| 土壤有机质可见光-近红外光谱预测样本优化选择 |
| |
| 引用本文： | 肖云飞,高小红,李冠稳.土壤有机质可见光-近红外光谱预测样本优化选择[J].土壤,2020,52(2):404-413. |
| |
| 作者姓名： | 肖云飞 高小红 李冠稳 |
| |
| 作者单位： | 青海师范大学地理科学学院,青海师范大学地理科学学院,青海师范大学地理科学学院 |
| |
| 基金项目： | 国家自然科学(41550003)和青海省科技厅自然科学(2016-ZJ-907)。 |
| |
| 摘 要： | 土壤有机质可见光-近红外光谱预测中建模样本的优化选择对提高有机质模型估算精度具有重要作用。本文以湟水流域土壤有机质为例,采用基于土壤单一属性信息考虑的建模样本选择方法:浓度梯度法、Kennard-Stone(KS)方法,以及基于土壤多种信息考虑的建模样本选择方法:Rank-KS(RKS)法、土壤类型结合浓度梯度法以及土壤类型结合KS法。通过偏最小二乘回归建模,探索可见光–近红外光谱预测青海湟水流域有机质的最优样本集。结果表明:不同级别样本数的最佳建模样本选择方法不同,整体表现为基于土壤多种信息挑选的建模样本集的模型精度相比土壤单一信息均较高,特别是KS方法结合土壤类型后的建模样本集模型精度明显提高且在样本数较少时更为明显。土壤类型可以优化建模样本选择方法提高模型预测精度。在保证固定验证样本模型预测精度的情况下,土壤类型参与建模样本的选择可以有效减少建模样本数,进而降低了建模成本。
|
| 关 键 词： | 土壤有机质 可见光-近红外光谱 土壤类型 建模样本构建 湟水流域 |
| 收稿时间： | 2018/5/24 0:00:00 |
| 修稿时间： | 2018/12/20 0:00:00 |
Optimal Selection of Calibration Sample Sets for Predicting Soil Organic Matter Contents from Visible and Near Infrared Reflection Spectrum |
| |
| XIAO Yunfei,GAO Xiaohong,LI Guanwen.Optimal Selection of Calibration Sample Sets for Predicting Soil Organic Matter Contents from Visible and Near Infrared Reflection Spectrum[J].Soils,2020,52(2):404-413. |
| |
| Authors: | XIAO Yunfei GAO Xiaohong LI Guanwen |
| |
| Institution: | College of Life and Geographical Sciences,Physical Geography and Environmental Process Key Laboratory of Qinghai Province,Qinghai Normal University,College of Life and Geographical Sciences,Physical Geography and Environmental Process Key Laboratory of Qinghai Province,Qinghai Normal University,College of Life and Geographical Sciences,Physical Geography and Environmental Process Key Laboratory of Qinghai Province,Qinghai Normal University |
| |
| Abstract: | The optimal selection of correction samples in the light - near infrared spectrum prediction of soil organic matter is very important to improve the accuracy of organic matter model. This paper is mainly taking the HuangShui basin soil organic matter as the research smple. based on single attribute information considering soil sample selection methods: concentration gradient method, Kennard - Stone (KS) method, and based on the soil of a variety of information to consider correction sample selection methods: Rank - KS method (RKS) method, soil type combined with concentration gradient method and soil type combined with KS. By partial least squares regression modeling, the optimal sample set for the prediction of organic matter by visible - near infrared spectrum is explored. Results show that In different sample sizes, the best method to select the correct sample is different, and the model precision of the calibration sample set with multiple information of soil is higher than that model precision of the calibration sample set with a soil information. especially after KS method combined with soil type obviously improve the precision of calibration sample set model and the periods of low number of samples is more apparent. Soil type can be optimized to improve the accuracy of model prediction. In guarantee under the condition of fixed samples and model prediction accuracy and soil type involved in the calibration sample selection can effectively reduce the calibration sample, makes the prediction of soil organic matter visible light - near infrared spectrum is more economic. |
| |
| Keywords: | Soil organic matter Visible near-infrared reflection spectra Soil type Optimization of calibration set |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《土壤》浏览原始摘要信息 |
| 点击此处可从《土壤》下载免费的PDF全文 |
|
