• A machine-learning model (using XGBoost) predicts monthly 10-m wind speeds over Michigan with high accuracy (R² ≈ 0.96, RMSE ≈ 0.12 m s⁻¹) 

• The most influential predictors of wind speed in this region are local/regional features such as distance to the nearest Great Lake, surface roughness, and surface (skin) temperature—while large‐scale teleconnections (e.g., ENSO, AO, NAO) play a secondary but still measurable role

• Over the ~70-year period (1950–2020) analyzed, no statistically significant long-term trend in annual mean wind speed was found for Michigan, but the interannual variability has decreased

Globally recognized for his pioneering work related to extreme weather and climate change, Dr. Laiyin Zhu, associate professor in Western Michigan University's School of Environment, Geography and Sustainability, is being celebrated for his growing contributions to his field.

Quote from Science Advances: ONLINE COVER NASA’s SeaWiFS satellite captures a Saharan dust storm blowing 1000 miles into the Atlantic Ocean. Tropical cyclone rainfall significantly impacts coastal communities, primarily through inland flooding. Leveraging 19 years of hourly satellite observations of precipitation and other meteorological variables, Zhu et al. developed a machine-learning model to predict tropical cyclone rainfall of individual storms. They found that dust optical depth is a key predictor of rainfall and also improves model performance. The influence of dust optical depth on rainfall is complex and nonlinear and improving modeling tools will extend our understanding of this process.

• The study shows that the heavy rainfall region of tropical cyclones is expanding outward, meaning the area receiving intense rain around a cyclone is getting larger over time.
  
  

• This expansion is linked to warmer ocean and atmospheric conditions, which allow storms to hold more moisture and spread rainfall further from the storm center.
  
  

• As a result, more inland and wider surrounding regions are becoming exposed to flooding risks, even if wind intensity near the storm center does not increase.