Comparison of jers-1 and radarsat synthetic aperture
Radar data for mapping mangrove and its biomass
Mazlan Hashim and Wan Hazli Wan Kadir
Department of Remote Sensing
Faculty of Geoinformation Science & Engineering
Universiti Teknologi Malaysia
81310 UTM, Skudai, Johor, Malaysia
Tel: +607-5502873, Fax: +607-5566163,
E-mail: mazlan@fksg.utm.my
Abstract
This paper has reviewed comparison of classification of mangrove forest at species-level,
and estimation of mangrove biomass using JERS-1 SAR and Radarsat SAR
(standard mode) data. Both of these comparisons were made at selected test site in
Sungai Pulai Mangrove Forest Reserve in Malaysia. The results demonstrated the
utility of SAR data as potential source in mapping mangrove classes and indicator for
biomass. Although there has been limited availability of exhaustive sampling points
done accessibility at the test site, but the results indicated the evidence of C and L band
utility for mangrove mapping and biomass estimation.
Introduction
Mangrove forests grow exclusively in the intertidal zone, where they are greatly
influenced by the coastal environment. Mangrove forests are becoming dwindling
resources because of their continued alienation for various land uses that are assumed
to be of greater economic values. In Malaysia alone mangrove forest area have
decreased by 46.8 percent of the original gazetted area, i.e. from 505, 300 hectares in
1980 to 269, 000 hectares in 1990 (Clough, 1993). Due to its nature, especially, of its
remoteness and limited accessibility, the detecting and mapping of these changes
using conventional technique are elaborately time consuming and very costly. In this
study, SAR data which is independent of to cloud cover and weather interference are
examined for mapping mangrove and estimation of mangrove biomass.
In recent years, SAR data have been used in classification of vegetation precisely
forest over tropical regions. However, only limited studies have been reported on
mapping mangrove forest (Mazlan Hashim, 1999) Moreover, none of these studies
have ever been attempted to examine the potential of SAR to classify mangrove forest
at species level. In this context, this paper is focused on two issues : (i) analyse
whether or not mangrove species can be categorized using typical satellite-based SAR
resolution, and (ii) retrieve of biomass information based on radar backscatter.
Apart from vegetation studies using SAR data, estimation of forest biomass has
widely been reported but again very little effort have been undertaken for mangrove
(Imhof, 1995). Previous studies have indicated that there exist strong correlation
between radar backscatter with forest biomass, particularly of those SAR data
acquired in L and P bands (Beaudoin et al., 1994). Based on these facts, it is also the
main objective of this paper to report on the estimation of mangrove biomass using
JERS-1 SAR and Radarsat SAR which were acquired in C band and L band,
respectively.
Material and Method
Study area.
In order to validate of SAR data in extracting information pertinent to classify
mangrove at species level and to estimate the biomass, a study area which is located
in the southwest of Johore, Malaysia (Figure 1) – the Sungai Pulai mangrove forest
reserve was selected. The study area covers approximately an area of 12.3 km x 18.0
km (centered at 103° 16’ E lat. and 1° 13’ N long.). In the past decade, this area
although has been demarcated as reserve forest but lately has also been given way to
conversion for land related development programs such as development of new port,
aquaculture, charcoal-making industry as well as residential area for supporting the
newly developed industries.
Figure 1 : The study area - Sungai Pulai Mangrove Forest Reserve and corresponding
JERS-1 and Radarsat SAR data of the area.
Digital Image Processing
The JERS-1 (processed at level 2.1 by NASDA- National Space Development Agency
of Japan) and Radarsat (SGF-Path Image) data were used in this study. Specification
of the data is tabulated in Table 1. The ancillary information used to support the study
which includes the corresponding area topographic map (1:50,000 scale), related
forestry records and documents were used as ground reference data. The extend of
mangrove boundary given by the topographic map were digitized into digital image
processing and used as “vector-overlay” in assisting the collection of training and
later used in the accuracy assessment.
Table 1 : Specification of JERS-1 and Radarsat SAR multi-temporal data employed in the study.
| Sensor | JERS-1 | Radarsat |
| Acquired date | Sept. 28, 1994 | Oct., 26, 1997 |
| Pixel size / resolution | 18 meter | 25 meter |
| Wavelength | 23.5 cm | 5.6 cm |
| Polarization | HH | HH |
Minimizing speckle
Minimization of speckle effects in SAR data are commonly carried out using adaptive
radar filters (Lopes et al, 1990). In this study, Lee-Sigma filter at window size 7x7
showed the best result over mangrove forest in both images. This selection were
made based on the analysis of the mean vectors before and after filtering operation as
well as the coefficient of variance (Paudyal and Aschbacter, 1993).
Image Classification
The extracted pixels within the mangrove boundary were classified using combined
unsupervised-supervised approach with maximum likelihood classifier. In this
approach, the spectral classes generated in the unsupervised approach is refined based
on the existing forestry records and ancillary data. Once the samples from all
available classes within the area are known, training areas and signature vectors of
these classes were then generated before supervised maximum likelihood
classification was performed.
Biomass Estimation
In this study, we focussed on the estimation of mangrove biomass from radar
backscattering of JERS-1 and Radarsat SAR data. Regression analysis of the sample
biomass measured in the field with radar backscatter coefficient of JERS-1 and
Radarsat SAR were examined using stepwise regression approach. Based on the
regression analysis, the parameters describing the relationship of mangrove biomass
to radar backscatter were used to calculate the biomass of the entire area. The
computed biomass were then compared with the recently surveyed biomass of the
area by Forestry Department (1996).
Ground truthings and analysis
Ground truthings were carried out for two reasons: (a) verifying the classified SAR
data for accuracy analysis, and (b) to make in-situ measurements for biomass
estimation. For verification, survey random samples were identified in the field
where the position and corresponding class were noted, which later used in
contingency matrix for classification assessments. Global positioning system are used
in recording the positions of samples collected. In the biomass estimation,
measurement of mangrove tree samples at selected sites for consist of the tree basal
area, dbh (diameter at breast height), biomass by parts and density of trees.
