April 9, 2025

About the Malleefowl

Source: JJ Harrison, CC BY-SA 4.0 <https://creativecommons.org/licenses/by-sa/4.0>, via Wikimedia Commons

Governments place stringent conservation requirements on industries that traditionally disrupt the habitat of protected species. One of these species is Leipoa ocellata, or the Malleefowl. Outline Global has helped several mining companies identify and protect the Malleefowl on their mining tenements by investing in our EQBE™ environmental quality block engine product.

This article explores how EQBE™ finds malleefowl nests using AI, revolutionizing detection methods to provide a more efficient and accurate solution for mining companies.

The Malleefowl is one of three birds belonging to the mound-building megapode species in Australia (the brush turkey is another). About the size of a domestic chicken, it constructs large nest mounds to incubate its eggs and generally keeps a low profile.

You can find Malleefowl all the way across the south of Australia from inland Victoria to southwest West Australia. This bird holds cultural significance to the indigenous people, featuring in certain ‘Dreaming’ sites and trails in central Australia.

Why it matters

As a result of agriculture, mining, feral predation, roadkill and inappropriate fire regimes, Malleefowl communities are now reduced to isolated areas of remnant habitat. Nationally, the Malleefowl is listed as Vulnerable under the Environment Protection and Biodiversity Conservation (EPBC) Act 1999 and is recognised as threatened wherever it occurs. In WA, they’re protected by the Biodiversity Conservation Act 2016 (WA) – under this act, you can be fined up to $500,000 for impacting threatened species.

Mining proposal submissions in Western Australia must include surveys of vegetation and fauna in the tenements as part of their application process. These surveys assess the full extent of threatened species populations, including Malleefowl, in the affected areas.

The challenge of manual detection

Source: https://www.malleefowlvictoria.org.au/about-malleefowl

How do we determine where these elusive birds are? The answer lies in the Malleefowl nests. A typical Malleefowl nest (or ‘mound’) is about four metres across and 75cm high but they can be constructed to an impressive size of 22m in circumference (7m radius entrance) and 1m high.

Traditionally, Malleefowl nests have been painstakingly mapped using costly and time-consuming ground surveys. Due to population decline, nests are becoming increasingly widely scattered, with an average distribution of 2 active mounds per km2. Mounds can also be abandoned or moved, and new mounds constructed. In addition, mounds are often camouflaged by soil and leaf matter and hidden underneath trees.

Malleefowl nests are too challenging for human surveys, too expansive for drone coverage, and too well-camouflaged for satellite detection. To confidently detect Malleefowl nests we need high accuracy and intelligent insights over vast areas.

This is Outline Global’s sweet spot — and a great application of our EQBE™ product.

How EQBE™ finds Malleefowl nests

Introduction to our EQBE™ technology

EQBE™, Outline’s quality block engine, is our turnkey solution for Malleefowl nest detection using AI. This advanced technology operates in six dimensions to analyse complex terrain data:

X, Y, Z: Creating a precise three-dimensional model of the terrain
Time: Tracking changes over different periods
Change: Identifying alterations in the landscape
Impact: Assessing the significance of detected changes

This multidimensional approach allows us to predict Malleefowl nest locations with high confidence, even in challenging terrains.

Benefits of Malleefowl nest detection

EQBE™ offers several key advantages for Malleefowl nest detection:

Comprehensive coverage: Efficiently surveys large areas, ideal for determining the full extent of Malleefowl populations.
Penetrative imaging: Utilises LiDAR technology to detect nests hidden under vegetation.
High accuracy: Combines aerial imagery and LiDAR data for precise nest identification.
Time efficiency: Significantly reduces the need for time-consuming ground surveys.
Cost-effective: Minimises fieldwork while maximising nest discovery rates.
Adaptability: Our machine learning models continuously improve, enhancing detection accuracy over time.

Methodology

Outline’s methodology for Malleefowl nest detection using AI involves several key steps that integrate high-resolution aerial imagery and LiDAR data with geospatial and computer vision techniques to train a deep-learning object-detection model.

EQBE™ three-step methodology

STEP 1: Data Collection

Aerial imagery and LiDAR capture: Our survey operations cover large areas quickly, ideal for determining the full extent of Malleefowl populations.
Data processing and DTM creation: LiDAR data is processed to create a Digital Terrain Model, penetrating canopy cover to reveal hidden nests.

STEP 2: Data Modeling

Training data preparation: The model is pre-trained on Outline’s vast datasets, enabling redirection to Malleefowl nest detection with reduced training data requirements. We create comprehensive training datasets using bounding boxes and customised normalisation techniques.

Data modelling – bounding boxes

Model training process: Our pre-trained model is fine-tuned for Malleefowl nest detection and optimised for small feature recognition in aerial data.

Model fine-tuning

STEP 3: Analytics

Prediction generation and confidence scoring: The model analyses the aerial imagery and LiDAR data to identify potential Malleefowl nests. It then assigns each potential nest a confidence score, indicating how likely it is to be a genuine nest. This scoring system helps experts prioritise which locations to review, making the process more efficient.

EQBE Malleefowl mound layers

Detection confidence

POI (Points of Interest) analysis: This secondary analysis validates the likelihood of detected mounds being actual Malleefowl nests.

This methodology significantly enhances nest detection efficiency, minimising fieldwork while maximising discovery accuracy.

Results and efficiency

Example of detection results are shown in the table below.

Detection results sample table

While not fully automating nest detection, the model significantly enhances detection efficiency, minimising fieldwork while maximising nest discovery.

What our customers are saying

We recently helped on the Kalgoorlie Nickel Project – Goongarrie Hub to identify Malleefowl nests on their tenements with high-confidence predictions.

“The EQBE system was ground-truthed by KNP, with an impressive accuracy rate.” Justine Hyams, Approvals Manager, Kalgoorlie Nickel

This real-world example highlights how EQBE™ delivers actionable insights for mining companies, streamlining environmental compliance and conservation efforts.

Why choose Outline Global?

Continuous improvement of ML pipeline through expertise and a wealth of data

Our team of experienced geospatial data scientists has been refining our machine learning pipeline for years, continuously improving our methodology to ensure optimal accuracy and efficiency in Malleefowl nest detection.

Our ML models leverage over a decade of high-quality geospatial data, allowing our algorithms to:

Recognize long-term patterns in nest construction and abandonment
Adapt to environmental changes affecting Maflleefowl habitats
Improve prediction accuracy by learning from a diverse set of historical examples

By integrating advanced data modelling, rigorous training processes, and sophisticated analytics, EQBE™ transforms raw geographical data into actionable insights for conservation efforts.

Discover how you can use EQBE™ to protect Malleefowl on your mining tenements.

Request a Demo of EQBE™ for Malleefowl Detection.