Wenjun Cui

research

Research interests

Multi-Scale Dynamics of Convection: Investigating the complex interplay between synoptic-scale forcing and local mesoscale processes to understand the mechanisms driving thunderstorm development and evolution.
Severe Weather Climatology & Trends: Analyzing long-term variability in storm characteristics and tracks to quantify changing risks and understand the climatological footprint of severe weather events.
Remote Sensing & AI/Machine Learning Applications: Leveraging advanced machine learning algorithms and multi-sensor remote sensing (radar, satellite) to enhance the detection, classification, and analysis of convective hazards.

Recent work

Reconstructing a Climatology of Mesoscale Convective System Hazards using Machine Learning

The interannual variability of hazards are captured by the model-simulated MCSs but it overestimates the event count

This work utilizes a 13-year dataset of storm reports and atmospheric reanalysis to train object-based machine learning models, developing an observationally based climatology of MCS hazards. This reconstructed baseline is then used to evaluate how accurately convection-permitting models simulate severe events, such as flash floods, wind, hail, and tornadoes.

Key Findings:

Machine Learning Performance: The developed models successfully predict and distinguish flash flood and general severe events, though a model aimed at higher-impact significant-severe hazards showed limited skill due to small sample sizes.
Reconstructing Climatology: By applying the successful ML models back to the full MCS archive, a more complete warm-season hazard climatology was generated, correcting for historical underreporting and missing data in physical storm reports.
Model Simulation Evaluation: When applied to convection-permitting model simulations, the spatial distribution of simulated hazards generally aligns with real-world observations. However, overall event frequencies differ considerably, which can be attributed to biases in the derived input variables and the simulation's depiction of physical MCS properties.

Synoptic Weather Patterns and Mesoscale Convective System Flash Flood Potential in the United States

Hierarchical clustering effectively captured the seasonal and spatial variability of MCS-related synoptic patterns, identifying two primary spring clusters and three summer clusters concentrated in the Great Plains and Midwest

This work investigates the environmental precursors to MCS-driven flash floods by applying a machine learning clustering algorithm to long-term storm data, flash flood reports, and atmospheric reanalyses (2007–2017). By tracking the short-term evolution of synoptic weather patterns prior to storm initiation (12hr to 6hr), the study isolates the primary drivers of extreme rainfall.

Key Findings:

Pattern Clustering: While synoptic clusters effectively reflect regional and seasonal differences in MCS occurrence, they do not cleanly separate flood-producing systems from those that do not produce floods.
Regional Moisture Transport: In specific regions like the southern Great Plains, an MCS becomes highly flood-prone when localized synoptic forcing drives strong water vapor transport from a nearby moisture source.
The Dominant Factor: Regardless of the broader synoptic weather setup, the overall spatial footprint of the storm—specifically, a broader precipitating area—is the single most critical variable dictating flash flood potential.

Changes in mesoscale convective system precipitation structures in response to a warming climate

MCS occurrences will become more frequent in the southern United States under a warming climate

This work applies a satellite- and radar-based tracking algorithm to convection-permitting climate models to evaluate how MCSs will evolve under a pseudo-global warming scenario. By comparing historical and future simulations, the research isolates specific changes in MCS precipitation structures and intensity.

Key Findings:

Macro Trends: Despite a slight historical underestimation in MCS summer frequency and area, future projections indicate notable increases in MCS frequency, overall precipitating area, and total warm-season precipitation, primarily in the southern U.S.
Stratiform Regions: Demonstrate minimal spatial expansion, but mean and maximum precipitation rates increase by 15% and 29%, respectively.
Convective Cores: Exhibit a highly robust response, with size expanding by 24%, while mean and maximum precipitation rates surge by 37% and 42%.
Environmental Drivers: The future intensification of convective rainfall is directly attributed to the compounding effects of increased atmospheric instability and greater moisture availability.

Page updated

Google Sites

Report abuse