Geophysical Research Letters

Physical mechanism of long-term drying trend over tropical North Africa

Authors

Errata

This article is corrected by:

  1. Errata: Correction to “Physical mechanism of long-term drying trend over tropical North Africa” Volume 37, Issue 21, Article first published online: 11 November 2010

Abstract

[1] Based on an approximated moisture budget equation, we investigate the physical mechanisms of a drying trend observed over tropical North Africa in the boreal summer during the 20th Century by analyzing datasets of several climate-model experiments forced with various combinations of natural and anthropogenic forcings. Increased anthropogenic aerosols thermodynamically induce a drying trend due to a tropospheric cooling and dynamically induce an additional drying trend due to an atmospheric local circulation change stirred up by the strong gradient of a sea surface temperature anomaly over the tropical Atlantic Ocean. Increased greenhouse gases, on the other hand, induce a drying trend through the large-scale dynamic effect, which is canceled out by the thermodynamically induced moistening trend due to tropospheric warming. Therefore, the drying trend observed over tropical North Africa during the 20th Century is strongly affected by the increased anthropogenic aerosols through both the dynamic and thermodynamic effects.

1. Introduction

[2] During the 20th Century, a drying trend was observed over the low latitude in the northern hemisphere (NH), while moistening trends were observed over the high latitude of NH and the low latitude in the southern hemisphere (SH) [Intergovernmental Panel on Climate Change (IPCC), 2007; Zhang et al., 2007]. By analyzing the observed precipitation and multi-model simulations statistically, Zhang et al. [2007] concluded that anthropogenic forcings had a detectable influence on the observed changes in zonal mean precipitation within latitudinal bands.

[3] The drying trend over the equatorial region of Africa (hereafter, tropical Africa) was one of the pronounced climatic changes in precipitation over the low latitude in the NH [e.g., Chappell and Agnew, 2008; Ting et al., 2009]. It has been pointed out that the African drying trend was directly attributable to changes in the sea surface temperature (SST) field. For instance, Giannini et al. [2003] suggested that the recent drying trend in the semi-arid Sahel was the results of a warmer-than-average low-latitude SST around Africa. Hoerling et al. [2006] indicated that the relatively larger cooling of the tropical North Atlantic Ocean than that of the tropical South Atlantic Ocean was the key driver of the drought over the Sahel.

[4] The individual impact of increasing aerosols and greenhouse gases (GHGs) on the drying trend in Africa has been identified by several sensitivity experiments with General Circulation Models (GCMs) [e.g., Rotstayn and Lohmann, 2002; Paeth and Feichter, 2006]. Paeth and Feichter [2006] reported that, in general, GHG forcing was causing most of Africa to become warmer and more humid, while the increased aerosols had the opposite effect. They also suggested that the impact of the increased aerosols on precipitation was larger than that of the increasing GHGs in Africa during the 20th Century. Rotstayn et al. [2000] and Takemura et al. [2005] suggested that the effect of the increased aerosols could cause a southward shift of the equatorial precipitation, which might be responsible for the drying trend over tropical Africa. Several factors affecting the African drying trend have been proposed in these previous studies, but the detailed mechanism has not been completely clarified.

[5] Here, we investigate the impacts of anthropogenic forcing on the long-term drying trend over the northern part of tropical Africa (hereafter, tropical NAF) using 20th Century simulations performed by a Coupled GCM (CGCM) generally known as the medium-resolution version of the Model for Interdisciplinary Research On Climate (MIROC) [Hasumi and Emori, 2004]. The atmospheric component of MIROC includes an explicit representation of the first and second kinds of indirect effects induced by soluble aerosols as well as the direct effects of all aerosols. We look into the relative contribution of individual anthropogenic forcing factors by analyzing datasets of several experiments forced with different combinations of external climate forcing factors. According to an analyzing method of Chou et al. [2009], we also diagnose the dynamic and thermodynamic terms in an approximated moisture budget equation to clarify the detailed physical mechanism of the precipitation changes.

2. Observed Precipitation Trend

[6] Table 1 shows the observed long-term precipitation trends for the tropical NAF. To confirm the reliability of the trend, we used three gridded observational datasets, which were derived from the Climate Research Unit (CRU) [New et al., 2000], the Global Historical Climatology Network (GHCN) [Vose et al., 1992], and the Global Precipitation Climatology Centre (GPCC) [Rudolf and Schneider, 2005]. The regional mean precipitation was calculated within the rectangular region (EQUATION-20N, 20W–50E) shown in Figure 1a. Decreasing trends of annual mean precipitation are observed over the tropical NAF by all the datasets; they are significant at the 99% level according to the Student's t-test. The drying trends are prominent in the boreal summer (hereafter summer). Here, we focus our analysis on summer. The time series of the regional mean summer precipitation show clear drying trends in the 20th Century (Figure 1e). Note that some decadal variability was found in the observational datasets.

