Abstract

In this study the simulation of a severe rainfall episode over Mumbai on 26 July 2005 has been attempted with two different mesoscale models. The numerical models used in this study are the Brazilian Regional Atmospheric Modeling System (BRAMS) developed originally by Colorado State University and the Advanced Research Weather Research Forecast (WRF-ARW) Model, version 2.0.1, developed at the National Center for Atmospheric Research. The simulations carried out in this study use the Grell–Devenyi Ensemble cumulus parameterization scheme. Apart from using climatological sea surface temperature (SST) for the control simulations, the impact of the Tropical Rainfall Measuring Mission (TRMM) Microwave Imager (TMI) SST on the simulation of rainfall is evaluated using these two models. The performances of the models are compared by examining the predicted parameters like upper- and lower-level circulations, moisture, temperature, and rainfall. The strength of convective instability is also derived by calculating the convective available potential energy. The intensity of maximum rainfall around Mumbai is significantly improved with TMI SST as the surface boundary condition in both the models. The large-scale circulation features, moisture, and temperature are compared with those in the National Centers for Environmental Prediction analyses. The rainfall prediction is assessed quantitatively by comparing the simulated rainfall with the rainfall from TRMM products and the observed station values reported in Indian Daily Weather Reports from the India Meteorological Department.

1. Introduction

Severe weather systems generally occur with strong gust winds and heavy precipitation. The numerical prediction of such events remains one of the most challenging problems in the field of meteorology. Most of the global models generally underestimate the total rainfall of any heavy precipitating event and commit errors in the timing and location of the event. For better prediction of flash floods it is necessary to understand the dynamics and physics of isolated heavy precipitation and other dynamical features associated with thunderstorms, tornados, and so on. Heavy rainfall occurs frequently around Mumbai during the summer monsoon season. However, the rainfall over the northern parts of Mumbai on 26 July 2005 was extremely heavy. On that day within a span of a few hours, northern Mumbai received unprecedented rainfall, with Santa Cruz recording 94.4 cm of rainfall for the day and more heavy rainfall of 104.9 cm at Vihar Lake located around 15 km northeast of Santa Cruz. Bhandup located southeast of Vihar Lake received 81.5 cm of rainfall. Colaba in southern Mumbai, on the other hand, received only 7 cm. This event disrupted life, besides causing heavy damage to property and human life. Prior knowledge of such an extreme event by a day or a few hours could have minimized the loss of life. Although some diagnostic studies of this extreme rainfall event (Jenamani et al. 2006; Shyamala and Bhadram 2006) and some modeling efforts (Bohra et al. 2006; Chang et al. 2007, manuscript submitted to Global Planet. Change; Vaidya and Kulkarni 2007) specifically for this event have been reported in a hindcast mode, none of the operational models had given an accurate real-time prediction.

With the advancement in numerical techniques for the assimilation of satellite-based observations, significant improvements have occurred in the accuracy and the reliability of numerical weather prediction. However, considerable effort is still needed to allow the prediction of extreme weather events like the Orissa Supercyclone (29 October 1999) or the extreme rainfall event of 26 July 2005 over Mumbai. Though some of the state-of-the-art nonhydrostatic and compressible mesoscale models like the fifth-generation Pennsylvania State University–National Center for Atmospheric Research Mesoscale Model (MM5) have been successfully used for the simulation of extreme events like tropical cyclones (Mandal et al. 2004; Singh et al. 2005), this model was not very successful at the prediction of the Mumbai rainfall event even in a hindcast mode.

Though Kershaw (1988) has shown the beneficial effect of sea surface temperature (SST) anomaly, as obtained from satellite sources on a prediction of the onset of the southwest monsoon over India, the relation between SST and tropical heavy rainfall events has not been studied in adequate detail. Rautenbach (1998) studied the relationship of SST and the frequency of extreme rainfall events over South Africa and found that a warm SST region that had developed over the southeast Atlantic Ocean gave rise to the extreme high rainfall over there. Meneguzzo et al. (2004) analyzed the sensitivity of the meteorological high-resolution numerical simulation of the extreme flood of the Arno River basin, Italy, to different representations of SST and showed that observed SST had a clear positive impact on simulations in the Regional Atmospheric Modeling System (RAMS) as compared to climatological SST. Lebeaupin et al. (2006) have also shown the sensitivity of torrential rain events over the western Mediterranean region to SST. They found, using a very high-resolution Méso-NH research model, that an increase (decrease) of SST by several degrees, on average, intensifies (weakens) the convection and the convection could even be stopped with a large decrease of SST. The potential effect of satellite-derived SST on mesoscale convection and heavy rainfall simulation is relatively unexplored in the tropical region, although some studies have reported the use of microwave satellite measurements for the prediction of intensity of tropical cyclones (Kidder et al. 2000; Wentz et al. 2000). The purpose of the present study is to examine the impact of SST derived from the Tropical Rainfall Measuring Mission (TRMM) Microwave Imager (TMI) on the simulation of the severe rainfall event of 26 July 2005 over Mumbai, using the Brazilian Regional Atmospheric Modeling System (BRAMS) and the Advanced Research Weather Research Forecast (WRF-ARW) modeling system. However, these limited experiments with these two different models definitely give some clues about the performance of NWP models over the tropical Indian region. The WRF Model has been used for the simulation of thunderstorms at Machilipatnam over the east coast of India, for a case of cyclonic circulation over Kerala, India (Vaidya 2007), and for the prediction of warm season rainfall forecasts over the central United States (Gallus and Bresch 2006).

The present study is organized in the following manner: section 2 describes the characteristic features related to Mumbai rainfall; brief descriptions of the models, data, and experiments conducted are given in section 3; and section 4 describes the results and discussion. Conclusions are given in section 5.

2. Characteristic of Mumbai rain

a. Synoptic features of active monsoon

Generally, during the active phase of the southwest monsoon in July and August, the regions to the windward side of the Western Ghats (a north–south mountain range in the western zone of India), like Konkan and Goa (including Mumbai), and coastal Karnataka (Fig. 1) get heavy rainfall because of the orographic effect (Rao 1976). The strong westerly moist wind from the Arabian Sea hits the hills of the Western Ghats and is lifted vertically upward during the active monsoon and causes very heavy rainfall there. During this period the strong westerly/southwesterly flow over the Arabian Sea also leads to the formation of an offshore trough over the sea off the west coast, causing very heavy rainfall activity along the west coast of India, including over Mumbai (George 1956). This westerly/southwesterly flow is also strengthened when the Arabian Sea branch of the monsoon is active or when a depression/low pressure area forms over the north Bay of Bengal and moves into central India. Because of the presence of a midtropospheric cyclone (MTC), the Gujarat-Konkan coast also experiences very heavy rainfall of up to 40 cm day−1. During the onset phase of the southwest monsoon, heavy rainfall of more than 20 cm day−1 is quite common at Mumbai. Generally it is caused by a convergence of the dry winds from the north with the advancing moist southwesterly monsoon winds. Heavy rain is also associated with the development of an onset vortex either over the Arabian Sea or over the Bay of Bengal. During the active phase of monsoon, some special features of the synoptic situation are conducive to the occurrence of very heavy rainfall over Mumbai, which are 1) the development of a low pressure area over the northwest Bay of Bengal, 2) the intensification of the monsoon trough and development of an embedded convective vortex over the central parts of India, 3) the strengthening of the Somali Jet, and 4) the superpositioning of a mesoscale offshore vortex over the northeast Arabian Sea and its northward movement. Thus, the synoptic-scale features are highly favorable at times for the occurrence of heavy to very heavy rainfall over the western coast of India including Mumbai.

Fig. 1.

A map of India showing important landmarks referred to in the paper.

Fig. 1.

