• Anderson, D. L. T., 1976: The low level jet as a western boundary current. Mon. Wea. Rev.,104, 907–921.

  • Ashok, K., D. Gregory, and P. M. Inness, 1995: Intraseasonal variability of the Indian summer monsoon simulated by the Hadley Centre climate model. Hadley Centre Tech. Note CRTN-60, Hadley Centre, Bracknell, United Kingdom, 18 pp. [Available from Hadley Centre for Climate Prediction and Research, Bracknell RG12 2SY, United Kingdom.].

  • Bannon, P. R., 1979: On the dynamics of the East African jet. PartII: Jet transients. J. Atmos. Sci.,36, 2153–2168.

  • ——, 1982: On the dynamics of the East African jet. Part III: Arabian Sea branch. J. Atmos. Sci.,39, 2267–2278.

  • Bedi, H. S., and M. M. S. Bindra, 1980: Principal components of monsoon rainfall. Tellus,32, 296–298.

  • Bess, T. D., G. L. Smith, T. P. Charlock, and F. G. Rose, 1992: Annual and interannual variations of Earth-emitted radiation based on a 10-year data set. J. Geophys. Res.97, 12825–12835.

  • Cadet, D., and M. Desbois, 1981: A case study of a fluctuation of the Somali jet during the Indian summer monsoon. Mon. Wea. Rev.,109, 182–187.

  • Cullen, M. J. P., 1993: The unified forecast/climate model. Meteor. Mag.,112, 81–94.

  • Esbensen, S. K., and Y. Kushnir, 1981: Heat budget of the global ocean: Estimates from marine observations. Climate Research Institute Rep. 29, 27 pp. [Available from Oregon State University, Corvallis, OR 97331.].

  • Ferranti, L., J. M. Slingo, T. N. Palmer, and B. J. Hoskins, 1997: Relationships between interannual and intraseasonal monsoon variability as diagnosed from AMIP integrations. Quart. J. Roy. Meteor. Soc.,123, 1323–1357.

  • Findlater, J., 1969: A major low-level air current near the Indian Ocean during the northern summer. Quart. J. Roy. Meteor. Soc.,95, 362–380.

  • ——, 1974: An extreme wind speed in the low-level jet-stream system of the western Indian Ocean. Meteor. Mag.,103, 201–205.

  • Gordon, C., and D. K. Wright, 1995: A comparison of modelled surface fluxes with climatological estimates. Hadley Centre Tech. Note CRTN-53, Hadley Centre, Bracknell, United Kingdom, 25 pp. [Available from Hadley Centre for Climate Prediction and Research, Bracknell RG12 2SY, United Kingdom.].

  • Gowariker, V., V. Thapliyal, S. M. Kulshresthra, G. S. Mandal, N. Sen Roy, and D. R. Sikka, 1991: A power regression model for long range forecast of southwest monsoon rainfall over India. Mausam,42, 125–130.

  • Hall, C. D., R. A. Stratton, and M. L. Gallani, 1995: Climate simulations with the Unified Model: AMIP runs. Hadley Centre Tech. Note CRTN-61, Hadley Centre, Bracknell, United Kingdom, 92 pp. [Available from Hadley Centre for Climate Prediction and Research, Bracknell RG12 2SY, United Kingdom.].

  • Hart, J. E., 1977: On the theory of the East African low level jet stream. Pure Appl. Geophys.,115, 1263–1282.

  • Hoskins, B. J., 1974: The role of potential vorticity in symmetric stability and instability. Quart. J. Roy. Meteor. Soc.,100, 480–482.

  • ——, and A. J. Simmons, 1975: A multi-layer model and the semi-implicit method. Quart. J. Roy. Meteor. Soc.,101, 637–655.

  • ——, and M. J. Rodwell, 1995: A model of the Asian summer monsoon. Part I: The global scale. J. Atmos. Sci.,52, 1329–1340.

  • ——, M. E. McIntyre, and A. W. Robertson, 1985: On the use and significance of isentropic potential vorticity maps. Quart. J. Roy. Meteor. Soc.,111, 877–946.

  • Inness, P. M., and D. Gregory, 1994: Simulation of the Indian monsoon and tropical intraseasonal variability by a general circulation model. Hadley Centre Tech. Note CRTN-52, Hadley Centre, Bracknell, United Kingdom, 50 pp.[Available from Hadley Centre for Climate Prediction and Research, Bracknell RG12 2SY, United Kingdom.].

  • Ju, J. H., and J. Slingo, 1995: The Asian summer monsoon and ENSO. Quart. J. Roy. Meteor. Soc.,121, 1133–1168.

  • Kishtawal, C. M., N. Gautam, S. Jaggi, and P. C. Pandey, 1994: Surface-level moisture transport over the Indian Ocean during southwest monsoon months using NOAA/ HIRS data. Bound.-Layer Meteor.,69, 159–171.

  • Krishnamurti, T. N., and H. N. Bhalme, 1976: Oscillations of a monsoon system. Part I: Observational aspects. J. Atmos. Sci.,33, 1937–1954.

  • ——, J. Molinari, and H.-L. Pan, 1976: Numerical simulation of the Somali jet. J. Atmos. Sci.,33, 2350–2362.

  • ——, V. Wong, H.-L. Pan, R. Pasch, J. Molinari, and P. Ardanuy, 1983: A three-dimensional planetary boundary layer model for the Somali Jet. J. Atmos. Sci.,40, 894–908.

  • Kulkarni, A., R. H. Kripalani, and S. V. Singh, 1992: Classification of summer monsoon rainfall patterns over India. Int. J. Climatol.,12, 269–280.

  • Kumar, S., 1992: Interaction of mid-latitude systems in the Southern Hemisphere with southwest monsoon. Mausam,34, 405–410.

  • ——, K. Sethumadhavan, and M. K. Rao, 1983: Influence of frontal systems moving across southern Africa and southwest Indian Ocean on the onset of southwest monsoon over west coast of India. Mausam,43, 37–42.

  • Matthews, A. J., B. J. Hoskins, J. M. Slingo, and M. Blackburn, 1996: Development of convection along the SPCZ within the Madden–Julian oscillation. Quart. J. Roy. Meteor. Soc.,122, 669–688.

  • Mooley, D. A., and B. Parthasarathy, 1983: Indian summer monsoon and El-Niño. Pure Appl. Geophys.,121, 339–352.

  • Rao, G. V., and J. L. Haney, 1982: Kinematic and thermal structure of two surges of flow in the northern Mozambique channel area. Quart. J. Roy. Meteor. Soc.,108, 957–974.

  • Rasmusson, E. M., and T. H. Carpenter, 1983: The relationship between eastern equatorial Pacific sea surface temperatures and rainfall over India and Sri Lanka. Mon. Wea. Rev.,111, 517–528.

  • Rodwell, M. J., and B. J. Hoskins, 1995: A model of the Asian summer monsoon. Part II: Cross-equatorial flow and PV behavior. J. Atmos. Sci.,52, 1341–1356.

  • ——, and ——, 1996: Monsoons and the dynamics of deserts. Quart. J. Roy. Meteor. Soc.,122, 1385–1404.

  • Saha, K. R., 1970: Air and water vapour transport across the equator in western Indian Ocean during northern summer. Tellus,22, 681–687.

  • ——, F. Sanders, and J. Shukla, 1981: Westward propagating predecessors of monsoon depressions. Mon. Wea. Rev.,109, 330–343.

  • Sikka, D. R., and S. Gadgil, 1980: On the maximum cloud zone and the ITCZ over Indian longitudes during the southwest monsoon. Mon. Wea. Rev.,108, 1840–1853.

  • Sperber, K. R., and T. N. Palmer, 1996: Interannual tropical rainfall variability in general circulation model simulations associated with the Atmospheric Model Intercomparison Project. J. Climate,9, 2727–2750.

  • Webster, P. J., 1983: Mechanisms of monsoon low-frequency variability:Surface hydrology effects. J. Atmos. Sci.,40, 2110–2124.

  • WMO, 1992: Simulation if interannual and intraseasonal monsoon variability. WCRP-68, WMO/TD-470, Report of workshop, NCAR, Boulder, CO, 236 pp. [Available from World Meteorological Organization, Case Postale No. 2300, CH-1211 Geneva 20, Switzerland.].

  • ——, 1993: Simulation and prediction of monsoons. WCRP-80, WMO/TD-546, Report of workshop, New Delhi, India, 83 pp. [Available from World Meteorological Organization, Case Postale No. 2300, CH-1211 Geneva 20, Switzerland.].

