Assessment of Airborne Lidar for Measuring the Structure of Forests and Woodlands in Queensland, Australia

Armston, John David (2013). Assessment of Airborne Lidar for Measuring the Structure of Forests and Woodlands in Queensland, Australia PhD Thesis, School of Geography, Planning & Env Management, The University of Queensland.

       
Attached Files (Some files may be inaccessible until you login with your UQ eSpace credentials)
Name Description MIMEType Size Downloads
s4043542_phd_submission.pdf Thesis full text application/pdf 36.19MB 0
Author Armston, John David
Thesis Title Assessment of Airborne Lidar for Measuring the Structure of Forests and Woodlands in Queensland, Australia
School, Centre or Institute School of Geography, Planning & Env Management
Institution The University of Queensland
Publication date 2013
Thesis type PhD Thesis
Supervisor Stuart Phinn
Richard Lucas
Peter Scarth
Total pages 197
Total colour pages 25
Total black and white pages 172
Language eng
Subjects 0909 Geomatic Engineering
0599 Other Environmental Sciences
Formatted abstract
Significant progress on quantifying state and trends in vegetation structure in Queensland, Australia, has been made by integrating in-situ measurements with satellite imaging datasets. Large field-based inventory and monitoring programs are expensive to maintain and there is often insufficient resources to accurately capture the spatial and temporal variability in structure, especially for scaling up to satellite remote sensing. Airborne lidar is an active remote sensing technology that provides quantitative information on the three dimensional structure of vegetation. Previous research has shown that lidar technology increases the scope and accuracy of metrics quantifying vegetation structure across a range of environments, however most commercial lidar systems are discrete return with different sensor and system properties between instruments. Details on these differences are often proprietary, which has made repeatable estimates of biophysical variables from different instruments difficult.

    There has been little investigation into the use of airborne waveform lidar technology for improving the estimation of canopy structure parameters for forest and woodlands in Queensland, Australia, or other areas. The aim of this research is to investigate methods for measuring the structure of forests and woodlands in Queensland, Australia, using waveform lidar remote sensing. This requires lidar derived estimates of vegetation structure parameters that are of comparable or higher accuracy than field measurements. These data are to be used to support the retrieval of biophysical attributes and detecting change from satellite and airborne remote sensing data over large areas. This thesis is divided into five chapters, with Chapters 2--4 each addressing a main research objective. The objective of Chapter 2 was to investigate if waveform lidar data can improve the estimation of canopy gap probability from airborne platforms. The objective of Chapter 3 was to determine the impact of sensor and survey properties on the estimation of vegetation structure parameters from waveform and discrete return lidar data. Finally, the objective of Chapter 4 was to develop and validate a practical approach for estimating foliage projective cover, leaf area index, and crown cover across multiple bioregions in Queensland, Australia.

    In Chapter 2 of this thesis, the estimation of canopy gap probability from small footprint waveform lidar was investigated using a RIEGL LMS-Q680i waveform lidar dataset acquired at multiple flying heights over a eucalypt dominated savanna in north eastern Australia. A method for estimating canopy/ground volumetric backscatter coefficients and canopy gap probability from small footprint waveform lidar data was presented, and compared to canopy gap probability estimates derived using published discrete return based methods. The waveform method produced consistently higher accuracy than discrete return methods in reference to field measurements of canopy gap probability, however the waveform method was sensitive to the statistical distribution of canopy and ground waveform integrals. Estimates of canopy gap probability from the waveform method also exhibited less variance than discrete return methods under varying flying heights. However, it was difficult to isolate the impacts of different sensor and airborne survey properties on the observed bias and variance. If validated, lidar simulations would have allowed a quantitative understanding of the limits on retrieval imposed by the interaction between sensor and survey properties, canopy structure and the signal-to-noise of waveforms. Such a modelling approach has not been developed for small footprint waveform lidar in Australian savannas.

    In Chapter 3, a simulation approach was developed to quantify the impact of airborne lidar sensor and survey properties on the estimation of gap probability profiles. Ground measurements of radiometry and vegetation structure at the leaf, plant and stand scales were used to construct geometrically explicit mixed tree-grass savanna scenes. Monte Carlo ray tracing was used to simulate the propagation of laser pulses through the scenes. Validation of scene structure parameters using field measurements of vegetation cover and area metrics showed close correspondence. There was greater variation in measured within-crown canopy gap probability than modelled. Further research is required to improve parameterisation of within-crown Eucalypt structure. Lidar simulations were run with sensor and survey properties matching a RIEGL LMS-Q680i waveform lidar dataset acquired at multiple flying heights. The simulated waveform and discrete return gap probability profiles replicated the direction and magnitude of bias in observed profiles. Simulation experiments performed across a realistic range of flying height, beam divergence and pulse rate values found that the sensitivity of discrete return gap probability profiles to footprint size was dependent on pulse energy transmission losses. Waveform gap probability profiles were insensitive to all sensor and survey properties, except when the signal-to-noise was very low. This was particularly evident at the short-wave infra-red wavelength of some commercial lidar instruments. Results showed that the modelling approach can be used to quantify bias in vegetation structure parameters derived from waveform and discrete return lidar datasets with disparate sensor and survey properties.

    Chapter 4 used existing gap probability theory to initiate development of a practical approach to the estimation of foliage projective cover, leaf area index and crown cover from airborne waveform lidar data. RIEGL LMS Q560 and Q680i waveform lidar transects were collated across eight bioregions in eastern Australia. Estimates of canopy/ground volumetric backscatter coefficients were derived from the waveforms and directional gap probability profiles were estimated. Field transect measurements were used to develop calibration factors for the estimation of foliage projective cover and leaf area index from these products. Crown cover was estimated by a model that accounts for the within crown and stand scale clumping of foliage. Comparisons were conducted for lidar derived estimates of foliage projective cover, leaf area index and crown cover with: (i) field transect estimates of foliage projective cover and crown cover; (ii) ground based lidar and hemispherical photography estimates of leaf area index; and (iii) lidar estimates from repeat flight paths. A key finding of the comparison was that the waveform and discrete return lidar estimates were repeatable and of comparable accuracy to the traditional field based techniques. However, spatial and temporal variation in the wood-to-total-plant-area ratio introduced error in some environments and in cases of tree mortality. Further research to separate wood and foliage plant area, ideally with multi-spectral lidar systems, is required to address this limitation.

    The key implication from the findings of this research for measuring the structure of forests and woodlands in Queensland, Australia, is that waveform lidar should be strongly considered in preference to discrete return data for estimating canopy gap probability, since biophysical models may then be transferable between waveform lidar datasets acquired using different instruments in different environments without the need for ground calibration. In the absence of field calibration, discrete return methods will introduce uncertainty into the detection of change. In addition, methods, tools and software libraries have been advanced to quantify the impact of airborne lidar sensor and survey properties and their interaction with canopy structure on waveform and discrete return lidar data products, and to apply airborne lidar data to the estimation of canopy gap probability and key vegetation cover metrics across multiple bioregions in Queensland, Australia. As these bioregions represent typical savanna vegetation communities, they may potentially be used in similar forest and woodland environments throughout the world.
Keyword Remote sensing
Savanna
Waveform
Discrete return
Lidar
Gap probability
Monte Carlo ray tracing

 
Citation counts: Google Scholar Search Google Scholar
Created: Thu, 19 Dec 2013, 12:00:52 EST by John Armston on behalf of Scholarly Communication and Digitisation Service