modulation of daily precipitation over southwest asia by the …blyon/papers/mwr_sw_asia_and... ·...

Modulation of Daily Precipitation over Southwest Asia by theMadden–Julian Oscillation

MATHEW BARLOW

Atmospheric and Environmental Research, Inc., Lexington, Massachusetts

MATTHEW WHEELER

Bureau of Meteorology Research Centre, Melbourne, Australia

BRADFIELD LYON

International Research Institute for Climate Prediction, Palisades, New York

HEIDI CULLEN

The Weather Channel, and Georgia Institute of Technology, Atlanta, Georgia

(Manuscript received 14 October 2003, in final form 18 March 2005)

ABSTRACT

Analysis of daily observations shows that wintertime (November–April) precipitation over SouthwestAsia is modulated by Madden–Julian oscillation (MJO) activity in the eastern Indian Ocean, with strengthcomparable to the interannual variability. Daily outgoing longwave radiation (OLR) for 1979–2001 is usedto provide a long and consistent, but indirect, estimate of precipitation, and daily records from 13 stationsin Afghanistan reporting at least 50% of the time for 1979–85 are used to provide direct, but shorter andirregularly reported, precipitation data. In the station data, for the average of all available stations, there isa 23% increase in daily precipitation relative to the mean when the phase of the MJO is negative (sup-pressed tropical convection in the eastern Indian Ocean), and a corresponding decrease when the MJO ispositive. The distribution of extremes is also affected such that the 10 wettest days all occur during thenegative MJO phase. The longer record of OLR data indicates that the effect of the MJO is quite consistentfrom year to year, with the anomalies averaged over Southwest Asia more negative (indicating more rain)for the negative phase of the MJO for each of the 22 yr in the record. Additionally, in 9 of the 22 yr theaverage influence of the MJO is larger than the interannual variability (e.g., the relationship results inanomalously wet periods even in dry years and vice versa).

Examination of NCEP–NCAR reanalysis data shows that the MJO modifies both the local jet structureand, through changes to the thermodynamic balance, the vertical motion field over Southwest Asia, con-sistent with the observed modulation of the associated synoptic precipitation. A simple persistence schemefor forecasting the sign of the MJO suggests that the modulation of Southwest Asia precipitation may bepredictable for 3-week periods. Finally, analysis of changes in storm evolution in Southwest Asia due to theinfluence of the MJO shows a large difference in strength as the storms move over Afghanistan, withapparent relevance for the flooding event of 12–13 April 2002.

1. Introduction

Southwest (SW) Asia, taken here as the region cen-tered on Iran, Afghanistan, and Pakistan, is a generally

arid region with high mountains. Except for northernPakistan, which is influenced by the summer monsoon,precipitation in the region is primarily associated witheastward-moving synoptic storms during the winter andearly spring (Martyn 1992). These relatively infrequentstorms are of high importance, both in terms of waterresources and agriculture and in terms of flooding andtransportation hazards. Previous analysis has shownthat tropical convection in the eastern Indian Ocean

Corresponding author address: Dr. Mathew Barlow, Universityof Massachusetts—Lowell, One University Avenue, Lowell, MA01854.E-mail: [email protected]

DECEMBER 2005 B A R L O W E T A L . 3579

© 2005 American Meteorological Society

can affect the local winds and precipitation of the re-gion at seasonal time scales during the cold season(Barlow et al. 2002). The tropical convection of theeastern Indian Ocean also has prominent variability onthe intraseasonal time scale of the Madden–Julian os-cillation (MJO) (e.g., Wang and Rui 1990). This, to-gether with the importance of synoptic time scales inSW Asia precipitation, suggests that important aspectsof the relationship between the two regions may beactive at daily to intraseasonal time scales. Theseweather–climate relationships may provide a basis forforecasting at the time scales between synoptic and sea-sonal.

The climate of SW Asia ranges from desert condi-tions in interior Iran, southwestern Afghanistan, andsoutheastern Pakistan, to semiarid steppe in much ofthe rest of the region, although the vegetation can berelatively lush along the southern shores of the Caspianand in the snowmelt-fed river valleys. In northern Pak-istan, the primary rainfall season is summer, associatedwith the northernmost advance of the Asian monsoon.For the rest of the region, winter and early spring (No-vember–April) is the primary rainfall period (Fig. 1a).The complex terrain of the high mountain ranges inIran and the Hindu Kush in Afghanistan and northernPakistan (Fig. 2b) plays a key role in the distribution ofprecipitation, which is heaviest on the windward slopes.Yearly precipitation can exceed 1 m in some areas ofthe high mountains of western Iran and the HinduKush, as well as along the southern shore of the Cas-pian Sea (not resolved in the gridded precipitationshown in Fig. 1).

During the Northern Hemisphere winter and springseasons, SW Asia is within the subtropical belt of up-per-air westerlies (e.g., Krishnamurti 1961), which bringthe moisture-bearing synoptic storms that are the pri-mary precipitation mechanism for much of the region.The maxima in these westerly winds, or jets, are asso-ciated with regions of synoptic storm activity, some-times called “cyclone belts,” “storm tracks” (Blackmonet al. 1977), or “baroclinic waveguides” (Wallace et al.1988). A local maximum of synoptic activity exists overthe Mediterranean (Petterssen 1958) in association witha baroclinic waveguide (Wallace et al. 1988); occasionalextensions of this storm track are associated with themovement of synoptic storms into the region.

On average, SW Asia lies between two maxima in thesubtropical westerlies: in the exit region of the NorthAfrica/Arabian jet (nomenclature varies) and near theentrance region of the east Asian jet. The upper-levelwesterlies have a complex relationship with the stormactivity. Shear, deformation, and maximum speed ofthe upper-level wind field are all known to affect tran-

sient activity. Jet exit regions are of particular interestfrom a number of perspectives. The strong horizontalgradients of wind speed at the entrance and exit regionsof the jet are typically balanced by vertical circulations,which can affect the amount and strength of storm ac-tivity (Blackmon et al. 1977). The exit region of a jetcan serve as a source or sink of energy, via barotropicinstability, for transients (e.g., Simmons et al. 1983;Branstator 1985), with the sign of the energy transferdependent on the horizontal deformation of the tran-sient activity. The jet exit is also a preferred region forbaroclinic energy conversion, as the decrease in shearallows the development of local instability modes (Caiand Mak 1990). Indeed, the net effect of the winds onthe storm track appears to result from competing con-tributions from baroclinic and barotropic energetics(Cai and Mak 1990; Whitaker and Dole 1995). The up-per-level westerly flow is also reinforced by the diabaticheating of the storms (Hoskins and Valdes 1990),though due to the relatively modest precipitationamounts of SW Asia, it may not be a large factor in thiscase. Finally, we note that analysis of cold surges, whichare also linked with the regional jet structure and tropi-cal rainfall in Indonesia, has suggested interactions be-tween the Asian jet and the North African/Arabian jet(Chang and Lau 1982).

This intermediate position with respect to the localjet features appears to render the local precipitationsensitive to large-scale variability over a wide geo-graphic range. Influences have been identified from At-lantic-based (Aizen et al. 2001) and Pacific-based vari-ability (Nazemosadat and Cordery 2000; Hoerling andKumar 2003; Tippett et al. 2003), as well as from localSST variability in the Persian Gulf (Nazemosadat1998). Tropical variability in the eastern Indian Oceanin particular has been identified as effective in forcingthe extratropical circulation (Sardeshmukh and Hos-kins 1988; Ting and Sardeshmukh 1993) and, specifi-cally, winter precipitation over SW Asia (Barlow et al.2002). Reduced SW Asian seasonal precipitation occursin conjunction with enhanced seasonal convection overthe eastern Indian Ocean. The eastern Indian Ocean isalso a preferred region for MJO activity and, indeed,examination of some previous analyses of MJO evolu-tion does show out-of-phase anomalies between thetwo regions [e.g., Fig. 3 in Jones et al. (1998) and Fig. 4in Lo and Hendon (2000)].

The observed pattern of the MJO winds is similar toa convectively coupled Kelvin–Rossby mode (e.g., Ruiand Wang 1990; Hendon and Salby 1994). While thedynamics of the MJO are not yet fully understood, thisinterpretation is supported by several theoretical stud-ies (e.g., Yamagata and Hayashi 1984; Hendon 1988;

3580 M O N T H L Y W E A T H E R R E V I E W VOLUME 133

Wang and Rui 1989; Hendon and Salby 1996; Moskow-itz and Bretherton 2000). The two Rossby gyres to thewest of the tropical rainfall anomalies have a first-baroclinic-mode structure with opposite-signed circula-

tions at upper and lower levels, and are very similar tothe steady linear response expected from the shallow-water equations with no mean wind (Gill 1980). Whenthe MJO tropical rainfall anomalies are in the eastern

FIG. 1. (a) Nov–Apr average precipitation, contoured at intervals of 20 cm, from the Newet al. (2000) gridded data. Red numbers show station locations for data used in section 3. (b)Topography, contoured at intervals of 1 km.


Indian Ocean, the Northern Hemisphere Rossby gyreextends over SW Asia; thus, the MJO wind anomaliescan affect SW Asia in terms of both changes to theupper-level winds as well as the low-level winds (and,hence, moisture transport).

The relationship between the Northern HemisphereRossby gyre of the MJO and the midlatitude westerlieshas some similarities to the boreal summer analysis ofRodwell and Hoskins (1996, 2001), where, in a simpli-fied model, the Asian monsoon heating forces a Rossbywave response to the west, which interacts with thewesterlies to produce descent over the eastern Saharaand Mediterranean and the Kyzylkum desert (encom-passing SW Asia, as defined here). Although the Asianmonsoon heating is considerably north of the equatorwhile the MJO heating is roughly equatorially symmet-ric, the summertime westerlies are also displaced north-ward of the wintertime westerlies and the MJO heatinghas considerable latitudinal extent, so that the intersec-tion of the Rossby wave response with the westerliesappears similar between the two cases.

The MJO has been shown to modulate synoptic-scaleweather and/or its associated precipitation in several

regions, including the Americas (Nogues-Paegle andMo 1997; Mo and Higgins 1998a,b,c; Mo 1999; Jones2000; Higgins et al. 2000; Nogues-Paegle et al. 2000;Whitaker and Weickmann 2001; Bond and Vecchi2003), Australia (Hendon and Liebmann 1990; Wheelerand Hendon 2004), and India (Goswami et al. 2003).This modulation, together with demonstrated predic-tive skill for the MJO out to 3 weeks (Waliser et al.1999; Lo and Hendon 2000; Wheeler and Weickmann2001; Mo 2001) and potentially a month (Waliser et al.2003), suggests the possibility of using the MJO as abasis for predictions at time scales between long-rangeweather forecasting and short-term climate prediction(Schubert et al. 2002). The nature of the relationshipbetween the MJO and Southwest Asia precipitation,however, and of any related predictability has not yetbeen explored.

Here we examine this relationship using daily winds,vertical velocity, and outgoing longwave radiation(OLR) for the regional analysis, complemented bydaily station precipitation for Afghanistan. The dataare described in section 2. The strength and consistencyof the precipitation relationship is analyzed in section 3.

FIG. 2. Difference in daily OLR and 200-hPa wind anomalies between the positive andnegative phases of the MJO in the eastern Indian Ocean (red box), for Nov–Apr 1979–2001.The OLR anomalies are contoured at intervals of 4 W m�2. Negative OLR anomalies, whichcorrespond to positive precipitation anomalies, are shaded green, while positive OLR anoma-lies, which correspond to negative precipitation anomalies, are shaded brown. As the com-posite is based on the differences between the two phases from daily data, the average dailyanomaly during a particular phase would be half that shown. To estimate the net anomalyresulting from a particular phase, the daily average should be multiplied by the length of thephase (approximately 21 days).


Changes in dynamic processes, including the jet struc-ture and the thermodynamic balance, are examined insection 4. The potential for predicting SW Asia precipi-tation based on the MJO is explored in section 5. Theimportance of accounting for synoptic timing when di-agnosing the influence of the MJO is examined in sec-tion 6. Finally, a summary and discussion are given insection 7.

2. Data

a. Precipitation estimates

Both OLR and station precipitation data from Af-ghanistan are used to estimate daily precipitation forNovember–April, the main precipitation period formuch of SW Asia. OLR provides an indirect but rela-tively long, continuous, and real-time proxy for precipi-tation, while the station reports provide direct, butshort, irregularly reported, and geographically limiteddata. OLR responds to cloud activity and is frequentlyused as part of precipitation estimation algorithms. Al-though OLR is most closely associated with deep tropi-cal convection, it is also related to precipitation in themidlatitudes and at daily time scales (Arkin and Meis-ner 1987; Huffman et al. 2001). We have used dailystation data in the region (described below) to furtherverify that daily OLR is a useful precipitation proxy.The daily OLR produced by the National Oceanic andAtmospheric Administration–Cooperative Institute forResearch in Environmental Sciences (NOAA–CIRES)Climate Diagnostics Center (Liebmann and Smith1996) at 2.5° � 2.5° latitude–longitude resolution isused for the 1979–2002 period, November–April. Thir-

teen stations reporting at least 50% of the time areavailable for 1979–85 for Afghanistan from the GlobalSummary of the Day (GSOD) (Lott 1998). The stationnames, locations, elevations, November–April averageprecipitation, and annual average precipitation aregiven in Table 1. Although the station data have noformal quality control, examination of consistency withOLR and wind anomalies as well as visual inspectionsuggests that the quality is satisfactory for the analysisundertaken here, although the number and irregularityof the missing values preclude confident assessment ofyear-to-year variability in the station data. The dailymean climatology has been removed from the OLR.

b. Winds and vertical velocity

Daily winds and vertical velocity are taken from theNational Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis (Kalnay et al. 1996) at 2.5° � 2.5°latitude–longitude resolution for November–April,1979–2002. The 200-hPa level has proven a useful diag-nostic level for both the response to tropical convectionanomalies and the jet-level dynamics of synopticstorms, which are the primary precipitation mechanismfor much of SW Asia (Martyn 1992). Both the totalfields and anomalies are considered; for the anomaliesthe daily mean climatology for the 1979–2001 periodwas removed.

c. MJO estimate

The estimates of the MJO are produced by Austra-lia’s Bureau of Meteorology Research Centre (BMRC)based on the technique of Wheeler and Weickmann

TABLE 1. Afghanistan stations used in study. The average precipitation values shown were calculated from the daily precipitationvalues for the 1979–85 period. Station locations are shown in Fig. 1. Note that precipitation in the region is very irregular, and averagesduring this period may not be representative of other periods.

