Construction of Vegetation Cover Map in Metro Manila for Urban Environmental Evaluation

Construction of Vegetation Cover Map in Metro Manila for Urban Environmental Evaluation

Donkyu Yun, Akira Hoyano Tsuneo Matsunaga and Takao Katoh
Dept. of Environmental Science and Technology, Tokyo Institute of Technology
4259 Nagatsuta-chao, Midori-ku, Yokohama, 226-8502, Japan
Tel: (81)-45-924-5501 FAX: (81)-45-924-5553
E-mail : dyun@depe.titech.ac.jp

Introduction
To plan the comfortable urban, revegetation planning is a factor which cannot be overlooked. Especially, environmental alleviating effect of vegetation has been notifying on discussing the urban environment. First of all, the essential work on the vegetation planning and management is understanding the vegetation existing. In this study, it has been adapted an algorithm for extracting the VCRP (Vegetation Cover Ratio in a Pixel) from LANDSAT/TM and SPOT/HRV on Metro Manila. It has been also aimed to verify the value of VCRP and identify the appropriateness of the algorithm on tropical region using field survey.

Calculating the VCRP with the remote sensing data
Metro Manila (Philippine) is tropical region located in 14 degrees' north latitude. There are rainy (June to November) and dry season (December to may). The used data are LANDSAT/TM (observed in 1997.5.15) and SPOT/HRV (observed in 1997.5.2). The reference map was a digital map itemized into road., rail road, river and so on. Before calculating VCRP, it had been pretreated the rectification of the remote sensing data and the digital map with accuracy of 1.2 pixel.

It is necessary to find the mean MRVI values of vegetation and non-vegetation so as to calculate VCRP by adapting algorithm of VCRP estimation to Metro Manila area. The estimation method with MRVI histogram proposed by authors (l) has been adapted to find the mean MRVI values of vegetation and non-vegetation. Figure 1 shows an example of MRVI histogram extracted from LANDSAT/TM. The solid line represent calibration curves of part A and B, respectively, which curves are normal distribution curves. The mean values of the curves have the mean value of vegetation (A, 0.34) and non-vegetation (B, 0.19), respectively. The relation equation between MRVI and VCRP has been extracted using the mean values previously mentioned, and then VCRP of all site has been calculated. On the same way, VCRP has been also extracted using SPOT/HRV.

Verification of vegetation cover in representative blocks
(1) Methodology of Verification
The field survey as well as satellite observation has done the last period on dry season, so that can be considered vegetation growth state to be almost near. Generally, the pixel wise comparison is known as the verification method for vegetation cover in a block (VCRB). However this method has some problem such as rectification between the satellite images and ground-position. As an alternative method, it can be thought the comparison of each block. Since the method has the each integration value from several tens1 to hundreds1 pixels, there is little influence of rectification error. On the basis of above reason, we selected the block wise comparison method.

(2)Selecting the representative blocks
We selected the method for comprising the representative blocks, because surveying and comparing all blocks requires many times and great efforts, though it is desirable. Selecting the representatives blocks, it should be considered the vegetation pattern and the position identification. The blocks satisfied the above-mentioned two conditions were choose using the VCRP image, street map and field reaffirmation. Consequently, 14 blocks which contained no, low, middle and all vegetation cover blocks have been selected.

(3)Calculating the vegetation cover in a block from field survey
The measuring items for extracting vegetation cover in a block (VCRB) is as follows. The measuring method was done by taking a photograph, measuring with tape measure and with eye.

(a) Species and forms of trees: The dimension of tree forms (height, width, clear length) was measured in condition that the species of trees can be distinguished. In the other case, each tree height was classified into tall tree, middle tree and shrub.

(b) The leaf area ratio on a tree crown: The binary image (Fig. 1 (b)) is extracted from the photograph Fig. 1 (a)) taken vertically on the ground. Finally, the ratio of leaf on the tree crown is obtained using binary image.

Plant name on tree form	Leaf cover ratio
Plant name on tree form	Mean	S. D.
Narra	0.87	0.12
Acacia	0.34	0.14
Tangile	0.62	0.28
Molave	0.71	0.14
Bakawan	0.71	0.28
Palm	0.50	0.28
Lawn (dense)	0.70	0.30
Lawn (sparce)	0.50	0.30
Tall tree	0.61	0.19
Middle tree	0.71	0.14
Law tree	0.62	0.28

Table. 1 Leaf cover ration on tree crown

Fig. 1. Photograph and binary image of tree crown looked upward on the ground looked upward on the ground in case of Acacia.

Fig. 2. Relationship between VCRP in Satellite data and that in field survey. (Parts of 0% and 100% in field survey are extended in order to prevent the confusion in multi-pointing.)
(4) Comparison of Vegetation Cover Ratio in each Block
Relationship between VCRB in verification block and VCRB is obtained by remote sensing data as shown in figure 2. In this figure, vegetation cover ratio extracted from two types is nearly same. In addition, we can see that VCRB obtained by field survey has been underestimated, because inside residential district or rooftop vegetation had not been observed.

Construction of Vegetation map with VCRP
Figure 3 shows the vegetation map with VCRP constructed using LANDSAT/TM and SPOT/HRV, respectively. Two distribution maps are similar overall, which has been confirmed that the different image shows about 80 percent concordance within 10 percent or less different. We can see that the small scale vegetation are extracted as approximately from 10 to 50 percent in south side of Rizal park where is covered building, pavement, and small vegetation.

Figure 3. Vegetation cover maps with VCRP in Metro Manila Using remote sensing data
Conclusion
The vegetation map with VCRP was constructed by adapting the vegetation mapping algorithm to LANDSAT/TM and SPOT/HRV in Metro Manila. To the purpose of comparing the VCRP map, field survey was also carried out in the same reason. It was clarified the adaptability for the algorithm by comparing the VCRB calculated from remote sensing data and from field survey.

Acknowledgments
This study is supported by "Impact analysis of metropolitan policies for development and environmental conservation in the Philippines" (1997~) as a research project on "research for the future program", JSPS (Japan Society for the Promotion of Science).

Reference

Donkyu Yun, Akira Hoyano, Naoki SHIRAI and Masahiro SETOJIMA : A study on putting to practical use of urban vegetation mapping with vegetation cover ratio in a pixel Part 1, A method for quantification of vegetation cover ratio in a pixel, Proc. of the 25^th Japanese Conference on Remote Sensing, Nov. 1998, Kyoto, Japan. (Japanese)