Temporal Normalization of UAV Thermal Infrared Data From Long-Duration Flights

Uncrewed aerial vehicle (UAV) thermal infrared (TIR) remote sensing is playing an increasingly important role in diverse applications such as agriculture, forestry, hydrology, and ecological monitoring. Moreover, UAV-based remote sensing significantly contributes to the understanding of fundamental remote sensing science issues, such as scale variation and its impacts on multisource data collaboration. To cover extensive areas, UAVs often need to fly multiple strips to ensure complete coverage, leading to significant intervals between the start and end of missions. This can cause notable changes in brightness temperature (BT) due to the different observation times, introducing considerable uncertainty into the final BT mosaic, which, in turn, directly impacts subsequent applications, such as the calculation of land surface temperature (LST) and other temperature-related analyses. This study presents a temporal effect removal of LST (TERL) method to effectively correct for differences due to observation time and thereby enhance the temporal consistency of BT data. The core of TERL involves three key processes: 1) modeling the temporal information of TIR mosaic pixels; 2) deriving the temporal dynamics related to specific surface features by classifying image sequences; and 3) capturing the temporal variation of BT differences and temperature compensation. Validation results indicate that TERL significantly improves both the temporal comparability of pixels and the consistency between UAV temperature data and ground observations. Specifically, the root-mean-square error (RMSE) of the corrected data is 22.50%–77.14% smaller than that of the uncorrected data, with an impressive average reduction of 51.09%. Compared to the digital number probability density function fitting and radiative transfer simulation-based (DRAT) method, which primarily addresses temperature drift, TERL achieves an average RMSE reduction of 29.65%, showcasing its better performance. Moreover, the corrected data better reflect the temperature variation trends of surface features and show strong correlations with ground observation data, with most correlation coefficients exceeding 0.5. Thus, TERL facilitates more accurate comparisons and analyses of UAV TIR data, ultimately enhancing not only the reliability and effectiveness of quantitative remote sensing research with UAVs but also advancing the understanding of fundamental issues like scale variation in remote sensing science.