ANNE) 
SPRUCE STAND MONITORING BASED ON THE SUCCESSIONAL SPECTRAL TRAJECTORY 
USING LANDSAT TM DATA 
Yoshio Awaya 
ApS Forestry and Forest Products Research Institute Tohoku Research Center, Morioka, Iwate 
00 020-01, Japan 
(Paper 
Commission VII, Working Group 3 
KEY WORDS: Forestry, Classification, Monitoring, Landsat, Spruce, Succession 
ABSTRACT: 
A forest monitoring method, which was based on the exponential function and the minimum 
distance classifier, was proposed. Averades and standard deviations of age classes were 
estimated and used for the classification of four Landsat Thematic Mapper images, which 
were obtained in winter and summer. The method was applied for spruce stands, and it 
was found that the method . reduced effects of seasonal spectral variation on 
classification. The results suggested that successional stages can be monitored quite 
well using summer images, and comparison of winter and summer imagery showed forest 
structure differences, which were mainly due to difference of spruce density in stands. 
  
1. INTRODUCTION boreal forest in southern Finland. Most 
Pakt of. forests: 1m the study site.is 
Vegetation changes its spectra seasonally artificial,’ and their. major, species, are 
and it: disturbs "consistent. vegetation spruce (Ezo spruce, Picea jezoensys Carr. 
monitoring using remote sensing. On the and Akaezo spruce, Picea glehnii Mast.), 
other hand, it has been recognized .for a fir (Abies sachalinensis Mast.) and 
few species that forests change their Japanese larch (Larix leptolepis Gord.). 
spectra in accordance with their growth AS the forests stand on a very flat area, 
stages. The relationship between there .is little : topographic effect ..on 
reflectance and stand age of species can satellite imagery. The soil is made of 
be expressed by an exponential function volcanic ash and It is thin, then tree 
in a Spectral measurement from a growth is poor and. wind throw happens 
helicopter, and it is called the often. Though Ezo spruce was. the main 
successional Spectral trajectory planting species, a very strong typhoon 
(Peterson and Nilson, 1993: Nilson and caused a severe damage on .the forests in 
Peterson, 1994). The successional early 1950's. Some tree species were 
trajectory can be expressed very well planted experimentally after that. 
through four” seaâsons “for reflectance lor Japanese larch was planted in late 1950's, 
radiance data of channel 3, 4 and 5 of however. a disease and mice caused severe 
Landsat Thematic Mapper  (TM3, TM4 and damages. Though Seotch pine (Pinus 
TM5) (Awaya and Tanaka, 1996: Awaya et al., ; : x 
sylvestris Linn.) was also. planted. in 
1996). The trajectory may be appeared by 
increase of leaf biomass, forest coverage 
(namely background effects), composition 
of planted and non-planted species, and 
so on (Nilson and Peterson, 1994: Awaya 
et al., 1996). Those ‘suggestithati forests 
in the same successional stage can be 
identified based on the trajectories 
using remote sensing images in any 
seasons. The objective of this paper is 
to demonstrate a monitoring method based 
on the minimum distance classifier and 
the successional spectral trajectories, 
and to make confirm usefulness of the 
trajectories using summer and winter TM 
late 1950's, it was suppressed by broad- 
leaved trees due to its declining. Fir 
Seems .to be quite successful, but sits 
commercial. value is not good... Then Ezo 
Spruce and Akaezo spruce become to be the 
major planting species since early 1960's. 
However, = they are planted in smaller 
areas making stripes interleaved by 
natural broad-leaved trees as shelters 
from winds. 
Four TM images, which were 2 images in 
Snowy season (March ' 12, 1985, D0312; 
March 9, 1993, D0309) and 2; images in mid 
imagery summer (August 10, 1985, D0810/ : July 8, 
: 1993, D0708), were used. Those images 
were resampled and overlaid on  D0708 
using first order. polynomials derived 
from control points in areas lower than 
400 meters in elevation to make 
registration of images in different paths 
best. The images were also geometrically 
corrected in the same manner of image 
overlaying using the Universal Transverse 
Mercator Map Projection. Forest planning 
2. TEST SITE AND DATA 
A test site was selected in the national 
forest near Tomakomai city in Hokkaido 
island in Japan. The test site is located 
in the border of boreal forest and cool 
temperate forest. The forest has very 
rich undergrowth comparing with the 
  
  
25 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996 
nd