The tropical precipitation response to precessional forcing
The general picture that emerges from the, essentially terrestrial, proxy-estimates shown in Fig. 1 is that maximum precipitation in the tropics follows local maximum summer insolation. When considering the annual, zonal mean change in net precipitation in our model simulations (Fig. 4a) a similar view emerges: in the \(\text {P}_{\text {MAX}}\) run the tropical precipitation maximum shifts towards the Southern Hemisphere, which has a stronger SH seasonal cycle than the \(\text {P}_{\text {MIN}}\) run. This is true for precipitation over land as well as over the ocean. While the oceanic precipitation change is mostly confined between 15°S and 15°N, and switches sign roughly at the equator, continental precipitation changes have a larger meridional extent due to the strong contributions of landmasses located away from the equator (e.g. Australia, South Africa, the Himalayas) and (in the zonal mean) switch from positive to negative around 10°S.
A better understanding of regional processes is obtained by studying the spatial distribution of the annual mean precessional precipitation changes (Fig. 4b). The regional hydroclimate responses are clearly much more complex than just the warmer summer hemisphere (SH, for \(\text {P}_{\text {MAX}}\)) becoming wetter. We find a small intensification of the SH monsoon systems (South America, Southern Africa and Australia), and a decrease in precipitation over the South and East Asian monsoon regions. The annual mean ITCZ over the Atlantic shifts towards the warmer summer hemisphere as well. But over the Pacific, there is an increase in precipitation over the northern equatorial Pacific (dateline to 110°W) in addition to a strong intensification south of the equator. The latter appears to be related to a reduction in the tilt of the South Pacific Convergence Zone. The wettening of the central Pacific is contrasted zonally with a drying of the Western Warm Pool region. Over the Indian Ocean there is a dipole pattern with a meridionally slanted zero-line. Finally there is a strong increase in annual mean precipitation over the Andaman and South China Seas. All together, Fig. 4b shows that the simple paradigm of precipitation shifting towards the warmer summer hemisphere (SH in \(\text {P}_{\text {MAX}}\))—although valid in a zonal mean sense—does not apply universally throughout the tropics. Zonal mean frameworks of understanding the ITCZ, such as the ones discussed by Schneider (1977), Lindzen and Hou (1988), Kang et al. (2009) and Schneider et al. (2014), will therefore not be able to explain the regional changes in tropical precipitation.
The Atlantic ITCZ
Position versus intensity
The differential response across the tropics of precipitation to precessional forcing suggests that regional dynamics play an important role, and it will not be possible to find a universal mechanism driving this response. In the remainder of this paper we will therefore focus our attention on one specific region, the Atlantic ITCZ. Throughout the literature, different definitions of the ITCZ are used; here we take ITCZ to refer to either the location of maximum precipitation or the location of zero meridional wind (confluence line).
Despite an annual mean insolation profile that is symmetric around the equator, the present-day climatological annual mean ITCZ over the Atlantic Ocean is displaced into the Northern Hemisphere (e.g. Mitchell and Wallace 1992). While the question of the hemispheric preference of the ITCZ has not been answered definitively, it has been attributed to the existence of the AMOC (e.g. Krebs and Timmermann 2007; Schneider et al. 2014) as well as the asymmetry of the continents surrounding the Atlantic Ocean and its resulting ocean-atmosphere interactions (Xie 1996, 2005; Xie and Saito 2001). Seasonally the Atlantic ITCZ migrates from a position close to the equator in boreal spring to a maximum northward position around 10°N in late summer (e.g Mitchell and Wallace 1992). This seasonal march of the ITCZ has been found to be sensitive to remote climate forcing on various timescales, such as the El Niño-Southern Oscillation (Chiang et al. 2002), the North Atlantic Oscillation (Marshall et al. 2001), Northern Hemisphere ice conditions (Chiang et al. 2003; Chiang and Bitz 2005; Broccoli et al. 2006), the Atlantic meridional mode (Nobre and Shukla 1996; Chiang et al. 2002; Xie and Carton 2004), and meltwater pulses and AMOC (Broccoli et al. 2006; Timmermann et al. 2007; Deplazes et al. 2013; Marzin et al. 2013; Menviel et al. 2014).
