Comparison of Different Sensors and Analysis Techniques for Tropical
Mangrove Forest Mapping
J. Aschbacher, P. Tiangco, C.P. Giri, R.S. Ofren, D. R. Paudyal, Y. K. Ang
European Commission, Joint Research Center,
IRSA, TP441, I-21020 Ispra (VA),Italy;
Tel: +39-332-785968, Fax: +39-332-785461,
E-mail: josef.aschbacher@jrc.it
Asian Institute of Technology (AIT), SERD-RSL,
GPO Box 2754, Bangkok 10501, Thailand
UNEP/EAP-AP. c/o AIT, GPO Box 2754, Bangkok 10501, Thailand
Abstract
The objective of this study is to compare different remote sensing sensors and analysis techniques for the purpose of mangrove mapping. A study area is Southern Thailand of approximately 40x30 km size was selected. A systematic assessment of strengths and limitations of data taken from different sensors, namely Landsat TM, Spot HRV, MOS MESSR, JERS-1 SAR, was carried out. The results of the investigation show that optical remote sensing data is highly suitable for mapping mangrove forests and can discriminate reasonably well four mangrove forest classes, namely homogeneous rhizophora, homogeneous nypa, mixed dense and mixed open mangrove forest. The classification accuracy is approximately 87%. The use of radar data alone resulted in a significantly lower classification accuracy, but on the other hand provided additional information related to the age distribution of rhizophora stands.
Introduction
A mangrove forest is a salt tolerant forest ecosystem of the inter-tidal regions along the coastlines. It is predominantly a tropical evergreen forest comprising trees and shrubs. Although the mangrove forest is characterised by very low floristic diversity it is considered as one of the most productive ecosystems. In the past year, mangrove forests into have been severely threatened by over-exploitation and, more recently, conversion of mangrove forests into shrimp ponds. Other factors of mangrove degradation are related to mining activities, establishment of industrial estates, coastal development and agricultural land expansion programmes. The environmental impact is irreversible and mostly severe with changes occurring with in a few months. The need for updated and continuous information is evident for a sound management approach.
Currently, monitoring is widely based on visual interpretation of aerial photography or optical satellite imagery. The latter is hampered by frequent cloud coverage along the tropical coastlines. This study will introduce radar imagery and digitally based analysis methods.
The work presented in this paper was carried out within the frame work of a research project sponsored by the Austrian Academy of Sciencess (AAS), With support from the European Space Agency (ESA), the National Research Council of Thailand (NRCT), the Thai Royal Forest Department (RFD), the Asian Institute of Technology (AIT), and the United Nations Environment Program (UNEP). The Austrian Association for Development & Co-operation (ADC) supported the administration of the project.
Objectives
The main objective of this study is to systematically compare different remote sensing sensor types for the purpose of mangrove monitoring. Data of common optical and radar sensors shall be compared , and a suitable methodology developed. Recommendations shall be made regarding the best approach for the use of remote sensing for mangrove mapping and monitoring.
Study Area and Data Basis
Study Area
A study area of approximately 30x40 km size was selected in Phanagnga Bay, Southwest Thailand. The center co-ordinates are 98.5 deg E, 8.4 deg N. The topography is generally flat, with a few single mountain cliffs reaching out from the Andaman Sea, or the land surface. The flat topography causes a very large tidal range with extended mud flats. The soil in the mangrove area comprises clayey and silty sediments., they are alkaline in nature and generally moderately fertile and high in organic matter. The climate is humid tropical with seasonal monsoon rainfall from April to September and around December [1,2]
The mangrove types observed in the study area are homogeneous stands of Nypa palms and Rhizophora, as well as mixed mangrove forest of varying density. Selective logging by the local people and tin mining activities are the main reason for deforestation, which is done in a reasonably controlled fashion. No conversion from mangrove forests into fishponds is observed in the study area.
Data Basis
The satellite images available for this study are listed in Table 1.
Table 1: Satellite data available for data analysis
| Sensor |
Acquisition date |
Spatial resolution |
| Spot HRV |
22-Mar-92 |
20m |
| Landsat TM |
20-Apr-93 |
30m |
| MOS-1 MESSR |
22-Feb-89 |
50m |
| Jers-1 SAR |
8-Sep-92 |
18m (3looks) |
| ERS-1 SAR |
16-Jul-92, 20-Aug-92, 29-Oct-92 |
25x30m (3looks) |
Among the three optical images only the spot data are practically cloud-free, while the Landsat and MOS images show scattered clouds over the mangrove study area. No recent image with less cloud cover could be obtained of this region. The ERS-1 radar images were available in 8-bit only, as processed by the Indian receiving station. Data calibration was carried out according to [3]. The JERS-1 SAR data were available in standard 16-bit format. The Spot and multi-temporal ERS-1 images are shown in Fig. 1a and 1b, respectively.
Figure 1: Mangrove study area of Phangnga Bay (Thailand), as shown in (a-left) Spot HRV imagery and (b-right) multi-temporal ERS-1 SAR imagery (image copyrights Spotlimage, ESA).
The interpretation of remote sensing data was supported by extensive ground measurements taken concurrently with ERS-1 overflights as well as topographic maps, soil maps, meteorological and tidal information and a digital elevation map.