Figure 1.

(a–b) Distributions of climatological summer precipitation averaged from 1961 to 1990, (c–d) linear trends of precipitation from 1902 to 1998, and (e–f) time series of the summer precipitation anomaly with respect to the climatology from 1961 to 1990 averaged over the land within the rectangular region shown in Figures 1a–1d. The left side shows the observations, and the right, the FULL experiments. In Figure 1e the CRU, the GHCN, and the GPCC are shown by red, green, and blue dots and lines, respectively. In Figure 1f shading represents the range between the maximum and minimum values in 4 ensemble members.

Table 1. Annual and Seasonal Mean Precipitation Trend Over Tropical North Africa, 1902–1998a
 CRUGHCNGPCC
  • a

    Values are (mm/day) per century. All trends are calculated from gridded data with 5.0° × 5.0° resolution. Boldface and italics represent 99% and 95% significant levels, respectively, according to the Student's t-test. DJF: December–January–February, MAM: March–April–May, JJA: June–July–August, and SON: September–October–November.

Annual0.2120.2720.190
DJF0.093−0.0430.077
MAM−0.0990.1510.125
JJA0.4300.5700.364
SON−0.1810.2890.195

3. Experimental Design

[7] Table 2 summarizes the combinations of external forcing factors used in all the experiments with MIROC. Details of each forcing factor are given by Nozawa et al. [2005]. In the basic experiment (referred to as FULL), the simulations were forced with both natural (changes in solar irradiance and stratospheric volcanic aerosols) and anthropogenic (changes in well-mixed GHGs, sulfate and carbonaceous aerosols, stratospheric and tropospheric ozone, and land use) forcing factors. We also conducted an experiment forced with natural forcing only (NTRL) and another one forced with anthropogenic forcing only (ANTH). Furthermore, we performed three additional experiments with individual anthropogenic forcing; one forced with increases in the concentrations of well-mixed GHGs only (GHGS), one with changes in the emissions of anthropogenic sulfate and carbonaceous aerosols only (ARSL), and one with changes in the concentrations of stratospheric and tropospheric ozone only (OZON). Each experiment had four ensemble runs with different initial conditions derived from a long-term control simulation under the pre-industrial condition. We analyzed the average of four ensemble members for each experiment.

Table 2. Forcing Factors Used in the Experiments Analyzed in This Studya
 Natural Forcing (Solar + Volcano)Anthropogenic Forcing
Aerosols (Sulfate + Carbon)Greenhouse GasesOzone
  • a

    The letter ‘x’ denotes the inclusion of a specific forcing factor.

FULLxxxx
NTRLx   
ANTH xxx
ARSL x  
GHGS  x 
OZON   x

4. Results and Discussion

4.1. Simulated Precipitation Trend

[8] The FULL experiment adequately simulates the distribution of summer precipitation derived from the CRU observation over the tropical NAF (Figures 1a1b). The large precipitation is realistically captured as a part of the Intertropical Convergence Zone (ITCZ) over the Atlantic Ocean. The drying trend is simulated over the tropical NAF, as shown in the CRU observations (Figures 1c1d). Moreover, the time series of the regional mean summer precipitation show a similar drying trend to those from the observations (Figures 1e1f). The magnitude of the simulated dying trend is, however, smaller than that of the observational datasets (Tables 1 and 3). The amplitude of interannual variation in the FULL is also comparable to that in the observations.

Table 3. Same As Table 1 Except the Annual and Boreal Summer Precipitation Trends Are Simulated By the Experiments Listed in Table 2
 FULLNTRLANTHARSLGHGSOZON
Annual0.194−0.0150.1720.2740.067−0.053
JJA0.2970.0800.3100.4000.050−0.062

[9] The ANTH and ARSL experiments simulate a significant drying trend in summer, as the FULL does, but the NTRL, GHGS, and OZON do not (Table 3). In other words, the experiments with historical changes in anthropogenic aerosols can simulate the significant drying trend. This result suggests that the historical changes in anthropogenic aerosols play an important role in the drying trend over the tropical NAF in the 20th Century.