As noted in Sect. 3.1, the annual mean Atlantic ITCZ responds to precession by shifting towards the warmer summer hemisphere (Fig. 4b), with more precipitation under \(\text {P}_{\text {MAX}}\) forcing occurring in the south, and less in the north. The resulting dipole is not centered on the equator, nor is it symmetric around its axis. Rather, the increase in precipitation on the southern side of the dipole is stronger than the decrease on its northern side. In the \(\text {P}_{\text {MAX}}\) run, total annual integrated precipitation over the Atlantic Ocean between 20°S and 20°N increases by \(2.28\times 10^{12}\hbox { m}^3\), or 10 % (Table 2). For comparison, global annual integrated precipitation increases by \(2.21\times 10^{12}\hbox { m}^3\), or 0.5 %
In the context of the precessional forcing, which has a strong seasonal component but no annual one (Fig. 2b), the changes in annual mean precipitation could derive from two components: A change in the seasonal march of the ITCZ (amplitude/position change), or a change in when precipitation occurs along this march (phase change). Figure 5 shows monthly mean precipitation and position of the Atlantic ITCZ as measured by the location of zero meridional wind, averaged between 40°W and 20°W (This longitude range was chosen to isolate the oceanic ITCZ response). In boreal spring and summer, when the \(\text {P}_{\text {MIN}}\) insolation in both hemispheres exceeds that of \(\text {P}_{\text {MAX}}\) (see Fig. 2b), the ITCZ travels further north, and at an earlier time of year, in the \(\text {P}_{\text {MIN}}\) simulation (Fig. 5b) than in the \(\text {P}_{\text {MAX}}\) simulation (Fig. 5a). In boreal fall and winter, when the \(\text {P}_{\text {MAX}}\) insolation exceeds that of \(\text {P}_{\text {MIN}}\), the reverse does not happen however: The \(\text {P}_{\text {MAX}}\) and \(\text {P}_{\text {MIN}}\) ITCZs reach approximately the same southward extent. This change in seasonal march of the confluence line—even without a change in seasonality—implies a northward shift of the annual mean Atlantic ITCZ in \(\text {P}_{\text {MIN}}\) compared to \(\text {P}_{\text {MAX}}\).
In addition to a change in meridional position, there is a change in phase of the precipitation: Under maximum precession (Fig. 5a), most of the precipitation over the tropical Atlantic falls between April and July, while under minimum precession (Fig. 5b) there are two precipitation maxima—a brief one in July and a more extended one between September and December. The oceanic precipitation maxima (Fig. 5a, b) thus occur several months after maximum TOA insolation (Fig. 2a).
Different mechanisms are thought to be responsible for driving changes in position and intensity of oceanic moisture convergence in the tropics. Changes in meridional position are caused by a mechanism originally proposed by Lindzen and Nigam (1987), relating convergence in the boundary layer to surface pressure gradients caused by meridional (cross-equatorial) SST gradients (see also Tomas et al. 1999). We will refer to this process as the ‘boundary layer mechanism’. An initial meridional displacement can be further amplified by a process known as the Wind-Evaporation-SST (WES) feedback (Xie and Philander 1994): changes in low-level convergence will affect latent heat fluxes and can feed back positively onto the anomalous SST gradient. Numerous studies have shown that the WES feedback is an important amplifier of meridional ITCZ shifts over the Atlantic Ocean (e.g. Braconnot et al. 2000; Biasutti et al. 2005; Zhao et al. 2005; Bosmans et al. 2012).
On the other hand, Gill (1980) and subsequent studies (e.g. Chiang et al. 2001; Biasutti et al. 2004) looked at the entire troposphere, and found that local diabatic forcing can change the stability profile of surrounding areas and drive anomalous adiabatic vertical motion elsewhere, resulting in divergence changes. We will refer to these processes as the ‘full troposphere mechanism’. Biasutti et al. (2004, 2005) showed that the precipitation response to remote diabatic forcing depends heavily on the background stability of the troposphere. Changes in diabatic heating are most likely to induce remote changes in regions where there is already precipitation occurring. The full troposphere mechanism therefore mostly impacts the intensity of precipitation.
In the following sections we more closely evaluate the precipitation response in individual seasons to see what drives precessional changes in the intensity and meridional position of the ITCZ. The main results are summarized schematically in Fig. 6. We argue that the precessional precipitation response over the Atlantic Ocean is a direct result of continental forcing. The continental response to precessional forcing, and the oceanic response to these continental changes, vary in time and space and thus rectify the annual mean change of the Atlantic ITCZ.
