In rangeland production systems, accurate data on land condition is vital for decision making to ensure long-term sustainability and profitability. In Mitchell grass rangelands of northwest Queensland, the amount and distribution of Mitchell grass tussocks is a key indicator of land condition. Traditional land condition assessments rely on subjective visual methods, which are limited to small areas. The increasing accessibility of lightweight, low-cost drones, combined with advances in deep learning, allows for efficient quantification of vegetation at larger scales relevant to rangeland management. This study developed a methodology to estimate Mitchell grass tussock density and distribution by integrating high-resolution drone imagery with deep learning. The model achieved a detection accuracy of 83.1% and was used to assess land condition based on tussock density. This approach offers valuable insights into drone survey methods and image classification, providing a reliable tool for land condition assessment and sustainable grazing management.
.
Authors:
Lucy Gardner
