Increasing PM2.5 (particulate matter smaller than 2.5 micrometers) poses a significant health risk to people. Understanding variables critical to PM2.5 spatial and temporal variation is a first step towards protecting vulnerable populations from exposure. Previous studies investigate variables responsible for PM2.5 variation but have a limited temporal span. Moreover, although land-use classes are often taken into account, the vertical environment's influence (e.g., buildings, trees) on PM2.5 concentrations is often ignored and on-road circle buffers are used. To understand variables most critical to PM2.5 concentration variation, an air pollution sensor and GPS unit were affixed to a bicycle to sample for variables over three seasons (spring, summer, fall). Samples were taken on a route during the weekdays at four targeted hours (7AM, 11AM, 3PM, and 7PM) and joined with meteorological data. 3D morphology was assessed using LiDAR data and novel wind-based buffers. Wind speed only, wind direction only, and wind speed and direction buffers were computed and compared for their performance at capturing micro-scale urban morphological variables. Zonal statistics were used to compute morphological indicators under different wind assumptions in seasonal ordinary least squares regression models. A comprehensive wind and buffer performance analysis compares statistical significance for spatial and temporal variation of PM2.5. This study identifies the best wind parameters to use for wind-based buffer generation of urban morphology, which is expected to have implications for buffer design in future studies. Additionally, significant exposure hotspots for UNT students to PM2.5 pollution are identified.