How Climate Models Keep Getting the Extremes Wrong
March 8, 2026 · Frisian News
Climate scientists have built ever more complex computer models to predict future weather, yet these systems consistently underestimate extreme heat, cold, and storms. New research shows the models miss crucial feedback loops that make disasters worse than their forecasts suggest.
Last summer, weather stations across the northern hemisphere recorded temperatures that climate models predicted would not occur until 2040 or later. Heat domes over North America, Europe, and Asia broke records that stood for decades, yet climate scientists found their most sophisticated models had simply not seen these extremes coming. The gap between forecast and fact raises a hard question: if the models fail on events happening now, how much faith can we place in their predictions for 2050 or 2100?
The problem runs deeper than mere calculation errors. Climate models work by dividing the Earth into a grid and computing how air, water, and heat move through each cell. But the grid boxes are huge, often covering an area the size of a small country. Local effects that intensify storms, lock in heat waves, and amplify rainfall get smoothed away. A thunderstorm that dumps five inches of rain in one town and nothing in the next becomes a model average of two inches of rain everywhere.
Feedback loops pose an even trickier issue. When sea ice melts, darker ocean water absorbs more heat, which melts more ice. When soil dries out, plants wilt, which changes how much sunlight reflects back to space, which dries the soil further. Models include some of these loops, but they guess at the strength and timing. Recent studies show that these feedback effects amplify extreme weather far more than the models predict, turning a bad season into a catastrophic one.
Scientists at several universities have begun building models with higher resolution and more detailed feedback rules. Early results suggest that extreme events will hit harder and more often than earlier forecasts warned. This should matter to policymakers who rely on climate models to justify billion-dollar spending on adaptation and energy policy. If the models underestimate extremes by a large margin, the public money spent on preventing them may be badly aimed.
The honest answer is that climate models remain useful tools, but they are blunt ones. They show the long-term trend correctly enough. They fail badly at predicting when and where the worst will strike. Until scientists fix these gaps, anyone claiming to know with certainty what the weather will do in 2070 is talking past the evidence.
Foarige simmer registrearren wearstasions yn it noardlik healrûn temperatueren dy't klimaatmodellen net soe foarsizze oant 2040 of letter. Hittekoepels oer Noard-Amerika, Europa en Aazje braken records dy't dessinnia standhâlden, mar klimaatwetenskippers stelden fêst dat harren meast útwurke modellen dizze ekstremmen ienfâld net hain sjen kommen. De slurte tusken foarsizzjen en werklikheid stelt in swiere fraach: as de modellen tekoart komme by barren dy't no bierren, hoefolle fertrouwen kinne wy stelle yn harren foarsizzingen foar 2050 of 2100?
It probleem gaat djipper dan allinne rekenfouten. Klimaatmodellen wurkje troch de ierde yn in raster yndiele en te berekkenje hoe lucht, wetter en warmte troch elk fak bewegen. Mar de rasterfakken binne reusegrutsk, faak grut genôch om in lyts lân te dekken. Lokale effekten dy't sturmen fersterke, hittebalken fêststelle en reinfall fergrutsje, wurde glêd makke. In tûnwaar dy't yn de iene stêd fiif sintimeter rein dumpe en nocharis niks bale, wird in modelgemiddelde fan twa sintimeter rein oeral.
Terochwurkingssulsjes foarmje in noch lestiger probleem. Wannear see-iis smelt, opslorbert donkerder oseaanwetter mear warmte, wat mear iis smelte lit. Wannear ierde útdrûge, ferwelke planten, wat feroaret hoefolle sinnelight nei de romte weromkeaatst, wat de ierde fierder útdrûge lit. Modellen betrüke guon fan dizze sulsjes yn, mar giskaast nei harren sterkte en timing. Resint ûndersyk tonet oan dat dizze terochwurkingseffekten ekstrem waar folle mear fersterke as modellen foarsizze, wêrtroch in slear sezoen yn in ramtsjalopper sezoen feroaret.
Wetenskippers oan ferskate universiteiten hawwe modellen mei hegere resolúsje en mear detailleare terochwurkingsregels boud. Earste resultaten suggerearje dat ekstreme barren hurder en faker sûnne swinge as earstere foarsizzingen warskoed. Dit soe wichtich wêze moatte foar beliedsmekers dy't fertrouwe op klimaatmodellen om miljarden-euro's te rjochfeartichje foar oanpassing en enerzjybelied. As de modellen ekstremmen mei in grutte marge ûnderskatte, kin it iepenbier jild dat foar befoarking útjûn wurdt slechts rjochte wêze.
It earlike antwurd is dat klimaatmodellen brûkbere ynstrumenten bliuwe, mar stompe. Se toarje de lange-termyntrek korrekt genôch. Se failje slecht by it foarsizzen fan wannear en dêr it ergste sûnne swingen sil. Oant wetenskippers dizze gatten dikke, lit elkenien dy't mei sekerheid bewearje kin te witten wat it waar yn 2070 dwaan sil tsjin it bewiis yn.
Published March 8, 2026 · Frisian News · Ljouwert, Fryslân