AI-powered measurement verification and reporting system for agroforestry trees to estimate carbon sequestration potential

Abstract

Nature-based climate solutions, such as agroforestry, offer potential for carbon sequestration while providing co-benefits. However, the lack of scalable and low-cost measurement, reporting, and verification (MRV) systems limits smallholder participation in carbon markets. This study presents the DiameterAlgorithm, a non-contact method for tree diameter estimation using semantic segmentation and two-dimensional photogrammetry. The fine-tuned model achieved a mean intersection over union (mIoU) of 0.937. The algorithm was tested on image datasets from managed trees settings in Kenya (n = 142) and Pennsylvania, USA (n = 40), with regression analysis showing high accuracy (R² = 0.97, RMSE = 2.20–2.23 cm). Bias analysis showed slight overestimation for small to medium trees (5–35 cm DBH) and underestimation for larger trees (>36 cm DBH), with an overall mean bias of +0.68 cm. Coupled with allometric equations, the DiameterAlgorithm enables scalable, site-level biomass estimation for carbon markets.

Code and data availability

The paper publicly releases its tree image dataset (calibration/evaluation images from Kenya and Pennsylvania) on ScholarSphere and the containerized diameter estimation tool on Docker Hub, both explicitly stated in the data availability statement.