Connections between the summer east African and Indian rainfall regimes



[1] Rainfall over parts of east Africa and India is observed to be positively correlated and associated with variations in the low-level height field over the northern Indian Ocean and the Arabian Peninsula. A regional climate model is used to understand the mechanisms of this covariability. SST anomalies in the Arabian Sea are imposed to generate anomalies to analyze. Variations in the monsoon trough, which is located over east Africa extending eastward to southeastern Asia, regulate the connection between east Africa and India and mediate the precipitation covariability in the model. When the monsoon trough is weak, northern Ethiopia and western India are dry, in association with a weaker Somali jet, but rainfall is enhanced over southern Ethiopia. Over east Africa, rainfall decreases over northern Ethiopia are related to increased dry air advection from the north. The atmospheric response when the monsoon trough is strong is not the opposite of the weak trough simulation as conditions are wetter over West India, but drier over all of Ethiopia. Hydrodynamical processes including latent heating are more influential in the vicinity of the western Indian Ocean when the monsoon is perturbed by warm Arabian Sea SSTAs.

1. Introduction

[2] Several observational studies reveal an association between boreal summer rainfall over east Africa and India. Walker [1910], Bhatt [1989], and Whetton and Rutherfurd [1994] found that when summer rainfall over India is high, flooding of the Nile is likely. Camberlin [1995] correlated summer monthly and seasonal rainfall indices over east Africa (i.e., Ethiopia, Kenya, and Uganda) derived from 237-station data set with gridded global rainfall amounts from the Hulme [1992] data set, and found a significant (at the 5% level) positive correlation between rainfall over east Africa and precipitation along the west coast of the Indian Peninsula. Camberlin [1997] observed that this relationship also occurs on intraseasonal timescales, with precipitation over India leading east African rainfall by 2–6 days.

[3] Relationships also exist between rainfall over east Africa and India and key atmospheric circulation features. Camberlin [1995, 1997] associated dry (wet) seasons over both east Africa and India with above (below) normal lower tropospheric heights over the Arabian Peninsula and Indian Ocean. Osman and Hastenrath [1969], Kidson [1977], Hulme and Tosdevin [1989], and Camberlin [1995] suggest that the upper tropospheric tropical easterly jet is weaker and located farther south during anomalously dry summers over east Africa.

[4] The purpose of this paper is to investigate the physical processes that lead to these observed correlations. Studies using a hydrostatic primitive equation model by Rodwell and Hoskins [1996, 2001] provide some insight about a particular atmospheric mechanism that may be relevant. They examined the remote effects of the diabatic heating associated with the Asian monsoon on the large-scale circulation, and found that the atmospheric response to the diabatic forcing is similar to the so-called Matsuno-Gill response [Matsuno, 1966; Gill, 1980], in which an eastward-propagating Kelvin wave occurs to the east of the heating region, and a westward-propagating Rossby wave to the west [see Rodwell and Hoskins, 2001, Figure 9]. Many of the region's lower tropospheric circulation features, including the Somali jet, cross-equatorial flow over the east coast of Africa, northerly flow over northeastern Africa, westerly flow over Central Africa, and subsidence over the Sahara can be associated with this classical response to Asian monsoon heating. Of particular interest is an elongated region of low heights, a monsoon trough, positioned over western India and extending westward to east Africa in the lower troposphere. Heating over east Africa deepens this trough, and our study suggests that its magnitude and positioning are key for understanding interactions between east African and Indian climate.

[5] We use a regional climate model to investigate the connections between east Africa and India. The model produces a good simulation of the regional climate, including the monsoon trough, so we are able to force perturbations of the trough and analyze the hydrodynamical response over India and east Africa. In the model and the real world, many forcings can induce variability in the monsoon trough, including sea surface temperature anomalies (SSTAs) (e.g., El Nino SSTAs in the Pacific [Palmer, 1986; Folland et al., 1986; Webster and Yang, 1992; Ju and Slingo, 1995; Webster et al., 1998; Slingo and Annamalai, 2000; Rowell, 2001] and Indian Ocean SSTAs [Yamazaki, 1988; Chandrasekar and Kitoh, 1998; Meehl and Arblaster, 2002], and changes in land surface conditions over India and Africa [e.g., soil moisture; Yang and Lau, 1998; Douville et al., 2001], and/or snow cover over the Himalayas [Barnett et al., 1989; Douville and Royer, 1996; Sankar-Rao et al., 1996; Bamzai and Shukla, 1999]). Here we choose to use Arabian Sea SSTAs because observational studies [Shukla, 1975; Weare, 1979; Alapaty et al., 1995; Webster et al., 1998] document a relationship between these SSTAs and the low-level height field in this region, and these provide validation opportunities for our modeling study. But this methodology should not be interpreted as a statement that Arabian Sea SSTAs act on the monsoon system as an immutable external forcing in the real world. In fact, summertime Indian Ocean SSTAs may be as much a “result” of low-level wind anomalies as they are a cause of the circulation structure [Webster et al., 1998, 1999; Clark et al., 2000; Camberlin et al., 2001]. We do not investigate the time-dependent adjustment, and we are not making statements about cause and effect between SSTAs and circulation anomalies. Rather, we examine the climatological hydrodynamics to understand the connections between east Africa and India, and to shed some light on their covariability.

[6] An overview of the boreal summer monsoon trough is presented in section 2. Section 3 discusses the regional climate model and the simulation design, while section 4 compares the model control simulation results with observations and the NCEP/NCAR reanalysis. Results from the model simulations are reported in section 5, including a simulation in which the monsoon trough is weak and one in which the trough is strong. In section 6 we look for the mechanisms found in the model in the reanalysis. Conclusions are provided in section 7.

2. The Boreal Summer Monsoon Trough

[7] The primary cause of the summer monsoon trough over India and Africa is differential heating of land and ocean. During boreal summer, solar insolation is at its northernmost position and the land surface over the Sahara, Arabia, India, and the Himalayas warms more than the Gulf of Guinea and Arabian Sea. The result is strong cross-equatorial height gradients between Asia and the southern Indian Ocean [Yanai and Li, 1994; Webster et al., 1998].

[8] Figure 1a shows the summer 850 hPa geopotential heights from the 1949–2001 NCEP/NCAR reanalysis climatology. The monsoon trough, denoted by the bold dashed line, extends westward from India and over the Arabian Peninsula to the west coast of Africa, with the lowest heights located over northern India and southern Pakistan between 20°N and 30°N. Heights are not as low over Africa, but the trough is still well defined between 10°N and 20°N. Note the strong negative meridional height gradient to the south of the trough axis over the northern Indian Ocean and India.

