双月刊

ISSN 1006-9895

CN 11-1768/O4

基于卫星降水和WRF预报降水的“6.18”门头沟泥石流事件的回报检验研究
作者:
作者单位:

作者简介:

通讯作者:

基金项目:

国家自然科学基金项目41605084,中国科学院国际合作局对外合作重点项目134111KYSB20150016


Hindcast Study of “6.18” Mentougou Debris-Flow Event Based on Satellite Rainfall and WRF Forecasted Rainfall
Author:
Affiliation:

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
    摘要:

    2017年6月18日北京门头沟地区突发泥石流,造成6人伤亡。短时强降水是这起事件的主要诱发因素,但常规气象观测并没有很好地观测到此次降水过程,可见降水数据的准确性对于滑坡泥石流的实时预警及预报至关重要。近年来,卫星遥感估算降水发展迅速,WRF(Weather Research and Forecasting Model)模式关于降水的预报技巧也逐渐提高。本文以自动站降水资料为参考,首先利用定性方法和泰勒图、TS(Threat Score)评分等定量的方法比较了CMORPH(CPC MORPHing technique)、GPM(Global Precipitation Measurement)和PERSIANN-CCS(Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks-Cloud Classification System)三种卫星降水资料以及不同起报时间的WRF预报降水对此次降水过程的表现能力,然后利用降水数据驱动滑坡泥石流统计预报模型,对此次事件进行了回报,分析不同降水数据在模型中的实际应用效果,最终为滑坡泥石流实时预警和预报系统的构建提供参考。结果表明,三种卫星降水资料基本上能反映出此次降水过程东北—西南向的带状空间分布形态,其中,CMORPH与自动站资料的空间相关性最好,命中率也最高,但对降水量有一定的高估,GPM对平均降水量的时间变化有较好的反映,体现了卫星降水在观测较少地区的良好利用价值,PERSIANN-CCS的表现则相对差些。WRF模式能预报出此次降水的带状空间分布特征,但降水中心的位置与实际有所偏差;此外,预报的最大降水量的峰值出现时间比实际上晚。由于此次降水的强局地性,只有空间分辨率均匀且质量相对较好的CMORPH卫星降水驱动模型可以回报出此次事件,而自动站点资料由于空间分布不均,则没有回报出此次事件,这表明了卫星降水在滑坡泥石流实时预警系统的构建中具有一定的优势。WRF模式降水驱动模型可以提前做出预警,虽然预报的事件发生时间与实际相比偏晚3~5 h,但WRF可以较好地预报72 h内的降水,因而可以延长灾害的可预见期。WRF模式预报降水的时间和空间精度都需要进一步提高,但是仍具有很好的参考意义。

    Abstract:

    On 18 June 2017, a debris-flow event suddenly occurred in Mentougou district of Beijing, and 6 people were dead or injured during this event. Short-term heavy rainfall is the main factor that triggered this event, while the routine meteorology observations didn't report this event very well. It can be seen that accurate rainfall observation and forecast are crucial for early warning of landslides and debris flows. In recent years, satellite remote sensing technology for estimating rainfall has been developed rapidly, and the forecasting skill of precipitation by WRF (Weather Research and Forecasting Model) has also been gradually improved. Based on rainfall data collected at automatic weather stations (AWS), the performance of three satellite rainfall products-CMORPH (CPC MORPHing technique), GPM (Global Precipitation Measurement) and PERSIANN-CCS (Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks-Cloud Classification System) and WRF model forecasted rainfall during this event are evaluated using qualitative and quantitative methods such as Taylor diagram and Threat Score. These rainfall data are then used to drive the statistical landslide and debris-flow forecasting model to hindcast this event. The applicability of different rainfall data in the model is analyzed, which provides references for the construction of landslide and debris-flow real-time warning and forecasting system. Results show that these three satellite rainfall products can well reproduce the rain belt, which extended from northeast to southwest in the study area. CMORPH has the highest spatial correlation coefficient with the observations and its probability of detection (POD) is also the highest, yet it overestimates the rainfall. GPM can well describe the temporal variability of area averaged precipitation, and shows the potential application of satellite remote sensing on rainfall observations in areas with less observational stations. The performance of PERSIANN-CCS is not as good as the former two. WRF can forecast the spatial distribution of the rain belt, but the simulated rainfall center has a bias compared with the actual position. Besides, the occurrence time of the forecasted maximum precipitation appeared later than its actual occurrence time. Due to the locality of the heavy rainfall, the landslide and debris-flow model can accurately hindcast the event only when driven by CMORPH, which has a uniform spatial distribution and good quality. The observational rainfall from AWS cannot make an accurate hindcast due to its uneven distribution, which indicates that the satellite rainfall has certain advantages in the construction of landslide and debris-flow real-time warning system. The landslide and debris-flow model driven by WRF simulation can make early warning of the event, although the forecasted event occurs 3-5 h late. WRF can well forecast precipitation within 72 h, and thus prolongs the predictable period of the event. The spatiotemporal accuracy of WRF model still needs to be improved, but it is still crucial in early warning of disasters.

    参考文献
    相似文献
    引证文献
引用本文

何爽爽,汪君,王会军.基于卫星降水和WRF预报降水的“6.18”门头沟泥石流事件的回报检验研究.大气科学,2018,42(3):590~606 HE Shuangshuang, WANG Jun, WANG Huijun. Hindcast Study of “6.18” Mentougou Debris-Flow Event Based on Satellite Rainfall and WRF Forecasted Rainfall. Chinese Journal of Atmospheric Sciences (in Chinese),2018,42(3):590~606

复制
分享
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:2018-04-12
  • 最后修改日期:
  • 录用日期:
  • 在线发布日期: 2018-05-31
  • 出版日期: