The long-lived monsoon depressions of 2006 and their linkage with the Indian Ocean Dipole

Transcription

1 INTERNATIONAL JOURNAL OF CLIMATOLOGY Int. J. Climatol. 31: (2011) Published online 6 May 2010 in Wiley Online Library (wileyonlinelibrary.com) DOI: /joc.2156 The long-lived monsoon depressions of 2006 and their linkage with the Indian Ocean Dipole R. Krishnan,* D. C. Ayantika, Vinay Kumar and Samir Pokhrel Climate and Global Modelling Division, Indian Institute of Tropical Meteorology, Pune , India ABSTRACT: A highlight of the 2006 boreal summer monsoon season was the enhanced activity of long-lived monsoon depressions and low-pressure systems (LPS) over the Indian region. Another important phenomenon during this period was the evolution of a positive Indian Ocean Dipole (PIOD) event. Although previous studies have investigated the impact of PIOD on the large-scale monsoon response, their influence on monsoon LPS activity is not well understood. Based on detailed diagnostic analyses of monsoon LPS during 2006, as well as those associated with other PIOD events during , the present work addresses two specific issues concerning the roles of PIOD-induced large-scale circulation changes and internal feedbacks between latent heating and dynamics, in sustaining the monsoon LPS activity. The results show that PIOD conditions generally favour increased propensity of long-lived (>5 days) LPS with long westward tracks extending into northwest India. The average contribution of long-lived monsoon LPS to the total is found to be approximately 12% higher during PIOD episodes as compared to non-piod. The PIOD events showed two important large-scale elements conducive for enhancement of LPS activity: strengthening of cross-equatorial moisture transport from south-eastern tropical Indian Ocean into the Bay of Bengal and enrichment of barotropic instability of monsoon flow. Estimates of latent-heating profiles from TRMM-satellite products during the 2006 LPS showed heating in the mesoscale updrafts above 600 hpa with maximum approximately 400 hpa; while cooling prevailed in lower levels. Stratiform precipitation covered approximately 70 85% of rain area during the prolonged LPS; and the large-scale monsoon Hadley-type circulation was found to be intensified with strong mid-level inflows entering the stratiform rain region. The overall findings suggest that the PIOD-induced background circulation together with internal feedbacks between mesoscale convective systems and large-scale circulation can effectively enhance the longevity of monsoon LPS. These results should serve as important inputs for numerical weather forecasting of extreme rainfall events associated with the regional monsoon phenomenon. Copyright 2010 Royal Meteorological Society KEY WORDS monsoon depressions; life period; Indian Ocean Dipole Received 19 March 2009; Revised 3 February 2010; Accepted 19 March Introduction Monsoon lows and depressions, which form over the north Bay of Bengal and move west-northwest (WNW) along the quasi-stationary trough across north-central India, are one of the primary rain-producing synopticscale disturbances of the Indian summer monsoon. The general synoptic features of these disturbances are well documented (Koteswaram and Rao, 1963; Rao, 1976). This study focuses on the monsoon depressions during the summer of This season was rather outstanding in terms of the large number of lows and depressions that produced widespread and heavy rainfall over central India (Jayanthi et al., 2006). Quite a few of them lasted for more than a week during the course of their westward traverse across the subcontinent. Although it is true that the long-lasting nature of the 2006 depressions is not * Correspondence to: R. Krishnan, Climate and Global Modelling Division, Indian Institute of Tropical Meteorology, Pashan, NCL Post, Pune , India. krish@tropmet.res.in; krish0365@yahoo.com unprecedented, it is worthwhile investigating the roles of the large-scale conditions, as well as the synoptic and mesoscale convective systems (MCS) in sustaining the monsoon depression activity. An important large-scale feature of 2006 was the evolution of a positive Indian Ocean Dipole (PIOD) event in the tropical Indian Ocean (Luo et al., 2006). The IOD is a natural mode of ocean atmosphere coupled instability of the Indian Ocean which exerts significant influence on the regional and global climate (Behera et al., 1999; Saji et al., 1999; Webster et al., 1999; Black et al., 2003; Yamagata et al., 2004, Kripalani et al., 2005). A PIOD episode is characterized by cold sea surface temperature (SST) anomalies in the southeastern tropical Indian Ocean (SETIO) off the Coast of Sumatra and warm anomalies in the western tropical Indian Ocean (WIO). The pattern of SST anomalies in the tropical Indian Ocean is reversed during the negative IOD phase. The ocean atmosphere coupling during PIOD events maintains below-normal rainfall over SETIO and abovenormal rainfall over WIO. Studies have shown that Copyright 2010 Royal Meteorological Society

