Commentary on the Paper of Barnett et al.

KEVIN E. TRENBERTH

National Center for Atmospheric Research

Dr. Barnett has given us a very "colorful" talk dealing with an ocean GCM that has very simple forcing. The nature of the forcing, as he mentioned, is critical in terms of its magnitude in particular, and its special character to a lesser degree. The first thing that I thought of—probably a minor point—with regard to the fresh-water forcing is the aspect dealing with runoff from rivers into the ocean. In fact, the precipitation exceeds evaporation over land in general. Is that in the model? If you just model E - P over the ocean, there will be imbalances, because there E exceeds P. What does that do to the overall fresh-water balance?

Dr. Barnett used three kinds of forcings. The main one he talked about was the white/white, which seems somewhat unrealistic. He used a global standard deviation value of 2 mm/day. That number may be reasonable globally, but should be much larger than that in the tropics and about half that in high latitudes. But large numbers at high latitudes were one of the critical things in getting this oscillation.

My other concern is that I consider the model to be unnaturally constrained. Any time you use an idealized atmosphere, the nature of the fluxes into the model ocean and the ability of the atmosphere to feed back and adjust in various ways are limited. For instance, can the climate system adjust to compensate for the fresh-water flux? The parameterization derived here for the red/red case, especially in the tropics, does indicate that the fresh-water flux can be altered substantially by changes in sea temperatures. Certainly other aspects of land-ocean differences introduce similar complexities. Therefore, the relevance of something like this to the real world is, I think, a very open question.

Dr. Barnett did not talk about mechanisms, although maybe other atmospheric modelers will. Perhaps, because the model integrations are made over wide areas, the spatial structure does not matter much; if the area mean is the main thing that counts, the result is a red spectrum whatever you do. Wide-area integration also implies that there is a random-walk process that will result in perturbations in the fresh-water flux. The perturbations will affect or even shut down the thermohaline circulation, which will alter the heat balance because of the heat transports involved, which in turn will cause changes in temperatures.

This process is reflected in the temperature variations the model exhibits at high latitudes. Dr. Barnett asks, "Is this natural variability something that is going to confound us when we look at the greenhouse effect?" But ultimately, shutting down the thermohaline circulation will change the temperatures enough that they will probably cause the thermohaline circulation to jerk back into action at some point and advect fresh-water around. Might this be part of the mechanism that results in an oscillation? I do not know. I think the bottom line is the question of these models' relevance to the real world.

Discussion

BARNETT: That last question does need to be kept in mind, but for a full-ocean GCM our model does quite a good job of reproducing the main features of the global ocean.

TRENBERTH: Isn't the forcing at high latitudes much higher than in the AMIP run?

BARNETT: Not much. We picked 2 mm/day because it's a fairly good global average, and I think it's fairly realistic. I believe some other models using a smaller figure still get the oscillation.

SARACHIK: Why did you choose to show the transport of the Antarctic circumpolar current, and what was the structure of the changes in it?

BARNETT: It's a simple diagnostic for the system. If you have no signal there, you won't have much anywhere.

BRYAN: I'd like to respond to Kevin's comment about runoff. For the North Atlantic, I believe that runoff is dominant at very high latitudes simply because the coastline is so extensive by comparison with the ocean area. If you also take into account the tremendous Arctic fresh-water discharge, the total runoff is much greater than the local net precipitation.

BARNETT: This runoff effect has been included in a couple of models for the Hamburg greenhouse runs, and it's my impression that it didn't make much difference. If it is large, though, it should be fairly simple to put into these kinds of models.

TRENBERTH: Runoff might affect the nature of feedbacks, and clearly on time scales that cover the melting of major ice caps it would become critical.

WEAVER: It seems to me that you do have a sort of parameteriza-



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement