Vegetation height plays a critical role in understanding landscape dynamics, wildfire risk, and environmental planning. To support these applications, Geoneon has developed a high-resolution vegetation height dataset for all of Tasmania, capturing estimated average vegetation height at a 10m resolution using Sentinel-2 satellite imagery as of January 2023. This dataset provides comprehensive, scalable, and up-to-date insights into Tasmania’s vegetation structure, helping stakeholders make informed decisions in wildfire risk assessment, conservation & biodiversity management, urban & regional planning, and carbon estimation & climate resilience. 
Understanding Vegetation Height Mapping
Traditionally, accurate vegetation height estimation required LiDAR or stereo-photogrammetry from aerial surveys, which are expensive and time-consuming. Geoneon’s approach leverages deep learning applied to Sentinel-2 satellite imagery, offering a scalable alternative that balances accuracy with cost-effectiveness.
Our method involves a deep learning model trained on high-quality reference datasets that link visual spectral patterns from Sentinel-2 images to known vegetation heights. To create a cloudless and temporally stable dataset, we used multiple Sentinel-2 images from the period October 2022 to December 2023, combining them into a cloud-free mosaic representing vegetation height as of January 2023. By analysing spectral and spatial information across diverse landscapes, our model can estimate above-ground vegetation height for each 10m pixel, providing valuable insights that can complement more expensive solutions, like LiDAR acquisitions.
Methodology: AI-Powered Height Estimation
Geoneon’s AI-driven model follows a process designed to ensure high accuracy and consistency across diverse terrains:
- Data Collection & Training
- Sentinel-2 satellite imagery serves as the primary input, providing multispectral data at 10m resolution.
- A cloudless mosaic was generated using multiple Sentinel-2 images captured between October 2022 and December 2023, ensuring high-quality and representative vegetation height estimation for January 2023.
- Deep Learning Model
- A custom encoder-decoder neural network processes spectral and spatial information to infer vegetation height.
- Training data includes thousands of reference points where ground-truth height measurements were available, ensuring robust model learning.
- The model is optimised using a combination of Mean Absolute Error (MAE) and Mean Squared Error (MSE) to ensure precise height estimations.
- Validation & Accuracy Assessment
- The model’s accuracy is evaluated using metrics such as absolute error, relative error, and Delta metrics.
- The final dataset provides a reliable estimate of Tasmania’s vegetation height at 10m resolution, suitable for large-scale environmental analysis.
Key Applications
The Tasmania-wide vegetation height dataset offers significant benefits across multiple sectors:
- Wildfire Risk Assessment: Identifies areas with tall vegetation that may contribute to fire fuel loads, helping emergency services prioritize fire mitigation strategies.
- Conservation & Biodiversity Management: Supports habitat mapping, reforestation monitoring, and ecological studies.
- Urban & Regional Planning: Helps policymakers assess green space quality, tree canopy coverage, and urban heat mitigation potential.
- Carbon Estimation & Climate Resilience: Aids in estimating above-ground biomass, supporting carbon sequestration assessments and climate action plans.
Looking Ahead
This dataset represents the first step in a continuous monitoring initiative. We are already working on the January 2024 update, which will allow us to detect vegetation changes over the past year.
Access
Geoneon’s vegetation height dataset is now available. If you are interested in accessing this data or discussing potential applications, get in touch with us!
📩 Contact us to explore how our vegetation mapping technology can support your projects.
🔗 Subscribe to our newsletter for more updates.
