Scholarlink

The National Centre for Medium Range Weather Forecasting (NCMRWF) has set up a 1 hrly updated 1.5 km High Resolution Rapid Refresh (HRRR) Data Assimilation (DA) system with regional NCUM-R forecast model that is configured for three specific domains over the Indian region. In this study, we considered the model domain configured over the northeast region. The main objective of the present study is to demonstrate the assimilation of DWR observations in the HRRR DA system for the simulation of thunderstorms over the northeast states of India. Two set-up numerical experiments viz CTL (without DA) and HRR (Assimilation of radial velocity and reflectivity observations along with surface, sonde and satellite observations in 1 hrly 4DVAR HRRR DA system) are carried out. The analysis innovation (O-A) is relatively less than the background innovation (O-B) for all variables, suggesting that the assimilation of observations is properly utilized in the assimilation cycle. The location and distribution of rainfall are better simulated in the HRR experiment than in the CTL simulation. But the CTL simulated a high amount of rainfall that slightly shifted northeastward corresponding the observations. The peak rainfall is well brought out in the HRR simulation with minimal time lag while compared with the CTL simulation. The CTL simulation shows a high amount of rainfall during the early hours which may be due to the model spin-up. The temporal evolution of the convective system is well captured by the HRR with a higher correlation than in the CTL experiment. The ETS at different rainfall thresholds clearly shows an improvement in the HRR simulation than the CTL simulation. The forecast skill (%) of HRR with respect to the CTL is more at higher thresholds of rainfall. It is deduced that the high intensity of precipitation is well simulated by the HRR experiment than in the CTL experiment. At the same time, frequency bias is also considerably less in the HRR simulation than the CTL simulation in all cases. The object-based verification of rainfall clearly explains that the HRR experiment simulated the location and pattern of rainfall well with minimal errors than in the CTL simulation. But the volume errors are slightly more in the HRR simulation than the CTL. It is deduced that the HRR produces high intensity rainfall over the domain corresponding to the observation. Overall, the total errors are reduced in the HRR experiment while compared with the CTL. The present study results suggested that the hourly updated cyclic assimilation in the high-resolution modeling approach with the assimilation of DWR observations has a beneficial impact on such convective weather systems.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Rapid Update Cycle Data Assimilation for Severe Storm Prediction

  • 章节

  • 发表日期：

    2025年

查看全文
  • Ashish Routray,
  • Devajyoti Dutta,
  • Shivaji Singh Patel,
  • John P. George,
  • V. S. Prasad

摘要

The National Centre for Medium Range Weather Forecasting (NCMRWF) has set up a 1 hrly updated 1.5 km High Resolution Rapid Refresh (HRRR) Data Assimilation (DA) system with regional NCUM-R forecast model that is configured for three specific domains over the Indian region. In this study, we considered the model domain configured over the northeast region. The main objective of the present study is to demonstrate the assimilation of DWR observations in the HRRR DA system for the simulation of thunderstorms over the northeast states of India. Two set-up numerical experiments viz CTL (without DA) and HRR (Assimilation of radial velocity and reflectivity observations along with surface, sonde and satellite observations in 1 hrly 4DVAR HRRR DA system) are carried out. The analysis innovation (O-A) is relatively less than the background innovation (O-B) for all variables, suggesting that the assimilation of observations is properly utilized in the assimilation cycle. The location and distribution of rainfall are better simulated in the HRR experiment than in the CTL simulation. But the CTL simulated a high amount of rainfall that slightly shifted northeastward corresponding the observations. The peak rainfall is well brought out in the HRR simulation with minimal time lag while compared with the CTL simulation. The CTL simulation shows a high amount of rainfall during the early hours which may be due to the model spin-up. The temporal evolution of the convective system is well captured by the HRR with a higher correlation than in the CTL experiment. The ETS at different rainfall thresholds clearly shows an improvement in the HRR simulation than the CTL simulation. The forecast skill (%) of HRR with respect to the CTL is more at higher thresholds of rainfall. It is deduced that the high intensity of precipitation is well simulated by the HRR experiment than in the CTL experiment. At the same time, frequency bias is also considerably less in the HRR simulation than the CTL simulation in all cases. The object-based verification of rainfall clearly explains that the HRR experiment simulated the location and pattern of rainfall well with minimal errors than in the CTL simulation. But the volume errors are slightly more in the HRR simulation than the CTL. It is deduced that the HRR produces high intensity rainfall over the domain corresponding to the observation. Overall, the total errors are reduced in the HRR experiment while compared with the CTL. The present study results suggested that the hourly updated cyclic assimilation in the high-resolution modeling approach with the assimilation of DWR observations has a beneficial impact on such convective weather systems.