4.2. Diagnosis of Terms in the Moisture Budget Equation

[10] The changes in GHGs and aerosols largely affected the climate changes in the 20th Century [IPCC, 2007], although the impact of the increasing GHGs on the African precipitation trend was small in our model (Table 3). To obtain a deeper insight into the physical mechanisms of the drying trend over the tropical NAF, we investigate the detailed impacts of the individual anthropogenic forcing factors on the precipitation trend with a focus on the dynamic and thermodynamic effects on the precipitation change. The dynamic and thermodynamic effects were diagnosed by the following approximated moisture budget equation by Chou et al. [2009]:

equation image
equation image

where 〈 〉 denotes a vertical mass integration through the troposphere, equation image means the climatology, and ( )′ means the departure from the climatology. P, E, ω, v, and q represent the precipitation, evaporation, vertical velocity, horizontal velocity, and specific humidity, respectively. The first term on the right-hand side of equation (2) is the dynamic effect associated with anomalous circulation ω′ (hereafter, a dynamic term), and the second term shows a thermodynamic effect associated with anomalous moisture q′ (hereafter, a thermodynamic term). The third and fourth terms show anomalous horizontal moisture transport and anomalous evaporation, respectively. More detailed descriptions and discussions are shown by Chou and Neelin [2004] and Chou et al. [2009]. In estimating each term, we use the difference in the variables between the early 20th Century (1901–1920) and the late 20th Century (1981–2000) for their anomaly. The climatology is defined as an average in the early 20th Century.

[11] Figure 2 shows the difference in the summer precipitation, which is almost equivalent to the long-term trend of precipitation, the dynamic and thermodynamic terms, and their sum. The sum of the dynamic and thermodynamic terms is comparable to the difference in the precipitation over the tropical NAF (Figures 2a2f), which means that the precipitation changes are mainly controlled by the two terms in all three experiments; this is because the large vertical motion is climatologically predominant. Over the Sahel region located at the northern boundary of the ITCZ, however, the anomalous horizontal moisture transport is comparable to the dynamic and thermodynamic terms (Figure S1 of the auxiliary material). This is consistent with the “upped-ante” mechanism suggested by Chou et al. [2009].

Figure 2.

(a–c) Difference in the summer mean precipitation between the early (1901–1920) and the late 20th Century (1981–2000), (d–f) changes in the mean summer precipitation estimated by the sum of the thermodynamic and dynamic effects, (g–i) the thermodynamic effect only, and (j–l) the dynamic effect only for the FULL (top), ARLS (middle), and GHGS (bottom) experiments.

[12] In the FULL experiment, a positively small thermodynamic effect and a negatively large dynamic effect produce a large decrease in precipitation over the tropical NAF (Figures 2d, 2g, and 2j). The ARSL shows large negative effects for both the dynamic and thermodynamic terms, resulting in a large decrease in precipitation (Figures 2e, 2h, and 2k). In contrast, the positive thermodynamic effect is almost canceled out by the negative dynamic effect in the GHGS, resulting in a small change in precipitation around the Gulf of Guinea (Figures 2f, 2i, and 2l). In the NTRL and OZON, these two effects on the precipitation change are much smaller than those in the GHGS and ARSL over the tropical NAF (not shown).

4.3. Physical Mechanisms of the Drying Trend Over Tropical North Africa

[13] The FULL and GHGS experiments showed a positive thermodynamic effect −〈equation imagepqequation image, while the ARSL showed a negative thermodynamic effect over the tropical NAF (Figures 2g2i). The specific humidity q increases with increasing temperature in the lower levels because the moisture usually concentrates in the lower troposphere, which means that the term ∂pq′ is positive with increasing temperature. Over the convergence zone, such as the ITCZ, the ascending wind is climatologically predominant, which means that the term equation image is negative. As a result, the FULL and GHGS, which induce the increasing temperature, result in the positive thermodynamic effect over the tropical NAF. The ARSL, which induces the decreasing temperature, results in the negative thermodynamic effect through the same process.

[14] On the other hand, the FULL, ARSL and GHGS experiments showed a negative dynamic effect −〈ω′∂pequation image〉 over the tropical NAF, although the dynamic effect was weakly positive over the fringes of the equatorial region (Figures 2j2l). In these experiments, the descending wind anomaly, which is a reduction of the ascending wind, causes the negative dynamic effect.

[15] The ARSL experiment shows the strong gradient of the SST anomaly over the tropical Atlantic Ocean, which modifies the local atmospheric circulation over tropical Africa (Figure 3a). The strong SST gradient is caused by the local radiative forcing due to increasing emissions of anthropogenic sulfate and carbonaceous aerosols from tropical Africa. The interhemispheric SST change over the Atlantic Ocean, which is caused by increasing anthropogenic aerosols from North America as well as tropical Africa, also enhances the atmospheric circulation change over tropical Africa. The contribution of aerosol forcings to the Atlantic SST is consistent with the results presented in previous studies [e.g., Rotstayn et al., 2000; Hoerling et al., 2006].