  • View in gallery

    Mean June–August 850-hPa wind vectors with shaded isotachs over the Indian Ocean region calculated using analyses from the European Centre for Medium-Range Weather Forecasts for 1992–95.

  • View in gallery

    Low-level wind vectors at 887 hPa with shaded isotachs over the Indian Ocean region taken from 1994 analyzed data from the European Centre for Medium-Range Weather Forecasts all at 12 UTC. (a) 11 June, injection; (b) 16 June, southward turning; and (c) 21 June, monsoon “break.”

  • View in gallery

    (a) The low-level northward wind component at 887 hPa averaged over the box indicated covering the Mozambique channel region using the same 1200 UTC data as in Fig. 1. (b) The daily all-India rainfall percentage of normal calculated operationally by the Indian Meteorological Department for the monsoon season of 1994, together with a 7-day running mean.

  • View in gallery

    Low-level winds vectors with shaded contours of potential vorticity at 887 hPa over the Indian Ocean region for the idealized model simulations. The left panels, (a)–(d), show a sequence of days from the model control simulation; the right panels, (e)–(h), show the same days for the perturbation simulation where a realistic anticyclonic weather system has been introduced into the Southern Hemisphere midlatitudes. The numbers in parenthesis indicate the number of days after the anticyclonic forcing was turned on. The shading units are PV units.

  • View in gallery

    Vorticity forcing used in the idealized model with contours 0.5, 1.0, and 1.5 × 10−5 s−1 day−1 shown. The forcing moves within the model along the track shown. Circles are printed at successive days.

  • View in gallery

    (a) and (b): The horizontal projection of three-dimensional particle trajectories within the idealized model simulations. The trajectories are calculated both backward and forward from day 11 when the particles are aligned along the west coast of India at σ = 0.89 (black dots). Circles are drawn every other day. The small gray filled circles indicate that a particle came too close to the earth’s surface for its path to be traced beyond that point. (a) Control simulation, (b) perturbed simulation also showing the vorticity forcing applied in this integration and its track. The contours shown for the forcing are 0.5, 1.0, and 1.5 × 10−5 s−1 day−1 and the circled numbers, 1 and 3, show its location 1 and 3 days from when it was turned on. (c) Height in hPa against time of all the particles shown in (a) and (b) except for those corresponding to the four northmost and three southmost black dots at day 11.

  • View in gallery

    Mean sea-level pressure from (a) the control integration and (b) the perturbation integration. Both panels correspond to the time 2 days after the anticyclonic forcing was turned on in the perturbation integration. The contour interval is 2 hPa; “MH” signifies the Mascarene high, “MT” the monsoon trough.

  • View in gallery

    Mean June–August 850-hPa wind vectors with shaded isotachs over the Indian Ocean region calculated from a 10-yr integration of the GCM forced with observed sea surface temperatures from 1979 to 1988.

  • View in gallery

    Low-level 850-hPa winds from GCM integrations. (a) Composite control run at day 9, (b) composite perturbation run at day 9, (c)–(f) anomalies at day 9 for the four individual perturbation runs, and (g) composite anomaly at day 9. The 20 m s−1 reference arrow applies to full fields and the 10 m s−1 reference arrow applies to anomaly fields.

  • View in gallery

    (a) Map showing 69.375°E meridian at which GCM modeled zonal moisture fluxes are evaluated. (b) Zonal moisture fluxes across the meridian shown in (a), vertically integrated from 1000 to 700 hPa at day 10 for composite control (solid) and composite perturbation (dashed) integrations. (c) Composite anomaly zonal moisture fluxes across the meridian shown in (a), vertically integrated from 1000 to 700 hPa at days 7–13.

  • View in gallery

    (a) GCM composite control run precipitation averaged over days 6 to 13. (b) Composite anomaly precipitation averaged over days 6 to 13. (c) Time series of anomaly precipitation averaged in the two boxes shown in (b) (dotted), together with 3-day running means (solid).

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 222 221 6
PDF Downloads 56 56 8

Breaks in the Asian Monsoon: The Influence of Southern Hemisphere Weather Systems

View More View Less
  • 1 Centre for Global Atmospheric Modelling, Department of Meteorology, University of Reading, Reading, United Kingdom
© Get Permissions
Full access

Abstract

Atmospheric model results suggest that chaotic weather systems in the Southern Hemisphere midlatitudes can trigger “breaks” in the Indian monsoon rainfall. Indeed, the mechanism may be able to trigger a more general break of the entire Asian monsoon. The mechanism proposed involves the injection of dry, high negative potential vorticity air from the Southern Hemisphere midlatitudes into the low-level monsoon inflow. Observations from the 1994 monsoon season tend to support this mechanism and, if true, it may imply some predictive skill for shorter-range forecasting. However, the mechanism proposed may also imply that an accurate seasonal forecast of monsoon rainfall is an impossible objective, with important consequences for the agricultural economies of the region. Results are presented from both an idealized model and a full general circulation model.

Corresponding author address: Dr. M. Rodwell, Hadley Centre, U.K. Meteorological Office, London Road, Bracknell RG12 2SY, United Kingdom.

Email: mjrodwell@meto.gov.uk

Abstract

Atmospheric model results suggest that chaotic weather systems in the Southern Hemisphere midlatitudes can trigger “breaks” in the Indian monsoon rainfall. Indeed, the mechanism may be able to trigger a more general break of the entire Asian monsoon. The mechanism proposed involves the injection of dry, high negative potential vorticity air from the Southern Hemisphere midlatitudes into the low-level monsoon inflow. Observations from the 1994 monsoon season tend to support this mechanism and, if true, it may imply some predictive skill for shorter-range forecasting. However, the mechanism proposed may also imply that an accurate seasonal forecast of monsoon rainfall is an impossible objective, with important consequences for the agricultural economies of the region. Results are presented from both an idealized model and a full general circulation model.

Corresponding author address: Dr. M. Rodwell, Hadley Centre, U.K. Meteorological Office, London Road, Bracknell RG12 2SY, United Kingdom.

Email: mjrodwell@meto.gov.uk

1. Introduction

During late spring, an intense rainfall phenomenon, known as the Asian summer monsoon, commences over southern Asia. The monsoon rains are essential for the annual replenishing of freshwater supplies for many countries in the region. Some parts of India receive up to 90% of their total annual rainfall in just 3 months (Mooley and Parthasarathy 1983). The rains bring relief from the hot, dry, and dusty conditions that precede them, but they also bring the danger of landslides and flooding. During a monsoon season, there is considerable variability in the winds and rains with wet “active” periods punctuated by dry “break” spells when rainfall can be strongly reduced. There may be several reasons for this active–break variability. This paper identifies one possible mechanism that links synoptic activity in the Southern Hemisphere midlatitudes to dry periods in the monsoon season. The study appeals to both observations and previous papers, using theory, simple modeling, and full GCM experiments to substantiate the proposed mechanism.

Figure 1 shows the June to August time-mean 850-hPa winds calculated using analyses from the European Centre for Medium-Range Weather Forecasts (ECMWF) for the years 1992–95, inclusive. There are generally easterlies over the tropical southern Indian Ocean, cross-equatorial flow at the east coast of Africa, and westerlies over the Arabian Sea and into India. These westerlies carry much of the moisture that sustains the monsoon rainfall and 60%–80% of this moisture may be accounted for by fluxes across the equator; see Saha (1970) and Kishtawal et al. (1994). The low-level cross-equatorial flow, sometimes referred to as the “Findlater jet” (Findlater 1969), is therefore a crucial element of the monsoon and has been the focus of several modeling studies, including Krishnamurti et al. (1976), Krishnamurti et al. (1983), Anderson (1976), Hart (1977), and Bannon (1979, 1982). However, the jet represents by far the largest time-mean low-level cross-equatorial flow seen at any longitude and at any time of the year and this can pose problems regarding the vorticity of the air or, more elegantly, the potential vorticity (PV) of the air (Rodwell and Hoskins 1995; hereafter RH).

