Establishing Models for Predicting Above Ground Carbon Stock Based on Sentinel 2 Imagery for Evergreen Broadleaf Forests in South Central Coastal Ecoregion, Vietnam

In Vietnam, the models for estimating above ground biomass (AGB) for converting to carbon stocks prediction mostly based on diameter at breast height (DBH), tree height (H), wood density (WD) meanwhile the remote sensing application has considered as suitable method since improving accuracy and reducing cost. With this context, this study was conducted with aim to develop correlation equations among total above ground carbon (TAGC) and indices of Sentinel 2 images to directly predict carbon stock for assessing carbon emission and removal. In this study, remote sensing indices great influencing TAGC were determined by principal component analysis (PCA) and forest inventory factors from 115 sample plot was used to calculate the TAGC. Regression models were established by Ordinary Least Squares and Maximum Likelihood methods and validated by Monte Carlo cross-validation method. The study found out that NDVI, SAVI, NIR and three variable combination (NAVI, ARVI), (SAVI, SIPI), (NIR, EVI) have strongly influenced on TAGC. Total 36 linear and non-linear with weight models basing on above selected variables were established, in which quadratic models used NIR and variable combination (NIR, EVI) with AIC of 756.924, 752.493, R2 value of 0.86, 0.87 and MPSE of 22,04%, 21,63% respectively, were found as optimal models. Therefore, the study these models have recommended for predicting carbon stocks for Evergreen Broadleaf Forests in South Central Coastal Ecoregion, Vietnam.