Mapping of shallow water bottom type based on landsat tm data

To survey for mapping of the type, abundance and areal extent of the shallow water bottom habitat is very important for marine scientists and coastal planners Nevertheless, surveys utilizing ships may be very costly, time consuming, and inefficient for the areas requiring repeated inventories Remote sensing technology which provides data with a wide and repetitive coverage can help to solve these problems. Unfortunately, the use of remote sensing data for mapping bottom habitats is limited by water depth This study addresses to solve these problems and to assess the applicability of remote sensing technique for bottom type mapping in shallow water.

Takabonerate islands water is selected as the study area because of the status as the National Park and the significance of shallow water bottom habitats. Landsat Thematic Mapper (TM) data covering the study area of the band 1 (blue), 2 (green), 3 (red) and 4 (near infrared) are tested.

In order to remove the influence of the depth in the raw TM dara, some combinations of the band ratioing and a model developed based on the simple water reflectance model (SWRM) are employed. The band ratioing combinations based on the raw TM data as well as the haze-corrected data are used The SWRM is applied to the raw TM data and of the haze-corrected data.

The performance of the band ratioing and the developed model in enhancing the bottom reflectance is analyzed statistically by using the correlation analysis and ANOVA (one-way and two-way) The correlation analysis is used to know the relationship between the TM digital number (DN) and the depth as well as the one in the band ratioing combinations and the developed model Based on the results comparison is made for knowing whether there is any reduction of the depth influence to the DN or not. The one-way ANOVA is used to test the differences in the mean of DN among the depth groups as well as bottom type groups. The two-way ANOVA is used to know the differences in the mean DN of one group against the other groups of the depth groups as well as bottom type groups.

The correlation of the depth and the mean DN in the raw TM bands show that the mean DN tends to decline with increasing depth. The mean DN tends to increase with the increasing depth in both the raw TM band ratioing and the haze-corrected data ratioing The ANOVA test to the depth groups resulted that the mean DN among the depth groups is different significantly in the three TM visuble bands. The same test to the data derived from the SWRM shows that there are no significant differences among the mean DN.

The ANOVA test to the bottom type groups resulted that the three TM visible bands have significance differences in the mean DN. TM visible bands can explain up to 70 percents of the variance of the bottom type reflectance No improvement results are found in the band ratioing combinations. The significant improvement is found in the SWRM output using the haze-corrected data, which can explain up to 80 percents of the variance of the bottom type reflectance.

Based on the statistical analysis result, the SWRM output using the haze corrected data is selected and used as inputs to the supervised multispectral classification process. The classification accuracy assessment is conducted and found that the classification result has an accuracy of 87.1 percents. This result shows that the Landsat TM data can be used to map shallow water bottom types.