Boreal forest ecosystem characterization at site and landscape scales using multispatial resolution remote sensing data

Detailed forest ecosystem classifications have been developed for large regions of northern Ontario. These ecosystem classifications provide tools for ecosystem management that constitute part of a larger goal of integrated management of forest ecosystems for long-term sustainability. These classification systems provide detailed stand-level characterization of forest ecosystems at a local level. However, for ecological approaches to forest management to become widely accepted by forest managers, and these tools to be widely used, methods must be developed to characterize and map or model ecosystem classes at landscape scales for large regions.;In this study, the site-specific Northwestern Ontario Forest Ecosystem Classification (NWO FEC) was adapted to provide a landscape-scale (1:20000) forest ecosystem classification for the Rinker Lake Study Area located in the Boreal Forest north of Thunder Bay, Ontario. Multispatial resolution remote sensing data were collected using the Compact Airborne Spectrographic Imager (CASI) and analysed using geostatistical techniques to obtain an understanding of the nature of the spatial dependence of spectral reflectance for selected forest ecosystems at high spatial resolutions. Based on these analyses it was determined that an optimal size of support for characterizing forest ecosystems (i.e., optimal spatial resolution), as estimated by the mean ranges of a series of experimental semivariograms, differed based on (i) wavelength; (ii) forest ecosystem class (and at low altitude as a function of mean maximum canopy diameter (MMCD)); and (iii) altitude of the remote sensing system. In addition, maximum semivariance as estimated from the sills of the experimental semivariograms increased with density of understory.;Based on the estimates for optimal spatial resolutions for six landscape-scale forest ecosystem classes, a series of spectral-spatial features were derived from the high-altitude CASI data (4 metre spatial resolution) using spatial averaging. Linear discriminant analysis for various spectral-spatial and texture feature combinations indicated that a spatial resolution of approximately 6 m was optimal for discriminating the six-landscape scale ecosystem classes. Texture features, using second-order spatial statistics that were derived from the 4 m remote sensing data, also significantly improved discrimination of the classes over the original 4 m data. Finally, addition of terrain descriptors, particularly elevation within a local region, improved discrimination of the six landscape scale ecosystem classes. It has been demonstrated that in a low-relief boreal environment, addition of textural and geomorphometric variables to high-resolution CASI reflectance data provides improved discrimination of forest ecosystem classes. Although these improvements are statistically significant, the absolute classification accuracies are not at levels suitable for operational classification and mapping.;The analysis presented here represents the initiation of a complex modelling approach that is necessary for improving forest ecosystem characterization and prediction using additional primary datasets and derived datasets that possess various levels of measurement. Not only are optimal or multispatial resolution remote sensing data required, but also appropriately scaled terrain and landscape features depicting soil texture, nutrient and moisture regimes. Incorporation of these types of terrain-specific variables with reflectance data should provide further improvement in forest ecosystem classification and modelling at landscape scales.