The potential vorticity equation can be expressed (e.g., Hoskins et al. 1985) as
i1520-0469-54-22-2597-eq1
where, in particular, P is potential vorticity, ζa is absolute vorticity, θ̇ = /Dt is the diabatic potential temperature source, and K is the frictional force curl. Under idealized conditions, of adiabatic, frictionless flow, PV is conserved by a parcel of air and the Findlater jet may therefore be expected to transport not only moisture, but also negative potential vorticity from the Southern Hemisphere into the Northern Hemisphere. Rodwell and Hoskins (1995) demonstrated that if significant negative PV does get into the Northern Hemisphere, this may lead to instabilities (see also Hoskins 1974) and, importantly, to strong southward (anticyclonic) turning of the low-level flow so that the air tends to “avoid” India and even return to the Southern Hemisphere. One can speculate that this may have serious consequences for moisture fluxes into India and, therefore, to monsoon rainfall itself. The degree to which in reality, and in GCMs, material modification of PV does occur (and the variability of this modification) may therefore be fundamental to the mean monsoon (and to its variability).

Rodwell and Hoskins (1995) showed that the frictional torque exerted by the East African Highlands on the jet is an important mechanism for modifying PV. They also showed that the height of maximum diabatic heating over the southern Indian Ocean can be a critical factor. If the heating maximum occurs below the low-level inflow, as indeed the June–August mean heating appears to do, then the modification of PV is in the “correct” sense to send a (negative) PV anomaly towards zero. When the heating maximum was placed above the inflow, the sense of modification was wrong and PV anomalies were magnified. In this latter case, however, the easterlies were completely cut off and the mass source for the inflow became the Southern Hemisphere midlatitude westerlies with northward flow from the Southern Hemisphere midlatitudes along the entire east coast of Africa. Naturally, such a change in circulation will alter the whole inflow problem. The PV signature of midlatitude air, being so much stronger than that of tropical air, means that even with modification, significant negative PV may be expected to occur over the Arabian Sea. This change to a midlatitude mass source for the inflow is observed periodically during the monsoon season; although it may be more associated with in situ synoptic weather events (Bannon 1979; Cadet and Desbois 1981) than to changes in the vertical profile of heating over the tropical Indian Ocean. Figure 2a shows the 887-hPa winds for 11 June 1994 from ECMWF analyses. Here there is a good example of a midlatitude “injection” of air into the monsoon inflow, and this appears to be associated with an anticyclone over southern Africa. Southerly winds of up to 15 m s−1 are seen over the Mozambique channel and, using this windspeed, one can estimate a timescale of about 5 days for a PV anomaly to be advected from southern Africa to the eastern Arabian Sea.

Figure 2b shows the wind field 5 days later, on 16 June, and indeed there is southward turning of the flow, particularly to the south and west of the tip of the Indian peninsula. This southward turning might not seem very strong or significant but one must remember that the mean westerlies exist throughout the monsoon season and it is these relatively small changes that occur alongside the rather larger intraseasonal variability in rainfall. Although the Eulerian picture might be little changed, it is the Lagrangian perspective that is important for moisture fluxes, and in this respect the flow may be very different indeed.

The southward turning is still seen a further 5-days later, in Fig. 2c, on 21 June. At this time, the westerly flow over India is significantly weaker. Through the linearity arguments of Hoskins and Rodwell (1995), reduced atmospheric latent heating over southern Asia, and therefore reduced rainfall, would be consistent with this weaker westerly flow.

Figure 3a shows the June–August 1994 time series of the 887-hPa meridional wind component near Madagascar, averaged over the box indicated. The 10 m s−1 “spike” seen on 10–11 June corresponds to the midlatitude injection seen in Fig. 2a. The lower panel, Fig. 3b, shows the time series of the all-Indian rainfall percentage of normal and its 7-day running mean. Generally during this monsoon season, the rainfall appears to be above normal (100%) and, indeed, the seasonal mean all-India rainfall for 1994 was about 10% above average, consistent with the stronger than normal monsoon trough seen in ECMWF analyses (not given here). This simple percentage of normal index is particularly efficient at identifying intraseasonal variability because longer timescales associated with the establishment and withdrawal of the mean monsoon are effectively eliminated. Active phases of the monsoon are associated with heavier than normal rainfall, whereas monsoon breaks are characterized by reduced rainfall. The 40% drop to 60% of normal in the running mean on 21 June signifies a strong break and is consistent with the wind field in Fig. 2c. Such a break, of which there are generally two or three in a season (Krishnamurti and Bhalme 1976), can have severe consequences for agriculture and domestic water use. A 40% drop in rainfall over an 8-day period represents a reduction of about 4% or 5% in the seasonal mean all-India rainfall, or about half the 9.7% total standard deviation of seasonal mean rainfall, as given in, for example, Rasmusson and Carpenter (1983). A further 10 m s−1 spike is seen on 28 July and this too is followed by southward turning (not shown here) and again, after about 11 days, a break in monsoon rainfall around 8 August.

What is of interest here is not the Southern Hemisphere midlatitude synoptic systems per se; it is the injections that some such weather systems can produce. There are many more than two weather systems during the monsoon season, but only two during 1994 led to injections of nearly 10 m s−1 averaged over the box shown in Fig. 3a. To achieve such an strong injection, a weather system has to be sufficiently strong and sufficiently equatorward.

Clearly there could be many other processes that may account for this rainfall time series, but these observations with two injections, each followed by a rainfall break after the same length of time, do at least suggest that further study of the mechanism proposed here is required.

Other studies have demonstrated that ridges in the Southern Hemisphere midlatitudes can cause surges in the cross-equatorial flow (Bannon 1979; Kumar et al. 1983) and that these surges are correlated with increased rainfall over northwestern India (Findlater 1969). Although the present study suggests that the most significant response to a midlatitude injection may be the triggering of a break in monsoon rainfall, there is, due to a difference in timing, no contradiction with these previous papers. This point will be discussed again later.

The present study concentrates primarily on the Indian monsoon, although it will be demonstrated that the mechanism might have a more general influence on the entire Asian summer monsoon. The aim of this study is to try to isolate and evaluate the midlatitude injection mechanism from all the other processes. This will be done with the help of an idealized global primitive equation model in section 2 and the United Kingdom Meteorological Office GCM, the Unified Model, in section 3. A discussion of the implications of this mechanism is given in section 4.

2. Idealized modeling

a. Model description

The model used in this investigation is explained in Hoskins and Rodwell (1995) and only a brief description will be given here. It is a time-dependent, global, hydrostatic, primitive equation model derived from Hoskins and Simmons (1975). It is nonlinear, spectral in the horizontal, and uses finite differences in σ coordinates in the vertical. The horizontal resolution is T31 and there are 15 levels in the vertical. The model is initiated with the zonal mean of a 6-yr June–August climatology derived from initialized analyses from ECMWF for 1983–88. A smoothed earth orography is raised over the first 5 days of an integration and hydrostatic adjustments are made to the temperature and surface pressure so that a near steady-state solution to topography is achieved as quickly as possible. The idealized model uses a simple linear drag in the lowest two levels, σ = 0.887 and 0.967, on timescales of 5 days and 1 day, respectively, over the oceans, and 5/4 and 1/4 days, respectively, over land areas. A ▿6 horizontal hyperdiffusion, with a timescale of 6 h on the smallest resolved horizontal scale, is applied to the vorticity, divergence, and temperature.

From day 5, a realistic time-mean three-dimensional diabatic forcing is applied. This forcing is identical to that used in the standard integration of RH. Although the model does not contain moisture or clouds explicitly, this heating mimics the effects of, for example, latent heat release and radiation. This simplification, although intuitively rather severe when considering a monsoon, is very useful as it decouples moisture supply complications from the circulation response to heating. Experiments with this model are essential to bridge the gap between theoretical ideas and the real atmosphere as represented in a GCM. The GCM will be used later to explore the fully coupled system.