Station name

WorldMeteorologicalOrganization(WMO) ID Lon Lat

Elevation(m)

Avg Nov–Aprprecipitation

(cm)

Avg annualprecipitation

(cm)

1) Faizabad 40904 70°31�E 37°07�N 1200 42.82 50.202) Mazari Sharif 40911 67°12�E 36°42�N 378 18.64 21.783) Kunduz 40913 68°55�E 36°40�N 433 18.48 24.034) Maimana 40922 64°45�E 35°55�N 815 25.13 27.715) North Salang 40930 69°01�E 35°19�N 3366 43.44 50.396) Herat 40938 62°13�E 34°13�N 964 33.58 34.437) Chakahoharan 40942 65°16�E 34°32�N 2230 15.44 20.328) Kabul 40948 69°13�E 34°33�N 1791 25.57 30.189) Jalalabad 40954 70°28�E 34°26�N 580 17.37 19.59

10) Ghazni 40968 68°25�E 33°32�N 2183 21.73 26.8311) Farah 40974 62°11�E 32°22�N 700 7.02 7.0312) Bust 40988 64°22�E 31°33�N 780 22.61 23.0713) Kandahar 40990 65°51�E 31°30�N 1010 25.04 25.24


(2001), which filters daily OLR data for the character-istic eastward-propagating zonal wavenumbers (1–5)and periodicities (30 to 96 days) of the MJO. Two es-timates are used: a more accurate “diagnostic” estimatethat includes both past and future data in calculatingthe value for a given time and so cannot be produced inreal time, and a “monitoring” or 0-day forecast, whichis the estimate that would have been operationallyavailable in real time. For the former we have used theperiod from 1979 to 2001, and for the latter, the periodfrom 1980 to 1995. An index of MJO convection in theeastern Indian Ocean is created by averaging the datafor the rectangular region 15°S–10°N, 80°–100°E.

3. SW Asia precipitation and the sign of the MJO

A very simple measure of the phase of the MJO is thesign, positive or negative, of the associated tropical con-vection anomaly averaged over a given area, using theWheeler and Weickmann (2001) data described in theprevious section. The period of the MJO is roughly30–60 days (e.g., Madden and Julian 1994), so it main-tains the same sign for 2–3 weeks at a time. We beginour analysis by averaging all the days from Novemberto April 1979–2001 into two groups based on whetherthe MJO convection anomaly in the eastern IndianOcean (averaged 15°S–10°N, 80°–100°E) is positive ornegative. The averaging region is similar to the “IPX”[Indian Ocean precipitation extension] region of Bar-low et al. (2002) but modified slightly to better matchthe local maximum in MJO activity. The exact defini-tion of the averaging region does not substantially af-fect the results. For the purposes of this analysis, werefer to enhanced convection (negative OLR anoma-lies) in the eastern Indian Ocean as the “positivephase” and suppressed convection (positive OLRanomalies) as the “negative phase.” Figure 2 shows thedifference in daily OLR and 200-hPa winds betweenthe positive and negative phases of the MJO in theeastern Indian Ocean (red box) for November–April1979–2001. Here we use the diagnostic MJO informa-tion; use of the information that would have been avail-able in real time results in only a modest reduction insignal.

The pattern of OLR and winds in Fig. 2 is consistentwith previous research, with upper-level winds having alarge response to the north and west of the tropicalconvection, providing a direct link between the easternIndian Ocean and SW Asia in the wind field. There isalso good correspondence with the steady-state linearresponse to diabatic heating in the eastern IndianOcean shown in Barlow et al. (2002). The larger am-

plitude of the Rossby gyre in the Northern Hemisphereis consistent with the stronger mean winds in the winterhemisphere (Jin and Hoskins 1995). For the negativephase (suppressed tropical convection), the circulationanomalies are opposite in sign. The lower-level winds(not shown) are also consistent with a Rossby responseand have a weak circulation opposite to that of theupper-level winds.

Although modest in strength compared to the tropi-cal anomaly, the OLR signal over SW Asia (blue box)has high statistical significance: when the same calcula-tion as for Fig. 2 was based on 1000 randomly generatedMJO time series, the same magnitude of signal wasobtained less than 0.1% of the time. (The random timeseries were generated by maintaining the power spec-trum of the MJO time series while randomizing thephase of each spectral component. This provides timeseries that are random but have the same temporalcharacteristics as the original signal.)

Even though the OLR anomalies over SW Asia ap-pear small, the fact that the average precipitation in theregion is also small means that MJO-related anomaliesmay still represent a significant modulation of the localprecipitation. This is confirmed in the station data,where the average over all stations during the negativephase of the MJO (1.66 mm day�1) is increased by 23%relative to the mean, while the average value during thepositive phase (1.07 mm day�1) is decreased by 21%relative to the mean. This relationship was also verifiedseparately for the 1979–81 and 1982–85 subperiods ofthe available station data. This difference is also con-sistent among the stations (Fig. 3), with 11 of the 13showing the same relationship, 7 stations showingchanges of at least 20%, and 3 stations differing bymore than 50% (more than a factor of 2 differencebetween phases). As expected from the pattern of OLRanomalies in Fig. 2, there is considerable geographicvariation in the signal over the country—the largestdifferences are in those stations on the west and south-ern flanks of the Hindu Kush (station positions areshown in red in Fig. 1), where the OLR anomalies arealso largest, with considerably smaller differences at theother stations, which are in the north and east of Af-ghanistan. Although the shortness of record and num-ber of missing values in the station data necessitate acautious interpretation, the available station data sug-gest that the MJO has a large effect and the OLRanomaly pattern suggests that these Afghanistan sta-tion differences may not even reflect the largest part ofthe signal. Unfortunately, the unavailability of compa-rable station data in the other countries currently pre-cludes an assessment of this; obtaining more stationdata will be a focus of future efforts.


Finally, we consider the influence of the MJO withrespect to the wettest days in the average over the 13stations. Of the 10 wettest days, all occurred during thenegative phase of the MJO. No 2 of the 10 wettest daysoccurred within 7 days of each other, so the relationshipis not just the result of one or two events. Based oncalculations with 1000 MJO-like random time series(generated as described previously), the probability ofthe 10 wettest days all falling within the same phase bychance is 0.1%.

