Abstract
D. Phil. (Energy Studies)
Woody biomass resource, mostly in the form of fuelwood and charcoal, is the predominant source
of basic domestic energy for low-income rural and urban households in sub-Saharan Africa. In
most developing economies, quantitative information on available woody biomass resources, at
scales appropriate for energy planning purposes is often lacking. The continued reliance on
biomass resources to meet the sustenance and livelihood needs in poor economies is exerting
unsustainable pressure on the resources. The VW Foundation initiated a multi-institutional and
inter-disciplinary bioenergy modelling project that sought to provide stakeholders with quantitative
information on available woody biomass and its sustainable utilisation in rural communal
woodlands in three countries in southern Africa. The overall project themes related to:
(i) remote sensing approaches for quantifying woody biomass
(ii) modelling rural energy at the village level
(iii) biomass conversion technological pathways
(iv) environmental and socio-economic analysis of fuelwood consumption
(v) spatial-temporal analysis of the communal woodland dynamics.
The exploitation of traditional biomass resources, for cash and mercantile purposes, is
leading to accelerated losses of carbon sinks, natural forests and biodiversity, as well as creating
local scarcity of woody biomass. Although communal woodlands are important sinks for carbon
sequestration, for which developed countries are willing to pay, many poor communities are often
caught between conserving communal woodlands and meeting their immediate domestic energy
needs. Carbon stocks in communal woodlands are becoming crucial input for the reporting
requirements of international conventions such as Reducing Emissions from Degradation and
forest Deforestation (REDD) and the Kyoto Protocol on Climate Change (KPCC).
This research investigated the capability of using full polarimetric spaceborne ALOS
PALSAR retrievals for mapping and quantifying above ground woody biomass in semi-open
savanna woodlands in South Africa, Mozambique and Zambia. Existing allometric equations were
used to estimate above ground biomass densities from tree parameter measurements in selected
training plots. The optimum polarisation channels for estimating standing woody biomass in
savanna woodlands were ascertained by investigating the correlation between above ground
biomass densities and normalised backscattering coefficient ( σ o ) from retrievals acquired using
the horizontal transmit and horizontal receive (HH), horizontal transmit and vertical receive (HV),
and vertical transmit and vertical receive (VV) polarisation states, under both wet and dry
conditions. The training datasets were bootstrapped since the number of the training plots was
limited. Regression and prediction equations have been established between the above ground
biomass densities and backscatter intensities for the resampled training dataset, with the highest
correlation coefficient for each polarisation.
The method developed in this work identifies woody vegetation from the interaction of full
polarimetric radar signals with terrain scattering mechanisms and maps the distribution of woody
vegetation at any required scale. Terrain scattering mechanisms were classified by (i) performing
an unsupervised entropy/alpha (H/α) Wishart classification procedure on ALOS PALSAR full
beam imagery, based on the Cloude-Pottier decomposition, and (ii) using scattering classes from
the unsupervised classification as training input for the maximum likelihood classification
procedure on Freeman decomposition data. The classification results were used to delineate woody
and non-woody vegetation classes. Equations to predict above ground biomass densities were
developed by inverting the regression equations, which were established from the relationship
between backscatter intensities and above ground biomass densities from selected training plots.
The predicted biomass densities were reclassified into five categories and the mean biomass
density values for the categories were used to compute the available woody biomass resources at
the desired scale. The woody vegetation classes were used to mask biomass densities estimated
from prediction equations.
The research has contributed to an improved understanding of the interpretation and analysis
of full polarimetric spaceborne retrievals acquired over African semi-open savanna woodlands by
extending approaches developed previously for boreal and temperate region forests. The research
work has succeeded to map and to some extent, quantify above ground woody biomass at
landscape scale, using full polarimetry spaceborne ALOS PALSAR retrievals validated against
plot-scale measurements. The results contribute to the estimation of woody biomass resources for
national and global energy and carbon sequestration initiatives. However, the training dataset was
limited; hence, the resultant biomass estimation equations are site specific. The approach developed
in this work needs refinement before it can be utilised for operational monitoring of savanna
woodlands and extrapolation of landscape scale biomass. Estimating woody biomass from the
polarimetric retrievals is an improvement on techniques based on optical remote sensing methods
because the response of the radar signal is responsive to the physical parameters being surveyed
(mass/volume of woody biomass) rather than a surrogate (top of canopy greenness).
The purpose of the VW Foundation Bioenergy Modelling project was to estimate available
and accessible biomass within village precincts. Results from this work were used to develop
GIS-based spatial models for estimating fuelwood collection times and the associated 'least-cost'
collection routes between households and selected woodlands. The models took into account the
constraints imposed by land tenure systems and geophysical factors such as terrain and
watercourses to compute the time spend on fuelwood collection. The fuelwood collection effort is
used to estimate the balance between rates of exploitation and woodland regeneration in order to
determine the point at which harvesting fuelwood becomes unsustainable. An important output
from the GIS models is a woodlands at risk map, which ranks the vulnerability of woody biomass
resources in terms of travel times from surrounding villages.