2.4.1 RADAR

Several satellite-based RADAR sensors are currently available for biomass studies. There are many papers available on this subject and the reader is advised to study the references provided to gain more insight into the application of RADAR data in biomass studies. Past studies showed that RADAR backscatter does not correlate well with stand parameters like DBH, height or even basal area (Hyyppa et al. 2000). This section thus provides an overview of the technology and application to bulk biomass estimates.

Theory — confirmed by studies — indicate that backscatter signatures at different RADAR frequencies, as well as polarisations of backscatter, result from scattering from different portions of the tree canopy and the ground surface, while slope and aspect also affect backscatter. The portion of the tree with which the RADAR energy interacts is a function of wavelength. Wavelength is described in bands: L-band SAR with 24 cm wavelength (e. g., JERS-1 satellite, HH polarisation), C-band SAR with 5.6 cm wavelength (e. g. ERS-1 and ERS-2 satellites, VV polarisation), and P-band SAR (airborne sensors) are the bands used in remote sensing applications. Longer wavelengths, like L-band and P-band, penetrate deeper into the vegetation canopy, and scattering of the radiation originates from trunks and large branches. At shorter wavelengths, like C-band, scattering occurs in the upper layers of the canopy from leaves and small branches.

SAR interferometry (InSAR) provides information on the topography of the surface and on temporal changes in certain land surface properties. This technique can retrieve both structural information of natural targets, which in some cases can be converted to biophysical parameters, and digital elevation models (DEM), which can be used for geocoding and radiometrically correcting SAR backscatter imagery. Interferometry is based on the principle that two SAR sensors image an area with the same sensor characteristics from different viewing positions. The two sensors are separated by a spatial baseline. From the two signals an interferogram can be computed from which two parameters can be derived: (i) the interferometric coherence as a measure of the correlation between the two signals; and (ii) the interferometric phase, which is related to topographic height (Balzter et al. 2002).

Kasischke et al. (1995) provide detailed discussions on scattering in SAR and the components of woody biomass. They summarise that the variation in stem biomass accounts for more of the variability in RADAR signature in the P-band HH polarisation, while highest correlation in VV polarisation occurred in the case of total biomass in the canopy layer. The authors also concluded that, as RADAR frequencies increases, overall sensitivity to variations in biomass decrease, which is from L-band to P-band to C-band. L-band exhibits similar biomass scattering in HH and VV polarisation to P-band. They conclude that multi-polarisation C-, L-, and P-band SAR data can be used to estimate biomass in pine forests with total aboveground biomass up to 40 kg m“2.

Carreiras et al. (2013) reported that their study used a machine learning algorithm to establish a relationship between in situ forest aboveground biomass (AGB) in Miombo woodlands in Mozambique and L-band Synthetic Aperture Radar (SAR) backscatter intensity (gamma nought, y°) data obtained from the Phased Array L-band SAR (PALSAR) sensor, on-board the Advanced Land Observing Satellite (ALOS). The algorithm used, unique bagging stochastic gradient boosting (BagSGB), as it also allows the production of spatially explicit estimates of prediction variability and an indication of the importance of each predictor variable. Estimates of forest AGB with a root mean square error (RMSE) of 5.03 Mg ha_1, based on a tenfold cross validation, were produced with their modelling approach. Also, the coefficient of correlation (r) between the observed and predicted forest AGB value was 0.95, again based on tenfold cross validation. The variable contributing the most to this model was the mean backscatter intensity for the HH polarisation, which was explained by the low tree canopy cover characterising Miombo savannah woodlands, thus invoking scattering mechanisms associated with this polarisation (e. g., trunk-ground scattering).

Saatchi et al. (2007) used RADAR remote sensing data to map biomass distribution of the Amazon basin. The RADAR data were combined with forest inventory plot data and optical remote sensing at 1 km resolution and ranges up to 400 Mg ha_1. Sun et al. (2003) also reported results from the fusion of LiDAR and RADAR data, which hints at more accurate measurement of biomass distribution on a worldwide scale. Studies by Santoro et al. (2002) concluded that stem volume retrieval was possible up to 350 m3 ha_1 in boreal forests, but that forest density still presents a challenge and that this was not possible at the stand level.