The 5th Mesoscale Meteorological Seminar

Update: 06. 12, 2018

(Jun. 15-16, 2018)

Research Meeting in Kashiwa Campus

Date: June.15.2018（Fri）9:00-17:30
       June.16.2018（Sat）9:00-17:30
Venue: Auditorium
       Atmosphere and Ocean Research Institute, The University of Tokyo
       5-1-5 Kashiwanoha, Kashiwa-shi,Chiba 277-8564, JAPAN
Invited Speakers: Dr. Russ Stanley Schumacher, Colorado State University
Contact Person: Dr. Takashi Unuma, meso.discuss@gmail.com
Contact Person at AORI: Prof. Keita Iga, meso.discuss@gmail.com

June 15th
 09:30--11:30    The ingredients for extreme rainfall, and how mesoscale convective systems (MCSs) can bring them together: Part I (Dr. Russ S. Schumacher, Colorado State University, United States)
 11:30--13:00     Lunch
 13:00--14:00     The ingredients for extreme rainfall, and how mesoscale convective systems (MCSs) can bring them together: Part II (Dr. Russ S. Schumacher, Colorado State University, United States)
 14:00--14:30    An overview of mesoscale convective systems (MCSs) producing heavy rainfall in Japan (Mr. Hiroshige Tsuguti, Chief Scientist at Meteorological Research Institute, Japan)
 14:30--15:00     Environmental properties of quasi-stationary convective events during the warm seasons in Japan (Dr. Takashi Unuma, Japan Meteorological Agency, Japan)
 15:15--15:45     Discussion
 15:45--16:00     Break
 16:00--17:30     Poster presentations by participants

June 16th
 09:30--10:30    The sensitivity of modeled MCSs to low-level moisture (Dr. Russ S. Schumacher, Colorado State University, United States)
 10:30--11:00     Real-time data assimilation of ground-based microwave radiometer network for cloud-scale short-range rainfall forecast in Tokyo Metropolitan region (Dr. Ryohei Kato, Dr. Shingo Shimizu and Dr. Ken-ichi Shimose, National Research Institute for Earth Science and Disaster Resilience, Japan)
 11:00--11:30     Discussion
 11:30--13:00     Lunch
 13:00--14:00     Advancing probabilistic rainfall forecasts through ensemble prediction and application of machine learning (Dr. Russ S. Schumacher, Colorado State University, United States)
 14:00--14:30     Ensemble_based analysis for heavy rains and tornadoes in Japan (Dr. Sho Yokota, Meteorological Research Institute, Japan)
 14:30--15:00     Discussion
 15:00--15:15     Break
 15:15--16:15     Oral presentation by participants
 16:30--17:30     General discussion

Invited Speaker's Summaries:
1.  The ingredients for extreme rainfall, and how mesoscale convective systems (MCSs) can bring them together
Heavy precipitation can be summarized by the equation sometimes called the “first law of quantitative precipitation forecasting,” that the total precipitation at a point is simply the product of the average rain rate and the rainfall duration. Extreme rainfall accumulations can occur when the rain rate, the duration, or both are very large. Yet despite the simplicity of this equation, understanding and (especially) predicting extreme rain events can be a great challenge. The ingredients for large rain rates include water vapor and upward motion, which can be reasonably represented in observations and models, but cloud microphysical processes, which are not well understood, can either enhance or mitigate rain rates. Rainfall duration is a function of the size, speed, and organization of the rain-producing storms. When they are organized and move in particular ways, mesoscale convective systems (MCSs) can produce both very high rain rates for a long period of time.

This presentation will explore the processes in MCSs that support extreme precipitation, including insights from observations and numerical models. We will address questions such as: what distinguishes extreme-rain-producing MCSs from those producing more modest amounts? What processes lead to slow-moving, long-lasting MCSs that often produce heavy rain? What role do vortices (including tropical cyclones, mesoscale convective vortices, and mesocyclones) play in heavy-rain production? What are some of the key remaining questions in the understanding of extreme-rain-producing MCSs?

2.  The sensitivity of modeled MCSs to low-level moisture
Although numerical models often faithfully represent the structures and processes associated with MCSs, they often have large errors in the placement and rainfall production of those MCSs. Here, I will show how this may be related to errors and uncertainties in the profile of water vapor in the MCS environment. Two sets of simulations, with varying levels of complexity, show that small changes to the near-surface moisture profile can lead to large changes in the resulting MCS. Unsurprisingly, small additions or subtractions of water vapor change the rainfall rates and accumulations from the MCS, but more surprisingly they can also result in large differences in the location and spatial distribution of rainfall.

Radiosonde observations from the 2015 Plains Elevated Convection At Night (PECAN) field campaign show substantial moisture differences from operational model analyses used to drive convection-allowing forecasts. Analysis of the 24-25 June 2015 PECAN MCS reveals similar sensitivities as the idealized simulations: moisture errors in models lead to displacement errors in MCS precipitation. Differences between the observed moisture in PECAN soundings and the analyzed moisture in the operational Rapid Refresh analysis will be presented across a large sample of cases from 2015, and potential implications for quantitative precipitation forecasting will be discussed.

3.  Advancing probabilistic rainfall forecasts through ensemble prediction and application of machine learning
Heavy precipitation has inherently limited predictability, owing to the small spatial scales on which it occurs, and uncertainty in the atmospheric initial conditions and representation of physical processes. As such, it is best to present forecasts of heavy precipitation probabilistically. Ensemble forecasting provides a potential remedy to this problem, yet many limitations remain: ensembles can be computationally expensive to run, they tend to be underdispersed, and there are challenges in presenting low-probability but potentially high-impact forecast scenarios. In this presentation, I will outline some recent advances in post-processing for ensemble forecasts of heavy precipitation that involve the application of machine-learning algorithms to a long record of model forecasts and resulting observations. The method incorporates both model QPF and heavy-rainfall ingredients, and outputs probabilities of exceeding various recurrence interval thresholds. It results in improvements over the "raw" ensemble output in all areas, but especially in regions where most heavy rainfall comes from convective systems. I will also discuss our collaboration with operational forecasters to include this machine-learning model output as an input to their forecast process.