Menu
Research Questions/Objectives:
The overarching aim of this project is to test and calibrate different drone survey methods and vegetation filtering software to produce accurate and high-resolution digital terrain models of alluvial gully systems.
Brief Description of the Project: The Spyglass Beef Research Facility cattle station suffers from severe alluvial gully erosion due largely to its sodic soils. Optimising the digital terrain models by testing and calibrating drone survey methods and vegetation filtering software is a crucial step towards enabling accurate measurement of erosion rates and mapping of hydrology through the gullies. This will, in turn, enable the larger ongoing project to more accurately monitor and assess the effectiveness of varying remediation techniques on erosion of alluvial gullies, as well as their impact on the quality of water flowing out of the Burdekin catchment.
Background and Significance of the Research Question to drought risk, vulnerability, preparedness, or resilience: Critical to the drought resilience of any landscape are retention of water and vegetation. Gullies reduce a landscape’s ability to retain moisture needed to sustain plant growth and therefore support grazing production.
In something of a vicious cycle, the loss in drought resilience due to the presence of gullies exacerbates further erosion and gully formation as even more of the vegetation critical to moisture retention and erosion control is lost during periods of drought. The gullies become paths of least resistance for runoff, forming effective hydrological conduits that rapidly transport water out of the landscape, further decreasing the amount of water retained.
This loss of water, coupled with the loss of soil and nutrients due to erosion, means that management of these gullies is crucial for maintaining productivity and drought resilience while simultaneously having the benefit of reducing negative downstream impacts.
My research project focusses on novel techniques for quantifying how land surface condition (vegetation) and topography will be key inputs in our understanding of how water (and the erosion it causes) moves through the landscape.
The outputs and findings of this project will be broadly applicable to many landscapes, but especially relevant to all areas of Queensland with sodic soils – which include large swathes of the Burdekin Catchment, which contains the TNQ Hub agricultural areas with both the highest GVPs and the most drought susceptible landscapes.
Given that climate change is predicted to increase the intensity and duration of droughts nationwide, as well the intensity and frequency of extreme weather events including flood events that play major roles in alluvial erosion, the significance of this work on improving drought resilience will only increase.
This project is directly relevant to addressing 4 of the 5 priority themes of the TNQ drought hub:
Academic and research experience relevant to the honours project: This project will be the final component of my Master of Science (Professional) here at JCU, where I have focused my studies on skill-based subjects with a current GPA of 6.3.
As a result, I have acquired confidence with GIS (completing Introduction to GIS and Advanced GIS with high distinctions), coding in R, data management, scientific writing and broad field skills related to soil science, terrestrial and marine biology and hydrology.
I also completed a soil science subject taught by Jack, became familiar with the process of environmental and social impact assessments, and gained a high distinction for a subject on it.
Prior to this degree I completed a Bachelor of Science at the University of Melbourne, majoring in Marine Biology while almost meeting the full requirements of a botany major as well. The broad model of the degree gave me strong foundation of knowledge in ecology, botany, data management, scientific writing, conservation, biology, and chemistry.
I assisted field scientists on a David Lindenmayer project in the Victorian high country involving soil and leaf sampling in hardwood plantations before COVID resulted in it being terminated.
Principal Supervisor’s skills and experience in relation to this project topic: Dr. Jack Koci’s skills and experience in relation to this project are simultaneously broad and specialized and the key reason the academic component of this DAF remediation project is covering such an extensive array of fields and variables (such as water quality, hydrology, geomorphological changes, soil monitoring, multiple erosion monitoring techniques, and rainfall, among others).
Dr. Koci completed his PhD on gully erosion, has extensive experience with agriculture throughout his academic career and published extensively on hydrology, geomorphology, soil, erosion, grazing and gully systems.
Background
High-resolution topographic information is invaluable for the sustainable management of water in water-limited rangeland production systems, such as the gullied savanna rangelands of north-east Queensland. High-resolution topographic information, for example, enables assessment of where water is likely to move, be stored and be lost in the landscape. Topographic information can also be used to identify erosion hotspots and inform targeted remediation strategies. Topographic data can be derived from satellite products, but often the spatial resolution of readily available data is too large to enable detailed assessments at the paddock scale. High-resolution topographic information can be derived from aerial and ground-based Light Detection and Ranging (LiDAR) surveys, however, the high cost associated with these surveys precludes frequent data capture. Increasingly, low-cost drones are being used to produce high-resolution topographic information, through a technique known as structure-from-motion with multi-view stereo photogrammetry. While increasingly applied, few studies have quantified error and uncertainty in the topographic models produced in gullied savanna rangelands, and there remains uncertainty as to optimal surveying techniques. The angle at which images are captured, for example, can have a major influence on the accuracy of the topographic models produced, but few studies have investigated this in detail.
Aim
The overarching aim of this project was to evaluate the effect that camera angle has on the accuracy of drone-derived digital terrain models (DTMs).
Method
The study was conducted at two study sites in the Upper Burdekin Catchment, Queensland, Australia. At each study site, DTMs were constructed from images captured at camera angles including 90° (‘nadir’), 80°, 70° and 60°, across two flight altitudes (50 m and 25 m). A DJI Phantom 4 was the drone used for the surveys, with an image overlap of 80% front and side to ensure sufficient coverage of the sites. The images were then processed in the program Agisoft Metashape using photogrammetry to produce 3D topographical maps or digital terrain models (DTMs) of the sites as well as ortho-mosaics; high-resolution images of the entire site made by stitching all the drone images together. Elevation accuracies of the DTMs were assessed by comparing point elevations from the drone-derived DTM, with on-ground validation points, collected with a real-time kinematic global navigation system (RTK GNSS).
Key findings
Our study revealed intermediate oblique imagery, captured at camera angles of 70° and 80°, consistently yielded the lowest Root Mean Square Errors (RMSE) for both study sites and altitudes. These RMSE values ranged from 0.11 m to 0.14 m, consistently surpassing those of the 90° (‘nadir’) DTMs by 1-3 cm. The DTMs proved highly effective at reconstructing cross-sections from the RTK validation points, achieving RMSEs as low as 0.02 m. This accomplishment was achieved in a fraction of the time required for RTK GNSS surveys. Most notably, interpolated smooth patches in the DTMs, frequently encountered in areas with deep undercuts and dense vegetation, exhibited significant reduction. Specifically, between the 90° (‘nadir’) and 70° DTMs, we observed a reduction of approximately 12% for DTMs at a 50 m altitude and 56% for those at a 25 m altitude.
However, DTMs produced with a 60° oblique camera angle consistently exhibited RMSE values 2-3 cm higher compared to DTMs produced from the intermediate oblique camera angles (70°). Additionally, the 60° DTMs featured larger interpolated smooth patches.
In summary, our findings suggest that DTM elevation accuracy improves with the incorporation of intermediate oblique camera angle imagery. However, excessive angles moving away from the nadir position seem to negate these benefits.
Practical Applications
In summary, this research offers practical applications that extend beyond the scope of this study, encompassing gully remediation, advanced hydrological modelling, infrastructure planning, and soil health. These applications collectively contribute to enhancing drought resilience in Tropical North Queensland.
Acknowledgements
This project was made possible through invaluable support and funding from the Tropical North Queensland Drought Hub and the Queensland Department of Agriculture and Fisheries, reflecting a shared commitment to drought resilience in the region.
