| dc.description.abstract | Greenhouse gases (GHGs) are major contributors to the earth's warming, causing the well-known greenhouse effect. Among these gases, carbon dioxide (CO2) accounts for 76% of GHG emissions and is generated by various human activities. Trees play a crucial role in mitigating GHGs by absorbing CO2 from the atmosphere and storing it as aboveground carbon (such as stems, branches, and leaves) and belowground carbon (including roots). This process not only aids in ecosystem restoration but also minimizes the impact of GHGs. Agroforestry has emerged as a promising approach to ecosystem restoration and reducing the impact of greenhouse gases through carbon sequestration. However, traditional remote sensing classification techniques often find it difficult to accurately identify agroforestry practices within different land covers, such as grassland and bushland/shrubland. This study proposes the use of individual tree mapping to determine agroforestry practices. Deep learning algorithms, specifically the deepforest algorithm, combined with kernel density point pattern analysis, were employed to determine agroforestry practices. Additionally, the InVEST carbon storage and sequestration model was used to assess the carbon sequestration potential of the agroforestry practices. The results revealed the presence of home gardens (mixed dense), strip plantation (alley cropping), and silvopastoral systems i.e., integrating pasture crops and trees (sparse dense) within the study area, collectively storing 455.61Mg of carbon. The deepforest algorithm effectively delineated agroforestry zones by accurately detecting small trees, although its accuracy was moderately affected by limited training data. This approach enabled the identification of three specific agroforestry practices and the estimation of their carbon sequestration using the InVEST model, surpassing the capabilities of traditional methods. | en_US |
