Individual Tree-Level Water Status Inference Using High-Resolution UAV Thermal Imagery and Complexity-Informed Machine Learning

This chapter introduces an innovative methodology for inferring the water status of individual trees through the integration of high-resolution Unmanned Aerial Vehicle (UAV) thermal imagery and Complexity-informed Machine Learning (CIML)Complexity-informed machine learning (CIML). The chapter opens with an insightful introduction, emphasizing the critical role of individual tree-level water status assessment for digital agriculture. Materials and methods are meticulously described, covering the experimental site, irrigation management practices, ground truth data collection through infrared canopy and air temperature measurements, acquisition of thermal infrared remote sensing data using UAVs, and the application of Complexity-Informed Machine Learning (CIML). The principle of tail matching is introduced as a key component of the methodology, along with various machine learning classification algorithms employed for analysis. Results and discussion present a comprehensive analysis, including a comparison of canopy temperature per tree based on ground truth and UAV thermal imagery, the relationship between temperature differentials and irrigation treatment, and the classification performance of CIML on different irrigation treatment levels. These findings provide valuable insights into the efficacy of the proposed methodology in accurately assessing the water status of individual trees. The chapter concludes with a succinct summary, encapsulating key findings and their implications.