cost effective, reliable method of monitoring these changes over time and over relatively small areas. Those instruments offering high spatial detail, such as Landsat Thematic Mapper or Enhanced Thematic Mapper (TM or ETM+), typically have small swath widths and long repeat times that result in compositing intervals that are too large to resolve accurate time scales for many of these changes. Obtaining multiple scenes and producing maps over very large, forested areas is further restricted by high processing costs and the small window of acquisition opportunity. Coarse spatial resolution instruments such as the Moderate Resolution Imaging Spectroradiometer (MODIS) or the Advanced Very High Resolution Radiometer (AVHRR) typically have short revisit times (days rather than weeks), large swath widths (hundreds of kilometres), and in some cases, hyperspectral resolutions, making them prime candidates for multiple-scale change detection research initiatives. <p>In this thesis, the effectiveness of 250m spatial resolution MODIS data for the purpose of updating existing large-area, 30m spatial resolution Landsat TM land cover map product is tested. A land cover polygon layer was derived by segmentation of Landsat TM data using eCognition 4.0. This polygon layer was used to create a polygon-based MODIS NDVI time series consisting of imagery acquired in 2000, 2001, 2002, 2003, 2004 and 2005. These MODIS images were then differenced to produce six multiple-scale layers of change. Accuracy assessment, based on available GIS data in a subregion of the larger map area, showed an overall accuracy as high as 59% with the largest error associated with change omission (0.51). The Cramers V correlation coefficient (0.38) was calculated using the GIS data. This was compared to the results of an index-based Landsat change detection, Cramers V=0.67. This thesis research showed that areas greater than 15 hectares are adequately represented (approximately 75% accuracy) with the MODIS-based change detection technique. The resulting change information offers potential to identify areas that have been burned or extensively logged, and provides general information on those areas that have experienced greater change and are likely suitable for analysis with higher spatial resolution data.

Across boreal forests and resource rich areas, human-induced change is rapidly occurring at various spatial scales. In the past, satellite remote sensing has provided a cost effective, reliable method of monitoring these changes over time and over relatively small areas. Those instruments offering high spatial detail, such as Landsat Thematic Mapper or Enhanced Thematic Mapper (TM or ETM+), typically have small swath widths and long repeat times that result in compositing intervals that are too large to resolve accurate time scales for many of these changes. Obtaining multiple scenes and producing maps over very large, forested areas is further restricted by high processing costs and the small window of acquisition opportunity. Coarse spatial resolution instruments such as the Moderate Resolution Imaging Spectroradiometer (MODIS) or the Advanced Very High Resolution Radiometer (AVHRR) typically have short revisit times (days rather than weeks), large swath widths (hundreds of kilometres), and in some cases, hyperspectral resolutions, making them prime candidates for multiple-scale change detection research initiatives. <p>In this thesis, the effectiveness of 250m spatial resolution MODIS data for the purpose of updating existing large-area, 30m spatial resolution Landsat TM land cover map product is tested. A land cover polygon layer was derived by segmentation of Landsat TM data using eCognition 4.0. This polygon layer was used to create a polygon-based MODIS NDVI time series consisting of imagery acquired in 2000, 2001, 2002, 2003, 2004 and 2005. These MODIS images were then differenced to produce six multiple-scale layers of change. Accuracy assessment, based on available GIS data in a subregion of the larger map area, showed an overall accuracy as high as 59% with the largest error associated with change omission (0.51). The Cramers V correlation coefficient (0.38) was calculated using the GIS data. This was compared to the results of an index-based Landsat change detection, Cramers V=0.67. This thesis research showed that areas greater than 15 hectares are adequately represented (approximately 75% accuracy) with the MODIS-based change detection technique. The resulting change information offers potential to identify areas that have been burned or extensively logged, and provides general information on those areas that have experienced greater change and are likely suitable for analysis with higher spatial resolution data.