MAPPING THE AIR MOISTURE CHANGE IN UNDER CANOPY TREES USING A HEMISPHERICAL AND AERIAL PHOTOGRAPH BASED ON MACHINE LEARNING APPROACHES

Abstract

The essential roles of trees in controlling the local climatic variation, such as air moisture, are still interesting to observe. Therefore, this study must deliver knowledge of the benefits of growing trees and enhance people's awareness of climate change adaptation. Here, the analysis requires several data fields such as hemispherical photography, an aerial photograph of a UAV, and air temperature collected using a wet and dry bulb thermometer, which has converted to air moisture. All these are considered to understand the air moisture change under the trees' canopy during a day observation. The hemispherical photography and aerial photograph of a UAV are processed to measure the tree's canopy size and then used together with interpolated air moisture to map the variation in air moisture distribution in under-canopy trees using random forest (RF) and Artificial Neural Network (ANN). The result shows that hemispherical photography describes the ability to control the air moisture change. As its size increases, the air moisture level tends to be higher. It was maintained at more than 70% compared to the area with lower canopy cover. This characteristic is similar to the pattern shown by the RF and ANN. However, the SVM has better results as it can separate air humidity in vegetated and non-vegetated areas.