English
引用本文:曹秀佳,谷健,马宁宁,等.基于粒子群优化小波神经网络模型的春玉米生育阶段干旱预测[J].灌溉排水学报,0,():-.
CAO Xiu-jia,GU Jian,MA Ning-ning,et al.基于粒子群优化小波神经网络模型的春玉米生育阶段干旱预测[J].灌溉排水学报,0,():-.
【打印本页】   【下载PDF全文】   查看/发表评论  【EndNote】   【RefMan】   【BibTex】
过刊浏览    高级检索
本文已被:浏览 1074次   下载 0  
分享到: 微信 更多
基于粒子群优化小波神经网络模型的春玉米生育阶段干旱预测
曹秀佳1, 谷健2, 马宁宁2, 刘泳圻3, 王子豪2, 尹光华2
1.中国科学院沈阳应用生态研究所 沈阳;2.中国科学院沈阳应用生态研究所;3.沈阳农业大学
摘要:
【目的】开展区域性作物生长季气候干旱预测,指导春玉米高效节水补灌生产。【方法】采用皮尔逊相关系数方法选取了与干旱指数最相关的因子,利用阜新市阜蒙县1965—2019年逐日气象数据,探索建立了粒子群算法优化的小波神经网络模型(PSO-WNN),采用春玉米不同生育阶段的水分亏缺指数验证了模型精度,并利用模型模拟预测了未来5年干旱发生情况。【结果】通过模型验证,春玉米5个生育阶段(播种-出苗阶段、出苗-拔节阶段、拔节-抽雄阶段、抽雄-乳熟阶段、乳熟-成熟阶段)的均方根误差(RMSE)分别为0.0419、0.0174、0.0481、0.0297、0.0421,决定系数R2分别为0.8402、0.9853、0.8990、0.9575、0.9177,且预测结果与实际干旱等级相符。未来5年该地区春玉米在播种-出苗阶段可能无旱或轻旱,出苗-拔节阶段可能发生中旱甚至特旱,生育后期干旱程度逐渐减弱,拔节-抽雄和抽雄-乳熟两个阶段出现轻旱概率较高,乳熟-成熟阶段出现干旱的概率较低,程度较小。【结论】文中构建的模型适用于阜新地区春玉米干旱预测,未来5年该地区春玉米生产应该更多关注出苗-拔节阶段的旱情。
关键词:  干旱预测;小波神经网络;春玉米;作物水分亏缺指数;模型
DOI:
分类号:S423
基金项目:“十三五”国家重点研发计划项目(2017YFD0300704,2018YFD0300301);辽宁省自然科学( 20180550617);中国博士后(2018M641708);
Drought Forecasting in Different Growth Stages of Spring Maize Using Particle Swarm Optimization-Wavelet Neural Network Model
CAO Xiu-jia,GU Jian,MA Ning-ning,et al
1.Institute of Applied Ecology,Chinese Academy of Sciences;2.Shenyang Agricultural University
Abstract:
【Background】Drought is one of the meteorological disasters in China and has occurred frequently In recent years. The area affected by droughts accounts for the largest proportion of meteorological disasters, which has a huge impact on society, economy and ecology. Drought in Northeast is more frequent, longer lasting, and more severe. In the context of global warming, increasing temperature and decreasing precipitation will cause droughts to occur more frequently and severely. Therefore, drought forecasting has become an important research topic in agricultural production.【Objective】This paper aimed to carry out drought forecast and guide the efficient water-saving supplemental irrigation production of spring maize better. 【Method】The Pearson correlation coefficient method was used to select the factors most relevant to the drought index among daily meteorological data of Fumeng County, Fuxin City from 1965 to 2019. The crop water deficit index at different growth stages of spring maize was forecasted using particle swarm optimization-wavelet neural network model (PSO-WNN). 【Result】The results show that, root mean square error (RMSE) are 0.0419、0.0174、0.0481、0.0297、0.0421, and the determination coefficients (R2) are 0.8402、0.9853、0.8990、0.9575、0.9177 respectively of the five growth stages (sowing-seedling, seedling-joining, jointing-tasseling, tasseling-milking, milking-maturity), and the forecast results are consistent with the actual drought levels. There may be no drought or slight drought in the sowing-seedling stage of the drought in this region, moderate or even extreme drought may occur in the seedling-jointing stage, and the degree of drought in the later growth stage Gradually weakening, the two stages of jointing-tasseling and tasseling-milking have a higher probability of slight drought, and the probability of drought in the milking-maturity stage is lower and less severe. 【Conclusion】This paper proves that the model is suitable for drought forecasting in Fuxin area. And more attention should be paid to drought of maize in the seedling-jointing stage in this region in the next five years.
Key words:  Drought forecasting; wavelet neural network; spring maize; crop water deficit index; model