光谱学与光谱分析, 2019, 39 (11): 3514, 网络出版: 2019-12-02  

不同含水量黑土土壤光谱反射率半经验模型构建

A Semi-Empirical Model for Reflectance Spectral of Black Soil with Different Moisture Contents
袁静 1,2王鑫 1,2颜昌翔 1
作者单位
1 中国科学院长春光学精密机械与物理研究所, 吉林 长春 130033
2 中国科学院大学, 北京 100049
摘要
土壤含水量的变化情况与时空分布对热量平衡、 农业墒情等具有显著的影响。 利用反射率光谱信息反演土壤含水量的研究, 可为实现土壤含水量速测、 揭示土壤含水量时空变异规律提供科学依据。 构建不同含水量黑土土壤反射率光谱半经验模型, 深入探究土壤重量含水量与反射率光谱的关系。 制备了12种不同湿度的土壤样品。 采用ASD Field Spec Pro 3地物波谱仪对制备的不同湿度梯度的黑土土壤进行反射率光谱测量。 利用菲涅耳反射率建立土壤表面反射模型; 在以往的研究中, Kubelka-Munk (KM)模型中的漫反射率R∞通常被视为对于给定材料和照明波长的常数或需要反演的参数。 通过研究发现, 漫反射率R∞不仅与材料和波长有关, 还与土壤含水量相关。 利用与土壤含水量相关的吸收系数及散射系数描述了土壤含水量与漫反射率R∞的关系, 并基于KM理论对体散射分量进行建模; 进而构建不同含水量黑土土壤反射率光谱半经验模型。 根据实际测量数据选用最小二乘算法对模型参数进行反演, 并通过分析反演参数简化模型。 最后, 将未参与建模的不同含水量梯度的数据代入模型中, 验证模型的有效性。 结果表明: 对比不同含水量土壤反射率光谱的模拟值与实测值在400~2 400 nm波段范围内的模拟精度发现, 含水量为200 g·kg-1的土壤反射率光谱的均方根误差最大, 为0.008, 含水量为40 g·kg-1的土壤反射率光谱的均方根误差最小, 为0.000 6, 不同含水量下土壤样品反射率光谱的均方根误差的均值是0.005 1。 在400~2 400 nm波段范围内, 不同波长下黑土土壤反射率光谱的预测均方根误差基本低于0.008, 1 920 nm波长处的预测均方根误差最小, 为0.002 062。 采集长春地区的土壤检验模型的可靠性, 配制15个不同含水量样品并对其进行反射率光谱测量。 选取9个样品数据用于建模, 6个样品数据用于验证。 结果表明: 在400~2 400 nm波段范围内, 不同波长下的长春土壤反射率光谱的预测均方根误差基本低于0.015, 525 nm波长处的预测均方根误差最小, 为0.000 922 5。 综上所述, 所建立的模型具有很高的预测精度, 可很好地适用于不同含水量黑土土壤反射率光谱的模拟。
Abstract
The variation and the spatial-temporal distribution of soil moisture content have significant effects on heat balance, agricultural moisture, etc. Research on the inversion of soil moisture content using reflectance spectral information can provide a theoretical basis for realizing rapid test of soil moisture content and revealing the spatial-temporal variation of the soil moisture content. In this paper, a semi-empirical model of reflectance spectral of black soil with different moisture content was built to thoroughly explore the relationship between the soil moisture content and the reflectance spectral. First, 12 soil samples with different moisture contents were prepared. Secondly, reflectance spectral of the black soil with different moisture content gradients was measured by ASD Field Spec Pro 3 spectrometer. Then, the soil surface reflection model was built by using the Fresnel reflectivity; In previous studies, the diffuse reflectance in the Kubelka-Munk (KM) model was often considered as a constant for a given material and illumination wavelength or a parameter that needed to be inverted. It has been found through research that diffuse reflectance is related not only to material and wavelength, but also to soil water content. By using the absorption and scattering coefficients which related to the soil moisture content, this model described the relationship between the soil moisture content and the diffuse reflectance. Besides, a model of volume scattering component was built based on KM theory; Furthermore, a semi-empiricalmodel of reflectance spectral of black soil with different moisture content was built. Next, according to the measurement data, the least squares algorithm was used to invert the model parameters, and the model was simplified by analyzing the inversion parameters. Finally, the data of different moisture content gradients that were not used for modeling were substituted into the model to verify the validity of the model. The results showed that compared with the spectral simulation accuracy in the range of 400~2 400 nm under different moisture contents, the root mean square error (RMSE) of reflectance spectra of soil with moisture content of 200 g·kg-1 is 0.008 which is the largest, and the RMSE of reflectance spectra of soil with moisture content of 40 g·kg-1 is 0.000 6 which is the smallest, the mean value of the RMSE of reflectance spectraof soil under different moisture contents is 0.005 1. In the range of 400~2 400 nm, the root mean square error of prediction (RMSEP) of black soil reflectance spectra at different wavelengths is generally less than 0.008. The RMSEP at 1 920 nm band is 0.002 068 which is the smallest. The soil in Changchun was collected to test the reliability of the model, and reflectance spectra of 15 soil samples with different moisture content were measured. Six samples were selected for model validation, and the remaining samples were selected as a calibration dataset for model calibration. The results showed that in 400~2 400 nm band, the RMSEP of reflectance spectra at different wavelengths is generally less than 0.015. The RMSEP at 525 nm band is 0.000 922 5 which is the smallest. In conclusion, the established model has a high prediction accuracy and can be well applied to simulate the reflectance spectra of black soil with different moisture contents.

袁静, 王鑫, 颜昌翔. 不同含水量黑土土壤光谱反射率半经验模型构建[J]. 光谱学与光谱分析, 2019, 39(11): 3514. YUAN Jing, WANG Xin, YAN Chang-xiang. A Semi-Empirical Model for Reflectance Spectral of Black Soil with Different Moisture Contents[J]. Spectroscopy and Spectral Analysis, 2019, 39(11): 3514.

关于本站 Cookie 的使用提示

中国光学期刊网使用基于 cookie 的技术来更好地为您提供各项服务,点击此处了解我们的隐私策略。 如您需继续使用本网站,请您授权我们使用本地 cookie 来保存部分信息。
全站搜索
您最值得信赖的光电行业旗舰网络服务平台!