2 MONSOON DEPRESSIONS OF 2006 AND INDIAN OCEAN DIPOLE 1335 PIOD events tend to favour enhanced summer monsoon precipitation over the Indian landmass through supply of moisture from the SETIO region towards the Bay of Bengal and the Indo-Gangetic plains (Behera et al., 1999; Ashok et al., 2004). In a recent study, Ajayamohan and Rao (2008) reported that PIOD events can modulate extreme rainfall events over central India by favouring enhanced genesis of low-pressure systems (LPS) over the Bay of Bengal. The point that is central to this discussion pertains to the longevity of monsoon lows and depressions which can often be responsible for extensive rainfall leading to excessive flooding and widespread damage as witnessed over central India in 2006 (Jayanthi et al., 2006). The 2006 LPS were long-lasting, it is not obvious if the anomalous conditions during PIOD events were responsible for the longevity of monsoon disturbances. Although previous studies (Behera et al., 1999; Ashok et al., 2004) have noted the importance of PIOD events in enhancing the monsoon precipitation over the Indian region, their influence on the monsoon LPS activity is not well understood. One of the questions addressed in this article is the role of the background large-scale conditions during PIOD events in prolonging the monsoon LPS activity over extended periods (approximately a week or so). In order to delineate the possible large-scale control of PIOD events on the monsoon disturbances, a detailed analysis of LPS tracks, rainfall and atmospheric circulation for the period has been additionally carried out. The second issue examined in this study pertains to the role of internal feedbacks among latent heating, the synoptic and large-scale motions in sustaining the monsoon disturbances. This issue is relevant because moist processes and organization of MCS are known to be crucial for the growth of monsoon depressions (Krishnamurti et al., 1976, Houze and Churchill, 1987; Stano et al., 2002; Houze, 2004). Here, the emphasis is on gaining deeper understanding of the feedbacks between latent heating and dynamics which can enhance the longevity of monsoon LPS. For this purpose, various high resolution satellite products based on the Tropical Rainfall Measurement Mission (TRMM) data in conjunction with the National Centers for Environmental Prediction (NCEP) / National Center for Atmospheric Research (NCAR) reanalysis have been analyzed for the 2006 monsoon LPS. The details of the datasets are described in Section Datasets The analysis of the 2006 monsoon LPS is based on atmospheric parameters using the NCEP/NCAR reanalysis (Kistler et al., 2001), as well as satellite-based and surface observations. The SST is based on the National Oceanic and Atmospheric Administration (NOAA) Optimum Interpolated (OI) SST (Reynolds and Smith, 1994). The LPS tracks were obtained from India Meteorological Department (IMD) daily weather reports. Rainfall during the 2006 depressions have been examined using data from the TRMM Microwave Imager (TMI) which is a passive sensor that measures microwave radiation at different wavelengths emitted by objects (e.g. land, oceans, clouds, precipitation, etc.) based on the temperature of the objects. The TMI works in the frequencies 10.65, 19.35, 21, 37 and 85.5 GHz and scans conically over a swath of 760 km with an instantaneous field of view of varying sizes for different frequencies. The complete description of the TRMM sensor package is given by Kummerow et al. (1998). The TMI rainfall used here is the merged TRMM 3B42 product which utilizes the TMI rain estimates to adjust infrared (IR) rain rates at high temporal resolution ( 3.0 h) over daily gridded latitude longitude boxes (Huffman et al., 1995). The study also utilizes rainfall from the Climate Prediction Center (CPC) Merged Analysis of Precipitation (CMAP) dataset ( ) which is a product of merging rain gauge observations, precipitation estimates from satellites and numerical models (Xie and Arkin, 1997). In addition, latent heating from the TMI-V6 2A-12 dataset has also been analysed for the 2006 monsoon LPS. The latentheating data is based on the vertical profiles of hydrometeors generated from TMI brightness temperatures by blending radiometric data with dynamical cloud models (Kummerow et al., 1996). Finally, inferences have been drawn about the precipitation-type of monsoon depressions using the TRMM precipitation radar (PR) 2A-23 dataset which has high horizontal (approximately 4 km) and vertical (approximately 250 m) resolutions (Iguchi et al., 2000). The PR was the first space-borne instrument designed to provide three-dimensional structure of reflectivity and rainfall over land as well as oceans. This instrument measures backscattered radiowave signal from hydrometeors, as the radar signal travels through clouds to the surface and back to the TRMM satellite. The observation band frequency has two channels and GHz, which allows cross-track scanning and these measurements yield valuable information on the intensity and distribution of rain, as well as the rain type (stratiform, convective, etc.). For the past PIOD cases atmospheric winds, temperature, vertical velocity and specific humidity from NCEP reanalysis are employed for the period ; and the HadISST dataset (Rayner et al., 2003) is used to understand the anomalous SST conditions. Information about monsoon LPS tracks is based on two sources: LPS tracks during from Mooley and Shukla (1987) who made a comprehensive compilation of LPS based on daily weather reports and annual summary of storms and depressions published by IMD and LPS tracks during the remaining period , the information about monsoon LPS are obtained from Sikka (2006) and also from the IMD daily weather reports (IDWRs). In addition, for the analysis of past PIOD cases, the (1 1 ) gridded rainfall dataset from IMD (Rajeevan et al., 2006) which is available for the period has been utilized.