Initially, after turning on the heating, the modeled circulation evolves rapidly. The low-level westerly winds over India, for example, increase in an explosive fashion somewhat like the real monsoon onset. Some 5 days later, the circulation pattern begins to asymptote to a flow very similar the “observed” June–August mean flow (Hoskins and Rodwell 1995) Figures 4a–d show the 887-hPa winds and shaded PV field at days 11, 15, 21, and 25, respectively, for this control simulation. The fact that the flow is very similar to the observed flow suggests that, if moisture were included in the model, then the moisture supply to the monsoon region would be consistent with the imposed heating. Beyond about day 25 of the simulation, growing, mobile waves, presumably associated with baroclinic instability, begin to affect the Indian Ocean region. Such waves, which would actually be unrealistic in this idealized model, are of no explicit interest in the following experiment. There is sufficient time before day 25 to test the hypothesis of this paper.

b. Experimental design

The model’s simplicity means that the effects of changes made to it can be relatively easily understood. Here a comparison is made of the control model’s simulation and that of the same model but with a realistic anticyclonic weather system in the Southern Hemisphere midlatitude westerlies. This later simulation is referred to as the perturbation simulation. The anticyclone in the “Roaring Forties” is transient in nature, forced by a moving vorticity source, Fig. 5, which is turned on at day 9 of the simulation and follows the track shown in Fig. 5 until day 13 when it is turned off. The vertical structure of the forcing is equivalent barotropic, being constant with height up to 400 hPa and then gradually reducing to zero by 125 hPa. The maximum central vorticity forcing is 2 × 10−5 s−1 day−1. The track, vertical and horizontal structure, and magnitude of the forcing were designed to recreate the evolution of the anticyclone seen in ECMWF analyses that leads to the injection of midlatitude air on 10 June 1994, as seen in Fig. 2a. A future extension of this work may be to consider other forcing structures, with different vertical profiles, but this is beyond the scope of the present study. Although this southern hemispheric forcing is anticyclonic and therefore has a positive central forcing on vorticity, it is the air on the eastern flank of the anticyclone that is envisaged entering the monsoon inflow. The hypothesis, which will be confirmed, is that this air will retain its strong negative PV signature.

c. Idealized model results

Figures 4e–h shows the low-level winds and shaded potential vorticity for the same days, 11, 15, 21, and 25, of the perturbation simulation. The darkest shading corresponds to strongly negative PV, generally in the Southern Hemisphere. The shading becomes lighter as negative PV gets closer to zero. The medium dark to light shading shows progressively more positive PV, generally in the Northern Hemisphere. The numbers in parenthesis in the figure represent the number of days from when the forcing was turned on and it is this dating convention that will be used exclusively from now on (note that the forcing itself is turned off at day 4). At day 2 (Fig. 4e), the anticyclone is clearly evident in both the wind vectors and PV field near southern Africa. South of Madagascar, in agreement with the northward winds and the conservation properties of potential vorticity, the PV contours are displaced more north than in the control simulation (Fig. 4a). Farther north, there is little difference in the potential vorticity field between control and perturbation simulations at this time.

Between days 2 and 6, a surge in cross-equatorial flow occurs in the perturbation simulation, reaching its maximum around day 4 when cross-equatorial flow is about 12 m s−1 compared with only 7.5 m s−1 in the control simulation. Such surges have been observed (Findlater 1969; Rao and Haney 1982) and modeled (Bannon 1979) in past studies. If this surge represents a pulse of tropical air, ahead of the air of midlatitude origin, then one might expect to see a temporary increase in rainfall over India.

By day 6 (Fig. 4f) potential vorticity contours over the western Indian Ocean are displaced considerably farther north, with substantially more negative PV north of the equator and stronger winds over the Arabian Sea than in the control simulation. The flow over the Arabian Sea is diffluent with much of the air turning south and avoiding India. In contrast, the flow of the control simulation (Fig. 4b) does not turn south. This result suggests that, in the real atmosphere, the stronger PV signature of the Southern Hemisphere midlatitudes may indeed get across the equator and could influence the course of the monsoon. Note also the cutoff of the easterly flow east of about 60°E over the tropical southern Indian Ocean and, at the same time, the strengthening of the southeasterly winds from northern Madagascar toward the east coast of Africa. These latter winds are clearly associated with the anticyclone and occur in a similar location to where Findlater (1974) found very strong southeasterlies during 1972.

There is little evidence left by day 12 (Fig. 4g) of the anticyclone in the Southern Hemisphere, although there is still enhanced negative potential vorticity over the Arabian Sea and southward diversion of the flow on approaching India. The intensity of the model’s monsoon remains unaltered because the imposed (latent) heating is unaltered. Hence the diversion of the flow in this model really must be a consequence of the now-dissipated weather system that occurred a week earlier far away in the Southern Hemisphere.

Day 16, Figs. 4d and 4h, is included to demonstrate that the perturbed flow does eventually revert back to something very similar to the control simulation. The similarity perhaps implies that the previous changed circulation is not an unrealistic and growing instability within the model.

The fact that the mass source of the low-level monsoon inflow really is changing fundamentally in response to the anticyclone can be demonstrated more conclusively by examining the “back trajectories” shown in Figs. 6a and 6b. The trajectories, which are calculated in three-dimensional space, show the origins of individual air particles that cross the west coast of India at level σ = 0.89 on day 11 of the two model simulations and where they subsequently went.

If, in the real world, deep convection was to occur at some point along the axis of the inflow, then these model trajectories may not be representative of “real” trajectories. However, infrared satellite pictures, not presented here, show virtually no deep convection along the inflow axis during the monsoon season. The present study implicitly makes use of this fact. There may be several reasons for the lack of deep convection along the axis of the inflow. The fact that the inflow is gliding down the sloping isentropes towards the intertropical convergence zone means that air parcels are continually being warmed adiabatically, acting so as to reduce relative humidity. Also, during the established monsoon, sea surface temperatures under the axis of the jet are strongly reduced relative to the tropical ocean areas to the east.

The many circles over the southern Indian Ocean in the control simulation (Fig. 6a) show that the air particles spend a considerable amount of time in this region, about 2 weeks in fact, and might be expected to adopt the characteristics of warm, moist tropical air with the mean observed moisture content (Saha 1970; Kishtawal et al. 1994) of about 16 g kg−1 and with ambient (small) PV signature. Indeed, in the control simulation, as these particles cross the tropical Indian Ocean, their potential vorticity changes smoothly in response to diabatic and frictional effects so that they really do become “tropical air particles.” In contrast, the perturbation simulation shows many of the air particles originating farther south, Fig. 6b, and higher up, Fig. 6c, in the atmosphere at about 650 hPa where they would be expected to have a stronger negative PV signature and be much drier (Saha 1970) with a moisture content of perhaps 1 g kg−1. The speeds and three-dimensional trajectories of these “perturbation particles” mean that they have at most 4 days in the boundary layer over the tropical Ocean–Arabian Sea and this may not be enough time for them to attain the mean 16 g kg−1 of moisture. If one assumes a sustained moistening of the planetary boundary layer of 2 g kg−1 day−1, corresponding to a realistic 125 W m−2 latent heat flux (Esbensen and Kushnir 1981; Gordon and Wright 1995), then this would imply that about 7 or 8 days are required for the particles to attain the mean observed humidity. The idea that such midlatitude air may retain its dry signature is consistent with observations in Rao and Haney (1982), who found a mixing ratio minimum of only 6 g kg−1 at 850 hPa in the cross-equatorial jet core during an injection in 1977.

Figures 6a and 6b show that the particles of the perturbation simulation come into closer proximity with the Ethiopian Highlands than those of the control simulation and the lateral torque exerted on the flow by the frictional effects of the model’s mountains is nearly twice as strong as in the control simulation, not shown here. This is an example of the proposed “compensation phenomenon” (see RH), whereby reduced upstream PV modification or increased initial PV signature may be compensated for by increased interaction with the east African Highlands. The compensation may imply that the negative PV of the air is changed enough that it does not completely avoid the Indian peninsula and this may well be important for ensuring that the monsoon never completely fails. In terms of moisture fluxes, one might expect that some of the moisture carried by the air would be deposited over Kenya and the Horn of Africa (as part of the rain of the Ethiopian “Kremt” rainy season, for example) and that much of the moisture for the Indian monsoon rainfall would have to be picked up subsequently over the Arabian Sea.

In agreement with the wind vectors (Figs. 4f and 4g), the air particles of the perturbation simulation do turn more sharply south on approaching India. The combined effect of the deflection of the flow and the fact that the air could be somewhat drier must be that the moisture fluxes across the west coast of India are reduced. It is speculated that this reduction in moisture flux into India may be able to trigger a break in monsoon rainfall, which, in the real case with an interactive monsoon, could perhaps be perpetuated by suppressed latent heating over the land and enhanced latent heating over the equatorial Indian Ocean.

Other trajectories (not given here) show that on about day 7, particles arriving at the west coast of India originated farther south, and had a stronger negative PV signature, than those of the control run. Particles of true midlatitude origin begin to arrive from about day 9. These results suggest that the break effect of the southern hemispheric disturbance on the rainfall of India will begin to occur about a week after the initial injection.

One could speculate that during this Indian monsoon break, the increased mass flux south of India may increase moisture fluxes into southeast Asia. Although generally attributed to an El Niño relationship with the Asian summer monsoon, an east–west dipole in rainfall (and outgoing radiation) has been observed as a dominant mode of variability (Rasmusson and Carpenter 1983; Bess et al. 1992) with enhanced rainfall over southeast Asia and eastern India accompanying reduced rainfall over western India and vice versa. However, it will be shown that GCM results do not link this mode with midlatitude injections.

