Estimating more reliable measures of forest canopy temperatures using thermal imaging

Abstract:

Leaf temperature is critical to plant function, and it can be used to examine forest responses to droughts, heat waves, and storm events. The recent development of thermal infrared (TIR) imaging techniques has offered indirect measurement of forest canopy skin temperature, and it allows for extensive temporal and spatial sampling compared to direct thermocouple-based measurements. However, the accuracy of TIR indirect canopy temperature is not well understood, as few studies have evaluated how TIR-derived temperatures compare to other approaches. The objectives of this study are: (1) to monitor canopy temperatures of a coniferous forest canopy using a TIR camera and in situ sensors; (2) to evaluate the reliability of TIR canopy temperatures by comparing against leaf temperatures measured by thermocouples; (3) to develop and examine methods for improving TIR measures based on corrections of camera’s default parameters (“Recalculation”) and records of sensitivity by parameter changes (“Data-training”). This study showed the canopy temperatures varied from -5 and 30°C, and the patterns of changes between the TIR and thermocouple measures corresponded well. Overall, TIR canopy temperatures were underestimated against the direct thermocouple measurements with mean absolute error (MAE) of 0.83-1.38°C and root mean square error (RMSE) of 1.11-1.53°C for the study period. The modified TIR temperatures from the “Recalculation” method exhibited MAE of 0.56-0.95°C and RMSE of 0.83-1.15°C, and those by the “Data-training” method resulted in MAE of 0.32-0.50°C and RMSE of 0.53-0.83°C. Our results demonstrate that the TIR technique includes small errors for canopy temperature measurements; however, the range of errors is smaller when correction methods are applied.