联合GF-6和Sentinel-2红边波段的森林地上生物量反演

光谱反射率能反映地物差异，是森林地上生物量（Aboveground Biomass，AGB）遥感反演的理论基础。红边波段处于近红外与红光波段交界处快速变化的区域，能对植被冠层结构和叶绿素含量的微小变化做出快速反应，对植被生长状况较敏感。研究以GF-6和Sentinel-2多光谱影像作为数据源，结合野外调查AGB数据，构建落叶松和樟子松AGB线性和非线性估测模型，通过比较模型精度选择最优模型进行森林AGB反演和空间分布制图。结果表明：GF-6和Sentinel-2影像红边波段反射率与落叶松、樟子松AGB均呈显著相关（P<0.05），红边波段对AGB估测较敏感。多变量估测模型整体估测效果优于单变量模型，所有模型中多元线性回归模型取得了最优的决定系数（落叶松R²=0.66，樟子松R²=0.65）和最低的均方根误差（落叶松RMSE=31.45 t/hm²，樟子松RMSE=54.77 t/hm²）。相比单个数据源，联合GF-6和Sentinel-2影像构建的多元线性回归模型估测效果得到了显著提升，模型RMSE对于落叶松和樟子松AGB估测分别最大降低了22.9%和11.2%。增加红边波段进行AGB估测能显著提高模型估测精度，三组数据源分别加入红边波段信息后进行建模，模型RMSE得到了显著降低。GF-6拥有800 km观测幅宽和高效的重访周期，可以快速地提供大尺度时间序列数据，在森林地上生物量反演和动态监测方面有着很大潜力。

Retrieving the forest aboveground biomass by combining the red edge bands of Sentinel-2 and GF-6

Spectral reflectance can reflect the difference of ground objects, which is the theoretical basis of forest aboveground biomass (AGB) in remote sensing inversion. Red edge band is located in the special wavelength and scope which change fast at the junction of near-infrared and red bands, which can quickly respond to the small changes of vegetation canopy structure and chlorophyll content. Compared with other bands, red edge band is more sensitive to vegetation growth and chlorophyll change. In this study, GF-6 and Sentinel-2 multispectral images were used as data sources, and the linear and nonlinear models were constructed on the basis of Larch and Scotch pine field survey data for AGB estimation. The estimation accuracy of all models was compared, and the model with the highest estimation accuracy was regarded as the optimal model for the final AGB mapping in the study area. The results showed that the red edge reflectance of GF-6 and Sentinel-2 images were both significantly correlated with AGB (P<0.05), and red edge bands were more sensitive to AGB estimation than other bands. On the whole, the estimation effect of multivariable estimation model was better than that of univariate model. Multivariate linear regression (MLR) model obtained the highest determination coefficient (R²=0.66 for Larch and 0.55 for Scotch pine) and the lowest root mean square error (RMSE=31.45 t/hm² for Larch and 54.77 t/hm² for Scotch pine). Compared with single data source (GF-6 or Sentinel-2), the estimation effect of multiple linear regression model constructed by combining GF-6 and Sentinel-2 images was significantly improved, and the RMSE of AGB estimation model for Larch and Scotch pine decreased by 22.9% and 11.2% in highest measure, respectively. Taking the red edge band as additional variables for AGB estimation can significantly improve the estimation accuracy and effect of the model. The RMSE of the model was significantly reduced by adding red edge band information to three groups of data sources which are GF-6, Sentinel-2 and Sentinel-2 combined with GF-6. GF-6 has an observation width of 800 km and efficient revisit period, which can provide large-scale time series data quickly. As a remote sensing data source, GF-6 has great potential in forest aboveground biomass estimation and dynamic monitoring.