4. Changes in dynamic process

a. Jet structure

As seen in Fig. 2, the MJO has an effect on the jet-level winds over the region, and so may affect the jetdynamics. The magnitude of the wind field at 200 hPa isshown for the two phases of the MJO in Fig. 4, indicat-ing variations in the jet structure over SW Asia. Al-though the jet-level winds over SW Asia increase dur-ing the positive phase of the MJO, the local precipita-tion decreases. While the MJO anomalies increase thelocal wind speed aloft, these anomalies occur in the exitregion of the North Africa/Arabian jet maximum andresult in a decrease in the gradient of wind speed, pro-

ducing a more diffuse jet exit region extending fromsouthern Iran through northern India (Fig. 4a). Theopposite occurs during the negative MJO, when thenegative wind anomalies enhance the jet exit region(Fig. 4b). The presence of an exit region is associatedwith a local mode of baroclinic instability (Cai and Mak1990) and so we may expect that an enhanced exit re-gion would be associated with an increase in transientactivity and, hence, increased precipitation.

b. Vertical motion and thermodynamic balance

Large-scale vertical velocity can have an importantinfluence on both tropical and extratropical precipita-tion (e.g., Cotton and Anthes 1989; Bluestein 1993).The changes in vertical velocity associated with theMJO are shown in Fig. 5 at 500 and 300 hPa. Midtro-pospheric levels are appropriate for diagnosing verticalmotions associated with both tropical variability (e.g.,Peixoto and Oort 1992) and synoptic storms (e.g., Limand Wallace 1991). For convenience, the following dis-cussion is with respect to the positive phase of the MJO(enhanced tropical convection in the eastern IndianOcean and suppressed synoptic precipitation overSouthwest Asia); the same arguments apply, with op-posite sign, to the negative phase.

FIG. 3. Afghanistan average daily rainfall (mm day�1) for 13 stations, averaged duringpositive (red) and negative (blue) phases of the MJO in the eastern Indian Ocean, Nov–Apr1979–85. All stations have daily reports for at least 50% of the days. The station locations areshown in Fig. 1.


As expected, vigorous upward motion occurs in theregion of enhanced tropical convection, with largestvalues at 300 hPa, consistent with previous analysis ofvertical velocity associated with Tropical-type convec-tion in the NCEP data (e.g., Barlow et al. 1998).Anomalous subsidence is anticipated over SW Asia, aslocal precipitation has decreased and this is, indeed,observed. Interestingly, however, the magnitude of thedescent is quite large, comparable in size to the tropicalanomalies. Given that changes in regional precipitation,while an important fraction of the local average, aremodest in terms of absolute value, the vigor of the sub-sidence is surprising, as is the occurrence of the largestvalues at 300 hPa. The large values of descent are alsoconsiderably in excess of what might be expected by“Hadley”-type forcing from the tropical convection.While the upper-level horizontal winds associated withthe MJO (Fig. 2) are in good agreement with the simpleshallow-water framework of Gill–Matsuno, the ex-pected descent in the idealized case for equatorially

symmetric heating is very diffuse, with largest valuesonly 1/6 of the maximum ascent (Gill 1980), much lessthan observed here.

What, then, is causing the large values of subsidence?To investigate the dynamics of the descent region, weexamine the thermodynamic energy balance. The hy-drostatic thermodynamic energy equation may be writ-ten (e.g., Holton 1992)

�T

�t� �V · �T � Sp� �

J

cp,

where T is the temperature, V is the horizontal windvector, Sp is the static stability parameter, cp is the spe-cific heat of dry air, and J represents diabatic heating.Static stability is proportional to the vertical gradient oftemperature, so Sp� represents the vertical advection oftemperature, sometimes called the adiabatic term.

FIG. 4. Composite of 200-hPa wind speed, based on total fields,for (a) the positive phase of the MJO in the eastern Indian Ocean,and (b) the negative phase. The contour interval is 5 m s�1.

FIG. 5. Difference between positive and negative phases of theMJO as in Fig. 2, but for vertical velocity at (a) 500 and (b) 300hPa. The contour interval is 2 Pa s�1. The vertical velocity is inpressure coordinates, so negative values (shaded blue) representupward motion and positive values (shaded red) represent down-ward motion.


When the tendency term is small with respect to theother terms and the static stability is approximatelyconstant (as assumed in quasi-geostrophy), vertical ve-locity is in balance with temperature advection and dia-batic heating. In the Tropics, scale considerations sug-gest that vertical velocity is typically balanced by dia-batic heating, while in the extratropics vertical velocityis typically balanced by temperature advection (Hos-kins and Karoly 1981). In the subtropics, both diabaticheating and temperature advection must be considered(Hoskins 1986). Here we examine the MJO variabilitywith respect to the thermodynamic balance by calculat-ing the terms of the thermodynamic equation from thedaily reanalysis data, and then averaging into MJOpositive and MJO negative cases and subtracting, asdone previously with OLR, winds, and precipitation.

The diabatic heating term is calculated as a residual.Comparisons of diabatic heating residually derivedfrom the NCEP reanalysis and ECMWF reanalysis, andfrom NCEP reanalysis 6-h model forecasts (Barlow etal. 1998) suggest that, although differences among theestimates indicate some uncertainties in magnitude and

vertical profile, they are broadly consistent, particularlyat midlevels. The fidelity of modeled circulations forcedby residually derived diabatic heating (e.g., Nigam1994, 1997) lends further credence to the residual cal-culation. Formally, the thermodynamic equation ap-plies to instantaneous values and frequently is appliedto 6-h data for residual diagnosis. Here we are usingdaily data, so we have validated the calculations byrepeating them with one year of 6-h data and comparedthe resulting MJO composites with those from dailydata: they are in close agreement.

Figure 6 shows the difference between positive andnegative MJO for each of the four terms in the ther-modynamic equation, using the same contour intervalthroughout. The tendency term (Fig. 6a) is negligiblecompared with the other terms, so the vertical velocityterm (Fig. 6b) is balanced by diabatic heating (Fig. 6c)and temperature advection (Fig. 6d). The similarity be-tween the vertical velocity field (Fig. 5b) and the ver-tical velocity term (Fig. 6b) shows that static stability isapproximately constant. As expected, the tropical bal-ance is largely with diabatic heating, and the extratrop-

FIG. 6. Difference between positive and negative phases of the MJO as in Fig. 2, but for the terms of thethermodynamic equation at 300 hPa: (a) the tendency term, (b) the vertical velocity term, (c) the diabatic heatingterm, and (d) the temperature advection. The contour interval is 0.3 K day�1 throughout, with the zero contouromitted.


ical balance is largely with temperature advection, al-though there is some contribution from diabatic heat-ing. The vigorous upward motion in the tropical easternIndian Ocean (Fig. 6b) is in balance with the diabaticheating in that area (Fig. 6c), associated with the strongtropical convection of the MJO. The descent overSouthwest Asia (Fig. 6b), in contrast, is primarily bal-anced by temperature advection (Fig. 6d), with only asmall contribution from diabatic heating (Fig. 6c).

To further consider the MJO-related changes in tem-perature advection, we examine the two main contri-butions: advection of the anomalous (MJO-related)temperature by the mean wind and the advection of themean temperature by the anomalous (MJO-related)wind. (We have verified that the product of the anoma-lous terms is small). These are shown in Fig. 7, with theadvection shown in (a) and (b) and the componentwinds and temperatures shown in the (c) and (d). Thetemperature anomalies associated with the MJO (Fig.7c) are consistent with the simple Gill–Matsuno frame-work: the temperature anomalies have approximately

the same spatial pattern as the upper-level winds, inkeeping with the first-baroclinic-mode structure of thewinds and the thermal wind relationship. Both tem-perature advection terms are important over SouthwestAsia, as the vigorous westerlies intersect the extratrop-ical MJO temperature anomalies and the extratropicalMJO wind anomalies intersect the mean midlatitudethermal gradient. Both mechanisms force subsidenceover Southwest Asia throughout the middle and uppertroposphere.

