15 NOVEMBER 2014 M C C R A R Y E T A L . 8323

Simulations of the West African with a Superparameterized Climate Model. Part II: African Easterly Waves

RACHEL R. MCCRARY National Center for Atmospheric Research,* Boulder, Colorado

DAVID A. RANDALL Department of Atmospheric Science, Colorado State University, Fort Collins, Colorado

CRISTIANA STAN Department of Atmospheric, Oceanic and Earth Sciences, George Mason University, Fairfax, Virginia, and Center for Ocean–Land–Atmosphere Studies, Calverton, Maryland

(Manuscript received 6 November 2013, in final form 17 April 2014)

ABSTRACT

The relationship between African easterly waves and convection is examined in two coupled general circu- lation models: the Community Climate System Model (CCSM) and the ‘‘superparameterized’’ CCSM (SP-CCSM). In the CCSM, the easterly waves are much weaker than observed. In the SP-CCSM, a two- dimensional cloud-resolving model replaces the conventional cloud parameterizations of CCSM. Results show that this allows for the simulation of easterly waves with realistic horizontal and vertical structures, although the model exaggerates the intensity of easterly wave activity over West . The simulated waves of SP-CCSM are generated in East Africa and propagate westward at similar (although slightly slower) phase speeds to observations. The vertical structure of the waves resembles the first baroclinic mode. The coupling of the waves with convection is realistic. Evidence is provided herein that the diabatic heating associated with deep con- vection provides energy to the waves simulated in SP-CCSM. In contrast, horizontal and vertical structures of the weak waves in CCSM are unrealistic, and the simulated convection is decoupled from the circulation.

1. Introduction difficulty representing the mean annual cycle and vari- ability of precipitation over West Africa (e.g., Cook and The West African monsoon (WAM) involves many Vizy 2006; Roehrig et al. 2013). This problem is often at- complicated interactions between the atmosphere, ocean, tributed to sea surface temperature (SST) biases in the and land surface on a wide range of temporal and spatial equatorial Atlantic and their influence on the intertropical scales, from individual events to global atmospheric convergence zone (e.g., Vizy and Cook 2002). Even when dynamics. Coupled global circulation models (CGCMs), forced with observed SSTs, however, the models struggle which are used to project future climate changes, have to capture key rain-making weather systems such as Af- rican easterly waves (AEWs) (Sander and Jones 2008; Ruti and Dell’Aquila 2010). Denotes Open Access content. AEWs are synoptic-scale disturbances with periods of approximately 3–6 days, wavelengths in the range of 21 * The National Center for Atmospheric Research is sponsored 2000–6000 km, and westward phase speeds of 7–9 m s by the National Science Foundation. (e.g., Burpee 1972, 1974; Reed et al. 1977). AEWs are the dominant mode of variability over West Africa during the summer monsoon season [June–September (JJAS)] Corresponding author address: Rachel R. McCrary, National Center for Atmospheric Research, P.O. Box 3000, Boulder, CO and are strongly linked to rainfall and convection (Reed 80307. et al. 1977; Duvel 1990: Mathon et al. 2002; Fink and E-mail: [email protected] Reiner 2003; Gu et al. 2004; Kiladis et al. 2006; Mekonnen

DOI: 10.1175/JCLI-D-13-00677.1

Ó 2014 American Meteorological Society 8324 JOURNAL OF CLIMATE VOLUME 27 et al. 2006). It is well established that AEWs act as seed AEWs (e.g., Seo et al. 2008). One limitation of these disturbances for Atlantic hurricanes (Carlson 1969; studies is that they have focused primarily on the kine- Duvel 1990; Avila and Pasch 1992; Thorncroft and matic properties of AEWs, neglecting the role that Hodges 2001; Hopsch et al. 2010). convection and diabatic heating play in AEW devel- Despite decades of research, a complete un- opment and maintenance. As discussed above, it is now derstanding of the initiation, development, and main- believed that convection is an important component tenance of African easterly waves continues to elude us. of AEW dynamics and must be considered in model In the past, AEWs were thought to develop solely due to evaluations. the combined barotropic/baroclinic instability of the Our goal in this study is to examine and compare the midtropospheric (AEJ) (e.g., Burpee role that convection plays in the simulation of AEWs in 1972; Thorncroft and Hoskins 1994a,b). Current theory two CGCMs. The first is the standard Community Cli- suggests, however, that AEW initiation and growth mate System Model version 3 (hereafter CCSM3) cannot be accounted for without an additional energy (Collins et al. 2006) and the second is the super- source (Hall et al. 2006; Thorncroft et al. 2008; Hsieh parameterized CCSM3 (SP-CCSM) (Stan et al. 2010). and Cook 2005, 2007, 2008). In particular, it has been With superparameterization, conventional cloud param- argued that diabatic heating associated with convection eterizations are replaced by a simplified cloud-resolving is needed to account for the dynamics of AEWs. Both model (CRM) embedded in each grid column (e.g., theory (Thorncroft et al. 2008) and observational evi- Grabowski 2001; Khairoutdinov and Randall 2001). The dence (Berry and Thorncroft 2005; Mekonnen et al. superparameterization does not increase the resolution 2006; Kiladis et al. 2006) suggest that convective heating of the host model; it is simply a more physically based in the vicinity of the Darfur mountains acts as a finite- way to parameterize heating and drying rates. Super- amplitude perturbation to the regional circulation, trig- parameterized GCMs can serve as a stepping stone be- gering AEWs. The waves then propagate westward, tween traditional GCMs and the global CRMs, which drawing additional energy from the barotropic/baroclinic cannot yet be used in climate studies. With super- instability of the AEJ. It has also been suggested that parameterization, we can study the coupling between the AEWs can be initiated by the intense convection asso- large-scale circulation and cloud systems without the ciated with the ITCZ (Hsieh and Cook 2005). Convec- computational cost of global CRMs. tive heating within the ITCZ creates potential vorticity As shown in Part I of this paper (McCrary et al. 2014, (Schubert et al. 1991) and the reversal of the meridional hereafter Part I), the implementation of the super- potential vorticity gradient triggers the Charney–Stern parameterization improves many aspects of the West relationship, which provides energy to the waves (Hsieh African monsoon (WAM) simulated by CCSM3, in- andCook2005; Berry and Thorncroft 2012). Berry and cluding the mean position and intensity of monsoon Thorncroft (2012) showed that the organized deep con- rainfall and the meridional displacement of the ITCZ vection embedded within the AEWs themselves is criti- during the annual cycle. Part I shows that many of the cally important for the overall energetics of the waves and changes to the simulated WAM system can be attributed that without the convective heat source AEWs cannot to the improved vertical structure of atmospheric heating persist over Africa for as long as is observed. and moistening simulated by the SP-CCSM. By examin- A number of studies have examined AEW activity in ing the 2–6-day bandpass-filtered eddy kinetic energy CGCMs (Ruti and Dell’Aquila 2010; Daloz et al. 2012; (EKE), Part I also shows that the SP-CCSM exaggerates Skinner and Diffenbaugh 2013), uncoupled GCMs (Fyfe the intensity of easterly wave activity over Africa, while 1999; Chauvin et al. 2005; Reale et al. 2007; Pohl and CCSM3 exhibits little to no wave activity in the same re- Douville 2011), and regional climate models (Seo et al. gion(seeFig.12inPart I). The difference between these 2008; Sylla et al. 2013). Both Ruti and Dell’Aquila models is not their resolution, but in the way subgrid-scale (2010) and Skinner and Diffenbaugh (2013) demon- cloud processes are represented. strate that the CGCMs used in phase 3 of the Coupled The purpose of this paper is to identify why the two Model Intercomparison Project (CMIP3) exhibit dis- models exhibit such different wave activity. Our em- parate wave activity, with some showing overly robust phasis is on trying to understand how changing the wave activity and others showing little to no wave ac- physics influences AEW structure. We do this by ex- tivity. Skinner and Diffenbaugh (2013) go on to highlight ploring the relationship between AEWs and convection that synoptic-scale precipitation variability is strongly in both CCSM3 and SP-CCSM. The paper is organized correlated with AEW activity. Uncoupled GCM and re- as follows. Section 2 discusses the models, data sources, gional climate model experiments suggest that increased and methodology used to examine AEWs. Section 3 horizontal resolution may improve the simulation of presents a brief overview of the aspects of the climatology 15 NOVEMBER 2014 M C C R A R Y E T A L . 8325

TABLE 1. The models and observation-based datasets used in this study.