Figure 1.

June–September geopotential heights and winds at (a) 850 hPa and (b) 200 hPa from the 1949–2001 NCEP reanalysis climatology. Height contours are in units of meters and the wind vector scale is 5 m s−1. The bold dashed line represents the monsoonal trough axis from Africa to India.

[9] The 850 hPa wind from the NCEP reanalysis climatology is also shown in Figure 1a. The flow is geostrophic to first order near and to the north of the trough, since the trough stays north of 10°N latitude. North (south) of the trough axis low-level flow is northerly or northeasterly (westerly or southwesterly), extending from West Africa to the Indian Peninsula.

[10] The southwesterly flow over the Arabian Sea is the northern branch of the Somali jet, and is associated with the strong meridional height gradient over the Arabian Sea [Findlater, 1969]. Over east Africa, the Somali jet is associated with lower tropospheric divergence in the vicinity of the Horn of Africa, as this region is located at the entrance of the jet. Westerly zonal flow from central Africa, northeasterly flow from the Saharan and Arabian deserts, and southerly cross-equatorial flow from the southern Indian Ocean all converge over Ethiopia and Kenya as the low-level flow accelerates eastward into the Somali Jet. Over India, the jet flows directly over the Indian Peninsula.

[11] Another important low-level flow feature over east Africa is the low-level Turkana jet [Kinuthia and Asnani, 1982; Kinuthia, 1992; Indeje et al., 2001]. Southeasterly flow from the Indian Ocean penetrates over east Africa through a narrow break in topography between the Ethiopian and east African Highlands, transporting moisture past the mountain barrier. This feature is not well resolved in the NCAR/NCEP reanalysis (Figure 1a) since the width of the Turkana Channel is only about 300 km across with elevations rapidly dropping from over 2 km to approximately 0.5 km.

[12] Figure 1b shows geopotential height and wind fields at 200 hPa from the NCEP/NCAR reanalysis summer climatology. The vertical structure of the monsoon trough is baroclinic, with an elongated anticyclone in the upper troposphere is located over the low-level cyclones of southern Asia and eastern Africa. The associated flow is westerly north of 30°N, and easterly between the equator and 20°N. This easterly flow, known as the tropical easterly jet, extends westward from east Asia to Africa and can be thought of as part of the upper tropospheric outflow from the Asian monsoon.

[13] Figure 2 shows summer rainfall rates over east Africa and India for various observational climatologies including the 1979–2001 GPCP satellite-gauge precipitation climatology from Huffman et al. [1997] (Figure 2a), the Climate Prediction Center Merged Analysis of Precipitation (CMAP) climatology from Xie and Arkin [1997] (Figure 2b), the Legates and Willmott [1990] rain gauge climatology (Figure 2c), and the Hulme [1992] 1900–1998 rain gauge climatology (Figure 2d). Modeling studies indicate that the presence of a flat version of the African continent generates a large-scale precipitation maximum over east Africa, and smaller-scale structure is introduced by the topography [Semazzi and Sun, 1997; Cook, 1997]. All of the rainfall climatologies in Figure 2 indicate a precipitation maximum ranging between 6 and 12 mm d−1 over the Ethiopian Highlands, yielding 600–1000 mm of rain for the summer season. Lower troposphere westerly flow from western and central Africa converges with relatively dry northerly flow from the Sahara and moist southeasterly flow from the southern Indian Ocean over the Ethiopian Highlands [Griffiths, 1972]. To the east of the Highlands, precipitation rates are considerably smaller over the coastal plains of Somalia and Kenya due to strong divergence associated with inflow into the Somali jet. These four components are the important moisture sources for determining rainfall over Ethiopia during the boreal summer.

Figure 2.

June–September precipitation for the (a) 1979–2001 GPCP satellite-gauge precipitation climatology [Huffman et al., 1997], (b) 1979–2001 CPC Merged Analysis of Precipitation (CMAP) climatology [Xie and Arkin, 1997], (c) Legates and Willmott [1990] climatology, and (d) Hulme [1992] 1900–1998 climatology. Contours are every 2 mm d−1.

[14] Over India, rainfall is largest along the west coast, as the onshore flow of the Somali jet encounters the Ghat Mountains. The magnitude of this maximum ranges between 10 and 20 mm d−1 in various data sets shown in Figure 2. east of the Ghats, rainfall rates fall to 4–7 mm d−1. Precipitation rates are also lower over northern India, with magnitudes ranging from 5 to 10 mm d−1 among the different climatologies. Another precipitation maximum is found in the vicinity of the Ganges River delta (∼10–20 mm d−1).

[15] The monsoon region between east Africa and India experiences a high degree of variability on various timescales. On the intraseasonal timescale, the monsoon region can experience “breaks” in activity where rainfall is reduced and the circulation associated with the monsoon trough is weakened including the Somali jet [Gadgil and Asha, 1992; Camberlin, 1997; Webster et al., 1998]. On interannual timescales, the monsoon system tends to exhibit a biennial cycle, where wet years tend to be followed by dry years [Webster et al., 1998; Meehl and Arblaster, 2002]. This biennial signal is not limited to just rainfall, as the signal also appears in the pressure, wind, and SST fields over the Indian Ocean basin. The monsoon is also linked with the El Niño-Southern Oscillation [Yamazaki, 1988; Webster et al., 1998; Meehl and Arblaster, 2002], as El-Niño (La Niña) conditions are typically associated with a weak (strong) monsoon system.

3. Model and Simulations

[16] The PSU/NCAR MM5 (v3.4) model [Grell et al., 1993] is used for our seasonal climate study over the tropics. MM5 is a limited area grid point model with finite differences centered in space and time. The model is nonhydrostatic and uses semiexplicit time integration to eliminate sound wave modes. The vertical sigma coordinate system is terrain following, and is defined σ = (ppt)/(pspt), as where p is pressure, ps is surface pressure, and pt is the pressure at the top of the atmosphere. We use 24 σ-levels, with the top of the atmosphere set at 50 hPa, and 120 km grid spacing with a model time step of 2 min. Our domain is quite large (Figure 3) to encompass the regions of interest and to place the lateral boundaries far from these regions.

Figure 3.

One hundred twenty kilometer resolution domain and topography for summer seasonal model simulations. Shading interval is every 500 m.

