Utilizing remotely sensed imagery and GIS for mapping ecological and social attributes in Sustainable Forest Management and rural livelihoods

The goal of this thesis is to utilize remotely sensed imagery and GIS for delineation of ecological information and social attributes in Sustainable Forest Management (SFM) and rural livelihood improvement. The thesis focuses on two case studies: Case Study 1---Boreal Forest of Manitoba, Canada and Case Study 2---Tropical Rain Forest of Kalimantan Timur, Indonesia. The specific objectives of Case Study 1 are to provide SFM the required information at the ecosite level through the use of existing GIS inventory data and remotely sensed imagery. The second Case Study aims to develop a geomatics methodology to identify rattan in the rain forests of Kalimantan Timur using Radarsat imagery.;SFM planning requires an ecological approach to terrestrial and wetland ecosystem classification and mapping. However, within the existing Forest Resource Inventory (FRI) database critical ecological attributes required for the accurate delineation of ecosites are not available using traditional aerial photograph interpretation techniques. Remotely sensed imagery is examined to determine if it is able to provide understory ecoelements by overcoming scaling property problems. Principal Components Analysis (PCA) is preformed on multitemporal Landsat 7 ETM+ imagery to identify understory phonological changes at two different scales. A Canonical Correlation Analysis (CanCor) is used to quantify the relationship between optical interpreted understory imagery and Landsat ETM+ data. Utilization of a Digital Information Model (DEM) allows topology measurements to be made for the generation of six enduring landscape features: (1) Peak, (2) Ridge, (3) Pass, (4) Plane, (5) Channel and, (6) Pit. Identifying relationships between boreal tree and wetland species to their surrounding landform features enables mapping forested areas at the ecosite scale. A Canonical Correspondence Analysis (CCA) is used to determine relationships between FRI species and landform features.;The traditional livelihoods of the Dyak tribal people in Kalimantan Timur, Indonesia depend on a diverse income portfolio that includes raw rattan as a significant component. The government of Indonesia has placed a ban on the export of rattan due to a perceived shortage. This policy was not supported by a quantitative analysis of the rattan stock as there are currently no tools to provide accurate estimates. Continual cloud cover make it nearly impossible to utilize optical remote sensing imagery in isolated tropical regions while low level aerial photo acquisition is simply too expensive. Radarsat-1 imagery possesses cloud penetrating ability as it utilizes microwave radiation and is relatively inexpensive. However, image speckle is inherent in Synthetic Aperture Radar (SAR) data making it difficult to interpret. Five filtering techniques are evaluated using varying kernel sizes to determine which algorithm reduces speckle within Radarsat-1 imagery and maintains spatial properties of dry rice fields (Ladang) within Kalimantan Timur, Indonesia. Multiple Discriminate Analysis (MDA) and PCA are presented as a quantitative evaluation of filter speckle suppression based on segmentation and data recovery ability at varying kernel sizes for a multitemporal multi-incident angle SAR dataset.;For Case Study 1, new geomatic techniques were able to delineate ecological attributes including understory characteristics and enduring landform features. This information combined with the existing FRI layer may be used to identify ecosites in a Decision Support System (DSS) for ecosite delineation. Once ecosites can be physically identified, the boreal forest can be managed in a spatial and ecological manner. For Case Study 2, the Gamma filter at the 11 x 11 kernel sized proved to be the most effective filtering technique to remove inherent speckle from SAR imagery. Identification of rattan stock may now take place to generate an accurate inventory in Kalimantan Timur, Indonesia. With proper management of this essential resource the rural livelihoods of the Dyak tribal people should be improved. The filtering evaluation methodology may also be used to determine the optimal filter for other small scale cultural features in tropical environments when using SAR imagery.