Examining effective use of data sources and modeling algorithms for improving biomass estimation in a moist tropical forest of the Brazilian Amazon

Abstract

Previous research has explored the potential to integrate lidar and optical data in aboveground biomass (AGB) estimation, but how different data sources, vegetation types, and modeling algorithms influence AGB estimation is poorly understood. This research conducts a comparative analysis of different data sources and modeling approaches in improving AGB estimation. RapidEye-based spectral responses and textures, lidar-derived metrics, and their combination were used to develop AGB estimation models. The results indicated that (1) overall, RapidEye data are not suitable for AGB estimation, but when AGB falls within 50–150 Mg/ha, support vector regression based on stratification of vegetation types provided good AGB estimation; (2) Lidar data provided stable and better estimations than RapidEye data; and stratification of vegetation types cannot improve estimation; (3) The combination of lidar and RapidEye data cannot provide better performance than lidar data alone; (4) AGB ranges affect the selection of the best AGB models, and a combination of different estimation results from the best model for each AGB range can improve AGB estimation; (5) This research implies that an optimal procedure for AGB estimation for a specific study exists, depending on the careful selection of data sources, modeling algorithms, forest types, and AGB ranges.

Keywords

• lidar
• rapideye
• aboveground biomass
• moist tropical forest
• support vector regression
• random forest
• linear regression
• stratification