Estimating mangrove aboveground biomass in the Colombian Pacific coast: A multisensor and machine learning approach

Abstract

The Colombian Pacific Coast is renowned for its exceptional biodiversity and hosts vital mangrove ecosystems that benefit local communities and contribute to climate change mitigation. Therefore, estimating mangrove aboveground biomass (AGB) in this region is crucial for planning and managing these coastal forest covers, ensuring the long-term sustainability of the essential environmental services provided by the Colombian Pacific Coast (CPC). This study employed a spatial estimation approach to assess mangrove AGB, evaluating various parametric and non-parametric models using a multisensor combination and machine learning on the Google Earth Engine (GEE) platform within the CPC. Synthetic aperture radar (SAR) satellite imagery (ALOS-2/PALSAR-2, SRTM, NASADEM, and ALOSDSM) and optical data (Landsat 8) were utilized to quantify mangrove AGB in 2022 across the four departments of the CPC. The Random Forest model exhibited superior predictive performance compared to the other models evaluated, achieving values of R2 = 0.783, RMSE = 38.239 [Mg/ha], MAE = 27.409 [Mg/ha], and BIAS = 0.164. Our findings reveal that the mangrove AGB map for the CPC exhibits a mean ± standard deviation of 181.236 ± 28.939 [Mg/ha] across eight classes, ranging from 88.622 [Mg/ha] to 378.21 [Mg/ha]. This research provides valuable information to inform and strengthen various management strategies and decision-making processes for the mangrove forests of the CPC.