Origin/ Horizontal Temporal Vertical Selected Dataset reference resolution resolution levels variables CCSM3 Model T42 2.8832.88 Daily mean 26 levels interpolated Precipitation, OLR, , (Collins 25 years to standard temperature, specific et al. 2006) pressure levels humidity, geopotential height SP-CCSM Model T42 2.8832.88CRMs; Daily mean 30 levels interpolated Precipitation, OLR, winds, (Stan 4 km in north–south 25 years to standard temperature, specific et al. 2010) direction pressure levels, humidity, geopotential CRM levels height collocated with GCM levels TRMM Satellite and 0.25830.258 3 hourly Surface Precipitation (3B42) rain gauge averaged (Huffman to daily et al. 2007) means 1998–2010 NOAA Satellite 2.5832.58 Daily mean from Top of atmosphere OLR OLR (Liebmann 1979–2010 and Smith 1996) ERA-Interim Radiosonde, 1.5831.58 4 times daily 25 levels 1000–100 hPa Winds, specific humidity, satellite, model averaged to temperature, geopotential forecast (Dee daily means height et al. 2011) 1979–2010 over West Africa that are important for AEWs. The Gent 2002) and the Community Land Model version 3 variability of the simulated AEW activity is compared to (CLM3, Bonan et al. 2002). The cloud-resolving models observations in section 4. Section 5 discusses the hori- are coupled to the land surface at the GCM scale. More zontal and vertical structures of the simulated waves. information about the superparameterization can be The energetics of the simulated AEWs is explored in found in Grabowski and Smolarkiewicz (1999), Grabowski section 6. The paper ends in section 7, with a summary of (2001), Khairoutdinov and Randall (2001, 2003),and the results and a discussion. Khairoutdinov et al. (2005, 2008). The ocean coupling is described by Stan et al. (2010). Randall et al. (2014) pro- 2. Data and methods vide an overview of the superparameterized framework highlighting the influence of superparameterization on a. Models, observations, and reanalysis tropical variability. A detailed description of the models used in this study Table 1 also summarizes the observation-based data- can be found in Part I, Stan et al. (2010), and DeMott sets used in this study. Briefly, precipitation is from the 3- et al. (2011, 2013). The experimental designs are sum- hourly Tropical Rainfall Measure Mission (TRMM) marized in Table 1. Briefly, the atmospheric models 3B42 precipitation dataset (Huffman et al. 2007). Out- were run at T42 resolution with a semi-Lagrangian dy- going longwave radiation (OLR) is from the daily mean namical core. Each simulation is 25 years in length. The National Oceanic and Atmospheric Administration in- CCSM3 simulation used 26 levels, while the SP-CCSM terpolated OLR dataset (Liebmann and Smith 1996). All had 30 levels. In CCSM3, deep convection is parame- other meteorological fields are from the Interim Euro- terized following Zhang and McFarlane (1995).In pean Centre for Medium-Range Weather Forecasts SP-CCSM, 2D CRMs with periodic lateral boundary (ECMWF) Re-Analysis (ERA-Interim, herein ERA-I) conditions replace the conventional parameterization (Dee et al. 2011). All observational data have been av- of subgrid-scale cloud processes. The embedded CRMs eraged to daily time scales. The apparent heat source, Q1, have a horizontal spacing of 4 km, 28 levels, and are is estimated from ERA-I and both models using the ap- oriented in the east–west direction. Since the CRMs are proach of Lin and Johnson (1996): Q1 is a measure of 2D, convective momentum feedback from the CRM to diabatic heating and includes the influence of both radi- the large scale is not included. In both simulations the ative heating and latent heating. Although ERA-I is an- atmospheric model is coupled to the 38 version of the chored to observations through data assimilation, Q1 is Parallel Ocean Program (POP) ocean model (Smith and also influenced by precipitation in ERA-I, which is 8326 JOURNAL OF CLIMATE VOLUME 27 heavily model dependent. The Q1 estimates from ERA-I Figures 1a–c show the JJAS climatological pre- are likely consistent with observations, but do not rep- cipitation, 925-hPa winds, and the monsoon air mass resent the exact heating profiles. Janiga and Thorncroft from observations and the models. The simulated cli-

(2013) show that, while profiles of Q1 over West Africa matology is based on the entire 25-yr record from each vary among different reanalysis products, they have model, in observations the climatology of TRMM is qualitatively similar structures. The barotropic and bar- from 1997–2010, and ERA-I is derived from 1979–210. oclinic eddy kinetic energy conversion terms shown in In the observed monsoon, JJAS represents the peak section 6 are also estimated from the models and ERA-I. monsoon season when the largest rainfall amounts of Again, while the results from ERA-I are anchored by data the year occur over West Africa. During this season, the assimilation, they are influenced by the model used. ITCZ has moved northward, shifting from over the ocean during boreal winter to over the continent during b. Methods JJAS. Embedded within the observed monsoon are There are a number of ways to identify AEWs. The three distinct precipitation maxima: one over the Ethi- synoptic approach focuses on the kinematic properties opian Highlands, one near the Bight of Bonny near of AEWs and involves tracking anomalies of , Cameroon, and a third near the Guinea Highlands that pressure, and vorticity to generate track statistics and extends out over the . These precip- composites based on the passage of many individual waves itation maxima are all collocated with regional topog- (e.g., Hodges et al. 2003; Berry and Thorncroft 2005; raphy. The southwesterly low-level monsoon winds Janiga and Thorncroft 2013). An alternative approach cross the equator, and bring moisture onto the conti- focuses on the convective signal associated with the nent. The monsoon winds converge with the dry north- waves and uses space–time filtering to build a statistical erly harmattan winds in what is referred to as the picture of a ‘‘typical’’ disturbance using regression intertropical discontinuity (indicated by the thick back 2 analysis (Kiladis et al. 2006; Serra et al. 2008). The re- line in Fig. 1), just to the north of the 1 mm day 1 pre- sults shown here use the latter approach and present cipitation contour. Monsoon rainfall is qualitatively a statistical depiction of AEWs from observations and similar in other precipitation products such as those of the two models, using convection as an indicator of wave the Climatic Research Unit (CRU) and Global Pre- activity. This method allows us to focus on the coupling cipitation Climatology Project (GPCP). between convection and AEWs, which is important When compared to CCSM3, SP-CCSM better repre- since ultimately we want to determine how well the sents both the magnitude and the spatial patterns of models simulate precipitation over West Africa. summer monsoon precipitation over West Africa (Fig. 1). Using the space–time filtering methods of Wheeler The region of maximum precipitation is shifted from over and Kiladis (1999) and the regression methods described the Gulf of Guinea in CCSM3 (not realistic) to over the in Kiladis et al. (2006, hereafter K06), we examine continent in SP-CCSM. SP-CCSM captures the local a composite view of the horizontal and vertical struc- maximum in precipitation over the Atlantic as well as the tures of the simulated waves and compare them with maximum over the Ethiopian Highlands. SP-CCSM does observations. Details of the methods are given in sec- not simulate the maximum in precipitation near Cameroon tions 4 and 5 of this paper. nor does it capture the dry region along the Guinea Coast While some studies have attempted to use similar between the two coastal precipitation maxima. Rain rates space–time filtering methods to examine the energetics in the SP-CCSM tend to be much larger than observed of passing AEWs, we found that that it was more useful and rainfall during the peak monsoon period does not to separate the different energy sources using a time penetrate as far northward as observed. In CCSM3, rain filter alone. Details are given in section 6. rates over the continent are weaker than observed, and the simulated precipitation maximum occurs over the Gulf of Guinea. It is likely that the rainfall biases along 3. Monsoon climatology the Guinea Coast in SP-CCSM and over the Gulf of Guinea in CCSM3 are linked to warm SST biases in both a. Precipitation, OLR, and low-level winds models and a misrepresentation of the Atlantic equato- The observed seasonal cycle of the WAM has been well rial cold tongue (see Part I). Monsoon winds do not documented in previous studies (e.g.,Liebmann et al. penetrate as far northward as observed in either model, 2012). Part I gives a detailed discussion and examination and the intertropical discontinuity is positioned farther of the seasonal cycle as simulated by both CCSM3 and south than observed in both models. These differences SP-CCSM. Here we show the key features of the summer have important implications for the energetics analysis monsoon that are related to AEW activity. presented in section 6. 15 NOVEMBER 2014 M C C R A R Y E T A L . 8327

21 21 FIG. 1. JJAS mean precipitation (mm day , filled contours) and 925-hPa winds (m s , vectors) from (a) observations (TRMM and ERA-I), (b) SP-CCSM, and (c) CCSM averaged between June and September. The thick black line marks the zero contour of the zonal 2 wind and delineates the monsoon air mass. Also shown are JJAS OLR (W m 2) from (d) NOAA, (e) SP-CCSM, and (f) CCSM, and JJAS mean zonal wind at 600 hPa from (g) ERA-I, (h) SP-CCSM, and (i) CCSM.

