Atlantic Meridional Mode (AMM) – Watts Up With That?

By Andy May

The Atlantic Meridional Mode Index (AMM) describes meridional variability in the tropical Atlantic. The area of interest is the ocean area inside 32°N to 21°S and from 75°W to the West African coastline (~15°E). Sometimes the boundaries are given as: 22°S-32°N and 74°W to the West African coast. This is the region where the Intertropical Convergence Zone (ITCZ) exists as it moves north and south with the seasons (Chiang & Vimont, 2004).

There is a possible weak analogue in the Pacific, but the Pacific is so dominated by ENSO that it may not be a real meridional mode related to ITCZ movements (Chiang & Vimont, 2004). Basically, the AMM can be considered unique in the world (Chang, Ji, & Li, 1997).

The AMM region overlaps with the AMO and WHWP regions. The fact that these three oscillations are among the top six that correlate well with HadCRUT5, and together they explain 74% of the variability in HadCRUT5, either means the Atlantic region is very important in determining the global mean surface temperature (GMST) or the measurements of global surface temperature are biased toward the North Atlantic. I suppose both are possible. Figure 1 shows the best regression of the three Atlantic oscillations (AMO, WHWP, and AMM) against HadCRUT5.

The AMM can be expressed either as a wind anomaly or an SST (sea surface temperature) anomaly in its region. Here we will mostly use the SST anomaly, but the wind anomaly is similar, as shown in figure 2. Both full-year and winter anomalies are plotted. The close relationship between wind and SST in this region is easily seen in figure 2.

In general, the negative AMM phase is associated with cool Northern Hemisphere and warm Southern Hemisphere tropical Atlantic anomalies, and the positive phase is the reverse (Patricola, Saravanan, & Chang, 2014). It represents a cross-equatorial SST gradient that flips from negative to positive and vice-versa. It has a strong decadal to multidecadal signal. Annually the SST AMM has a positive peak value in October and its lowest negative value in April. Both the negative and positive modes are associated with shifts in the ITCZ and the associated winds and storms. The AMM is driven by positive wind-evaporation-sea surface temperature feedback where wind anomalies lead SST anomalies by ~2 months (Xia, Zuo, Sun, & Liu, 2023). These air-sea heat fluxes can be explained by thermodynamic and dynamic (aka wind) processes (IPCC, 2021, p. 2168), (Chang, Ji, & i, 1997), and (Patricola, Saravanan, & Chang, 2014).

The AMM to hurricane relationship

The AMM works in concert with ENSO to influence Atlantic tropical cyclone activity as shown in figure 3. There are other factors that influence hurricane activity and may be independent of the AMM and ENSO, at least in part, such as Saharan dust, African easterly waves, and unrelated upper tropospheric temperature, but the AMM and ENSO are very strong influences on Atlantic hurricane activity (Patricola, Saravanan, & Chang, 2014). A neutral or El Niño ENSO state and a negative AMM produce the smallest number of major hurricanes and the smallest ACE (accumulated cyclone energy). The latest current ENSO state (May, 2025) is neutral and the current AMM (May, 2025) is negative, this suggests fewer and weaker Atlantic/Caribbean hurricanes than normal this summer. A strong La Niña and a strongly positive AMM produce the most hurricanes and the largest ACE. The AMM state is a stronger influence on hurricane activity and strength, but the ENSO state matters.

During the positive phase of the AMM, the Atlantic Intertropical Convergence Zone (ITCZ) is displaced northward, often causing drought in Northeast Brazil. Brazilian rainfall is more strongly correlated with the AMM than ENSO (Chang, Ji, & Li, 1997).

A key component of the AMM is the positive feedback between the ocean surface and atmosphere. During a positive phase of the AMM, SSTs become warmer than normal in the tropical North Atlantic and cooler than normal in the tropical South Atlantic. Surface air pressure responds to the SST anomalies, becoming higher than normal over the anomalously cold Southern Hemisphere SSTs and lower than normal over anomalously warm Northern Hemisphere SSTs. The pressure differences influence the wind pattern, which enhances changes in SST. The AMM wind and SST anomalies are mapped for the region in figure 3, which illustrates a positive AMM pattern.

Modeling the AMM

The AMM is strongly correlated to seasonal hurricane activity in the Atlantic on both decadal and interannual time scales (Vimont & Kossin, 2007). Vimont & Kossin tell us that the AMO excites the AMM on decadal time scales and through the AMM has a decadal influence on hurricane activity. The AMM is reasonably well understood, compared to other oscillations. It is also very easily observed in weather data.

We have already discussed how the AMO is only weakly modeled in the IPCC CMIP6 models (IPCC, 2021, p. 504). AR6 also reports that the CMIP6 models of the AMM are “Low Performance” (IPCC, 2021, p. 115). AR6 states that they have:

“… low confidence in projected changes of the Tropical Atlantic Variability (TAV) because of the general failure of climate models to simulate main aspects of this variability such as the northward displaced ITCZ.”

