Scaling mangrove aboveground biomass from site‐level to continental‐scale

AIM: We developed a set of statistical models to improve spatial estimates of mangrove aboveground biomass (AGB) based on the environmental signature hypothesis (ESH). We hypothesized that higher tidal amplitudes, river discharge, temperature, direct rainfall and decreased potential evapotranspiration explain observed high mangrove AGB. LOCATION: Neotropics and a small portion of the Nearctic region. METHODS: A universal forest model based on site‐level forest structure statistics was validated to spatially interpolate estimates of mangrove biomass at different locations. Linear models were then used to predict mangrove AGB across the Neotropics. RESULTS: The universal forest site‐level model was effective in estimating mangrove AGB using pre‐existing mangrove forest structure inventories to validate the model. We confirmed our hypothesis that at continental scales higher tidal amplitudes contributed to high forest biomass associated with high temperature and rainfall, and low potential evapotranspiration. Our model explained 20% of the spatial variability in mangrove AGB, with values ranging from 16.6 to 627.0 t ha⁻¹ (mean, 88.7 t ha⁻¹). Our findings show that mangrove AGB has been overestimated by 25–50% in the Neotropics, underscoring a commensurate bias in current published global estimates using site‐level information. MAIN CONCLUSIONS: Our analysis show how the ESH significantly explains spatial variability in mangrove AGB at hemispheric scales. This finding is critical to improve and explain site‐level estimates of mangrove AGB that are currently used to determine the relative contribution of mangrove wetlands to global carbon budgets. Due to the lack of a conceptual framework explicitly linking environmental drivers and mangrove AGB values during model validation, previous works have significantly overestimated mangrove AGB; our novel approach improved these assessments. In addition, our framework can potentially be applied to other forest‐dominated ecosystems by allowing the retrieval of extensive databases at local levels to generate more robust statistical predictive models to estimate continental‐scale biomass values.