OLR is often used as a proxy for rainfall because it is seem to suggest that more deep convection forms over more normally distributed than precipitation and the West Africa in CCSM3. In fact, however, the cold cloud time record for daily OLR covers a much longer period tops in CCSM3 are not associated with much rainfall. (e.g., ;30 yr for the NOAA OLR dataset compared to b. The African easterly jet 13 yr for TRMM precipitation). We use OLR as a proxy for rainfall in this study, but with some misgivings, as As discussed in the introduction, the AEJ is important discussed below. Figures 1d–f show that the JJAS OLR for AEW dynamics. While instabilities associated with in the observations and both models corresponds well the AEJ are not sufficient to trigger AEWs (Hall et al. with regions of enhanced rainfall. The observed OLR 2006), the jet is believed to be important for the prop- climatology is based on the 1979–2010 time period; the agation and overall development of AEWs. The jet de- simulated climatology is from the 25-yr simulation. In velops during the summer season owing to the strong the models, the OLR is influenced not only by the cloud low-level meridional temperature and soil moisture parameterizations, which generate the simulated clouds, gradients between the Gulf of Guinea and the Sahara. but also by the microphysics parameterizations and the way Figures 1g–i show the JJAS mean zonal wind at 600 hPa. 2 that radiation is parameterized. For example, the colder In ERA-I, the peak winds reach about 12–14 m s 1 cloud tops found in CCSM3, compared to SP-CCSM, might (Fig. 1h). The winds are found at about 158N, at the 8328 JOURNAL OF CLIMATE VOLUME 27

FIG. 2. Average JJAS signal-to-noise space–time spectra averaged between 158S and 158N at all longitudes for disturbances that are (a)–(c) symmetric and (d)–(f) antisymmetric about the equator from (left) observations, (middle) SP-CCSM, and (right) CCSM.

600-hPa level, and extend from the west coast to ap- (top) and antisymmetric across the equator (bottom). proximately 58E. Easterly winds also extend down to These figures were created using the methods outlined about 850 hPa, where they give way to the westerly mon- in Wheeler and Kiladis (1999). Anomalies are generated soon winds (not shown). The AEJ is reasonably well po- by removing the first three harmonics of the seasonal sitioned in both CCSM3 and SP-CCSM. The jet tends to be cycle, the June–September signal is calculated by sub- slightly weaker than observed in SP-CCSM and extends dividing each summer into two 96-day segments that farther east than observed, to about 208E. As mentioned in overlap by 60 days, and the red-noise spectrum is based Part I, the AEJ may be weaker than observed in SP-CCSM on the data from all months of the year. because of the overly active simulated easterly wave ac- In the observations, considerable power is found in tivity. The strength of the AEJ may also be influenced by the tropical depression (TD) range, which includes west- the differences in the low-level meridional temperature ward propagating waves with wavelengths 2000–7000 km gradients found in each model as well as differences in the and periods 2–10 days. The TD range has been shown representation of the diabatically forced circulations as- to correspond well with AEW activity (K06). CCSM3 sociated with the ITCZ and the Saharan heat low (Part I). exhibits little to no power in ranges associated with the Madden–Julian oscillation (MJO), equatorial Rossby (ER) waves, or inertia–gravity waves. It does have sub- 4. African easterly waves stantial power in the Kelvin wave range, although at To examine the organization of tropical convection shorter wavelengths than observed. There is no coherent during boreal summer, space–time spectra of the OLR easterly wave power in CCSM3. for JJAS have been calculated for the observations and Implementation of the superparameterization alters both models. Figure 2 shows the average June–September the interactions between the large-scale dynamics and signal-to-noise ratio power spectra of OLR averaged be- cloud processes in such a way that the simulated spectrum tween 158Sand158N for disturbances that are symmetric of tropical wave activity is more realistic. In particular, 15 NOVEMBER 2014 M C C R A R Y E T A L . 8329 the SP-CCSM simulates TD activity, suggesting that AEW activity should be present in this model. Easterly waves are predominantly a Northern Hemisphere phe- nomenon, so significant power is also found in the TD region of the antisymmetric power spectra (Figs. 2d–f). Figure 3 shows the variance of TD filtered OLR from observations and for both models. During JJAS, the observed easterly wave activity occurs broadly across most of the tropics and is strongest over West Africa, the eastern Atlantic, and the intra-American seas (Fig. 3a). Easterly waves are also observed over the central and western Pacific. Over West Africa, TD filtering captures approximately 20%–30% of the total observed OLR variance (McCrary 2012). In SP-CCSM, the simulated easterly wave activity is greatly exaggerated over West Africa and the east Atlantic, where the variance in TD-filtered OLR is more than 4 times larger than observed (Fig. 3b). In this region, the simulated TD-filtered OLR contains 35%– 45% of the total variability (McCrary 2012). SP-CCSM also simulates a southward shift in the TD-filtered OLR over Africa, compared to observations. This corresponds 22 2 to the regional maximum of precipitation in SP-CCSM FIG. 3. JJAS mean variance of TD-filtered OLR [(W m ) ] from (Figs. 1b,e). As in the observations, easterly wave activity (a) observations, (b) SP-CCSM, and (c) CCSM. occurs broadly over the tropics in SP-CCSM, but it is weaker than observed over the central and western Pa- cific. We speculate that the exaggerated easterly wave and simulated waves. Following K06, we use the TD- activity over Africa may be due in part to differences in filtered time series of OLR at the point 108N, 108W (our the simulated convective processes over land compared base point) as an index of AEW activity. This base point to oceans. time series of AEW variability serves as the reference As expected from the space–time spectra, easterly time series against which all dynamic, thermodynamic, wave activity in CCSM3 is weaker than observed (Fig. 3c). and convective variables are regressed. The TD-filtered Over Africa, the only signal in the CCSM3-simulated TD- AEW index is regressed against unfiltered daily anom- filtered variance is found over the Gulf of Guinea, where alies of all circulation- and convection-related variables precipitation is a maximum. This variance is less than half of interest at all grid points and levels covering the West of the observed and accounts for less than 15% of the total African domain. These variables include OLR, stream- variance (McCrary 2012). function, winds, vertical velocity, temperature, humid- We obtained similar results using traditional kine- ity, and diabatic heating. In the following figures, the matic indices of AEW activity, such as the variance in regressed value of each variable is plotted at each grid 2–6-day bandpass-filtered 700-hPa meridional wind (see point and level for a 21 standard deviation change in the Part I). Again, SP-CCSM shows exaggerated AEW ac- filtered AEW index. This allows us to examine the cir- tivity, while CCSM3 exhibits much less activity than is culation associated with a typical AEW disturbance observed (see Part 1). passing over the point 108N, 108W. Lag regressions are also calculated to examine the circulation on the days before and after the wave passes the base point. Re- 5. The structure of African easterly waves gressions are calculated for the JJAS time period using Next, we compare the horizontal and vertical struc- data from all years available from each dataset and tures of the simulated AEWs with observations during model. JJAS. To do this we use lagged linear regression tech- The base point 108N, 108W was chosen because it was niques similar to those found in K06 and references used in K06 and is in a region of high OLR variability. It within. These techniques are often used to study con- is south of the AEJ, where barotropic energy conver- vectively coupled waves and allow us to compare the sions occur. Other base points have also been examined, composite horizontal and vertical structure of the observed and results are discussed by McCrary (2012). 8330 JOURNAL OF CLIMATE VOLUME 27

