Canadian Forest Service Publications
Effect of topographic correction on forest change detection using spectral trend analysis of Landsat pixel-based composites. 2016. Chance, C.M.; Hermosilla, T.; Coops, N.C.; Wulder, M.A., White, J.C. International Journal of Applied Earth Observation and Geoinformation. Vol. 44, pp. 186-194.
Issued by: Pacific Forestry Centre
Catalog ID: 36514
Availability: PDF (download)
Available from the Journal's Web site. †
† This site may require a fee
tPixel-based image compositing enables production of large-area surface reflectance images that arelargely devoid of clouds, cloud shadows, or haze. Change detection with spectral trend analysis usesa dense time series of images, such as pixel-based composites, to quantify the year, amount, and magni-tude of landscape changes. Topographically-related shadows found in mountainous terrain may confoundtrend-based forest change detection approaches. In this study, we evaluate the impact of topographiccorrection on trend-based forest change detection outcomes by comparing the amount and location ofchanges identified on an image composite with and without a topographic correction. Moreover, weevaluated two different approaches to topographic correction that are relevant to pixel-based imagecomposites: the first corrects each pixel according to the day of year (DOY) the pixel was acquired, whilstthe second corrects all pixels to a single reference date (August 1st), which was also the target date forgenerating the pixel-based image composite. Our results indicate that a greater area of change is detectedwhen no topographic correction is applied to the image composite, however, the difference in change areadetected between no correction and either the DOY or the August 1st correction is minor and less than 1%(0.54–0.85%). The spatial correspondence of these different approaches is 96.2% for the DOY correctionand 97.7% for the August 1st correction. The largest differences between the correction processes occurin valleys (0.71–1.14%), upper slopes (0.71–1.09%), and ridges (0.73–1.09%). While additional tests underdifferent conditions and in other environments are encouraged, our results indicate that topographiccorrection may not be justified in change detection routines computing spectral trends from pixel-basedcomposites.
Plain Language Summary
In this manuscript, we review and synthesize recent literature on four key remote sensing technologies: airborne laser scanning (ALS), terrestrial laser scanning (TLS), digital aerial photogrammetry (DAP), and high spatial resolution (HSR)/very high spatial resolution (VHSR) satellite optical imagery. Our intention is to demonstrate the capacity of these technologies in the context of forest inventory and synthesize some of the issues and opportunities associated with the current and future application of these technologies.