The situation with respect to the use of remote sensing for quantification of 
thinning removals is, at this time, quite different. No reliable, satellite-based 
method exists today for detecting thinnings or distinguishing between 
thinnings of different intensities. Areas thinned can be confused with areas 
subject to disease or pest damage according the work of Hame (1988). 
Using optical wavelengths with today’s satellite sensors, it is probably 
possible to separate heavily thinned stands (e.g., greater than 60 percent 
basal area removed) from other categories without the use of physical plant 
canopy illumination modelling. However, most thinning in Sweden occurs at 
lower rates of stem removal. Thus, physically-based modelling techniques 
similar to those described by Li and Strahler (1985 and 1986) may have to be 
applied at least during technique development. Their use could also have 
significant payoff in forest damage assessment. 
Other important elements of a successful approach to thinning detection and 
mapping will likely involve: (1) image segmentation into stands or portions 
of stands; work here in Scandinavia (Hame 1987 and 1988, Hagner 1989, 
Olsson 1989, and Tomppo 1989) shows that stand-level image segmentation 
produces cover type and cover type change results of generally superior 
quality, and that affordable and consistent algorithms are now available for 
this purpose; (2) evaluation of multitemporal aerial photography and/or 
video imagery on a sampling basis to estimate thinned area (e.g., building on 
technology of the type reviewed by Spencer and Hall (1988) in Canada; and 
(3) evaluation of the use of radar. 
Determination of Early Regeneration Success 
Sylvandcr (1985) has reported success in estimating trees per hectare in 
four year-old plantations in central Sweden using very large scale (1:200) 
aerial photography obtained from a helicopter. Photography was obtained 
in a sampling mode in conjunction with the use of a laser altimeter. Nearly 
all trees over 0.1 meter tail were detected, though high reliability of the 
method required that the mean seedling height be at least 0.4 to 0.5 meters. 
Trees that were missed were located in areas where dry grass had fallen over 
them. Costs were on the order of 70 SEK per hectare sampled. 
Similar results over two year old plantations showed that the number of trees 
was significantly underestimated according to corresponding ground data. 
The reasons for this result appear to be the small size of the trees and the 
problem of larger plants being covered by grass. Sylvander suggests that a 
sample of ground plot data could be used to calibrate the larger sample of 
aerial photograph observations. Spencer and Hall (1988) describe an 
approach used in Canada were this kind of ground calibration of large scale 
photo seedling counts was used with success. 
Thus, for the purpose of assessing early regeneration stocking success (as 
required by Swedish forestry practice law), the use of very large scale aerial 
photography represents a viable measurement alternative. This is 
particularly true where improved local (e.g., commune) statistics are 
required, and a cost-effective way must be found to expand sample size. By 
contrast, the current generation of satellite systems does not appear to 
provide data useful for assessment of regeneration stocking success in very 
young plantations. 
412