Edinburgh Research Explorer

Spatial Wavelet Statistics of SAR Backscatter for Characterizing Degraded Forest: A Case Study From Cameroon

Research output: Contribution to journalArticle

Related Edinburgh Organisations

Open Access permissions

Open

Original languageEnglish
Pages (from-to)3572 - 3584
Number of pages13
JournalIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Volume8
Issue number7
DOIs
Publication statusPublished - 7 May 2015

Abstract

Forest degradation is an important issue in global environmental studies, albeit not yet well defined in quantitative terms. The present work addresses the problem, by starting with the assumption that forest spatial structure can provide an indication of the process of forest degradation, this being reflected in the spatial statistics of synthetic aperture radar (SAR) backscatter observations. The capability of characterizing landcover classes, such as intact and degraded forest (DF), is tested by supervised analysis of ENVISAT ASAR and ALOS PALSAR backscatter spatial statistics, provided by wavelet frames. The test is conducted in a closed semideciduous forest in Cameroon, Central Africa. Results showed that wavelet variance scaling signatures, which are measures of the SAR backscatter two-point statistics in the combined space-scale domain, are able to differentiate landcover classes by capturing their spatial distribution. Discrimination between intact and DF was found to be enabled by functional analysis of the wavelet scaling signatures of C-band ENVISAT ASAR data. Analytic parameters, describing the functional form of the scaling signatures when fitted by a third-degree polynomial, resulted in a statistically significant difference between the signatures of intact and DF. The results with ALOS PALSAR, on the other hand, were not significant. The technique sets the stage for promising developments for tracking forest disturbance, especially with the future availability of C-band data provided by ESA Sentinel-1.

Download statistics

No data available

ID: 19692515