The JJA ITCZ
In JJA, precipitation over the Atlantic Ocean in \(\text {P}_{\text {MAX}}\) increases between 5°S–5°N, while it decreases between 5°N and 10°N (Fig. 7a, contours). These simulated changes in precipitation could be driven by temperature-induced changes in specific humidity (‘thermodynamics’), by changes in the mean circulation (‘dynamics’), by changes in transient eddy moisture convergence or by changes in evaporation. While Merlis et al. (2013) in an aquaplanet simulation attributed precessionally induced changes in the ITCZ mostly to thermodynamic processes, a decomposition of the moisture budget in our simulations (along the lines of e.g. Trenberth and Guillemot 1995; Clement et al. 2004; Held and Soden 2006; Seager et al. 2010) indicates that the dominant contribution to JJA precipitation changes over the Atlantic are changes in the mean circulation, in particular its convergence (see “Appendix”). We suggest that in this season, the meridional position of the Atlantic ITCZ is impacted by surface temperature changes over Africa, while its intensity is affected by changes in the African summer monsoon.
Between March and August there is a \(\text {P}_{\text {MAX}}\)-induced anomalous reduction in TOA shortwave radiation throughout the entire tropics (Fig. 2b). Figure 7a shows the resulting changes in JJA 2m temperature and precipitation. Over northern Africa (20°N), where climatological soil moisture is low, the reduction in insolation leads to strong continental cooling. South of 20°N, it causes a weakening of the climatological African summer monsoon and equatorial precipitation: total precipitation over the African continent in \(\text {P}_{\text {MAX}}\) is reduced by 35 % in MAM and 81 % in JJA (see Table 2; Fig. 7a). This weakening is accompanied by a reduction in cloudiness (Fig. 7b) and evaporative cooling that creates a net surface forcing opposite in sign to the original TOA forcing. This results in a reduction of the \(\text {P}_{\text {MAX}}\) cooling between 10°S–10°N, and even a slight warming between 10°N–20°N (Fig. 7a).
As shown by Chiang et al. (2001), the North Atlantic trade winds strongly respond to remote influences such as ENSO and continental forcing (unlike the Atlantic cross-equatorial winds, which are very sensitive to changes in the cross-equatorial SST gradient). The reduction of continental temperatures over northern Africa drives wind changes that advect dry air southward, and strengthen the North Atlantic trades in \(\text {P}_{\text {MAX}}\) (Fig. 7b). This intensification of the trade winds sets up an anomalous meridional SST gradient that alters the meridional position of the Atlantic ITCZ through the boundary layer mechanism.
Figure 8 shows the \(\text {P}_{\text {MAX}}\) and \(\text {P}_{\text {MIN}}\) zonal mean Atlantic SSTs and their difference, averaged between 40°W and 20°W. Generally the maximum precipitation, maximum SST and the confluence line are closely co-located, with the notable exception of \(\text {P}_{\text {MIN}}\) JJA, when the precipitation maximum is located in between the confluence line and SST maximum (Figs. 5, 8). Previous work has indicated that SST and precipitation maxima do not co-locate when meridional pressure gradients play an important role in driving moisture convergence (e.g. Tomas and Webster 1997). Indeed, Fig. 8c indicates that the anomalous northward ITCZ shift in \(\text {P}_{\text {MIN}}\) JJA coincides with the onset of an anomalous meridional SST gradient.
