郑州邙山黄土剖面粒度和总有机碳的高光谱反演

Hyper-spectral inversion of grain size and total organic carbon in loess profile of Mangshan,Zhengzhou

黄土自身的发生和发展过程记录了丰富的历史信息,其粒度和总有机碳指标能较准确地反映出气候环境的演变.为探究高光谱遥感技术在获取黄土环境信息方面的应用,该研究以郑州邙山枣树沟村黄土剖面为研究对象,结合高光谱技术,通过对平滑处理后的原始光谱、一阶微分(FD)、二阶微分(SD)、去包络线(CR)和倒数对数(Log(1/R))与黄土剖面粒度和总有机碳开展相关性分析,选出相关系数R较大的波段作为特征波段建立基于PLSR(偏最小二乘回归)模型进行分析.研究发现:郑州邙山枣树沟黄土剖面中粒度和总有机碳变化指示了研究区全新世中期约5400 a.BP至今经历了冷干—暖湿—冷干的气候旋回;黄土不同地层单元的反射光谱特征虽在整体上曲线趋势相似,但其光谱反射率表现为黄土层L_(0-2)>黄土层L_(0-1)>过渡层Lt>古土壤层S_(0-1)>表土层T_(S)的规律;反演模型中经FD变换后的光谱为自变量的PLSR模型是反演黄土剖面平均粒径最佳模型,SD光谱变换为自变量的PLSR模型为反演黄土剖面TOC最佳模型.

The occurrence and development of loess has recorded abundant historical information,and its mean grain size and total organic carbon(TOC)can accurately reflect the environmental evolution.In order to explore the application of hyperspectral remote sensing technology in obtaining loess environmental information,Zaoshugou loess profile in Zhengzhou was taken as the research object.Combining hyperspectral technology,spectral data were mainly correlated with macro element of loess profile by smoothed original spectra,first-order differential(FD),second-order differential(SD),de-envelope(CR)and reciprocal logarithm(Log(1/R).A PLSR(Partial Least Square Regression)model was established to analyze the larger band of correlation coefficient R as characteristic band.The results were as follows:The changes of grain size and total organic carbon in Mangshan Loess profile indicated that the study area experienced a climate cycle of cold dry-warm wet-cold dry since about 5400 a.BP in the middle Holocene;The reflectance spectra of loess in different stratigraphic units showed their own characteristics and differences.The law of spectral reflectance was L_(0-2)>L_(0-1)>L_(t)>S_(0-1)>T_(S);Among the inversion models,the PLSR model with the spectrum transformed by FD as the independent variable was the best model for retrieving the average grain size of loess profile,and the PLSR model with the spectrum transformed by SD as the independent variable is the best model to retrieve the TOC of loess profile.