Kardakov, A., Kiviste, A. and Peterson, U. 2009. Assessment of Driving Factors Affecting Radiance Changes in Regenerating Mixed Forest Clearcuts Using Landsat Winter Imagery.   Baltic Forestry, 15 (2): 216-225

This paper describes an approach for estimating the effect of stand variables on radiance of forest patches that were regenerating after the disturbance of clearcut felling. Analysis was performed on a difference image of a two-date Landsat Thematic Mapper (TM) image pair. Landsat TM images from late winter made in plain snow cover conditions – a non-traditional season for forest mapping – were used in the study. Seasonal snow cover in winter in boreal and hemi-boreal latitudes presents specific conditions with high target (stems and branches of shrubs and trees) to background radiance contrast and is therefore a well-timed season for the detection of subtle changes in shrub and tree cover in patch areas regenerating after disturbance.
The results of the study show that, in all spectral bands studied (TM1–TM4), there is a significant effect of patch age (i.e.) time passed since disturbance of clearcut logging, on the radiance difference of regenerating patches on a difference image. Radiance change in none of the Landsat TM bands 1–4 was found more sensitive than others to patch variables. The marginal effect of Spruce index, a binary variable denoting whether the patch was managed as a young spruce stand or a deciduous stand, means that the chance for discrimination of young spruce-dominated stands from deciduous-dominated stands in winter images is low in most cases. No significant effect of stand variables on patch-wise differences was found between the patch areas classified from difference image to that represented in the forestry database. There was a significant effect of the type of neighbours – either bright or dark – on the estimation of the patch area on a difference image.

Key words: forest change detection, Landsat TM winter images, clearcut regeneration