The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008 
4 
We plot absolute trend differences of CC between the two years 
(i.e. CC 2 oo4 - CC 20 o2). This histogram approximates a normal 
distribution, but the mean (p) is slightly negative (-0.057) 
instead of 0. We therefore use statistics to obtain the standard 
deviation (a = 0.22) and the following thresholds for 5 
differential classes: severe decrease (-1 to p-2c), slight decrease 
(p-2o to p-o), indifferent (p-a to p+a), slight increase (p+a to 
p+2a) and severe increase (p+2a to 1). 
The spatial change of the derived CC between 2002 and 2004 in 
the forested area of the Three Gorges region is mapped in 
Figure 2. It clearly shows that areas with an increasing and 
decreasing CC are not randomly distributed. As a result, most 
of the forest areas show an increase of CC between 2002 and 
2004. In some counties of Chongqing reservoir region, like the 
range from Wuxi in the east to Shizhu in the west, more than 
5% coverage with the class of severe increase are detected. This 
general increasing trend of the forest CC in the observed period 
is not only due to an expected natural increase of CC in trees 
but also because of some policies implemented in the Three 
Gorges region. However, due to the rural resettlement and 
urban relocation in the Three Gorges region resulted from the 
Dam project, an increase in resource needs is observed, such as 
the demand of arable farmland and wood removal, which 
directly leads to forest destruction and therefore decreasing CC. 
This is particularly visible in two counties, Xingshan and 
Yichang. The identified ‘severe decrease’ regions are the most 
important areas requiring a solid and sustainable forest resource 
protection in the Three Gorges region. 
Figure 2. Change map of CC between 2002 and 2004 in the 
Three Gorges region. 
5. CONCLUSION AND OUTLOOK 
This study indicates that: (1) The Li-Strahler geometric-optical 
model can be successfully inverted for estimating the forest 
crown closure (CC) and the pixel-based fraction of sunlit 
background scene component (K g ) is the most important input 
parameter that influences the accuracy of the Li-Strahler model 
inversion; (2) The regional scaling-based endmember extraction 
method can upscale the information from high spatial resolution 
data to low spatial resolution data by means of inverting a linear 
unmixing model in the overlapping region for two images. It 
can also expand the information from local to regional. The 
inverted Li-Strahler model combined with the scaling method 
makes it possible to estimate CC in large areas. By using multi 
temporal data, a change detection can be carried out; and (3) 
Spatial interpolation techniques can properly deal with the 
missing estimates resulting from the Li-Strahler model 
inversion based on the endmembers calculated by the scaling 
approach. With the support of a statistical based interpolation, 
the inverted Li-Strahler model combined with the scaling 
method can finally yield spatially continuous maps representing 
CC in the whole study area. 
Although this method contains several uncertainties, it provides 
a remotely sensed technique to detect changes of the forest 
structure. This study gives the basis for understanding the 
changes of the forest between 2002 and 2004 in the Three 
Gorges region. It also points out important issues for further 
work in an effort to develop a forest monitoring and change 
detection protocol for the Three Gorges region based on multi 
temporal satellite observations. The Three Gorges Dam is 
expected to be fully operational in 2009. The eco- 
environmental impact resulting from the construction of the 
dam will require a long-term monitoring approach, beyond the 
one presented here. In addition, besides the forest crown closure, 
other forest structural and biophysical properties, such as tree 
height, age, leaf area index, and canopy chlorophyll will be of 
future research interest for forest monitoring in this region. 
ACKNOWLEDGMENTS 
We gratefully acknowledge financial support from the 
Knowledge Innovation Program of the Chinese Academy of 
Sciences (KZCX1-YW-08-01-01) and (KZCX3-SW-334). We 
appreciate supports from Zhang Lei, Zhu Liang, Dong Lixin, 
Wei Yanchang, Li Jinye and Liu Xin for participation in the 
field campaign. 
REFERENCES 
Brown, L., Chen, J.M., Leblanc, S.G. and Cihlar, J. 2000. A 
shortwave infrared modification to the simple ratio for LAI 
retrieval in boreal forests: An image and model analysis. 
Remote Sensing of Environment, Vol. 71, pp. 16-25. 
Chambers, J.Q., Asner, G.P., Morton, D.C., Anderson, L.O., 
Saatchi, S.S., Espirito-Santo, F.D.B., Palace, M. and Souza Jr, 
C. 2007. Regional ecosystem structure and function: ecological 
insights from remote sensing of tropical forests. Trends in 
Ecology and Evolution, Vol. 22, pp. 414-423. 
Cressie, N.A.C. 1993. Statistics for spatial data. Wiley series in 
probability and mathematical statistics. Wiley, New York. 
Deutsch, C.V. and Joumel, A.G. 1998. GSLIB: Geostatistical 
software library and user's guide. Applied geostatistics series. 
Oxford University Press, New York, 369 pp. 
Franklin, J. and Strahler, A.H. 1988. Invertible canopy 
reflectance modeling of vegetation structure in semiarid 
woodland. IEEE Transactions on Geoscience and Remote 
Sensing, Vol. 26, pp. 809-825. 
Gemmell, F. 1999. Estimating conifer forest cover with 
thematic mapper data using reflectance model inversions and 
two spectral indices in a site with variable background 
characteristics. Remote Sensing of Environment, Vol. 69, pp. 
105-121. 
Goodwin, N., Coops, N.C. and Stone, C. 2005. Assessing 
plantation canopy condition from airborne imagery using