3 1336 R. KRISHNAN et al. 3. Monsoon LPS activity and large-scale conditions during The monsoon depressions of 2006 The 2006 summer monsoon season witnessed as many as seven lows and eight depressions as defined by the IMD (Jayanthi et al., 2006). The IMD classification is based on the intensity of vortex, around the central region of low pressure, measured by the strength of surface winds. LPS systems with wind speeds up to 8.5 ms 1 are referred to as lows, while LPS with winds ranging between ms 1, ms 1 and ms 1 are referred to as depressions, deep-depressions and cyclonic storms, respectively (Das, 1968). Vorticity budget studies suggest that the westward movement of LPS involves vorticity generation in the western sector by the divergence term (Daggupaty and Sikka, 1977; Rajamani and Sikdar, 1989). Figure 1 shows the tracks of four westward moving LPS, which include lows, depressions and deep-depressions, during 2006 each lasted for more than a week. In each of the above four cases, the LPS remained as either a deep-depression or a depression for at least three consecutive days (Jayanthi et al., 2006). The example in Figure 2 (i) illustrates the evolution of the TMI rainfall and NCEP reanalysis 850 hpa streamlines associated with the LPS during 31 July August The important point is the co-evolution of the synoptic-scale circulation approximately 2000 km, and the mesoscale precipitation extending over a few hundred kilometers. Note that the maximum rainfall is mostly located in the southwest sector of the cyclonic circulation, which corresponds to the area of maximum lowlevel convergence and vertical motion, and is consistent with earlier studies (Pisharoty and Asnani, 1957; Mooley, 1973; Rajamani and Rao, 1981; Stano et al., 2002). The Figure 1. Tracks of the 4 westward moving LPS during 2006 are shown [(31 Jul 11 Aug); (30 Jun 06 Jul); (15 22 Aug); (27 Aug 03 Sep)]. The LPS tracks include lows, depressions and deep-depressions as defined by the IMD and obtained from IMD Daily Weather Report (IDWR). In each of the above four cases, the LPS remained either as a deep-depression or depression for at least 3 continuous days (Jayanthi et al., 2006). This figure is available in colour online at wileyonlinelibrary.com/journal/joc enhanced rainfall over the west-coast (Figure 2(e) (h)) is known to include contribution from orographic precipitation (Sarker, 1967). Figure 3 shows the rainfall, averaged for the period 31 July August 2006 and the vertically integrated moisture transport vector. Several studies have emphasized the importance of crossequatorial moisture transport from the southern tropical Indian Ocean on the monsoon precipitation over the subcontinent (Saha and Bavadekar, 1973, Krishnan, 1998, Ramesh Kumar et al., 1999). Figure 3 shows the largescale moisture transport into the subcontinent by the cross-equatorial monsoonal winds; together with a welldefined cyclonic circulation over the monsoon-trough region. An east west oriented rain band can also be seen in Figure 3, extending eastward from the west-coast of India into Burma and Southeast Asia Large-scale anomalies in the Indian Ocean region during the 2006 summer monsoon The anomalous large-scale background conditions during the 2006 summer monsoon can be inferred from the maps of NOAA OISST, and NCEP reanalysis winds, rainfall and moisture-transport anomalies shown in Figure 4. One of the striking features is the pattern of cold (warm) SST anomalies in the SETIO (WIO) associated with the PIOD (Figure 4). The SST cooling in the east is accompanied by equatorial easterly anomalies and south-easterly wind anomalies off Sumatra which promote upwelling in the eastern IO and are known to enhance the evaporative fluxes from the SETIO (Behera et al., 1999; Yamagata et al., 2004). Over the WIO, the near-equatorial wind anomalies show convergence around 50 E (Figure 4) and intensified cross-equatorial flow over the centraleastern Arabian Sea and the Bay of Bengal. The spatial map of rainfall anomaly during JJAS 2006 (Figure 4) shows below-normal precipitation over the SETIO indicating suppressed convection, while positive rainfall anomalies can be seen over WIO and the plains of central-north India. The anomalous moisturetransport vector (Figure 4) shows enhanced westerly transport over the Arabian Sea, Indian landmass and the Bay of Bengal. Following Behera et al. (1999), the moisture-transport vector was decomposed into rotational and divergent components (Figure 4(c) (d)). The rotational component clearly shows the dominance of the cross-equatorial moisture transport, while the divergent component shows anomalous outflow of moisture from the SETIO region, which essentially acts like a moisture-source during PIOD events and feeds moisture into the monsoon-trough region (Behera et al., 1999). 4. Monsoon LPS activity during past PIOD events If the large-scale circulation anomalies during PIOD episodes can influence the monsoon LPS activity, the same should be discernible from past records as well. Keeping this in view, the monsoon LPS tracks and

4 MONSOON DEPRESSIONS OF 2006 AND INDIAN OCEAN DIPOLE 1337 (c) (d) (e) (f) (g) (h) (i) Figure 2. (a i) Example illustrating the evolution of 850 hpa streamlines using NCEP/NCAR reanalysis winds and Tropical Rainfall Measurement Mission (TRMM) Microwave Imager (TMI) rainfall (mm h 1 ) associated with the LPS during 31 July August This LPS was seen as low in the Bay of Bengal on 31 July, which later intensified into a depression over the northwest Bay and moved in the WNW direction. This figure is available in colour online at wileyonlinelibrary.com/journal/joc Figure 3. TMI rainfall (mm hr 1 ) averaged during the period (31 Jul 08 Aug) The LPS track is shown by the thick black line. The vectors represent the vertically averaged water-vapor transport vector Q (kg m 1 s 1 )definedasq = g 1 Pu P s qvdp, where g is the acceleration due to gravity; q is the specific humidity; P s = surface pressure and P u is 300 hpa.. V is the horizontal wind vector on pressure levels. The transport vector is calculated using NCEP/NCAR reanalysis data. This figure is available in colour online at wileyonlinelibrary.com/journal/joc circulation features during PIOD and non-piod periods have been examined. For this purpose, equal cases of PIOD and non-piod years for the period have been selected objectively as follows: 1. The selection of PIOD cases is based on the Dipole Mode Index (DMI) a normalized index represented by the anomalous SST gradient between the western (10 S 10 N, E) and eastern (10 S Equator, E) Indian Ocean as defined by Yamagata et al. (2004) (source: research/d1/iod). Using the DMI time-series, a PIOD is selected if the DMI for a particular year (see Table I) exceeds 1 SD (σ = 0.375). 2. Additionally it is ensured that a selected PIOD episode has co-occurred during a period of above-normal monsoon rainfall over central-north India where the monsoon LPS activity dominates (Sikka, 2006). This criterion is important because the background largescale monsoon conditions can affect the clustering and frequency of monsoon synoptic disturbances (Rajeevan et al., 2000; Goswami et al., 2003). Keeping this in view, only those PIOD episodes which have cooccurred with monsoon seasons of above-normal (at least > +5% of normal) precipitation over centralnorth India (73 E 90 E; 20 N 30 N) computed from the IMD gridded rainfall dataset (Rajeevan et al., 2006) have been selected. 3. The above-mentioned objective criteria led to the identification of five PIOD cases (1961, 1994, 1997, 2006 and 2007). To have a bigger sample size, two additional PIOD events have been included, i.e (DMI approximately 0.36) and 1983 (DMI

