基于Enter-PLSR的高光谱晋华宫矿南山煤矸石山景区土壤有效氮估测

土壤有效氮是反应土壤肥力的一个重要指标。高光谱技术对土壤有效氮进行监测和评价可以为土地复垦和生态修复提供指导性的动态信息。针对国家矿山公园高光谱有效氮监测研究的空白和偏最小二乘回归(PLSR)计算效率问题，利用ASD FieldSpec 3光谱仪采集了晋华宫矿南山煤矸石山景区(晋华宫矿国家矿山公园组成部分)，30个碱化栗钙土土样350～2 500 nm波段的光谱曲线，研究了土壤光谱与土壤有效氮含量之间的关系。对土壤反射率光谱进行一阶微分和倒数的对数变换，然后在提取特征吸收波段的基础上，分别进行PLSR和进入法-偏最小二乘回归(Enter-PLSR)建立估算模型并检验。结果表明Enter-PLSR估算模型在保证与PLSR估算模型相近准确度的情况下，自变量从122个降低到12个，大大提高了计算效率，在填补国家矿山公园高光谱有效氮监测研究空白的同时弥补了PLSR方法计算效率的不足。

Hyperspectral Extraction of Soil Available Nitrogen in Nan Mountain Coal Waste Scenic Spot of Jinhuagong Mine Based on Enter-PLSR

Soil available nitrogen content is an important index reflecting soil fertility. It provides dynamic information for land reclamation and ecological restoration if soil available nitrogen can be monitored and evaluated using hyperspectral technology. Facing the study blank of soil available nitrogen in National Mine Park and the deficiency of poor computational efficiency of partial least squares regression (PLSR) method, the present paper presents the relationship between soil spectrum and soil available nitrogen based on spectrum curves (ranging from 350 to 2 500 nm) of 30 salinized chestnut soil samples, which were collected from southern mountain coal waste scenic spot, located in Jinhuagong mine in Datong city, Shanxi Province, China (one part of Jinhuagong national mine park). Soil reflection spectrum was mathematically manipulated into first derivative and inverse-log spectral curves, then a corresponding estimation model was built and examined by PLSR and Enter-partial least squares regression (Enter-PLSR) based on characteristic absorption. The result indicated that Enter-PLSR corresponding estimation model greatly increased the computation efficiency by reducing the number of independent variables to 12 from 122 in case of a close accuracy of PLS corresponding estimation model. By using hyperspectral technology and Enter-PLSR method, the study blank of soil available nitrogen in National Mine Park was filled. At the same time, the computation efficiency problem of PLSR was resolved.