Based on this analysis, we suggest the following in-terpretation, given for the positive MJO case (signs arereversed for the negative case). Consistent with previ-ous work, the MJO variability is associated with vigor-ous tropical convection in the eastern Indian Ocean andassociated first-baroclinic, Rossby-like wind and tem-perature anomalies extending over Southwest Asia.The interaction of the MJO circulation and the meanflow results in both advection of the MJO temperatureanomalies by the mean wind and advection of the meanthermal gradient by the MJO wind anomalies. Via the

FIG. 7. Difference between positive and negative phases of the MJO as in Fig. 2, but for the primary contributionsto temperature advection at 300 hPa. The temperature advection is shown for (a) the advection of anomaloustemperature by the mean wind, and (b) the advection of the mean temperature by the anomalous wind. Thecontour interval is 0.3 K day�1, as in Fig. 6. The associated constituents are shown as (c) the anomalous tempera-ture (shaded) and mean wind (vectors) and (d) the mean wind (contours) and the anomalous winds (vectors).


thermodynamic energy equation, both these tempera-ture advection terms contribute to subsidence overSouthwest Asia, and the subsidence suppresses localprecipitation.

A similar argument is made in Rodwell and Hoskins(1996, 2001), in the context of the monsoon heatingduring boreal summer. Using an idealized model, theyshow that the Asian monsoon heating produces aRossby wave packet and the interaction of the wester-lies with the warm Rossby thermal anomaly producesdescent (“isentropic downgliding”). Although theiranalysis focuses on boreal summer when the maximumtropical heating is north of the equator, the summer-time westerlies are also displaced to the north, and theinteraction of the Rossby thermal anomalies with thewesterlies appears similar to the present analysis. Theyalso suggest a feedback mechanism whereby the de-scent, which is occurring in a warm thermal anomaly,would both inhibit precipitation (and associated latentheat release) and increase longwave cooling, factorsthat would reinforce the descent—a “diabatic enhance-ment.” There is some suggestion of this with the MJOvariability, as a reduction in diabatic heating is ob-served over Iran (Fig. 6c), although the values aresmall.

5. Three-week forecast potential

The forecast potential for SW Asia resulting from theMJO relationship depends both on how consistent therelationship is from year to year and on how effectivelythe relevant part of the MJO can be forecast. For apreliminary assessment, we will use OLR anomalies av-eraged over SW Asia (blue box in Fig. 2) as the forecasttarget and a simple persistence scheme for forecastingthe sign of the MJO (positive or negative in the red boxin Fig. 2).

To examine the consistency of the relationship, theOLR anomalies that occurred during the two MJOphases were calculated for each year and averaged overSW Asia. The yearly anomalies associated with eachphase are shown in Fig. 8a, along with the total region-ally averaged yearly anomalies (which equals the sumof the two). The influence of the MJO on the regionalaverage is quite consistent from year to year. Addition-ally, the MJO influence is comparable to the interan-nual variability: in 9 of the 22 yr the anomalies of onephase of the MJO are of different sign than the seasonalaverage (black line). That is, in those years, the averageinfluence of the MJO results in wetter-than-average pe-riods even in drier-than-average years, and vice versa.

As a preliminary assessment of the degree to whichthis relationship might be forecast, we consider an in-

termediate step where we examine the year-to-yearconsistency as before but based on a persistence fore-cast of the sign of the MJO rather than on the observedsign. The time scale of the MJO is roughly 30–60 days(e.g., Madden and Julian 1994) and since a relationshipwith SW Asia precipitation has been shown based onlyon the sign of the activity, the influence on the SW Asiaprecipitation may be expected to last roughly a half-cycle, or about 2–3 weeks at a time. This may providesome useful information in the time scales betweendaily weather forecasts and seasonal climate predic-tions. As only the sign of the MJO is being consideredhere, and given its rather narrow frequency band, apersistence forecast can be made simply by assumingthat once the MJO index changes sign, it then maintainsthe same sign for 3 weeks, either suppressing or enhanc-ing SW Asia precipitation during that time. This ap-proach produces a 21-day forecast every time the MJOindex changes sign, resulting in 8–12 forecast periods inthe November–April period. This scheme could easilybe improved upon but provides a simple starting point.As we are estimating forecast potential, we use themonitoring (0-day forecast) MJO data, the data thatwould have been available in real time. Figure 8b showsthe yearly averages for both the positive and negativecases. Even with this simple scheme, the predicted3-week periods of enhanced and suppressed OLR showa clear separation between the two phases for 8 of the16 yr. Moreover, the anomalies for the two cases haveopposite signs in 5 of those 8 yr. Obtaining such differ-ences by chance is highly unlikely—in 1000 recalcula-tions of the differences based on an MJO-like randomtime series the average difference between the twocases never exceeded the observed average differenceof �3.42 W m�2.

This result does not represent a formal prediction noris the forecast target of OLR averaged over all SW Asiaexpected to be a necessarily useful predictand. How-ever, the demonstrated consistency of the signal in3-week averages based on just a simple persistenceforecast of MJO activity suggests that the forecast po-tential is worthy of further exploration.

6. Synoptic timing: Three-day evolution

Precipitation in SW Asia comes primarily from theinfrequent passage of synoptic storms. So, regardless ofthe time scale of an external influence, it can only affectlocal precipitation during the occasional passage of thestorms. That is, the influence is only realized during thebrief periods of precipitation associated with eachstorm and is extraneous the rest of the time. Therefore,we can expect the modulation of regional precipitation


to be best captured with respect to the timing of thelocal storms.

A typical track for regional storms passes throughsouthern Iran, Afghanistan, and Pakistan. To examinethe influence of the MJO activity in the eastern IndianOcean on such storms, Fig. 9 shows daily OLR anomalylags following the incidence of OLR anomaly less than�20 W m�2 in the northern Persian Gulf (red box inFigs. 9a,b). The lags are averaged separately into thosecases occurring during a positive MJO in the easternIndian Ocean (256 cases, shown in Fig. 9a) and thosecases during a negative MJO (467 cases, shown in Fig.9b) for 1979–2001. (The averaging region was chosenbased on the 3-day lead OLR composite to the Af-ghanistan station data; results are not strongly sensitiveto the exact definition of the box geography or thethreshold.) Despite the same geographic origin andthreshold criterion in both cases, the evolution is strik-ingly different depending on the sign of the MJO. Al-though the 0-day lags are similar in strength, the 1-daylag shows clear differences, and by the third day, dif-

ferences are dramatic. Even in the irregularly reported,shorter-record station data, a large difference ispresent, with Afghanistan precipitation at the 3-day lagtwice as large in the negative MJO case compared tothe positive case. Especially for the negative MJO case,where the tropical MJO convection anomalies are quiteprominent, the characteristic Rossby-like wind re-sponse and enhanced vertical motion over SW Asia(not shown) are present both in advance of and duringthe storm’s movement through the region.