FIG. 4. Horizontal structure of AEWs at the base point 108N, 108W from (left) observations, (middle) SP-CCSM, and (right) CCSM at lag (top) 22, (middle) 0, and (bottom) 12 days. Anomalies of OLR and 850-hPa circulation associated with one standard deviation of the 2 TD-filtered OLR time series at the base point from each dataset. Filled contours are OLR anomalies in W m 2. Line contours are the 2 850-hPa streamfunction contoured from 1 3 105 m2 s 1. Vectors are the 850-hPa winds. Only the statistically significant OLR and wind vectors at the 95% confidence level are shown. a. Horizontal structure correlations are determined to be statistically significant at the 95% confidence level using a two-tailed t test are The horizontal structure of AEWs is shown in Fig. 4. shown. The streamfunction is shown everywhere for This figure shows the OLR and circulation (stream- ease of interpretation of the overall circulation. For each function and winds) anomalies corresponding with a 21 model, the contours are scaled by the standard deviation standard deviation change in the TD-filtered OLR time of the TD-filtered time series at the base point, in order series at the base point 108N, 108W from observations to emphasize the differences in the structures of the (right), SP-CCSM (middle) and CCSM3 (left). With this waves rather than their magnitudes. In all cases, the we can examine the linear relationship between AEW magnitudes are larger than observed in SP-CCSM and activity at the base point and the corresponding spatial smaller than observed in CCSM3. patterns of the circulation and convection over all of Despite large differences in the strength of the AEWs West Africa. To examine the temporal progression of in the observations compared to SP-CCSM, the hori- the waves as they pass the base point, the regressions are zontal structure of the simulated waves is fairly realistic. calculated for lag 22 (top), 0 (middle), and 2 (bottom) At lag 0 (Figs. 4b,e), there is a relatively large area of days. For OLR and wind anomalies only values where negative OLR centered near the base point, indicating 15 NOVEMBER 2014 M C C R A R Y E T A L . 8331

FIG. 5. Horizontal structure of AEWs at the base point 108N, 108W from (left) observations, (middle) SP-CCSM, and (right) CCSM at lag (top) 22, (middle) 0, and (bottom) 12 days. Anomalies of precipitation and 850-hPa circulation associated with one standard de- viation of the TD-filtered precipitation time series at the base point from each dataset. Filled contours are anomalies of precipitation in 2 2 mm day 1. Line contours are the 850-hPa streamfunction contoured from 1 3 105 m2 s 1. Vectors are the 850-hPa winds. Only the statistically significant precipitation contours and wind vectors at the 95% confidence level are shown. Anomalies will be of opposite sign to those found in Fig. 4 because increased rainfall is typically associated with colder cloud tops. a region of enhanced convection. This convective region Some notable differences between the observed and is flanked on either side by regions of resolved-scale simulated waves are as follows. OLR anomalies in SP- subsidence. In both cases, the convective anomalies are CCSM are shifted to the south of the center of each collocated with anomalies of the 850-hPa circulation. circulation pattern, which is not observed. The observed Enhanced convection is associated with an anomalous waves originate over the Darfur Mountains or as far east cyclonic circulation, and suppressed convection is as- as the Ethiopian Highlands (K06), but in SP-CCSM the sociated with an anomalous anticyclonic circulation. At waves form farther west. As noted by K06, as the ob- lag 0, convection occurs within the trough of the wave in served waves propagate westward, the OLR minimum both cases. Following the development of these waves shifts from behind the trough over East Africa, to cen- at various lags, we see that they emanate from East tered within the trough over West Africa, and finally to Africa (Figs. 4a,d) and propagate offshore into the ahead of the trough over the Atlantic. A similar analysis Atlantic basin (Figs. 4c,f), where it is observed that done with TRMM rainfall data rather than OLR shows they sometimes act as seed disturbances for Atlantic that actual precipitation may occur ahead of the trough hurricanes. as AEWs propagate across West Africa (Fig. 5). In Fig. 5 8332 JOURNAL OF CLIMATE VOLUME 27 the horizontal structures of the waves from observations source (Q1, Figs. 6p–r) that correspond with a 21 stan- and both models are calculated using the TD-filtered dard deviation change in the TD-filtered OLR time se- time series of precipitation, rather than OLR, at the base ries at the base point 108N, 108W at lag 0 for the point. The TRMM data show that, as might be expected observations and both models. The thick black line in in midlatitude synoptic storms, precipitation tends to each panel highlights the position of peak convection, occur ahead of the trough axis as the waves propagate which can also be found in the OLR lag-regression plots across West Africa (Figs. 5a–c). The differences be- at the top of the panel (Figs. 6a–c). As usual, we expect tween precipitation and OLR may be due to either that the strongest convective activity roughly corre- persistence of high clouds after precipitation has ended sponds with a minimum in OLR. or the coarse resolution of the NOAA OLR product. The vertical structure of the observed AEWs is that The spatial relationship between convection and the expected for the first baroclinic mode, with meridional wave in SP-CCSM is different from observations, which winds changing sign at approximately 300 hPa (Fig. 6d). show that convection (both OLR and precipitation) There is very little tilt of the meridional winds with resides near the trough axis as the wave propagates height in the lower levels, indicating that barotropic across Africa (Figs. 4d,e and 5d,e) and falls behind the energy conversion is important for AEWs in this region. trough when the wave moves out over the Atlantic (Figs. The peak in convection occurs near the trough axis, 4f and 5f). Differences between the simulated and ob- where the winds change from southerly (positive) on the served waves are likely due to either the coarse resolu- east side of the trough to northerly (negative) on the tion of the global model, which spreads rainfall out west side. As expected, the region of peak convection is spatially, or incorrect timing of convection relative to associated with the strongest upward motion (Fig. 6g). the wave. The waves simulated by SP-CCSM also prop- As pointed out by K06, there are two distinct maxima in agate more slowly than observed, which may be due to the vertical profile of omega, at 700 and 400 hPa. These the overly strong convection found in the model. are thought to correspond with two different cloud In CCSM3, the regression method does show a weak populations, namely shallow convection at low levels convective signal along 108N, but there is no statistically and deep convection aloft. Vertical profiles of specific significant circulation pattern associated with the OLR humidity (Fig. 6m) and diabatic heating Q1 (Fig. 6p) or precipitation anomalies (Figs. 4g–i and 5g–i). This show peaks similar to the two peaks in omega. The rel- indicates that convection is decoupled from the circu- ative minima in omega, specific humidity, and Q1 found lation on the TD time and space scales. at 600 hPa are near the freezing level. As mentioned in the introduction, a number of pre- Ahead of the peak in convective activity, the lower vious studies have shown that observed AEWs tend to troposphere is anomalously warm and moist (Figs. 6j,m), follow two different tracks, one to the south of the AEJ most likely due to the effects of shallow cumulus con- and another to the north. In this section we have focused vection ahead of the peak in deep convection (K06), on the southern track of the observed AEWs, as the although warm advection from northerlies ahead of the direct link with moist convection is clearer south of the trough may influence temperatures as well. Where AEJ. When we examined AEWs using base points to convective activity is strongest, the layer between 850 the north of the AEJ, we found that the northern track is and 500 hPa is cool and moist. These cool, moist condi- not well captured by the SP-CCSM. Further discussion is tions also trail the peak in convection by about 108 given by McCrary (2012). As shown in the next section, longitude. The anomalously cool temperatures behind diabatic heating appears to drive the waves simulated by the peak in convection may be due to adiabatic cooling SP-CCSM. The northern track of observed AEWs tends associated with the peak in upward motion found in the to be dry and not as closely associated with convection as same layer, or they may be due to the evaporation of the southern track (Reed et al. 1977). In SP-CCSM, any falling rain. A second cool region is found near the wave activity to the north of the AEJ appears to ema- surface. This low-level cool layer is likely due to con- nate out of the region of maximum convection, which is vective downdrafts and cold air advection from the Gulf not observed (McCrary 2012). of Guinea. The secondary peak in vertical motion found in the upper troposphere between 200 and 400 hPa b. Vertical structure closely corresponds with anomalously warm tempera- The vertical structures of the observed and simulated tures. This warming is thought to be due to the diabatic AEWs are depicted in Fig. 6, which shows vertical cross heating associated with deep convection. The positive sections, along 108N, of anomalies of meridional wind correlation between temperature and Q1 at the same (Figs. 6d–f), omega (Figs. 6g–i), temperature (Figs. 6j–l), level suggests that heating acts to strengthen the dis- specific humidity (Q, Figs. 6m–o), and the apparent heat turbance. Behind the main convective signal, at upper 15 NOVEMBER 2014 M C C R A R Y E T A L . 8333

