Large fractions of urban populations around the globe are under an increased threat of extreme heat events. Anthropogenic climate change, coupled with rapid urbanization, have exacerbated these threats. Most of the current urban heat studies have relied upon conventional data sources such as satellites and weather stations to map and analyze the urban heat islands (UHIs). However, these data sources lack the spatial and temporal resolutions required to accurately capture the temperature variations in space and time. Towards overcoming these limitations, this thesis explores the challenges of harnessing modern data collection paradigms, namely crowdsourcing (through human-borne sensors) and drive-by sensing (through vehicle-borne sensors) for UHI analysis. This thesis proposes a three-tier framework called Smart Community-Centric Urban Thermal Sensing (SCOUTS) for efficiently gathering temperature data through crowdsourcing and drive-by sensing, integrating them with data from satellite and weather stations and performing innovative analysis to map and study UHIs. While crowdsourcing and drive-by sensing are inexpensive data collection strategies, harnessing them in an efficient manner for UHI analysis poses several research challenges. This thesis addresses two major challenges in crowdsourcing and drive-by sensing for UHI analysis, respectively. The first is to detect human-borne temperature sensors that are placed anomalously and hence fail to accurately represent the actual outdoor environment. The proposed scheme for detection of anomalously placed sensors is based on our novel feature selection and classification design.The second major challenge that we address is to select public transportation vehicles (city buses) for sensor deployment so as to maximize the spatio-temporal coverage value of the data collected through the drive-by sensing paradigm for UHI analysis. In this regard, we make two unique research contributions: formulating the bus selection problem as an optimization problem and introducing our cost-aware approaches to enhance the spatiotemporal coverage. This thesis reports a series of experiments demonstrating the benefits and limitations of our approaches for detecting anomalously placed sensors in thermal crowdsensing and for sensor deployment in drive-by sensing.