5 1338 R. KRISHNAN et al. (c) 2 50 (d) Figure 4. Anomaly maps for JJAS-2006 sea surface temperature (SST) ( C) and surface-wind (ms 1 ) Rainfall (mm day 1 ) and vertically integrated moisture transport vector Q (kg 1 s 1 ) (c) Rotational component (d) Divergent component of moisture transport (kg m 1 s 1 ). The contours in (d) represent the potential function (Behera et al., 1999). The winds and moisture are from NCEP/NCAR reanalysis; rainfall from CMAP dataset and the SST anomalies are from NOAA OI SST V2. This figure is available in colour online at wileyonlinelibrary.com/journal/joc approximately 0.33) although the DMI for these 2 years cases were slightly short of 1σ. Table I gives the values of the DMI and rainfall anomaly over central-north India for the period It can be noted that the seasonal monsoon rainfall over centralnorth India was relatively higher during most of the non-piod cases as compared to the PIOD cases. Although the reasons for the precipitation differences in the two cases are not fully clear, it has been verified from mean sea-level pressure maps that the monsoon trough was relatively stronger in the non- PIOD composite as compared to the PIOD composite (figures not shown). 4. For the non-piod cases, all the positive IOD events during were first excluded. Out of the remaining years, the seven top above-normal monsoon rainy seasons having excess (> +5% of normal) precipitation over central-north India have been selected. The seven non-piod cases are 1958, 1970, 1971, 1973, 1975, 1980 and It may be mentioned that 1958 and 1980 coincided with negative IOD events. Although some instances of negative IOD events have been accompanied by monsoon droughts over India (Krishnan et al., 2006), the criterion of (> +5% normal) monsoon precipitation over central-north India eliminates such instances from our analyses. 5. Basically, the above objective procedure allowed selection of seven PIOD and seven non-piod cases. The idea behind ensuring above-normal monsoon rains over the plains of central-north India in both these categories is to enable a fair comparison of monsoon disturbances in the two cases with and without positive IOD background conditions. The LPS dates for the two categories are given in Tables II and III, respectively Monsoon LPS tracks The monsoon LPS tracks for the PIOD and non-piod cases are shown in Figure 5 and (c), respectively. It must be mentioned that the tracks in Figure 5 include all those cases in which the LPS lasted for 3 days or more. The idea of including the shorter duration LPS (i.e. 3 and 4 days) in this analysis is to understand their contributions to the total LPS activity during PIOD and non-piod years. Furthermore, the reason for not including very short-lived LPS (i.e. 1 day and 2 days) is based on the understanding that the interactions between the very short-lived LPS and the monsoon large-scale circulation is expected to be minimal; and therefore not likely to make tangible differences to the present results. Note that the tracks in both the cases are concentrated around the monsoon trough. Tables II and III show that the total number of LPS in the PIOD years is approximately 86 and relatively higher as compared to the non-piod years of approximately 75. Also, the number of LPS with duration (>5 days) is significantly higher in the PIOD (approximately 42) category, as compared to the non-piod (approximately 29) case. In other

