首页 >  , Vol. , Issue () : -

摘要

全文摘要次数: 81 全文下载次数: 157
引用本文:

DOI:

10.11834/jrs.20222120

收稿日期:

2022-03-18

修改日期:

2022-04-09

PDF Free   EndNote   BibTeX
联合ICESat-2和GEDI星载激光雷达数据的浙江省森林地上生物量估算
孟鸽, 赵旦, 许聪, 陈俊华, 李秀纹, 郑朝菊, 曾源
中国科学院空天信息创新研究院 遥感科学国家重点实验室
摘要:

本研究重点对比分析ICESat-2和GEDI两种星载激光雷达数据用于森林地上生物量估算的适用性,首先在古田山典型研究区采用线性逐步回归方法分别构建基于ICESat-2的条带和基于GEDI的光斑尺度森林地上生物量估算模型,再结合MODIS光学遥感数据和ASTER GDEM地形数据,采用随机森林法在浙江省分森林类型构建逐像元尺度的森林地上生物量外推模型,根据40个森林地上生物量地面调查样点的精度验证探讨大区域尺度森林地上生物量的最优估算方法,实现了2020年浙江省森林地上生物量成图。研究结果表明:(1)在古田山典型研究区基于ICESat-2的条带尺度森林地上生物量估算模型精度(R2=0.7057,RMSE=0.3571 ln(t/ha))高于基于GEDI的光斑尺度模型(R2=0.5186,RMSE=0.2805 ln(t/ha));(2)外推到浙江省的森林地上生物量估算结果表明基于ICESat-2的精度(R2=0.59,RMSE=31.2525 t/ha)明显优于GEDI(R2=0.4113,RMSE=39.2652 t/ha),且研究区的高程对两种数据的估算差异影响最大,仅采用高程在600m以下的GEDI光斑数据来构建外推模型,发现估算精度有显著提高(R2=0.5387,RMSE=25.4017 t/ha);(3)将ICESat-2条带与高程在600m以下的GEDI光斑联合起来构建外推模型,是利用两种星载激光雷达数据进行浙江省森林地上生物量估算的最优方法(R2=0.678,RMSE=27.3592 t/ha)。该研究融合地面、机载与星载遥感数据,以浙江省为例开展森林地上生物量的区域尺度估算研究,可为森林碳储量的动态监测与固碳能力评估提供方法借鉴与应用示范。

Forest aboveground biomass estimation in Zhejiang Province combining ICESat-2 and GEDI spaceborne LiDAR data
Abstract:

Objective : Forest aboveground biomass (AGB) plays an important role in the study of carbon cycle and global change. Spaceborne LiDAR can provide information about forest vertical structures that is advantageous in AGB estimation, among which ICESat-2 and GEDI are the latest available spaceborne data. In this study, we investigated the applicability of ICESat-2 and GEDI for forest AGB estimation at regional scale, and analyzed the effect of data fusion of ICESat-2 and GEDI to find an optimal method to map the spatial distribution of forest AGB accurately in Zhejiang Province. Method: First, we built footprint-level forest AGB estimation models by stepwise regression in the typical study area of Gutian Mountain based on ICEsat-2 and GEDI spaceborne LiDAR data, respectively. Then, combined with MODIS data and ASTER GDEM terrain information, forest AGB estimation models with spatial continuity at 250m pixel scale for different forest types were constructed by Random Forest algorithm throughout Zhejiang Province. Estimation results were validated using 40 forest AGB field plots. Finally, by comparing validation results of AGB estimation based on ICESat-2 or GEDI solely and the combination of the two spaceborne LiDAR data, the optimal method of forest AGB scaling was selected and the spatial distribution of forest AGB of the year 2020 was mapped in Zhejiang Province. Result: The accuracy of segment-level forest AGB estimation based on ICESat-2 (R2=0.7057, RMSE=0.3571 ln(t/ha)) outmatches footprint-level forest AGB estimation based on GEDI (R2=0.5186, RMSE=0.2805 ln(t/ha)) in the typical study area of Gutian Mountain. Validation accuracy of forest AGB estimation result based on ICEsat-2 (R2=0.59, RMSE=31.2525 t/ha) is superior to GEDI (R2=0.4113, RMSE=39.2652 t/ha) in Zhejiang Province. The difference of forest AGB estimation performance between ICESat-2 and GEDI is mainly related to elevation, validation accuracy based on GEDI is higher when filtering footprints that are acquired in high elevation areas with an elevation threshold of 600m (R2=0.5387, RMSE=25.4017 t/ha). Combining ICESat-2 and GEDI data (elevation ≤ 600m) to build scaling model is the optimal method to estimate forest AGB in Zhejiang Province (R2=0.678, RMSE=27.3592 t/ha). Conclusion: We have obtained a reliable estimation of forest AGB in Zhejiang Province based on ICESat-2 and GEDI data, which is a significant practice of regional scale forest AGB estimation. Our study can provide an effective method for forest carbon dynamic and sequestration potential monitoring using spaceborne LiDAR data.

本文暂时没有被引用!

欢迎关注学报微信

遥感学报交流群 分享按钮