Modelling and Estimation of Above Ground Biomass and Carbon Stock of
Pinus Roxburghii Dominated Forest Using Sentinel-2 Imagery in Shreenagar
Hill, Nepal
Abstract
Aim: To model and estimate the total above ground biomass (AGB) of
forest with the best model out of five different regression models
Location: Shreenagar Hill Forest, Tansen Municipality, Nepal Time
Period: During the month of July, 2023 Major Taxa Studies: Pinus
roxburghii Methods: Sentinel-2 satellite imagery and field-measured AGB
at plot level were used. Field data were collected from a total of 26
sample. Randomly chosen 18 sample plots (SPs) (70%) were used to
generate the model and remaining 8 SPs (30%) for validation of
developed model. Using various bands with 10m spatial resolution, eleven
VIs were calculated & correlated with field measured AGB at plot level.
Results & Main Conclusions: Evaluating the fit statistics, quadratic
regression model using NDVI with correlation coefficient (R) 0.92,
coefficient of determination (R^2) 0.86, AIC (161.13) & BIC (164.69)
was found as the best model. Predicted value of AGB from best model and
observed value of AGB from field were used for model validation. Root
mean square error (RMSE), R & R^2 were found as 13.3594 t.ha-1per
plot, 0.9597 and 0.9211 respectively during the model validation.
Therefore, the quadratic regression model with NDVI as best fit model
was used to estimate the total AGB and carbon stock (CS) of study area.
The average value of AGB & CS (including no vegetation area) for total
study area were found 192.403 & 90.429 t.ha-1 respectively. The value
of AGB & CS range from 0 to 233.451 & 0 to 109.722 t.ha-1 per pixel
respectively. The benefits, possibilities, and effectiveness of
combining Sentinel-2 VIs with field data to forecast biomass are
demonstrated by this work. To reduce the estimation error & make wider
application of research, very large sample size can be chosen by future
researchers.