Characteristics of wet and dry spells during southwest monsoon season over southeast India a diagnostic study

Characteristics of wet and dry spells during southwest monsoon season over southeast India a diagnostic study Thesis submitted to Andhra University Visakhapatnam, Andhra Pradesh, India for the award of the degree in Doctor of Philosophy in Atmospheric Science T. Mohan Satyanarayana National Atmospheric Research Laboratory Gadanki -517 112 A.P., India. October 2011

Declaration I declare that the research work presented in this thesis has been carried out by me under the joint supervision of Dr. T. Narayana Rao, Scientist, National Atmospheric Research Laboratory (NARL), Gadanki and Prof. O. S. R. U. Bhanukumar, Andhra University, Visakhapatnam. I further declare that this work is original and has not been formed the basis for the award of any Degree, Diploma, Associateship, Fellowship or other similar title in any other Institution or University. (T. Mohan Satyanarayana) Certified by Dr. T. Narayana Rao Prof. O. S. R. U Bhanukumar Dept. of Meteorology and Oceanography. Andhra University Visakhapatnam 530 003

Dr. T. Narayana Rao Certificate This is to certify that the thesis entitled Characteristics of wet and dry spells during southwest monsoon season over southeast India a diagnostic study. submitted by Mr. T Mohan Satyanarayana, is a record of research work done by the candidate under my supervision during the period of his stay. I further declare that this work is original and has not been formed the basis for the award of any Degree, Diploma, Associateship or any other similar title in any University or Institution. Date: (T. Narayana Rao)

Andhra University Visakhapatnam 530 003, India Prof. O.S.R.U Bhanukumar Dept. Meteorology and Oceanography Fax: Phone: Certificate This is to certify that the thesis entitled Characteristics of wet and dry spells during southwest monsoon season over southeast India a diagnostic study. submitted by Mr. T Mohan Satyanarayana, is a record of research work done by the candidate under my joint supervision during the period of his stay. I further declare that this work is original and has not been formed the basis for the award of any Degree, Diploma, Associateship or any other similar title in any University or Institution. Date: (Prof O.S.R.U. Bhanukumar)

Acknowledgements A considerable effort has been taken to bring this thesis to the current form. During the journey, many people have helped with support, advice, and ideas, both directly and indirectly. I take this opportunity to express my sincere gratitude to the people who have been involved in bringing out this work to the current level. It is difficult to overstate my gratitude to my Ph.D. supervisor, Scientist, NARL. With his enthusiasm, his inspiration, and his great efforts to explain things clearly and simple helped me in understanding the subject very elaborately. I am very much fortunate to have him as my guide. He is so unruffled and his attitude and dedication made me to work hard. Without his continuous support and patience, it would have been extremely difficult to complete this work. Throughout my thesis-writing period, he provided encouragement, sound advice, good teaching, good company, and lots of good ideas. I am extremely thankful to him for sparing his valuable time and going through my thesis and making useful corrections and suggestions. I would have been lost without him. First of all I would like to express my sincere thanks to, Andhra University, Visakhapatnam, who has been my supervisor since the beginning of my study. He devoted his precious time and provided many helpful suggestions during the course of this work, which indeed helped me in improving the thesis. Without his constant support, readiness to provide the required facilities and tolerance, it would have been extremely difficult to finish this work. I would like to thank, Director, NARL, for permitting me to utilize the resources at the center to the maximum extent throughout the course of my work. I would like to acknowledge the advice and encouragement of Scientist, NARL, during my thesis work. Occasional discussions with him have helped me in gain knowledge. I would like to thank the valuable technical support and help provided by, and Mr. I am extremely happy to express my thanks to,, for their kind cooperation., and all other engineers, scientists and technical staff of NARL I would also like to express my deep sense of gratitude to Head, Department of Meteorology and Oceanography, Andhra University, i

