|
Servel Miller, Malgorzata Leszczynska
Evaluation of the use high resolution satellite
Imagery to map slope instability in a tropical environment: St. Thomas, Jamaica
| Conference Information: |
9th International Conference “Environmental Engineering”, 22–23 May 2014, Vilnius, LITHUANIA |
| Source: |
ICEE-2014 - International Conference on Environmental
Engineering |
| Book Series: |
International Conference on Environmental Engineering
(ICEE) Selected papers |
| ISSN: |
eISSN 2029-7092 online |
| ISBN: |
eISBN 978-609-457-640-9 |
| ISBN: |
ISBN 978-609-457-690-4 CD |
| Year: |
2014 |
| Publisher: |
Vilnius Gediminas Technical University Press Technika |
View full text in PDF format
Abstract
Landslides are a major natural hazard in Jamaica, and have
resulted in loss of life, major economic losses, social disruption
and damage to public and private properties. There is a need to
delineate areas that are prone to slope instability in order to
mitigate their effects. The first and most important stage for
the creation of a landslide risk maps is the collection of accurate
landslide data in a timely manner. However the type of terrain makes
landslide mapping particularly difficult. Aerial Photographs have proven
to be an effective way of mapping landslides but acquiring new photographs
to map recent landslides is very expensive. High resolution satellite imagery
were evaluated for their effectiveness in delineating landslides. The landslides
on a whole had no distinctive spectral property; hence no one classification
technique could be used to identify them. This research developed integrative
methods utilising a combination of: edge enhancement to delineate the scarps
area; Wetness Index to identify back titling blocks and debris flow lobes
where moisture is higher; shape classification (to distinguish from e.g.
ground cleared for agriculture); and slope curvature to map scarps. The
information from the image classification was combined in a GIS and automated
to determine the probability of the presence and or absence of a landslides.
Data derived was validated against detailed field mapping at a scale
of 1:5000. For more recent landslides, the modelling proved to be
effective, accurately identifying 91% of the landslide both in
terms of the location and extent. For the older landslides Pre
2000) the mapping was less effective, with misclassification as
high as 24% particularly for smaller landslides. However, the
use of these imagery does have great potential as they prove
useful for mapping new landslides quickly and efficiently after
landslide disaster and are much cheaper and quicker to acquire.
Keywords: natural hazard; landslide risk maps; landslide data; Geographical Information System.
|
