摘要
植株群体叶面湿润时间(LWD)与病害发生密切相关,为预防叶面湿度过高引起日光温室番茄病害的发生,建立合理准确地日光温室番茄群体纵向尺度叶面湿度分布的预测模型十分必要。通过测定温室外光照强度、温室内温度、相对湿度及番茄植株不同高度的叶面湿润时间,分析温室环境因素对番茄群体纵向尺度叶面湿润时间的影响规律,并利用BP神经网络和多元线性回归建立番茄植株叶面湿润时间的预测模型。结果表明:叶面湿润时间与光照强度、温度呈现负相关关系,与湿度呈现正相关关系;植株不同高度的叶面湿润时间具有一定差异性,表现为植株群体上部叶片湿润时间变化最为明显,中部叶片变化次之,下部叶片变化最慢。距地面150,90和30cm高度处叶面湿润时间BP神经网络模型均方根误差(RMSE)分别为0.9262,1.3275和1.5568,多元线性回归方程的RMSE分别为2.0349,2.8907和3.4359,进一步的分析发现模拟值和实测值之间具有良好的对应关系。叶面湿润时间与光照强度、温度和湿度关系密切,且在群体内不同高度所受影响存在差异性,本研究建立的BP神经网络叶面湿润时间模型优于多元回归方程,可用于长季节栽培番茄群体叶面湿润时间的预测。
Leaf wetness duration （LWD） of plants population is closely related to the happening of disease, to prevent the happening of disease triggered by high tomato leaf humidity in solar greenhouse, establishing reasonable and accurately prediction model of LWD distribution in tomato population at a vertical scale in solar greenhouse is very necessary. Through measure light intensity outside the greenhouse, temperature and relative humidity inside the greenhouse and LWD, this study analyzed the influence of environmental factors on LWD at different heights of tomato population, and then established prediction model of LWD by BP neural network and multiple linear regression. Results showed that the LWD were negatively related with light intensity and temperature, and positively correlated with humidity. LWD was difference at different plant heights. The change of LWD was most obvious in upper leaves, the second was at middle leaves, the weakest change was at lower leaves. The Root Mean Square （RMS） errors of LWD by BP neural network model at 150cm, 90cm and 30cm were 0.9262, 1.3275 and 1.5568, respectively. The errors of RMS by multiple linear regression equation was 2.0349, 2.8907 and 3.4359, respectively. Further analysis found that simulated values had good corresponding relationship with measured values. LWD closely related to light intensity, temperature and humidity, and the impact at different heights was diversity. The LWD prediction model of BP neural network established in this study was superior to the multiple linear regression equation, which could be used for LWD prediction of greenhouse tomato cultivation for a long season.
出处
《沈阳农业大学学报》
CAS
CSCD
北大核心
2016年第6期647-653,共7页
Journal of Shenyang Agricultural University
基金
“十三五”国家重点研发计划项目(2016YFD0201004)