The CMIP5 models did a poor job with the AMM, and with regard to wind, the CMIP6 models do a little better. But both the CMIP5 and CMIP6 models reproduce the SSTs in the AMM region poorly (Xia, Zuo, Sun, & Liu, 2023). This SST problem in the tropics is a general and persistent problem with all climate models (IPCC, 2021, p. 444). The poor observation-model SST match in the models is global, but most obvious in the tropics, especially in the tropical mid-troposphere (McKitrick & Christy, 2020), (McKitrick & Christy, 2018), and (IPCC, 2021, p. 444). The wind/SST processes that drive the AMM are not modeled very well in most of the CMIP6 models, and some would say nearly all the models are poor. Xia et al. call this process the wind-evaporation-SST or “WES” feedback and suggest this is the part of the AMM that CMIP6 models get wrong (Xia, Zuo, Sun, & Liu, 2023). There are a few individual models that do a decent job, for example E3SM1 and ESM1, but the variability in reproducing the AMM in the models is large and inter-model unresolved regional contradictions are common (IPCC, 2021, pp. 1393-1394). More details on the model match to observations, or lack of it, can be seen in Xia, et al. (Xia, Zuo, Sun, & Liu, 2023) and in AR6 WGI, especially section 10.3.3 page 1393, figure 10.6 is brutal. We see in AR6 WGI:

“Model performance varies strongly from model to model, but also between ensembles. These biases are an expression of model error that leads to misrepresented phenomena and processes, and thus limit the confidence in future projections of regional climate.” (AR6, p 1395)

I wish I had written that.

Discussion

A positive AMM with a simultaneous La Niña signals a strong hurricane season, but if the models fail to simulate the AMM, how can they predict hurricane activity? I don’t think we should take model projections of more hurricanes in the future seriously until both the AMO and the AMM can be modeled properly.

About half the CMIP6 models can roughly reproduce the observed spatial pattern of the AMM (see figure 4). Observations show that the wind anomalies lead the SST anomalies by about two months as shown in figure 5. Note the Y axis in figure 5 is not the anomaly itself, but the variance.

Many of the models are successful in reproducing a spring wind peak, but most do not reproduce the very logical SST peak that should follow, or if they do they get the time difference wrong. Some models get it spectacularly wrong, like NorCPM1 and ACCESS-ESM1-5. Xia studied the models that got the WES feedback wrong and found that they had thicker ocean mixed layer depths. The thicker mixed layer led to weaker ocean-atmospheric coupling and a weaker SST response to the wind forcing which screwed up the spacing between the wind and the SST changes. Xia and colleagues were unable to ascertain why some of the models could not even reproduce the spatial pattern of the AMM.

From a physics and thermodynamic perspective, the AMM is one of the simpler and most obvious oscillations and has a regular pattern over the calendar year. One would think it should be simpler to model than the others. The fact that both the CMIP5 and CMIP6 models have such a hard time with it is telling.

One more important point, that I’ve written about before is that it makes little sense to build model ensembles and average them, this has been called Model Democracy (IPCC, 2021, p. 226). This procedure is made worse because AR6 and CMIP6 do not use the same models in each of their “model ensembles.” They sometimes select models based on “performance,” (IPCC, 2021, p. 226) which typically means how well each one compares to observations. Thus, the ensemble approach allows the modelers to pick and choose the results. It is noteworthy that the sensitivity of the results to model selection is “rarely performed” (IPCC, 2021, p. 1425).

One can study multiple models and build criteria to evaluate them, but once you have found the one that best matches observations, there is no need to average its results with inferior models and produce a climate model ensemble. This is especially true if the members of the various ensembles used change.

All models have errors and observation mismatches; results do not improve by averaging various models. Averaging various models runs from the same model with different parameters and initializations makes sense, averaging different semi-independent models does not. All this is especially true with regard to the climate oscillations discussed in this series, especially the AMM. They are well documented climate features and only the models that can reproduce them properly should be considered and the climate model to be used for IPCC evaluations should be the best one, with the closest match to reality.

The next post in this series will be about the North Pacific Index or NPI.

Previous posts in this series:

Musings on the AMO

The Bray Cycle and AMO

Climate Oscillations 1: The Regression

Climate Oscillations 2: The Western Hemisphere Warm Pool (WHWP)

Climate Oscillations 3: Northern Hemisphere Sea Ice Area

Climate Oscillations 4: The Length of Day (LOD)

Climate Oscillations 5: SAM

Chang, P., Ji, L., & Li, H. (1997). A decadal climate variation in the tropical Atlantic Ocean from thermodynamic air-sea interactions. Nature, 385, 516-518. doi:10.1038/385516a0

Chiang, J. C., & Vimont, D. J. (2004). Analogous Pacific and Atlantic Meridional Modes of Tropical Atmosphere–Ocean Variability. Journal of Climate, 17(21), 4143 – 4158. doi:10.1175/JCLI4953.1

IPCC. (2021). Climate Change 2021: The Physical Science Basis. In V. Masson-Delmotte, P. Zhai, A. Pirani, S. L. Connors, C. Péan, S. Berger, . . . B. Zhou (Ed.)., WG1. Retrieved from https://www.ipcc.ch/report/ar6/wg1/

McKitrick, R., & Christy, J. (2018, July 6). A Test of the Tropical 200- to 300-hPa Warming Rate in Climate Models, Earth and Space Science. Earth and Space Science, 5(9), 529-536. doi:10.1029/2018EA000401

McKitrick, R., & Christy, J. (2020). Pervasive Warming Bias in CMIP6 Tropospheric Layers. Earth and Space Science, 7. doi:10.1029/2020EA001281

Patricola, C. M., Saravanan, R., & Chang, P. (2014). The Impact of the El Niño–Southern Oscillation and Atlantic Meridional Mode on Seasonal Atlantic Tropical Cyclone Activity. Journal of Climate, 27(14), 5311-5328. doi:10.1175/JCLI-D-13-00687.1

Vimont, D. J., & Kossin, J. P. (2007). The Atlantic Meridional Mode and hurricane activity. Geophysical Research Letters, 34(7). doi:10.1029/2007GL029683

Xia, F., Zuo, J., Sun, C., & Liu, A. (2023). The Atlantic Meridional Mode and Associated Wind–SST Relationship in the CMIP6 Models. Atmosphere, 14(2). doi:10.3390/atmos14020359

Related