【摘要】 为探究不同海拔条件对土壤有效钾含量高光谱反演的影响以及筛选效果最好的光谱指标。采集118个土壤样本后进行其室内理化分析、光谱测量与处理等一系列工作,在土壤原始光谱(R)处理的基础上提取了反射率倒数一阶微分((1/R)’)、反射率倒数的对数一阶微分((log(1/R))’)和反射率对数的倒数一阶微分((1/(log R))’)三种光谱变换指标,分析土壤原始光谱和三种变换后的光谱指标与不同海拔区土壤有效钾含量的相关性,并运用偏最小二乘回归法(PLSR)建立不同海拔条件下土壤有效钾的高光谱预测模型。结果表明:(1)比较土壤原始光谱和三种变换后的光谱指标,基于(log(1/R))’变换结果构建的PLSR模型在土壤有效钾的反演效果最好,其决定系数(R2)最高,为0.89,均方根误差(RMSE)为12.45 mg kg-1;(2)相比全区域而言,依据海拔分区所建立的模型能够更好的预测土壤有效钾的含量。该结果对今后地形复杂区域土壤养分的光谱预测具有一定的指导作用。更多还原

【Abstract】 In order to analyze the influence of altitude conditions on the hyperspectral inversion of soil available potassium(AK) and to select the best spectral indicators, 118 soil samples were collected in three sub-regions that divided by altitudes and their AK concentrations and spectral data were measured. After that, the relationships between the concentrations of soil AK and the four spectral indices, including the raw spectral reflectance(R),first-order differential from reciprocal reflectance(1/R)’, first-order differential from logarithm of reciprocal reflectance(log(1/R))’ and first-order differential from reciprocal of logarithm reflectance(1/(log R))’, were conducted in the whole region and the three sub-regions. Furthermore, partial least-squares regression(PLSR) method was used to build quantitative inversion model of AK based on the results of relationship analysis. The results showed that the spectral data that transformed by(log(1/R))’ was the best indicator to predict soil AK concentrations with the highest value of determination coefficient(R2) and relative percent deviation(RPD) and the lowest value of root mean square error(RMSE) compared with the other three spectral transformation indices. The values of R2 and RMSE of the prediction model by using spectral data that transformed by(log(1/R))’ were 0.89 and 12.45 mg kg-1, respectively. Models predicting for soil AK concentrations built in sub-regions were better than those in the whole region. The results play an instructing role on hyperspectral estimation of soil nutrients in the complicated terrain areas.更多还原