[17] The model is run in climate mode. Lateral boundary conditions are climatological winds, temperature, and moisture at each vertical level from the NCEP reanalysis [Kalnay et al., 1996] updated at 12-hour intervals. Similar to the method used by Vizy and Cook [2002], monthly means from the NCEP reanalysis are set to represent the middle of the month, and a temporal linear interpolation is used to generate lateral boundary conditions for every 12 hours. These lateral boundary conditions include the influence of a seasonal-scale cycle, but shorter timescales, such as the diurnal cycle, are filtered out. Vizy and Cook [2002] assessed the influence of using such climatological lateral boundary conditions and found that the elimination of these shorter timescale cycles did not influence the results in the interior of their model domain. The domain for our simulations (Figure 3) is larger than the domain used in the previous study.

[18] Ocean surface temperatures and soil moisture availability are also prescribed and updated every 12 hours. The monthly SST climatology from Shea et al. [1992] and the monthly soil moisture climatology from Willmott et al. [1985] are used to generate 12-hourly SSTs and daily values for soil moisture availability for both simulations. Monthly means are taken to represent the middle of the month, and daily values are linearly interpolated from these means.

[19] Extensive testing and validation were used to select the boundary locations and conditions, and physical parameterizations that provide an optimal simulation for this region. The radiation package chosen is the Rapid Radiative Transfer Model (RRTM) longwave scheme [Mlawer et al., 1997], which combines with the cloud-radiation shortwave scheme and the simple ice explicit moisture scheme [Dudhia, 1989] to determine shortwave radiation and cloud ice. RRTM uses a correlated-K model to account for the effects of CO2, O3, and water vapor in determining longwave radiative fluxes. RRTM also interacts with the precipitation and cloud fields in the MM5. The RRTM radiation parameterization was chosen because it produces a much more realistic surface radiation budget in our simulation compared to other radiation parameterizations available, including the CCM2 radiation parameterization. (Previous regional climate simulation studies [Giorgi et al., 1993; Vizy and Cook, 2002; Cook et al., 2003] note that when using the CCM2 radiation parameterization, the model consistently produced high cloud fractions in the lower troposphere, strongly influencing the surface radiation.)

[20] Other parameterizations chosen for our simulations include the Kain-Fritsch cumulus parameterization [Kain and Fritsch, 1993], shallow cumulus parameterization [Grell et al., 1993], and the high resolution Blackadar planetary boundary layer scheme [Blackadar, 1979; Zang and Anthes, 1982].

[21] Three 139-day simulations are utilized: a control, a cold SSTA case, and a warm SSTA case. Each simulation is initialized on 00Z 15 May using climatological (1949–2001) May NCEP reanalysis conditions [Kalnay et al., 1996] and run through the end of September. The first 17 days are discarded for model spin-up. The remaining 3-hourly output is averaged to form a summer (June–September) seasonal climatology.

[22] The cold Arabian Sea SSTA is shown in Figure 4; the warm SSTA is identical except for the sign. The Gaussian-shaped SSTA, with a maximum of ±2 K, is located in the central Arabian Sea and represents a basin-wide cooling (warming) of the Arabian Sea. The magnitude of the anomaly is roughly twice as large as those typically observed on interannual timescales. This guarantees a clear atmospheric signal, including in noisy fields such as precipitation and moisture convergence. Realistically, the largest SSTAs tend to occur off coastal east Africa during the boreal summer, but particular summers do show evidence of basin-wide cooling (e.g., the summer of 1986).

Figure 4.

Position and absolute magnitude of the sea surface temperature anomalies for the model simulations. Units are in Kelvin.

4. Validation of the Model Climatology

[23] Figure 5a shows the June–September rainfall climatology for the control simulation. Rainfall maxima are present along the Western Ghats of India (∼18 mm d−1), over northeastern India and Bangladesh (∼15 mm d−1), over Ethiopia (∼12 mm d−1), in the vicinity of Lake Victoria (∼8 mm d−1), and over the southeastern equatorial Indian Ocean (∼14 mm d−1).

Figure 5.

June–September precipitation rates for the control simulation on the (a) 120-km resolution grid and (b) interpolated to a 2.5° resolution grid. Units are in millimeters per day.

[24] While these precipitation maxima are similar to the observed maxima (Figure 2), rainfall rates are generally larger. Some of the discrepancies could be related to resolution differences. For example, Figure 5b shows the modeled rainfall rates interpolated to the 2.5° × 2.5° grid of the GPCP product (Figure 2a). In the smoothed model precipitation, maximum rainfall rates are significantly lower and more similar to the observations. However, the difference in resolution does not fully explain the discrepancies between the control and the observations. The model tends to overproduce rainfall over the equatorial southwest Indian Ocean, and rainfall rates are much larger than the observations along the southwestern coast of the Arabian Peninsula. The latter discrepancy is associated with a large surface latent heat flux from the Red Sea in model, at least twice as large as the values in the NCEP/NCAR climatology. Rainfall is also underproduced over central Africa, with modeled rainfall rates up to 3 mm d−1 lower than the observations. This discrepancy is associated with a drier than observed southwesterly West African monsoon flow. This model is able to produce a very good simulation of the West African monsoon system [Vizy and Cook, 2002], but we were unable to optimize the parameter choice and boundary conditions to produce a good simulation in both east and West Africa at the same time for this domain.

[25] There are also significant differences in the distribution of rainfall among the different observed rainfall climatologies shown in Figure 2. For example, rainfall rates over the eastern Bay of Bengal are larger and further south in the Legates and Willmott climatology compared to the GPCP and CMAP climatologies. Over east Africa, rates are 2 mm d−1 larger in the Legates and Willmott climatology than in the GPCP, CMAP, or Hulme climatologies.

[26] Figures 6a and 6b show June–September geopotential heights and winds for the control simulation on the model's 120-km resolution grid at 870 hPa and interpolated to a 2.5° × 2.5° grid at 850 hPa, respectively. The model captures the features of the low-level circulation fairly well when compared to the NCEP reanalysis climatology (Figure 1a). One discrepancy is that the monsoon trough and the Somali jet over the Arabian Sea are approximately 3°–5° of latitude further south than in the NCEP reanalysis. The meridional height gradient in our control is stronger further south than in the reanalysis, hence the shift in the trough equatorward. This difference is also associated with a stronger lower tropospheric cyclone and the overproduction of precipitation (Figure 5) over the southwestern tropical Indian Ocean in the model.