22 21 FIG. 6. Zonal cross section of (a)–(c) OLR (W m ) and zonal–height cross sections of (d)–(f) anomalous meridional wind (m s ), 21 21 21 (g)–(i) omega (Pa s ), (j)–(l) temperature (K), (m)–(o) specific humidity (g kg ), and (p)–(r) the apparent heat source Q1 (K day ) along 108N from (left) observations, (middle) SP-CCSM, and (right) CCSM at lag day 0. The position of peak convection is highlighted with the thick black line. 8334 JOURNAL OF CLIMATE VOLUME 27 levels, anomalously moist conditions persist, likely as- sociated with trailing stratiform convection. This picture of AEWs closely follows what we might expect from any convectively coupled equatorial wave (e.g., Kiladis et al. 2009). The vertical structure of AEWs in SP-CCSM is markedly similar to that observed. The meridional winds have a ‘‘first baroclinic mode’’ structure, changing sign at approximately 300 hPa (Fig. 6e). At low levels, the slight eastward tilt of the meridional winds suggests that baroclinic energy conversions may be important for AEWs near 108N, more in the model than in the ob- servations (this will be examined further in section 6). As in ERA-I, the region of maximum convection is as- sociated with enhanced upward motion (Fig. 6h), al- FIG. 7. Schematic of Lorenz energy diagram highlighting the though it is an order of magnitude stronger in the model main production terms of eddy kinetic energy associated with than in the reanalysis. Also, there is only one maximum African easterly waves. Primes are deviations from a 10-day run- v in the omega profile, centered at about 400 hPa. The ning average. Note that T is daily mean temperature, vertical velocity, Q the apparent heat source, V the horizontal wind vertical profiles of convection associated with AEWs in 1 H speed, and p pressure. Also, g 5Gd/(Gd 2G), where Gd and G are this model do not show the observed relative minima at the dry adiabatic and observed lapse rates respectively; cp is the 600 hPa. The vertical profiles of omega (Fig. 6h), specific heat capacity at constant pressure, and R is the dry air gas constant. humidity (Fig. 6k), and diabatic heating (Fig. 6n) also only show one vertical peak. It may be that the vertical resolution of the model is too coarse to capture the associated with weak positive anomalies of moisture and finescale vertical structures associated with deep con- temperature at the upper levels, suggesting top-heavy vection in the tropics. The simulated vertical structure of convection. This top-heavy convective heating profile the temperature (Fig. 6k) is comparable to observations, may help to explain the deeper minima of the OLR in with a warm signal ahead of the peak in convection, cool CCSM3, relative to SP-CCSM. anomalies in the mid to lower troposphere trailing the peak in convection, and warming between 200 and 6. The energetics of AEWs 400 hPa. There is also a weak warming and moistening ahead of the deep convection, which may be associated We now examine the energy sources and trans- with shallow convection. The diabatic heating associated formations that are important for the AEWs. The Lorenz with the waves in SP-CCSM is a factor of 4 stronger than energy diagram provided in Fig. 7 shows the processes that calculated from ERA-I. As discussed in the next section, can potentially contribute to the eddy kinetic energy diabatic heating appears to play a critical and perhaps (EKE) associated with AEWs. Three processes dominate exaggerated role in generating AEWs in SP-CCSM. the energy cycles of the waves: barotropic conversion, In CCSM3, the simulated vertical profiles of the me- baroclinic conversion, and diabatic heating. In this section, ridional wind (Fig. 6f), omega (Fig. 6i), temperature primes represent departures from a 10-day running aver-

(Fig. 6l), humidity (Fig. 6o), and Q1 (Fig. 6r) are quite age, so can be interpreted as synoptic-scale anomalies. This different from those observed. Convection in CCSM3 method does not separate ‘‘wave’’ from ‘‘no-wave’’ dis- always occurs ahead of the trough axis in a region of turbances and we will actually be focusing on the energy anomalous northerlies. As expected from the maps of associated with all synoptic-scale disturbances. The current AEWs, the meridional wind anomalies associated with analysis in terms of temporal anomalies is comparable to convection are much weaker than observed (Fig. 6). At that of Hsieh and Cook (2007), who analyzed AEWs in lag 0 there is noticeable eastward tilt with height in the terms of departures from a zonal average. meridional winds below 800 hPa. The profiles of omega Figure 8 shows meridional cross sections of the baro- are very top-heavy, with the maximum uplift above tropic and baroclinic sources of EKE for ERA-I and the 400 hPa. The vertical profiles of diabatic heating are also two models, averaged between 208W and 108E. This top-heavy, with positive values of Q1 only above 600 hPa. range of longitudes is associated with amplified AEW While the signature of shallow convection at low levels activity in both ERA-I and SP-CCSM. In these figures, ahead of the peak in deep convection is strong in the warm colors indicate regions where AEWs gain EKE, humidity and temperature plots, the peak in convection is and cool colors show regions where EKE is lost. 15 NOVEMBER 2014 M C C R A R Y E T A L . 8335

FIG. 8. Meridional–height cross sections of the (left) baroclinic and (right) barotropic conversions to eddy kinetic energy averaged between 208W and 108E from (a),(d) ERA-I, (b),(e) SP-CCSM, and (c),(f) CCSM. Energy terms are 2 2 in m2 s 2 day 1.

In ERA-I and both models, there are three regions of both ERA-I and SP-CCSM3, this production region is EKE production due to barotropic conversions (Figs. centered at ;108N and is south of the AEJ, near the 8a–c). The first, found at upper levels, is associated with ITCZ. Given the exaggerated strength of the AEWs in the tropical easterly jet. The second is in the middle SP-CCSM, it is not surprising that the barotropic energy troposphere for the observations and both models. In conversion term is larger in the model than in ERA-I. 8336 JOURNAL OF CLIMATE VOLUME 27

CCSM3 also produces a source in the midtroposphere, We now investigate potential sources of eddy avail- but it is markedly weaker and positioned far to the south able potential energy (EAPE). Figure 9 shows the of the jet, over the Gulf of Guinea. The third production meridional cross sections of the two other terms asso- region is found at low levels in the observations and both ciated with the eddy available potential energy equa- models, near the location of the intertropical depression tion. The first term represents the conversion of mean (ITD), where the southwesterly monsoon winds converge potential energy to eddy available potential energy with the northeasterly harmattan winds. In ERA-I, this owing to the eddy heat flux along the horizontal mean source region is centered at 17.58N, whereas in both SP- temperature gradient (Figs. 9a–c). Given that the me- CCSM and CCSM3 it is found at 158N. Recall from sec- ridional temperature gradient in ERA-I and both tion 3 that in both models the monsoon winds do not models is largest over North Africa, it is not surprising penetrate as far northward as observed, so the simulated that this term is positive north of 108N at low levels. confluence region of the ITD tends to be displaced In this region, mean available potential energy is southward relative to the observations. Hsieh and Cook converted to EAPE and then to EKE via baroclinic (2007) found that the low-level production of EKE is processes. partially offset by the loss of EKE due to frictional dis- The second term (Figs. 9d–f) represents the genera- sipation. Unfortunately, neither ERA-I nor the models tion or destruction of EAPE due to diabatic processes. have the diagnostics needed to perform a complete en- We will refer to this term as the ‘‘generation term,’’ ergy budget, so that possibility cannot be evaluated here. denoted by G. In ERA-I diabatic processes generate Baroclinic conversion is also important for AEWs. In EAPE north of 158N at low levels. Farther south, cen- ERA-I and both models south of 128N, positive con- tered at 108N, G is negative at low levels (750 hPa) but versions of EKE due to baroclinic processes are found in positive at upper levels (350 hPa) (Fig. 9d). The close the lower and upper troposphere, with a region corre- correspondence of G with the baroclinic term at 108N sponding with the destruction of EKE in between (Figs. suggests that the EKE consumed (generated) by the 8d–f). For ERA-I, the region of positive baroclinic en- baroclinic term is in part generated by heating and ergy conversion at low levels extends from 128 to 258N cooling. What this tells us is that, at low levels, cold and from the surface to 600 hPa. As discussed by Reed anomalies created by adiabatic cooling of rising air are et al. (1977) and Hsieh and Cook (2007), the positive balanced by convective heating (Q1 . 0). At upper baroclinic conversions found in this region are associ- levels, warm descending is offset by evaporative cooling ated with the ascent of warm dry air from the north and (Q1 , 0). At upper levels, a positive covariance of the descent of cool moist air from the south. In both SP- temperature and heating acts to generate EAPE, sup- CCSM and CCSM3, the baroclinic term is stronger at porting wave development. When integrated over the low levels than in ERA-I, and extends farther to the entire depth of the atmosphere, it appears that G weakly south. In both models, the temperature perturbations tends to increase the EAPE between 158 and 258N, es- are stronger than observed (McCrary 2012). That the pecially over the ocean (Fig. 10). baroclinic conversion to EKE extends farther south in Convection plays an exaggerated role in the energet- SP-CCSM supports the idea that baroclinic conversions ics of AEWs in SP-CCSM (Fig. 9e). When averaged play an important role in dynamics of AEWs at low between 208W and 108E, the simulation is similar to levels along 108N (see section 5b). the observations except that the overall magnitudes of In both ERA-I and SP-CCSM, strong baroclinic the sources and sinks of EAPE are stronger than ob- conversion is also found between 500 and 200 hPa cen- served. Near 78N, 750 hPa we see a large region where tered over 108N(Figs. 8d,e). This is most likely due to G is negative. Above this, centered at 400 hPa, there is the ascent of warm air associated with the latent heat a region where G is positive. Averaged throughout the released owing to the convection embedded within the depth of the troposphere, G acts as a source of EAPE AEWs. Below 500 hPa the baroclinic term is negative (Fig. 10b). (v0T0 , 0), which can be due to ascending cold air or In CCSM3, the only substantial values of G are at descending warm air. It has been suggested that dy- upper levels (Fig. 9f). Remember from section 5b that namical forcing within the waves may cause ascending convection in this model is fairly top-heavy, so we would cold air (Yanai 1961; Diedhiou et al. 2002; Hsieh and expect diabatic heating to be larger at upper levels. Cook, 2007). The variances of vertical velocity and In summary, we have shown that in ERA-I and SP- temperature in CCSM3 are much weaker than those in CCSM diabatic heating is a source of energy for AEWs. either ERA-I or SP-CCSM (McCrary 2012). At upper Overall, the diabatic heating term plays a larger role in levels, positive baroclinic conversion occurs only above SP-CCSM. Convection does not appear to be important 350 hPa, and is shifted south, centered over the equator. for the limited wave activity found in CCSM3. 15 NOVEMBER 2014 M C C R A R Y E T A L . 8337

