19
boreal summer thus is: Opposing rainfall biases in South
America and Africa produce a zonal surface pressure gradient along the equator that is weaker than in observations. The
resulting winds in the equatorial western Atlantic are too
weak in magnitude and cannot reproduce the observed distribution of upper ocean heat content. Consequently, the seasonally induced equatorial upwelling in early boreal summer
is not sufficient to produce the observed cooling that finally
triggers the Bjerknes feedback.
In agreement with these mechanisms, a number of studies
have found that a physically sound way to reduce the equatorial Atlantic warm bias is to improve the atmospheric models. Tozuka et al. (2011) showed that tweaking the convection
scheme can project strongly on the ability of the models to
simulate the correct distribution of climatological SSTs in
the equatorial Atlantic. Harlaß et al. (2015) conducted a
number of experiments with the KCM that varied both the
horizontal and vertical resolution of the atmospheric GCM,
while keeping a constant coarse resolution for the ocean
GCM. For sufficiently high atmospheric resolutions, the
western equatorial wind bias strongly decreased and the
equatorial Atlantic warm bias nearly vanished. The seasonal
cycle as a whole greatly improved. In a follow-up study,
Harlaß et al. (2017) found that sea level pressure and precipitation gradients along the equator are not sensitive to the
atmospheric resolution. Nevertheless, the wind bias in their
study decreased significantly. To explain this, they propose
that the position of maximum precipitation and zonal
momentum transport play an important role in giving rise to
the zonal wind bias. Zonal momentum can be either transported by mixing it from the free troposphere into the boundary layer or by meridional advection into the western
equatorial Atlantic (Zermeño-Diaz and Zhang 2013; Richter
et al. 2014b, 2017). These findings agree with the study of
Richter et al. (2014a), who found that zonal wind variability
in the western equatorial Atlantic is strongly related to vertical momentum transports in the overlying atmosphere.
Further studies by Voldoire et al. (2014), Wahl et al. (2011),
and DeWitt (2005) confirm the importance of the atmospheric component of a CGCM to properly simulate the
complex tropical Atlantic climate system.
Outlook: Implications for the Usability
of CGCMs in the Equatorial Atlantic
Using the KCM, a CGCM that simulates the tropical Atlantic
in a manner very similar to a wide range of state-of-the-art
CGCMs, we have shown exemplary that coupled global
climate models currently struggle to simulate a realistic
equatorial Atlantic climate system. The dominant feature of
this problem is that CGCMs struggle to simulate the defining
feature of the seasonal cycle – the formation of the Atlantic
cold tongue in early boreal summer. An important cause of
this bias is a strong and seasonally varying westerly wind
bias in equatorial zonal wind in atmospheric models that is
present even in the absence of atmosphere-ocean coupling.
While much progress has been made in understanding and
reducing the equatorial Atlantic warm bias, many models
still produce a profoundly unrealistic seasonal cycle in the
equatorial Atlantic. How does this shortcoming affect the
usefulness of coupled models in the equatorial Atlantic?
A key task of climate models is to forecast deviations
from the expected climate state. For seasonal predictions, the
expected climate state is the climatological seasonal cycle.
Some of these deviations are generated randomly and are, by
definition, unpredictable. Others are the product of – sometimes potentially predictable – climate variability.
In the tropical Atlantic, the dominant mode of year-toyear SST variability is the Atlantic Niño
13
(Zebiak 1993).
The Atlantic Niño is essentially a modulation of the seasonal
formation of the cold tongue (Burls et al. 2012). This modulation can manifest in a range of different cold tongue measures. For example, cold tongue growth might set in earlier
(or later), the cold tongue might cool more strongly, or it
might, in its mature phase, occupy a larger area in the tropical Atlantic than usual. Caniaux et al. (2011) argued that all
of these measures reveal an aspect of cold tongue variability,
but that they do not vary consistently with each other.
Still, the Atlantic Niño is generally described in terms of
Atl3 summer SSTs. While the seasonal cycle of Atl3 SSTs
spans a range of roughly 5 °C, interannual variations of Atl3
SST between May and July rarely exceed amplitudes of 1 °C
(Fig. 10a). The seasonal cycle of the tropical Atlantic is by
far the dominant signal in Atl3 SSTs (Fig. 10b). It is the
background against which the interannual variability of the
Atlantic Niño plays out.
Even though the Atlantic Niño constitutes only a relatively small deviation from the seasonal cycle, its effects on
adjacent rainfall patterns can be substantial (e.g., Giannini
et al. 2003; García-Serrano et al. 2008; Polo et al. 2008;
Rodríguez-Fonseca et al. 2011). A key demand of African
countries, where food security heavily relies on agriculture,
is hence to be able to reliably predict the amplitude of the
Atlantic Niño a few months, ideally even more than a season, ahead. Only such relatively long-ranged forecasts
would allow African farmers to adapt their farming strategy
13 The name “Atlantic Niño” refers to the Pacific El Niño, because the
pattern of Atlantic Niño SST anomalies is similar to the Pacific El Niño.
Apart from this, a number of differences exist between the two phenomena (discussed for example in Keenlyside and Latif (2007), Burls et al.
(2011), Lübbecke and McPhaden (2012), Richter et al. (2013), and
Lübbecke and McPhaden (2017)). Nnamchi et al. (2015, 2016) argued
that the Atlantic Niño might not be dynamical in nature, but a product
of atmospheric noise forcing. Alternative names for the Atlantic Niño
are Atlantic Zonal Mode or Atlantic Cold Tongue Mode.