The role of the strengthening trade winds in driving the onset of this anomalous meridional SST gradient can be illustrated through a calculation of the mixed layer heat budget. The time evolution of the vertically integrated mixed layer temperature is given by the SST equation:
$$\begin{aligned} c_p\rho _0H\frac{\partial T}{\partial t}&= D_o + Q_{\text {net}}\nonumber \\&= D_o + Q_{\text {SW}} + Q_{\text {LW}} + Q_{\text {LH}} + Q_{\text {SH}}. \end{aligned}$$
(7)
Here T is SST, \(c_p\) is the specific heat capacity at constant pressure, \(\rho _0\) the density of seawater, and H the seasonally and spatially varying mixed layer depth. \(Q_{\text {SW}}\), \(Q_{\text {LW}}\), \(Q_{\text {LH}}\), and \(Q_{\text {SH}}\) are the surface heat fluxes from shortwave and longwave radiation, latent heat and sensible heat respectively. \(D_o\) represents the changes in mixed layer temperature due to ocean heat transport (three-dimensional advection and mixing). We can calculate the change in mixed layer heating rate from
$$\begin{aligned} \delta \left( c_p\rho _0H\frac{\partial T}{\partial t} \right)&= \delta Q_{\text {SW}} + \delta Q_{\text {LW}} + \delta Q_{\text {LH}} \nonumber \\&\quad + \delta Q_{\text {SH}} + \delta D_o , \end{aligned}$$
(8)
which after some manipulation becomes
$$\begin{aligned} \delta \frac{\partial T}{\partial t} =&\frac{1}{c_p\rho _0}\frac{1}{H_{{\text {P}_{\text {MAX}}}} } \left\{ \delta Q_{\text {SW}} + \delta Q_{\text {LW}} + \delta Q_{\text {LH}} \right. \nonumber \\&\left. + \,\,\delta Q_{\text {SH}} + \,\delta D_o \right\} - \frac{\delta H}{H_{{\text {P}_{\text {MAX}}}}} \left( \frac{\partial T}{\partial t}\right) _{{\text {P}_{\text {MIN}}}}, \end{aligned}$$
(9)
where \(\delta\) is the difference operator \(\text {P}_{\text {MAX}}\)–\(\text {P}_{\text {MIN}}\).
Figure 9 shows these various contributions to the change in heating rate. Note that the negative latent heat and sensible heat flux changes are plotted so that blue and red correspond to cooling and heating respectively. The ocean heat transport term in Fig. 9f was calculated as the residual from the surface heat fluxes and the \(c_p\rho _0H\frac{\partial T}{\partial t}\) term. Over the ocean, there is generally a 1-2 month lag between insolation forcing and surface temperature changes. In order to understand the onset of the anomalous SST gradient in May, as seen in Fig. 8c, we therefore need to understand the contributions to \(\delta \frac{\partial T}{\partial t}\) in March (referenced by the solid green line in Fig. 9).
The TOA insolation forcing shown in Fig. 2b shows little meridional gradient in boreal spring. At the surface, net shortwave radiation \(\delta Q_{SW}\) (Fig. 9b) shows a strong cooling between 20°S and 10°N which is due to a \(\text {P}_{\text {MAX}}\) increase in cloudiness over these latitudes (Fig. 7b) following the precipitation changes of Fig. 5c. By contrast, the total anomalous heating rate plotted in Fig. 9a shows strongest cooling in boreal spring over the North Atlantic. This cooling is the result of an anomalous \(\text {P}_{\text {MAX}}\) increase in the surface latent heat flux (Fig. 9d) which cools the North Atlantic starting in March, while a decrease in latent heat flux warms the Atlantic just south of the equator. Throughout boreal spring and summer, the changes in latent heat flux continue to warm the Atlantic south of 5°N and cool it north of that in \(\text {P}_{\text {MAX}}\). The dipole in latent heat flux changes is formed by the \(\text {P}_{\text {MAX}}\) strengthening of the trade winds over the North Atlantic, and a weakening of the cross-equatorial winds in boreal spring and summer (Fig. 7b and contours in Fig. 9d). It is a clear indication that the WES feedback is at work: The \(\text {P}_{\text {MAX}}\) cooling of the North Atlantic caused by the boreal spring strengthening of the trades reduces the meridional SST gradient, which keeps the region of convergence close to the equator, weakening the cross-equatorial winds, and warming the equatorial region (Fig. 9d). This warming further reduces the meridional SST gradient (Fig. 8c) and keeps the ITCZ south. The opposite is the case in the \(\text {P}_{\text {MIN}}\) simulation, where an increased meridional SST gradient pushes the Atlantic ITCZ north (Fig. 5c).
