Department of Geosciences, Georgia State University, 34 Peachtree Center Avenue, Atlanta, GA 30302, USA.
Department of Geography, Environment and, Spatial Sciences, Michigan State University, East Lansing, MI, USA; Remote Sensing and GIS Research and Outreach Services, Michigan State University, East Lansing, MI, USA.
Sci Total Environ. 2022 Mar 15;812:151432. doi: 10.1016/j.scitotenv.2021.151432. Epub 2021 Nov 5.
This study constructs two biophysical metrics; one based on Land Surface Temperatures (LST) and an integrated spectral index. The latter is an aggregate of Normalized Difference Vegetation Index (NDVI), Normalized Difference Bareness Index (NDBaI), Normalized Difference Water Index (NDWI), and Normalized Difference Built-up Index (NDBI). The goal is to determine how disparate weighting techniques, data transformation approaches, and spatial visualization pathways influence the computation of composite heat metrics. Using composite images made of aggregated images from late May to Early September within Google Earth Engine, we generated four composites by combining biophysical metrics with SoVI using equal and Eigen-based weightings informed by Principal Component Analysis (PCA). We compared equal interval classification, global and local Moran's as pathways for spatial visualization of hotspots. We utilized several data transformation techniques in a Geographic Information System (GIS), including rescaling, reclassification, zonal statistics, and spatial weighting. Mann Kendall and Sen's Slope detected and quantified monotonic trends in each spectral index. The results show that the LST biophysical metric and its composites indicate increased heat susceptibility over time, with disproportionately exposed core metro counties. The integrated spectral index and its proxies showed reduced vulnerability hence not a good proxy for LST. At the same time, the Mann Kendall and Sen's Slope found persistent increases in NDVI and NDWI and decreases in NDBI and NDBaI. However, opposite trends were evident in core city counties. The LST-based composites and spectral indices-based composites varied in the spatial-temporal distribution of hotspots. Disparate weighting mechanics, data transformation techniques, and visualization alternatives influence the magnitude and spatial-temporal distribution of heat hotspots.
本研究构建了两个基于地表温度（LST）和综合光谱指数的生物物理指标。后者是归一化差异植被指数（NDVI）、归一化差异光秃指数（NDBaI）、归一化差异水指数（NDWI）和归一化差异建筑指数（NDBI）的综合。目的是确定不同的加权技术、数据转换方法和空间可视化途径如何影响复合热指标的计算。我们使用 Google Earth Engine 中 5 月下旬至 9 月初聚合图像生成的复合图像，通过结合生物物理指标和 SoVI 使用基于主成分分析（PCA）的等权重和特征权重生成了四个复合图像。我们比较了等间隔分类、全局和局部 Moran 作为热点空间可视化的途径。我们在地理信息系统（GIS）中使用了几种数据转换技术，包括重缩放、重新分类、区域统计和空间加权。Mann Kendall 和 Sen's Slope 检测和量化了每个光谱指数的单调趋势。结果表明，LST 生物物理指标及其复合指标表明随着时间的推移，热敏感性增加，核心大都市县的暴露程度不成比例。综合光谱指数及其代理显示出脆弱性降低，因此不是 LST 的良好代理。与此同时，Mann Kendall 和 Sen's Slope 发现 NDVI 和 NDWI 持续增加，NDBI 和 NDBaI 持续减少。然而，核心城市县的趋势则相反。基于 LST 的复合指标和基于光谱指数的复合指标在热点的时空分布上存在差异。不同的加权机制、数据转换技术和可视化选择会影响热热点的大小和时空分布。