6 MONSOON DEPRESSIONS OF 2006 AND INDIAN OCEAN DIPOLE 1339 Table I. Time-series of Dipole Mode Index (DMI) and rainfall departure from normal (%) over north-central India (73 E 90 E; 20 N 30 N) for the period ( ). The DMI is a normalized anomalous sea surface temperature (SST) gradient between the western (10 S 10 N, E) and eastern (10 S Equator, E) as defined by Yamagata et al. (2004) (Ref: The markings in columns 5 6 correspond to the positive Indian Ocean Dipole (PIOD) and non-piod cases selected for the present analysis. Year DMI North-central India rainfall anomaly (%) PIOD Non- PIOD words, the percentage contribution of long-lived monsoon LPS to the total is about 50% during PIOD years and about 38% during non-piod years. It has also been confirmed that most of the negative IOD events, with the exception of 1980, were associated with significantly fewer cases of long-lived monsoon LPS as compared to PIOD events. From the LPS tracks described above, additional information has been deduced by computing LPS density values on grid boxes by counting the number of LPS passing through any particular grid box. Maps of LPS density for the PIOD and non-piod cases are shown in Figure 5 and (d), respectively. It can be observed that the LPS density near the Bay of Bengal and vicinity is relatively higher in the PIOD case as compared to the non-piod. Furthermore, note that the LPS densities extend more westward into northwest India in the PIOD case as compared to the non-piod Background large-scale conditions The large-scale ocean atmosphere conditions during the PIOD and non-piod periods show interesting differences as evidenced from the JJAS anomaly composites in Figure 6 based on NCEP reanalysis products and HadISST dataset. The PIOD composite depicts anomalous SST cooling in the SETIO and warm anomalies in the WIO, together with strong equatorial easterly and south-easterly wind anomalies off Sumatra-Java. An intensified monsoon cross-equatorial flow can be seen over the Bay of Bengal and eastern Arabian Sea in Figure 6. In contrast, the SST cooling in the non-piod composite is mostly seen in the Arabian Sea with maximum cooling off the Somali and Arabian coasts; and enhanced cross-equatorial flow over the west-central Arabian Sea (Figure 6(c)). The rainfall anomalies in the non- PIOD case (Figure 6(d)) show increased rainfall over the Indian region. In the PIOD composite, enhanced precipitation can be noticed along the monsoon-trough extending from northwest India into northern Bay of Bengal; along with the near-equatorial east west pattern of positive rainfall anomalies over the WIO and negative anomalies over the SETIO and eastern Indian Ocean (Figure 6). Also note that the intensified cross-equatorial moisture transport across the Bay of Bengal in the PIOD case is accompanied by a north south gradient in the precipitation anomalies around 90 E (Figure 6) Rainfall and circulation anomalies during LPS Composites of rainfall and 850 hpa wind anomalies associated with LPS days for the PIOD and non-piod cases are shown in Figure 7. The rainfall anomalies are computed with respect to daily climatological normals based on the daily data of IMD rainfall (Rajeevan et al., 2006), and the wind anomalies are based on daily data from NCEP reanalysis (Kistler et al., 2001). The LPS days for the PIOD and non-piod cases are given in Tables II and III, respectively. The rainfall increase during LPS periods is largely seen over central

7 1340 R. KRISHNAN et al. Table II. Dates of monsoon low-pressure systems (LPS) during the seven PIOD cases. The total number of LPS cases is 85; and the number of long-lived (>5 days) LPS cases is 42. The LPS dates are obtained from (Mooley and Shukla, 1987; Sikka, 2006 and IMD daily weather reports) June June June June June June June June 28 June 01 July June June June 30 June 06 July 26 June 03 July June 31 July 03 August July June July July July July August July 25 June 01 July 28 July 02 August July July July 28 August 03 September July July August 27 July 01 August August July September August July August August August August September August July 26 August 01 September August August August September August July September August September 31 Aug 05 September September September July September August September September September 28 July 02 August 28 August 03 September September September September August September September August September September August September August September 26 August 01 September September September September September

8 MONSOON DEPRESSIONS OF 2006 AND INDIAN OCEAN DIPOLE 1341 Table III. Same as TableII except for non-piod years. The total number of LPS cases is 75; and the number of long-lived (>5 days) LPS cases is July June 02 04June June June June June July 28 June 4 July June June June June June August July June July 25 June 01 July June June 27 August 4 September July June July July July July September August July July 30 July 01 August July July September August July July August July July September August July August August July September 23 August 2 September August August August August September September August August 31 August 8 September August August September September August 25 August 1 September September August 30 August 7 September August September September September September September September India (Figure 7 and (c)). The rainfall anomaly over the west-coast may include contributions partly from orographic effects due to the Western Ghat mountains (Sarker, 1967); although non-orographic regions like Gujarat can experience significant precipitation increase under the influence of the long-lived LPS particularly during the PIOD years (Figure 11). The 850-hPa wind anomalies show stronger monsoon cross-equatorial flow and stronger cyclonic anomalies over the monsoon-trough region in the PIOD composite (Figure 7) relative to the non-piod composite (Figure 7(d)). The enhanced cyclonic vorticity to the north of 15 N spans a wider longitudinal extent (60 E 120 E) in the PIOD case as compared to the non-piod case (Figure 7(d)). More importantly, the PIOD composite (Figure 7) exhibits stronger equatorial easterly anomalies and south-easterly anomalies over the SETIO, and the enhanced crossequatorial flow gives rise to strong westerly anomalies over the Arabian Sea and the Bay of Bengal. Basically, one can note a very prominent meridional gradient of the low-level wind anomalies in the PIOD composite Evolution of circulation anomalies during LPS Examination of the time-evolution of circulation anomalies starting from the day of LPS formation (i.e. Day 1) is helpful in comprehending the longevity of LPS. Figure 8 shows composite plots of relative vorticity and wind anomalies at 850 hpa on Days 1, 5, 9 and 13 for the PIOD and non-piod cases. The circulation anomalies are with respect to daily climatological normals computed from daily winds for the period In both the PIOD and non-piod cases, cyclonic vorticity anomalies can be noticed over the Bay of Bengal and along the monsoon trough from Day 1 to Day 5. However, in the PIOD case (Figure 8(c)), the positive vorticity anomalies along the monsoon trough appear to be more persistent even beyond Day 9. Furthermore, it is interesting to note that the PIOD composite shows persistent anticyclonic vorticity anomalies on the southern side over the Equator 10 N latitude belt extending across 60 E 120 E. The pattern of cyclonic vorticity on the northern side (15 N 30 N) and anticyclonic vorticity to the south of 10 N in the IOD composites (Figure 8 (d)) is a key point to this discussion. Several studies in the past have examined the stability characteristics of summer monsoon zonal-flows and possible dynamical instabilities associated with monsoon depressions (Keshavamurty et al., 1978; Goswami et al., 1980; Mishra and Salvekar, 1980; Satyan et al., 1980; Dash and Keshavamurty, 1982). In particular, various instability mechanisms (viz., barotropic, baroclinic, combined barotropic baroclinic instability) arising due to horizontal and vertical shears in the zonal-flows have been investigated. A necessary condition for instability of sheared flows in a channel is that the gradient of potential vorticity (PV) of the flow field must vanish somewhere in the channel. Earlier studies have observed that the meridional gradient of PV over the summer monsoon region vanishes at almost all

