bul 2004 
further 
above, 
are. 
jp. 66- 
Vesley, 
-April, 
3) 
2003) 
- 14772 
r. Volker 
for the 
hout the 
partment 
ary data, 
/e would 
titute for 
ware and 
ially Mr. 
QUALITY ASSESSMENT OF GLOBAL MODIS LAI PRODUCT FOR THE REGIONAL 
SCALE APPLICATIONS 
Sun-Hwa Kim' and Kyu-Sung Lee 
Department of Geoinformatic Engineering, Inha University, Incheon, S. KOREA 
22032128 @inhavision.inha.ac.kr 
  
KEY WORDS: Remote Sensing, Ecology, Forestry, Accuracy, Quality, Algorithm 
ABSTRACT: 
MODIS LAI product has been increasingly important for analyzing the process and productivity of terrestrial ecosystems at global 
scale. This study was aimed to assess the quality of global MODIS LAI product for applications in regional and even in local scales. 
To examine the quality of MODIS LAI data, we produced a reference LAI map that was derived by relating the ground-measured 
LAI to Landsat ETM+ reflectance over the forest of the Kyongan watershed in central Korean peninsula. After the reference LAI 
map was rescaled at 1 km spatial resolution, it was compared with the MODIS LAI product that was generated about the same time 
with the reference data. From the comparison between the MODIS LAI and the reference LAI, it was found that MODIS LAI values 
were slightly higher at the forestland than the reference map and they were generally lower in grasslands and croplands. Specially, 
MODIS LAI pixels estimated by the backup algorithm by using NDVI were obviously under-estimated comparing with the reference 
LAI value. The MODIS LAI pixels that were attenuated by cloud cover also showed large difference from the reference LAI value. 
The discrepancy between the MODIS LAI and the reference LAI can be also caused by the misclassification of MODIS land cover 
product that is one of the input variable to create MODIS LAI The quality of MODIS LAI product was largely dependent on the 
estimation algorithm, cloud cover, and the accuracy of MODIS land cover product. e 
INTRODUCTION 
Although it has been relatively new since we are able to get 
MODIS LAI product, the estimation algorithm development 
started since the early 1990's. After launching the EOS terra 
satellite in 1999, the Moderate Resolution Imaging 
Spectrometer (MODIS) has begun to deliver 1km global scale 
LAI products. EOS program progressed the validation of 
MODIS products with production of MODIS data. After 2003, 
the papers about validation of provisional MODIS products 
were began to introduce. Main products of validation were 
MODIS land cover and LAI products. The cases of validation 
of MODIS LAI product have two parts, which are the 
validation of LAI Radiative Transfer algorithm itself and the 
accuracy of LAI product by using empirical methods (Tian et 
al., 2002; Cohen et al., 2003). Since MODIS LAI product was 
primarily designed at global and continental scales, the 
validation of various geographic areas must be continued in 
local or global scale. The validity of MODIS global LAI 
product is being analyzed at several sites over the world, but it 
is rare to find any studies of analyzing at temperate forest in 
northeast Asia. 
Leaf Area Index (LAI) has been an important parameter that is 
directly related to the photosynthesis, evapotranspiration, and 
the productivity of plant ecosystem (Bonan, 1993). 
Measurement of LAI in the field is very difficult, and requires a 
great amount of time and efforts (Gower et al., 1999). The 
mapping of LAI in large geographic area may be impossible 
when we rely on the field measurement. To solve this problem, 
there have been continuing efforts to develop methodologies to 
estimate LAI using remote sensor data (Turner et al., 1999). 
The normalized difference vegetation index (NDVI) was the 
most commonly used, however, showed the saturation 
phenomenon in high LAI value (Chen and Cihlar, 1996; 
Carlson and Reley, 1997). Although empirical modeling is 
relatively easy and useful method for relating field measured 
  
% 
Corresponding author 
27 
LAI to remote sensor data, several factors have certain 
influence empirical model (Cohen et al., 2003). The vegetation 
type, canopy structure, background, atmospheric conditions, 
and topographic conditions were affects the empirical model 
(Tian et al., 2000; Panferov et al., 2001). 
In this study, we are attempted to determine the validity of 
MODIS global LAI product at temperate forest in northeast 
Asia. Terrestrial ecosystem in this region is vulnerable to 
several factors of environmental perturbations. MODIS LAI 
product can be an invaluable source of information for 
analyzing the status and activity of vegetation at regional and 
local scale. 
METHODS 
Study area 
The study area, the Kyongan Watershed, is located near the 
Seoul metropolitan area in central Korean peninsula. The study 
area covered about 1,070km? area and 6796 of them is forest. 
Two third of the forest is natural forest of mixed deciduous 
species, in which the dominant species are oaks (Quercus 
mongolica) mixed with natural pine species. The remaining one 
third of the forest is coniferous plantation stands of Korea pine 
(Pinus karaiensis), Pitch pine (Pinus rigida), and Larch (Larix 
leptolepis). Except for a few plantation forest stands, most 
forests have very dense canopy closure over 80% and average 
tree age ranges from 20 to 50 years. The study area has been an 
experimental watershed for many years to study local-scale 
runoff and water quality monitoring, in which site-specific LAI 
estimate can be a variable of interest for the hydrological 
studies (Lee et al., 2003).