A map of India showing important landmarks referred to in the paper.

b. Extreme rainfall event of 26 July 2005: Observed synoptic features

On 26 July 2005, the summer monsoon was active over major parts of the country, and a low pressure area formed over the north Bay of Bengal on 24 July and became a well-marked “low” when it reached inland. Because of this, the monsoon trough moved south of its normal position and a strong cross-equatorial flow developed. As the cross-equatorial flow concentrated at the west and mixed with the low-level jet, the strengthened westerly winds hit the Konkan and Goa coasts resulting in heavy rainfall over these regions. On the next day (25 July) the rainfall band moved north, and large parts of Maharashtra, including Mumbai, received heavy rainfall on 26 July. Jayaram (1965) showed that a diurnal pressure gradient of about 4–8 hPa along the west coast of India between 15° and 20°N is necessary for the occurrence of a heavy rainfall event in the Mumbai belt. This condition is more effective when the pressure gradient is combined with a trough off the Konkan coast or a depression over the Bay of Bengal. Sharma (1965) has shown that during large variations in rainfall over Mumbai, the westerly winds are on the order of 15–20 m s−1 with depth of 3 km. In the present case, the 3-hourly surface chart has also shown the pressure gradient of 4–6 hPa along the west coast between 15° and 20°N from 0000 to 1200 UTC 26 July (Jenamani et al. 2006); westerly winds were also on the order of 15–25 m s−1 with depth of 5.8 km. Thus the synoptic-scale features were highly favorable for the occurrence of heavy rainfall, but the localized nature of the event and its strong intensity could not be anticipated.

On this day Santa Cruz recorded 94.4 cm of rainfall; however, even higher rainfall of 104.9 cm was recorded at Vihar Lake (Jenamani et al. 2006; Shyamala and Bhadram 2006). Although the Met Office (UKMO) model was one of the few models that predicted up to 80 cm of rainfall in retrospective mode over a smaller area covering Mumbai on 26 July, 24 h in advance, most of the numerical models, both global and mesoscale, failed to simulate this catastrophic event in operational as well as in hindcast mode (Bohra et al. 2006). Here an attempt has been made to simulate this event using two mesoscale models. In particular, the impact of TMI SST on the performance of these models is investigated.

3. The model, data, and experimental design

The mesoscale models used in this study are (i) the WRF-ARW Model, version 2.0.1, developed primarily at the National Center for Atmospheric Research (NCAR) in collaboration with different agencies like the National Oceanic and Atmospheric Administration (NOAA), the National Centers for Environmental Prediction (NCEP), NOAA’s Earth System Research Laboratory (NESRL), University of Oklahoma, and many others, and (ii) the BRAMS originally developed by Colorado State University from RAMS, version 5.04, with several additional features to make the model suitable for tropical situations. Table 1 summarizes the salient dynamical and physical features of these models as used in the present study.

Table 1.

A summary of characteristics of the WRF and BRAMS models.

A summary of characteristics of the WRF and BRAMS models.
A summary of characteristics of the WRF and BRAMS models.

The aim of the WRF modeling system is to serve both operational forecasting and atmospheric research (http://www.wrf-model.org). The WRF is a limited-area, nonhydrostatic primitive equation model with multiple options for various physical parameterization schemes (Skamarock et al. 2005). This version employs Arakawa C-grid staggering for horizontal grid and a fully compressible system of equations. A terrain-following sigma coordinate is used in the vertical. The time-split integration uses a third-order Runge–Kutta scheme with a smaller time step for acoustic and gravity wave modes. In combination with multiple-nest capability, a large number of physics options makes the model capable of performing simulations on any scale, limited only by data resolution, quality, and computer resources. Physics options used in this study include the Grell–Devenyi Ensemble (GDE; Grell and Devenyi 2002) cumulus parameterization scheme and the WRF single-moment 6-class Graupel (WSM6) microphysics scheme. The planetary boundary layer (PBL) is parameterized using the advanced version (nonlocal gradient) of the Medium-Range Forecast Model (MRF) PBL scheme (Hong and Pan 1996), and for the soil model the multilayer Noah land surface model (LSM) is used. The longwave radiation scheme is based on the Rapid Radiative Transfer Model (RRTM) and shortwave radiation is based on Dudhia (1989).

Another regional model used in this study is the BRAMS (http://www.cptec.inpe.br/brams). The BRAMS is a new version of the RAMS (Cotton et al. 2003; Pielke et al. 1992) tailored to the tropics. The BRAMS/RAMS is a multipurpose numerical prediction model designed to simulate atmospheric circulations spanning in scale from hemispheric scales down to large-eddy circulations in the planetary boundary layer. Among the additional possibilities of BRAMS relative to RAMS, version 5.04, are the ensemble version of shallow cumulus and deep convection parameterizations (Grell and Devenyi 2002; Freitas et al. 2005) and the Land Ecosystem Atmosphere Feedback (LEAF-2) model of surface parameterizations. The cloud microphysics is the single moment bulk scheme from Walko et al. (1995), which includes five categories of ice: pristine ice crystals, snow, aggregates, graupel, and hail.

The GDE cumulus parameterization scheme consists of an ensemble of cumulus schemes in which multiple schemes are run within each grid box and the results are averaged to give the feedback to the model. All cumulus schemes in the GDE are of mass-flux nature, differing in updraft and downdraft with entrainment and detrainment parameters, and precipitation efficiencies. The dynamic control closures are based on either CAPE or low-level vertical velocity or moisture convergence. Those based on CAPE either balance the rate of change of CAPE or relax the CAPE to a climatological value, and those based on moisture convergence balance the cloud rainfall to the integrated vertical advection of moisture.

The measurement of SST through clouds by satellite microwave radiometers has been an elusive goal for many years. The early radiometers in the 1980s, like the Scanning Multichannel Microwave Radiometer (SMMR), were poorly calibrated, and later radiometers, like the Special Sensor Microwave Imager (SSM/I), lacked the low-frequency channels needed by the retrieval algorithm. Finally, in November 1997, the TMI radiometer with a 10.7-GHz channel was launched aboard the TRMM satellite. Unlike the Defense Meteorological Satellites Program (DMSP) platforms of the SSM/I, the TRMM satellite travels west to east in a semiequatorial orbit. The TMI data used in this study are three-day mean centered on 23 July 2005 with a pixel resolution of 0.25°. This produces data collected at changing local times for any given earth location between 40°S and 40°N.

For this study four experiments are designed, namely, WRFCON and WRFTMI using the WRF model and RAMCON and RAMTMI using the BRAMS model. WRFCON and RAMCON are the control experiments with climatological SST; WRFTMI and RAMTMI are the experiments with TMI SST as the boundary condition. All four experiments were carried out over identical single domains, with 15-km horizontal resolution having 260 × 200 grid points, covering the region [2.6°S–24°N, 55.2°–90.8°E]. In the vertical, both the models have 31 vertical layers with the top model layer at 50 hPa. For each simulation, a 54-h integration was started from 0000 UTC 25 July 2005. Figures 2a–c shows the climatological SST for 0000 UTC 25 July 2005, the observed SST from TMI, and their difference. This clearly shows that there is not much gradient in the SST field along the western coast of India in the climatology (Fig. 2a), whereas the TMI SST has a higher gradient structure (Fig. 2b). Over the central Indian Ocean, Bay of Bengal, and Arabian Sea, TMI SST is 1°C warmer relative to its climatology (Fig. 2c).

Fig. 2.

Initial SST (°C) valid at 0000 UTC 25 Jul 2005: (a) climatology, (b) TMI, and (c) climatology − TMI. Contour levels are 23°, 25°, 26°, 26.5°, 27°, 27.5°, 28°, 29°, and 30°C in (a) and (b) and 1°C in (c).

Fig. 2.

Initial SST (°C) valid at 0000 UTC 25 Jul 2005: (a) climatology, (b) TMI, and (c) climatology − TMI. Contour levels are 23°, 25°, 26°, 26.5°, 27°, 27.5°, 28°, 29°, and 30°C in (a) and (b) and 1°C in (c).

