Urban heat mitigation effects and greenery justice in Washington, D.C.

The deterioration of urban heat exposure risks is highlighting the benefits of urban greenery for creating healthy and resilient cities. Using Washington D.C. as a case, this study demonstrates the urban temperature mitigation effects and effective ranges of GVI (Green View Index) and NDVI (Normalized Difference Vegetation Index) at street-scale, examines greenery justice in residential zones and affordable housing surroundings, the differentiated cooling effects of different genera. Other Street View Indices (SVIs) includes Road View Index (RVI), Building View Index (BVI) and Sky View Index (SKVI) were also applied to test their contribution to urban heat because the machine learning models have relatively high accuracy when identifying these objects. In methodology, this study first compared the performance of two machine learning models: (1) an OpenMMLab pre-trained large model; and (2) a self-trained small model. I then applied the latter one to conduct semantic segmentation analysis for 81, 053 Google Street View images from the Auto Arborist Dataset and generated SVIs, the metrics representing the proportions of objects from the street- and side-view. I visualize all SVIs and use Global and Local Moran’s I to conduct spatial autocorrelation analysis with the exception of discovering clusters. SVIs, together with NDVI are then applied to conduct Pearson’s correlation analysis with Land Surface Temperature (LST) to examine their contributions, either mitigation or exacerbation effects, to urban heat. This study found that LST shows statistically significant negative correlations with GVI and NDVI at p<0.001 level. The correlation coefficients between GVI, NDVI, and LST are -0.48 and -0.69 measured in the Circular Unit when the diameter is 30 meters, and the coefficients between GVI, NDVI, and LST are -0.47 and -0.66 measured in the Rectangular Unit when the width is 30 meters. The distributions of GVI, RVI, BVI, and SKVI follow patterns of strong spatial autocorrelation at 99.99% level. The clusters of GVI and SKVI are spatially complementary, while clusters of RVI and BVI are sort of spatially overlapped. Urban greenery in different residential zones was socioeconomically unfair, affordable housing projects tend to have less street greenery in vulnerable neighborhoods in ‘Local South’ wards, these gaps were also manifested from the differentiated cooling effects of tree genera. The thesis then discussed the advantages and disadvantages of street-level indices compared with remote sensing approaches, the issue of Greenery Justice and DC’s urban forest policies, relevant improvement strategies and the adaptability of two Machine Learning models. I hope these critical and dialectic discussions could inspire more related research. In conclusion, this study add to the literature by revealing the relationship between street greenery measured by GVI and LST, I argue that enhancing urban street green environments and achieving equitable cooling effects is of practical importance, which can be achieved by monitoring GVI, balancing tree genus, and improving the vertical quality of tree canopies, given GVI’s controllability and human-centered characteristics.

Name thesis:degree_name

M.U.P.

Level thesis:degree_level

Thesis

Discipline thesis:degree_discipline

Urban Planning

Grantor

University of Illinois Urbana-Champaign

Year dc:date

2025

Author dc:creator

• Zhang, Zhenpeng

Contributors dc:contributor

• Fang, Fang
• Greenlee, Andrew

• Urban Heat Island
• Temperature Mitigation
• Built Environment
• Greenery Justice
• Machine Learning
• Image Semantic Segmentation

Statement dc:rights

• Copyright 2025 Zhenpeng Zhang

Language dc:language

en, eng

Handle dc:identifier

https://hdl.handle.net/2142/129327

OAI record oai:identifier

https://www.ideals.illinois.edu/oai-pmh?verb=GetRecord&metadataPrefix=etdms&identifier=oai%3Awww.ideals.illinois.edu%3A2142%2F129327