2 21 0 0 $ 1 21 0 0 FIG. 9. Meridional–height cross sections of the (right) T cpgVH T H T and (left) T gQ1T conversion to eddy kinetic energy averaged between 208W and 108E from (a),(d) ERA-I, (b),(e) SP-CCSM, and (c),(f) CCSM. 2 2 Energy terms are in m2 s 2 day 1.

7. Summary and discussion examined the variability, structure, and energetics of AEWs in two CGCMs: the standard CCSM3, which The primary goal of this paper was to investigate the uses traditional convective parameterizations, and the sensitivity of AEWs to model physics and the subgrid- SP-CCSM, where convective parameterizations are re- scale representation of convective processes. We have placed by embedding a 2D CRM into each grid box. The 8338 JOURNAL OF CLIMATE VOLUME 27

advantage of the SP-CCSM is that convective clouds and basic mesoscale dynamics are explicitly represented, without the immense cost of a global cloud-resolving model. The important thing to remember in this study is that both models are run at a coarse T42 resolution and that the significant differences found in the simulation of AEWs in each model is due to their different repre- sentations of moist physics. Through this work we have shown that the incor- poration of the superparameterization into CCSM3 greatly modifies how tropical convection is organized and enables the simulation of AEWs, although they are stronger than observed. CCSM3 produces little to no easterly wave activity over West Africa, and it appears that convection is decoupled from the circulation in that model. Diabatic heating profiles are top-heavy in CCSM3, and a large dry bias in the midtroposphere may inhibit deep convection. In the SP-CCSM, on the other hand, AEWs are a robust feature of the variability over West Africa. Although AEWs in SP-CCSM are much stronger than observed, their horizontal and vertical structures are similar to observations. The waves in SP- CCSM develop in central/eastern Africa, as observed, and propagate westward at slightly slower phase speeds than found in observations. The vertical structure of the waves has the form of the first baroclinic mode, with warm moist conditions found at low levels ahead of the region of deep convection, and some indication that stratiform convection may follow the region of deep convection with rain falling through dry conditions and cooling the air through evaporation. This structure is similar to that described in Kiladis et al. (2009), who studied the vertical profiles of convectively coupled equatorial waves. The AEWs found in SP-CCSM are associated with intense convection, and therefore excessive diabatic heating. Diabatic heating generates eddy available po- tential energy, which serves as an important source of energy for the simulated AEWs. It appears that AEWs in SP-CCSM are maintained by the combined effects of the hydrodynamic instability of the AEJ and the con- vection imbedded within the waves. Additional work is needed to fully diagnose the processes that generate and maintain the waves in SP-CCSM. Forecasting experi- ments might be useful. Results from the SP-CCSM are consistent with a number of other studies that have pointed to convec- tion as an important component of wave initiation and FIG. 10. Vertical averages of the diabatic heating term in the growth (e.g., Hall et al. 2006). Our analysis of the ERA-I eddy available potential energy equation from (a) ERA-I, (b) SP- CCSM, and (c) CCSM between 925 and 200 hPa. Energy terms are data shows that diabatic heating is only a weak energy 2 22 21 in m s day . source for AEWs. Keep in mind, however, that Q1 is not directly available from ERA-I and was approximated for

use in this study (section 2). In addition, Q1 for ERA-I is 15 NOVEMBER 2014 M C C R A R Y E T A L . 8339 influenced by the model’s parameterizations. Perhaps Q1 Because the SP-CCSM can simulate both the MJO is underestimated in ERA-Interim and diabatic heating (Benedict and Randall 2009) and the Asian summer does play a more critical role in wave development. monsoon (DeMott et al. 2011, 2013) with reasonable

A comparison of the climatologies of Q1 across multiple fidelity, it may be possible to use the model to examine reanalysis products (Janiga and Thorncroft 2013)shows the teleconnections between the Indo-Pacific and West that they have qualitatively similar patterns, but quanti- Africa, thereby improving our understanding of sea- tative differences that could influence the results. A sec- sonal and annual rainfall variability over Africa. ond issue is that cumulus momentum transport is not included in SP-CCSM and may act in nature as a fric- Acknowledgments. George Kiladis provided helpful tional influence that reduces the amplitude of the waves. input as this research was carried out. Our research has We also note that Diaz and Aiyyer (2013) present evi- been supported by the National Science Foundation dence that upstream development associated with the Science and Technology Center for Multi-Scale Mod- energy dispersion from a previous AEW may act to eling of Atmospheric Processes (CMMAP), managed by trigger subsequent AEWs over East Africa, suggesting Colorado State University under Cooperative Agree- that convection is important for the amplification of ment ATM-0425247. Cristiana Stan was also supported AEWs but not their genesis. with the NSF Grant AGS-1211848. We acknowledge the It has been argued that higher-resolution simulations support of the Computational and Information Systems are needed to simulate the complex WAM system and Laboratory (CISL) at NCAR for providing computer African easterly waves. AEWs are relatively small fea- time for this work. tures (wavelengths 2000–6000 km) that propagate quickly 2 (phase speeds 7–9 m s 1) across West Africa so it does seem likely that the simulated AEWs would benefit from REFERENCES increased spatial resolution. A moderate increase in res- olution does not lead to more realistic physics, however. Alaka, G. J., Jr., and E. D. Maloney, 2012: The influence of the The point of our work is that the AEWs are also sensitive MJO on upstream precursors to African easterly waves. J. Climate, 25, 3219–3236, doi:10.1175/JCLI-D-11-00232.1. to model physics. We show that an improved representa- Avila, L. A., and R. J. Pasch, 1992: Atlantic tropical systems tion of subgrid-scale cloud processes allows AEWs to de- of 1991. Mon. Wea. Rev., 120, 2688–2696, doi:10.1175/ velop even in a coarse-resolution CGCM. Not only do 1520-0493(1992)120,2688:ATSO.2.0.CO;2. AEWs exist in the low-resolution SP-CCSM, but their Benedict, J. J., and D. A. Randall, 2009: Structure of the Madden– horizontal and vertical structures are remarkably similar to Julian oscillation in the superparameterized CAM. J. Atmos. Sci., 66, 3277–3296, doi:10.1175/2009JAS3030.1. observations. Berry, G. J., and C. D. Thorncroft, 2005: Case study of an intense Based on our analysis, we wonder if the varied wave African easterly wave. Mon. Wea. Rev., 133, 752–766, doi:10.1175/ activity found in the CMIP3 models identified by Ruti MWR2884.1. and Dell’Aquila (2010) and Skinner and Diffenbaugh ——, and ——, 2012: African easterly wave dynamics in a meso- (2013) may be a function of convective parameterization scale numerical model: The upscale role of convection. J. At- mos. Sci., 69, 1267–1283, doi:10.1175/JAS-D-11-099.1. and the way that temperature and heating profiles are Bonan, G. B., K. W. Oleson, M. Vertenstein, S. Levis, X. Zeng, represented in CGCMs. We have demonstrated that Y. Dai, R. E. Dickinson, and Z.-L. Yang, 2002: The land sur- moist physics is important for the simulation of AEWs face climatology of the Community Land Model coupled to the and are currently investigating the role that convection NCAR Community Climate Model. J. Climate, 15, 3123–3149, , . plays in AEW variability in other CGCMs. doi:10.1175/1520-0442(2002)015 3123:TLSCOT 2.0.CO;2. Burpee, R. W., 1972: The origin and structure of easterly waves in Taking our study together with others published the lower troposphere of North Africa. J. Atmos. Sci., 29, 77–90, earlier (e.g., Benedict and Randall 2009; DeMott et al. doi:10.1175/1520-0469(1972)029,0077:TOASOE.2.0.CO;2. 2011), SP-CCSM has been shown to improve the ——, 1974: Characteristics of North African easterly waves during simulation of tropical precipitation and variability the summers of 1968 and 1969. J. Atmos. Sci., 31, 1556–1570, , . over many regions of the tropics. In terms of West doi:10.1175/1520-0469(1974)031 1556:CONAEW 2.0.CO;2. Carlson, T. N., 1969: Synoptic histories of three African distur- African climate, this is important because AEW ac- bances that developed into Atlantic hurricanes. Mon. Wea. tivity varies on both interannual (e.g., Grist 2002)and Rev., 97, 256–276, doi:10.1175/1520-0493(1969)097,0256: intraseasonal (Janicot et al. 2011; Alaka and Maloney SHOTAD.2.3.CO;2. 2012) time scales. There is evidence that the Madden– Chauvin, F., J.-F. Royer, and H. Douville, 2005: Interannual vari- Julian oscillation (Matthews 2004)andtheAsian ability and predictability of African easterly waves in a GCM. Climate Dyn., 24, 523–544, doi:10.1007/s00382-004-0507-9. summer monsoon (Janicot et al. 2011, 2009) modulate Collins, W. D., and Coauthors, 2006: The Community Climate African rainfall and AEW activity on intraseasonal System Model version 3 (CCSM3). J. Climate, 19, 2122–2143, time scales. doi:10.1175/JCLI3761.1. 8340 JOURNAL OF CLIMATE VOLUME 27