Figure 6.

June–September geopotential heights and winds for the control simulation (a) on the 120-km resolution grid at 870 hPa and (b) interpolated to the NCEP's 2.5° resolution grid at 850 hPa. Height contours are in units of meters and wind vectors are in units of meter per second.

[27] The Turkana jet is better resolved in the regional model than in the NCAR/NCEP reanalysis. The Turkana Channel is represented by six grid points in our 120-km resolution domain, which is triple the two grid point coverage in the reanalysis. Southerly and southeasterly flow associated with the cross-equatorial flow converges with westerly flow from central Africa in the vicinity of this valley.

[28] Figure 7a shows the zonal component of the 150 hPa flow for the NCEP/NCAR reanalysis' June–September climatology. Easterly flow prevails in the tropics in this domain, with a maximum greater than 30 m s−1 south of India. The jet weakens to 20–25 m s−1 over east Africa. As in the NCEP reanalysis, the modeled tropical easterly jet (Figure 7b) maximum is positioned over the southern Arabian Sea, but magnitudes are approximately 5 m s−1 larger than the reanalysis. The easterly maximum also extends further westward over Africa and the zonal gradient over western and central Africa is approximately 40% stronger than in the reanalysis. The jet in the model at 120-km resolution (not shown) is similar to the jet in the model interpolated to a 2.5° × 2.5° grid shown in Figure 7b.

Figure 7.

June–September 150 hPa zonal wind component for the (a) 1949–2001 NCEP reanalysis climatology and (b) the MM5 control simulation interpolated to the NCEP's 2.5° resolution grid. Winds units are in meter per second. Negative values are shaded and denote easterly flow.

[29] The model simulates the east African and Indian boreal summer monsoon systems with sufficient accuracy to justify further analysis. Under the range of uncertainty in the observations themselves (Figure 2), rainfall rates are reasonable over much of the domain. At the very least, the results from the model are an improvement over lower resolution GCM simulation results. When compared to various GCM precipitation climatologies [Gadgil and Sajani, 1998, Figures 2b and 9], the regional climate model presented here provides an improved simulation of the summer monsoons. It does particularly well over India, a region where most GCMs have difficulty in correctly capturing the summer monsoon rainfall. Jha et al. [2000] found that this poor performance is closely tied to the relatively coarse resolution used by GCMs, and that the use of higher horizontal resolution improves the monthly rainfall rates in the GCMs. In addition to having higher horizontal resolution than is typical in GCMs, the regional model also allows one to select the set of physical parameterizations (discussed in section 3) that produce the best simulation of precipitation over east Africa and India, instead of having to consider the entire globe in optimizing as in a GCM.

[30] Characteristics of the large-scale summer circulation of the east African and Indian monsoons are also represented in the control simulation, suggesting that the model is useful for diagnosing the mechanisms responsible for the connection in precipitation between east Africa and India.

5. Results

5.1. Weak Monsoon Trough Simulation

[31] The monsoon trough weakens between east Africa and India when the model is forced by cold Arabian Sea SSTAs as the lower troposphere cools in response to a decrease in both sensible and latent heating from the ocean surface. Evaporation rates decrease by approximately 3 mm d−1 (∼40%), reducing the lower tropospheric moisture content of the atmosphere by about 15%. In the central and eastern Arabian Sea, the sensible heating flux from the ocean to the atmosphere also decreases, while in the upwelling regions of the western Arabian Sea, sensible cooling of the ocean surface increases. The net result is that the lower atmosphere cools over the Arabian Sea by approximately 1.5 K, decreasing the meridional height gradients over the Arabian Sea by approximately 25%.

[32] Figure 8a shows the 870 hPa wind and height field for the cold SSTA simulation. The monsoon trough (Figure 6a) is still present, extending from the Bay of Bengal westward to West Africa. Unlike the control, however, the lowest geopotential heights are positioned over the northeastern coast of India and the northeastern Bay of Bengal. Heights are also lower along the coast of the Arabian Peninsula and over northeastern Africa. As in the control simulation, the lower tropospheric flow is approximately geostrophic everywhere except over western and central Africa.

Figure 8.

June–September 870 hPa (a) geopotential heights and winds for the weak monsoon trough simulation and (b) geopotential height and wind differences between the weak monsoon trough simulation and the control. Heights are in meters and winds are in meter per second. Shading denotes negative values.

[33] Figure 8b shows the lower tropospheric wind and height differences between the cold SSTA simulation (Figure 8a) and the control (Figure 6b). The imposition of cold Arabian Sea SSTAs leads to positive height anomalies to the north and west of the cooled region, and just south of the equator in the western Indian Ocean. The magnitudes of the height anomalies over the Arabian Sea and Arabian Peninsula are much larger than in the southwestern Indian Ocean. Over the Bay of Bengal and southeastern Indian Ocean, height anomalies are negative. Again, the magnitudes of the height anomalies are larger in the Northern Hemisphere.

[34] The circulation and height anomalies for the simulation with cold Arabian Sea SSTAs show some similarities to the theoretical results from Matsuno [1966] and Gill [1980]. They resemble a Rossby wave response to the surface forcing [Matsuno, 1966, Figure 4c]. Over the western Indian Ocean (i.e., west of 70°E) the anomalous anticyclones straddling the equator are associated with anomalous westerly flow along the equator west of 70°E, anomalous northerly flow over western India, and anomalous southerly flow over northeastern Africa. East of 70°E, anomalous lower tropospheric flow along the equator is easterly, similar to a stationary Kelvin wave response to cold SSTA forcing. Additionally, weak twin cyclonic anomalies straddle the equator.

[35] The positive height anomalies over the Arabian Sea and Arabian Peninsula indicate that the monsoon trough weakens from east Africa to India when SSTAs are cold in the Arabian Sea; the negative height anomalies over eastern India and the Bay of Bengal indicate a strengthening. The Somali jet weakens, as the lower tropospheric meridional height gradient over the Arabian Sea and India is reduced by approximately 25%. The zonal height gradient over India increases due to a stronger monsoon trough over the northern Bay of Bengal and a weaker trough over western India and the Arabian Sea. In association with this stronger zonal gradient, low-level flow becomes northerly over much of northern and central India.

