Carbon emissions assesment of post selective logging impacts using unmanned aerial vehicle and high-resolution satellite imagery

Abstract

A Selective Management System (SMS) is an effective management strategy in logging, especially for quantifying carbon emissions. Selective logging may cause a significant loss of carbon to the atmosphere if it is not properly monitored. This study was intended to investigate the potential of integrating remote sensing approaches and ground measurement to assess carbon emissions from selective logging in Ulu Jelai Reserve Forest, Pahang, Malaysia. Data from various platforms, such as unmanned aerial vehicle and satellite remote sensing, were used to extract selective logging impact indicators. This study used two main approaches to estimate carbon emissions: (i) ground measurement integrated with spatial data and (ii) remote sensing techniques using machine-learning approaches. A carbon calculator from Winrock’s International was used to quantify emission factors and estimate the total carbon emissions. Meanwhile, a linear regression model and a series of machine learning models consisting of Support Vector Machine (SVM), Random Forest, and K-Nearest Neighbour were tested to produce the best post-selective logging impact carbon emission model. By applying the emission factor of 1.305 Mg C m-3 , the estimated total carbon emission from the research area obtained was 7050.54 Mg C, with an average of 84.95 Mg C ha-1 . Remote sensing data were used to extract the attributes of forest logging, such as stumps and their structures, logging infrastructure, logging damage, and the density of forest cover before and after logging practises. Stepwise regressions were performed to select the most highly correlated variables among the thirteen potential variables. The findings indicated that the best dependent variables were logging gaps, road area, and road length. Among the three machine learning models tested, SVM provided the best accuracy, with a root mean squared error (RMSE) of 21.10%, 0.23% bias, and an adjusted R-squared (R2 ) of 0.80. The linear model also showed a promising result with an RMSE of 22.14%, a bias of 0.72%, and an adjusted R2 of 0.75. The emission factor generated from this study can be used to assess the total carbon emission over a larger area, and the developed carbon emission model can be applied to analyse carbon emissions from selective logging in other tropical areas with similar logging practises. It is also expected that this study will be useful in assisting related authorities such as the forestry department, researchers, and many more in optimising logging practises to sustain forest carbon sequestration and mitigate climate change.