Atlantic forest woody carbon stock estimation for different successional stages using Sentinel-2 data

• DOI: 10.1016/j.ecolind.2023.109870
• Publication Date: 2023-01-09T22:55:58Z
• Authors: Otávio Miranda Verly; Rodrigo Vieira Leite; Ivaldo da Silva T...; Samuel José Silva...; Hélio Garcia Leite; José Marinaldo Gl...; Maria Paula Miran...; Valéria de Fatima...; Carlos Moreira Mi...; Angelica Plata-Rueda; Bárbara Monteiro ...; José Cola Zanuncio; Laércio Antônio G...
• Keywords: Semideciduous seasonal forest; Forest succession; Artificial neural networks; Ecology; Passive remote sensing; 0401 agriculture, forestry, and fisheries; Non-parametric modeling; 04 agricultural and veterinary sciences; Semideciduous Seasonal Forest; Artificial Neural Networks; QH540-549.5

The Atlantic Forest is one of the most threatened biodiversity hotspots and environmental impacts has made its landscape fragmented heterogeneous. heterogeneity fragments a challenge for characterization quantification forest resources, such as stock biomass carbon. Methodologies based on remote sensing have been used, to improve these estimates without compromising execution costs. objective was estimate, with high spatial resolution passive sensing, aboveground carbon in different successional stages Forest. Forests were classified into initial, intermediate, advanced stages. In each stratum, 10 plots (20x50 m) established, calculated by adjusted Schumacher Hall model. reflectances blue, green, red, near-infrared bands vegetation indices (VIs) obtained dry rainy seasons, from MSI/Sentinel-2 images, m. Artificial Neural Networks (ANN), combinations variables, trained validated simulated reflectance values. Carbon estimated ANN best performance training validation. average strata 24.99, 35.79 82.28 Mg ha−1, respectively, general 47.68 ha−1. better season. addition VIs did not performance. spectral data consistent adequate validate selected ANN. total stock, modeled 41,962.15 Mg, ranging 6.68 108.29 an 48.70 stratum more than three times that observed initial they efficiently using high-resolution multispectral data, season, inputs.