Figure 3.

Changes in SST, surface winds, radiative forcing at the tropopause and vertical circulation over the Atlantic Ocean between the early and the late 20th Century for (a) the ARSL and (b) the GHGS experiments. Horizontal panel: distribution of SST anomaly (shading), 925 hPa winds (vectors) and radiative forcing at the tropopause (contour). The contour intervals are 2.0 W/m2 (solid lines) and 1.0 W/m2 (broken lines). Vertical panel: zonal-mean cross sections of the vertical wind anomalies (shading) and meridional and vertical winds (vectors) over the Atlantic Ocean (60W–20E).

[16] The gradient of the SST anomaly is, on the other hand, unclear over the Atlantic Ocean in the GHGS experiment in spite of the strong descending wind anomaly over the tropical NAF (Figure 3b). An analysis of the change in the upper level circulation (Figure S2) indicated that the GHG-induced warming weakened the global-scale atmospheric circulation, as pointed out by Knutson and Manabe [1995], which resulted in the ascending anomaly over the South Atlantic Ocean. The cyclonic circulation anomaly at the surface corresponds to this ascending wind anomaly. The strong descending wind anomaly over the tropical NAF was associated with the ascending anomaly over the South Atlantic Ocean through a regional-scale circulation between the South Atlantic Ocean and tropical Africa (Figure 3b), but the detailed mechanisms remain to be investigated.

[17] The dynamic effect in the FULL experiment is affected by changes in both GHGs and aerosols. The SST increases in the FULL as well as the GHGS but has a strong gradient over the tropical Atlantic Ocean, as shown in the ARSL (Figure S3). These changes consequently produce the negatively large dynamic effect over the tropical NAF.

[18] Our results suggest a future moistening over the tropical NAF through the positive thermodynamic effect because GHG-induced warming will be accelerated and the surface cooling by anthropogenic aerosols will be reduced. This inference is consistent with Hoerling et al. [2006], who suggested that the precipitation would increase over the Sahel region in the 21st Century using the 18-model average of CGCMs. It would be useful to separate the time periods from the 1970s when determining the precipitation trend because the sulfate emissions have been decreasing and drying trend observed in the tropical NAF has been moderate since that time. In addition, we must consider natural decadal variations, such as the Atlantic Multidecadal Oscillation, which would play a significant role on the future precipitation trend in the tropical NAF as well as anthropogenic forcings [Ting et al., 2009].

5. Conclusion

[19] Several climate-model experiments forced with external natural and anthropogenic forcing factors were conducted to investigate the relative contribution of individual anthropogenic forcing factors to climatic changes in the 20th Century. The analysis of these experiments shows that the increased anthropogenic aerosols affect the drying trend observed over the tropical NAF in the boreal summer during the 20th Century. The diagnosis of dynamic and thermodynamic terms in an approximated moisture budget equation clarifies the physical mechanisms of the drying trend.

[20] Over the tropical NAF, increased anthropogenic aerosols, which induce surface cooling, result in a negative thermodynamic effect, while the increasing GHGs, which induce warming, result in a positive thermodynamic effect. On the other hand, the increased aerosols cause a negative dynamic effect over the tropical NAF corresponding to the descending wind anomaly caused by the strong gradient of the SST over the tropical Atlantic Ocean and also the inter-hemispheric SST change over the Atlantic Ocean. The increasing GHGs bring the negative dynamic effect over the tropical NAF, which is related to the change in the global-scale atmospheric circulation. Consequently, the increased aerosols cause the drying trend due to both negative dynamic and thermodynamic effects over the tropical NAF, while the increasing GHGs result in an unclear precipitation trend because the positive thermodynamic effect is cancelled out by the negative dynamic effect in the 20th Century.

[21] As the above consideration is based on one CGCM, MIROC, we need further studies using the other CGCMs to confirm our results of the changes in the atmospheric circulation which contribute to the dynamic effect. Although the dynamic effect of the increasing GHGs has some uncertainty, the thermodynamic effect brings moistening over the tropical NAF due to GHG-induced warming. This means that the contribution of the increasing GHGs to the drying trend would be small. For this reason, we strongly suggest that the drying trend observed over the tropical NAF is affected by the increasing anthropogenic aerosols.

Acknowledgments

[22] We thank two anonymous reviewers for their helpful comments. This work was partly supported by the Japanese Ministry of Education, Culture, Sports, Science, and Technology through the Innovative Program of Climate Change Projection for the 21st Century. The Earth Simulator at JAMTEC and NEC SX-6 and SX-8R at NIES were employed to perform the CGCM simulation. The GFD-DENNOU Library was used to draw the figures.

Ancillary