As individual storm tracks and speeds vary consider-ably, and as only the sign of the MJO is considered, thissimple compositing technique provides a somewhatblurred picture. Nonetheless, the timing and pattern ofthe composite evolution still appears to be relevant toindividual events. Figure 9c shows the daily evolutionfor a synoptic storm associated with severe flooding inAfghanistan for the period 10–13 April 2002 (contourinterval is twice that of Figs. 9a,b). Note that these datawere not included in the compositing and so provide anindependent comparison. Although the event has

FIG. 8. (a) The Nov–Apr means of SW Asia (blue box in Fig. 2) OLR anomalies (��1) are shown as a black line.The Nov–Apr means are also split into the values that occurred during the positive MJO (red line) and the negativeMJO (blue line). As the MJO was either positive or negative for every day in the record (no values of exactly zerooccurred), the lines corresponding to the two phases sum exactly to the total Nov–Apr anomalies. The sign of theOLR anomalies has been reversed to correspond to the sign of the implied precipitation anomalies. (b) Thecalculation is repeated, but for 3-week averages following a change in sign of the MJO. This is equivalent tocalculating 3-week averages of SW Asia OLR based on a “persistence” forecast of the sign of the MJO.


stronger magnitudes overall, as well as considerablymore precipitation in Saudi Arabia than the compos-ites, it has several similarities in timing and pattern,including the position and timing of the maximumtracking through Iran into Afghanistan, as well as thecoherence of the MJO signal in the eastern IndianOcean. A negative MJO composite with magnitudescloser to the April 2002 event can be obtained by com-positing with a threshold (the MJO criterion in Fig. 9bwas solely based on sign, without accounting for themagnitude of the MJO signal). This finding that therelationship is stronger when the timing of individualstorms is considered is consistent with the fact that, aspreviously noted, all 10 of the wettest days in the stationdata occur during the negative phase of the MJO. In-deed, compositing for the 3 days previous to the 10wettest days results in an evolution similar to Fig. 9b:negative anomalies are present throughout the evolu-

tion in the eastern Indian Ocean, while positive anoma-lies start in the northern Persian Gulf and trackingthrough Afghanistan.

Of particular interest for the April 2002 floodingevent, the identifying aspects for the enhanced compos-ite were clearly present 2 days in advance of the flood-ing in Afghanistan. A synoptic storm was visible track-ing over the Persian Gulf on 10 April (Fig. 9c) and theBMRC real-time MJO monitoring data did diagnose ahighly negative MJO phase (suppressed convection) inthe eastern Indian Ocean (not shown). We emphasizethat these identifying aspects for the enhanced case wereall present in data that are readily available in real time.

7. Summary and discussion

Using both daily OLR and station precipitation data,we have shown that MJO activity in the eastern Indian

FIG. 9. The 4-day evolution of OLR anomalies is shown for (a) 0–3-day lag composites to OLR ��20 W m�2 in western Iran (redbox), when the MJO is positive in the eastern Indian Ocean; (b) 0–3-day lag composites to OLR anomalies ��20 W m�2 in westernIran, when the MJO is negative in the eastern Indian Ocean; and (c) daily OLR anomalies for an Afghan flooding event in Apr 2002.Negative OLR anomalies are shaded green, corresponding to positive precipitation anomalies, while positive OLR anomalies are shadedbrown, corresponding to negative precipitation anomalies. The contour interval in (a) and (b) is 5 W m�2, and in (c) is 10 W m�2.


Ocean has a considerable influence on the precipitationof SW Asia during November–April. Average dailyprecipitation, from 13 stations in Afghanistan for 1979–85, decreases by 23% relative to the mean during theperiods when the MJO enhances convection in the east-ern Indian Ocean (positive phase) and increases by21% when the MJO suppresses convection in the east-ern Indian Ocean (negative phase). Individually, 7 ofthe 13 stations change by at least 20% and 3 change bymore than 50% (more than a factor of 2 differencebetween the precipitation during the two phases). Thedistribution of extremes is also affected such that the 10wettest days all occur during the negative MJO phase.The longer OLR record, 1979–2001, shows that the sig-nal is very consistent, with averaged SW Asia anoma-lies positive (less precipitation) during the MJO posi-tive phase and negative (more precipitation) during thenegative phase for every season in the 1979–2001 rec-ord. In 9 of the 22 yr, the average OLR anomaly duringone of the phases was of opposite sign with respect tothe seasonal anomaly—corresponding to wetter-than-average periods during a drier-than-average year andvice versa.

Dynamically, the MJO-related changes in tropicalconvection affect both the jet-level winds and the ver-tical motion at mid- and upper levels. The changes tothe upper-level winds affect the sharpness of the jet exitregion over the region, consistent with changes in localbaroclinic instability and the observed anomalies ofprecipitation. The vertical motion field is also changedby thermodynamic processes. The interaction of theMJO circulation and the mean flow result in both ad-vection of the MJO temperature anomalies by themean wind and advection of the mean thermal gradientby the MJO wind anomalies, both terms contributing tovigorous changes in vertical velocity over SouthwestAsia, consistent with the changes in precipitation.

The predictability of the MJO and the link betweenthe MJO and SW Asia precipitation suggests the po-tential for prediction of the SW Asia precipitation. Asa preliminary step, we have shown that, for an areaaverage, the MJO is very consistent year to year and asimple persistence forecast shows some ability to dis-tinguish between enhanced and suppressed SW Asiaprecipitation at 3-week time scales.

The influence of the MJO on SW Asian precipitationcan be seen more clearly when keyed to the infre-quently occurring local storms. A simple threshold ap-proach using OLR averaged over the northern PersianGulf (which could be calculated operationally) capturesa primary storm track into Afghanistan. Compositingthe storm evolution based on the MJO phase showslarge differences by the second day, as the storm enters

Afghanistan. This MJO influence appears to have rel-evance to the flooding event of April 2002, when anexceptionally vigorous storm moved through the regionin concert with well-defined negative MJO anomalies inthe eastern Indian Ocean.

Two factors appear to be important in explaining thesensitivity of the SW Asia precipitation to the MJO: theproximity of SW Asia to the most active region of theMJO, so that SW Asia is within the direct wind re-sponse to the MJO tropical convection anomalies whenthey are at their largest values, and the vigorous windresponse to tropical forcing in that same region (Tingand Sardeshmukh 1993). The structure of the meanflow is important both to the vigor of the wind responseand to the thermodynamic interaction of the wind re-sponse with the mean flow, which results in changes tothe vertical velocity over SW Asia. Additional factors,including wave–mean flow interaction and the role ofmoisture transport, may be important, and further re-search is warranted.