The model uses various terrestrial datasets for terrain, land use, soil type, soil temperature, vegetation fraction, snow albedo, monthly albedo, and so on, from the WRF users’ Web site (http://www.mmm.ucar.edu/wrf/users/) in both the WRF and BRAMS model initializations. Apart from these the NCEP final analyses (FNL) 1° × 1° 6-hourly data are used for preparing both the initial and boundary conditions, for both the model integrations.

For verification of the model results, India Meteorological Department (IMD) station and TRMM rainfall data are used. TRMM is a joint mission between the National Aeronautics and Space Administration (NASA) and the Japan Aerospace Exploration Agency (JAXA) to monitor tropical and subtropical precipitation and to estimate its associated latent heating. TRMM provides systematic visible, infrared, and microwave measurements of rainfall in the tropics as key inputs to weather and climate research. The NCEP FNL analyses are used to verify the simulated large-scale circulation patterns and thermodynamic fields.

4. Results and discussion

Two sets of forecast results are obtained in each model: one control experiment with climatological SST and the other using TMI SST. The purpose of the present study is to see the criticality of TMI SST for the simulation of this extreme rainfall event. The performance of each model is assessed with respect to upper- and lower-level wind, rainfall, moisture field, their thermal structure, different fluxes, and also by calculating the CAPE.

a. Circulation features

In July during the active Indian summer monsoon conditions, the strong westerly or southwesterly flow over the Arabian Sea feeds the formation of the offshore trough over the west coast of India. This triggers very heavy rainfall activity along the west coast of India, including Mumbai. The strengthening of this westerly or southwesterly takes place with either the Arabian Sea branch of the Indian monsoon being active or a depression or low pressure system forming over the northern part of the Bay of Bengal and moving to central India. Both the lower- and upper-level large-scale flow patterns over the Indian region are analyzed with 850- and 300-hPa winds valid at 1200 UTC 26 July 2005 from the NCEP analyses as shown in Figs. 3a,b. This shows the presence of strong cyclonic circulation, which extended up to the midtroposphere over Orissa and the eastern part of Madhya Pradesh. This is associated with a well-marked low pressure system near the surface. It also shows the presence of strong westerly or northwesterly winds with speeds of 20 m s−1 at the lower level at 850 hPa (Fig. 3a) over the western coast of India including Mumbai. This lower-level strong westerly or northwesterly wind is overlain by easterly or northeasterly wind at 300 hPa (Fig. 3b) with speeds of 6–12 m s−1 over Mumbai.

Fig. 3.

The NCEP analyses winds (m s−1) at 1200 UTC 26 Jul 2005 at (a) 850 and (b) 300 hPa. (a) Contour levels are 3, 6, and 12 m s−1; winds of 20 and 25 m s−1 are shaded. (b) Contour levels are 3, 6, and 9 m s−1, and winds of 12, 15, and 20 m s−1 are shaded.

Fig. 3.

The NCEP analyses winds (m s−1) at 1200 UTC 26 Jul 2005 at (a) 850 and (b) 300 hPa. (a) Contour levels are 3, 6, and 12 m s−1; winds of 20 and 25 m s−1 are shaded. (b) Contour levels are 3, 6, and 9 m s−1, and winds of 12, 15, and 20 m s−1 are shaded.

The atmospheric divergent circulation associated with the vertical motion field can be best understood by the divergent component of wind. Generally the atmospheric heating associated with convection will induce centers of divergence. The divergence fields (Figs. 4a–d) at different times of 26 July from the analysis indicate that upper-level divergence overlying low- and midlevel convergence had existed around Mumbai for some time and during 26 July there was a substantial increase in convergence below 700 hPa. This seems consistent with the observed changes in the rainfall intensity. It is also noted that low-level air over Mumbai was extremely moist while midlevel air was relatively dry, and once the change to enhanced low-level convergence had occurred, there was a supply of moisture to feed the rain system. Figures 4e–h show the vertical cross section of zonally averaged wind (72.5°–73.5°E) from the analysis for 26 July at different times. The zonal wind sections clearly illustrate the development of a low-level jet, which reaches a maximum strength of around 20 m s−1 near 700 hPa at 1200 and 1800 UTC 26 July. The corresponding subtropical jet is also evident from a wind maximum of around 30 m s−1. The convective system that produced the intense rainfall developed over Mumbai just poleward of the low-level jet. The vertical velocity (omega) fields essentially represent the distribution of convective heating. Figures 5 and 6 show, respectively, the time–height cross section of vertical velocity (omega) averaged at 18.5°–19.5°N, 72.5°–73.5°E for 0000 UTC 25 July to 0600 UTC 27 July 2005 (at intervals of 3 h) and latitude–height cross section of relative vorticity at 73°E valid for 1200 UTC 26 July obtained from the global analysis. This shows that the intense rainfall occurred during a period of sustained upward vertical motion. Thus the large-scale flow patterns over Mumbai present in the analysis were highly favorable from 24 July onward. This large-scale flow together with a local environmental setting influences the occurrence of this extreme event.

Fig. 4.

The latitude–height cross section of (top) analyzed divergence and (bottom) zonal wind for every 6 h starting from 0000 to 1800 UTC 26 Jul 2005 (at average longitude 72.5°–73.5°E). Contour intervals are 1 × 10−5 s−1 for divergence and 5 m s−1 for zonal wind.

Fig. 4.

The latitude–height cross section of (top) analyzed divergence and (bottom) zonal wind for every 6 h starting from 0000 to 1800 UTC 26 Jul 2005 (at average longitude 72.5°–73.5°E). Contour intervals are 1 × 10−5 s−1 for divergence and 5 m s−1 for zonal wind.

Fig. 5.

Time–height cross section of analyzed vertical velocity (omega) over 18.5°–19.5°N, 72.5°–73.5°E for the period 0000 UTC 25 Jul–0600 UTC 27 Jul 2005 with 3-h interval; contour interval is 0.1 hPa s−1.

Fig. 5.

Time–height cross section of analyzed vertical velocity (omega) over 18.5°–19.5°N, 72.5°–73.5°E for the period 0000 UTC 25 Jul–0600 UTC 27 Jul 2005 with 3-h interval; contour interval is 0.1 hPa s−1.

Fig. 6.

The latitude–height cross section of analyzed relative vorticity at longitude 73°E valid at 1200 UTC 26 Jul 2005. Contour interval is 1 × 10−5 s−1.

Fig. 6.

The latitude–height cross section of analyzed relative vorticity at longitude 73°E valid at 1200 UTC 26 Jul 2005. Contour interval is 1 × 10−5 s−1.

Figure 7 presents the simulated 850-hPa winds valid at 1200 UTC 26 July 2005 over the Indian region from the different experiments. All experiments predicted strong westerly or northwesterly winds over the western coast of India with maximum speeds of 20 m s−1. Both the models show the presence of strong cyclonic circulation over Orissa and the eastern part of Madhya Pradesh, though the center of the low pressure system is simulated a bit eastward as compared to the analyses. Figure 8 shows the simulated 300-hPa winds valid at 1200 UTC 26 July 2005 from different experiments. All experiments have simulated the upper-level easterly or northeasterly flow reasonably well, with speeds of approximately 12 m s−1 over Mumbai. As a whole, both the upper- and lower-level large-scale circulation features are simulated reasonably well by all schemes when they are compared with the corresponding NCEP analyses. Figure 9 represents the vertical cross sections of simulated divergence and zonal wind at longitudes 72.5°–73.5°E valid for 1200 UTC 26 July from different experiments. The north–south gradient in Figs. 9a–h is much larger, when compared with the corresponding analysis in Figs. 4c,g. This difference can be attributed to horizontal resolution. In the analysis the horizontal resolution is 1° (110 km) whereas in the simulation it is 15 km. The order of magnitude of the simulated zonal wind matches that in the analysis, whereas the simulated divergence is much stronger than the observed one. Thus the patterns of all major features described in the analysis are reproduced reasonably well in the simulation, though there are some discrepancies in the magnitude. It is also observed that the WRF model differs from the BRAMS because of the presence of more structure at small scales. The low-level jet in RAMCON and RAMTMI is relatively stronger as compared to observation. The upper-level divergence and low-level convergence are reproduced quite well in both RAMTMI and RAMCON. The midlevel (at 600 hPa) and upper-level (300 hPa) convergence (Fig. 4c) present in the analysis is reproduced slightly in RAMTMI, though the strength is less, which is not present in RAMCON. The large-scale features in WRFCON and WRFTMI do not differ much.

