Developing machine learning algorithms to identify forest structural characteristics from earth observation data on a regional scale, Australia

The global environment is currently facing the challenge of climate change, which has severe consequences on ecosystems, biodiversity, and human societies. Forest ecosystems are particularly vulnerable to climate change, and their accurate monitoring and management are crucial to mitigate its impact. In recent years, Earth Observation (EO) data and machine learning (ML) algorithms have emerged as powerful tools to monitor and manage forest ecosystems. Therefore, the main goal of this project is to develop machine learning algorithms that can accurately estimate forest structural characteristics from the Global Ecosystem Dynamics Investigation (GEDI) lidar data. Specifically, the aim is to extract biophysical metrics from each GEDI waveform and map the forest structural characteristics using open-access Earth Observation (EO) data and machine learning algorithms on a regional scale, Australia. The results of this project can be used for sustainable forest management, carbon accounting, and biodiversity conservation. Overall, this project will contribute to our understanding of how machine learning algorithms and EO data can be used to monitor and manage forest ecosystems in Australia.

Below are some specific projects and tasks related to this broad topic that can be pursued, and further discussions can be brainstormed around them.

  1. Model Configuration Optimization: The student can focus on optimizing the configuration of machine learning models used to estimate forest structural characteristics from GEDI lidar data. They can experiment with different model architectures, hyperparameters, and optimization techniques to determine the most accurate and efficient configuration.
  2. Fuel Type Analysis: The student can investigate the impact of different fuel types on the estimation of forest structural characteristics. By focusing on specific fuel types such as grassland, shrubland, forest or woodland etc, they can analyze how the machine learning algorithms perform and identify any specific challenges or improvements needed for accurate estimation.
  3. Integration of Multiple Datasets: The student can explore the integration of additional datasets, such as high-resolution remote sensing data or Synthetic-aperture radar (SAR), to complement GEDI lidar data. By combining multiple data sources, they can assess whether the inclusion of additional information improves the accuracy and reliability of forest structural characteristic estimates.
  4. Validation with Ground Measurements: The student can validate the estimated forest structural characteristics using ground measurements. They can compare the results obtained from machine learning algorithms and EO data with the collected ground measurements datasets. This analysis will provide insights into the accuracy and reliability of the estimation techniques.

Keywords: Biophysical metrics; Earth observation; Forest structure; GEDI lidar data; Machine learning; Remote sensing

Updated:  25 May 2023/Responsible Officer:  Director, Fenner School/Page Contact:  Webmaster, Fenner School