High-resolution mangrove mapping in Guatemala using PlanetScope imagery and a SWIR-free moisture index

Abstract

Mangrove forests are ecologically and socioeconomically important ecosystems that require accurate and timely monitoring. This study presents a high-resolution mangrove mapping approach along the Pacific coast of Guatemala using PlanetScope imagery integrated with a machine-learning framework. Sentinel-2 and Landsat data were used to derive the Modular Mangrove Recognition Index (MMRI), which depends on shortwave infrared (SWIR) bands. Given the absence of SWIR in PlanetScope, four alternative multispectral indices (MWDI1–MWDI4) were developed using blue, green, red, and near-infrared bands. Their performance was evaluated against the Modified Normalized Difference Water Index (MNDWI) using Pearson correlation, and temporal stability was assessed under seasonal variability. A Random Forest classifier trained with 40 field-validated plots (181,278 pixels) achieved high classification accuracy (93.9–99.7%), with Kappa values between 0.90 and 0.99 and F1 scores above 0.94. MWDI2 showed the best agreement with MNDWI, demonstrating its suitability as a SWIR-independent alternative for mangrove mapping.