Fig. 7.

The simulated 850-hPa winds (m s−1) along with magnitude valid at 1200 UTC 26 Jul 2005 from the different experiments: (a) WRFCON, (b) WRFTMI, (c) RAMCON, and (d) RAMTMI. Contour levels are 3, 6, and 12 m s−1; winds of 20 and 25 m s−1 are shaded.

Fig. 7.

The simulated 850-hPa winds (m s−1) along with magnitude valid at 1200 UTC 26 Jul 2005 from the different experiments: (a) WRFCON, (b) WRFTMI, (c) RAMCON, and (d) RAMTMI. Contour levels are 3, 6, and 12 m s−1; winds of 20 and 25 m s−1 are shaded.

Fig. 8.

The simulated 300-hPa winds (m s−1) valid at 1200 UTC 26 Jul 2005 from the different experiments: (a) WRFCON, (b) WRFTMI, (c) RAMCON, and (d) RAMTMI. Contour levels are 3, 6 and 9 m s−1; winds of 12, 15, and 20 m s−1 are shaded.

Fig. 8.

The simulated 300-hPa winds (m s−1) valid at 1200 UTC 26 Jul 2005 from the different experiments: (a) WRFCON, (b) WRFTMI, (c) RAMCON, and (d) RAMTMI. Contour levels are 3, 6 and 9 m s−1; winds of 12, 15, and 20 m s−1 are shaded.

Fig. 9.

North–south cross section of (top) simulated divergence and (bottom) zonal wind at 1200 UTC 26 Jul 2005 from the different experiments (at 72.5°–73.5°E). Contour intervals are 5 × 10−5 s−1 for divergence and 5 m s−1 for zonal wind.

Fig. 9.

North–south cross section of (top) simulated divergence and (bottom) zonal wind at 1200 UTC 26 Jul 2005 from the different experiments (at 72.5°–73.5°E). Contour intervals are 5 × 10−5 s−1 for divergence and 5 m s−1 for zonal wind.

The role of convective heating and feedback between this process and the large-scale flow is explained by showing the vertical cross section of vertical velocity (omega) and derived quantities like relative vorticity. Figures 10 and 11 show the time–height cross section of vertical velocity (omega) over 18.5°–19.5°N, 72.5°–73.5°E and the relative vorticity at 73°E valid for 1200 UTC 26 July from different experiments. The maximum vertical velocity in WRFTMI is simulated around 1800 UTC 26 July, which is 6 h after the time of intense rainfall, whereas RAMCON and RAMTMI have simulated it at 1200 UTC 26 July. The timing of vertical velocity in WRFCON is delayed by 6 h. The WRFTMI shows the strong vertical motion a bit southward from Mumbai as compared to observation (figure not shown). Here the figures represent the area averages of vertical velocity. Both the models have overestimated the vertical velocity when compared with the analysis (Fig. 5). The relative vorticity (Fig. 11) is both a measure of the wind field response to the heating as well as an index of inertial instability, the latter being a measure of the resistance to horizontal motions. In RAMCON and RAMTMI experiments, the maximum simulated vorticity occurs over Mumbai, whereas in WRFCON and WRFTMI its position is a bit south of Mumbai during the storm. The magnitudes of maximum relative vorticity simulated by both the models are one order of magnitude higher when compared with the analysis (Fig. 6). This can also be attributed to the difference of horizontal resolution between simulations and analysis. Thus it is apparent that the observed large-scale flow, present in the analysis, was simulated reasonably well in both the models, though there are some differences in the magnitude of divergence and relative vorticity.

Fig. 10.

Time–height cross section of averaged simulated vertical velocity (omega) over 18.5°–19.5°N, 72.5°–73.5°E for the period 0000 UTC 25 Jul–0600 UTC 27 Jul 2005 with 3-h interval. Contour intervals are 0.5 hPa s−1 in (a) and (c) and 1 hPa s−1 in (b) and (d).

Fig. 10.

Time–height cross section of averaged simulated vertical velocity (omega) over 18.5°–19.5°N, 72.5°–73.5°E for the period 0000 UTC 25 Jul–0600 UTC 27 Jul 2005 with 3-h interval. Contour intervals are 0.5 hPa s−1 in (a) and (c) and 1 hPa s−1 in (b) and (d).

Fig. 11.

North–south cross section of the simulated relative vorticity (at 73°E) at 1200 UTC 26 Jul 2005 from different experiments. Contour interval is 10 s−1.

Fig. 11.

North–south cross section of the simulated relative vorticity (at 73°E) at 1200 UTC 26 Jul 2005 from different experiments. Contour interval is 10 s−1.

b. Thermodynamic features

The NCEP temperature and relative humidity fields are also critically analyzed at the lower and upper levels for the period of 25–27 July 2005 to find the exact nature of the thermodynamic structure during this event. Figures 12a,b show the 850- and 300-hPa temperature valid at 1200 UTC 26 July 2005. It is found that warm isotherms are lying to the north and west of Mumbai in the Arabian Sea at lower levels. Also, advection from the west at these levels is due to the strong westerlies (Fig. 3a) from the Arabian Sea, thus bringing the warm and moist air from the west. The latitude–height cross section of relative humidity from NCEP analyses shows (Figs. 13a,b) that the relative humidity was more than 90%–95% at various levels up to 400 hPa around Mumbai (average along 72.5°–73.5°E) between 0600 and 1200 UTC 26 July 2005. At 1200 UTC 26 July 2005, air was relatively dry in the midtroposphere at 750–550 hPa in the south of Mumbai, which resembles the actual observations as reported by IMD. Strong vertical wind shear along with an updraft in the low levels might have influenced the development of additional lifting needed for the occurrence of this severe localized rainfall at Mumbai. The study by Jenamani et al. (2006) shows, with different observational data from IMD, that the vertical uplifting of moist air mass was the highest at 1200 UTC 26 July 2005, among all observations between 24 and 27 July 2005. The rainfall rate was the largest over Mumbai during 0900–1500 UTC 26 July 2005.

Fig. 12.

Analyzed temperature (K) of 26 Jul 2005 at (a) 850 hPa and (b) 300 hPa.

Fig. 12.

Analyzed temperature (K) of 26 Jul 2005 at (a) 850 hPa and (b) 300 hPa.

Fig. 13.

Analyzed latitude–height cross section of the relative humidity (%) at 72.5°–73.5°E at (a) 0600 and (b) 1200 UTC 26 Jul 2005.

Fig. 13.

Analyzed latitude–height cross section of the relative humidity (%) at 72.5°–73.5°E at (a) 0600 and (b) 1200 UTC 26 Jul 2005.

