利用随机森林方法优选光谱特征预测土壤水分含量 下载: 1057次
Prediction of Soil Moisture Content by Selecting Spectral Characteristics Using Random Forest Method
1 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
2 绿洲生态教育部重点实验室, 新疆 乌鲁木齐 830046
图 & 表
图 1. 光谱采集实验示意图
Fig. 1. Schematic of spectral acquisition experiment
下载图片 查看原文
图 2. 不同SMC土壤样本光谱的反射率曲线及吸收特征曲线。(a)光谱反射率曲线;(b)吸收特征曲线
Fig. 2. Spectral reflectance curves and spectral absorption characteristic curves of sample with different soil moisture contents. (a) Spectral reflectance curves; (b) spectral absorption characteristic curves
下载图片 查看原文
图 3. 18种光谱吸收特征参数对SMC影响的贡献度
Fig. 3. Contribution degree of 18 spectral absorbance characteristic parameters on SMC
下载图片 查看原文
图 4. SMC实测值与预测值的比较
Fig. 4. Measured and predicted values of SMC
下载图片 查看原文
表 1土壤样本含水量统计特征
Table1. Statistical characteristics of soil sample moisture content
Sample set | Sample size | Mean value | Standard deviation | Maximum value | Minimum value | CV /% |
---|
Whole set | 38 | 14.59 | 5.76 | 23.94 | 1.48 | 39.48 | Calibration set | 25 | 15.10 | 5.44 | 23.94 | 1.48 | 36.01 | Validation set | 13 | 13.61 | 6.45 | 21.15 | 1.95 | 47.38 |
|
查看原文
表 2光谱吸收特征参数与SMC的相关性分析
Table2. Correlation analysis between spectral absorption characteristic parameters and SMC
Spectral absorptioncharacteristic parameter | SMC absorptionband /nm | Correlationcoefficient |
---|
D | 1400 | 0.90** | A | 1400 | 0.95** | DA | 1400 | -0.68** | La | 1400 | 0.95** | Ra | 1400 | 0.93** | S | 1400 | -0.13 | D | 1900 | 0.86** | A | 1900 | 0.73** | DA | 1900 | 0.49** | La | 1900 | 0.72** | Ra | 1900 | 0.71** | S | 1900 | -0.05 | D | 2200 | 0.93** | A | 2200 | 0.90** | DA | 2200 | -0.04 | La | 2200 | 0.90** | Ra | 2200 | 0.90** | S | 2200 | 0.30 |
|
查看原文
表 3表层SMC的模拟精度
Table3. Simulation accuracy of SMC at surface layer
Model | Training set | Test set |
---|
R2 | eRMSE | R2 | eRMSE |
---|
Random forest | 0.87 | 1.82 | 0.83 | 2.46 |
|
查看原文
表 4SMC的预测结果
Table4. Predicted SMC
Model | Regression equation | Calibration set | Validation set |
---|
| eRMSEC | | eRMSEP | RPD |
---|
MLSR | Y=3.68+2.16A2200+96.29D2200 | 0.88 | 2.08 | 0.89 | 2.21 | 2.80 |
|
查看原文
包青岭, 丁建丽, 王敬哲. 利用随机森林方法优选光谱特征预测土壤水分含量[J]. 激光与光电子学进展, 2018, 55(11): 113002. Qingling Bao, Jianli Ding, Jingzhe Wang. Prediction of Soil Moisture Content by Selecting Spectral Characteristics Using Random Forest Method[J]. Laser & Optoelectronics Progress, 2018, 55(11): 113002.