[36] In association with the weaker monsoon trough and meridional height gradient, westerly flow decreases by approximately 10% from central Africa to the Horn of Africa. Additionally, the zonal height gradient over northeastern Africa and the Arabian Peninsula weakens in association with the weaker trough and northerly flow over Egypt, Libya, Sudan, and Saudi Arabia decreases by about 25%. Cross-equatorial flow along the east African coast also decreases by 5–10%, as positive height anomalies over the southwestern Indian Ocean are associated with weaker zonal height gradients along the coast.

[37] In the upper troposphere, heights are 10–40 m lower than in the control over southwestern Asia. Near the equator, the tropical easterly jet east (west) of 60°E is approximately 5–8 m s−1 stronger (weaker) than in the control. The change in the sign of the anomaly at 60°E in the upper troposphere corresponds roughly to the low-level sign changes.

[38] The rainfall response is shown in Figure 9. Rainfall decreases along the west coast of India by up to 9 mm d−1, over northern and northeastern India by 3–11 mm d−1, and over northwestern Ethiopia and the southern Arabian Peninsula by approximately 3–10 mm d−1. Rainfall increases over southern Ethiopia and along the Malay Peninsula by up to 4 mm d−1.

Figure 9.

June–September weak monsoon trough simulation minus control rainfall differences. Units are in millimeters per day. Shading denotes negative values.

[39] The correlations between Ethiopian and Indian rainfall in our model results for the weak monsoon trough case do not agree precisely with Camberlin's [1995] observational analysis. Calculating the same summer (June–September) rainfall indices that Camberlin [1995, 1997] used reveals a positive (negative) relationship between rainfall over northwestern (southwestern) Ethiopia and western India for the weak monsoon simulation. Understanding why our relationship does not agree with the limited observational evidence is difficult because of the spatial and temporal differences between their statistical-based studies and our regional climate-mode modeling study. However, it is possible to understand why rainfall in our simulation covaries the way it does over east Africa and India when the SSTAs are cold.

[40] In the weak monsoon (cold SST) case, the reduced westerly flow on the southern flank of the monsoon trough (Figure 8) is associated with decreased low-level wind convergence, moisture convergence, and rainfall along the western coast of India of up to 60%. Further east, the reduced westerly low-level flow into central India is accompanied by a reversal in direction (from southerly to northerly) of the meridional flow, reducing the wind convergence, moisture convergence, and rainfall over the interior by up to 80%.

[41] The decrease in rainfall over northwestern Ethiopia (near 14°N, 38°E) results from changes in all four of the components identified in section 2 as being moisture sources for this region. Westerly inflow weakens, but westerly outflow into the Somali outflow into the Somali jet decreases by approximately 2 m s−1 more, so there is a net decrease (increase) in zonal wind divergence (convergence) on the eastern (western) slopes of the northern Ethiopian Highlands. Northerly flow from northeastern Africa also decreases when the zonal height gradient weakens, resulting in a decrease in meridional wind convergence over northern Ethiopia. The net result over northwestern Ethiopia is that the lower tropospheric wind convergence increases modestly, causing a 2 mm d−1 increase in moisture convergence. However, this increase in moisture convergence is offset by a 4 mm d−1 increase in dry air advection. The localized low-level flow over northern Ethiopia and over the Red Sea basin becomes less westerly and more northerly when the monsoon trough is weaker due to weaker inflow into the Somali jet, resulting in an increase in advection of drier air from the Arabian and Saharan deserts into northwestern Ethiopia and a precipitation decrease.

[42] Rainfall increases over southwestern Ethiopia when the monsoon trough is weaker (Figure 9). This increase is primarily associated with an increase in wind and moisture convergence (∼3 mm d−1) over the southern slopes of the Ethiopian Highlands. Lower tropospheric westerly flow over southern Ethiopia and the Somali jet over the Arabian Sea weaken with the decrease in the meridional height gradient. The cross-equatorial flow penetrates further west, into northern Kenya and southern Ethiopia through the Turkana Channel, instead of being incorporated into the westerly Somali jet. This model result suggests that there is a negative relationship between the strength of the southeasterly flow entering the Turkana Channel and the strength of the westerly Somali jet. Moisture laden air transported via this cross-equatorial flow converges over the southern slopes of the Ethiopian Highlands, increasing the moisture convergence and rainfall rates by over 3 mm d−1 over southern Ethiopia.

5.2. Strong Monsoon Trough Simulation

[43] The 870 hPa wind and height fields for the strong monsoon simulation and their differences from the control simulation are shown in Figures 10a and 10b, respectively. With a warm Arabian Sea, the monsoon trough between east Africa and India deepens by up to 35 m and the Somali jet becomes stronger. Over the southwestern equatorial Indian Ocean, low-level heights decrease by approximately 6 m, about half the magnitude of the weak monsoon simulation response (Figure 8b). Unlike the weak monsoon trough simulation, anomalous flow is predominantly southerly over the western equatorial basin, indicating a stronger Somali jet over the Arabian Sea.

Figure 10.

June–September 870 hPa (a) geopotential heights and winds for the strong monsoon trough simulation and (b) geopotential height and wind differences between the strong monsoon trough simulation and the control. Heights are in meters and winds are in meter per second. Shading denotes negative values.

[44] When warm SSTAs are imposed in the Arabian Sea, the lower troposphere warms and the monsoon trough deepens. However, the distribution of temperature change (not shown) is not the same as the weak monsoon trough (cold SSTA) case. Between 870 and 600 hPa, the magnitude of the warming is about twice as large as the cooling in the weak monsoon trough case. This asymmetry has two causes. One is the nonlinearity of the Claussius-Clapeyron equation, which causes the response in the evaporation rates to be larger in the warm SST case than in the cold SST case, but the latent heat release is in the middle troposphere and not near the surface. Another reason is a positive feedback between the low-level wind convergence and precipitation fields, which enhances the low-level wind and evaporation rates.

[45] There are also structural differences in the weak and strong trough cases. Below 870 hPa, the largest temperature anomalies (up to 2.5 K) are located over the western Arabian Sea, whereas the largest anomalies in the weak monsoon case (∼−2 K) are found over the eastern Arabian Sea.

[46] Warming from the surface is delivered to the lower atmosphere by sensible heating. Nonlinearities in the Blackadar boundary layer surface sensible heat flux parameterization [Blackadar, 1979; Zang and Anthes, 1982; Grell et al., 1993] cause the response of the sensible heating to be larger when the low-level flow is enhanced and the roughness length is larger. Thus the magnitude of the lower atmosphere heating is greater in the strong monsoon trough case than the weak trough case. Frictional velocities increase by about 40% over the Arabian Sea in the warm case, driving an enhancement of the sensible heat flux to the atmosphere of 25 W m−2. In the cold case, frictional velocity decreases by 20%, so the sensible heat flux decreases by only 10 W m−2 and the overall change in the low-level temperature is less.