The simulated 850-hPa temperatures valid for 1200 UTC 26 July 2005 (Figs. 14a,d) from all experiments resemble well the observed features. It is observed that the temperature in WRFTMI differs from that in the BRAMS because of the presence of more structure in small scales over the southern Bay of Bengal. The simulated 300-hPa temperatures valid for 1200 UTC 26 July 2005 (figure not shown) also show the warmer isotherm in the upper level when compared with analysis. Figure 15 shows the simulated latitude–height cross section of relative humidity from different experiments around Mumbai (average along 72.5°–73.5°E) for 1200 UTC 26 July 2005. All experiments have simulated the observed features of relative humidity reasonably well with a maximum up to 95% occurring up to 400 hPa over Mumbai. One interesting feature is that both RAMCON and RAMTMI have simulated a closed contour of the relatively dry region in the midtroposphere to the south of Mumbai at 1200 UTC. Actually this midlevel dryness along with strong vertical wind shear resulting in high conditional instability over Mumbai was reported by IMD at 0000 UTC 26 July. Thus the large-scale thermodynamic features over Mumbai City were also highly favorable for the occurrence of this event, and both the models are quite successful in simulating the observed thermodynamic structure for this specific event.

Fig. 14.

The simulated 850-hPa temperatures (K) valid at 1200 UTC 26 Jul 2005 from the experiments (a) WRFCON, (b) WRFTMI, (c) RAMCON, and (d) RAMTMI. Contour interval is 1 K.

Fig. 14.

The simulated 850-hPa temperatures (K) valid at 1200 UTC 26 Jul 2005 from the experiments (a) WRFCON, (b) WRFTMI, (c) RAMCON, and (d) RAMTMI. Contour interval is 1 K.

Fig. 15.

The latitude–height cross section of relative humidity (%) valid at 1200 UTC 26 Jul 2005 from the different experiments (average along 72.5°–73.5°E): (a) WRFCON, (b) WRFTMI, (c) RAMCON, and (d) RAMTMI. Contour interval 10%.

Fig. 15.

The latitude–height cross section of relative humidity (%) valid at 1200 UTC 26 Jul 2005 from the different experiments (average along 72.5°–73.5°E): (a) WRFCON, (b) WRFTMI, (c) RAMCON, and (d) RAMTMI. Contour interval 10%.

c. Rainfall

The rainfall, which is one of the most important parameters for the tropical weather systems, is discussed here. Figures 16a,b give the 24-h observed rainfall valid at 0300 UTC 27 July 2005 over Mumbai and neighboring stations as recorded by IMD and observed by TRMM. The observatory at Santa Cruz in north Mumbai received rainfall of 94.4 cm during the 24-h period ending at 0300 UTC 27 July 2005 and an even higher rainfall of 104.9 cm was recorded at Vihar Lake Observatory. However, Colaba in south Mumbai received merely 7.3 cm (Fig. 16a). The pattern of the rainfall as recorded by IMD was oriented along the northwest–southeast direction. The maximum rainfall measured by TRMM was only 32 cm (Fig. 16b); however, the position is captured very well when compared with the rainfall pattern as recorded by IMD. The rainfall recorded by IMD is a point measurement, while the resolution in TRMM rainfall is 0.25° both along scan and pixel; this can be one of the reasons for their differences in magnitude. Though the TRMM rainfall is a merged derived product from both IR and microwave measurements, sometimes TRMM underestimates the rainfall near the surface because of warm clouds. Table 2 shows the temporal distribution (every 3 h) of rainfall at Santa Cruz starting from 0300 UTC 26 July to 0300 UTC 27 July. This clearly shows that the maximum intensity of rainfall was from 0900 to 1500 UTC 26 July.

Fig. 16.

The 24-h cumulative rainfall (cm) valid at 0300 UTC 27 Jul 2005 observed by (a) IMD at different stations near Mumbai and (b) TRMM.

Fig. 16.

The 24-h cumulative rainfall (cm) valid at 0300 UTC 27 Jul 2005 observed by (a) IMD at different stations near Mumbai and (b) TRMM.

Table 2.

Temporal distribution of rainfall at Santa Cruz during 26–27 Jul 2005.

Temporal distribution of rainfall at Santa Cruz during 26–27 Jul 2005.
Temporal distribution of rainfall at Santa Cruz during 26–27 Jul 2005.

Figure 17 presents the simulated 24-h cumulative rainfall from different experiments valid at 0300 UTC 27 July 2005 over Mumbai. Both the models have simulated the observed features very well with TMI SST as compared to their counterparts with climatological SST, though there are some discrepancies in the location. Figures 17a,b show that the maximum rainfall simulated in WRFCON and WRFTMI is 96 cm in both cases, though the spatial distribution of the maximum rainfall in WRFTMI is wider compared to the distribution of the observed rainfall. In both the cases the exact location of maximum intensity is about 20–25 km (for WRFTMI) and 110–120 km (for WRFCON) south as compared to the observation. The intensity of rainfall is also simulated well in RAMTMI (Fig. 17d), as compared to the control simulation RAMCON (Fig. 17c). The intensity of maximum rainfall in RAMCON and RAMTMI is 24 and 96 cm, respectively, with the location being 10–15 km north of the true location. The orientation of rainfall is simulated well in RAMTMI, elongated along the northwest–southeast direction, though a bit northward. In the WRF model the orientation is circular with a tongue extending to the north. As a whole, the rainfall simulation in both the models improved a lot when TMI SST was used as the lower boundary condition.

Fig. 17.

The simulated 24-h cumulative rainfall (cm) valid at 0300 UTC 27 Jul 2005 from the different experiments: (a) WRFCON, (b)WRFTMI, (c) RAMCON, and (d) RAMTMI. Contour levels are 1, 2, 4, 8, 12, 16, 24, 32, 48, 64, 80, and 96 cm.

Fig. 17.

The simulated 24-h cumulative rainfall (cm) valid at 0300 UTC 27 Jul 2005 from the different experiments: (a) WRFCON, (b)WRFTMI, (c) RAMCON, and (d) RAMTMI. Contour levels are 1, 2, 4, 8, 12, 16, 24, 32, 48, 64, 80, and 96 cm.

Figure 18 shows 6-hourly accumulated rainfalls from TRMM over Mumbai starting from 0600 UTC 25 July to 0000 UTC 27 July 2005. It clearly shows a localized area of heavy precipitation directly over Mumbai at 1200 and 1800 UTC 26 July 2005 and 0000 UTC 27 July 2005, on the order of 8–16 cm (Figs. 18f–h). The TRMM pictures show the location of intense rainfall very accurately; however, the magnitude is underestimated.

Fig. 18.

The 6-hourly accumulated rainfall (cm) from the TRMM: (a) 0600, (b) 1200, (c) 1800 UTC 25 Jul 2005, (d) 0000, (e) 0600, (f) 1200, and (g) 1800 UTC 26 Jul 2005, and (h) 0000 UTC 27 Jul 2005. Contour levels are 1, 2, 4, 8, 16, 32, and 48 cm.

Fig. 18.

The 6-hourly accumulated rainfall (cm) from the TRMM: (a) 0600, (b) 1200, (c) 1800 UTC 25 Jul 2005, (d) 0000, (e) 0600, (f) 1200, and (g) 1800 UTC 26 Jul 2005, and (h) 0000 UTC 27 Jul 2005. Contour levels are 1, 2, 4, 8, 16, 32, and 48 cm.

Figures 19 and 20 give the simulated 6-hourly rainfall starting from 0600 UTC 26 July 2005 using the WRF and BRAMS models. The WRF model has simulated rainfall along the west coast of India and failed to simulate the rainfall along 22°–24°N in both the cases (WRFCON and WRFTMI) at 1800 UTC 26 July and 0000 UTC 27 July, respectively, when compared with TRMM. The BRAMS model has captured the intense rainfall location near Mumbai at 0600 and 1200 UTC 26 July, whereas it has failed to simulate the other rainfall band such as rainfall along the Western Ghats. In the simulations using the BRAMS model, the intensity of maximum rainfall is well captured when TMI SST is used. In both the WRF experiments, the intensity is very well captured, but the position is even farther south of the true location with climatological SST than with TMI SST. Overall, both the models are found capable of simulating this specific event even with a horizontal resolution of 15 km.

