Leaf Area Index (LAI) is an important parameter as assimilation and validation data of the global ecosystem models. Recently several organizations have developed the satellite-based global LAI datasets that are publicly available. However their algorithms, definitions, and uncertainties are slightly different. The objective of this paper is to review the definition, uncertainty, and validation method of the currently available global LAI datasets derived from satellite (ISLSCPII, Boston Univ., GLOBCARBON, MOD15, CYCLOPES). The LAI estimation approaches (vegetation index, and radiative transfer inversion) are firstly described. Then LAI estimation algorithms and validation methods are summarized. The review of the validation activities point out that the recent validation studies are still insufficient both to cover globally and to warrant the LAI seasonality, indicating further validations and algorithm refinements.

View full abstract

This paper presents a GA(Genetic algorithms)-based band selection procedure for producing the color composite image using hyperspectral data (e.g., hyperion data). The more number of hyperspectral bands, the more difficult efficient and effective processing for producing color composite images. In producing color composite images, one of the requirements is to maximize the amount of information in the images (i.e., image entropy). Through GA operations, the increasing of entropy (i.e., fitness value) was confirmed, which corroborates the GA operations can be applied for band selection in producing color composite images. Toward the end of the run, three bands in case of maximizing image entropy are selected, and to produce color composite image, those are assigned to red, blue, and green plane, respectively. Compared with general color composite images (i.e., natural, false, true and ultraviolet color), we conclude that the produced composite image with maximum entropy are useful for image interpretation in terms of the amount of information as well as the image features.

View full abstract

In recent years, various models were developed to evaluate forest ecosystem productivity. Leaf area index (LAI) and fraction of photosynthetically active radiation (FPAR) are key parameters in these models. Remotely sensed data is used to identify these parameters. However, these parameters derived from remotely sensed data have been treated as that of forest canopy. Understory plant, especially dwarf bamboo, plays an important role to the carbon balance in a cool-temperate forest ecosystem. Understanding the areas included the effect of dwarf bamboo is therefore needed for precise evaluation of forest ecosystem productivities using remotely sensed data. Logistic regression model based on the knowledge of ecological research was applied to the mapping of the distribution of dwarf bamboo using satellite remotely sensed data and digital elevation model (DEM). Light condition on the forest floor is the main factor affecting dwarf bamboo. In this study, relative direct solar radiation, NDVI in leaf constant period and difference in NDVI between pre- and post-leaf fall period were valid parameters for mapping the distribution of dwarf bamboo. The logistic regression model developed by this study indicated an overall accuracy of 86.11%.

View full abstract

In recent years, distribution of forests is changing in many countries due to desertification and planting. Satellite remote sensing provides detailed distribution of the vegetation changes. In particular, the new satellite sensors, such as MODIS and SPOT-Vegetation, have provided more accurate data than old sensors such as NOAA-AVHRR for such purpose. In this research, we analyzed vegetation changes from year 2000 to 2005 in eastern Asia with data taken by SPOT-Vegetation and Terra-MODIS. In order to remove cloud contaminations, we applied annual MVC (maximum value composite) to time series in NDVI (Normalized Difference Vegetation Index) taken by each sensor. The annual MVC made it easy to see inter-annual variability because it is hardly influenced by seasonality of vegetation. Then we calculated the linear trend of inter-annual NDVI at each pixel. We applied these procedure to each sensor's NDVI data independently for the sake of cross-validation. As a result, both sensors consistently showed increase of NDVI in north-eastern to central China, whereas decrease in eastern Mongolia. By extracting the trends of NDVI with statistical significance of p=0.05 for both sensor's data, we estimated that NDVI increased in an area of 177,000km², whereas decreased in an area of 63,000km². Most of the increase happened in China. This estimation was consistent with FRA2005 (Global Forest Research Assessment 2005), in which Chinese forest reportedly increased in more than 200,000km² from 2000 to 2005.

View full abstract