Exploring the factors behind the discrepancy between two-dimensional and three-dimensional indicators of greenspace exposure

Abstract

Assessing greenspace exposure is vital in environmental health research due to its impact on human health. Advances in technology, such as Light Detection and Ranging (LiDAR), enable precise three-dimensional (3D) greenspace assessments, measuring the exact volume of greenspace exposure. Despite these advancements, traditional two-dimensional (2D) methods like the NDVI remain prevalent due to extensive research and data availability. Understanding the relationships and spatial discrepancies between 2D and 3D greenspace indicators is essential for improving public health strategies, as current gaps hinder comprehension of their impact on health outcomes and the application of 3D indicators in research. This study addresses this gap by explaining the inconsistent spatial patterns between the NDVI and greenspace volume, identifying landscape-related factors associated with these inconsistencies, and elucidating the implications of differences between the NDVI and volume on greenspace exposure measurement. We curated a set of landscape factors based on prior research and used an explainable machine learning technique to explore the associations between 2D and 3D greenspace indicators and these landscape elements. Our findings reveal that in highly developed urban and woodland areas, the greenspace exposure measured by vegetation volume tends to be higher than the greenspace exposure measured by the NDVI, while in areas dominated by low-growing vegetation, NDVI-based measurements are higher compared to the measure based on vegetation volume. Additionally, our study also indicates that the differences between the 2D and 3D greenspace indicators may be influenced by topographic factors. These insights offer strategic guidance for the application of 3D greenspace indicators in environmental health studies and inform future urban planning and policy decisions.