Talk given at the AMS meeting in Phoenix. 14 Feb 1998

Modeling the response of the Pacific to Central American mountain-gap winds

Figures from a talk given at the AMS meeting in Phoenix. 14 Feb 1998

William S. Kessler and Zuojun Yu

The dramatic effects on the Pacific of the strong winds that blow through the mountain gaps in Central America have been known for more than 35 years. Strong down-gradient winds occur through the three passes 10-12 times each boreal winter. Wind speeds can be as large as 20 m/s, and the events typically last for 5-7 days. These winds are due to winter synoptic high pressure events over the Great Plains, extending south over the Gulf of Mexico. Probably the most climatologically important aspect of this phenomenon is the rapid cooling that occurs in boreal fall when strong winds blow on this region with its very shallow summer thermocline. However, the winds also generate long-lived anticyclonic eddies that have been explained as the result of offshore winds exciting an initially symmetric pair of eddies, but immediately destroying the cyclonic (upwelling) one through entrainment under strong winds (McCreary et al, 1989, JMR p81). The purpose of this talk is to suggest that other processes may well contribute to the process of eddy formation; in particular there seems to be a connection to the reflection of intraseasonal equatorial Kelvin waves.

  1. Mean and standard deviation of SST in the north-eastern tropical Pacific
    Even in this very highly averaged mean picture, the SST in the Pacific west of the three mountain gaps is 0.5 to 1°C cooler than surrounding areas. If only the climatological winter season is considered, the influence of the mountain gaps is even more pronounced. This signal is a significant aspect of the regional SST climatology. Note that the regions of low mean SST are also regions of high variance of SST.

  2. Standard deviation of NSCAT wind speed (High-passed)
    The high-passed wind variability is also large west of the mountain gaps, and this variability extends well offshore. It can be shown that at synoptic timescales, high atmospheric pressure at Houston is well correlated with northerlies at the Gulf of Tehuantepec.

  3. Time series of NSCAT winds along 95°W during February-March 1997
    This time series, extending from the Mexican coast in the Gulf of Tehuantepec to the equator and cutting across the two high-SST-variance regions of Fig. 2, shows the episodic occurrence of Tehuantepec and Papagayo winds. Four event were seen during this period, the largest of which, in mid-February, had speeds in excess of 10 m/s (as measured by NSCAT, which is probably too low; see discussion of following figure). Note the apparent correlation between Tehuantepeckers (10°-15°N) and Papagayos (4°-10°N). Both appear to be driven by the same high-pressure forcing.

  4. Time series of meridional winds at 14°N, 95°W from NSCAT, ECMWF and FSU monthly winds
    This time series shows that about 10 wind events were seen during boreal winter 1996-97, just offshore of the Gulf of Tehuantepec. ECMWF and NSCAT winds were highly correlated, although ECMWF winds were stronger. This is due to the NSCAT winds being averaged over 5 days; if a 5-day average is taken of ECMWF winds then the magnitudes are comparable. (See also comparisons of NSCAT swath and gridded magnitudes, and NSCAT vs ECMWF winds during the strongest Tehuantepec event of the NSCAT period). If the episodic strong wind events are important to the evolution of the ocean in this region, then one would expect that model solutions forced with the FSU monthly winds would produce a different solution than that forced with either of the other two products.

    The Gent/Cane OGCM was forced with the three wind products. This is a high time and space resolution version of the model (20 minute timestep, 20 km grid spacing in the Central American region (see the grid spacing for the fine-grid runs)). The model was first spun up with 3 years of the FSU 1961-91 climatological winds. For the ECMWF and FSU runs, these winds were imposed beginning 1 January 1996 (model year 4). The NSCAT run was started from the results of the ECMWF run on 1 November 1996. All the runs were carried through the NSCAT period (to 26 June 1997). Results discussed below are from these runs.

    Based on the differences among the wind products shown in Fig. 4, we expected to see substantial differences among the model solutions. Surprisingly, this was not the case. Figs. 5 and 6 show the model SST and sea level during February 1997, for the three wind products.

  5. SST during Feb 1997
  6. Sea level during Feb 1997
    There is relatively little to choose from among these three runs. In particular, all three showed the warm SST/high sea level signature of an apparent Tehuantepec eddy near 12°N, 103°W. In all runs, the eddy was a sea level rise of about 30-40 cm from adjacent regions.

  7. Sea level on 20 February 1997 from TOPEX and the model (NSCAT run)
    This snapshot shows that the model representation of the eddies during the strong wind period was fairly realistic. The altimeter shows 2 strong and 2 weaker anticyclonic eddies. The model has a somewhat more jumbled picture. It is easy to identify the largest eddy at 12°N, 104°W, and the smaller satellite eddy to its northwest with the corresponding features in the altimeter. The model splits the other large eddy at 12°N, 91°W into two, but the other small eddy to the northwest is approximately correct.

    Comparing the observed and model fields in a time-longitude plot shows that the model representation of the time evolution is fairly realistic:

  8. TOPEX and model sea level at 10°-15°N during July 1996-May 1997
    This figure (and the succeeding 3 as well) shows the maximum sea level between the latitudes 10° and 15°N. This shows all the eddies in the band together (as in Fig. 7, two large ones are near 12°N and two smaller ones near 14°N. In both model and observations, a large eddy is seen gaining amplitude from about 97°W in November 1996 and propagating west to 115°W by May 1997. However, the initiation of this eddy can be traced back to July 1996 at about 90°W, as the model makes clear. The large amplitude increase occurs in November, though. Similarly, the second large observed eddy gains amplitude in February 1997 at about 92°W. The model splits this eddy in two (as seen above in Fig. 7), and both of these are seen to emanate from the boundary, one in November and one in January. A suggestion that the observed eddy extends back to the boundary in November is barely visible. Thus the picture seems to be that the eddies blossomed during the high wind period (roughly October-March, see Fig. 4), but weak antecedents can be traced back several months before.

