结合分数阶微分技术与机器学习算法的土壤有机碳含量光谱估测 下载: 1107次
Combination of Fractional Order Differential and Machine Learning Algorithm for Spectral Estimation of Soil Organic Carbon Content
1 新疆大学绿洲生态教育部重点实验室, 新疆 乌鲁木齐 830046
2 新疆大学资源与环境科学学院智慧城市与环境建模自治区普通高校重点实验室, 新疆 乌鲁木齐 830046
3 广东省生态环境技术研究所, 广东 广州 510650
图 & 表
图 1. SOC含量统计
Fig. 1. SOC content statistics
下载图片 查看原文
图 2. 不同有机碳含量的土壤光谱反射率曲线
Fig. 2. Spectral reflectance curves of soil with different organic carbon content
下载图片 查看原文
图 3. 相关系数通过0.01显著性检验水平的波段数量
Fig. 3. Number of bands whose correlation coefficient passes significance test level of 0.01
下载图片 查看原文
图 4. SOC含量与不同阶数在每个波段处的相关系数热力图
Fig. 4. Thermal map of correlation coefficient between SOC content and different orders at each bands
下载图片 查看原文
表 1ELM建模算法的模拟结果
Table1. Simulation results of ELM modeling algorithm
Order | R2 | RMSE | RPD |
---|
0 | 0.538 | 2.176 | 1.347 | 0.2 | 0.560 | 2.161 | 1.283 | 0.4 | 0.641 | 1.923 | 1.516 | 0.6 | 0.361 | 2.529 | 1.061 | 0.8 | 0.581 | 2.229 | 1.323 | 1.0 | 0.274 | 2.873 | 0.917 | 1.2 | 0.361 | 2.872 | 0.997 | 1.4 | 0.345 | 2.667 | 1.027 | 1.6 | 0.247 | 2.850 | 0.961 | 1.8 | 0.062 | 3.396 | 0.801 | 2.0 | 0.196 | 2.976 | 0.905 |
|
查看原文
表 2RF建模算法的模拟结果
Table2. Simulation results of RF modeling algorithm
Order | R2 | RMSE | PD |
---|
0 | 0.763 | 1.169 | 2.389 | 0.2 | 0.780 | 1.132 | 2.440 | 0.4 | 0.799 | 1.049 | 2.640 | 0.6 | 0.807 | 1.905 | 2.313 | 0.8 | 0.799 | 1.997 | 2.614 | 1.0 | 0.803 | 1.131 | 2.274 | 1.2 | 0.816 | 1.152 | 2.458 | 1.4 | 0.826 | 1.130 | 2.588 | 1.6 | 0.828 | 1.014 | 2.858 | 1.8 | 0.816 | 1.189 | 2.259 | 2.0 | 0.821 | 1.142 | 2.466 |
|
查看原文
表 3MARS建模算法的模拟结果
Table3. Simulation results of MARS modeling algorithm
Order | R2 | RMSE | RPD |
---|
0 | 0.621 | 2.047 | 1.470 | 0.2 | 0.710 | 1.724 | 1.774 | 0.4 | 0.795 | 1.522 | 1.923 | 0.6 | 0.709 | 1.708 | 1.670 | 0.8 | 0.830 | 1.464 | 2.113 | 1.0 | 0.829 | 1.471 | 2.149 | 1.2 | 0.846 | 1.218 | 2.604 | 1.4 | 0.867 | 1.071 | 2.783 | 1.6 | 0.845 | 1.231 | 2.535 | 1.8 | 0.847 | 1.175 | 2.656 | 2.0 | 0.844 | 1.172 | 2.681 |
|
查看原文
表 4Elastic Net算法的模拟结果
Table4. Simulation results of Elastic Net modeling algorithm
Order | R2 | RMSE | RPD |
---|
0 | 0.586 | 2.095 | 1.325 | 0.2 | 0.590 | 2.081 | 1.367 | 0.4 | 0.626 | 1.977 | 1.466 | 0.6 | 0.778 | 1.507 | 1.989 | 0.8 | 0.801 | 2.706 | 1.141 | 1.0 | 0.837 | 1.204 | 2.169 | 1.2 | 0.846 | 1.149 | 2.207 | 1.4 | 0.849 | 1.130 | 2.548 | 1.6 | 0.869 | 1.035 | 2.798 | 1.8 | 0.848 | 1.120 | 2.663 | 2.0 | 0.843 | 1.122 | 2.562 |
|
查看原文
表 5GBRT建模算法的模拟结果
Table5. Simulation results of GBRT modeling algorithm
Order | R2 | RMSE | RPD |
---|
0 | 0.788 | 1.916 | 2.162 | 0.2 | 0.793 | 1.883 | 2.275 | 0.4 | 0.814 | 1.712 | 2.618 | 0.6 | 0.830 | 1.526 | 2.707 | 0.8 | 0.841 | 1.381 | 2.848 | 1.0 | 0.846 | 1.276 | 2.267 | 1.2 | 0.847 | 1.290 | 2.679 | 1.4 | 0.857 | 1.257 | 2.882 | 1.6 | 0.878 | 1.125 | 3.142 | 1.8 | 0.851 | 1.282 | 2.798 | 2.0 | 0.848 | 1.264 | 2.813 |
|
查看原文
表 6不同波段处受土壤内部分子化学键的影响情况对比
Table6. Comparison of effects of different bands on molecular chemical bonds in soil
Band | Visible | Near infrared |
---|
Wavelength /nm | 500 | 620 | 890 | 1400 | 2200 | 2300 | Chemical bond | Fe—O、C—H | C—H | O—H | Al—OH | C—H |
|
查看原文
赵启东, 葛翔宇, 丁建丽, 王敬哲, 张振华, 田美玲. 结合分数阶微分技术与机器学习算法的土壤有机碳含量光谱估测[J]. 激光与光电子学进展, 2020, 57(15): 153001. Qidong Zhao, Xiangyu Ge, Jianli Ding, Jingzhe Wang, Zhenhua Zhang, Meiling Tian. Combination of Fractional Order Differential and Machine Learning Algorithm for Spectral Estimation of Soil Organic Carbon Content[J]. Laser & Optoelectronics Progress, 2020, 57(15): 153001.