An analysis of large-scale forest cover disturbance in Canada (1998-2004) based on multi-temporal coarse resolution data

Abstract

A procedure was developed to map large-scale forest disturbances at annual, continental scales using 1-km resolution data from a combination of satellite sensors and ancillary spatial data. The method, dubbed Change Screening Analysis Technique (Change-SAT), creates a probability of change map using multiple logistic regression and multi-temporal change metrics. The probability map is converted to binary change map and a decision tree model applied to attribute the most likely cause of change among burning, harvesting, flooding, or defoliation. This paper presents the results of applying Change-SAT over Canada for the period 1998-2004. A variety of interesting change examples is demonstrated, including insect defoliation, flooding related to a hydroelectric project, and widespread damage and die-off resulting from drought and a snow/wind storm. Although the method is generally not well suited to providing quantitative estimates of change, it identifies large disturbances that can be investigated in greater detail based on field visits or higher resolution imagery.