The anticyclone does, locally, intensify the Mascarene high and it might seem counterintuitive that an increase in the north to south pressure gradient could occur alongside a break in rainfall. If the Mascarene high (Fig. 7a) were to intensify on an oceanwide scale and for a prolonged period of time, then clearly this would be consistent with strengthened (low PV) easterlies and perhaps a sustained increase in rainfall over India. Here, however, the Mascarene high is seen to be modulated on a shorter length scale (Fig. 7b) and the easterlies are replaced by a (high PV) mass source from the south. Ahead of this midlatitude air, there could well be a pulse of tropical air that may lead to a temporary increase in Indian rainfall. In the real atmosphere, both processes could be competing with each other. There will be further discussion of this topic in section 4.

3. GCM modeling

a. Model description

The GCM used is the “HadAM 2B with 3CV physics” version of the United Kingdom Meteorological Office Unified Model (UM). An overview of the model is given in Cullen (1993). It is a gridpoint, hybrid-σ, p-vertical coordinate model and, in the climate mode used here, has a horizontal resolution of 2.5° × 3.75° in latitude and longitude with 19 layers in the vertical. The split–explicit time step is 30 min. The model has prognostic cloud water and ice and uses mean orography. There is a mass-flux convection scheme with stability closure. Importantly (Inness and Gregory 1994), convective downdrafts are included, which, by reducing the stability of the tropical (oceanic) boundary layer, lead to the establishment of realistic monsoon precipitation. Sea surface temperatures, sea ice, and the annual solar cycle are prescribed.

The model produces a very reasonable monsoon simulation both in terms of mean and intraseasonal variability (Inness and Gregory 1994; Ashok et al. 1995; Hall et al. 1995). Figure 8 shows the mean 850-hPa wind vectors and isotachs calculated from a 10-yr integration of the same model using observed sea surface temperatures from 1979 to 1988. The main monsoon errors are that the low-level flow is generally between 2 and 5 m s−1 too strong, rainfall over the Bay of Bengal and eastern peninsular India is perhaps a little too high and the monsoon onset is too rapid. While these errors should be noted, they are not considered serious for the present investigation.

b. Experimental design

Depressions tracking westward along the “monsoon trough” (Saha et al. 1981), complicated interactions with the land surface and many other processes may lead to variability of monsoon rainfall in reality and in GCMs. It would be very difficult to isolate the mechanism proposed here (and to justify how it was isolated) without having first investigated the response of the idealized model. Another problem is that there is a continuous stream of perturbation activity in the Southern Hemisphere midlatitudes; if the anticyclonic forcing is applied one top of a preexisting cyclone, then they tend to cancel each other and therefore strongly reduce the strength of the injection. The approach taken here is to identify the track of a preexisting anticyclone within a control integration and then, in the perturbation integration, to go back and enhance the anticyclone along this track for a period of 4 days with the vorticity forcing.

The vorticity forcing used in the UM has the same vertical structure and similar horizontal structure as in the idealized model, but it is applied to the gridpoint winds rather than to a spectral vorticity field. Its precise form is given in the appendix.

Whereas one perturbation integration was sufficient for the idealized model, this will clearly not be the case here. As will be demonstrated, four independent GCM integrations were found to be sufficient for the purposes of this study. Two control runs were made with sea surface temperatures from the 1986 monsoon season; one initiated on 1 May and the other on 1 June, using initial conditions from a longer integration of the model forced with observed sea surface temperatures. These two control integrations provided a sufficient number of preexisting anticyclones for the perturbation runs. Three anticyclones were identified and used from the 1 June control run. Correspondingly, the perturbation forcings were applied from 12 June, 27 June, and 2 July in three separate model runs. The fourth perturbation integration used an anticyclone from the 1 May integration and here the forcing was applied from 21 June. In all cases the monsoon was well established prior to the forcing being turned on. Care was taken to ensure that the perturbation runs were numerically identical to the appropriate control prior to the perturbation being applied.

Since the positions and strengths of the final anticyclones will differ, one might expect to see differences in the timing of any response over India. However, if the mechanism does have a critical role to play in the monsoon, then one might hope that a simple composite would be adequate to isolate the model response. In the composite given here, day x is the mean of the four perturbation integrations x days after the forcing was turned on. A composite of the control integrations for the same corresponding days was also calculated.

c. GCM results

Initially, a similar anticyclone is seen over southern Africa to that in the idealized model given in Fig. 4e. At day 4 of the composite there is a pulse in the cross-equatorial flow with meridional windspeed increased by about 4 m s−1, again similar to that in the idealized case.

Figure 9a shows the 850-hPa winds for day 9 of the control composite and Fig. 9b shows the same for the perturbation composite. Day 9 is chosen because this is the time when the anomaly composite (i.e., perturbation minus control, given in Fig. 9g) over India has maximum amplitude. After day 9, the individual integrations tend to become less coherent with each other and the anomaly composite is strongly reduced. Day 9 also agrees well with the timing suggested by the idealized model results.

The main differences between the control and perturbation winds in the Indian region (Figs. 9a and 9b) would appear to be a southward shift of the westerly flow over India, with increased easterly flow along the southern foothills of the Himalayas. The anomaly composite (Fig. 9g) shows these features. The anomalous southward flow over the western Arabian Sea indicates that the inflow is turning more sharply in the perturbation composite and that the influence of the drier Saudi Arabian air may also be stronger. Over the southern Arabian Sea, the anomalous westerly flow indicates a strengthening in the mean westerlies, and the anomalous northward flow over the Bay of Bengal implies stronger northward turning of the flow into the head of the bay. Another feature to note in the anomaly composite is the reduction in westerly flow over Malaysia and this, together with the strengthened westerlies over the southern Arabian Sea and anomalous northward flow across the equator into the Bay of Bengal, suggests increased convergence south of India and over the southern Bay of Bengal. There is also a significant onshore anomaly into eastern equatorial Africa, similar to that seen in the idealized integration in Fig. 4f.

To demonstrate that the somewhat small differences between control and perturbation composites are “significant,” the anomaly composite has been broken down into the four individual anomalies. It is seen (not shown here) that the compositing procedure does indeed filter out much of the spurious “noise,” particularly in the Southern Hemisphere midlatitude westerlies and various other small-scale activity even in the Northern Hemisphere. What remains, it is hoped, are only the coherent features common to all or most of the integrations. One can see in the four individual anomalies, an1–4 in Figs. 9c–f, that, despite the differences in timing of the response warned of above, the reduction in westerlies over northern India is a feature common to all four runs. Similarly, the southward anomaly over the western Arabian Sea and the strengthening of the westerly flow over the southern Arabian Sea are common to three of the integrations on this day. In fact, these anomaly features are the dominant northern hemispheric features to emerge from the composite. It is also worth noting that the anomalous onshore flow into eastern equatorial Africa is a common feature to all integrations, although it is particularly strong in one of them, an2 Fig. 9d. Hence it would appear that this composite anomaly in the Indian Ocean–southern Asia region shown in Fig. 9g is a coherent response and that four integrations is a sufficient number.

Figure 10b shows the day 10 control and perturbation composite zonal moisture fluxes, uq, across the model longitude of 69.375°E (indicated in Fig. 10a), vertically integrated from 1000 to 700 hPa. Here u is zonal wind and q is specific humidity. One sees a southward shift from control to perturbation of about 2°–4° latitude. Figure 10c shows the composite anomaly moisture flux at days 7–13. There is a clear progression over days 7, 8, 9, and 10 with moisture fluxes into northern India, north of about 12°N, reducing. South of about 12°N, the opposite is the case with moisture fluxes generally increasing. This seesaw in moisture fluxes is consistent with the southward shift of the westerly flow seen in Fig. 9. At the model latitude of 18.75°N (approximate latitude of Bombay) on day 10 (Fig. 10b), there is a 34% reduction in moisture flux and this percentage increases rapidly farther north so that by about 21°N there is a full 100% reduction in moisture flux. After day 10, Fig. 10c shows the seesaw tips back again with moisture fluxes increasing in the north and decreasing in the south.

