Spatiotemporal dynamics of aboveground biomass in a managed forest, central Mexico

ABSTRACT Background: Quantifying aboveground biomass (AGB) is crucial for studying the carbon cycle and estimating mitigation potential of climate change. Combining field inventory data and remote sensing such as Landsat imagery, is a common approach for landscape-Level AGB analysis. However, uncertainties in biomass estimations persist, highlighting the need for improved statistical methods. The objectives of this study were (i) model the AGB of temperate forests managed for timber production using Landsat 8 data and three regression algorithms (linear regression, generalized additive models [GAM], and random forests), and (ii) quantify interannual AGB variations (2013-2022) across a forest landscape. Predictor variables included spectral bands, vegetation indices (VI), textural metrics, and stand age. Results: The RF algorithm showed the best performance with accurate estimates, explaining 76% of the AGB variability. It also showed an RMSE of 32.93 Mg ha-1 when stand age was included as a predictor variable. The AGB showed a spatial variation from 9 to 289 Mg ha-1, and an inventory of 113,408.81 Mg (±11,663.13 Mg) in a landscape of 823.6 ha, ranging from 101,904.70 Mg in 2013 to 127,915.60 Mg in 2022. The 10-12-year-old stands showed the highest increment of biomass after a decade, increasing from 71.06 Mg ha-1 (±19.81) in 2013 to 153.37 Mg ha-1 (±14.13) in 2022. Conclusion: The study evaluated a practical methodology to estimate the spatiotemporal variation of AGB in managed temperate forests. This approach can be implemented to support the evaluation of the potential contribution of managed forests to climate change mitigation.