Abstract:In order to explore and quantify the uncertainty caused by the thermal infrared imaging system for the measurement of walnut canopy temperature, the A310f thermal imager was used to perform a high-frequency and continuous observation of the canopy temperature of three sample trees for 20 days in the walnut plantation. During the period, the canopy temperature was measured in different directions (east, west, south and north) and at different angles (10°, 30°, 45°, 60° and 80°). First of all, the sensitivity analysis of five important parameters showed that the canopy temperature was most affected by leaf emissivity (εleaf), followed by the ambient reflection temperature (Trefl), and was less affected by air temperature (Ta) and air relative humidity (RH). It was not sensitive to distance (D) changes. Then, two-way analysis of variance was performed between the canopy temperature of three sample trees and four directions. The results showed that there was a significant level (P<0.05) between the directions, and the difference between the samples was not significant. Further, through multiple comparison methods, significant differences (P<0.05) were found between the south and the north directions, and the differences between the other directions were not significant. Analysis of the different angles showed that there was no significant temperature difference between the five angles. The external temperature of the canopy was higher than the internal temperature of the canopy that was directly observed by profile gradient method. The temperature frequency histogram reflected the bimodal distribution characteristic of the pixel points, and the peak temperature of the canopy pixel was 25.1℃. Finally, the analysis of variance was performed on the temperature from inside and outside of the canopy. The results showed that there was an extremely significant difference (P<0.01) between them, and the external temperature of the canopy reached the maximum value at 13:00. In summary, the uncertainty analysis of canopy temperature would provide a theoretical basis for reducing measurement error and improving measurement accuracy.