The composite, vertically integrated uq anomaly appears to be predominantly due to changes in u and not to changes in q. One of the four runs that make up the composite (corresponding to an1 in Fig. 9c) did, however, show a southward shift of the humidity pattern on the 69°E meridian, perhaps implying more influence of the drier Saudi Arabian air. The fact that the perturbation composite air is not drier (as had been speculated from simple calculations in section 2b above) might be in part due to the fact that the modeled latent heat flux from the Arabian Sea is sustained for much of the monsoon season at 500 W m−2, quite a lot larger than the value used in the calculation above (and indeed seen in the climatology maps).

This southward shift in moisture fluxes is impressively reflected in the rainfall field. Figure 11b shows the composite anomaly rainfall averaged over days 6–13. Anomaly rainfall is a particularly difficult field to display; here it is shown on a logarithmic scale in both positive and negative regions. Days 6–13 represent a rather long 8-day period, similar to the duration of the two breaks in the observed all-India rainfall given in Fig. 3b and centered on days 9 and 10 when moisture flux anomalies are most extreme. Figure 11a shows the composite control rainfall averaged over the same days. Perhaps by chance, this 8-day mean control rainfall is very similar to the model’s 5-yr climatology, given in Inness and Gregory (1994).

Compositing removes much of the noise in the anomaly field (Fig. 11b) and the monsoon response emerges as by far the strongest and spatially coherent feature in the Northern Hemisphere. Indeed, the monsoon rainfall anomaly is stronger in magnitude than that in the Southern Hemisphere midlatitudes themselves. There is also a strong response in the South Pacific convergence zone—perhaps a tropical response related to a coherent feature moving in the extratropical midlatitudes (Matthews et al. 1996).

Figure 11b shows a clear decrease in rainfall north of about 12°N over central and western India and an increase south of 12°N. This pattern is quantitatively consistent with the moisture flux anomalies seen in Fig. 10c and inline with what this midlatitude injection mechanism would predict. A greater than −2.5 mm day−1 reduction is seen over a large part of central and northwestern India and, since the control gives a total of less than about 6 mm day−1 in this region, this anomaly represents a 40% decrease in rainfall, sustained over an 8-day period. This is similar in magnitude to the breaks seen in the all-India rainfall series during 1994 in Fig. 3b. Over the southern tip of the Indian peninsula and the equatorial Indian Ocean, anomalies of +4 mm day−1 on top of a control of around 8 mm day−1 represent a 50% increase in rainfall.

In Fig. 11b, there is an indication of increased rainfall along the southern edge of the Himalayas (line joining CD). This is consistent with the increased easterly flow seen in Fig. 9b compared to Fig. 9a, and is interesting because it is widely believed that the monsoon trough shifts north during a break period. This northward shift can be seen, for example, in rainfall empirical orthogonal functions associated with break conditions in Bedi and Bindra (1980) and Kulkarni et al. (1992).

In Fig. 11b, the decrease in rainfall over northwestern India may be seen as part of a more general decrease along the line joining AB and, as might be expected from the nonnegative nature of the rainfall field, the largest decreases occur where there is largest rainfall in the control. This more general decrease, which was not suspected from the idealized model results, suggests that the cross-equatorial PV mechanism may apply to the entire Asian monsoon and not just to the Indian part of it. Further study is required on this generalization.

The decrease in rainfall over western southeast Asia was particularly unexpected from the idealized results. It is perhaps related to the stronger northward turning of the flow over the Bay of Bengal and the apparent northward shift of the monsoon trough to the foothills of the Himalayas (Fig. 9). These features are strongly constrained in the idealized model by the prescribed heating field.

Increased rainfall over the Kenyan coast (>8 mm day−1 increase) on top of a control of 4 mm day−1 represents a 200% increase. The stronger onshore flow was noted above but whether the magnitude of this increase is realistic or due to model instabilities at the equator or to the precise nature of the prescribed vorticity forcing is not clear. However, Findlater (1974) reported 88 mm of rainfall in exactly this location (at Mombasa, Kenya) on 5 October 1972; this occurred 2 days after the very strong southeasterlies of 65 m s−1, mentioned above, were observed north of Madagascar.

Figure 11c shows the daily time series of anomalous rainfall averaged in the two boxes drawn in Fig. 11b, together with a 3-day running mean. Despite reflecting the day 6–13 anomaly in Fig. 11b and therefore perhaps likely to maximize during this period, these indices do give some extra insight into the problem.

Over the first 3 days, there is little signature in these indices, but by day 5, the daily composite for the northern box shows an increase to +2 mm day−1 anomaly rainfall. In fact, this is a feature common to all four of the integrations and may be related to a coherent “pulse” of tropical air ahead of the midlatitude injection, associated with the temporary increase in cross-equatorial flow speed mentioned above. This result is consistent in its timing with previous studies but, unlike former studies, appears here not to be the major consequence of a ridge moving into the Mascarene high.

After day 5, one sees the emergence of the anomaly pattern of Fig. 11b. The length of time that this anomaly persists appears to agree with the speculation above that the break may be triggered by a midlatitude injection but that, in part at least, it is perpetuated by the subsequent lack of latent heating over the land.

4. Discussion

Oscillations in the strength of the Mascarene high have been linked to variability of monsoon rainfall in previous studies (e.g., Krishnamurti and Bhalme 1976). Kumar et al. (1983) argued that a pulse in cross-equatorial flow in the second half of May, as a ridge passed over the eastern coast of South Africa, could lead to the onset of the Indian monsoon and that the lack of such a pulse may result in a delayed onset. This is consistent with the present results, which show such a cross-equatorial pulse of tropical air and a coherent, temporary increase in rainfall over much of the west coast of India (in this case during the established monsoon). Other authors (Findlater 1969; Cadet and Desbois 1981) have found correlations between the strength of the cross-equatorial flow and Indian rainfall during the established monsoon. One of the interesting results of the present study, however, is that this increase in rainfall may not be the dominant effect of a ridge in the Southern Hemisphere midlatitudes. The present results suggest that the rainfall over central and northwestern India could, subsequently, fall by as much as 40% averaged over an 8-day period in response to a midlatitude injection of air into the low-level monsoon inflow, with a corresponding 50% increase in rainfall over the southern tip of the Indian peninsula and the equatorial Indian Ocean. These changes in precipitation are consistent with the southward shift of the westerly flow and zonal moisture fluxes over the western coast of India.

Interestingly, when one considers the timescales suggested by the present study, figures given in Kumar (1992), for the 1979 MONEX data, reveal a similar time lag between pressure ridge passing over southern Africa and monsoon break. The quasi biweekly oscillation of Krishnamurti and Bhalme (1976) also showed a minimum in monsoon rainfall lagging about 9 days behind the maximum intensity of the Mascarene high. A statistical survey using the time lag found in this study would be helpful to further ascertain the significance of this mechanism. It is also interesting to note that the latitudinal oscillation of the monsoon westerly moisture fluxes seen in this study in Fig. 10c may have some relation to the northward propagation of monsoon cloud bands, observed by Sikka and Gadgil (1980) and modeled by Webster (1983).

The band of reduced rainfall (AB in Fig. 11b) from northwestern India to the Philippines suggests that perturbations in the Southern Hemisphere midlatitudes may be able to influence more than just the Indian part of the Asian summer monsoon. Intraseasonal modes of variability can be reflected in the seasonal anomalies (Ferranti et al. 1997), and, indeed, the seasonal anomaly for 1994 (from ECMWF analyses not given here) shows the same band AB, but with the opposite sign, implying a strengthening of the entire monsoon trough across northern Indian and southeastern Asia. In the light of the present study, this could be partly a consequence of diminished Southern Hemisphere synoptic activity or a poleward shift of the Southern Hemisphere midlatitude westerlies during the June–August season, thereby reducing the ability of synoptic activity to inject high PV air into the monsoon inflow.

A southward error in the location of the monsoon is a feature common to several GCMs at present (Sperber and Palmer 1996). In addition to possible local reasons for this error, such as land–surface feedbacks, for example, it could reflect too much perturbation activity in the modeled Southern Hemisphere midlatitudes or insufficient material modification of PV over the Indian Ocean.

A southward shift of the monsoon in reality, or in a numerical weather prediction model, may have implications outside the monsoon system. Rodwell and Hoskins (1996) showed that the monsoon can force remote descent to its west and northwest. The very dry summertime climate of the Mediterranean and surrounding lands may be strongly related to this. They also showed that this descent is highly dependent on the latitude of the monsoon heating and a southward shift, for example, may lead to wetter weather, or weather forecasts, for southern Europe. There is also an interesting coherent and large increase in rainfall over eastern equatorial Africa that is consistent with reports made by Findlater (1974).