    This raises the question of what generated these eddies. It is usually thought that the mountain-gap winds produce the eddies (through the process explained by McCreary et al, 1991). But then why were there more than 10 strong wind events but only 2 large and 2 small eddies during the winter of 1996-97? And why do the eddies seem to be traceable back to well before the wind events occurred?

    To further explore these questions, a model run was made in which the winds forcing the model east of 120°W were fixed to their 1 November 1996 values from that date onwards. Wind forcing variability continued as observed west of 120°W, but in the eastern Pacific it remained constant thereafter. Thus there were no mountain-gap winds forcing the model during boreal winter 1996-97. Fig. 9 compares the full model solution (same as the right panel of Fig. 8) with the fixed-wind solution, and the difference between the two..

  9. Solutions with full forcing and fixed east Pacific winds

    Surprisingly, the solution without any mountain-gap winds still retains most of the same eddies in recognizable form, though with amplitude reduced by about a factor of two. Apparently the model eddies, at least, are not entirely generated by the mountain-gap winds. A second surprise is that the difference field is quite similar to the fixed wind solution. In a linear sense, one can think of the middle panel of Fig. 9 as representing the effect of the west Pacific winds, and the right panel as the effect of the east Pacific winds. Why should these be similar? Why should both elements of the wind field have a similar effect on the eastern ocean? A third surprise is that the eddies in both elements of the solution grow as the eddies move west.

    A possibility for surprise number one is suggested by Fig. 10, which shows that the three "Tehuantepec eddies" during boreal winter 1996-97 emanated from the boundary apparently as the reflection of equatorial Kelvin waves. The Rossby wave speed appears to be about 11 cm/s, which would imply that the long gravity wave speed c is about 2.2 m/s in this region. However, it is clear from the second of Figs. 10 (with overlaid crest lines), that the propagation speed is not perfectly constant, but appears to increase slightly to the west (consistent with the westward deepening of the thermocline?). In addition, there is an ambiguity about the starting point of Rossby reflection, since these figures show the maximum sea level between 10° and 15°N, and due to the tilt of the coast the boundary longitude changes from about 85°W to 93°W in this range. Since the eddies appear at different latitudes (e.g. Fig. 7), it is not clear where the signal leaves the boundary, and therefore estimating where the reflection crest lines should go in Figs. 10 is uncertain.
    Nevertheless, the hypothesis suggested by Figs. 9 and 10 is that the reflected (high sea level) Kelvin waves are a necessary element for the generation of Tehuantepec eddies. This would explain why there were many more mountain-gap wind events than resulting eddies. Perhaps the anticyclonic relative vorticity of the high sea level reflecting Kelvin waves preconditions the ocean to spin up eddies more efficiently. This would also explain why the two elements of the solution shown in Fig. 9 have very similar phase. This might also help to explain why only anticyclonic eddies are observed (e.g. McCreary et al 1991).

  10. Model surface height on the equator and at 10°-15°N
    Same as above with Kelvin and Rossby crest lines

    The corresponding picture from the TOPEX altimeter is somewhat less obvious, especially since (as also seen in Figs. 7 and 8 above) the model has two distinct eddies at 95°W-88°W where the altimeter shows one large eddy:

  11. TOPEX surface height on the equator and at 10°-15°N
    Same as above with Kelvin and Rossby crest lines from the model

Some remaining questions:
Why does the amplitude of the eddies grow westward (in both model and altimeter)? This is especially interesting in the model solution with no east Pacific wind variations (Fig. 9, middle panel), where any increase in magnitude must be taking energy from the background state. Is this evidence of a baroclinic instability process within the NEC? (Is there even a NEC this far east?)
Another mystery is why the reflected Rossby wave becomes so concentrated into a single eddy (e.g. Fig. 8), rather than the curved crest that might be expected from equatorial wave theory. In the model, one sees a normal-looking Rossby crest begin to form, but then focus itself into a single growing spot. Does this have something to do with the shape of the coast? Or is it the overlaid forcing of the Tehuantepec and Papagayo winds?

Other useful things ...
(watch this space for coming additions) (requests accepted)

10-day maps of TOPEX sea level anomalies in the region SW of Central America

Weekly maps of Reynolds SST in the same region These SSTs are rather disappointing for the Tehuantepec aficionado, wouldn't you say? I don't see any evidence for the "climatologically important" cooling under the strong winds. For example there are big blasts of gas through Tehuantepec all during Nov-Feb (Figs. 3 and 4), but the coastal region there remains slightly warmer than climatology throughout. Here's a detail example:
Reynolds SST at 14.5°N, 95.5°W during Nov 96 through Mar 97


To make these runs resolving the Pecker eddies, I ran the model with a roughly 20km grid spacing in the eastern region. The CFL condition then required a 30-minute timestep. So I spent a lot of time waiting for the computer to grind through, and filled up a lot of disk space, too. Here's documentation of the grid:
Back to the main Tehuantepecker page
Go to a page showing idealized experiments with Pecker-like bursts, etc, etc
Back to Kessler's home page