Coincident with the latent heat flux anomalies, changes in ocean heat transport warm the Atlantic south of 5°N in \(\text {P}_{\text {MAX}}\) (Fig. 9f). They further contribute to the \(\text {P}_{\text {MAX}}\) weakening of the meridional SST gradient in boreal spring and summer. The longwave, sensible heat, and mixed layer terms of Eq. (9), shown in Fig. 9c, e, and g respectively, contribute little to these heating rate changes. These results thus indicate that through the boundary layer mechanism and amplified by the WES feedback, precessional changes in African temperature and resulting surface wind changes alter the meridional position of the Atlantic ITCZ. While the entire South American continent cools in response to the \(\text {P}_{\text {MAX}}\) forcing (Fig. 7a), this cooling does not induce surface wind changes that can impact the ITCZ position (Fig. 7b).
Figure 10 shows JJA cross-sections of vertical velocity changes averaged over various latitude bands. Between 15°S and 5°N, vertical velocity changes over the ocean are opposite in sign to those over the continents (esp. Africa), showing that convection over the ocean decreases when it increases over land. The vertical profile of change in this area is very different from that north of 5°N, where over the Atlantic Ocean (50°W–20°W) the response is mostly limited to the boundary layer. Contours in Fig. 10 show potential temperature anomalies with respect to vertical mean changes. Throughout the tropics, the upper troposphere cools compared to the lower troposphere, indicating reduced stability. Together, the vertical velocity and potential temperature changes suggest that the mechanisms described by Biasutti et al. (2004) are at work: the decrease in continental precipitation cools the upper troposphere due to a reduction in diabatic heating, which then reduces the large-scale stability profile and increases convection in areas where the background stability is weak. Between 5°S and 5°N, the diabatic heating (not shown) resulting from the continentally-driven increase in oceanic precipitation partially offsets the large-scale cooling in the upper troposphere. Because climatological precipitation over South America is low in this season, and precessional changes in precipitation are small (Fig. 7a, Table 2), continental changes on the western side of the Atlantic hardly impact the Atlantic ITCZ.
The mechanisms described above are illustrated in the top panel of Fig. 6. The position of the Atlantic ITCZ is affected by \(\text {P}_{\text {MAX}}\) continental cooling over northern Africa, which increases the North Atlantic trades, and through the WES-feedback and wind-induced changes in ocean heat transport reduces the cross-equatorial SST gradient, keeping the \(\text {P}_{\text {MAX}}\) ITCZ south. A \(\text {P}_{\text {MAX}}\) weakening of the African summer monsoon affects the large-scale stratification profile by cooling the upper troposphere, increasing precipitation in areas that are already unstable, thus changing the intensity of the Atlantic ITCZ. As we shall see in the next section, the magnitude of the processes described above is not seasonally symmetric, thus rectifying the precipitation response in the annual mean.
The DJF ITCZ
The response of the Atlantic ITCZ to DJF forcing is markedly different from the JJA changes described above. As seen in Fig. 5, the \(\text {P}_{\text {MAX}}\) rainfall maximum attains the same southernmost position as that of \(\text {P}_{\text {MIN}}\). Figure 11a shows that in DJF \(\text {P}_{\text {MAX}}\) precipitation decreases over much of the Atlantic, except close to the African continent. The magnitude of the DJF precipitation response is smaller than that in JJA. An analysis of the DJF moisture budget components indicates that again, the precipitation changes over the ocean are dominated by changes in wind convergence rather than specific humidity (“Appendix”, Fig. 15). Our coupled model simulations suggest that in this season, changes in continental temperature fail to alter trade winds significantly (Fig. 11) and initiate the WES feedback. Changes in precipitation over both Africa and South America affect the intensity of precipitation over the Atlantic but not its location.
The TOA \(\text {P}_{\text {MAX}}\)–\(\text {P}_{\text {MIN}}\) shortwave forcing in boreal winter (Fig. 2b) is almost the exact mirror image of the boreal summer forcing. From approximately October to March, incoming shortwave radiation in \(\text {P}_{\text {MAX}}\) exceeds that in \(\text {P}_{\text {MIN}}\) throughout the entire tropics, with the strongest forcing over the Southern Hemisphere. In response to this anomalous \(\text {P}_{\text {MAX}}\) increase in insolation in boreal fall and winter (Fig. 2b), the entire African continent warms up, as shown in Fig. 11a for DJF. Between 10°N–20°N the increase in insolation is amplified by an increase in downward longwave radiation (not shown). South of 10°N, the increase in shortwave radiation leads to an increase in precipitation over the African continent (by as much as 86 % in DJF (Table 2)). As for JJA, precipitation changes over continental Africa in DJF mostly derive from changes in tropospheric convergence, with smaller contributions from evaporation changes and, over southern Africa, a nonlinear and surface term (Fig. 15; see explanation in “Appendix”). The resulting increase in cloudiness (Fig. 11b) and evaporative cooling partially (but not entirely) offset the expected warming (compare Figs. 7a, 11a).
