Evaluation of C-band SAR data for wetlands mapping

Citation
N. Baghdadi et al., Evaluation of C-band SAR data for wetlands mapping, INT J REMOT, 22(1), 2001, pp. 71-88
Citations number
23
Categorie Soggetti
Earth Sciences
Journal title
INTERNATIONAL JOURNAL OF REMOTE SENSING
ISSN journal
01431161 → ACNP
Volume
22
Issue
1
Year of publication
2001
Pages
71 - 88
Database
ISI
SICI code
0143-1161(200101)22:1<71:EOCSDF>2.0.ZU;2-C
Abstract
This publication reports results of an experiment carried out to examine th e potential of polarimetric C-band Synthetic Aperture Radar (SAR) for mappi ng various wetland classes found in the Mer Bleue region (near Ottawa, Cana da). The Mer Bleue region was surveyed by the C-band (5.3 GHz) polarimetric (HH, HV, VH, VV) SAR of the Canada Centre for Remote Sensing ( CCRS) at th ree times within the vegetation season: 16 June (spring flush for vegetatio n), 6 July (mature growth stage for vegetation) and 3 October 1995 (senesce nce). Signatures of six different cover types (forested and nonforested pea t bog, marsh, open water, clearing and forests) have been derived as a func tion of incidence angle. Separability between various classes was used to d etermine the relationships between season(s) and polarization(s) needed to differentiate various wetland classes. A supervised classification was used for wetlands mapping by means of multipolarization data. These investigati ons demonstrate some of the capabilities of SAR at C-band for mapping wetla nds. The cross-polarization data provided the best separation between the o bserved classes. The October dataset was better suited for discriminating b etween the classes present than the other periods observed. The overall acc uracies of the classification are 73% for June, 73% for July and 86% for Oc tober. Classification using a single polarization has been investigated and the results have shown that the HH and cross-polarizations are better than VV polarization. For October, the percentage of all pixels correctly class ified is 74% for HH polarization, 76% for cross-polarization, and 59% for V V polarization. Investigations were carried out to determine whether temporal changes can b e used to increase the information content of single polarization C-band SA R data, which are now available from ERS-2 and RADARSAT satellites. They de monstrated that the use of multitemporal data acquired in June, July and Oc tober do not provide a substantial amelioration of the classification of we tlands when the differentiation is not possible in any single period.