堆栈集成学习联合多源数据反演草地土壤水分

Inversion of Soil Moisture in Grassland Based on Stacking Ensemble Learning with Multi-source Data

为了有效提高机器学习模型反演土壤水分的稳定性和准确性,提出了一种堆栈集成学习模型,组合多种机器学习模型的优势,提升模型的精度和泛化能力。首先,通过相关性分析,得出入射角为42.5°时,观测数据与土壤水分相关性最好;其次,使用水云模型和τ-w模型构建模拟数据库,与观测数据共同组成反演数据集;最后,利用4种机器学习和堆栈集成学习模型反演稀疏草地的土壤水分。实验结果表明:采用主被动协同比单一主动或被动微波数据的反演结果具有更高的精度;最优堆栈集成学习方法反演结果的决定系数达到0.9714,均方根误差和平均绝对误差分别达到0.0136 cm^(3)/cm^(3)和0.0102 cm^(3)/cm^(3),均优于最优单一机器学习方法,验证了该方法的有效性。

Aiming at the problem of how to effectively improve the stability and accuracy of machine learning models for inversion of soil moisture,the paper proposes a stacking ensemble learning model to improve the accuracy and generalization ability of the model by combining the advantages of multiple machine learning models.Firstly,through correlation analysis,the observation data correlate best with soil moisture when the incidence angle is 42.5°.Secondly,a simulation database is constructed using the water cloud model and theτ-w model,together with the observation data form the inversion dataset.Lastly,the soil moisture of sparse grassland is inverted using the four types of machine learning and the stacking ensemble learning model.The experimental results show that:the inversion results using active-passive synergistic microwave data have higher accuracy than that of single active or passive microwave data;the coefficient of determination of the inversion results of the optimal stacking ensemble learning method reaches 0.9714,and the root mean square error and the average absolute error reach 0.0136 cm^(3)/cm^(3) and 0.0102 cm^(3)/cm^(3),which are better than that of the optimal single machine learning method,which proves the effectiveness of the proposed method.