Fig. 19.

The simulated 6-hourly cumulative rainfall (cm) valid at 0600 UTC 26 Jul–0000 UTC 27 Jul 2005 from WRFCON and WRFTMI. Contour levels are 1, 2, 4, 8, 16, 32, and 48 cm.

Fig. 19.

The simulated 6-hourly cumulative rainfall (cm) valid at 0600 UTC 26 Jul–0000 UTC 27 Jul 2005 from WRFCON and WRFTMI. Contour levels are 1, 2, 4, 8, 16, 32, and 48 cm.

Fig. 20.

The simulated 6-hourly cumulative rainfall (cm) valid at 0600 UTC 26 Jul–0000 UTC 27 Jul 2005 from RAMCON and RAMTMI. Contour levels are 1, 2, 4, 8, 16, 32, and 48 cm.

Fig. 20.

The simulated 6-hourly cumulative rainfall (cm) valid at 0600 UTC 26 Jul–0000 UTC 27 Jul 2005 from RAMCON and RAMTMI. Contour levels are 1, 2, 4, 8, 16, 32, and 48 cm.

Figure 21 shows the time series of 24-h cumulative observed rainfall over Santa Cruz and Colaba and the corresponding 24-h cumulative rainfall simulated by different experiments at corresponding grid points nearest to Santa Cruz. It clearly indicates that WRFTMI and RAMTMI are able to capture the sharp increase in rainfall distribution as in observation. Both the models with TMI SST have simulated the total cumulative rainfall of about 96 cm for RAMTMI and 100 cm for WRFTMI, whereas with climatological SST, in both WRFCON and RAMCON the 24-h cumulative rainfall near Santa Cruz is about 6–8 cm only. In RAMTMI a sharp increase has started 6 h before the actual increase, and in WRFTMI it started 6 h behind the actual start. Thus, on the whole, both the models have simulated the observed features of the rainfall quite satisfactorily when TMI SST is used as the boundary condition.

Fig. 21.

The observed 24-h cumulative rainfall (cm) time series over Santa Cruz and Colaba from 0300 UTC 26 Jul to 0300 UTC 27 Jul 2005 and the corresponding simulated rainfall (cm) near Santa Cruz from the different experiments.

Fig. 21.

The observed 24-h cumulative rainfall (cm) time series over Santa Cruz and Colaba from 0300 UTC 26 Jul to 0300 UTC 27 Jul 2005 and the corresponding simulated rainfall (cm) near Santa Cruz from the different experiments.

d. Fluxes

The study by Stensrud (1996) has shown the mechanisms that are responsible for the formation of low-level jet (viz. inertial oscillation, shallow baroclinicity, terrain effects, vertical parcel displacements, etc.). However, in the coastal regions, significant changes in surface characteristics (viz. the horizontal differences in the sensible and latent heat fluxes) produce strong low-level baroclinicity within the PBL, which in turn produces a low-level jet. In coastal regions significant diurnal changes in the strength of the low-level jet occur because of diurnal changes in the surface fluxes. The low-level jet is enhanced in many heavy precipitation events because of the gradients in sensible and latent heat fluxes, due to gradients in soil moisture (Paegle et al. 1996; Bernardet et al. 2000). Mo et al. (1995) found that heavy rainfall over the Mississippi River basin may partly be explained by the strong moisture transport because of strong low-level jets. Since Mumbai is situated on the west coast of India, here an attempt is made to assess whether the above factors can give some clue about this heavy precipitating event by analyzing the simulated fluxes. Figure 22 shows the sensible heat fluxes at 0600 UTC 26 July 2005 from all experiments. The simulations using the WRF model show strong gradients in sensible heat fluxes both in control as well as TMI SST experiments around Mumbai. The TMI SST experiment in the BRAMS model shows larger gradient in sensible heat flux as compared to that in the control experiment, which resembles the above findings. This strong gradient in sensible heat generates the horizontal pressure gradient in the boundary layer, which enhances the low-level jet at the top of the boundary layer. The presence of strong low-level convergences (Figs. 9a–d) exactly over Mumbai in the simulations supports this feature.

Fig. 22.

The simulated sensible heat flux valid at 0600 UTC 26 Jul from the different experiments: (a) WRFCON, (b) WRFTMI, (c) RAMCON, and (d) RAMTMI. Contour interval is 50 W m−2; values greater than 200 W m−2 are shaded.

Fig. 22.

The simulated sensible heat flux valid at 0600 UTC 26 Jul from the different experiments: (a) WRFCON, (b) WRFTMI, (c) RAMCON, and (d) RAMTMI. Contour interval is 50 W m−2; values greater than 200 W m−2 are shaded.

The interaction of land surface and the atmosphere can produce larger spatial variability in latent heat fluxes. It is another important parameter for producing variability in tropical rainfall. Figures 23a–d show the latent heat fluxes at 0600 UTC 26 July 2005 from different experiments. The general distribution of latent heat flux is similar in both the control and TMI experiments of both the models, with a larger gradient of latent heat flux along the Western Ghats in the WRF simulations and in the southwest of Mumbai in the BRAMS simulations ranging from 70 to over 280 W m−2 in a very small distance. However, in the model intercomparison, the structures are somewhat different. This difference is basically due to the difference of land surface model used in the BRAMS and WRF models. Over central India the simulated latent heat flux is higher in the BRAMS model as compared to that in the WRF model. The variability in latent heat flux found from grid cell to grid cell may result from rainfall–soil moisture feedback. Thus the patterns of simulated fluxes in the two models are not the same; however, with TMI SST they both have produced the maximum rainfall of 96 cm around Mumbai.

Fig. 23.

The simulated latent heat flux valid at 0600 UTC 26 Jul from the different experiments: (a) WRFCON, (b) WRFTMI, (c) RAMCON, and (d) RAMTMI. Contour interval is 70 W m−2; values greater than 280 W m−2 are shaded.

Fig. 23.

The simulated latent heat flux valid at 0600 UTC 26 Jul from the different experiments: (a) WRFCON, (b) WRFTMI, (c) RAMCON, and (d) RAMTMI. Contour interval is 70 W m−2; values greater than 280 W m−2 are shaded.

e. CAPE

Variation of convection in the atmosphere depends on the dynamics as well as thermodynamic instability index, which in turn depends on CAPE in the moist condition. The CAPE is one of the most critical parameters for atmospheric convection in the moist atmosphere and it measures the vertical instability. The CAPE is defined as the work done by the buoyancy force on a parcel lifted through the atmosphere moist adiabatically. Deep clouds can develop as a result of the ascent of air from a given level, only if its CAPE is greater than zero. When disturbances occur, heavy precipitation, strong winds, and downdrafts decrease the energy of the air from a given level—again only if its CAPE is greater than zero. The CAPE over Santa Cruz is calculated using the sounding data for 25–26 July 2005. It shows the highest value 4341 J kg−1 valid at 0000 UTC 25 July, just one day before the occurrence of the rainfall, and the value decreased to 3267 J kg−1 at 0000 UTC on the day of the occurrence of the rainfall and further reduced to 252 J kg−1 at 0000 UTC 27 July. Thus CAPE increased by 3000–4000 J kg−1 before convection and decreased by a similar magnitude following the convection. These large values of observed CAPE suggest that strong atmospheric thermodynamic instability was present along with large-scale favorable synoptic features over Santa Cruz during this localized event.