[47] The Gill/Matsuno response is difficult to identify in Figure 10b. Anomalous easterly flow along the equator east of 70°E may be a stationary Kelvin wave response to the anomalous heating, while the anomalous cyclonic circulation may be identifiable with that of a Rossby wave response. However, in the weak trough (cold SSTA) simulation, the zonal response is not clear. This may be because the response is weaker so the Gill/Matsuno response is not cleanly expressed, or because moisture processes play a more dominant role in the warm SSTA case.

[48] The asymmetry between the strong monsoon trough (warm SSTAs) and weak monsoon trough (cold SSTAs) cases persists into the upper troposphere. While the sign of the response in the strength of the Tropical easterly jet is the opposite of the weak monsoon simulation, the magnitudes of the anomalies are roughly 2 m s−1 smaller over equatorial Africa and the Indian Ocean.

[49] The anomalous rainfall for the strong monsoon trough simulation is shown in Figure 11. Precipitation rates increase over much of the Arabian Sea and along the western coast of India by 4–8 mm d−1. These are similar in structure and magnitude to the weak trough case (Figure 9), except in the central Arabian Sea where they are up to 4 mm d−1 larger. Elsewhere in the model domain, the precipitation response is not the opposite of the weak monsoon case. Over the Arabian Peninsula, the strong and weak monsoon cases have the same sign, and the north/south dipole structure seen over Ethiopia in the strong monsoon case is replaced by drying in both the north and the south in the weak monsoon case. Over India, a north/south dipole response replaces the fairly uniform response in Figure 9.

Figure 11.

June–September strong monsoon trough simulation minus control rainfall differences. Units are in millimeters per day. Shading denotes negative values.

[50] Differences in rainfall anomalies between the strong and the weak monsoon trough simulations are consistent with differences in the large-scale flow. Over central India (between 15°N and 20°N), the zonal height gradient is enhanced in the strong monsoon trough case, and is accompanied by an increase in southerly low-level flow. Moisture convergence and rainfall are reduced over this region, as some of the westerly onshore flow associated with the Somali jet instead is diverted approximately 5° of latitude to the north (Figure 10), leading to an enhancement of moisture convergence and rainfall over northern India by up to 5 mm d−1. The opposite atmospheric response occurs in the weak monsoon trough simulation (Figure 8), but the enhancement of rainfall over central India is not nearly as strong due to the overall reduction in the low-level flow and moisture content for this case.

[51] The rainfall decreases over the southwestern Arabian Peninsula and northern Ethiopia both in the strong and weak trough monsoon simulations have different causes. For the strong trough simulation, westerly flow into the Somali jet is approximately 3–6 m s−1 stronger over the Arabian Peninsula and east Africa. Wind convergence decreases by approximately 30% over the Ethiopian Highlands and over the southwestern Arabian Peninsula, as low-level flow is accelerated eastward toward India and incorporated into the stronger Somali jet. The increase in westerly flow over east Africa for this simulation is so dominant that even an increase in dry, northerly flow from the eastern Mediterranean is unable to penetrate southward over the Ethiopian Highlands. Instead, this northerly flow is incorporated into the westerly inflow over northern Sudan and Ethiopia and the meridional wind convergence and rainfall decrease over northern Ethiopia and the southwestern Arabian Peninsula. As previously discussed for the weak monsoon trough simulation, dry air advection from the north and east is primarily responsible for the decrease in rainfall over northern Ethiopia and the southwestern Arabian Peninsula, suggesting that the low-level zonal flow is most influential in influencing rainfall over northern Ethiopia when the monsoon trough is strong, while the meridional flow becomes more influential when the trough is weak and the zonal flow decreases in strength.

[52] The rainfall response over southwestern Ethiopia is the opposite of the weak monsoon trough case and is associated with a decrease in low-level wind convergence. Southeasterly flow over east Africa is weaker over northern Kenya and southern Ethiopia, including in the Turkana Channel, when the westerly Somali jet is strong.

[53] The differences in the atmospheric responses for the weak and strong monsoon trough simulations indicate the importance of nonlinearities within the model physics and, presumably, the climate system, in determining the response. The results from the weak trough simulation are not the exact opposite of the strong trough simulation. The perturbation of the trough is larger when the SSTAs are warm, but some of the remote responses are weaker. For example, the magnitude of the low-level height anomaly is about 5 times larger over the Bay of Bengal and the southeastern Indian Ocean in the weak monsoon trough case. There are other regions over the Indian Ocean where the magnitudes of the anomalous response are larger than in the opposite case, corresponding to regions of enhanced latent heating from the ocean. The well-known tropical connection between the latent heating and wind fields emerges in all cases, but the mechanisms that perturb the latent heating field can be different between the warm and the cold cases.

6. Comparison With the NCEP Reanalysis and Observations

[54] After developing an understanding of the central role of the summer monsoon trough in the covariation of rainfall over east Africa and India in the model, it is useful to relate the model results back to the observations. Many factors conspire to determine any year's rainfall distribution, and it is often difficult to isolate individual mechanisms. However, having used model simulations to identify a physical connection between the two regions (i.e., variations of the monsoon trough) that links the east African and Indian precipitation fields, summers from 1979 to 2001 (the GPCP and CMAP precipitation record period shown in Figure 2) in the NCEP/NCAR reanalysis were examined to find examples at work. There were two especially weak monsoon trough summers (1982 and 1986); however, there was not an especially strong trough summer similar to our model results during this time period. The closest example is the summer of 1988.

[55] In our simulations, we control the strength of the monsoon trough by manipulating Arabian Sea SSTs. The weak trough in the reanalysis in 1986 is accompanied by SSTAs similar to those of our weak monsoon trough simulation (Figure 4), but in 1982 SSTAs were up to 1 K warmer than the climatology. The SSTAs that accompany the strong trough are positive (up to 0.75 K) in the Arabian Sea, the Bay of Bengal (∼0.25 K), and the tropical Indian Ocean (∼0.30 K). Cold SSTAs up to 0.5 K are located just off the east African coast.

