Limiting External Absorptivity of UAV-Based Uncooled Thermal Infrared Sensors Increases Water Temperature Measurement Accuracy

Abstract

Thermal mapping of surface waters and the land surface via UAVs offers exciting opportunities in many scientific disciplines; however, unresolved issues persist related to accuracy and drift of uncooled microbolometric thermal infrared (TIR) sensors. Curiously, most commercially available UAV-based TIR sensors are black, which will theoretically facilitate heating of the uncooled TIR sensor via absorbed solar radiation. Accordingly, we tested the hypothesis that modifying the surface absorptivity of uncooled TIR sensors can reduce thermal drift by limiting absorptance and associated microbolometer heating. We used two identical uncooled TIR sensors (DJI Zenmuse XT2) but retrofitted one with polished aluminum foil to alter the surface absorptivity and compared the temperature measurements from each sensor to the accurate measurements from instream temperature loggers. In addition, because TIR sensors are passive and measure longwave infrared radiation emitted from the environment, we tested the hypotheses that overcast conditions would reduce solar irradiance, and therefore induce thermal drift, and that increases in air temperature would induce thermal drift. The former is in contrast with the conceptual model of others who have proposed that flying in overcast conditions would increase sensor accuracy. We found the foil-shielded sensor yielded temperatures that were on average 2.2 °C more accurate than those of the matte black sensor (p < 0.0001). Further, we found positive correlations between light intensity (a proxy for incoming irradiance) and increased sensor accuracy for both sensors. Interestingly, light intensity explained 73% of the accuracy variability for the black sensor, but only 40% of the variability in accuracy deviations for the foil-shielded sensor. Unsurprisingly, an increase in air temperature led to a decrease in accuracy for both sensors, where air temperature explained 14% of the variability in accuracy for the black sensor and 31% of the accuracy variability for the foil-shielded sensor. We propose that the discrepancy between the amount of variability explained by light intensity and air temperature is due to changes in the heat energy budget arising from changes in the surface absorptivity. Additionally, we suggest fine-scale changes in river-bed reflectance led to errors in UAV thermal measurements. We conclude with a suite of guidelines for increasing the accuracy of uncooled UAV-based thermal mapping.

Keywords

• absorptivity
• emissivity
• river temperature
• thermal infrared
• thermal refuges