To assess this observed finding of thermodynamic instability, the simulated CAPE is calculated by assuming a vertical profile at each grid point as a sounding for the region (18°–20°N, 72°–74°E) covering Mumbai. Figures 24 and 25 show the simulated CAPE from different experiments for 0000 UTC 26 and 27 July. The WRFTMI run has simulated the CAPE around 3000 J kg−1 near Mumbai on 26 July, whereas WRFCON has also simulated the same amount, but a bit southwest of Mumbai. The RAMCON and RAMTMI simulations show around 2100 and 3600 J Kg−1 of CAPE on that day around 30–50 km northwest of Mumbai. Basically when disturbances attenuate, the air–sea fluxes increase the energy of the surface air, while the temperature of the air aloft decreases because of radiative cooling. These factors destabilize the atmosphere and build up CAPE. On the following day the WRF model simulated more than 2000 J Kg−1 CAPE around Mumbai; however the BRAMS model simulated approximately 1000 J Kg−1 in the same place. While both values are higher than the observed one, the decreasing tendency of CAPE over Santa Cruz as in observation is present in the simulations. This is because deep cloud makes the upper troposphere warmer, and as a result the atmosphere becomes less unstable and CAPE is substantially reduced during or after the disturbances. Thus, on the whole, both the models are able to simulate the strong thermodynamic instability, which is one of the favorable conditions for heavy precipitation, covering Mumbai with some spatial differences in CAPE.

Fig. 24.

The simulated CAPE valid at 0000 UTC 26 Jul 2005 from the different experiments: (a) WRFCON, (b) WRFTMI, (c) RAMCON, and (d) RAMTMI. Contour interval is 300 J kg−1.

Fig. 24.

The simulated CAPE valid at 0000 UTC 26 Jul 2005 from the different experiments: (a) WRFCON, (b) WRFTMI, (c) RAMCON, and (d) RAMTMI. Contour interval is 300 J kg−1.

Fig. 25.

The simulated CAPE valid at 0000 UTC 27 Jul 2005 from the different experiments: (a) WRFCON, (b) WRFTMI, (c) RAMCON, and (d) RAMTMI. Contour interval is 300 J kg−1.

Fig. 25.

The simulated CAPE valid at 0000 UTC 27 Jul 2005 from the different experiments: (a) WRFCON, (b) WRFTMI, (c) RAMCON, and (d) RAMTMI. Contour interval is 300 J kg−1.

5. Conclusions

A qualitative assessment of diagnostic investigation and the numerical simulation of the unprecedented rainfall event of 26 July 2005 over Mumbai are attempted here with the BRAMS and WRF models using the GDE cumulus parameterization scheme. The criticality of TMI SST for simulating this extreme event is also ascertained. The large-scale circulation and different thermodynamical features both at the upper and lower levels in all simulations resemble well the diagnostic analyses. The localized heavy precipitation around Mumbai is captured reasonably well in both the models when TMI SST is used. In WRFTMI and RAMTMI the intensity of rainfall is captured very well though the position is around 20 km southeast in the WRF simulations and 10 km north of the actual position in the BRAMS model. This is quite encouraging with the horizontal resolution of 15 km. Both the models simulated large CAPE (Figs. 24b,d) with TMI SST along with strong low-level convergences near Mumbai (Figs. 9b,d), which is reflected in the high rainfall rate. A model run with 15-km grid spacing is not expected to have better precision. The area of maximum intensity of rainfall in the WRFTMI simulation is wider as compared to that in WRFCON. If we consider the intensity of rainfall, then both the control simulations (with climatological SST) have failed to capture the observed intensity at the position of interest. In WRFCON though a large CAPE is simulated (Fig. 24a), the low-level convergence (Fig. 9a) is somewhat south of Mumbai. The large-scale low-level convergence (Fig. 9c) with a bit smaller CAPE (Fig. 24c) simulated in RAMCON results in low rainfall. Thus it can be inferred that a strong low-level jet along with strong thermodynamic instability can induce such heavy rainfall events, and that climatological SSTs fail to produce these whereas TMI SSTs do. It is very encouraging that as a whole both the models have the predicting capability to capture such an intense rainfall so accurately with TMI SST; however, the BRAMS model seems to be more sensitive to the use of observed SST than the WRF model. Comparing Figs. 19 and 20, it is noted that in the BRAMS model the maximum values changed substantially with TMI SST (also noticeable in Fig. 17). However, the WRF values remain virtually unaltered, and the main effect of TMI SST is the relocation of the maximum rainfall significantly to the north. This finding is corroborated by Fig. 24, which shows that the impact of TMI SST on CAPE is larger on the BRAMS model than on the WRF model. Some issues remain unanswered in this study: the sensitivity of the forecasts to the land surface processes needs to be investigated, and additional observational and model diagnostics and budget calculations are required to find the interaction between the strong convective heating and the large-scale flow, and also that between the convection and subtropical jet, particularly with regard to extreme rain events. The testing of more case studies with higher and improved versions of the WRF model as well as other models by incorporation of satellite data and Doppler radar data through data assimilation system is required. Probably the most encouraging aspect of this study is the very significant forecast skill displayed, though in hindcast mode, on the very complex system of the extreme rainfall event of Mumbai, both in the WRF model as well as the BRAMS model with TMI SST and demonstration of the specific role of this last dataset. Although the above conclusions are based on a very limited number of experiments, this study can provide some insight to the WRF model users over the Indian region and prompt the modeling community to pursue and evaluate real-time quantitative precipitation forecasting, with the boundary conditions obtained from suitable satellite sensors.

Acknowledgments

The authors thank the two anonymous reviewers for their critical and insightful comments/valuable suggestions, which were helpful in substantially improving the content and quality of the presentation of this manuscript. The Director, Space Applications Centre (SAC), and the Deputy Director, RESA SAC, ISRO Ahmedabad, are thanked for their encouragement and help. The authors acknowledge the use of the WRF Model, the BRAMS, and the WRF User Support Group for useful suggestions during the model installation. The use of reanalyzed data from NCEP, TRMM rainfall data (http://trmm.gsfc.nasa.gov), and the station rainfall from the IMD is thankfully acknowledged. The authors are very thankful to Dr. P. S. Desai, retired Chief Scientist (Remote Sensing Applications), for his scientific input and for taking pains in reviewing this paper.

