成都西部山区森林火险气象等级预测方法评估
DOI:
CSTR:
作者:
作者单位:

1.四川省成都市气象局;2.四川省彭州市气象局

作者简介:

通讯作者:

中图分类号:

基金项目:

成都市气象局业务科技研究课题[2024 - 1 (17) ]


Evaluation of Forest Fire Risk Meteorological Rating Prediction Methods for the Western Mountains of Chengdu
Author:
Affiliation:

1.Chengdu Meteorological Office of Sichuan Province;2.Pengzhou Meteorological Station of Sichuan Province

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    【目的】进一步加强对彭州森林火险的认识,准确预测森林火险气象等级,提升当地气象服务水平。【方法】利用 2014-2023 年彭州国家基本气象站及 49 个区域气象站小时观测数据分析森林火险时空分布,基于最近邻回归(KNN)和随机森林算法预测森林火险气象等级。【结果】森林火险气象指数受气象因素影响大,较大值无明显周期性,其数值与海拔呈正相关(山区需防范时间较丘区更长,基本覆盖春、秋、冬季);机器学习算法模型预测结果整体趋势与实际观测一致,可信度高,月指数均方根误差 < 6. 2, 平均绝对误差 < 4. 52, 对于森林火险气象等级的预测基本无影响;KNN 算法在山区表现更佳,而随机森林算法在丘区表现更佳。【结论】揭示了彭州森林火险气象等级的时空分布特征,展现了机器学习在局地森林火险气象等级预测中的优良效果,有助于提升当地气象业务服务能力。

    Abstract:

    This paper aims to further enhance the understanding of forest fire risk s in Pengzhou, accurately predict the forest fire risk meteorological grades, and improve the service capability of the local meteorological department. Using hourly observation data from the Pengzhou National Basic Meteorological Station and 49 regional meteorological stations from 2014 to 2023, this paper analyzes the spatial and temporal distribution of forest fire danger in Pengzhou and predicts the forest fire risk meteorological rating based on the KNN and Random Forest algorithms. The results indicate that the forest fire risk meteorological index is significantly influenced by meteorological factors, with larger values showing no clear periodicity. Its values are positively correlated to altitude. The mountainous areas require a longer time of prevention compared to hilly areas in spring, autumn and winter. The overall trend of the prediction results by machine learning algorithm models is consistent with actual observations, with a high reliability, monthly index root mean square error < 6. 2, and average absolute error < 4. 52, which has essentially no impact on the prediction of forest fire risk meteorological rating. Among the two algorithms, the KNN algorithm performs better in mountainous areas, while the Random Forest algorithm performs better in hilly areas. This study has revealed the spatio - temporal distribution characteristics of the forest fire risk meteorological rating in Pengzhou, demonstrating the excellent effectiveness of machine learning in predicting local forest fire risk meteorological rating, and laying a scientific foundation for strengthening local meteorological service capabilities.

    参考文献
    相似文献
    引证文献
引用本文

刘若岚,罗坤,高梦醒,等.成都西部山区森林火险气象等级预测方法评估[J].山地气象学报,2026,50(2):22-30.

复制
分享
相关视频

文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2025-01-18
  • 最后修改日期:2025-06-20
  • 录用日期:2025-06-27
  • 在线发布日期: 2026-05-20
  • 出版日期:
文章二维码