Eigen RadarScience
Analysis

EGGS-WG reproduces weather patterns, with gaps in heavy rain

Alternative weather sequences generated by EGGS-WG reproduced many average and seasonal conditions in a peer-reviewed evaluation across 5 regions. Rainfall extremes, dew-point variability and minimum temperatures were harder to represent. The statistical model draws on historical ERA5-Land data; its results describe simulated climates rather than a forecast for a particular future storm.

Science··Evening
Overhead view of a stainless-steel rain gauge and temperature sensor shield surrounded by rainwater pools during heavy rain in New Zealand.

Historical climate supplies alternative weather

EGGS-WG, an open-source statistical weather generator developed by Alex Schuddeboom and colleagues, creates alternative sequences of weather for land regions. Its starting point is ERA5-Land, a historical weather reanalysis dataset. Rather than predicting the next actual storm, it reproduces statistical relationships in the earlier climate. It links conditions across a grid of locations and generates hourly variations in air and dew-point temperature.[1]

Five regions test seasonal and daily patterns

The peer-reviewed evaluation covered 3 regions in New Zealand and one each in Australia and New Guinea. Tests compared rainfall occurrence, precipitation totals, seasonal changes, dry and wet spells, and relationships between locations. Average conditions and the daily temperature cycle generally matched the underlying data well. Regions with strong monsoon influences proved harder for the model to reproduce accurately.[1]

Heavy rainfall remains the difficult tail

Clearer problems appeared among rainfall events in the top 1 per cent, alongside weaknesses in minimum temperatures and dew-point variability. The generator can overproduce extreme rainfall, while its underlying weather data can underestimate extremes; the two errors require separate assessment. The study did not test large simultaneous events across extensive domains or distant climate connections such as El Niño. Its validation was confined to the selected regions.[1]

References

  1. News sourceGeoscientific Model DevelopmentEGGS-WG matched averages but struggled with rainfall extremes↩1↩2↩3