Complementarity of X-, C-, and L-band SAR Backscatter Observations to Retrieve Forest Stem Volume in Boreal Forest

Maurizio Santoro,Oliver Cartus,Urs Wegmüller,Johan E S Fransson

doi:10.3390/rs11131563

Abstract

The simultaneous availability of observations from space by remote sensing platforms operating at multiple frequencies in the microwave domain suggests investigating their complementarity in thematic mapping and retrieval of biophysical parameters. In particular, there is an interest to understand whether the wealth of short wavelength Synthetic Aperture Radar (SAR) backscatter observations at X-, C-, and L-band from currently operating spaceborne missions can improve the retrieval of forest stem volume, i.e., above-ground biomass, in the boreal zone with respect to a single frequency band. To this scope, repeated observations from TerraSAR-X, Sentinel-1 and ALOS-2 PALSAR-2 from the test sites of Remningstorp and Krycklan, Sweden, have been analyzed and used to estimate stem volume with a retrieval framework based on the Water Cloud Model. Individual estimates of stem volume were then combined linearly to form single-frequency and multi-frequency estimates. The retrieval was assessed at large 0.5 ha forest inventory plots (Remningstorp) and small 0.03 ha forest inventory plots (Krycklan). The relationship between SAR backscatter and stem volume differed depending on forest structure and environmental conditions, in particular at X- and C-band. The highest retrieval accuracy was obtained at both test sites at L-band. The combination of stem volume estimates from data acquired at two or three frequencies achieved an accuracy that was superior to values obtained at a single frequency. When combining estimates from X-, C-, and L-band data, the relative RMSE for the 0.5 ha inventory plots at Remningstorp was 31.3%. For the 0.03 ha inventory plots at Krycklan, the relative RMSE was above 50%. In a retrieval scenario involving short wavelength SAR backscatter data, these results suggest combining multiple frequencies to ensure the highest possible retrieval accuracy achievable. Retrievals should be undertaken to target spatial scales well above the size of a pixel.

Highlights

The increasing number of orbiting platforms acquiring an unprecedented number of images of the Earth’s land surfaces fosters mapping and monitoring applications that cannot be achieved in timely manner with in situ observations
There is an interest to understand whether the wealth of short wavelength Synthetic Aperture Radar (SAR) backscatter observations at X, C, and L-band from currently operating spaceborne missions can improve the retrieval of forest stem volume, i.e., above-ground biomass, in the boreal zone with respect to a single frequency band
We investigated the properties of the Water Cloud Model (WCM) for each dataset (Section 5.2) and quantified the retrieval error per dataset based on a combination of estimates of stem volume (Section 5.3)

Summary

Introduction

The increasing number of orbiting platforms acquiring an unprecedented number of images of the Earth’s land surfaces fosters mapping and monitoring applications that cannot be achieved in timely manner with in situ observations. The SAR backscattered intensity from a forest combines (i) structural information related to the arrangement of the trees on the ground and the trees architecture, (ii) dielectric information related to the water content of the tree, (iii) properties of the soil underneath the forest in terms of surface roughness and moisture, and (iv) additional specific conditions of the environment at the time of image acquisition (e.g., presence of snow cover). The interplay of such factors depends on the frequency band at which the radar operates, the viewing geometry and the polarization of the microwave. The development of thematic mapping applications based on SAR data, requires careful consideration of these multiple factors

Objectives

Methods

Results

Conclusion