|
|||||
|
|
| 基于土壤水分和气象要素的林火预报研究——以广东省为例 | |
| 引用本文： | 蔡霁初,邱建秀,王大刚,林凯荣,阳坤,曾庆峰.基于土壤水分和气象要素的林火预报研究——以广东省为例[J].地理科学,2021,41(9):1676-1686. |
| 作者姓名： | 蔡霁初 邱建秀 王大刚 林凯荣 阳坤 曾庆峰 |
| 作者单位： | 1.中山大学地理科学与规划学院/广东省城市化与地理环境空间模拟重点实验室，广东 广州 510275 2.中山大学土木工程学院，广东 广州 510275 3.清华大学地球系统科学系，北京 100084 4.广东省林火卫星监测中心，广东 广州 510060 |
| 基金项目： | 国家自然科学基金项目(41971031);国家自然科学基金项目(51779278) |
| 摘 要： | 基于国内现行的森林火险气象指数和单因子火险贡献度模型，以及逻辑回归模型和随机森林模型，在林火预报中引入微波遥感土壤水分信息，使用MCD14DL火点数据集和地面气象观测资料对广东省不同时间尺度的林火发生概率进行预测。结果表明：逻辑回归模型和随机森林模型构建的林火预测模型显著优于现行的森林火险气象指数和单因子火险贡献度模型，预测精度提升约20%。其中，随机森林模型对林火频数的解释程度最高（两者相关系数为0.476）。此外，加入微波土壤水分信息后，相较原有的基于气象要素的林火预测模型，2种机器学习模型的预测精度均略有提升，体现了表层土壤水分信息在林火预报中的重要性。研究可为高效提取对地观测信息，以改进华南地区不同时间尺度的林火预报工作提供参考。 |
| 关 键 词： | 广东省 森林火灾预测 土壤水分 逻辑回归模型 随机森林模型 |
| 收稿时间： | 2020-10-15 |
| 修稿时间： | 2021-01-20 |
Forest Fire Prediction Based on Soil Moisture and Meteorological Factors:Taking Guangdong Province As An Example |
|
| Cai Jichu,Qiu Jianxiu,Wang Dagang,Lin Kairong,Yang Kun,Zeng Qingfeng.Forest Fire Prediction Based on Soil Moisture and Meteorological Factors:Taking Guangdong Province As An Example[J].Scientia Geographica Sinica,2021,41(9):1676-1686. | |
| Authors: | Cai Jichu Qiu Jianxiu Wang Dagang Lin Kairong Yang Kun Zeng Qingfeng |
| Institution: | 1. Guangdong Provincial Key Laboratory of Urbanization and Geo-simulation, School of Geography and Planning, Sun Yat-sen University, Guangzhou 510275, Guangdong, China 2. School of Civil Engineering, Sun Yat-sen University, Guangzhou 510275, Guangdong, China 3. Department of Earth System Science, Tsinghua University, Beijing 100084, China 4. Guangdong Forest Fire Satellite Monitoring Center, Guangzhou 510060, Guangdong, China |
| Abstract: | In this article, four methods including current operational forest fire danger index, single-factor contribution model, logistic regression model and random forest model, are inter-compared for their prediction accuracies of forest fire probability at daily and monthly timescales in Guangdong Province. Excepting for the daily meteorological observations and MCD14DL Active Fire product, the microwave-based soil moisture dataset is also included in the latter two machine learning models, in order to evaluate their potential utilities in forest fire prediction. The results show that the logistic regression and random forest models significantly outperform the current forest fire danger index and the historical single-factor contribution model, increasing the accuracy by approximately 20%. The normalized forest fire probability from random forest model prediction are strongly correlated with the normalized active fire number (MCD14DL), showing correlation coefficient of 0.476. In addition, inclusion of soil moisture information in the meteorological factors-based model slightly increases model accuracy, which evidences the importance of surface soil moisture in forest fire prediction. The results of this study could provide reference for efficiently mining earth observations to improve forest fire prediction at different time scales, and therefore improve regional disaster preparedness measures. |
| Keywords: | Guangdong Province forest fire prediction soil moisture logistic regression model random forest model |
| 本文献已被 CNKI 万方数据 等数据库收录！ | |
| 点击此处可从《地理科学》浏览原始摘要信息 | |
| 点击此处可从《地理科学》下载免费的PDF全文 | |