Researchers at North Carolina State University have developed a novel machine learning approach, named Complete Density Correction using Normalizing Flows (CDC‑NF), to enhance the accuracy of global climate models, particularly for compound extreme events like back‑to‑back heatwaves and heavy rainfall. These events, which involve simultaneous or sequential extremes in multiple variables (e.g. temperature and rainfall), have historically been poorly captured by standard climate models. When applied to the five most widely used global models, CDC‑NF significantly improved projections at both global and regional levels—including across the continental U.S., making model outputs align more closely with observed real‑world patterns.
Researchers at North Carolina State University have developed a novel machine learning approach, named Complete Density Correction using Normalizing Flows (CDC‑NF), to enhance the accuracy of global climate models, particularly for compound extreme events like back‑to‑back heatwaves and heavy rainfall. These events, which involve simultaneous or sequential extremes in multiple variables (e.g. temperature and rainfall), have historically been poorly captured by standard climate models. When applied to the five most widely used global models, CDC‑NF significantly improved projections at both global and regional levels—including across the continental U.S., making model outputs align more closely with observed real‑world patterns.
