An Empirical Analysis of Above-Ground Biomass and Carbon Sequestration Using UAV Photogrammetry and Machine Learning Techniques

Abstract

This research aims to analyze above-ground biomass and carbon sequestration using unmanned aerial vehicle (UAV) photogrammetry and machine learning methods, focusing on a case study of the dry dipterocarp forest in the Ban Hin Lat and Hin Lat Phatthana Community Forests. The methodology involved conducting field surveys and data analysis to estimate biomass using allometric equations and UAV photogrammetry data. The estimated biomass from both methods was then analyzed to determine carbon sequestration. Field survey results identified a total of 1,241 trees across 39 species. The analysis using allometric equations found a total above-ground biomass of 454,310.54 kg (454.31 tons), with a carbon sequestration of 213,525.95 kgCO2e (213.52 tCO2e). In contrast, the machine learning analysis using the Deepness technique from UAV data estimated an above-ground biomass of 463,689.13 kg (463.68 tons), with a carbon sequestration of 217,933.89 kgCO2e (217.93 tCO2e). The difference in carbon sequestration estimates between field data and UAV photogrammetry was 4.4 tons, indicating a relatively low error margin of 9.39%. Additionally, the results for the assessment data across different histogram intervals revealed a detection accuracy of tree crowns using UAV photogrammetry at 0.594, with a precision of 0.798, recall of 0.699, and F1 score of 0.745.

Keywords

• Remote sensing
• aerial image processing
• machine learning
• above ground biomass
• carbon sequestrations