Acknowledgements Visakhapatnam. I express my deep sense of gratitude and profound thanks to the best teachers I have ever had and, for their encouragement and support. Thanks to one and all for pushing me. In my office campus at NARL, I was surrounded by knowledgeable and friendly people who helped me daily. My sincere acknowledgements to each and every person in the Administrative, Account, Electrical, AC plant, Canteen and Security personals for all the help and cooperation. On a more personal note and whole heartedly I would like to show appreciation to for her ever-present support, help and encouragement in hard times. She often provided me with critical remarks, discussions, questions, suggestions which have helped me in improving in many ways. I am indebted to my many colleagues for providing a stimulating and fun environment in which to learn and grow. I am especially grateful to my friends (my roommate and colleague), I wish to thank my friends, colleagues, my post-graduate classmates at Andhra University, for helping me get through the difficult times, and for all the emotional support, camaraderie, entertainment, and caring they provided. I am very much thankful to the office staff of Department of Meteorology and Oceanography for their kind support. Acknowledgements to my close friends, which are too many to mention, always stood by my side asking over and over again When will you get it done? Next week? Next Month? When? Special thanks to my sweet little sister, Scientific Officer, IGCAR, Kalpakkam. Occasional visits to her place were a great and welcome relief from the hectic research work. She gave me unconditional love and support throughout my life and has always encouraged me in good and bad times. Lastly, and most importantly, I wish to thank my parents,. Without their support and love this thesis could not be done. and ii

Preface Preface indian Summer Monsoon Rainfall (ISMR) exhibits pronounced intraseasonal variability (ISV) on time scales ranging from; 3-7 days (super synoptic oscillations) to 10-20 days and to 30-60 days (Madden and Julian Oscillations - MJO). Fluctuations in sub-seasonal rainfall are mainly characterized by active and break spells of the Indian summer monsoon. These epochs are an important component of the rainfall variability and has a large impact on agricultural production and hence the economy of the country. The active and break spells exhibit large spatial variability. In particular, the spells over monsoon zone (north and central India) are often in opposite phase with those observed over southeast peninsular India. While active and break spells over the monsoon zone in relation to ISMR are studied by several researchers, characteristics of spells over the southeast peninsular India are not well documented. Further, several interesting observations are made, recently, on draft cores and their vertical structure over Gadanki (13.5 N 79.2 E), a rural station in the southeast peninsular India. But the causative mechanisms for the observed differences are not spelled out in those studies. The present study, therefore, documents intriguing differences in the thermal and dynamical characteristics and energetics of the atmosphere between contrasting phases of the monsoon over southeast peninsular India. Also, the variations in rainfall characteristics in different spells of the monsoon are also studied. The observed differences between spells are used to understand the occurrence of draft cores and their vertical structure in different spells of the monsoon. An extensive overview on intraseasonal oscillations (ISO), their characteristics, causative mechanisms, predictability and spatial variability is given in chapter 1. A detailed discussion on different techniques for the identification of active and break spells is also included. Gap areas in our knowledge on the spatial variability of ISO are identified. The scope of the thesis, i.e., how the present study fills this void?, is also presented. A detailed description of observations obtained from different remote sensing instruments, both ground-based and space-borne, used in the present study, their accuracy iii

