Abstract
With the expected launch of polarimetric spaceborne Synthetic Apaerture Radar (SAR) sensors over the next few years, namely ENVISAT ASAR (C band) and PALSAR (L band), there is an opportunity for regional estimation of forest and woodland biomass in Australia. However, due largely to political pressures, there is an increasing need to provide reliable estimates of Australia's greenhouse gas emissions resulting from changes in vegetation biomass within a relatively short timeframe. For this reason, the use of existing Landsat-based methods for biomass estimation is likely to continue, despite the known limitations of optical sensors for this purpose and the demonstration internationally of the benefits of SAR. This approach is, however, justified given the uncertainty in the potential use of future SAR sensors.
To address this uncertainty, research was undertaken to investigate how current airborne SAR sensors may be used to assess the potential of future SAR sensors for biomass estimation. This was undertaken by comparing relationships established with above ground/component biomass and NASA JPL AIRSAR data with those established with existing spaceborne SAR data at similar wavelengths and polarisations.
The study was conducted in an area near Talwood, southern Queensland. Ground estimates of above ground and component biomass were generated by applying a range of allometric equations to tree size measurements collected, in 1998 and 1999, from 44 fixed and variable area plots. Relationships were then established between total/component biomass and Japanese Earth Resources Satellite (JERS1) SAR L band HH, Space Shuttle Imaging Radar (SIR-C) SAR C band VV and L band VV, and NASA JPL AIRSAR C band VV, L band VV and fully polarimetric P band data, all acquired over the period 1994 to 1996.
The study indicated of C, L and P band data at all polarisations occurred at an above ground biomass of approximately 20-30 Mg ha-1, 60 Mg ha-1 and 80-100 Mg ha-1 respectively. Different SAR wavelengths were also shown to interact with different components of the biomass. Furthermore, the saturation levels and relationships observed were relatively consistent between sensors. The study concluded that the AIRSAR can be used to determine the likely potential of future spaceborne SAR sensors for biomass estimation. Furthermore, the added benefits of using SAR for biomass estimation, and hence for better quantifying greenhouse gas emissions from land use change and forestry, were demonstrated.
1. Introduction
Although a number of studies have demonstrated relationships between SAR data and vegetation biomass, the majority have focused on forests in tropical regions and the northern hemisphere. The research presented in this paper addresses the potential of SAR for quantifying the total above ground biomass (TAGB) of Australia's Eucalyptus woodlands.
2. Specific Objectives
The study aimed to determine whether
a) statistically significant relationships exist between polarimetric C, L and P band data and component (leaf, branch and trunk) biomass, and
b) relationships are consistent regardless of whether TOPSAR and SIR-C SAR C- and L-band data are used.
The research formed part of a pilot study aimed at determining whether a more rigorous investigation into the use of SAR for biomass estimation was warranted. For the study, only historical SAR data sets were available.
3. Study Area
The study was conducted in Queensland where, for the period 1990 to 1995, 56 % (1 million ha) of the total area of native woody vegetation clearance in Australia has occurred, averaging 268,060 ha
yr
-1. The study region, near Talwood (28º29'S, 149º28'E) supported extensive areas of open woodland consisting largely of Poplar Box (Eucalyptus populnea), Silver-leaved Ironbark (E. melanaphloia), Belah (Casuarina cristata), Brigalow (Acacia harpophylla) and White Cypress Pine (Callitris glaucophylla). Common understorey species included Wilga (Geijera parviflora) and Sandalwood Box (Eremophila mitchelli). Due to the complex nature of land use and management practices, the landscape consisted of a mosaic of cleared fields and woodland communities in various stages of degradation and/or regeneration.
4. Methods
Space Shuttle Imaging Radar (SIR-C SAR; C-and L-band VV polarisations), and NASA JPL TOPSAR (C-band HH, L-band VV and fully polarimetric P-band) were available for the study area for the years 1994 and 1996 respectively. Pre-processing included georeferencing of all SAR data to Australian Map Grid (AMG) coordinates, resampling to a pixel resolution of 12.5 m, and converting data to the backscatter coefficient
(
so ), defined as the average radar cross section per unit area of the individual scattering elements. For display purposes,
so was expressed in decibels (dB).
5. Field Data Collection
In October, 1998, field data were collected from 29 plots sited in woodlands in varying states of degradation and/or regeneration. In June, 1999, a further 14 plots were established largely in areas of younger regeneration. Although the field and remotely sensed data were collected several years apart, the rates of growth within these woodlands are relatively low and changes in biomass over the intervening periods were not considered to be substantial. Woodlands cleared or degraded since 1994 were not sampled.
Sampling was undertaken using fixed area plots in the early regeneration woodlands and variable area plots, sampled using the prism wedge method [1], in older regenerating and intact, albeit degrading, woodlands. The AMG co-ordinates of the centre of plots were obtained using a Global Positioning System with an accuracy of ± 10 metres. All trees > 3 cm diameter (D, at c 30 cm) were measured for D (at both 30 cm and 130 cm). Trees with D < 3 cm (at c 30 cm) were counted and their height (H, cm) was estimated. All trees measured were identified to genus or species. For selected genera/species, relationships were established between D and H. For all trees in each plot, the TAGB and component biomass was estimated using allometric equations [2,3], which required D and/or H as input, for Eucalpytus and understorey species. In all calculations, a bias correction factor was applied to avoid systematic bias when antilogging estimates to arithmetic units [4]. As allometric equations for biomass estimation were unavailable for Casuarina cristata and Callitris glaucophylla, the few plots containing these species were excluded from the analysis. For all plots, the component biomass was scaled-up to generate an estimate of component biomass per hectare (Mg ha
-1). For regions centred on each plot location, SIR-C SAR and TOPSAR data were extracted, averaged and related to component biomass.
