Response from resident wacko environmentalist follows.

First, we have to separate two ways that science views the world: Observational-correlative and causational-experimental.

From grade school onward we tend to focus on the causational-experimental as the 'proper' role of science. This is the labcoats, test tubes, and rats type science. It's where you hold many factors constant and only allow a few parameters to vary then you note the response. There is a direct cause-effect relationship. This is reductionary and mechanistic. It is the most reliable method of science possible. It builds bridges, spacecraft, stents, antibiotics, televisions and nuclear bombs.

The second is observational correlative. Charles Darwin would be an extreme example of that method. He painstakingly observed natures variations in organisms and reasoned relationships between differences and environmental factors to derive his theory of evolution. Evolution, cosmology, psychology, sociology, and anthropology are all examples of observational-correlative science.

Most sciences can be viewed using both paradigms. In fact, both paradigms are often used to address the same problem. They compliment each other.

In the weather sciences these two approaches can be used to describe and predict future events. A combination of the methodologies provides people with 5-day forecasts. Your local weather woman uses a large bluescreened map to demonstrate the mass flow of patterns while behind teh scenes she may have run a model or used models provided by consultants. These models are based on physics--the prince of causative science.

Climate is like weather but looks at long term trends rather than short term events. The step from weather forecasting to climatological modeling is huge. But the two science paradigms are still at play.

Climatological studies have models that are statistically based, observational and correlative. They also have models that are statistically validated but physical science based. Let's call the first kind (stat/obs/corr) a statistical model and the other model we'll call a physical model, though both share attributes and methods of the other and advances in either advance both.

The statistical model uses patterns of past occurences and present data to extrapolate future trends. Data comes from coral growth rings, ice depositional patterns, tree rings, stratigraphic layering, isotopic ratios and other things. These factors vary according local weather conditions. By gathering data world-wide scientists can remove local phenomena and look at base trends. As more data is gathered the model gets refined and improved.

The physical model uses knowledge of atmospheric chemistry and physics to derive extrapolations.

Both models use hindcasting to help gain confidence in the model. They will "predict" what the climate should look like at some historical date and then test the model by evaluating physical evidence.

The link between climatology and ocean circulation has long been known (observationally) but has only recently been attacked by physical science. The linkage is not strong for any given moment of time but it is pretty strong for long term trends.

The temperature difference between poles and equator causes a difference in salinity. Also, the temperature difference that varies with depth affects the salinities. The dissociation constants of salts are directly temperature dependent. Also, the density of water varies with temperature. Dense, cold, salty water is heavy.

Whenever there is a temperature difference in the climate and on the ocean there is a potential for currents, both atmospheric and oceanic. The linkage is real.

However, this particular model is data-sparse. There are very few data sets of real measurements of the ocean that are sufficiently comprehensive to provide reliable model inputs. So the model spits out scary information but it is not well founded.

As an aside, the evidence for a global warming trend is far richer. There is ample, compelling evidence that serious, large changes in temperature are underway. What is less well known is the degree to which these changes maybe anthropogenic and also the precision of the model output.

Bot