Improved estimation of forest stand volume by the integration of GEDI LiDAR data and multi-sensor imagery in the Changbai Mountains Mixed forests Ecoregion (CMMFE), northeast China

Abstract

Forest stand volume is a vital indicator of productivity and carbon storage. Conventionally, stand volume is estimated from field samples, Synthetic Aperture Radar (SAR), and optical imagery, which suffer saturation problems. Although Light Detection and Ranging (LiDAR) technique degrades the signal saturation, its large-scale application is hindered by spatial continuity. To address this issue, recently released Global Ecosystem Dynamics Investigation (GEDI) LiDAR data, Sentinel-1 SAR, Sentinel-2 Multispectral Instrument (MSI), and Advanced Land Observing Satellite (ALOS) digital surface model (DSM) imagery were integrated for volume modeling and estimation under a point-line-polygon framework. The footprint-level LiDAR variables, as a linear bridge, were adopted to link field plots to the full-cover multi-sensor imagery. Results showed that volume of the Changbai Mountains Mixed Forests Ecoregion (CMMFE) displayed variations along the elevation gradient, ranging from 47.56 to 277.30 m3/ha with a mean value of 151.39 m3/ha. Additionally, the accuracy comparison based on independent validation samples indicated that integrating GEDI LiDAR data under a point-line-polygon framework performed better than the traditional point-polygon approach, which directly linked field samples to multi-sensor imagery. The corresponding estimated error declined from 22.08% to 15.21%. The canopy cover and tree height from LiDAR, elevation from L band InSAR, and spectral indices of MSI red-edge bands were key for stand volume mapping in heterogeneous temperate forests. This comparison also showed that the integration of LiDAR by a point-line-polygon framework adopted 2/3 of the modeling points but acquired more accurate estimation than a traditional approach only based on multi-sensor imagery, which implied less field sampling work was needed for similar research. Consequently, as a pioneering exploration of GEDI LiDAR data combined with multi-sensor imagery under the point-line-polygon framework, this study provides an efficient methodology for the volume estimation of heterogeneous forests.