Can Climate Models Simulate the Observed Strong Summer Surface Cooling in the Equatorial Atlantic?
boreal summer thus is: Opposing rainfall biases in South
America and Africa produce a zonal surface pressure gradient along the equator that is weaker than in observations. The
resulting winds in the equatorial western Atlantic are too
weak in magnitude and cannot reproduce the observed distribution of upper ocean heat content. Consequently, the seasonally induced equatorial upwelling in early boreal summer
is not sufficient to produce the observed cooling that finally
triggers the Bjerknes feedback.
In agreement with these mechanisms, a number of studies
have found that a physically sound way to reduce the equatorial Atlantic warm bias is to improve the atmospheric models. Tozuka et al. (2011) showed that tweaking the convection
scheme can project strongly on the ability of the models to
simulate the correct distribution of climatological SSTs in
the equatorial Atlantic. Harlaß et al. (2015) conducted a
number of experiments with the KCM that varied both the
horizontal and vertical resolution of the atmospheric GCM,
while keeping a constant coarse resolution for the ocean
GCM. For sufficiently high atmospheric resolutions, the
western equatorial wind bias strongly decreased and the
equatorial Atlantic warm bias nearly vanished. The seasonal
cycle as a whole greatly improved. In a follow-up study,
Harlaß et al. (2017) found that sea level pressure and precipitation gradients along the equator are not sensitive to the
atmospheric resolution. Nevertheless, the wind bias in their
study decreased significantly. To explain this, they propose
that the position of maximum precipitation and zonal
momentum transport play an important role in giving rise to
the zonal wind bias. Zonal momentum can be either transported by mixing it from the free troposphere into the boundary layer or by meridional advection into the western
equatorial Atlantic (Zermeño-Diaz and Zhang 2013; Richter
et al. 2014b, 2017). These findings agree with the study of
Richter et al. (2014a), who found that zonal wind variability
in the western equatorial Atlantic is strongly related to vertical momentum transports in the overlying atmosphere.
Further studies by Voldoire et al. (2014), Wahl et al. (2011),
and DeWitt (2005) confirm the importance of the atmospheric component of a CGCM to properly simulate the
complex tropical Atlantic climate system.
Outlook: Implications for the Usability
of CGCMs in the Equatorial Atlantic
Using the KCM, a CGCM that simulates the tropical Atlantic
in a manner very similar to a wide range of state-of-the-art
CGCMs, we have shown exemplary that coupled global
climate models currently struggle to simulate a realistic
equatorial Atlantic climate system. The dominant feature of
this problem is that CGCMs struggle to simulate the defining
feature of the seasonal cycle – the formation of the Atlantic
cold tongue in early boreal summer. An important cause of
this bias is a strong and seasonally varying westerly wind
bias in equatorial zonal wind in atmospheric models that is
present even in the absence of atmosphere-ocean coupling.
While much progress has been made in understanding and
reducing the equatorial Atlantic warm bias, many models
still produce a profoundly unrealistic seasonal cycle in the
equatorial Atlantic. How does this shortcoming affect the
usefulness of coupled models in the equatorial Atlantic?
A key task of climate models is to forecast deviations
from the expected climate state. For seasonal predictions, the
expected climate state is the climatological seasonal cycle.
Some of these deviations are generated randomly and are, by
definition, unpredictable. Others are the product of – sometimes potentially predictable – climate variability.
In the tropical Atlantic, the dominant mode of year-toyear SST variability is the Atlantic Niño
13
(Zebiak 1993).
The Atlantic Niño is essentially a modulation of the seasonal
formation of the cold tongue (Burls et al. 2012). This modulation can manifest in a range of different cold tongue measures. For example, cold tongue growth might set in earlier
(or later), the cold tongue might cool more strongly, or it
might, in its mature phase, occupy a larger area in the tropical Atlantic than usual. Caniaux et al. (2011) argued that all
of these measures reveal an aspect of cold tongue variability,
but that they do not vary consistently with each other.
Still, the Atlantic Niño is generally described in terms of
Atl3 summer SSTs. While the seasonal cycle of Atl3 SSTs
spans a range of roughly 5 °C, interannual variations of Atl3
SST between May and July rarely exceed amplitudes of 1 °C
(Fig. 10a). The seasonal cycle of the tropical Atlantic is by
far the dominant signal in Atl3 SSTs (Fig. 10b). It is the
background against which the interannual variability of the
Atlantic Niño plays out.
Even though the Atlantic Niño constitutes only a relatively small deviation from the seasonal cycle, its effects on
adjacent rainfall patterns can be substantial (e.g., Giannini
et al. 2003; García-Serrano et al. 2008; Polo et al. 2008;
Rodríguez-Fonseca et al. 2011). A key demand of African
countries, where food security heavily relies on agriculture,
is hence to be able to reliably predict the amplitude of the
Atlantic Niño a few months, ideally even more than a season, ahead. Only such relatively long-ranged forecasts
would allow African farmers to adapt their farming strategy
13 The name “Atlantic Niño” refers to the Pacific El Niño, because the
pattern of Atlantic Niño SST anomalies is similar to the Pacific El Niño.
Apart from this, a number of differences exist between the two phenomena (discussed for example in Keenlyside and Latif (2007), Burls et al.
(2011), Lübbecke and McPhaden (2012), Richter et al. (2013), and
Lübbecke and McPhaden (2017)). Nnamchi et al. (2015, 2016) argued
that the Atlantic Niño might not be dynamical in nature, but a product
of atmospheric noise forcing. Alternative names for the Atlantic Niño
are Atlantic Zonal Mode or Atlantic Cold Tongue Mode.
Can Climate Models Simulate the Observed Strong Summer Surface Cooling in the Equatorial Atlantic?