基于支持向量机方法建立土壤湿度预测模型的探讨

支持向量机(Support Vector Machine简称SVM)方法,是通过核函数实现到高维空间的非线性映射,适宜于解决非线性问题,具有算法简单、计算量小、易于实现等优点。本文运用支持向量机方法建立了不同土层土壤湿度预测模型,0~10cm土层土壤湿度预测模型有较好的推广能力,10~50cm处的各层预测模型预报能力相对较弱。分析土壤湿度历史监测资料,发现同一时刻0~10cm土层与其它各土层土壤湿度具有较高的相关关系,基于此建立了预报精度较高的各土层土壤湿度的预测模型,实现了运用前期环境气象因子对各土层土壤湿度的预测。

Prediction Model of Soil Moisture Based on Support Vector Machines

The method of support vector machines(SVM),which can achieved non-liner mapping to high dimension space,is suitable for solving the problems of non-liner regressions.It also has advantages in simple and small scale calculating and easy to accomplish.The models for soil moisture prediction in different soil layers are put forward,and it is found,through real application,that the soil moisture forecast model established by SVM in 0-10 centimeter has a good popularizing ability,while there are some differences in 20-50 cm.By analyzing the historical data,we obtained that there is a higher relevance between soil moisture in 0-10 cm and the other soil layers,and then highly precise forecast models can be obtained.Therefore,the soil moisture in 10-50 cm can be forecasted with environmental and meteorological factors of early days.