Forest Monitoring Framework At Regional Level Using Multi-Resolution Satellite Data
With Comdination Of Optical And Thermal Bands
Vegetation Classification Methodology Using Multi-Resolution Data with Combination of Optical and Thermal Bands
The main theme of the integrated classification is to use NT-NDVI with the threshold for each land class from high resolution data to improve the thresholds used low resolution model to yield the thresholds used for low resolution data.
The most important issue that is the main advantage of multi-resolution classification is the primary forest, secondary forest, and active agriculture can be separated with ease by using the calibration of NT-NDVI thresholds from Landsat TM NT-NDVI. This comes from the fact that is very difficult the thresholds from low resolution data alone to differentiate primary forest, secondary forest, and active agriculture.
Low resolution data (NOAA AVHRR) with support of calibrated high resolution data (Landsat TM) were used to produce forest map in Indo-China Peninsula on the assumption that similar characteristics of the ground truth area can be extended to other area. Even though the verification was difficult to undertake due to lack of reference data, this methodology could be an estimate for forest monitoring in Indo-China peninsula.
The classified result from developed methodogy from NOAA Mosaic of southeast Asia (1990) from NIES is illustrated in Figure 3. The thresholds used for the classification are in Table 1. The result from error analysis from thailand' Forest map is illustrated in Table 2. The statistic of the classified result for each country in Indo-China Peninsula using calibrated NT-NDVI is illustrated in Table 3.
Figure 3 Classified result from Calibrated NT-NDVI (Indo-China Penensula, 1990)
Table 1 Thresholds applies for each land class
| |
Land Class |
NT-NDVI |
Calibrated NT-NDVI |
| 1 |
Primary Forest |
0.70 |
0.6821 |
| 2 |
Secondary Forest |
0.59 |
0.5913 |
| 3 |
Active Agriculture |
0.41 |
0.41 |
| 4 |
Harvested Land |
0.29 |
0.29 |
| 5 |
Bare Soil |
-0.05 |
-0.05 |
Table 2 The accuracy checking from the error analysis in Thialand's area
| NT-NDVI |
Accuracy |
| Central Plain |
92.55% |
| Northern Part |
78.58% |
| Southern part |
85.88% |
| Averaged |
85.67% |
Form the calculated statistic of forest area classified from NIES (1990) data in Table 3, Thailand's forest area is 2.03%. From the statistic of Thai's Royal Forest Department, Thailand's forest area is 26.64% (RED, 1992). The different is 1.61 %. Also, the statistical result for only primary forest class is 10.32 %. This statistical result is close to the statistic from RED that stated 12.2 % is Tropical Evergreen Forest (RED, 1992), which by the definition of this study categorizes this type of forest to be primary forest. This result shows that this new classification methodology can achieve the proposed objective.
Table 3 Statistic of the classified result for each country in Indo-China Peninsula using Calibrated NT-NDVI
| Land Type |
Thailand |
Mayanmar |
Laos |
Cambodia |
Vietnam |
| Water Area |
0.26% |
0.19% |
0.08% |
0.94% |
1.49% |
| Bare Soil |
16.69% |
3.26% |
2.74% |
12.24% |
24.18% |
| Harvested Land |
26.97% |
9.40% |
5.90% |
12.65% |
14.77% |
| Active Agriculture |
31.05% |
17.69% |
23.63% |
29.63% |
27.79% |
| Secondary Forest Area |
14.71% |
19.72% |
23.02% |
19.04% |
13.15% |
| Primary Forest Area |
10.32% |
49.74% |
44.43% |
25.50% |
18.63% |
Forest Monitoring Framework for Indo-China peninsula
By using the proposed vegetation classification methodology from multi-resolution satellite data with combination of optical and thermal bands, it is possible to monitor the primary and secondary forest area in Indo-China Peninsula regularly because the acquiring rate from low resolution satellite is rather high. And the proposed methodology can be applied to build an automated monitoring system.
The proposed forest monitoring framework is illustrated in Figure 4.
Figure 4 The Proposed Forest Monitoring Framework
Reference
-
LERTLUM, SURAT, Vegetation Classification Methodology From Multi-Resolution Data By The Combination of Optical And Thermal Bands, Doctor of Technical Science Dissertation, Asian Institute of Technology, Bangkok, Thailand, 1997.
- RED, 1992, Forestry Statistics of Thailand.
- YASUOKA, Y., SUGITA, M., YAMAGATA, Y., TAMURA, M., SUHAMA, T., 1995 Scaling Between NOAA AVHRR Data and Landsat TM Data for Monitoring Wetland, processing of the International symposium on Vegetation Monitoring, Chiba, Japan.