The positive \(\text {P}_{\text {MAX}}\) shortwave forcing enhances the austral summer monsoon over South America. Total precipitation over the South American continent increases by 36 % in DJF (Table 2). The increase in precipitation largely derives from an increase in wind convergence, but is partially offset by a decrease in moisture advection as increased evaporation over the continent decreases the land-ocean humidity gradient (see “Appendix”, Fig. 15). The entire South American continent warms in response to the positive shortwave forcing (Fig. 11a), although here as well, continental warming is limited by increased cloudiness (Fig. 11b) and latent heat flux.
Even though the \(\text {P}_{\text {MAX}}\) increase in TOA insolation is strongest over the Southern Hemisphere (Fig. 2b), continental warming is most pronounced over northern Africa (Fig. 11a). This warming causes a thermal low that slightly weakens the \(\text {P}_{\text {MAX}}\) North Atlantic trades (Fig. 11b). As evidenced by the heat flux analysis in Fig. 9, this weakening leads to a reduction in oceanic latent heat flux (Fig. 9d) and shallowing of the mixed layer (Fig. 9f) that warm the North Atlantic. This keeps the \(\text {P}_{\text {MAX}}\) ITCZ north (Fig. 5) at a time when the climatological ITCZ is moving south. At the same time, the southern tropical Atlantic is warming (Fig. 9a) due to the \(\text {P}_{\text {MAX}}\) increase in shortwave radiation (Fig. 9b). As a result, no meridional heating gradient develops, and the WES-feedback cannot enhance the meridional shift; between December and May the entire tropical Atlantic warms up (Fig. 8c). The overall warming is opposed by ocean heat transport changes (Fig. 9e) that cool the equatorial and South Atlantic. Changes in longwave radiation and sensible heat flux (Fig. 9c, e respectively) again play a negligible role. The boundary layer mechanism that was shown to be essential for the JJA meridional ITCZ shift is thus of minor importance in DJF. This is also shown in Fig. 12 which shows that there is very little boundary layer response in this season, indicating that a full troposphere response to diabatic heating dominates over the boundary layer mechanism.
Figure 12 shows DJF cross-sections of vertical velocity changes in various latitude bands over the tropical Atlantic. Much like in JJA, vertical velocity changes between 15°S and 5°N over the Atlantic ocean are opposite in sign to those over the continents. The strengthening of the austral summer (DJF) South American and African monsoon systems under maximum precession is associated with increased upward motion over land, in particular over Africa between 5°S and 5°N (Fig. 12). The associated convective heating (not shown) warms the upper troposphere (contours in Fig. 12), making it more stable and reducing convection over the ocean in \(\text {P}_{\text {MAX}}\) (Figs. 5, 11a). Biasutti et al. (2004) showed that convection changes over South America and southern Africa play comparable roles in driving precipitation anomalies in the Atlantic ITCZ. The precipitation changes over the South American and African continents in DJF are smaller than those in JJA (Table 2; Figs. 7a, 11a) so the full troposphere response in this season is weaker as well (compare Figs. 10, 12).
Even though the DJF \(\text {P}_{\text {MAX}}\)–\(\text {P}_{\text {MIN}}\) change in TOA shortwave forcing is almost the exact mirror image of the JJA forcing (Fig. 2b), the ITCZ response to this forcing is markedly different. The mechanisms described in this section are summarized in the bottom panel of Fig. 6. Despite stronger shortwave forcing in the Southern Hemisphere, the strongest continental heating is found over Northern Africa, leading to a slight weakening of the North Atlantic trades and warming of the North Atlantic. Because the South Atlantic is warming at the same time, meridional SST gradients are muted and the WES feedback does not operate. The net effect is that the position of the ITCZ does not change. Atlantic precipitation is mostly impacted by an increase in continental precipitation, which through diabatic heating and resulting changes in large-scale stratification decreases the intensity of oceanic precipitation.