论文部分内容阅读
已有土壤有机质的光谱预测模型其适用性受建模样本的采样尺度、土壤类型及光谱参数限制,需要在大尺度及范围上进一步检验适用性,并比较分析不同建模方法的建模效果以寻求适用性更好、精度更高的定量模型。在黑河上游大尺度范围采得225个土壤样品,进行了土壤有机质(SOC)及光谱反射率测定后将样本划分为建模集(180个土样)与验证集(45个土样)。将土壤光谱反射率(R)变换处理后得到连续统去除(CR)、倒数(REC)、倒数之对数(LR)、一阶微分(FDR)及Kubelka-Munck变换系数共6种指标,针对建模集分别采用逐步线性回归与偏最小二乘回归方法建立12种光谱指标与SOC的数学模型,并采用验证集进行模型预测效果评价。结果表明:(1)采用逐步线性回归或偏最小二乘回归方法建模,LR指标对SOC变化的解释效果都是最好,是SOC的最优预测因子。(2)基于LR指标建立的SOC模型中,采用偏最小二乘回归模型比逐步线性回归模型的预测精度更好,相较于黑河上游已有的经验模型,偏最小二乘回归法建立的模型的预测效果也更好。(3)采用本实验的225个土壤样品对比验证了黑河上游仅有的SOC模型。该模型的SOC预测值与实测值通过了均值T检验且Pearson相关系数达0.826,表明在局部典型区域建立的SOC预测模型,可以应用到更大尺度上的土壤有机质预测研究。
The applicability of existing spectral prediction models for soil organic matter is limited by the sampling scale, soil type and spectral parameters of the modeling samples, and the applicability needs to be further examined on a large scale and in a range, and the modeling effects of different modeling methods are comparatively analyzed Seeking a more quantitative, more accurate quantitative model. A total of 225 soil samples were collected from a large scale in the upper reaches of the Heihe River. Soil organic matter (SOC) and spectral reflectance were measured and the samples were divided into modeling set (180 soil samples) and validation set (45 soil samples). After the soil spectral reflectance (R) was transformed, a total of 6 indexes including CR, REC, reciprocal logarithm (LR), first derivative (FDR) and Kubelka-Munck transform coefficient In the modeling set, the mathematical models of 12 kinds of spectral indexes and SOC were established by using stepwise linear regression and partial least-squares regression respectively. The validation set was used to evaluate the effect of model prediction. The results show that: (1) The method of stepwise linear regression or partial least-squares regression modeling, LR indicators explain the changes of SOC are the best, is the best predictor of SOC. (2) The partial least squares regression model is better than the stepwise linear regression model in the SOC model based on the LR indicator. Compared with the existing empirical model in the upper reaches of the Heihe River, the model established by partial least squares regression The prediction effect is better. (3) The comparison of 225 soil samples in this experiment verified the only SOC model in the upper Heihe River. The predicted and measured SOC values of the model have passed the mean value T test and the Pearson correlation coefficient reaches 0.826, indicating that the SOC prediction model established in the local typical region can be applied to the larger-scale prediction of soil organic matter.