Above- and belowground tree biomass models for three mangrove species in Tanzania: a nonlinear mixed effects modelling approach

Abstract

Key message: Tested on data from Tanzania, both existing species-specific and common biomass models developed elsewhere revealed statistically significant large prediction errors. Species-specific and common above- and belowground biomass models for three mangrove species were therefore developed. The species-specific models fitted better to data than the common models. The former models are recommended for accurate estimation of biomass stored in mangrove forests of Tanzania. Context: Mangroves are essential for climate change mitigation through carbon storage and sequestration. Biomass models are important tools for quantifying biomass and carbon stock. While numerous aboveground biomass models exist, very few studies have focused on belowground biomass, and among these, mangroves of Africa are hardly or not represented. Aims: The aims of the study were to develop above- and belowground biomass models and to evaluate the predictive accuracy of existing aboveground biomass models developed for mangroves in other regions and neighboring countries when applied on data from Tanzania. Methods: Data was collected through destructive sampling of 120 trees (aboveground biomass), among these 30 trees were sampled for belowground biomass. The data originated from four sites along the Tanzanian coastline covering three dominant species: Avicennia marina (Forssk.) Vierh, Sonneratia alba J. Smith, and Rhizophora mucronata Lam. The biomass models were developed through mixed modelling leading to fixed effects/common models and random effects/species-specific models. Results: Both the above- and belowground biomass models improved when random effects (species) were considered. Inclusion of total tree height as predictor variable, in addition to diameter at breast height alone, further improved the model predictive accuracy. The tests of existing models from other regions on our data generally showed large and significant prediction errors for aboveground tree biomass. Conclusion: Inclusion of random effects resulted into improved goodness of fit for both above- and belowground biomass models. Species-specific models therefore are recommended for accurate biomass estimation of mangrove forests in Tanzania for both management and ecological applications. For belowground biomass (S. alba) however, the fixed effects/common model is recommended.