GEOSCAN Search Results: Fastlink


TitleAssessment of convolution neural networks for wetland mapping with landsat in the central Canadian boreal forest region
AuthorPouliot, D; Latifovic, R; Pasher, J; Duffe, J
SourceRemote Sensing vol. 11, 7, 772, 2019., Open Access logo Open Access
Alt SeriesNatural Resources Canada, Contribution Series 20190631
Mediapaper; on-line; digital
File formatpdf
Subjectsgeophysics; remote sensing; vegetation
ProgramCanada Centre for Remote Sensing Divsion
Released2019 03 31
AbstractThe Environment and Sustainable Development Indicators (ESDI) Initiative was introduced to track Canada's overall wealth in the form of natural and human capital, in additionto familiar economic data such as the gross domestic product (GDP). One of the six ESDIs is the Forest Cover Indicator (FCI). In this paper we define FCI, outline the overall method for deriving FCI, and report results for addressing four key technical issues in carrying out this overall method. The FCI is defined as interannual variations of Canada's forest area with the middle-summer crown closure (CC) ? 10%. Crown closure is the percentage of the ground surface covered by a downward vertical projection of the tree crowns. Theoverall monitoring method is mainly based on coarse resolution remote sensing data because of the need to cover Canada's extensive landmass during the middle-summer months and toupdate the results annually. Medium resolution satellite data, field measurements, and modeling approaches were used for calibration, correction, validation, and down-scaling, with a focus on the following 4 key technical issues: (1) correcting understory non-tree vegetation effect on CC, (2) downscaling forest cover area from 1-km to 100-m spatial resolution as required by the FCI definition, (3) detecting the changes of CC caused by disturbances, and (4) detecting changes in CC caused by forest regrowth. Methods and results for addressing these technical issues are described in the paper. While these results indicate that the key technical issues can be solved by integrating satellite remote sensing data/products and other data, there are clear needs for further development, especially testing against field measurements. © 2006 Rocky Mountain Mathematics Consortium.

Date modified: