Developing a framework for studying fine resolution impacts of climate change on forest ecosystems at regional scale using remote sensing and artificial intelligence

Abstract

Forests are one of the largest and most important terrestrial ecosystems that cover approximately 30% of the Earth’s landmass. Globally, forests are important carbon pools and provide a variety of ecosystem services. However, they are undergoing rapid changes due to climate change. The forests in California’s Sierra Nevada and Interior Alaska, for example, are under the threat of massive forest mortality, shift in species composition, and forest conversion to nonforest with corresponding changes in ecological processes. To overcome these challenges and build resistant and resilient forests under changing climate, we need relevant ecological assessment approaches and effective management practices that address the problems from tree scale to larger extents. To achieve this goal, fine-scale monitoring of forest conditions is required to detect potential type conversions and guide management interventions. One such approach is remote sensing-based methods validated with field data, which can reliably extend the field-based measurements across larger landscapes. To fully utilize the potential of the wide range of newly available remote sensing data, cutting-edge artificial intelligence techniques are required. In my research, I built a robust framework designed to generate high-resolution prediction maps of forest conditions at regional scale by integrating remote sensing and artificial intelligence techniques. In my dissertation I 1) developed an automated method for identifying the forest conditions (type and mortality status); 2) applied the method to examine which remote sensing data modality performs best for detection of the forest conditions; 3) produced forest condition maps at regional scale; 4) used prediction maps to derive metrics that is useful to managers and answer ecological questions. This work addresses open questions about large scale, fine resolution ecosystem patterns monitoring under climate change. The developed framework enables us to perform repeat, fine scale monitoring of forest condition which can be used to detect relevant ecological changes and assists managers in developing effective management strategies to mitigate climate change induced impacts.