基于偏相关分析的烟台市土壤温度影响因素及预测模型研究  

Influencing Factors and Prediction Models of Soil Temperature in Yantai Based on Partial Correlation Analysis

作　　者：崔素芳[1] 张振华[1] 姚付启[1] 张燕[1] 任尚岗[1] 

机构地区：[1]鲁东大学地理与规划学院,山东烟台264025

出　　处：《山东农业科学》2010年第1期18-21,共4页Shandong Agricultural Sciences

基　　金：山东省教育厅科技计划(J07YF16);鲁东大学大学生科技创新基金项目资助

摘　　要：根据烟台市气象局2005年的气象数据,利用偏相关分析方法分析了表层土壤温度与相对湿度、绝对湿度、大气温度、平均风速、日照时数、降水量6个气象因子之间的相关性,进而建立了基于相对湿度、日照时数、大气温度、绝对湿度4气象因子的多元线性回归模型和BP人工神经网络模型。结果表明:在6个气象因子中,相对湿度、日照时数、大气温度、绝对湿度与土壤温度存在极显著相关关系,平均风速、降雨量与土壤温度相关关系不显著;晴天时,BP神经网络模型的决定系数R2为0.9740,多元线性回归模型的决定系数R2为0.9739;阴天时,BP神经网络模型的决定系数R2为0.9881,多元线性回归模型的决定系数R2为0.9877,因此建立的神经网络模型具有很高的精度,能很好地满足土壤温度的预测要求。Based on the meteorological data in 2005 from Yantai Bureau of Meteorology, the correlation between the soil surface temperature and relative humidity, absolute humidity, air temperature, average wind speed, sunshine hours, rainfall was analyzed using partial correlation analysis method. The multivariate linear regression model and BP artificial neural network model were established based on relative humidity, sunshine hours, air temperature, absolute humidity. The results showed that there was significant correlation between soil temperature and relative humidity, sunshine hours, air temperature, absolute humidity, while the correla- tions between soil temperature and average wind speed and rainfall were not significant. In sunny day, the determination coefficient R^2 of BP neural network model was 0. 9740, and that of multiple linear regression model was 0.9739. In cloudy day, the determination coefficient R^2 of BP neural network model was 0.9881, and that of multiple linear regression model was 0. 9877. So the neural network model had a higher accuracy,which could well meet the forecast requirements of soil temperature.

关 键 词：土壤温度 偏相关分析 气象因子 预测模型 

分 类 号：S152.8[农业科学—土壤学]