Improving the accuracy of forest aboveground biomass using Landsat 8 OLI images by quantile regression neural network for Pinus densata forests in southwestern China

• DOI: 10.3389/ffgc.2023.1162291
• Publication Date: 2023-04-18T04:21:26Z
• Authors: Xiaoli Zhang; Lu Li; Yanfeng Liu; Yong Wu; Jing Tang; Weiheng Xu; Leiguang Wang; Guanglong Ou
• Keywords: Quantile; Quantile regression

It is a challenge to reduce the uncertainties of underestimation and overestimation forest aboveground biomass (AGB) which common in optical remote sensing imagery. In this study, four models, namely, linear stepwise regression (LSR), artificial neural network (ANN), quantile (QR), (QRNN) were used estimate Pinus densata AGB data by collecting 146 sample plots combined with Landsat 8-Operational Land Imager (OLI) images Shangri-La City, Yunnan Province, southwestern China. The results showed that compared LSR, R 2 mean square error (RMSE) ANN, QR, QRNN had improved significantly. particular, was able significantly improve situation when we estimated biomass, highest (0.971) lowest RMSE (9.791 Mg/ha) for whole segment. Meanwhile, through model validation, found (0.761) (6.486 on segment <40 Mg/ha. Furthermore, it (0.904) (9.059 >160 Mg/ha, offered great potential improving estimation accuracy AGB. conclusion, QRNN, combining advantages QR provides reducing precision influence caused using data.