[56] The NCEP reanalysis 850 hPa wind and height anomalies for the summers of 1982 and 1986 are shown in Figures 12a and 12b, respectively, in which lower tropospheric heights between east Africa and India are 4–12 m higher than the 1979–2001 climatological average. As in the weak monsoon trough simulation, the zonal and meridional low-level height gradients over northeastern Africa and India, the Somali jet, and the Tropical easterly jet are all weaker than normal. These circulation anomalies are associated with reduced wind convergence over India and northern Ethiopia.

Figure 12.

NCEP reanalysis June–September 870 hPa wind (m s−1) and height (m) anomalies for the summer of (a) 1982 and (b) 1986. Shading denotes negative values.

[57] The GPCP satellite-gauge rainfall anomalies for the summers of 1982 and 1986 are shown in Figures 13a and 13b, respectively. Similar to our weak monsoon trough simulation (Figure 9), rainfall rates were anomalously low along the west coast of India (∼1–3 mm d−1), over northeastern India (∼1 mm d−1), and over northern Ethiopia (∼1–2 mm d−1), and anomalously high over southern Ethiopia (∼1 mm d−1). Rainfall anomalies in the CMAP [Xie and Arkin, 1997] and Hulme [1992] data sets are similar in structure with some differences in magnitudes.

Figure 13.

June–September GPCP satellite-gauge rainfall anomalies for the summers of (a) 1982 and (b) 1986. Contours are every 0.5 mm d−1 and negative values are shaded.

[58] Figure 14a shows the 850 hPa height and wind anomalies from the reanalysis for the summer of 1988. The lower tropospheric trough deepens by 6–12 gpm between east Africa and India. Similar to our simulated strong monsoon trough simulation (Figure 10b), heights are anomalously low over northern Africa and the tropical Indian Ocean, and slightly higher (∼2 m) over the Bay of Bengal and Southeast Asia. Zonal and meridional height gradients are also stronger over the Arabian Sea and India during the summer of 1988, respectively, and are associated with a stronger Somali jet, and increased moisture convergence over western India.

Figure 14.

Summer (June–September) of 1988 (a) 870 hPa wind (m s−1) and height (m) anomalies and (b) precipitation anomalies (mm d−1) from GPCP satellite-gauge precipitation data set. Shading denotes negative values.

[59] The GPCP satellite-gauge rainfall anomalies (Figure 14b) show that the summer of 1988 is wet along the west coast of India (∼3 mm d−1), over southern and northwestern India (∼1–2 mm d−1), and over all of Ethiopia (0.5–1.5 mm d−1). The rainfall response is similar to our strong monsoon trough simulation (Figure 11) over India. Drier conditions in our simulation are positioned further north and east during the summer of 1988. Wet conditions along the coastal regions of the Arabian Sea are also in good agreement with the model results; however, the magnitudes are smaller in the GPCP precipitation estimates.

[60] Over Ethiopia the rainfall anomaly is opposite to the modeled response. Wind and moisture convergence over Ethiopia during this summer are much larger than in our simulation, hence the increase in rainfall over the region. One reason for this discrepancy may be related to the fact that the perturbation of the monsoon trough in our simulation is 3 times as large as the perturbation for the summer of 1988 and centered about 8°–10° of latitude further north. The stronger height gradient and Somali jet over the Arabian Sea are associated with a reduction in wind and moisture convergence over Ethiopia in our strong monsoon trough simulation, but this is not evident in 1988.

[61] This difference may also be associated with distribution of the SSTAs in the Indian Ocean. For example, studies by Yamazaki [1988], Chandrasekar and Kitoh [1998], and Meehl and Arblaster [2002] show that warm SSTAs in the equatorial Indian Ocean are associated with a reduction in rainfall over the southern Asian land regions and a localized increase in precipitation over the tropical Indian Ocean. Meehl and Arblaster tested the sensitivity further by warming the SSTs of the entire tropical Indian Ocean, and found that rainfall rates for this particular case are enhanced over both the Indian Ocean and southern Asia.

[62] Another possible reason for the discrepancy is the presence of other forcing mechanisms that influence the large-scale circulation. For example, SSTs in the tropical Atlantic SSTAs were anomalously warm during the summer of 1988. SSTAs in this region are influential in perturbing the West African monsoon [Vizy and Cook, 2001, 2002]. While the Somali jet is stronger in 1988, westerly flow from West Africa is enhanced more, increasing the moisture convergence and rainfall over Ethiopia. It should also be mentioned that our model simulations in their current configuration have a dry bias when simulating the West African monsoon, which could further impact rainfall over Ethiopia by reducing the amount of moisture available from the west. SSTs in the tropical Pacific were also anomalously cool during this summer as La Niña conditions were developing over the Pacific and may also be influential [Yamazaki, 1988; Ju and Slingo, 1995; Meehl and Arblaster, 2002].

7. Summary and Conclusions

[63] A regional climate model developed from a version of the PSU/NCAR MM5 is used to explore the mechanisms that lead to observed correlations between east African and Indian rainfall during the boreal summer on interannual timescales.

[64] The model is initialized with climatological values taken from the NCAR/NCEP reanalysis. The lateral boundaries are also constrained with the NCAR/NCEP climatology. Monthly mean values of winds, temperature, and specific humidity are fixed on the lateral boundaries for the 15th of each month, and a linear interpolation is used to generate appropriate daily values. Surface features, such as SST and land characteristics, are also prescribed and vary daily. Integrations begin in the middle of May and run through the end of September; the first two weeks are discarded as a spin-up adjustment.

[65] The model domain is unusually large for a regional model, and the resolution is coarse at 120 km grid spacing. The domain stretches from 70°W to 150°E in longitude, and 35°S to 52°N in latitude, in order to cover the areas of interest (east Africa and India), include regions that may be influential (such as West Africa and the Mediterranean), and to keep the model boundaries reasonably far from the area of interest (see Figure 3). The large size of the domain is partially responsible for the success of the integration with climatological lateral boundaries, since these boundary conditions eliminate the propagation of transient disturbances into the model.

[66] The model generates a reasonable simulation of the east African and Indian precipitation and circulation climatologies, offering a significant improvement over GCM simulations (see, for example, the GCM intercomparision study of Gadgil and Sajani [1998]). The combination of large domain with resolution fine enough to resolve features of the east African topography allows an identification of the circulation features that are important for determining the precipitation climatology in this region. Ethiopia provides a good example. The lower tropospheric flow is quite complex. Moist westerly and southwesterly flow from West Africa and the Congo Basin converges with moist southeasterly flow from the southern Indian Ocean and dry northerly flow from northeastern Africa and the Arabian Peninsula over the Ethiopian Highlands. Moisture convergence over the Highlands is further regulated by the amount of outflow from the region into the Somali jet.

