UAV thermal remote sensing for land surface temperature mapping in complex urban environments

Urban heat risk assessments increasingly require land surface temperature (LST) and near-surface air temperature at spatial scales that resolve microclimatic drivers such as material heterogeneity, shading, and complex terrain. While satellite thermal products and stationary air temperature observations provide essential regional and temporal context, their spatial resolution and coverage, as well as satellite revisit frequency, limit the quantification of surface thermal variability within urban blocks and campus-scale environments. Unmanned aerial vehicle (UAV) thermal imagery can bridge this scale gap, but quantitative LST retrieval remains sensitive to radiometric calibration, emissivity assumptions, local viewing geometry, geolocation accuracy, and acquisition-time atmospheric conditions.

This contribution develops and demonstrates a reproducible UAV thermal remote sensing workflow that converts raw thermal imagery into georeferenced LST mosaics over complex urban surfaces. Using a DJI Matrice 4T thermal sensor over a university campus in Sheffield, UK, thermal data were collected through multiple field surveys combining UAV flights with ground measurements collected alongside the flights. UAV flights were conducted in late June 2025, with flight planning targeting approximately 80% forward and side overlap. Raw thermal imagery derived from UAV was batch-converted using documented acquisition parameters informed by on-site conditions. Key factors include target distance, relative humidity, emissivity, and reflected apparent temperature,  applied consistently within each survey to support cross-frame comparability.  This research: (1) converts raw thermal imagery to georeferenced thermal outputs using ground-informed acquisition parameters (i.e. distance, humidity, emissivity, and reflected apparent temperature) to stabilise cross-frame temperature consistency; (2) reduces spatial distortions through co-registration with high-resolution basemaps, with a digital terrain model (DTM) used as an additional terrain reference; (3) accounts for surface emissivity variability by integrating land use/land cover and material proxies derived from complementary geospatial datasets, with high-resolution RGB orthomosaics used to derive land cover or material proxies (e.g., vegetation and pavements) that inform thermal processing parameters and support consistent interpretation of microscale thermal patterns.

The workflow delivers thermal remote sensing products at centimetre-level ground sampling distances and is designed to be transferable to other urban sites using standard UAV surveys and widely available geospatial datasets. By foregrounding calibration, emissivity handling, and quality control, this study strengthens the methodological basis for integrating UAV thermal observations into environmental remote sensing in urban settings, enabling more robust cross-scale interpretation of urban thermal patterns and supporting evidence-based decision making.