Relating multi-incidence angle RADARSAT-2 data to vegetation characteristics in the Lower Paraná River floodplain (Argentina)

NATALIA MORANDEIRA; FRANK AHERN; FRANCISCO GRINGS; BRIAN BRISCO; MATÍAS BARBER; PATRICIA KANDUS

Montreal

Workshop; Advanced SAR 2019 Workshop; 2019

Canadian Space Agency

The Paraná Riverfloodplain comprises large extensions of freshwater wetlands atsubtropical and temperate latitudes. In its lower portion, mosaics ofherbaceous vegetation dominate, dotted by shallow lakes, rivers andstreams. The resulting backscatter is due to the interaction amongleaves, emergent stems and open waters. A total of 11 fullpolarimetric RADARSAT-2 C-Band scenes were acquired in February 2018,differing in their orbit direction and beams (FQ5 to FQ29, meanincidence angles between 24.4° and 47.4°). Scenes were calibratedto the coherence matrix and geocoded. Single polarizationbackscattering (σ0HH, σ0HV, σ0VV), the HH-VV phase difference andthe contribution of the Freeman-Durden (FD) components (double bouncescatter, volume scatter, odd scatter) were extracted. During theacquisition period, the hydrometric level of the Paraná River wassimilar and relatively low. Field sampling was conducted in floodedsites dominated by the herbaceous species Schoenoplectus californicusor Ludwigia peruviana (4 and 3 monospecific sites, respectively).These species show contrasting architectures: Schoenoplectus hasalmost vertical stems and no leaves; while Ludwigia usually has amain stem with ramifications and leaves. Generalized linear models(GLM) were conducted in order to assess the relation betweenfield-obtained biophysical measures and SAR variables.ForSchoenoplectus, the overall pattern shows that σ0HH is higher thanσ0VV, and σ0HV is 5 to 10 dB lower than σ0HH. According to GLM,σ0HH decreases with increasing incidence angle and increases withabove-ground green biomass, mean height, moisture content and meanstem diameter. Models explain up to 72.8% of the total variation ofσ0HH. The backscattering decrease with incidence angle was notsignificant for σ0VV and σ0HV. σ0HV was affected mainly by meanheight, mean stem diameter and the number of green stems, with up to78.7% of explanatory power. σ0VV was mainly explained by moisturecontent and stem diameter, explaining 68.0%. The mean HH-VV phasedifference decreased with biomass, height and moisture content. FDdecomposition attributes the majority of backscatter to volumescattering. The contribution of double-bounce scatter is low and canbe explained up to 15.9% by moisture content and mean height; whilevolume scatter increases with all the vegetation variables, with meanheight explaining 45.0% of the total variation. Odd bounce scatterdecreases with incidence angle. For Ludwigia, σ0VV is only 1-3 dBbelow σ0HH, which is less than for Schoenoplectus, and σ0HV isstill 10 dB below σ0HH. GLM have less explanatory power than forSchoenoplectus. The decrease of σ0HH with increasing incidence angleis also observed. Incidence angle and above-ground green biomass ormean height, explain up to 44.5% of the total variation in σ0HH.Biomass also contributes to explain σ0HV and the number of greenstems contributes to σ0VV. Odd bounce scatter decreases withincidence angle.These results would lead to an improved understandingof the backscatter physics between C-Band SAR signal and herbaceouswetland stands.p { margin-bottom: 0.1in; line-height: 120%; }