Cook, K. H., and E. K. Vizy, 2006: Coupled model simulations of Hopsch, S. B., C. D. Thorncroft, and K. R. Tyle, 2010: Analysis of the West African monsoon system: Twentieth- and twenty-first- African easterly wave structures and their role in influencing century simulations. J. Climate, 19, 3681–3703, doi:10.1175/ . Mon. Wea. Rev., 138, 1399–1419, doi:10.1175/ JCLI3814.1. 2009MWR2760.1. Daloz, A. S., F. Chauvin, K. Walsh, S. Lavender, D. Abbs, and Hsieh, J.-S., and K. H. Cook, 2005: Generation of African easterly F. Roux, 2012: The ability of general circulation models to wave disturbances: Relationship to the African easterly jet. simulate tropical cyclones and their precursors over the North Mon. Wea. Rev., 133, 1311–1327, doi:10.1175/MWR2916.1. Atlantic main development region. Climate Dyn., 39, 1559– ——, and ——, 2007: A study of the energetics of African easterly 1576, doi:10.1007/s00382-012-1290-7. waves using a regional climate model. J. Atmos. Sci., 64, 421– Dee, D. P., and Coauthors, 2011: The ERA-Interim reanalysis: 440, doi:10.1175/JAS3851.1. Configuration and performance of the data assimilation system. ——, and ——, 2008: On the instability of the African easterly jet Quart. J. Roy. Meteor. Soc., 137, 553–597, doi:10.1002/qj.828. and the generation of African waves: Reversals of the potential DeMott, C. A., C. Stan, D. A. Randall, J. L. Kinter III, and vorticity gradient. J. Atmos. Sci., 65, 2130–2151, doi:10.1175/ M. Khairoutdinov, 2011: The Asian monsoon in the super- 2007JAS2552.1. parameterized CCSM and its relationship to tropical wave ac- Huffman, G. J., and Coauthors, 2007: The TRMM Multisatellite tivity. J. Climate, 24, 5134–5156, doi:10.1175/2011JCLI4202.1. Precipitation Analysis: Quasi-global, multiyear, combined- ——, ——, and ——, 2013: Northward propagation mechanisms sensor precipitation estimates at fine scale. J. Hydrometeor., of the boreal summer intraseasonal oscillation in the ERA- 8, 38–55, doi:10.1175/JHM560.1. Interim and SP-CCSM. J. Climate, 26, 1973–1992, doi:10.1175/ Janicot, S., F. Mounier, N. M. Hall, S. Leroux, B. Sultan, and G. N. JCLI-D-12-00191.1. Kiladis, 2009: Dynamics of the West African monsoon. Part Diaz, M., and A. Aiyyer, 2013: The genesis of African easterly IV: Analysis of 25–90-day variability of convection and the waves by upstream development. J. Atmos. Sci., 70, 3492– role of the Indian monsoon. J. Climate, 22, 1541–1565, 3512, doi:10.1175/JAS-D-12-0342.1. doi:10.1175/2008JCLI2314.1. Diedhiou, A., S. Janicot, A. Viltard, and P. de Félice, 2002: Ener- ——, and Coauthors, 2011: Intraseasonal variability of the West Af- getics of easterly wave disturbances over West Africa and the rican monsoon. Atmos. Sci. Lett., 12, 58–66, doi:10.1002/asl.280. tropical Atlantic: A climatology from the 1979–95 NCEP/ Janiga, M. A., and C. D. Thorncroft, 2013: Regional difference in the NCAR reanalyses. Climate Dyn., 18, 487–500, doi:10.1007/ kinematic and thermodynamic structure of African easterly s00382-001-0195-7. waves. Quart. J. Roy. Meteor Soc., 139, 1598–1614, doi:10.1002/ Duvel, J. P., 1990: Convection over tropical Africa and the Atlantic qj.2047. Ocean during northern summer. Part II: Modulation by east- Khairoutdinov, M. F., and D. A. Randall, 2001: A cloud resolving erly waves. Mon. Wea. Rev., 118, 1855–1868, doi:10.1175/ model as a cloud parameterization in the NCAR Community 1520-0493(1990)118,1855:COTAAT.2.0.CO;2. Climate System Model: Preliminary results. Geophys. Res. Fink, A. H., and A. Reiner, 2003: Spatiotemporal variability of the Lett., 28, 3617–3620, doi:10.1029/2001GL013552. relation between African easterly jets and West African squall ——, and ——, 2003: Cloud resolving modeling of the ARM lines in 1998 and 1999. J. Geophys. Res., 108, 4332, doi:10.1029/ summer 1997 IOP: Model formulation, results, uncertainties, 2002JD002816. and sensitivities. J. Atmos. Sci., 60, 607–625, doi:10.1175/ Fyfe, J. C., 1999: Climate simulations of African easterly waves. 1520-0469(2003)060,0607:CRMOTA.2.0.CO;2. J. Climate, 12, 1747–1769, doi:10.1175/1520-0442(1999)012,1747: ——, ——, and C. DeMott, 2005: Simulations of the atmospheric CSOAEW.2.0.CO;2. general circulation using a cloud-resolving model as a super- Grabowski, W. W., 2001: Coupling cloud processes with the large- parameterization of physical processes. J. Atmos. Sci., 62, scale dynamics using the cloud-resolving convection parame- 2136–2154, doi:10.1175/JAS3453.1. terization (CRCP). J. Atmos. Sci., 58, 978–997, doi:10.1175/ ——, C. DeMott, and D. Randall, 2008: Evaluation of the simulated 1520-0469(2001)058,0978:CCPWTL.2.0.CO;2. interannual and subseasonal variability in an AMIP-style sim- ——, and P. K. Smolarkiewicz, 1999: CRCP: A cloud resolving ulation using the CSU multiscale modeling framework. convection parameterization for modeling the tropical con- J. Climate, 21, 413–431, doi:10.1175/2007JCLI1630.1. vecting atmosphere. Physica D, 133, 171–178, doi:10.1016/ Kiladis, G. N., C. D. Thorncroft, and N. M. J. Hall, 2006: Three- S0167-2789(99)00104-9. dimensional structure and dynamics of African easterly waves. Grist, J. P., 2002: Easterly waves over Africa. Part I: The seasonal Part I: Observations. J. Atmos. Sci., 63, 2212–2230, doi:10.1175/ cycle and contrasts between wet and dry years. Mon. Wea. JAS3741.1. Rev., 130, 197–211, doi:10.1175/1520-0493(2002)130,0197: ——, M. C. Wheeler, P. T. Haertel, K. H. Straub, and P. E. Roundy, EWOAPI.2.0.CO;2. 2009: Convectively coupled equatorial waves. Rev. Geophys., Gu, G., R. F. Adler, G. J. Huffman, and S. Curtis, 2004: African 47, RG2003, doi:10.1029/2008RG000266. easterly waves and their association with precipitation. J. Geo- Liebmann, B., and C. A. Smith, 1996: Description of a complete phys. Res., 109, D04101, doi:10.1029/2003JD003967. (interpolated) outgoing longwave radiation dataset. Bull. Hall, N. M. J., G. N. Kiladis, and C. D. Thorncroft, 2006: Three- Amer. Meteor. Soc., 77, 1275–1277. dimensional structure and dynamics of African easterly waves. ——, I. Bladé, G. N. Kiladis, L. M. V. Carvalho, G. Senay, D. Allured, Part II: Dynamical modes. J. Atmos. Sci., 63, 2231–2245, and S. Leroux, 2012: Seasonality of African precipitation doi:10.1175/JAS3742.1. from 1996 to 2009. J. Climate, 25, 4304–4322, doi:10.1175/ Hodges, K. I., B. J. Hoskins, J. Boyle, and C. D. Thorncroft, 2003: JCLI-D-11-00157.1. A comparison of recent reanalysis datasets using objective fea- Lin, X., and R. H. Johnson, 1996: Kinematic and thermodynamic ture tracking: Storm tracks and tropical easterly waves. Mon. Wea. characteristics of the flow over the western Pacific warm pool Rev., 131, 2012–2037, doi:10.1175/1520-0493(2003)131,2012: during TOGA COARE. J. Atmos. Sci., 53, 695–715, doi:10.1175/ ACORRD.2.0.CO;2. 1520-0469(1996)053,0695:KATCOT.2.0.CO;2. 15 NOVEMBER 2014 M C C R A R Y E T A L . 8341

