, S)File last modified 3 December 1998
When it comes to evaluating the results from a GCM run, this evaluation essentially comes down to a comparison of the output with what we think we know about how the real world ocean works. But because there are a lot of assumptions built into a model as complex as a GCM we don't always expect to get an exact match between the model output and observations. This is where a good understanding of the physics behind the model makeup is important in guiding the model experimenter in deciding what are significant and trivial similarities and differences.
, S)In the model experiments conducted by Toggweiler et al. (1989a&b), and in general of most
GCM runs, the following was observed. The prognostic model of operation did a superior job at
reproducing the circulation, but the 
and S distributions compared more poorly to Levitus. In
general the bottom salinities were too fresh (approximately 0.25% too freash) and the
temperature deviation at mid-depths were too warm by up to 4oC. There was no salinity
minimum in the model output where the temperature deviation was the greatest.
The robust diagnostic version does a better job of matching Levitus (by almost a factor of two
over the prognostic mode), but the weak restoring terms in the interior (
) suppress convection
and other vertical motions causing major disruptions in deep sea ventilation. Diagnostic
calculations with a constant Ahv = 1.0 cm2 s-1 are approximately 50% higher in their deviations.
Wereas diagnostic calculations done with Ahv = f (z) have temperature deviations approximately
50% lower than the prognostic version. At great depth 
and S deviations are very small because
yr-1 is particularly effective where velocities are small.
Typically coarse resolution models have the following well known deficiencies. Either they have weak thermohaline circulations or their thermoclines are too warm. Additionally the model isotherms are usually several 100 meters too deep. The deep salinities of the model are too fresh and smooth numerical solutions require high eddy diffusivity parameterizations. Finally, western boundary currents do not separate at the proper latitude accompanied by spurious upwelling landward of the western boundary current.
In Toggweiler et al.'s model there was no chemical or biological transformations of 14C. The 14C
values are usually reported in 
, i.e. in ppt deviation from a standard 14C/12C ratio in a
standard reference material (in this case 19th century tree-rings). In the model the 14C values are
reported in a n arbitrary scale and the atmosphere is held constant at 100 of these units. To
convert between model to standard 14C scales use:
Keep in mind that 
is a ratio not a concentration, but this is one of those few cases where -
because this model has no chemical or biological transformations of the 14C - the ratio can be
treated as a concentration.
Toggweiler et al. performed five experiments on three different flow fields (prog, rdiag1, and rdiag2) with three different treatments of gas exchange.
| Experiment | Flow Field | Transfer Velocity | Vertical Eddy Diffusivity |
| A | RDIAG1 | 20 mol C m-2 yr-1 | Ahv = f (z) |
| B | RDIAG1 | k = f (U10) | Ahv = f (z) |
| C | RDIAG2 | k = f (U10) | Ahv = 1.0 |
| P | PROG | k = f (U10) | Ahv = f (z) |
| P' | PROG | k = f (U10) x 1.2 | Ahv = f (z) |
The surface concentration of C was set at 2 mol-C m-3 and in experiment A:
where DZ1 is the depth of the surface box in meters and 
is the rate of change of 14C at the
surface. This change takes place only at the surface grid cells and is propagated into the interior
of the model by the primitive equations. In experiments B through P':
where U10 is wind speed in m/s measured at 10 m above the surface of the ocean.
One of the first things one should do after running a GCM and getting some output is look at the circulations patterns. As you can see in Fig. 23.4.1, there are some important differences between the prognostic and diagnostic experiments.
You can see by examining Fig. 23.4.1 that the two shallow overturning cells are straddling the equator due to Ekman divergence (most of this activity is in the Pacific). Note also the mid-latitude overturning cells have a marked difference between the prognostic and diagnostic experiments (weaker in the diagnostic because the interior restoring parameter dampens vertical convection). You will also note that there is "mid-depth" flow from the N to the S compensated by a return flow above 1000 m, due mostly to the N. Atlantic deep water. This raises a problem for the model's 14C because the NADW is mostly above 2500 m and the data show it deeper. In examining the flow field one finds that the prognostic integrations are about 50% more vigorous than the diagnostic integrations in forming NADW. The abyssal northward flow, compensated by return flow above 3500 m (AABW), flows most strongly in the Pacific. Direct comparison to actual data is difficult due to problems with the time-mean meridional observations we mentioned in the last lecture.
When we look at the 14C profiles (globally averaged) at model steady state we observe the
following in comparison to real data and between the experiments. The bottom 14C in the model
is older than the 14C measured as part of GEOSECS, but the prognostic runs do a better than the
diagnostic runs in this respect. The prognostic runs also have a minimum at mid-depth as was
observed during the GEOSECS program, the diagnostic simulations do not. In experiment C (Ahv
= 1.0 cm2 s-1) they found that vertical mixing has a strong effect on the deep water 14C values.
One interesting outcome of all these experiments was that, apparently, global spatial variation of
gas exchange has little effect on the deep 14C values (which is why using 14C from the deep to
calibrate new gas exchange algorithms won't work, too insensitive). The prognostic experiment's
and S deviations do not, apparently, degrade its ability to ventilate the interior (does a better
job than the diagnostic runs). All of the experiments failed to produce enough NADW to fill the
N. Atlantic basin and so AABW penetrates too far north.
If you restrict yourself to looking at just the bomb-14C, you begin to see some additional features. The model's (all runs) bomb-14C inventory at GEOSECS time is low by about 16%. It is presumed that low model gas exchange rates are responsible for this phenomena. The model's upper ocean vertical mixing rate, especially in the recirculation regions of the subtropical gyres and cooler temperate and subpolar areas, appears to be too weak. Upwelling inshore of the western boundary currents causes glaring deficiencies in bomb-14C inventory distributions. The model also moves bomb-14C into the deep N. Atlantic interior too slowly.
Many of these flows were not known until 14C distributions were compared to GCM output.
GoTo Next Lecture