Allometric models for improving aboveground biomass estimates in West African savanna ecosystems

• Published in: Trees, Forests and People (Online) Forests and People (Online), 2021-06, Vol.4, p.100077, Article 100077
• Main Authors: Ganamé, Moussa, Bayen, Philippe, Ouédraogo, Issaka, Balima, Larba Hubert, Thiombiano, Adjima
• Format: Article
• Language: English
• Subjects: Burkina Faso; Dominant species; Forest carbon; Local allometric model; Ordinary least squares; Savanna; Seemingly unrelated regression
• ISSN: 2666-7193; 2666-7193
• DOI: 10.1016/j.tfp.2021.100077

•Mixed allometric models were developed for site-level dominant woody species.•Both Ordinary Least Square techniques and Seemingly Unrelated Regression methods were tested.•Seemingly Unrelated Regression method better predicted aboveground biomass.•Biomass predictors were specific to each woody species. West African Sudanian savanna ecosystems greatly contribute to local peoples’ livelihoods and climate change mitigation. Yet, the contribution of these ecosystems to carbon storage remains poorly documented due to the lack of accurate biomass predictive tools. Therefore, biomass allometric models developed at both the species-level and site-level may greatly improve carbon stock estimation. In this study, we developed allometric models for estimating aboveground biomass (AGB) at the tree component-, species- and site-levels in Burkina Faso. Five woody species with high socio-economic significance (Anogeissus leiocarpa, Combretum nigricans, Isoberlinia doka, Mitragyna inermis and Pterocarpus erinaceus) were selected at two forest sites based on their dominance in the savanna ecosystem. A total of 150 trees (30 trees per dominant species) spanning a wide range of diameter at breast height (DBH) were destructively sampled. Models to predict tree component biomass at the species-level were independently built with the Ordinary Least Squares (OLS) technique. Allometric models to predict tree total aboveground biomass (TAGB) for each species and all species were developed using both the OLS technique and the Seemingly Unrelated Regression (SUR) method. Biomass estimates were regressed with DBH as a single predictor, DBH and height (H) as interacted variables, and DBH, H and wood density (ρ) as three independent-input variables. The performance of the validated allometric models were compared with the generalized pantropical equation model for African tropical forests (Chave et al., 2014). The findings revealed that biomass predictors varied between the five species. The goodness-of-fit statistics revealed that both the OLS and SUR methods provide accurate biomass allometric equations, with the SUR method being the most accurate. The developed allometric models provide more accurate estimation of AGB than the pantropical model. Therefore, we recommend the use of these local models to improve the quantification of AGB and carbon stocks in Sudanian savanna ecosystems. Furthermore, the established species-specific and mixed-species allometric equations may constitute