Use of Radarsat-2 Polarimetric SAR Images for Fuel Moisture Mapping in Alaska Boreal Forests and South Africa Savannahs
Monday, 15 December 2014
The study reported a comparison between two Radarsat-2 polarimetric SAR (polSAR) images from extreme dry versus wet conditions are compared in an effort to determine the value of using polarimetric SAR data for estimating fuel moisture over South Africa savannahs and Alaska boreal forests. The savannahs study area is located into the Kruger National Park area and has 36 sites of lowveld savannas from bare overgrazed sites to medium-dense savannahs. The boreal forest study area has a chronosequence of black spruce ecosystems (recent burns, shrub-dominated regenerating forests , open canopied forests, moderately dense forest cover). Both study areas have a fairly level topography suitable for radar studies. The polSAR images were acquired using the same beam mode (FQ5 (23-25° incidence angle over the boreal sites, FQ15 (34.47–36.05° incidence angle) over the savannahs sites). Over each study area, soil moisture and vegetation structural data were measured in situ concurrently to the acquisition of the SAR imagery. The polSAR images were filtered for speckle noise using a Lee sigma filter and several polarimetric products were computed, such as those directly derived from the images (single linear and polairzed backscatters, polarimetric discriminators) and from target decompositions (Freeman-Durden, new van Zyl, Cloude-Pottier). Because most of these variables have a different unit, a normalized difference (in %) for each variable was calculated using the median values of the dry and wet dates for easier comparison of variable changes between the dates. Over both study areas, the normalized difference between wet and dry conditions was lower when higher tree canopy occurs. Results show utility of C-HH and C-RR polarized backscatters. Several polarimetric discriminators (dmin, Pr max, Pr min, Smax, Smin) were also significantly affected by the soil wetness. The Freeman Durden and van Zyl decomposition parameters outperformed the Cloude-Pottier decomposition parameters.