ORCID Identifier(s)

0000-0001-5481-2408

Graduation Semester and Year

Summer 2024

Language

English

Document Type

Dissertation

Degree Name

Doctor of Philosophy in Civil Engineering

Department

Civil Engineering

First Advisor

Jessica Eisma

Abstract

The global urban population is increasing, and it is anticipated that approximately 70% of people will reside in urban areas by 2050. Urbanization changes land use and land cover, altering local climatology. For example, various urban centers across the globe are experiencing extreme rainfall events, resulting in widespread damage to life and property with possible linkages to urbanization. The use of artificial materials in urban areas brings significant changes to the surface temperatures. Due to the high heat capacity of most of the construction materials, the temperature of the urban area can increase substantially compared to the rural areas. This results in urban heat islands. However, with the expansion of metropolitan areas, land use has diversified, creating micro-urban heat islands (MUHIs). This can cause severe thermal discomfort that is detrimental to human health. Moreover, the urban heat island also fosters rainfall and can initiate rainfall in urban areas. Therefore, investigating the urban influence on temperature and rainfall is vital, considering rapid urbanization, frequent extreme rainfall, and increasing temperatures.

The next three paragraphs will summarize the overall three broad objectives of the study: 1) Using a small unmanned aerial vehicle (UAV) to capture the influence of diverse land uses on the formation of micro-urban heat islands in different seasons, 2) Investigating extreme rainfall variations in response to urbanization using a high-resolution radar rainfall product, and 3) Leveraging tree-based machine learning methods to understand the climatological and land use effect on rainfall.

Land surface temperature (LST) measurement at high spatial resolution is indispensable to detect MUHIs, mitigate heat due to MUHI, and adapt mitigation strategies, which is difficult to obtain through coarser satellite observations (CSOs). As a potential alternative in remote sensing, unmanned aerial vehicles (UAVs) using thermal cameras can detect MUHIs by measuring LST at high spatial and on-demand temporal resolution. The first objective used the Zenmuse H20T onboard a UAV providing LST at ∼8 cm resolution to evaluate MUHIs in an area with diverse and contiguous LUs including three urban built-up LUs: 1) residential high cost (RHC), 2) residential low cost (RLC), 3) industrial area (IA) and one natural area (i.e., park area (PA)) within the Dallas Fort Worth (DFW) metroplex. The study was conducted in two seasons: fall (October 2022) and summer (June-July 2023). The loss of information between coarser (Landsat-derived LST) and finer spatial resolution (UAV-derived LST) was also assessed. A maximum MUHI of 25.54◦C and 15.85◦C was identified in the summer and fall seasons between 15:30 and 16:20 from the UAV. Landsat underestimated MUHI hotspots in the RHC by 15.62◦C and 10.86◦C in the summer and fall seasons, respectively. The Landsat not only underestimated the LST but also overestimated LST at places with lower LST. The UAV-derived LST was validated with on-ground measured LST, which showed good results in both seasons. UAVs are a practice tool that can be used to detect localized MUHI, which CSOs overlook. The results of this objective could be useful in mitigating MUHIs by leveraging the high-resolution temperature data available through UAVs, particularly considering ongoing climatic and environmental changes.

In addition to the deleterious impacts of MUHIs, extreme rainfall (ER) causes disasters in cities worldwide, resulting in far-reaching economic, public health, and infrastructural impacts. Characterizing the contribution of urban areas in altering ER is essential for resilience planning for future ER events under climate change. ER disparities in a metroplex due to heterogeneous development using a high spatial-temporal radar rainfall product (Multi-Sensor Precipitation Estimates) were shown in the second objective of this dissertation. Low synoptic days exceeding the 95th percentile of daily rainfall were analyzed from 2000-2016 within equal-sized grids in the DFW area and classified into four clusters based on urbanization. Urbanization and cumulative maximum rainfall were found to be highly correlated (R2=0.92), and the maximum rainfall occurred during the daytime. In contrast to existing studies indicating a downwind concentration of ER, the Urban Core cluster received the most ER overall. After a rapid urbanization period (2000-2008), the highest increase in ER was observed downwind under the dominant wind regime (southwesterly). Under southwesterly winds, the downwind area received approximately 5% higher ER than the upwind area. For low-speed southwesterly winds, this trend was reversed. Under high and low wind speeds for southwesterly winds, storms were tracked hourly, which showed that some storms were initiated by UHI. Peak hourly rainfall occurred in the urban core area 2-4 hours after the maximum UHI value. These results may help devise effective measures for informed urban planning to tackle future ER events.

Furthermore, recent ER events in different global metropolitan areas necessitate investigating the influence of different factors (i.e., climatological and land cover) on rainfall. For the third and last objective of the study, four tree-based machine learning (ML)methods were used to analyze the influence of eight variables on observed rainfall from 2015 to 2021. The study area (DFW) was divided into three urban development clusters (low, medium, and high developed land area change) for local-scale analysis. Random Forest (RF) outperformed other ML models when examined under different scenarios. At the global scale (the entire study area), the RF model achieved R2, root mean squared error, and mean absolute error of 0.93, 2.57 mm, and 0.70 mm on the test data. Temperature remained the most influential variable, followed by wind speed in most scenarios, and the importance of these two variables was increased when considering extreme rainfall (days exceeding 95th percentile of daily rainfall). The upwind located developed land types around the downtown influenced extreme rainfall downwind, reflecting a spatial lag effect of the urban development. Increasing the areas of high and medium-intensity developed land types triggered extreme rainfall, and the opposite was noted for the low and open-space land types. The findings show the application of ML retrospectively in understanding the small-scale impacts of land use and climatological features on rainfall in urban areas. The results can help inform decisions for urban planning and designing, eventually mitigating and adapting to extreme rainfall events in the future.

Overall, the three objectives of the dissertation focus on the influence of urbanization on rainfall and temperature using the latest remote sensing and ML techniques, considering DFW as the study area. While the study has been performed at a metroplex scale, the results could apply to areas with similar climatology and related features around the world.

Keywords

Urbanization, extreme rainfall, urban heat islands, heatwaves, drones, radar, machine learning, urban development

Disciplines

Civil Engineering | Hydraulic Engineering | Other Civil and Environmental Engineering

License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.

Available for download on Wednesday, August 12, 2026

Share

COinS