[67] Observational studies of the correlations between east African and Indian summer rainfall associate variations of geopotential height over the Arabian Sea with the correlation between Indian and Ethiopian rainfall. Consistent with those findings, we focus on the monsoon trough as a means of communication between the two rainfall systems. The Somali jet, which can be viewed as the southern edge of the monsoon trough, plays a prominent role in connecting east Africa and western India since the jet entrance region is over the former, and the jet exit is over the latter. In the entrance region, strong westerly and southwesterly inflow into the jet limits the amount of lower tropospheric convergence over east Africa and the southern Arabian Peninsula. The westerly flow transports moisture eastward toward India where the jet structure breaks down over the topography along the west coast.

[68] SSTAs are introduced into the Arabian Sea in the model as a way of modifying the strength of the trough and representing interannual variability within the system. Localized cold and warm SSTAs in the Arabian Sea are used to generate cases of a strong monsoon trough and a weak monsoon trough, respectively.

[69] Reduced lower tropospheric cyclonic flow in the weak trough case weakens the Somali jet over the Arabian Sea, northerly flow over northeastern Africa, and southerly flow over India. Rainfall decreases over much of India in association with reduced low-level wind convergence in the west, and enhanced dry air advection over central India.

[70] Over east Africa, wind convergence is increased over all of Ethiopia in association with reduced inflow into the Somali jet. However, dry air advection over northern Ethiopia increases even more in the weak trough simulation, resulting in a net decrease in rainfall. Over southern Ethiopia, southeasterly flow from the Indian Ocean, which is normally accelerated eastward into the Somali jet, instead converges with westerly flow over the southern Ethiopian Highlands and in the vicinity of the Turkana Valley. This increase in moisture convergence is associated with an increase in rainfall over this region.

[71] The importance of Turkana Channel and the associated low-level Turkana jet to the large-scale summer climate over southern Ethiopia and the Turkana Channel needs to be further explored with higher resolution simulations. While our model provides an improved simulation over the NCAR/NCEP reanalysis, it is still not sufficient for making firm conclusions about the inland penetration of the southeasterly flow and pinpointing the location of the convergence between the southeasterly flow and the westerly flow from Central Africa. In our 120-km simulations southeasterly flow converges with strong westerly flow in the Turkana Channel, but in a 30-km simulation run focused over east Africa using the same model, the flow is southeasterly throughout the entire Turkana Channel and converges with westerly flow on the western side of the Highlands. Indeje et al. [2001] produced a realistic southeasterly jet through this channel using a 60-km regional climate model during the October to December period. They concluded that the Turkana jet is an important factor in the climate of east Africa during this transitional period, and suggest that a similar study is necessary during the boreal summer. Additionally, our results indicate a negative relationship between the strengths of the Somali and Turkana jets. Whether this result would occur in a higher resolution simulation still needs to be investigated.

[72] When the monsoon trough is strong, the Somali jet over the Arabian Sea, northerly flow over northeastern Africa, and southerly flow over India are enhanced as well. Anomalously strong rainfall over western India is supported by moist southwesterly flow from the Arabian Sea. The increase in southwesterly flow is also associated with an increase in low-level wind convergence and rainfall over northwestern India. Over Central India, the wind convergence and rainfall decrease. The low-level flow becomes less convergent as more of the westerly flow associated with the Somali jet is redirected northward over northern and northwestern India.

[73] Rainfall over all of Ethiopia is reduced in the strong monsoon case, primarily in association with enhanced inflow into the Somali jet. Northerly flow from northeastern Africa and southerly cross-equatorial flow are accelerated eastward into the westerly Somali jet instead of converging with westerly flow over Ethiopia.

[74] Our results are not wholly consistent with the few observational studies that find positive correlations between Ethiopian and Indian rainfall [e.g., Camberlin, 1995, 1997]. A positive correlation between rainfall over northwestern Ethiopia and western India occurs in our simulation when the trough is weak, but the correlation between southwestern Ethiopian and western Indian rainfall is negative. When the trough is weak, rainfall is enhanced over much of Ethiopia (except for a strong precipitation reduction in the north), but reduced over all of India. It is difficult to ascertain why the model results differ from these statistical analyses of the observations because of the different spatial and temporal scales involved in each. For example, Camberlin [1995] used a measure of all-India rainfall, while we find that India does not necessarily behave as a unit, especially in the weak trough case. Also, northern Ethiopia seems to be more sensitive to the trough variations in our model than central and southern Ethiopia.

[75] The NCEP reanalysis and the GPCP, CMAP, and Hulme precipitation data sets were surveyed to find examples of strong and weak trough years to compare with the model results. Anomalously weak troughs occurred during the summers of 1982 and 1986, and rainfall rates are anomalously low over India and northern Ethiopia and high over southwestern Ethiopia during these years. The Somali jet over the Arabian Sea, northerly flow over northeastern Africa, and southwesterly flow over India are weaker than normal and accompanied by a decrease in lower tropospheric wind convergence over India and northern Ethiopia, and stronger dry air advection over northern Ethiopia. This precipitation and circulation differences are similar to the modeled weak monsoon case.

[76] It was difficult to find a good example of a strong monsoon trough summer in the reanalysis. The best example identified was the summer of 1988, which featured relatively high precipitation rates over India and all of Ethiopia. The Somali jet over the Arabian Sea and southwesterly flow over India are stronger than normal in the reanalysis, and associated with enhanced lower tropospheric wind and moisture convergence over India and Ethiopia. These anomalies did not agree with the modeled strong trough simulation, which may or may not be relevant since many causes of variability operate in any given year.

[77] This study contributes to the development of a dynamical framework for better understanding covariations of east African and India rainfall, but much more work in this area is needed. We have concentrated on the role of variations in the monsoon trough in establishing this covariability, but the complexity of the hydrodynamics in both of these regions, and discrepancies between our results and observational studies, suggests that there is more to the story. Additional modeling and diagnostic studies, and more observational studies that examine precipitation fields on more regional scales, are needed to fully understand the marked interannual variability of these two important rainfall systems.


[78] This research was supported by NASA grant NA16GP1621 and NSF grant ATM-9815419 and represents a portion of EKV's Ph.D. requirements at Cornell University. The authors would also like to thank the comments and suggestions made by the anonymous reviewers.