Hoe klimaatmodellen de ekstremen bliuwe misse
May 19, 2026 · Frisian News
A new analysis of global climate models reveals they systematically underestimate the frequency and intensity of extreme weather events. The gap between predictions and reality raises questions about the tools governments use to set climate policy.
FryskIt Intergovernmental Panel on Climate Change brocht ferline moanne in wiidweidich ûndersyk út dêr't de útfier fan klimaatmodellen oer tritich jier yn ferlike waard mei werklike waarsgegevens. De konklúzje slacht hurd oan: modellen foarsizden ien geval fan ekstreme hitte yn elk desenium foar de measte regio's, mar yn de wurklikheid kamen trije oant fjouwer kear safolle foar. Dizze kleau docht derta omdat oerheden wrâldwiid koalstofbudzjetten en plannen foar útjeften by rampen opstelle op basis fan dizze projeksjes.
It probleem begjint by hoe wittenskippers dizze modellen bouwe. Se brûke supercomputers om de natuerkunde te simulearjen op in skaal fan sawat 100 kilometer yn elk rastersel, wat betsjut dat se lokaal terrein, wettermassa's en atmosfearike pûltsjes glêdstrike dêr't ekstreem waar werklik ûntstiet. As in model temperatuer foarsiist, foarsiist it in gemiddelde foar in gebiet sa grut as in lyts lân. De tongerslach dy't dyn stêd oerstreamt, de freak wyn dy't in bosk omsmyt, de hittekap dy't fee deamakket: dizze foarfallen falle tusken de gatten troch. Ûndersikers fan de Universiteit fan Melbourne ûntdutsen dat modellen ek weromkoppelingslussen te lyts weagje. As wiete grûn tidens in hittegolf útdrûget, ferdampt minder wetter, wat de hitte slimmer makket. De modellen behannelje dizze ketting fan reaksjes te swak.
Nimmen byhâldt dat de modellen weardeleas binne. Se lizze brede opwarmingstendensen krekt genôch fêst. Mar as beliedsmakkers in model oanhâlde om miljarden út te jaan oan klimaatadaptaasje, moatte se witte dat it model it gefaar wierskynlik ûnderskatte hat. In stúdzje út 2024 fan klimaatwittenskipper Kevin Trenberth toande oan dat fersekeringsbedriuwen no omtinken jouwe oan dizze kleau, en se ferheegje de premys flugger as it offisjele klimaatadvys suggerearret. De fersekeringssektor, dy't allinne winst makket as se risiko korrekt priist, hat besletten dat de modellen te sêft binne.
De ûndersyksfinansieringsstruktuer makket in perverse prikkel. Ûndersikers dy't modellen bouwe krije subsydzjes om se te ferbetterjen, net om ta te jaan dat se ekstremen net fêstlizze kinne. Oerheidsorganen finansierje stúdzjes dy't de needsaak oan klimaatmaatregels befêstigje, gjin stúdzjes dy't sizze dat ús warskôgingssystemen ûnder de maat prestearje. Partikuliere bedriuwen dy't klimaatrisikosoftware ferkeapje, hawwe gjin belang by it bagatellisearjen fan ûnwissichheid. Nimmen belutsen by de mainstream klimaatwittenskip wol klinke as in skeptikus, sels as de gegevens dat easkje.
It earlike paad foarút easket dat klimaatwittenskippers neist inoar publisearje: wat de modellen foarsizze, wat werklik barde, en dêr't de kleau it grutste is. Beliedsmakkers fertsjinje sifers dêr't se op fertrouwe kinne. No krije se projeksjes dy't as wissicheden foarjûn wurde, wat de ynstellings dy't se makke hawwe folle better tsjinnet as de minsken dy't werklik stoarmen temjitte gean.
The Intergovernmental Panel on Climate Change released a comprehensive review last month examining thirty years of climate model output against actual weather data. The conclusion hits hard: models predicted one extreme heat event per decade for most regions, but real temperatures produced three to four times as many. This gap matters because governments worldwide draft carbon budgets and disaster-spending plans based on these projections.
The problem starts with how scientists build these models. They use supercomputers to simulate physics at a scale of roughly 100 kilometers per grid cell, which means they smooth over local terrain, water bodies, and atmospheric pockets where extreme weather actually forms. When a model predicts temperature, it predicts an average for an area the size of a small country. The thunderstorm that floods your town, the freak wind that levels a forest, the heat dome that kills livestock: these events fall through the cracks. Researchers at the University of Melbourne found that models also underweight feedback loops. When wet ground dries during a heat wave, for instance, less water evaporates, which makes the heat worse. The models treat this chain reaction too weakly.
Nobody claims the models are worthless. They capture broad warming trends accurately enough. But when policymakers cite a model to justify spending billions on climate adaptation, they should know that the model probably underestimated the danger. A 2024 study by climate scientist Kevin Trenberth showed that insurance companies now pay attention to this gap, and they are raising premiums faster than official climate guidance suggests they should. The insurance industry, which profits only if it prices risk correctly, has decided the models are too soft.
The research funding structure creates a perverse incentive. Researchers who build models get grants to improve them, not to admit they cannot capture extremes. Government agencies fund studies that confirm the need for climate action, not studies that say our warning systems underperform. Private companies that sell climate risk software have no interest in downplaying uncertainty. Nobody involved in mainstream climate science wants to sound like a skeptic, even when the data demands skepticism.
The honest path forward requires climate scientists to publish side by side: what the models predict, what actually happened, and where the gap is widest. Policymakers deserve numbers they can trust. Right now, they get projections dressed up as certainties, which serves the institutions that produced them far better than it serves the people who face actual storms.
Published May 19, 2026 · Frisian News · Ljouwert, Fryslân