Forest Monitoring Framework At Regional Level Using Multi-Resolution Satellite Data
With Comdination Of Optical And Thermal Bands
Methodology
A solution that is proposed for forest classification system of Indo-China Peninsula is to use low resolution data as the main source for monitoring vegetation activity. In order to improve the classification method to achieve better accurate, a new vegetation index, which is the combination of optical and thermal bands, has been developed in this study. To minimize the disadvantage of low accuracy from using low-resolution data, high resolution data is also utilized to calibrate the classified result obtained from low resolution data, (LERTLUM, 1997)
New Vegetation Index
In this study, the difference of characteristic of visible (bands 1) and near-infrared (band 2) in the from of NDVI, and thermal data (band 4) from NOAA AVHRR (LAC 1.
Km.) of each land type in the test area are examined and compared. It is found that the combination of information from NDVI and thermal band is more useful to classify the land types than NDVI along as shown in Figure 1.
Figure 1 Scatter diagram of CCT Value of Band 4 versus NDVI from NOAA AVHRR
From the analysis, it is found that in order to effectively utilize thermal band, the product of the NDVI and Thermal ratio should formulate the new vegetation index from the combination of thermal ratio with NDVI, which is related to vegetation activity. The new vegetation index is named Normalized Thermal- NDVI (NT-NDVI). The thermal ratio should be able to reflect the difference of land types according to the characteristic of vegetation cover to brightness temperature. It should also represent the normalized brightness temperature factor of NT-NDVI that can be used for the whole region. From the experiment and analysis, the proposed thermal ratio in normalized according to the following formula:
| Normalized Thermal Ratio= |
1.0+ |
(Local Max Brightness Temp- Actual Brightness Temp)
---------------------------------------------
Local Max Brightness Temp |
Where
Local Max Brightness Temp : is the local maximum brightness
temperature of the selected pixel
Actual Brightness Temp : is the brightness temperature of
the selected pixel
then
NT-NDVI = NDVI * Normalized Thermal Ratio
The ratio od relative difference of local brightness temperature to the local maximum brightness temperature divided by the maximum brightness temperature of the area is used. The local maximum brightness temperature is selected from the highest brightness temperature value in the square area surrounding selected pixel. This thermal ratio will reduce the effect of the temperature difference according to the latitude of the location or order factors such as recording data and time, or the altitude of the location, and will lead to the stability of the thermal ratio. From the analysis of characteristic of the local maximum brightness temperature versus the size of the area used for the calculation of the local maximum brightness temperature, the most appropriate area size for the clculation of local maximum brightness temperature is 60 x 60 km
2.
The analysis for new vegetation index is also applied to high resolution data (Landsat TM), which responds with similar characteristic as low resolution data.
This new vegetation index makes it possible to achieve better accuracy of the classification of primary forest, secondary forest, active agriculture land, harvested land, and bare soil, which could not be clearly classified by existing NDVI. In addition, this new vegetation index is stable over are, so it is possible to apply same thresholds out the study area.
Multi-Resolution Integration Technique
This process is accomplished by defining a regression model between the same parameters of low and high resolution satellite data. This process had been demonstrated before by Yasuoka et al. (1995) on NDVI from NOAA AVHRR and Landsat TM.
In this study, the average of band 3, band 4, and band 6 from Landsat TM are calculated to produce the datasets that has the same resolution as 1.1 km of NOAA AVHRR LAC 1.1 km. To use for the calculation of NT-NDVI. Next, the georeference between low resolution and the average high resolution datasets is performed. Finally, the regression model is calculated between NT-NDVI from low resolution and the averaged high resolution (LERTLUM, 1997)
The analysis between the NT-NDVI from averaged TM and NOAA AVHRR is performed as shown in Figure 2. From the study on NT-NDVI from averaged Landsat TM and from NOAA AVHRR, the relationship can be described by the following equation:
Regression Analysis of Normalized Thermal NDVI from Landsat TM and
NOAA AVHRR
Figure 2 Scatter diagram of NT-NDVIavhrr versus NT-NDVItm
NT-NDVIavhrr = 0.5531*NT-NDVItm - 0.3437 with r = 0.84
Where
NT-NDVI
avhrr is Normalized Thermal -NDVI from NOAA AVHRR
NT-NDVI
tm is Normalized Thermal from Landsat TM
