Professor Alfonso Sutera, in response to the post Q&A “On GCMs, Weather, and Climate” has alerted us to his relevant paper on the subject of global climate model predictive skill. It is
Bordi, I., K. Fraedrich, and A. Sutera(2010), Northern Hemisphere climate trends in reanalysis and forecast model predictions: The 500 hPa annual means, Geophys. Res.Lett., 37, L11809, doi:10.1029/2010GL043217.
The abstract reads
The lead time dependent climates of the ECMWF weather prediction model, initialized with ERA‐40 reanalysis, are analysed using 44 years of day‐1 to day‐10 forecasts of the northern hemispheric 500‐hPa geopotential height fields. The study addresses the question whether short‐term tendencies have an impact on long‐term trends. Comparing climate trends of ERA‐40 with those of the forecasts, it seems that the forecast model rapidly loses the memory of initial conditions creating its own climate. All forecast trends show a high degree of consistency. Comparison results suggest that: (i) Only centers characterized by an upward trend are statistical significant when increasing the lead time. (ii) In mid-latitudes an upward trend larger than the one observed in the reanalysis characterizes the forecasts, while in the tropics there is a good agreement. (iii) The downward trend in reanalysis at high latitudes characterizes also the day‐1 forecast which, however, increasing lead time
“….use forecasts to study the impact of short‐term tendencies
on long‐term trends when compared with reanalysis…we believe
this to be a novel analysis approach that moves forward the
issue of climate forecast.”
The conclusion includes the important perspective that
On these grounds, we offer the speculation that climate extrapolation, by simply using climate models, may be affected by different trends that observations and models have even at short lead time such as 10 days. Efforts to correct these discrepancies are in order. We wish to point out the great benefit gained in understanding climate behaviour by considering both reanalysis and short‐term forecasts obtained by a General Circulation Model (GCM). If climate has to be forecasted by deterministic means, in fact, the pitfalls of the method lie in the understanding of the statistics of the short‐term tendencies rather than in the forecast accuracy of long‐term averages. Whether a time dependent evolution of the atmospheric aerosol loadings may counterbalance the warming trends (higher geopotential) occurring in midlatitudes remains an open question. This counterintuitive effect may justify the cooling over the Pacific, but it is arduous to account for the warming trend over Eurasia.
The message that “[i]f climate has to be forecasted by deterministic means, in fact, the pitfalls of the method lie in the understanding of the statistics of the short‐term tendencies rather than in the forecast accuracy of long‐term averages” needs to be listened to by the IPCC community.