Growing stock volume estimation from L-band ALOS PALSAR polarimetric coherence in Siberian forest

This paper investigated the potential of L-band ALOS PALSAR full polarimetric radar data for forestry applications. In order to characterize the polarization response as a function of growing stock volume (GSV), weather conditions and different tree species in Siberian forests, the polarization phase difference between HH and VV polarizations and the magnitude of the polarimetric HHVV-coherence have been analyzed. Moreover, the retrieval accuracy of forest growing stock volume (GSV) has been evaluated. The investigations were conducted on forest stand level and maximum GSV was 350 m(3)/ha. Using multi-temporal L-band ALOS PALSAR polarimetric coherence data and extensive sets of the forest inventory data at a stand level, we observed a clear decrease of polarimetric coherence with an increase in GSV, with Pearson's correlation coefficient between -0.74 and -0.79. The correlation improved to -0.87 for the forest stands with high relative stocking >70% (i.e. more homogeneous forest stand structure) and the sensitivity of the polarimetric coherence to GSV increased to similar to 250 m(3)/ha GSV. The polarization phase difference (PPD) also allowed forest to be discriminated from non-forest areas. Comparison of PPD during thawing and unfrozen weather conditions at the time of image acquisition revealed a higher contribution of a volume scattering at the latter conditions. In addition, differences in polarimetric coherence and phase difference were observed as a function of tree species. In particular, stands of larch typically exhibited a +0.1 polarimetric coherence under unfrozen conditions compared to aspen, birch and pine. There was also less variability in the PPD compared to other tree species, both under thawing and unfrozen conditions. An exponential model was used to characterize the polarimetric coherence behavior as a function of GSV. The retrieval of GSV was better under unfrozen conditions, with an RAISE of 33 m(3)/ha and R-2 of 0.73. The retrieval error was slightly higher (RMSE = 51 m(3)/ha and R-2 = 0.61) under thawing conditions. In the case of the forest stands with the area larger than 30 ha and the relative stocking of at least 70%, the GSV retrieval error decreased to RMSE of 16 m(3)/ha and to RMSE = 19 m(3)/ha respectively. (C) 2014 Elsevier Inc. All rights reserved.