摘要
【目的】为油茶花期低温阴雨灾损的客观评估提供方法,为油茶气象防灾减灾提供参考。【方法】利用2006—2016年湖南31块油茶样地的油茶测产资料和同期临近格点的气温、光照和降水等资料,采用相关分析、方差分析和SHAP值分析相结合的方法,筛选湖南油茶花期低温阴雨关键气候影响因子,并利用筛选出的关键因子采用XGBoost算法建立油茶花期低温阴雨灾损评估模型。【结果】在16个油茶花期低温阴雨灾害气候影响因子中,极端最低气温、平均最低气温、日平均气温小于10℃积温和寡照时间是造成油茶花期灾损的关键影响因子,这4个因子与油茶产量相关性较好,在增产与减产样本间的差异显著,在模型构建中的贡献较大。通过分析4个关键影响因子对增产或减产的贡献,发现当日平均气温小于10℃积温超过150℃,或寡照时间多于25 d,或平均最低气温不大于9℃时,往往会伴有低温阴雨灾害出现,这些临界值可用于低温阴雨灾害的监测,对油茶花期是否有灾损的发生具有一定的指示意义。所构建油茶花期低温阴雨灾损模型的预测准确率较高,为82.4%,受试者工作特征曲线下面积(AUC)值高达0.82,说明该模型具有较好的预测能力。2020年湖南油茶花期低温阴雨的灾损模型评估结果与花期气象条件综合分析的结果相吻合。【结论】所构建湖南油茶花期低温阴雨灾损模型具有较好的预测能力和较好的适用性,可用于油茶花期低温阴雨灾损评估业务工作中。
【Objective】To provide an objective method for the loss estimation of low temperature and overcast rain damage at florescence of Camellia oleifera and to provide technical support for meteorological disaster prevention and mitigation.【Method】Based on the yield data of C.oleifera from 31 C.oleifera sample plots of Hunan province during 2006-2016 and temperature,sunshine and precipitation data near grid points during the same period,the key climate influencing factors of low temperature and overcast rain damage at florescence of C.oleifera in Hunan were screened by correlation analysis,variance analysis and SHAP value analysis.According to the selected key factors,the loss estimation model of low temperature and overcast rain damage at florescence was established by using XGBoost algorithm.【Result】Among 16 climate factors affecting of low temperature and overcast rain damage at florescence of C.oleifera,extreme minimum temperature,average minimum temperature,daily average temperature less than 10℃active accumulated temperature and scant lighting days were the key factors causing the yield reduction of C.oleifera at florescence,and four factors had a good correlation with the yield.By analyzing the contribution of three key influencing factors to increase or decrease production,it was found that when the daily average temperature less than 10℃active accumulated temperature exceeded 150℃,or scant lighting days were more than 25 days,or the average minimum temperature was not greater than 9℃,there would be low temperature and overcast rain damage,and these index thresholds could be used for the monitoring of low temperature and overcast rain damage and had certain indicative significance for the occurrence of disaster damage at florescence of C.oleifera.The prediction accuracy of the damage model was 82.4%,and the AUC value was as high as 0.82.These evaluation indexes showed that the model had good prediction ability.The assessment results of the loss estimation model of low temperature and overcast rain damage at florescence of 2020 were consistent with the results of the comprehensive analysis of meteorological conditions in the flowering period.【Conclusion】The loss estimation model of low temperature and overcast rain damage at florescence of C.oleifera in Hunan province had a good prediction ability,and through the practical application of the model,it was found that the model had a good applicability,and could be used in the low temperature and overcast rain disaster assessment of C.oleifera in the flowering period.
作者
张超
谢佰承
ZHANG Chao;XIE Baicheng(Hunan Key Laboratory of Meteorological Disaster Prevention and Reduction,Institute of Meteorological Sciences ofHunan Province,Changsha 410118,Hunan,China)
出处
《经济林研究》
北大核心
2022年第4期218-227,共10页
Non-wood Forest Research
基金
湖南省自然科学基金项目(2020JJ4396)
湖南省气象局重点项目(XQKJ20A009)。
关键词
花期
低温阴雨
灾损模型
湖南
florescence
low temperature and overcast rain
loss estimation model
Hunan