Much interest is placed on being able to forecast the seasonal mean monsoon rainfall; see Gowariker et al. (1991), WMO (1992, 1993), Ju and Slingo (1995), Mooley and Parthasarathy (1983), and Rasmusson and Carpenter (1983). The hypothesis is that this mean rainfall may be influenced by slowly varying “boundary conditions,” such as sea surface temperatures, either directly or indirectly by determining the frequencies of more transient weather systems over southern Asia. Some GCMs forced with observed sea surface temperatures do, indeed, show success in simulating monsoon interannual variability, particularly of the large-scale seasonal mean flow (Ju and Slingo 1995). However, there may be limits to the accuracy of long-range forecasts, particularly if a single event can contribute significantly to the seasonal anomaly. For example, the rainfall peak from 12 to 14 July 1994 (Fig. 3b) can be attributed to a strong monsoon depression over northwest India and this single event can be identified in the seasonal anomaly fields. From Fig. 3b, it can be seen that a single break event might reduce the all-India seasonal total rainfall by perhaps 4% or 5%—about half its standard deviation.

Ultimately, moisture fluxes into southern Asia must play a dominant role in the hydrology budgets of the region. If the hypothesis proposed in this paper is correct, that perturbation activity in the Southern Hemisphere midlatitudes can trigger monsoon breaks by altering moisture fluxes, then, given the chaotic nature of such midlatitude systems, a long-range forecast of monsoon rainfall to within half a standard deviation (±one break event) may be an impossible objective. On the shorter timescale, however, there could be some predictive skill to be gained by improving the observational network over the southern Indian Ocean and considering midlatitude injections into the monsoon inflow.

Acknowledgments

I would like to thank the Meteorological Office of the United Kingdom for making their “Unified Model” GCM available for my use. Special thanks go to Prof. B. J. Hoskins, Dr. J. M. Slingo, and Dr. M. Blackburn for numerous discussions on this topic. I would like to record my gratitude to Dr. P. Berrisford, Dr. J. Cole, and Dr. C. Jones for their considerable help with datasets and models.

REFERENCES

  • Anderson, D. L. T., 1976: The low level jet as a western boundary current. Mon. Wea. Rev.,104, 907–921.

  • Ashok, K., D. Gregory, and P. M. Inness, 1995: Intraseasonal variability of the Indian summer monsoon simulated by the Hadley Centre climate model. Hadley Centre Tech. Note CRTN-60, Hadley Centre, Bracknell, United Kingdom, 18 pp. [Available from Hadley Centre for Climate Prediction and Research, Bracknell RG12 2SY, United Kingdom.].

  • Bannon, P. R., 1979: On the dynamics of the East African jet. PartII: Jet transients. J. Atmos. Sci.,36, 2153–2168.

  • ——, 1982: On the dynamics of the East African jet. Part III: Arabian Sea branch. J. Atmos. Sci.,39, 2267–2278.

  • Bedi, H. S., and M. M. S. Bindra, 1980: Principal components of monsoon rainfall. Tellus,32, 296–298.

  • Bess, T. D., G. L. Smith, T. P. Charlock, and F. G. Rose, 1992: Annual and interannual variations of Earth-emitted radiation based on a 10-year data set. J. Geophys. Res.97, 12825–12835.

  • Cadet, D., and M. Desbois, 1981: A case study of a fluctuation of the Somali jet during the Indian summer monsoon. Mon. Wea. Rev.,109, 182–187.

  • Cullen, M. J. P., 1993: The unified forecast/climate model. Meteor. Mag.,112, 81–94.

  • Esbensen, S. K., and Y. Kushnir, 1981: Heat budget of the global ocean: Estimates from marine observations. Climate Research Institute Rep. 29, 27 pp. [Available from Oregon State University, Corvallis, OR 97331.].

  • Ferranti, L., J. M. Slingo, T. N. Palmer, and B. J. Hoskins, 1997: Relationships between interannual and intraseasonal monsoon variability as diagnosed from AMIP integrations. Quart. J. Roy. Meteor. Soc.,123, 1323–1357.

  • Findlater, J., 1969: A major low-level air current near the Indian Ocean during the northern summer. Quart. J. Roy. Meteor. Soc.,95, 362–380.

  • ——, 1974: An extreme wind speed in the low-level jet-stream system of the western Indian Ocean. Meteor. Mag.,103, 201–205.

  • Gordon, C., and D. K. Wright, 1995: A comparison of modelled surface fluxes with climatological estimates. Hadley Centre Tech. Note CRTN-53, Hadley Centre, Bracknell, United Kingdom, 25 pp. [Available from Hadley Centre for Climate Prediction and Research, Bracknell RG12 2SY, United Kingdom.].

  • Gowariker, V., V. Thapliyal, S. M. Kulshresthra, G. S. Mandal, N. Sen Roy, and D. R. Sikka, 1991: A power regression model for long range forecast of southwest monsoon rainfall over India. Mausam,42, 125–130.

  • Hall, C. D., R. A. Stratton, and M. L. Gallani, 1995: Climate simulations with the Unified Model: AMIP runs. Hadley Centre Tech. Note CRTN-61, Hadley Centre, Bracknell, United Kingdom, 92 pp. [Available from Hadley Centre for Climate Prediction and Research, Bracknell RG12 2SY, United Kingdom.].

  • Hart, J. E., 1977: On the theory of the East African low level jet stream. Pure Appl. Geophys.,115, 1263–1282.

  • Hoskins, B. J., 1974: The role of potential vorticity in symmetric stability and instability. Quart. J. Roy. Meteor. Soc.,100, 480–482.

  • ——, and A. J. Simmons, 1975: A multi-layer model and the semi-implicit method. Quart. J. Roy. Meteor. Soc.,101, 637–655.

  • ——, and M. J. Rodwell, 1995: A model of the Asian summer monsoon. Part I: The global scale. J. Atmos. Sci.,52, 1329–1340.

  • ——, M. E. McIntyre, and A. W. Robertson, 1985: On the use and significance of isentropic potential vorticity maps. Quart. J. Roy. Meteor. Soc.,111, 877–946.

  • Inness, P. M., and D. Gregory, 1994: Simulation of the Indian monsoon and tropical intraseasonal variability by a general circulation model. Hadley Centre Tech. Note CRTN-52, Hadley Centre, Bracknell, United Kingdom, 50 pp.[Available from Hadley Centre for Climate Prediction and Research, Bracknell RG12 2SY, United Kingdom.].

  • Ju, J. H., and J. Slingo, 1995: The Asian summer monsoon and ENSO. Quart. J. Roy. Meteor. Soc.,121, 1133–1168.

  • Kishtawal, C. M., N. Gautam, S. Jaggi, and P. C. Pandey, 1994: Surface-level moisture transport over the Indian Ocean during southwest monsoon months using NOAA/ HIRS data. Bound.-Layer Meteor.,69, 159–171.

  • Krishnamurti, T. N., and H. N. Bhalme, 1976: Oscillations of a monsoon system. Part I: Observational aspects. J. Atmos. Sci.,33, 1937–1954.

  • ——, J. Molinari, and H.-L. Pan, 1976: Numerical simulation of the Somali jet. J. Atmos. Sci.,33, 2350–2362.

  • ——, V. Wong, H.-L. Pan, R. Pasch, J. Molinari, and P. Ardanuy, 1983: A three-dimensional planetary boundary layer model for the Somali Jet. J. Atmos. Sci.,40, 894–908.

  • Kulkarni, A., R. H. Kripalani, and S. V. Singh, 1992: Classification of summer monsoon rainfall patterns over India. Int. J. Climatol.,12, 269–280.

  • Kumar, S., 1992: Interaction of mid-latitude systems in the Southern Hemisphere with southwest monsoon. Mausam,34, 405–410.

  • ——, K. Sethumadhavan, and M. K. Rao, 1983: Influence of frontal systems moving across southern Africa and southwest Indian Ocean on the onset of southwest monsoon over west coast of India. Mausam,43, 37–42.

  • Matthews, A. J., B. J. Hoskins, J. M. Slingo, and M. Blackburn, 1996: Development of convection along the SPCZ within the Madden–Julian oscillation. Quart. J. Roy. Meteor. Soc.,122, 669–688.

  • Mooley, D. A., and B. Parthasarathy, 1983: Indian summer monsoon and El-Niño. Pure Appl. Geophys.,121, 339–352.

  • Rao, G. V., and J. L. Haney, 1982: Kinematic and thermal structure of two surges of flow in the northern Mozambique channel area. Quart. J. Roy. Meteor. Soc.,108, 957–974.