9 1342 R. KRISHNAN et al. PIOD (c) Non-PIOD (d) Figure 5. (a, c) LPS tracks (b, d) density maps of LPS. The left and right columns are for the PIOD and non-piod cases respectively. LPS density is computed on grid boxes by counting the number of LPS passing through a given grid box. The mean track of the LPS is shown by thick black line. The 7 PIOD cases are (1961, 1967, 1983, 1994, 1997, 2006, and 2007) and the 7 non-piod cases are (1958, 1970, 1971, 1973, 1975, 1980 and 1990). The track positions are taken from Mooley and Shukla (1987), Sikka (2006) and IDWR. This figure is available in colour online at wileyonlinelibrary.com/journal/joc levels from lower-troposphere to 250 hpa around 23 N latitude in the region of the monsoon-trough, so that the meridional gradient of the zonal-wind is a measure of the barotropic instability (Keshavamurty et al., 1978; Keshavamurty and Sankar Rao, 1992). By examining the latitudinal variation of wind anomalies in the PIOD and non-piod cases, one can infer about the barotropic instability of zonal-flows in the two cases. Figure 9 shows the Hovmoller (y-t) section of 850 hpa relative vorticity and zonal-wind anomaly composites averaged over the Indian longitudes (70 E 95 E) for the PIOD and non-piod cases. The time-axis in Figure 9 is from the starting day (Day 1) of the LPS and extends till Day 15. An important point to be noted is the cyclonic vorticity (positive) anomaly near 20 N (Figure 9 (c)), which is flanked by negative anomalies (anticyclonic vorticity) on the northern and southern sides. The change of sign of relative vorticity on either side near the 20 N latitude indicates the latitudinal gradient of relative vorticity and the zero vorticity line can be noticed near 25 N which is a necessary condition for barotropic instability of monsoon LPS (Keshavamurty et al., 1978). Basically, the latitudinal gradient of relative vorticity is largely related to the meridional shear of the zonal-wind and the same can be confirmed from the Havmoller sections of zonal-wind anomalies shown in Figure 9 and (d)). Note that the meridional shear of the zonal-wind around the 21 N latitude is clearly evident from the westerly anomalies between 10 N and 20 N and the easterly anomalies between 22 N and 30 N. It is further interesting to note that the meridional shear of the zonal-wind anomalies is greatly enhanced in the PIOD case due to the presence of strong easterly anomalies between 10 S and Equator, which are absent in the non-piod case. In fact, the strong anticyclonic anomalies to the south of 10 N in the PIOD case (Figure 9) can be seen extending beyond Day 10. The strong and sustained meridional shear of the zonal-wind anomalies in the PIOD case (Figure 9) relative to the non-piod case (Figure 9(d)) indicates that the barotropic instability of the large-scale monsoon flow is significantly enriched in the former, and therefore aids in the longer sustenance of LPS in the PIOD case as compared to the non-piod. At this point, it must be mentioned that while background conditions during PIOD events are generally more conducive for the monsoon LPS activity, it is realized that the genesis and life-period of monsoon LPS can be influenced by several other internal mechanisms as well. For example, although the 2008 summer monsoon was accompanied by a positive PIOD, the LPS activity in 2008 was not as vigorous

10 MONSOON DEPRESSIONS OF 2006 AND INDIAN OCEAN DIPOLE 1343 PIOD (c) Non-PIOD 1 1 (d) Figure 6. Anomaly composites for the JJAS season during PIOD (left column) and non-piod (right column) years (a, c) SST ( C) and surface-wind (ms 1 ) (b, d) rainfall (mm day 1 ) and vertically integrated moisture transport vector Q (kg m 1 s 1 ). The winds, humidity and rainfall are from NCEP/NCAR reanalysis; and SST is from the HadISST dataset. This figure is available in colour online at wileyonlinelibrary.com/journal/joc and long-lived as compared to those of 2006 and 2007 [During the 2008 summer monsoon, there were about 10 cases of monsoon LPS. Only three instances of the LPS during 2008 lasted beyond 5 days (i.e June, 7 12 July, September). Furthermore, only a very few LPS cases during 2008 intensified into either a depression or a deep-depression, while most of them remained as low or well-marked low. This information is based on the daily weather reports from the IMD. Another important aspect of the 2008 summer monsoon was the relatively low rainfall over central India and the weak monsoonal low-level flow. In fact, a ridge-like feature was observed around the monsoon-trough region (approximately 20 N) over the Indian longitudes (approximately 75 E) during JJAS 2008). In the following section, the discussion will focus on understanding the internal feedback processes between latent heating and dynamics that are crucial for the sustenance of the monsoon LPS activity. 5. Elevated latent heating, vortex stretching and large-scale monsoon Hadley circulation From the above discussions, it is seen that PIOD events are characterized by two important large-scale features which are conducive for the enhanced monsoon LPS activity: strong and persistent cross-equatorial moisture transport from the SETIO into the Bay of Bengal and the enrichment of barotropic instability of the monsoon zonal-flow, which essentially arises from an increase in the meridional shear of the mean zonal-wind. Although sheared instability mechanisms (barotropic, baroclinic or combined barotropic baroclinic) are necessary for generating instability of sheared flows, they are not sufficient to initiate growth of monsoon disturbances (Keshavamurty and Sankar Rao, 1992). The monsoon troposphere is known to be conditionally unstable with the presence of cyclonic vorticity in the lower levels, which allows development of convective instability and growth of monsoon disturbances. Thus in addition to the large-scale background conditions, it is necessary to understand internal mechanisms, involving latent heating and dynamics that can sustain the monsoon LPS. For this purpose, a detailed analysis has been performed using latent-heating profiles from TMI-V62A-12, together with precipitation from the TRMM radar which is used to infer about the rain type Vertical structure of relative vorticity There were totally 14 cases of depression and deepdepression days (Figure 10) associated with the four westward moving LPS of 2006 shown in Figure 1. Based on these 14 cases, the vertical profiles of latent heating and relative vorticity associated with the monsoon depressions are examined. The vertical profiles of relative vorticity (ζ ) averaged around the depression region are shown in Figure 10 for the 14 cases. The profiles in Figure 10 are daily profiles, corresponding to the 14 depression and deep-depression days, from each