Mathon, V., H. Laurent, and T. Lebel, 2002: Mesoscale convective model of the tropical Atlantic. J. Climate, 21, 1417–1431, system rainfall in the Sahel. J. Appl. Meteor., 41, 1081–1092, doi:10.1175/2007JCLI1906.1. doi:10.1175/1520-0450(2002)041,1081:MCSRIT.2.0.CO;2. Serra, Y. L., G. N. Kiladis, and M. F. Cronin, 2008: Horizontal and Matthews, A. J., 2004: Intraseasonal variability over tropical vertical structure of easterly waves in the Pacific ITCZ. Africa during northern summer. J. Climate, 17, 2427–2440, J. Atmos. Sci., 65, 1266–1284, doi:10.1175/2007JAS2341.1. doi:10.1175/1520-0442(2004)017,2427:IVOTAD.2.0.CO;2. Skinner, C. B., and N. S. Diffenbaugh, 2013: The contribution of McCrary, R. R., 2012: Seasonal, synoptic and intraseasonal vari- African easterly waves to monsoon precipitation in the CMIP3 ability of the West African monsoon. Ph.D. dissertation, ensemble. J. Geophys. Res. Atmos., 118, 3590–3609, doi:10.1002/ Colorado State University, 157 pp. jgrd.50363. ——, D. A. Randall, and C. Stan, 2014: Simulations of the West Smith, R. D., and P. R. Gent, Eds., 2002: Reference manual for the African monsoon with a superparameterized climate model. Parallel Ocean Program (POP): Ocean component of the Part I: The seasonal cycle. J. Climate, 27, 8303–8322, Community Climate System Model (CCSM2.0 and 3.0). Los doi:10.1175/JCLI-D-13-00676.1. Alamos National Laboratory Tech Rep. LA-UR-02-2484, 75 Mekonnen, A., C. D. Thorncroft, and A. R. Aiyyer, 2006: Analysis pp. [Available online at http://www.cesm.ucar.edu/models/ of convection and its association with African easterly waves. ccsm3.0/pop/doc/manual.pdf.] J. Climate, 19, 5405–5421, doi:10.1175/JCLI3920.1. Stan, C., M. Khairoutdinov, C. A. DeMott, V. Krishnamurthy, Pohl, B., and H. Douville, 2011: Diagnosing GCM errors over West D. M. Straus, D. A. Randall, J. L. Kinter III, and J. Shukla, Africa using relaxation experiments. Part I: Summer monsoon 2010: An ocean–atmosphere climate simulation with an em- climatology and interannual variability. Climate Dyn., 37, bedded cloud resolving model. Geophys. Res. Lett., 37, L01702, 1293–1312, doi:10.1007/s00382-010-0911-2. doi:10.1029/2009GL040822. Randall, D. A., C. DeMott, C. Stan, M. Khairoutdinov, J. Benedict, Sylla, M. B., I. Diallo, and J. S. Pal, 2013: West African monsoon in R. McCrary, and K. Thayer-Calder, 2014: Simulations of the state-of-the-science regional climate models. Climate Variability— Tropical General Circulation with a Multiscale Global Model. Regional and Thematic Patterns, A. Tarhule, Ed., InTech, Meteor. Monogr., Amer. Meteor. Soc., submitted. doi:10.5772/55140. Reale, O., J. Terry, M. Masutani, E. Andersson, L. P. Riishojgaard, Thorncroft, C. D., and B. J. Hoskins, 1994a: An idealized study of and J. C. Jusem, 2007: Preliminary evaluation of the European African easterly waves. I: A linear view. Quart. J. Roy. Meteor. Centre for Medium-Range Weather Forecasts’ (ECMWF) Na- Soc., 120, 953–982, doi:10.1002/qj.49712051809. ture Run over the tropical Atlantic and African monsoon region. ——, and ——, 1994b: An idealized study of African easterly Geophys. Res. Lett., 34, L22810, doi:10.1029/2007GL031640. waves. II: A nonlinear view. Quart. J. Roy. Meteor. Soc., 120, Reed, R. J., D. C. Norquist, and E. E. Recker, 1977: The structure 983–1015, doi:10.1002/qj.49712051810. and properties of African wave disturbances as observed during ——, and K. I. Hodges, 2001: African easterly wave variability and phase III of GATE. Mon. Wea. Rev., 105, 317–333, doi:10.1175/ its relationship to Atlantic activity. J. Cli- 1520-0493(1977)105,0317:TSAPOA.2.0.CO;2. mate, 14, 1166–1179, doi:10.1175/1520-0442(2001)014,1166: Roehrig, R., D. Bouniol, F. Guichard, F. Hourdin, and J.-L. AEWVAI.2.0.CO;2. Redelsperger, 2013: The present and future of the West Af- ——,N.M.J.Hall,andG.N.Kiladis,2008:Three-dimensional rican monsoon: A process-oriented assessment of CMIP5 structure and dynamics of African easterly waves. Part III: Gen- simulations along the AMMA transect. J. Climate, 26, 6471– esis. J. Atmos. Sci., 65, 3596–3607, doi:10.1175/2008JAS2575.1. 6505, doi:10.1175/JCLI-D-12-00505.1. Vizy, E. K., and K. H. Cook, 2002: Development and application Ruti, P. M., and A. Dell’Aquila, 2010: The twentieth century Af- of a mesoscale climate model for the tropics: Influence of rican easterly waves in reanalysis systems and IPCC simula- sea surface temperature anomalies on the West African tions, from intra-seasonal to inter-annual variability. Climate monsoon. J. Geophys. Res., 107 (D3), doi:10.1029/ Dyn., 35, 1099–1117, doi:10.1007/s00382-010-0894-z. 2001JD000686. Sander, N., and S. C. Jones, 2008: Diagnostic measures for assessing Wheeler, M., and G. N. Kiladis, 1999: Convectively coupled numerical forecasts of African easterly waves. Meteor. Z., 17, equatorial waves: Analysis of clouds in the wavenumber– 209–220, doi:10.1127/0941-2948/2008/0269. frequency domain. J. Atmos. Sci., 56, 374–399, doi:10.1175/ Schubert, W. H., P. E. Ciesielsky, D. E. Stevens, and H.-C. Kuo, 1520-0469(1999)056,0374:CCEWAO.2.0.CO;2. 1991: Potential vorticity modeling of the ITCZ and the Hadley Yanai, M., 1961: Dynamical aspects of typhoon formation. J. Me- circulation. J. Atmos. Sci., 48, 1493–1509, doi:10.1175/ teor. Soc. Japan, 38, 283–309. 1520-0469(1991)048,1493:PVMOTI.2.0.CO;2. Zhang, G. J., and N. McFarlane, 1995: Role of convective scale Seo, H., M. Jochum, R. Murtugudde, A. J. Miller, and J. O. Roads, momentum transport in climate simulation. J. Geophys. Res., 2008: Precipitation from African easterly waves in a coupled 100, 1417–1426, doi:10.1029/94JD02519.