  • Rasmusson, E. M., and T. H. Carpenter, 1983: The relationship between eastern equatorial Pacific sea surface temperatures and rainfall over India and Sri Lanka. Mon. Wea. Rev.,111, 517–528.

  • Rodwell, M. J., and B. J. Hoskins, 1995: A model of the Asian summer monsoon. Part II: Cross-equatorial flow and PV behavior. J. Atmos. Sci.,52, 1341–1356.

  • ——, and ——, 1996: Monsoons and the dynamics of deserts. Quart. J. Roy. Meteor. Soc.,122, 1385–1404.

  • Saha, K. R., 1970: Air and water vapour transport across the equator in western Indian Ocean during northern summer. Tellus,22, 681–687.

  • ——, F. Sanders, and J. Shukla, 1981: Westward propagating predecessors of monsoon depressions. Mon. Wea. Rev.,109, 330–343.

  • Sikka, D. R., and S. Gadgil, 1980: On the maximum cloud zone and the ITCZ over Indian longitudes during the southwest monsoon. Mon. Wea. Rev.,108, 1840–1853.

  • Sperber, K. R., and T. N. Palmer, 1996: Interannual tropical rainfall variability in general circulation model simulations associated with the Atmospheric Model Intercomparison Project. J. Climate,9, 2727–2750.

  • Webster, P. J., 1983: Mechanisms of monsoon low-frequency variability:Surface hydrology effects. J. Atmos. Sci.,40, 2110–2124.

  • WMO, 1992: Simulation if interannual and intraseasonal monsoon variability. WCRP-68, WMO/TD-470, Report of workshop, NCAR, Boulder, CO, 236 pp. [Available from World Meteorological Organization, Case Postale No. 2300, CH-1211 Geneva 20, Switzerland.].

  • ——, 1993: Simulation and prediction of monsoons. WCRP-80, WMO/TD-546, Report of workshop, New Delhi, India, 83 pp. [Available from World Meteorological Organization, Case Postale No. 2300, CH-1211 Geneva 20, Switzerland.].

APPENDIX

Vorticity Forcing for GCM

If C is the center of the forcing, then the momentum forcing at a point X is given by
i1520-0469-54-22-2597-eq2
where R =6371000 m is the radius of the earth, r is the arc angle in radians between X and C, V is the tangential wind at X about C, and the constants a = 103.53, and b = 1.51 × 10−5 s−1 day−1 determine the scale and strength of the forcing, giving a maximum tangential wind forcing of 10 m s−1 day−1 at an arc angle of 15.3°. At an arc angle of 30° the forcing is sent smoothly to zero so that there is no direct effect of the forcing further afield. Note that b represents the central (and maximum) forcing on vorticity.

Fig. 1.
Fig. 1.

Mean June–August 850-hPa wind vectors with shaded isotachs over the Indian Ocean region calculated using analyses from the European Centre for Medium-Range Weather Forecasts for 1992–95.

Citation: Journal of the Atmospheric Sciences 54, 22; 10.1175/1520-0469(1997)054<2597:BITAMT>2.0.CO;2

Fig. 2.
Fig. 2.

Low-level wind vectors at 887 hPa with shaded isotachs over the Indian Ocean region taken from 1994 analyzed data from the European Centre for Medium-Range Weather Forecasts all at 12 UTC. (a) 11 June, injection; (b) 16 June, southward turning; and (c) 21 June, monsoon “break.”

Citation: Journal of the Atmospheric Sciences 54, 22; 10.1175/1520-0469(1997)054<2597:BITAMT>2.0.CO;2

Fig. 3.
Fig. 3.

(a) The low-level northward wind component at 887 hPa averaged over the box indicated covering the Mozambique channel region using the same 1200 UTC data as in Fig. 1. (b) The daily all-India rainfall percentage of normal calculated operationally by the Indian Meteorological Department for the monsoon season of 1994, together with a 7-day running mean.

Citation: Journal of the Atmospheric Sciences 54, 22; 10.1175/1520-0469(1997)054<2597:BITAMT>2.0.CO;2

Fig. 4.
Fig. 4.

Low-level winds vectors with shaded contours of potential vorticity at 887 hPa over the Indian Ocean region for the idealized model simulations. The left panels, (a)–(d), show a sequence of days from the model control simulation; the right panels, (e)–(h), show the same days for the perturbation simulation where a realistic anticyclonic weather system has been introduced into the Southern Hemisphere midlatitudes. The numbers in parenthesis indicate the number of days after the anticyclonic forcing was turned on. The shading units are PV units.

Citation: Journal of the Atmospheric Sciences 54, 22; 10.1175/1520-0469(1997)054<2597:BITAMT>2.0.CO;2

Fig. 5.
Fig. 5.

Vorticity forcing used in the idealized model with contours 0.5, 1.0, and 1.5 × 10−5 s−1 day−1 shown. The forcing moves within the model along the track shown. Circles are printed at successive days.

Citation: Journal of the Atmospheric Sciences 54, 22; 10.1175/1520-0469(1997)054<2597:BITAMT>2.0.CO;2

Fig. 6.
Fig. 6.

(a) and (b): The horizontal projection of three-dimensional particle trajectories within the idealized model simulations. The trajectories are calculated both backward and forward from day 11 when the particles are aligned along the west coast of India at σ = 0.89 (black dots). Circles are drawn every other day. The small gray filled circles indicate that a particle came too close to the earth’s surface for its path to be traced beyond that point. (a) Control simulation, (b) perturbed simulation also showing the vorticity forcing applied in this integration and its track. The contours shown for the forcing are 0.5, 1.0, and 1.5 × 10−5 s−1 day−1 and the circled numbers, 1 and 3, show its location 1 and 3 days from when it was turned on. (c) Height in hPa against time of all the particles shown in (a) and (b) except for those corresponding to the four northmost and three southmost black dots at day 11.

Citation: Journal of the Atmospheric Sciences 54, 22; 10.1175/1520-0469(1997)054<2597:BITAMT>2.0.CO;2

Fig. 7.
Fig. 7.

Mean sea-level pressure from (a) the control integration and (b) the perturbation integration. Both panels correspond to the time 2 days after the anticyclonic forcing was turned on in the perturbation integration. The contour interval is 2 hPa; “MH” signifies the Mascarene high, “MT” the monsoon trough.

Citation: Journal of the Atmospheric Sciences 54, 22; 10.1175/1520-0469(1997)054<2597:BITAMT>2.0.CO;2

Fig. 8.
Fig. 8.

Mean June–August 850-hPa wind vectors with shaded isotachs over the Indian Ocean region calculated from a 10-yr integration of the GCM forced with observed sea surface temperatures from 1979 to 1988.

Citation: Journal of the Atmospheric Sciences 54, 22; 10.1175/1520-0469(1997)054<2597:BITAMT>2.0.CO;2

Fig. 9.
Fig. 9.

Low-level 850-hPa winds from GCM integrations. (a) Composite control run at day 9, (b) composite perturbation run at day 9, (c)–(f) anomalies at day 9 for the four individual perturbation runs, and (g) composite anomaly at day 9. The 20 m s−1 reference arrow applies to full fields and the 10 m s−1 reference arrow applies to anomaly fields.

Citation: Journal of the Atmospheric Sciences 54, 22; 10.1175/1520-0469(1997)054<2597:BITAMT>2.0.CO;2

Fig. 10.
Fig. 10.

(a) Map showing 69.375°E meridian at which GCM modeled zonal moisture fluxes are evaluated. (b) Zonal moisture fluxes across the meridian shown in (a), vertically integrated from 1000 to 700 hPa at day 10 for composite control (solid) and composite perturbation (dashed) integrations. (c) Composite anomaly zonal moisture fluxes across the meridian shown in (a), vertically integrated from 1000 to 700 hPa at days 7–13.

Citation: Journal of the Atmospheric Sciences 54, 22; 10.1175/1520-0469(1997)054<2597:BITAMT>2.0.CO;2

Fig. 11.
 Fig. 11.

(a) GCM composite control run precipitation averaged over days 6 to 13. (b) Composite anomaly precipitation averaged over days 6 to 13. (c) Time series of anomaly precipitation averaged in the two boxes shown in (b) (dotted), together with 3-day running means (solid).

Citation: Journal of the Atmospheric Sciences 54, 22; 10.1175/1520-0469(1997)054<2597:BITAMT>2.0.CO;2

Save