Science

For many outside the scientific community, climate models often appear as sources of detailed views of the future—a kind of “digital twin” of Earth capable of resolving everything from large-scale regional hazards like droughts to neighborhood-level flood risks. While that perception makes for excellent marketing, from a scientific standpoint, it is very exaggerated.

Regional climate models offer a physically grounded alternative to statistical downscaling by simulating local processes rather than simply re-gridding global data. Their added realism is beneficial for engineering, infrastructure, and risk applications where accurate local behavior and extremes matter.

Gridded weather datasets like ERA5 are widely used for engineering design, risk assessments, and resilience planning. Yet if these datasets are biased against station observations, those biases carry into future projections—risking under-designed infrastructure, underestimated exposure, and misguided adaptation strategies.