Was sind die Temperaturprognosen wirklich wert? Klimamodelle bei wichtigem Realitäts-Check glatt durchgefallen

Klimamodelle mussten in letzter Zeit viel Kritik einstecken. Weder hatten sie die gebremste Erwärmung der letzten 15 Jahre vorhergesagt, noch vermögen sie die natürlichen Wärmeperioden der letzten Jahrtausende nachzuvollziehen. Ein Team um Christopher O’Reilly hat nun überprüft, inwieweit die gängigen Klimamodelle die Wintertemperaturen der letzten 100 Jahre der Nordhemisphäre reproduzieren können. Die Forscher waren erstaunt, als die Prognoseleistung Mitte des 20. Jahrhunderts nahe Null fiel. Ein bitterer Fehlschlag. Irgendetwas scheint in den Modellen zu fehlen. Ist es angesichts der krassen Defizite eigentlich vertretbar, die Politik von ebensolchen Modellen leiten zu lassen und weitreichende Entscheidungen zu treffen? Hier der Abstract der Arbeit, die im Juni 2017 erschien:

Variability in seasonal forecast skill of Northern Hemisphere winters over the twentieth century
Seasonal hindcast experiments, using prescribed sea surface temperatures (SSTs), are analyzed for Northern Hemisphere winters from 1900 to 2010. Ensemble mean Pacific/North American index (PNA) skill varies dramatically, dropping toward zero during the mid-twentieth century, with similar variability in North Atlantic Oscillation (NAO) hindcast skill. The PNA skill closely follows the correlation between the observed PNA index and tropical Pacific SST anomalies. During the mid-century period the PNA and NAO hindcast errors are closely related. The drop in PNA predictability is due to mid-century negative PNA events, which were not forced in a predictable manner by tropical Pacific SST anomalies. Overall, negative PNA events are less predictable and seem likely to arise more from internal atmospheric variability than positive PNA events. Our results suggest that seasonal forecasting systems assessed over the recent 30 year period may be less skillful in periods, such as the mid-twentieth century, with relatively weak forcing from tropical Pacific SST anomalies.

Noch ein Beispiel, diesmal aus dem Bereich des tropischen Pazifiks. Dort fanden Coats und Karnauskas, dass die Modelle auch hier komplett an der Realität vorbeisimulieren. Die Autoren schlussfolgern, dass jegliche Prognosen auf Basis dieser Modelle im Pazifik wenig vertrauenswürdig sind. Die Arbeit erschien im Oktober 2017 in den Geophysical Research Letters:

Are Simulated and Observed Twentieth Century Tropical Pacific Sea Surface Temperature Trends Significant Relative to Internal Variability?
Historical trends in the tropical Pacific zonal sea surface temperature gradient (SST gradient) are analyzed herein using 41 climate models (83 simulations) and 5 observational data sets. A linear inverse model is trained on each simulation and observational data set to assess if trends in the SST gradient are significant relative to the stationary statistics of internal variability, as would suggest an important role for external forcings such as anthropogenic greenhouse gasses. None of the 83 simulations have a positive trend in the SST gradient, a strengthening of the climatological SST gradient with more warming in the western than eastern tropical Pacific, as large as the mean trend across the five observational data sets. If the observed trends are anthropogenically forced, this discrepancy suggests that state-of-the-art climate models are not capturing the observed response of the tropical Pacific to anthropogenic forcing, with serious implications for confidence in future climate projections. There are caveats to this interpretation, however, as some climate models have a significant strengthening of the SST gradient between 1900 and 2013 Common Era, though smaller in magnitude than the observational data sets, and the strengthening in three out of five observational data sets is insignificant. When combined with observational uncertainties and the possibility of centennial time scale internal variability not sampled by the linear inverse model, this suggests that confident validation of anthropogenic SST gradient trends in climate models will require further emergence of anthropogenic trends. Regardless, the differences in SST gradient trends between climate models and observational data sets are concerning and motivate the need for process-level validation of the atmosphere-ocean dynamics relevant to climate change in the tropical Pacific.

Es deutet viel darauf hin, dass die Modelle überhitzen.