REFERENCES

REFERENCES
Bernardet
,
L. R.
,
L. D.
Grasso
,
J. E.
Nachamkin
,
C. A.
Finley
, and
W. R.
Cotton
,
2000
:
Simulating convective events using a high-resolution mesoscale model.
J. Geophys. Res.
,
105
,
14963
14982
.
Bohra
,
A. K.
,
S.
Basu
,
E. N.
Rajagopal
,
G. R.
Iyengar
,
M.
Das Gupta
,
R.
Ashrit
, and
B.
Athiyaman
,
2006
:
Heavy rainfall episode over Mumbai on 26 July 2005: Assessment of NWP guidance.
Curr. Sci.
,
90
,
1188
1194
.
Chen
,
C.
, and
W. R.
Cotton
,
1983
:
A one-dimensional simulation of the stratocumulus-capped mixed layer.
Bound.-Layer Meteor.
,
25
,
289
321
.
Chen
,
F.
, and
J.
Dudhia
,
2001
:
Coupling an advanced land surface–hydrology model with the Penn State–NCAR MM5 modeling system. Part I: Model implementation and sensitivity.
Mon. Wea. Rev.
,
129
,
569
585
.
Cotton
,
W. R.
, and
Coauthors
,
2003
:
RAMS 2001: Current status and future directions.
Meteor. Atmos. Phys.
,
82
,
5
29
.
Dudhia
,
J.
,
1989
:
Numerical study of convection observed during the winter monsoon experiment using a mesoscale two-dimensional model.
J. Atmos. Sci.
,
46
,
3077
3107
.
Freitas
,
S.
, and
Coauthors
,
2005
:
Monitoring the transport of biomass burning emissions in South America.
Environ. Fluid Mech.
,
5
,
135
167
.
Gal-Chen
,
T.
, and
R. C. J.
Somerville
,
1975
:
On the use of a coordinate transformation for the solution of the Navier-Stokes equations.
J. Comput. Phys.
,
17
,
209
228
.
Gallus
Jr.,
W. A.
, and
J.
Bresch
,
2006
:
Comparison of impacts of WRF dynamic core, physics package, and initial conditions on warm season rainfall forecasts.
Mon. Wea. Rev.
,
134
,
2632
2641
.
George
,
P. A.
,
1956
:
Effect of offshore vortices on rainfall along the west coast of India.
Indian J. Meteor. Geophys.
,
7
,
225
240
.
Grell
,
G. J.
, and
D.
Devenyi
,
2002
:
A generalized approach to parameterizing convection combining ensemble and data assimilation technique.
Geophys. Res. Lett.
,
29
.
1693, doi:10.1029/2002GL015311
.
Hong
,
S. Y.
, and
H. L.
Pan
,
1996
:
Nonlocal boundary layer vertical diffusion in a medium-range forecast model.
Mon. Wea. Rev.
,
124
,
2322
2339
.
Jayaram
,
M.
,
1965
:
A preliminary study of an objective method of forecasting heavy rainfall over Mumbai and neighborhood during the month of July.
Indian J. Meteor. Geophys.
,
16
,
557
564
.
Jenamani
,
R. K.
,
S. C.
Bhan
, and
S. R.
Kalsi
,
2006
:
Observational/forecasting aspects of the meteorological event that caused a record highest rainfall in Mumbai.
Curr. Sci.
,
90
,
1344
1362
.
Kershaw
,
R.
,
1998
:
The effect of a sea surface temperature anomaly on a prediction of the onset of the south-west monsoon over India.
Quart. J. Roy. Meteor. Soc.
,
114
,
325
345
.
Kidder
,
S. Q.
,
M. D.
Goldberg
,
R. M.
Zehr
,
M.
DeMaria
,
J. F. W.
Purdom
,
C. S.
Velden
,
N. C.
Grody
, and
S. J.
Kusselson
,
2000
:
Satellite analysis of tropical cyclones using the Advanced Microwave Sounding Unit (AMSU).
Bull. Amer. Meteor. Soc.
,
81
,
1241
1259
.
Laprise
,
R.
,
1992
:
The Euler equations of motion with hydrostatic pressure as an independent variable.
Mon. Wea. Rev.
,
120
,
197
207
.
Lebeaupin
,
C.
,
V.
Ducrocq
, and
H.
Giordani
,
2006
:
Sensitivity of torrential rain events to the sea surface temperature based on high-resolution numerical forecasts.
J. Geophys. Res.
,
111
.
D12110, doi:10.1029/2005JD006541
.
Mahrer
,
Y.
, and
R. A.
Pielke
,
1977
:
A numerical study of the airflow over irregular terrain.
Beitr. Phys. Atmos.
,
50
,
98
113
.
Mandal
,
M.
,
U. C.
Mohanty
, and
S.
Raman
,
2004
:
A study of the impact of parameterization of physical processes on prediction of tropical cyclone over the Bay of Bengal with NCAR/PSU mesoscale model (MM5).
Nat. Hazards
,
31
,
391
414
.
Mellor
,
G. L.
, and
J.
Yamada
,
1982
:
Development of a turbulence closure model for geophysical fluid problems.
Rev. Geophys.
,
20
,
851
875
.
Meneguzzo
,
F.
,
M.
Pasqui
,
G.
Menduni
,
G.
Messeri
,
B.
Gozzini
,
D.
Grifoni
,
M.
Rossi
, and
G.
Maracchi
,
2004
:
Sensitivity of meteorological high-resolution numerical simulations of the biggest floods occurred over the Arno river basin, Italy, in the 20th century.
J. Hydrol.
,
288
,
37
56
.
Mlawer
,
E. J.
,
S. J.
Taubman
,
P. D.
Brown
,
M. J.
Iacono
, and
S. A.
Clough
,
1997
:
Radiative transfer for inhomogeneous atmospheres: RRTM, a validated correlated-k model for the longwave.
J. Geophys. Res.
,
102
,
16663
16682
.
Mo
,
K. C.
,
J.
Nogues-Paegle
, and
J.
Paegle
,
1995
:
Physical mechanisms of the 1993 summer floods.
J. Atmos. Sci.
,
52
,
879
895
.
Paegle
,
J.
,
K. C.
Mo
, and
J.
Nogues-Paegle
,
1996
:
Dependence of simulated precipitation on surface evaporation during the 1993 United States summer floods.
Mon. Wea. Rev.
,
124
,
345
361
.
Pielke
,
R. A.
, and
Coauthors
,
1992
:
A comprehensive meteorological modeling system—RAMS.
Meteor. Atmos. Phys.
,
49
,
69
91
.
Rao
,
Y. P.
,
1976
:
Southwest Monsoon. Synoptic Meteor.
Monogr., No. 1, Indian Meteorological Department, 367 pp
.
Rautenbach
,
C. J.
,
1998
:
The unusual rainfall and sea surface temperature characteristics in the South African region during the 1995/96 summer season.
Water SA
,
24
,
165
172
.
Sharma
,
V. V.
,
1965
:
A study of heavy rainfall occurrence over Santacruz (Mumbai) with special reference to supersonic flight operations.
Indian J. Meteor. Geophys.
,
20
,
263
266
.
Shyamala
,
B.
, and
C. V. V.
Bhadram
,
2006
:
Impact of mesoscale–synoptic scale interactions on the Mumbai historical rain event during 26–27 July 2005.
Curr. Sci.
,
91
,
1649
1654
.
Singh
,
R.
,
P. K.
Pal
,
C. M.
Kishtawal
, and
P. C.
Joshi
,
2005
:
Impact of bogus vortex for track and intensity prediction of tropical cyclone.
J. Earth Syst. Sci.
,
114
,
427
436
.
Skamarock
,
W. C.
,
J. B.
Klemp
,
J.
Dudhia
,
D. O.
Gill
,
D. M.
Barker
,
W.
Wang
, and
J. G.
Powers
,
2005
:
A description of the Advanced Research WRF Version 2. NCAR Tech. Note NCAR/TN-468+STR, 88 pp
.
Stensrud
,
D. J.
,
1996
:
Importance of low-level jets to climate: A review.
J. Climate
,
9
,
1698
1711
.
Tripoli
,
G. J.
, and
W. R.
Cotton
,
1980
:
A numerical investigation of several factors contributing to the observed variable intensity of deep convection over South Florida.
J. Appl. Meteor.
,
19
,
1037
1063
.
Vaidya
,
S. S.
,
2007
:
Simulation of weather systems over Indian region using mesoscale models.
Meteor. Atmos. Phys.
,
95
,
15
26
.
Vaidya
,
S. S.
, and
J. R.
Kulkarni
,
2007
:
Simulation of heavy precipitation over Santacruz, Mumbai on 26 July 2005, using mesoscale model.
Meteor. Atmos. Phys.
,
98
,
55
66
.
Walko
,
R. L.
,
W. R.
Cotton
,
M. P.
Meyers
, and
J. Y.
Harrington
,
1995
:
New RAMS cloud microphysics parameterization. Part I: The single-moment scheme.
Atmos. Res.
,
38
,
29
62
.
Walko
,
R. L.
, and
Coauthors
,
2000
:
Coupled atmosphere–biophysics–hydrology models for environmental modeling.
J. Appl. Meteor.
,
39
,
931
944
.
Wentz
,
F. J.
,
C.
Gentemann
,
D.
Smith
, and
D.
Chelton
,
2000
:
Satellite measurements of sea surface temperature through clouds.
Science
,
288
,
847
850
.

Footnotes

Corresponding author address: Dr. S. K. Deb, Atmospheric Sciences Division, Meteorology and Oceanography Group, Space Applications Centre, Indian Space Research Organization (ISRO), Ahmedabad 380015, India. Email: sanjib_deb@rediffmail.com