11 1344 R. KRISHNAN et al. PIOD (c) Non-PIOD (d) Figure 7. Anomaly composites based on LPS days during PIOD (left column) and non-piod (right column) cases (a, c) IMD Rainfall (mm day 1 ) (b, d) NCEP/NCAR 850 hpa winds (ms 1 ). The LPS dates for the two cases are given in Tables II and III. In preparing the composite maps, the anomalies are first computed relative to the daily climatological mean and then the composites are constructed. This figure is available in colour online at wileyonlinelibrary.com/journal/joc of the four LPS. The vertical profiles in Figure 10 are constructed by averaging ζ over (10 longitude 8 latitude) grids with respect to the depression centres. The mean of the 14 profiles is shown by the black solid line (Figure 10). In all the cases, the cyclonic vorticity (positive) extends vertically up to 300 hpa, and even to 250 hpa on several occasions. Figure 10 shows the vertical profiles of ζ corresponding to the climatological mean for a particular day when the 2006 depression occurred, with the averaging area being same as in Figure 10. In Figure 10, the climatological profile for a particular day is computed using the daily climatological mean values obtained from NCEP reanalysis daily data for the period Comparison of Figure 10 with Figure 10 allows interpretation of the vertical structure of ζ during depressions relative to climatology. It can be seen that the cyclonic vorticity, from surface to 300 hpa, shows a two- to threefold increase during depression days as compared to the climatology. Also note that the mean climatological cyclonic vorticity (Figure 10) does not extend above 300 hpa, while in the case of the 2006 depressions the vorticity field is stretched vertically above the 300 hpa level Vertical structure of latent heating The vertical profiles of latent-heating rate from TMI (Figure 11) show the occurrence of elevated heating associated with the monsoon depressions. It can be seen that the latent heating dominates mostly above 600 hpa (approximately 4 km), with the maximum being located near 400 hpa level (approximately 8 km above surface). Although the magnitude of heating varies among the 14 profiles (Figure 11), the mean profile (black line) shows a maximum heating rate of about 1.25 K h 1 (i.e. approximately 30 K day 1 ). In the lower levels, the mean profile shows cooling as much as 0.75 K h 1. The important point in Figure 11 is the intensified vertical gradient of heating (cooling in lower levels and heating above 600 hpa) occurring in monsoon depressions. Here, it must be pointed out that, on the one hand, the structure of latent-heating profiles (Figure 11) for the

12 MONSOON DEPRESSIONS OF 2006 AND INDIAN OCEAN DIPOLE 1345 PIOD Non-PIOD Day 1 (e) Day 1 (f) Day 5 Day 5 (c) (g) Day 9 Day 9 (d) (h) Day 13 Day 13 5 Figure 8. Time-sequence of the anomaly composites of NCEP/NCAR reanalysis relative vorticity ( 10 6 s 1 ) and 850 hpa winds (ms 1 ) associated with the monsoon LPS during PIOD (left column, and non-piod (right column) cases. The anomaly composites are shown for Days 1, 5, 9 and 13. Here, Day 1 corresponds to the initial day of the LPS (see Tables II and 3). In preparing the composites, we have considered only those LPS cases which lasted for 5 days or more. This figure is available in colour online at wileyonlinelibrary.com/journal/joc 2006 LPS closely resembles the heating associated with stratiform-type precipitation in which evaporation and mesoscale downdrafts dominate in the lower levels, whereas condensation occurs in the mesoscale updraft aloft (Houze, 1982, 2004; Johnson, 1984; Mapes, 1993; Schumacher et al., 2003, 2004). On the other hand, the convective-type heating is distributed through the full depth of the troposphere with maximum amplitude around 500 hpa. This issue will be discussed in greater detail in Section 5.4. The effect of latent heating can be seen in the temperature response around the depression region. Figure 11 shows the vertical profiles of temperature difference between the depression region and the large-scale environment for the 14 depression cases. The mean profile of temperature difference from the 14 cases is shown by the black line. It can be seen that the temperature response is consistent with the latent-heating vertical profiles, showing a cooling of about 1.5 C below 600 hpa and a warm response of similar magnitude (approximately +1.5 C) aloft, which corroborates earlier studies (Keshavamurty, 1972; Krishnamurti et al., 1975; Godbole, 1977; Keshavamurty et al., 1978). It has been suggested that the low-level cooling is associated with significant evaporation in the rain area on the western sector of depression (Keshavamurty, 1972). The

