Quantifying forest above ground biomass: a critical review of methods, challenges, and opportunities

This is the authors' abstract. We don't add key points for this paper.

• This review compares traditional and remote sensing methods for biomass estimation. • Traditional methods are more accurate but arduous, costly on a large-scale. • Remote sensing techniques are suitable for large-scale forest analysis. • Artificial Neural Networks performed best with coefficient of determination of 0.8. • Further exploration of artificial and convolutional neural networks is recommended. Forests are important in sequestering carbon as a climate change mitigation measure. The amount of carbon stored in forest above-ground biomass (AGB) can be measured using Traditional and Remote sensing (RS) approaches. However, these approaches may give different results of the sequestered carbon depending on the forest density, location, and user. The purpose of this review was to assess the strengths, performance, challenges, and feasibility of both traditional and RS techniques in the measurement of forest AGB to recommend the most reliable approach. In this review, over 60 articles from peer-reviewed journals related to forest biomass carbon sequestration were synthesized. Traditional methods that involve direct cutting of the tree stem were found to accurately measure AGB since 95 % of the tree stem diameter biomass is measured with minimal error. However, they are expensive, tedious, and time-consuming and do not apply to extensive analysis. Conversely, RS techniques include employing models that are trained to give reliable results on extensive analysis for biomass estimation. For instance, upon comparing coefficient of determination (R²) for different models, SVM achieved 0.61, linear regression (0.60), RF (0.59), and ANN (0.80), and all models met the verification criteria. ANN model performed best on average with R² of 0.80 using training data of biomass samples. Therefore, further studies concerning integrating ANN and CNN in AGB estimation of remote forests should be conducted to obtain reliable results.