| 林嘉豪1, 户雪敏1, 段维彤1, 杨俊杰2, 黄朝炎3, 黄卫利4, 李宇翔1, 胡小平1*.湖北小麦条锈病预测模型的构建及应用[J].植物保护,2025,51(1):161-168. |
| 湖北小麦条锈病预测模型的构建及应用 |
| Construction and application of wheat stripe rust prediction model for Hubei |
| 投稿时间：2024-02-07 修订日期：2024-04-07 |
| DOI：10.16688/j.zwbh.2024078 |
| 中文关键词: 小麦条锈病 发生面积 全子集回归 BP神经网络算法 |
| 英文关键词:wheat stripe rust occurrence area full subset regression BP neural network algorithm |
| 基金项目:国家重点研发计划(2021YFD1401000); 科技部国际合作项目(G2023172013L); 西北农林科技大学推广项目(TGZX2021-13); 西北农林科技大学高水平创新团队项目(XYTD2023-04) |
| 作者 | 单位 | E-mail | | 林嘉豪1, 户雪敏1, 段维彤1, 杨俊杰2, 黄朝炎3, 黄卫利4, 李宇翔1, 胡小平1* | 1. 西北农林科技大学植物保护学院, 植保资源与病虫害治理教育部重点实验室, 农业农村部西北黄土高原作物有害生物综合治理重点实验室, 杨凌 712100
2. 湖北省植物保护总站, 武汉 430070
3. 湖北省襄阳市植物保护站, 襄阳 441021
4. 西安黄氏生物工程有限公司, 西安 710065 | xphu@nwsuaf.edu.cn |
|
| 摘要点击次数: 404 |
| 全文下载次数: 84 |
| 中文摘要: |
| 湖北省是我国小麦条锈菌的重要冬繁区之一, 也是小麦条锈病由西北越夏区传播至华中小麦主产区的关键通道, 加强对湖北省小麦条锈病的准确预测和科学防控至关重要?本研究利用1995年-2023年的数据, 通过相关性分析, 结合滑动窗口法筛选出了与湖北省小麦条锈病发生面积相关的因子, 包括日平均气温?日平均最高气温?日平均最低气温?平均日照时数?日累积降水量, 并以1995年-2020年的数据构建了条锈病发生面积的基于全子集回归模型和BP神经网络模型?结果表明, 全子集回归模型1和2对1995年-2020年小麦条锈病发生面积回测准确度分别为88.7%和88.1%, 对2021年-2023年的预测准确度分别为89.8%和95.2%; BP神经网络模型1和2对1995年-2020年小麦条锈病发生面积回测准确度分别为96.5%和95.8%, 对2021年-2023年的预测准确度分别为91.6%和90.9%?因此, BP神经网络模型1是湖北省小麦条锈病发生面积的最佳模型? |
| 英文摘要: |
| Hubei province is a crucial winter propagation area for wheat stripe rust pathogen in China and serves as a key transmission route for the disease from the northwestern over-summering regions to the main wheat-producing areas in central China. Accurately predicting and scientifically controlling wheat stripe rust in Hubei province is very important. This study utilized data from 1995 to 2023, employing correlation analysis and the sliding window method to identify factors related to the occurrence area of wheat stripe rust in the region, including daily average temperature, daily average maximum temperature, daily average minimum temperature, average sunshine hours, and daily cumulative precipitation. Prediction models for the occurrence area of wheat stripe rust from 1995 to 2020 were constructed using full subset regression and BP neural network algorithm. The backtesting accuracies for the full subset regression models 1 and 2 were 88.7% and 88.1%, respectively, while the prediction accuracies from 2021 to 2023 were 89.8% and 95.2%, respectively. For the BP neural network models 1 and 2, the backtesting accuracies from 1995 to 2020 were 96.5% and 95.8%, respectively, and the prediction accuracies from 2021 to 2023 were 91.6% and 90.9%, respectively. Therefore, BP neural network model 1 is the best model for predicting the occurrence area of wheat stripe rust in Hubei province. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
