The tropical savannas that cover more than 25% of the Australian continent and about 75% of the State of Queensland are predominantly used for cattle grazing.
Obtaining regularly updated information about the total vegetation cover of these grazing systems is particularly critical for natural resource management (NRM) in the tropical semi-arid catchments in Queensland because they drain into the Great Barrier Reef lagoon. Queensland's resource management agencies have a demand for higher temporal frequency, synoptic information on the biophysical properties of the total vegetative cover to improve existing erosion modelling and to determine sediment loads to the Great Barrier Reef. In this context, biophysical information includes vegetation structure and physiological properties.
Global remotely sensed products, such as the 8-day composites of the Moderate Resolution Imaging Spectroradiometer (MODIS) Fraction of Photosynthetically Active Radiation (FPAR) absorbed by a canopy, provide high temporal frequency biophysical information. However, the potential of the MODIS FPAR to complement regionally developed remote sensing products and support erosion modelling in Australian savannas has not been determined.
The main aims of this thesis were: (1) to investigate the sensitivity of a time series of MODIS FPAR to vegetation structural properties and vegetation components mapped from Landsat Thematic Mapper (TM) 5 and Enhanced Thematic Mapper Plus (ETM+) 7 imagery; and (2) to determine the MODIS FPAR's ability to improve erosion modelling in a tropical semi-arid catchment adjacent to the Great Barrier Reef.
A simplified albedo-fraction of absorbed Photosynthetically Active Radiation (fAPAR)-cover-factor model was developed. This model demonstrated the suitability of the fAPAR to function as a predictor for the vegetative cover factor (vCf) commonly used in erosion models that vary with the green and non-green cover fractions. To investigate the suitability of MODIS FPAR for the purpose of deriving vCf in a the study area, a MODIS FPAR time series was quality controlled and assessed for its sensitivity to national land cover and regionally validated, remotely sensed vegetation products. Statistical analysis of a dry season MODIS FPAR imagery revealed a significant sensitivity to vegetation structural categories (VSC) derived from a Landsat TM/ETM+ woody foliage projective cover (wFPC) product. A significant inverse relationship between the MODIS FPAR and percentage bare ground estimates from MODIS for woodlands to grass- and herblands was found. Analyses of the MODIS FPAR time series within the VSC quantified (i) the phenological and biophysical variability of the VSC, and (ii) covariations with the Landsat TM/ETM+ based green and non-green ground cover fractions by multiple regression analysis. The main findings of the first part of the thesis showed that 75% of the biophysical variability in MODIS FPAR could be explained by the Landsat TM/ETM based VSC and green and non-green ground cover fractions in open forests to grass- and herblands.
In the second part of the thesis, a time series of vCf maps was produced from MODIS FPAR and then integrated into a modified form of the Universal Soil Loss Equation to predict soil loss at high temporal resolution in the study area. Spatial patterns of dry season vCf maps for the study area mirrored patterns of land use and regional ecosystems. Strong similarities were observed between the high temporal frequency vCf estimates based on the time series of the MODIS FPAR and a time series of predicted Landsat TM/ETM+ FPAR derived from the multiple regression analysis of the first part of the thesis. A trend of decreasing protective function of the total vegetation cover was determined for most of the study area. Average annual vCf estimates derived from this study agreed with published, long-term average vCf for grazing areas in a Queensland savanna. To assess the feasibility of these high temporal frequency vCf estimates for erosion modelling they were integrated into a modified form of the Universal Soil Loss Equation to predict soil losses at high temporal resolution in the study area. High levels of agreement were found between predicted soil losses from the catchment and some in stream measurements of total suspended solids and streamflow.
This dissertation presents an integrated remote sensing approach to improving erosion modelling in a tropical semi-arid catchment adjacent to the Great Barrier Reef, Australia. Although the soil loss predictions derived were only of qualitative value, the MODIS FPAR, vCf, green and non-green ground cover fractions, and total vegetation cover maps produced provide unique high temporal information sources that have the potential to assist future land and water management in the region. The thesis makes a significant contribution to improved ecological understanding of the spatio-temporal dynamics of tropical semi-arid savannas of the study area and delivers insight into the factors that affect the PAR absorption at MODIS FPAR pixel and landscape scales. Future developments of remotely sensed biophysical products, field protocols for FPAR measurements, and ultimately comprehensive earth system models as well as event-based erosion models might benefit from these findings.