Considering the CDI measurement scale is 500 m, the map is divided into 500 m
meshes, covering all CDIs on the coastline. The average CDI is calculated to
represent the CDI value in each mesh. For example, if there is only 1 CDI in
1 mesh, this CDI value represents this mesh; if there are 2 CDIs in 1 mesh, then the
average of the CDIs represents the CDI value in this mesh.
The authors made the 500 m mesh map of embankment locality by using data
from the documents preserved in provincial government offices. In the field study,
the authors measured the height of the sea wall, along with the relative heights of
sand dunes or sand ridges and the swampy lowlands between the ridges using
HANDLEVEL K50-1560 (Nobel) among Hai Hau, Nghia Hung, Thanh Hoa areas,
etc. The authors used the longest embankment to represent each mesh. There are
some geomorphological features along the coast: the tidal plain, former river
courses, sand ridges, sand dunes, and offshore beaches. The authors also used
HANDLEVEL to measure the height of the coastline and the inland area of
banks. The heights of areas were taken into account also. However, only CDIs,
the heights of the banks, and heights of land can be taken into consideration since
all the other information was quite fragmentary and incomplete.
The risk map now can be built by integrating those three major factors: CDIs,
banks, and height of land. CDIs are subtracted from immediately succeeding CDIs
using data from eight sheets of JERS-1 image, that is, these CDIs directly measure
the coastline changes of the terms. To measure the risk of the banks, a risk ranking
from 0 to 2 is assigned to each bank by using the height of the banks. The standard
height of bank in this study area is 4 m. Risk level 0 is assigned to those existing
banks with the height more than 4 m; risk level 1 is assigned to those banks less than
4 m; risk level 2 is assigned to those areas without any banks. The risk measurement
regarding sea level can be classified into two categories: risk ranking 0 represents
the land level of the coastal area being higher than sea level; risk ranking 1 results if
the land level is lower than sea level. Based on the above nine items (seven sets of
CDI data, bank data, and heights of land data), 74 meshes in 500 m squares were
subjected to cluster analysis.
10.3.4.2 Numerical Estimation Using Hierarchical Cluster Analysis
In this study, cluster analysis was used to categorize the meshes. Cluster analysis
has been used by evolutionary biologists as a tool for phylogenetic relationship
studies since the 1960s. Hierarchical cluster analysis is one of the classical methods
in dynamic programming (DP) in multivariate analyses and often is used as a
heuristic approach.
The results of cluster analyses are normally expressed as the dendrograms.
Similarities and dissimilarities can be applied to the all calculable dataset. From
the perspective of pattern recognition, hierarchical cluster analysis is categorized as
the uncensored learning method. The objective of hierarchical cluster analyses is
the discovery of the cluster by classification of the ranks. The rank is absolutely
determined by the range of the similarities or the dissimilarities. Commonly used
10 Mapping Coastal Erosion Risk in the Southern Red River Delta, Vietnam
205
meshes, covering all CDIs on the coastline. The average CDI is calculated to
represent the CDI value in each mesh. For example, if there is only 1 CDI in
1 mesh, this CDI value represents this mesh; if there are 2 CDIs in 1 mesh, then the
average of the CDIs represents the CDI value in this mesh.
The authors made the 500 m mesh map of embankment locality by using data
from the documents preserved in provincial government offices. In the field study,
the authors measured the height of the sea wall, along with the relative heights of
sand dunes or sand ridges and the swampy lowlands between the ridges using
HANDLEVEL K50-1560 (Nobel) among Hai Hau, Nghia Hung, Thanh Hoa areas,
etc. The authors used the longest embankment to represent each mesh. There are
some geomorphological features along the coast: the tidal plain, former river
courses, sand ridges, sand dunes, and offshore beaches. The authors also used
HANDLEVEL to measure the height of the coastline and the inland area of
banks. The heights of areas were taken into account also. However, only CDIs,
the heights of the banks, and heights of land can be taken into consideration since
all the other information was quite fragmentary and incomplete.
The risk map now can be built by integrating those three major factors: CDIs,
banks, and height of land. CDIs are subtracted from immediately succeeding CDIs
using data from eight sheets of JERS-1 image, that is, these CDIs directly measure
the coastline changes of the terms. To measure the risk of the banks, a risk ranking
from 0 to 2 is assigned to each bank by using the height of the banks. The standard
height of bank in this study area is 4 m. Risk level 0 is assigned to those existing
banks with the height more than 4 m; risk level 1 is assigned to those banks less than
4 m; risk level 2 is assigned to those areas without any banks. The risk measurement
regarding sea level can be classified into two categories: risk ranking 0 represents
the land level of the coastal area being higher than sea level; risk ranking 1 results if
the land level is lower than sea level. Based on the above nine items (seven sets of
CDI data, bank data, and heights of land data), 74 meshes in 500 m squares were
subjected to cluster analysis.
10.3.4.2 Numerical Estimation Using Hierarchical Cluster Analysis
In this study, cluster analysis was used to categorize the meshes. Cluster analysis
has been used by evolutionary biologists as a tool for phylogenetic relationship
studies since the 1960s. Hierarchical cluster analysis is one of the classical methods
in dynamic programming (DP) in multivariate analyses and often is used as a
heuristic approach.
The results of cluster analyses are normally expressed as the dendrograms.
Similarities and dissimilarities can be applied to the all calculable dataset. From
the perspective of pattern recognition, hierarchical cluster analysis is categorized as
the uncensored learning method. The objective of hierarchical cluster analyses is
the discovery of the cluster by classification of the ranks. The rank is absolutely
determined by the range of the similarities or the dissimilarities. Commonly used
10 Mapping Coastal Erosion Risk in the Southern Red River Delta, Vietnam
205