For future analysis, several key areas can be readilyidentified. More observational data of precipitation,both in terms of daily station records of precipitationfor the other countries of the region and in terms ofdaily records with more consistent reporting, would bevery helpful for better verifying the strength and extentof the signal suggested by the OLR, as well as for analy-sis of individual events. We are in the process of ob-taining such data. A more sophisticated approach tostorm tracking would be useful for considering otherstorm tracks through the region, as well for assessingwhether just the strength of the storms is affected bythe MJO, or whether the number or track of the stormschange as well. Consideration of the degree to whichnumerical weather prediction model are able to capturethe tropical linkages, which appear to be importanteven to day-2 and day-3 forecasting, would provide auseful test of regional short-term prediction skill andevaluate the utility of considering the MJO state inshort-term forecasts. Seasonality needs to be consid-ered, as the mean flow and synoptic activity undergoimportant changes within the November–April periodconsidered here.

With respect to estimating the MJO variability, thereis a considerable amount of both diagnostic and fore-casting information not yet made use of. The currentanalysis only made use of the sign of the MJO; this maybe easily extended to account for both the magnitudeand life cycle of the MJO, both of which are likelyimportant. For forecasting the MJO, a persistence ap-proach has been used here for simplicity, but a numberof considerably more sophisticated schemes have beenproposed for MJO forecasting (e.g., Waliser et al. 1999;


Lo and Hendon 2000; Mo 2001; Wheeler and Hendon2004; etc.). A range of monitoring and forecasting esti-mates are now available or in planning (Schubert et al.2002); these will allow a more robust estimation of theMJO influence and related forecast skill.

Finally, the underlying dynamics of the relationshipneed further exploration, including further analysis ofthe mechanisms outlined here and consideration oftheir relative importance. Consideration of the dynami-cal issues with respect to MJO evolution and how theextratropical anomalies change as the MJO propagatesfrom the central Indian Ocean into the western Pacificwill likely provide a useful perspective. A key aspect forfurther analysis is the low-level flow and moisturetransport in SW Asia, which is complicated by bothdata scarcity and the mountainous nature of the region.

Acknowledgments. Interpolated OLR data were pro-duced by the NOAA–CIRES Climate Diagnostics Cen-ter, Boulder, Colorado. The IRI data library was in-valuable for manipulating the data in this study; allplots were produced with GrADS software. We thankTony Barnston, Chet Ropelewski, and John Hendersonfor useful discussions, and David Salstein, RichardRosen, Jafar Nazemosadat, and the reviewers forthoughtful comments on the manuscript, all of whichgreatly improved the analysis and presentation. Thisresearch was supported by the National Science Foun-dation under Grant ATM-0233563.

REFERENCES

Aizen, E. M., V. B. Aizen, J. M. Melack, T. Nakamura, and T.Ohta, 2001: Precipitation and atmospheric circulation pat-terns at mid-latitudes of Asia. Int. J. Climatol., 21, 535–556.

Arkin, P., and B. Meisner, 1987: The relationship between large-scale convective rainfall and cold cloud over the WesternHemisphere during 1982–84. Mon. Wea. Rev., 115, 51–74.

Barlow, M., S. Nigam, and E. H. Berbery, 1998: Evolution of theNorth American monsoon system. J. Climate, 11, 2238–2257.

——, H. Cullen, and B. Lyon, 2002: Drought in central and south-west Asia: La Nina, the warm pool, and Indian Ocean pre-cipitation. J. Climate, 15, 697–700.

Blackmon, M., J. Wallace, N.-C. Lau, and S. Mullen, 1977: Anobservational study of the Northern Hemisphere wintertimecirculation. J. Atmos. Sci., 34, 1040–1053.

Bluestein, H. B., 1993: Synoptic-Dynamic Meteorology in Midlati-tudes. Vol. II, Observations and Theory of Weather Systems,Oxford University Press, 594 pp.

Bond, N., and G. Vecchi, 2003: The influence of the Madden–Julian oscillation on precipitation in Oregon and Washing-ton. Wea. Forecasting, 18, 600–613.

Branstator, G., 1985: Analysis of general circulation model seasurface temperature anomaly simulations using a linearmodel. Part I: Forced solutions. J. Atmos. Sci., 42, 2225–2241.

Cai, M., and M. Mak, 1990: On the basic dynamics of regionalcyclogenesis. J. Atmos. Sci., 47, 1417–1442.

Chang, C.-P., and K. M. Lau, 1982: Short-term planetary-scaleinteractions over the Tropics and midlatitudes during north-ern winter. Part I: Contrasts between active and inactive pe-riods. Mon. Wea. Rev., 110, 933–946.

Cotton, W. R., and R. A. Anthes, 1989: Storm and Cloud Dynam-ics. Academic Press, 883 pp.

Gill, A., 1980: Some simple solutions for heat induced tropicalcirculation. Quart. J. Roy. Meteor. Soc., 106, 447–462.

Goswami, B. N., R. S. Ajayamohan, P. K. Xavier, and D. Sen-gupta, 2003: Clustering of synoptic activity by Indian summermonsoon intraseasonal oscillations. Geophys. Res. Lett., 30,1431, doi:10.1029/2002GL016734.

Hendon, H. H., 1988: A simple model of the 40–50 day oscillation.J. Atmos. Sci., 45, 569–584.

——, and B. Liebmann, 1990: The intraseasonal (30–50 day) os-cillation of the Australian summer monsoon. J. Atmos. Sci.,47, 2909–2923.

——, and M. Salby, 1994: The life cycle of the Madden–Julianoscillation. J. Atmos. Sci., 51, 2225–2237.

——, and ——, 1996: Planetary-scale circulations forced by in-traseasonal variations of observed convection. J. Atmos. Sci.,53, 1751–1760.

Higgins, R. W., J. K. Schemm, W. Shi, and A. Leetmaa, 2000:Extreme precipitation events in the western United Statesrelated to tropical forcing. J. Climate, 13, 793–820.

Hoerling, M., and A. Kumar, 2003: The perfect ocean for drought.Science, 299, 691–694.

Holton, J., 1992: An Introduction to Dynamic Meteorology. 3d ed.Academic Press, 511 pp.

Hoskins, B. J., 1986: Diagnosis of forced and free variability in theatmosphere. Atmospheric and Oceanic Variability, H. Cattle,Ed., Royal Meteorological Society, 57–73.

——, and D. J. Karoly, 1981: The steady linear response of aspherical atmosphere to thermal and orographic forcing. J.Atmos. Sci., 38, 1179–1196.

——, and P. Valdes, 1990: On the existence of storm-tracks. J.Atmos. Sci., 47, 1854–1864.

Huffman, G. J., R. F. Adler, M. M. Morrissey, D. T. Bolvin, S.Curtis, R. Joyce, B. McGavock, and J. Susskind, 2001: Globalprecipitation at one-degree daily resolution from multisatel-lite observations. J. Hydrometeor., 2, 36–50.

Jin, F.-F., and B. J. Hoskins, 1995: The direct response to tropicalheating in a baroclinic atmosphere. J. Atmos. Sci., 52, 307–319.

Jones, C., 2000: Occurrence of extreme precipitation events inCalifornia and relationships with the Madden–Julian oscilla-tion. J. Climate, 13, 3576–3587.

——, D. Waliser, and C. Gautier, 1998: The influence of the Mad-den–Julian oscillation on ocean surface heat fluxes and seasurface temperature. J. Climate, 11, 1057–1072.

Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Re-analysis Project. Bull. Amer. Meteor. Soc., 77, 437–471.

Krishnamurti, T., 1961: The subtropical jet stream of winter. J.Meteor., 18, 172–191.

Liebmann, B., and C. A. Smith, 1996: Description of a complete(interpolated) outgoing longwave radiation dataset. Bull.Amer. Meteor. Soc., 77, 1275–1278.

Lim, G. H., and J. M. Wallace, 1991: Structure and evolution ofbaroclinic waves as inferred from regression analysis. J. At-mos. Sci., 48, 1718–1732.

Lo, F., and H. Hendon, 2000: Empirical extended-range predic-tion of the Madden–Julian oscillation. Mon. Wea. Rev., 128,2528–2543.


Lott, N., cited 1998: Global surface summary of day. NationalClimatic Data Center, Asheville, NC. [Available online athttp://www.ncdc.noaa.gov/cgi-bin/res40.pl?page�gsod.html.]

Madden, R. A., and P. R. Julian, 1994: Observations of the 40–50-day tropical oscillation—A review. Mon. Wea. Rev., 122,814–837.

Martyn, D., 1992: Climates of the World. Elsevier, 436 pp.Mo, K. C., 1999: Alternating wet and dry episodes over California

and intraseasonal oscillations. Mon. Wea. Rev., 127, 2759–2776.

——, 2001: Adaptive filtering and prediction of intraseasonal os-cillations. Mon. Wea. Rev., 129, 802–817.

——, and R. W. Higgins, 1998a: Tropical influences on Californiaprecipitation. J. Climate, 11, 412–430.

——, and ——, 1998b: Tropical convection and precipitation re-gimes in the western United States. J. Climate, 11, 2404–2423.

——, and ——, 1998c: The Pacific–South American modes andtropical convection during the Southern Hemisphere winter.Mon. Wea. Rev., 126, 1581–1596.

Moskowitz, B., and C. Bretherton, 2000: An analysis of frictionalfeedback on a moist equatorial Kelvin mode. J. Atmos. Sci.,57, 2188–2206.

Nazemosadat, M. J., 1998: Persian Gulf sea surface temperatureas a drought diagnostic for southern Iran. Drought NetworkNews, Vol. 10, No. 3, International Drought InformationCenter and the National Drought Mitigation Center, 10–12.

——, and I. Cordery, 2000: On the relationships between ENSOand autumn rainfall in Iran. Int. J. Climatol., 20, 47–61.

New, M. G., M. Hulme, and P. D. Jones, 2000: Representing twen-tieth-century space–time climate variability. Part II: Devel-opment of 1901–1996 monthly grids of terrestrial surface cli-mate. J. Climate, 13, 2217–2238.

Nigam, S., 1994: On the dynamical basis for the Asian summermonsoon rainfall–El Nino relationship. J. Climate, 7, 1750–1771.

——, 1997: The annual warm to cold phase transition in the east-ern equatorial Pacific: Diagnosis of the role of stratus cloud-top cooling. J. Climate, 10, 2447–2467.

Nogues-Paegle, J., and K. C. Mo, 1997: Alternating wet and dryconditions over South America during summer. Mon. Wea.Rev., 125, 279–291.

——, L. A. Byerle, and K. C. Mo, 2000: Intraseasonal modulationof South American summer precipitation. Mon. Wea. Rev.,128, 837–850.

Peixoto, J. P., and A. H. Oort, 1992: Physics of Climate. AmericanInstitute of Physics, 520 pp.

Petterssen, S., 1958: Introduction to Meteorology. 2d ed. McGraw-Hill, 327 pp.

Rodwell, M., and B. J. Hoskins, 1996: Monsoons and the dynamicsof deserts. Quart. J. Roy. Meteor. Soc., 122, 1385–1404.

——, and ——, 2001: Subtropical anticyclones and summer mon-soons. J. Climate, 14, 3193–3211.

Rui, H., and B. Wang, 1990: Development characteristics anddynamic structure of tropical intraseasonal convectionanomalies. J. Atmos. Sci., 47, 357–379.

Sardeshmukh, P., and B. Hoskins, 1988: The generation of globalrotation flow by steady idealized tropical divergence. J. At-mos. Sci., 45, 1228–1251.

Schubert, S., R. Dole, H. Van den Dool, M. Suarez, and D. Wa-liser, 2002: Proceedings from a Workshop on Prospects forImproved Forecasts of Weather and Short-Term ClimateVariability on Subseasonal (2 Week to 2 Month) Time Scales.Vol. 23, NASA Tech. Memo. 2002-104606, 171 pp.

Simmons, A., J. Wallace, and G. Branstator, 1983: Barotropicwave propagation and instability, and atmospheric telecon-nection patterns. J. Atmos. Sci., 40, 1363–1392.

Ting, M., and P. Sardeshmukh, 1993: Factors determining the ex-tratropical response to equatorial diabatic heating anomalies.J. Atmos. Sci., 50, 907–918.

Tippett, M., M. Barlow, and B. Lyon, 2003: Statistical correctionof central southwest Asia winter precipitation simulations.Int. J. Climatol., 23, 1421–1433.

Waliser, D., C. Jones, J. K. Schemm, and N. E. Graham, 1999: Astatistical extended-range tropical forecast model based onthe slow evolution of the Madden–Julian oscillation. J. Cli-mate, 12, 1918–1939.

——, K. M. Lau, W. Stern, and C. Jones, 2003: Potential predict-ability of the Madden–Julian oscillation. Bull. Amer. Meteor.Soc., 84, 33–50.

Wallace, J., G.-H. Lim, and M. Blackmon, 1988: Relationshipbetween cyclone tracks, anticyclone tracks and baroclinicwaveguides. J. Atmos. Sci., 45, 439–462.

Wang, B., and H. Rui, 1989: Some dynamic aspects of the equa-torial intraseasonal oscillations. East Asia and Western Pa-cific Meteorology and Climate, P. Sham and C.-P. Chang,Eds., World Scientific, 119–130.

——, and ——, 1990: Synoptic climatology of transient tropicalintraseasonal convection anomalies: 1975–1985. Meteor. At-mos. Phys., 44, 43–61.

Wheeler, M. C., and K. Weickmann, 2001: Real-time monitoringand prediction of modes of coherent synoptic to intraseasonaltropical variability. Mon. Wea. Rev., 129, 2677–2694.

——, and H. H. Hendon, 2004: An all-season real-time multivari-ate MJO index: Development of an index for monitoring andprecipitation. Mon. Wea. Rev., 132, 1917–1932.

Whitaker, J., and R. Dole, 1995: Organization of storm tracks inzonally varying flows. J. Atmos. Sci., 52, 1178–1191.

——, and K. Weickmann, 2001: Subseasonal variations of tropicalconvection and week-2 prediction of wintertime westernNorth American rainfall. J. Climate, 14, 3279–3288.

Yamagata, T., and Y. Hayashi, 1984: A simple diagnostic modelfor the 30–50 day oscillation in the Tropics. J. Meteor. Soc.Japan, 62, 709–717.


modulation of daily precipitation over southwest asia by the …blyon/papers/mwr_sw_asia_and... ·...

Documents