摘要
盐渍化土壤在我国分布的面积大,范围广。由于人为的不合理灌溉等,土壤次生盐渍化日益严重。为了能够及时、精准、动态地监测盐渍土农田含水率的变化,以内蒙古河套灌区农田为研究对象,应用高光谱分析技术,采用偏最小二乘回归方法(PLSR)分析土壤反射光谱特征值与水分、盐分含量间的关系,建立盐渍化农田含水率高光谱预测模型,并对模型的稳定性和预测能力进行检验。结果表明:利用非盐渍土建立偏最小二乘模型能够有效预测低含盐量(含盐量小<0.4%)土壤的含水率。水分预测模型中土壤盐分含量小于等于0.4%时,R2大于0.8,RMSE小于4%,预测精度较好;土壤盐分含量大于等于0.5%时,预测值R2不足0.7,预测精度较差。结果表明土壤中盐分含量较大时,对含水率预测模型的估算精度均会产生影响。
The area of saline soil in our country is large and wide. The secondary salinization is becoming more and more serious be- cause of the unreasonable irrigation. In order to timely, accurately and dynamically monitor of soil moisture changes of salinized farmland, this paper applied partial least squares regression （PLSR） method to analysis soil spectral reflectance characteristics with moisture and salt content using spectral analysis technology in Inner Mongolia Hetao irrigation area. The hyperspectral prediction model of salt-affected soil moisture was established and the stability and prediction ability of model were cheeked. The results showed that the partial least squares （PLS） model could be used to predict the moisture in the low salt-affected soil （salt content 〈0. 4%）. When the soil salt content was less than or equal to 0. 4%, the value of R2 was more than 0. 8, the RMSE was less than 4%, the prediction accuracy was good. When the soil salt content was more than or equal to 0. 5 %, the value of R2 was less than 0. 7, and the prediction accuracy was poor. The results show that the estimation accuracy of the prediction model of salt-affected soil moisture would be affected when the soil salinity is relatively large.
出处
《中国农村水利水电》
北大核心
2016年第8期73-75,82,共4页
China Rural Water and Hydropower
基金
国家自然科学基金资助项目(50376031
51279142)
内蒙古自治区水利科技计划项目([2014]117-2)
中央高校基本科研业务费专项资金(2042015kf0185)
关键词
土壤含水率
盐渍土
高光谱
预测模型
偏最小二乘回归
soil moisture
saline soil
hyperspectral
prediction model
partial least squares regression