13 1346 R. KRISHNAN et al. PIOD (c) Non-PIOD Time (days) Time (days) (d) Figure 9. Hovmoller (latitude time section) plot of relative vorticity and zonal-wind anomaly composites starting from Day 1 of the LPS for the PIOD (left column) and non-piod (right column) cases (a, c) Relative vorticity ( 10 6 s 1 ) (b, d) 850 hpa zonal-wind (ms 1 ). The anomalies are zonally averaged between 70 E and 95 E. Data used is from NCEP/NCAR reanalysis. This figure is available in colour online at wileyonlinelibrary.com/journal/joc lower-tropospheric cooling in the field of monsoon depressions was also noted from aircraft measurements of wind and temperature during MONEX-1979 by Joseph and Chakravorty (1980). They found the existence of large temperature gradients (approximately 2 C per degree latitude) over the depression field at 850 and 700 hpa Association between vortex stretching and latent heating It was earlier seen from the vertical profiles of vorticity (Figure 10) that the monsoon LPS were associated with vorticity stretching in the vertical direction. The vorticity-stretching term [ (ζ + f).v], which involves the wind divergence, is the dominant term for vorticity generation during monsoon depressions (Rajamani and Sikdar, 1989). In the following discussions, it will be shown that the vorticity-stretching field is closely related to latent heating associated with the monsoon depressions. The contour plot in Figure 12 is the composite of the vertically integrated vorticity-stretching term for the monsoon depression days during The shading in Figure 12 is the composite of latent heating at the level of maximum heating around 400 hpa. Since the depression centres vary from day to day and from case to case, the stretching term and latent heating have been mapped with respect to a common depression centre with co-ordinates (0, 0). It can be seen that the maximum vorticity stretching is located to the west of the depression centre (0, 0). This is consistent with the study of Keshavamurty (1972) which reported a slight westward tilt of the cyclonic centre with height in the longitude-height plane. Furthermore, it can be noticed that the maximum latent heating is located to the south-west of the depression centre (0, 0). An examination of the time-sequence of vortex-stretching and latent-heating terms showed that the two fields co-evolved as the depression traversed in the WNW direction (figures not shown) Precipitation-type The separation of precipitation into convective and stratiform components is possible from PR measurements

14 MONSOON DEPRESSIONS OF 2006 AND INDIAN OCEAN DIPOLE 1347 Figure 10. Vertical profiles of relative vorticity ( s 1 ) averaged over (10 8 ) latitude - longitude domain w.r.t the depression centers Same as except for daily climatological mean. The profiles in are shown for 14 depressions and deep-depressions days during These 14 days are included in the 4 LPS tracks (see Fig. 1) and different marks are used to differentiate the 4 LPS tracks [Square (02 05 Aug); Circle (02 04 Jul); Triangle (16 18 Aug); Diamond (29 Aug 01 Sep)]. The black solid line is the mean of the 14 profiles. The climatological mean is considered for the period Data used is from NCEP/NCAR reanalysis. This figure is available in colour online at wileyonlinelibrary.com/journal/joc Figure 11. same as Fig.10 except for TMI latent-heating (K hr 1 ) averaged over (4 4 ) domain w.r.t rainfall centers during the 14 cases of depression and deep-depression days of The TMI rainfall data is used to identify the rainfall center. The TMI latent-heating profiles which were originally available on z-levels have been converted to pressure levels Vertical profiles of temperature difference (K) between the depression region and the environment. The environmental temperature is obtained by averaging over a large domain (64 E 98 E; 5 N 32 N). The temperature around a depression is computed by averaging over a (10 8 ) domain w.r.t depression centers (see Fig. 1) using NCEP/NCAR reanalysis data. This figure is available in colour online at wileyonlinelibrary.com/journal/joc as they incorporate information on the vertical crosssections of hydrometeors. Basically, convective precipitation is characterized by large horizontal reflectivity gradients, strong vertical motions and high rainfall rate. Smaller horizontal reflectivity gradients, weaker vertical motions and lower rain rates on average characterize stratiform precipitation (Zafar and Chandrasekar, 2004). The presence of a bright-band typically indicates stratiform region. The bright-band is considered as a zone in the upper levels, which separates liquid water (below) and frozen water (above). The TRMM- PR rain classification (2A23) algorithm uses two different methods for classifying rain type: one is vertical profile method (V-method) and the other is horizontal pattern method (H-method). Both methods classify rain into three categories: stratiform, convective and (c) other (Steiner et al., 1995; Awaka et al., 1998). Using the PR 2A-23 data, the rain types in the region of monsoon depressions during 2006 have been examined. The left column in Figure 13 shows examples of total rain (convective + stratiform + others) when the PR passes were located well within the monsoon depression region. The right column in Figure 13 shows the corresponding stratiform rain for those cases. Using the total and stratiform rain rates, the stratiform rain fraction (SF) was determined and the area-averaged SF was computed over (5 longitude 5 latitude) grid boxes around the rainfall centres. Totally, there were