Search

This dissertation, largely funded by the County of Los Angeles to optimize its urban forest planning and management, conducts a critical examination of the implications, applications, and limitations of remote sensing technology in applied science. Leveraging technologies such as Light Detection and Ranging (LiDAR) and high-resolution optical data, remote sensing has offered unparalleled insights into the Earth's surface and atmosphere. Specific applications, such as tree crown segmentation and creation of precise 3D environmental models, have proven invaluable in fields including ecology and urban planning. However, this technology is not without its challenges. An over-reliance on remote sensing data, absent corroborative ground-truthing, can lead to flawed conclusions. Furthermore, using low-resolution imagery for policy decisions concerning fine-scale, on-the-ground issues could create discrepancies between the data and actual conditions, undermining the effectiveness of interventions. In my first and second chapter of this dissertation, I demonstrate that remote sensing technology has become indispensable, particularly for monitoring individual species over time for urban forest management. By identifying and delineating individual tree crowns, this technology enables accurate estimation of species composition and condition. This detailed information facilitates species health monitoring and early disease detection, thereby aiding effective management interventions. However, the effectiveness of remote sensing in urban forestry necessitates continuous data update and integration of ground-based observations for validation. In my third chapter, I transition from the practical applications of remote sensing in urban forestry, and lean on previous research experience, including my collaboration with Los Angeles County, to delve into the social dimensions of this technology. The concept of "socializing the pixel" provides a critical framework for understanding how remote sensing operates within and is influenced by sociopolitical contexts. This perspective recognizes that remote sensing is not a purely objective tool but is embedded within social structures and power dynamics. Stakeholder engagement is thus crucial to consider diverse perspectives in data interpretation and application, enhancing the technology's effectiveness and local relevance. Conversely, power imbalances between remote sensing experts and non-experts, and denial of local knowledge can lead to biased results or misinterpretation of data. Hence, returning to the concept of "socializing the pixel" is critical to ensure effective and equitable use of remote sensing technology.

Tropical dry forests have been estimated to comprise 42% of all tropical forested biomes and are believed to be one of the world's most endangered ecosystems. There is a growing interest in identifying forest extent and forest change in tropical dry forest regions, especially to identify dry forest that deserve a high conservation priority at a global spatial scale. There is currently a debate concerning the classification and extent of tropical dry forest at the global scale. We identify the extent of tropical dry forest regions based on commonly used climatic definitions and datasets to improve global estimates of tropical dry forest extent. We compare climatic definitions of tropical dry forest (Murphy and Lugo, FAO, Dryflor, Aridity Index) using Worldclim, CHELSA, and Global Aridity and PET climatic datasets (1 km) and compare results to the World Wildlife Fund's Terrestrial Ecoregions (Tropical and Subtropical Dry Broadleaf Forest), as well as 579 field plots identified as tropical dry forest. Understanding the best method to estimate global tropical dry forest extent gives both researchers and policy makers a vital tool to begin protecting this critically endangered and valuable resource. We identify methods that most accurately predicted tropical dry forest extent. The global extents of tropical dry forest regions varied significantly with the Aridity Index predicting the largest extent, Murphy and Lugo and FAO predicting similar extents, and DryFlor predicting the smallest extent regardless of climatic dataset used. Globally, there was low agreement between climatic definitions and WWF Ecoregions. FAO and the Aridity Index climate definitions had the highest agreement with WWF Tropical and Subtropical Dry Broadleaf Forest Ecoregions (57%) while FAO (76%) and Murphy and Lugo (69%) definitions had the highest agreement with field plots. However, extents and accuracy varied significantly by regions, biodiversity hotspots, and island archipelagos. Tropical dry forest region extent varies significantly based on climatic definition but not climatic datasets at a global spatial scale. Nearly half of all tropical dry forests will be missed when only analyzing WWF Ecoregion boundaries and climatic definitions will be needed to estimate dry forest cover and change. There was high heterogeneity among climatic definitions at a regional and local spatial scale suggesting that climate definition can only provide a first order hypothesis about the distribution of dry forests and data on phenology, forest structure, and composition are still needed to compare local tropical dry forest extent.