A Satellite Based Monitoring of Changes in Mangroves in Krabi, Thailand Materials and methods Materials
Three dates of satellite imagery were acquired. Landsat TM data were obtained 15 December 1995 and 12 February 2000. Landsat ETM+ data was acquired on 9 February 2002. The wavebands representing of near-infrared and visible-red region were extracted from each Landsat dataset. The Normalized Difference Vegetation Index (NDVI) is used to transform multi-spectral data into a single image band which representing vegetation distribution. The NDVI values indicate the amount of green vegetation present in the pixel. Higher NDVI values indicate more green vegetation. In the ENVI system, NDVI were computed according to the standard algorithm: NVDI = (NIR-Red)/ (NIR+Red) Valid results fall between -1 and +1. The NDVI which computed from each of the three years (1995-2000-2002), were applied to ISODATA unsupervised classification. Change detection was taken an account on the changes which happen on 1995 to 2002. A statistic test (t-test single factors) was applied in order to compare 2 means on NDVI value among years. Change areas were significantly different at 95 % (p<0.05). The methods were used in this study presented in figure 2.  Figure 2 Methods Results and discussions Change detection and monitoring involve the use of multi-date images to evaluate differences in land cover due to environmental conditions and human actions between the acquisition dates of images. Successful use of satellite remote sensing for land use/cover change detection depends upon an adequate understanding of landscape features, imaging systems, and information extraction methodology employed in relation to the aims of analysis. In this study, The Vegetation Index techniques of change detection were applied. The Vegetation Index is represented to the vegetation distribution which transformed according the spectral reflectance characteristic. The mangrove change detections was done by clustering algorithm on unsupervised classification (ISODATA) of the 3 vegetation index band of 1995, 2000, and 2002. An advantage of this technique is to avoid the error in classification due to overlap between classes of the training areas (Brook and Kennel, 2002). Change detection using Normalized Difference Vegetation Index The wavelength bands representing to the near-infrared and visible-red region of the electromagnetic spectrum were extracted from each LANDSAT TM and ETM+ datasets. The Normalized Difference Vegetation Index (NDVI) images had been generated. Green vegetation absorbs more red light and less-infrared than other surfaces. High NDVI values represents as leaf biomass or leaf area increased. On the grey scale NDVI image the vegetated area appeared in bright tones. NDVI histograms (figure 3) showed 3 categories of datasets of 1995, 2000, and 2002. Green vegetation shown high values on the histogram curve. Water bodies have shown low NDVI values as minus values. The histogram of NDVI acquired on 15 December 1995, presented the highest ranging in NDVI value between -0.677 to 0.733. Vegetated area appeared in the range of 0.388 to 0.733. Settlement area and built up land ranged 0.1174 to 0.388. Barren dry land and dry shrimp pond ranged -0.065 to 0.1174. Mud flat area appeared the low NDVI value ranged-0.065 to -0.189. Water bodies and mangrove canal were shown the lowest NDVI value ranged-0.189 to -0.677. The histogram of NDVI on 12 February 2000, provided the narrower NDVI value than the 1995, varied from -0.579 to 0.645. Vegetated area was ranged from 0.314 to 0.645. Barren lands and settlement areas ranged from -0.057 to 0.314. Mud flats varied from -0.057 to -0.196. Water bodies and mangrove canal were also shown the lowest NDVI value ranged from -0.196 to -0.579. The histogram of NDVI on 9 February 2002, presented the narrowest NDVI value than the other images, ranged from 0.086 to 0.462. Although, vegetated area also appeared the highest NDVI value than other objects, varied from 0.086 to 0.462. But most of vegetated area contained the low NDVI value of 0.1988 to 0.3246. Barren lands and built-up areas ranged from -0.270 to 0.086. Water Bodies were also shown up the lowest NDVI value of -0.545 to -0.270.  Figure 3 NDVI histogram calculated from 1995, 2000 and 2002 imagery Considerate on the NDVI histogram on each year, it presented the result similar as described according to Lillesand and Kiefer (2000). They stated that the vegetation areas present high value of the index because of proportion of high near infrared (band 4) reflectance and low visible reflectance (band 3). While water bodies had higher reflectance value on visible band than near infrared band, thus it shown negative on the index value. Consequences, rock and soil areas had the similar reflectance value on both wavelength and cause the index near zero. Three features were extracted from the 3 NDVI histograms. It resulted that vegetation area occupied the higher value of the histogram. In contrast, water bodies also had the negative value and at near zero point which was the transition zone among water bodies and vegetation areas. It found that was urban area and barren land. Vegetation areas and water bodies were occupied most area of the image which could be accommodate to the real feature that terrestrial with cover by both mangrove forest and agriculture plants cover the upper portion of this area, while water bodies of sea features and canals were found cover the lower portion of the study area. By visually comparison the 3 histogram, both NDVI values on 1995 and 2000 consist of 2 main curves of vegetation areas and water body area. Urban area and barren were found in small proportion. While, the new peak of near zero value was emerged in the 2002 histogram. Because it carried the value or urban and barren soil, it could be considerate that there had new development of this feature happened. However, ground truth is needed in order to determine the exactly feature. |