Development and application of automated tools for high resolution gully mapping and classification from LiDAR data
Led by: Dr Andrew Brooks, GU
Project Summary
Accurately mapping gullies at high resolution and quantifying their key attributes is the critical first step in the process of prioritising and designing rehabilitation solutions. At least 40% of the accelerated erosion that is contributing to poor water quality in the Great Barrier Reef (GBR) Lagoon is sourced from gully erosion, demanding effective management and rehabilitation of these features. Current gully maps across the GBR are low resolution and overly simple, providing no differentiation between gully type. Airborne Light Detection And Ranging (LiDAR) is now widely recognised as being the best way to accurately map gullies at a landscape scale at a suitable resolution for management planning. Given the large volume of LiDAR data now becoming available, this project will develop and apply automated tools to enable the location of gullies to be extracted from LiDAR Digital Elevation Models (DEMs), along with key attributes of the gullies enabling them to be grouped into classes of similar gully types to aid prioritisation, management and catchment modelling.
Project Publications
Project Webinar – 4th December 2020
International Symposium on Gully Erosion, 21st-27th July 2019, Townsville
Project Description
Project Overview
This project builds on project 4.9 which is currently developing a typology of gullies to use as a basis for improving catchment water quality modelling, prioritising gully rehabilitation efforts and communicating to non-experts the diversity of gullies that exist in GBR catchments and across Queensland (Figure 1). A recent workshop attended by a wide range of catchment modellers, scientists, managers and policy advisors, has endorsed the draft typology and associated attribute database developed through National Environmental Science Program (NESP) Project 4.9. In this new project, we propose to further refine automated gully mapping approaches currently under development and develop new tools that will enable us to automate the attribute extraction and assignment of types to the mapped gullies from high-resolution LiDAR DEM data.
Gully erosion contributes roughly 40% of the fine suspended sediment delivered to the GBR, and consequently, gully rehabilitation and management is a high-priority activity across all GBR Natural Resource Management (NRM) regions. Both the Queensland and Australian Governments have recently invested significant resources into major gully rehabilitation programmes, through the Reef Trust 2 and 4 Programmes (AG); and the Queensland Government’s Major Integrated Project in the Bowen, Broken, Bogie (BBB), the Springvale Station Erosion Management Plan, and the Gully Innovation project (with Greening Australia). With these major investments (which in total amount to around $60M), and with new investments through the Great Barrier Reef Foundation (GBRF) fund, there is a pressing need amongst all parties to identify different gullies (and different gully processes) in the landscape and to prioritise management efforts and resources. This will ensure that appropriate treatments are applied to different gullies in the most cost-effective manner.
In light of the Australian Government’s recent $1M investment in a major LiDAR acquisition to support its Reef Trust gully and streambank rehabilitation programme, and the rapidly increasing areas of LiDAR data being acquired across Qld (e.g. the ROAMES data acquired by Fugro across Qld under contract to the Ergon), there is a pressing need to develop automated tools to map gullies within the newly acquired LiDAR data, according to gully types. It is now widely accepted that mapping gullies from LiDAR, particularly where coupled with high resolution multi-spectral imagery, provides a far superior product to that which can be obtained via manual and visual mapping from satellite imagery, so this project will map gullies and channels across a large extent of LiDAR data acquired in the Reef Trust project. The project will apply and further refine the typology developed in Project 4.9, and build on initial efforts applying this typology within the Qld Govt. Landholders Driving Change (LDC) Erosion Prioritisation Project in the BBB. In the Normanby catchment where the Australian Government has invested > $250K in LiDAR data acquisition over the last 10 years, change analysis of the full repeat LiDAR dataset will be undertaken to enable comparisons to be made of decadal-scale erosion rates according to the applied typology. This will not only provide a major update of the gully sediment yield data in the Normanby catchment but will enable erosion rates to be estimated in the Stewart, Annan and Jeannie River catchments, which have not been reflown since they were first acquired in 2009. Even in the Normanby catchment, only 30% of the original 2009 acquisition has subsequently been reflown. Ground validation will be undertaken in each region as part of the project, including significant in-kind from DNRME. Several workshops will be held throughout the project to review the gully typology and mapping results and approach.
Nature of Problem and Project Rationale
amongst agencies and organisations responsible for GBR catchment water quality management. There is a general consensus that we need to move beyond current manual and over-simplified gully mapping that underpins catchment models and GBR-wide sub-catchment water quality prioritisation approaches. The current gully mapping has only one variable (presence/absence spatial extent at 100m grid resolution) and otherwise the inherent assumption is that all gullies are equal in terms of sediment yield. This assumption introduces significant error into catchment models and reduces our capacity to prioritise rehabilitation expenditure.
With extensive areas of high resolution LiDAR data becoming widely available the opportunity now exists to improve the resolution of gully mapping by up to two orders of magnitude (metre resolution), and to classify these mapped gullies in a systematic way based on a range of automatically extracted and quantified metrics. This will dramatically improve our ability to prioritise gully management, improve catchment models and develop rehabilitation plans.
Approach
In this project we will further develop a variety of approaches that have been used in other parts of the world for mapping gullies or other landscape features (e.g. Castillo et al., 2014; James et al., 2007; Korzeniowska et al., 2018; Liu, et al., 2017; Passalacqua et al., 2010; Patten et al., 2018; Vendrusculo, 2014; Yang et al., 2017; Zahra et al., 2017). Given the variety of landscapes and gully types in the GBR catchments, initial work shows that no single existing technique is appropriate for mapping all gully types in all landscape settings. Consequently, we will be developing novel methods that combine aspects of approaches used elsewhere with new elements we have introduced based on our experience of GBR gully mapping. Initial progress shows that both alluvial and hillslope gullies can be mapped with a high degree of precision using these approaches (Figure 2 and 3), and thereby provide the basis for quantifying a range of gully metrics such as: width, depth, area, length, volume, slope, planform shape and cross sectional shape. These data, combined with other spatial datasets, will form the basis for the attribution of gullies according to the typology developed in Project 4.9.
NESP 2017 Research Priority Alignment
This project aligns with Tropical Water Quality (TWQ) Hub Research Theme 1 (Improved understanding of the impacts, including cumulative impacts, and pressures on priority freshwater, coastal and marine ecosystems and species) and specifically to Sub-theme 1.2 (Develop practical improvements to land management practices that will influence behavioural change and improve outcomes for tropical water quality and ecosystem health) in the NESP 2017 Research priorities.
Project Keywords
Gully database; Gully field guide; Gully classification; Gully mapping; LiDAR data analysis; machine learning.
Project Funding
This project is jointly funded through GU, Queensland Department of Environment & Science (Landscape Sciences) and the Australian Government’s National Environmental Science Program.
