<p>Every year, temperature fluctuation with improper management result in extensive damage to apple tree farms in West Azerbaijan Province, as a major hub for apple production in Iran. Conventional temperature measurement methods (e.g., infrared thermometer) are known to be too costly to the farmers. Accordingly, in the present research, a total of 12 satellite images acquired by Landsat-8 during the 2023 crop year were subjected to apple tree vegetation temperature estimation by means of mono-window (MW), split-window (SW), single-channel (SC), liner regression (LR), and decision tree (DT) methods, and the results were compared to field-measured data obtained by a TFA 1134 IR thermometer. The outputs showed that the MW, SW, SC, LR, and DT techniques returned average vegetation temperatures that were 3.3°C different from one another. The air and vegetation temperatures were found to be correlated at an <InlineEquation ID="IEq1"> <EquationSource Format="TEX">${R}^{2}$</EquationSource> <EquationSource Format="MATHML"><math> <msup> <mrow> <mi>R</mi> </mrow> <mn>2</mn> </msup> </math></EquationSource> </InlineEquation> of 0.77, while a very high <InlineEquation ID="IEq2"> <EquationSource Format="TEX">${R}^{2}$</EquationSource> <EquationSource Format="MATHML"><math> <msup> <mrow> <mi>R</mi> </mrow> <mn>2</mn> </msup> </math></EquationSource> </InlineEquation> value of 0.97 was achieved when correlating the vegetation temperature to the land surface temperature (LST), indicating that the LST is a better predictor of the vegetation temperature across the apple farms in the study area, as proved by RMSEs below 1°C. Results further showed that, among the considered remote sensing methods, the SW algorithm is the best estimator of the LST, producing an <InlineEquation ID="IEq3"> <EquationSource Format="TEX">${R}^{2}$</EquationSource> <EquationSource Format="MATHML"><math> <msup> <mrow> <mi>R</mi> </mrow> <mn>2</mn> </msup> </math></EquationSource> </InlineEquation> of 0.9799 and an RMSE of 0.12 °C.</p>

Comparison of Different Remote Sensing and Machine Learning Methods for Estimating Apple Tree Vegetation Temperature

期刊论文
发表日期：

2025年7月22日

查看全文

Mehdi Asadi

摘要

Every year, temperature fluctuation with improper management result in extensive damage to apple tree farms in West Azerbaijan Province, as a major hub for apple production in Iran. Conventional temperature measurement methods (e.g., infrared thermometer) are known to be too costly to the farmers. Accordingly, in the present research, a total of 12 satellite images acquired by Landsat-8 during the 2023 crop year were subjected to apple tree vegetation temperature estimation by means of mono-window (MW), split-window (SW), single-channel (SC), liner regression (LR), and decision tree (DT) methods, and the results were compared to field-measured data obtained by a TFA 1134 IR thermometer. The outputs showed that the MW, SW, SC, LR, and DT techniques returned average vegetation temperatures that were 3.3°C different from one another. The air and vegetation temperatures were found to be correlated at an ${R}^{2}$ $R^{2}$ of 0.77, while a very high ${R}^{2}$ $R^{2}$ value of 0.97 was achieved when correlating the vegetation temperature to the land surface temperature (LST), indicating that the LST is a better predictor of the vegetation temperature across the apple farms in the study area, as proved by RMSEs below 1°C. Results further showed that, among the considered remote sensing methods, the SW algorithm is the best estimator of the LST, producing an ${R}^{2}$ $R^{2}$ of 0.9799 and an RMSE of 0.12 °C.