博斯腾湖西岸湖滨带土壤盐分高光谱分析

Hyper-spectral quantification of soil salinity in the west lakeside of Bosten Lake

以博斯腾湖西岸湖滨带为研究区，利用ASD Field Spec 3便携式光谱仪对土壤样品反射率进行测量，结合土壤盐分数据，建立研究区土壤含盐量估算模型，探讨土壤盐分含量和反射率曲线之间关系。结果表明：研究区土壤盐分类型以硫酸盐为主；土壤样品光谱曲线形态特征相似，且基本平行，土壤光谱曲线在400～1500、1900～2100和2100～2300nm区域出现明显的吸收带；在350～1350、1430～1750和1950～2400nm区域土壤光谱反射率与土壤盐分含量相关系数较其他波段大，曲线平缓，相关系数均值为-0．35，且光谱反射率的一、二阶导数微分均可以增强反射率与土壤盐分含量之间的相关性；由土壤光谱发射率一阶导数微分建立的多元回归模型可以较好地预测土壤盐分含量，相关系数为0．90，均方根误差为0．22。

Using the ASD Field Spec 3 portable spectrometer to measure the reflectance of soil samples, combined with soil salinity data, soil salinity estimation model for investigating the relationship between soil salt content and reflectivity curve was established. The results showed that the soil salinity in the study area dominated by sulfates. The morphological spectral curves were similar and substantially parallel to each other. In 400-1 500, 1 900-2 100 and 2 100-2 300 nm region, the soil spectral curve had significant absorption bands. In 350 1 350, 1 430-1 750 and 1 950-2 400 nm region, the correlation coefficients between soil spectral reflectance and soil salt content were lager than others. Reflectance curve was gentle. The mean correlation coefficient was -0.35. And the first derivative and the second derivative differential of spectral reflectance enhanced the correlation between salt content and soil reflectivity. The multivariate regression model based on the first derivative of soil spectral reflectance could be used to predict soil salt content, and the correlation coefficient reached 0. 90, and average root mean square error was 0.22.