Preface and important specifications of the systems are given in chapter 2. The methodology for the identification of active (wet) and break (dry) spells over southeastern peninsular India is also presented in this chapter. Differences in thermal characteristics (temperature, humidity, stability, CAPE etc) of the atmosphere during wet and dry spells of the Indian summer monsoon over Gadanki are studied in chapter 3. Some of the unresolved issues related to the draft core statistics and their vertical structure in different phases of the monsoon are investigated, using vertical profiles of vertical velocity and buoyancy obtained from Indian Mesosphere Stratosphere Troposphere Radar (IMSTR) and radiosonde, respectively. Chapter 3 documented important differences in the thermal and stability parameters between wet and dry spells over Gadanki. To examine whether these differences are localized or seen over a large region (say southeast peninsular India), data from different models are first compared with observations. Several indices are used for the comparison in chapter 4. The model data are then used to extend the study region to better understand the variation of CAPE between wet and dry spells. The possible physical/dynamical mechanisms for the observed differences in CAPE between spells and their affect on convection are also investigated. Since the thermal and stability parameters are different in different spells, their forcing on the atmosphere will also be different. To understand how the forcing in different spells affect the wind structure and its diurnal variation, long-term observations of winds obtained with several ground-based remote sensing and in-situ instruments at Gadanki are utilized. Chapter 5, therefore, discusses differences in the vertical structure of the mean wind and its diurnal variation between different spells of the monsoon. A special emphasis is put on the low level jet and tropical easterly jet, the two most conspicuous components of the Indian summer monsoon. The differences in the thermal, dynamical and energetics of the atmosphere between wet and dry spells will certainly affect rainfall patterns and their intensity. In chapter 6, how the above parameters are affecting rainfall and the growth of draft cores iv

Preface in different spells of the monsoon are discussed. Also, several issues related to the rainfall are addressed in this chapter by making use of satellite rainfall observations. For example, how much seasonal rainfall occurs in each spell of the monsoon? What kind of rainfall occurs in different spells of the monsoon? How different types of rain vary diurnally in different spells of the monsoon? The most important results obtained on the thermal, dynamical and energetics of the atmosphere and rainfall characteristics in wet and dry spells are summarized in chapter 7. Though the present study addressed several intriguing issues, it also opened up new scientific problems. They are briefed in this chapter as future work. v

Contents Acknowledgements Preface Contents List of Figures List of Tables Page 1. Introduction. 1 1.1 Indian monsoon 1.1.1 Intraseasonal Oscillations 1.1.2 Characteristics of ISO 1.1.3 Causative mechanism of ISO 1.1.4 Predictability of ISO 1.2 Active and Break spells 1.2.1 Identification of active and break spells 1.2.2 Spatial variability of active and break spells 1.3 Scope of the thesis 2. Data analysis and Instrumentation. 21 2.1 Introduction 2.2 Surface observations 2.2.1 Automatic Weather Station 2.2.2 Optical rain gauge 2.3 Upper air observations 2.3.1 Radiosonde 2.4 Radar 2.4.1 Scattering mechanisms 2.4.1.1 Bragg scattering 2.4.1.2 Rayleigh scattering 2.4.2 SODAR 2.4.3 LAWP 2.4.4 Indian MST radar 2.4.5 Wind vector computations 2.5 Satellite observations 2.5.1 TRMM 2.5.1.1 2A25 2.5.1.2 3B42 2.6 Model data sets 2.6.1 NCEP/NCAR reanalysis data 2.6.2 ECMWF Interim data 2.6.3 WRF model data 2.7 IMD high resolution daily gridded rainfall data. 2.8 Identification of wet and dry spells vi

Contents 3. Thermal structure during wet and dry spells 46 3.1 Introduction. 3.2 Data used and meteorological background 3.3 Results and discussion 3.3.1 Variation of temperature and humidity on the surface and aloft. 3.3.2 Variation of significant levels 3.3.3 Variation of stability and instability indices 3.4 Discussion 3.5 Summary and Conclusions 4. Energitics during wet and dry spells 67 4.1 Introduction 4.2 Data base 4.3 Comparison of CAPE estimated by radiosonde and models 4.3.1 Time series and statistical comparisons 4.3.2 Comparison of models sensitivity to lifting parcel properties 4.3.3 Comparison of Intraseasonal variability and vertical structure of buoyancy and CAPE 4.4 Variation of CAPE and buoyancy between dry and wet spells over southeast India and associated implications 4.5 Summary and Conclusions 5. Dynamics during wet and dry spells 86 5.1 Introduction 5.2 Data and instrumentation 5.3 Results and discussion 5.3.1 Differences in mean wind between wet and dry spells at the surface and aloft 5.3.2 Differences in the spatial distribution of LLJ and TEJ between wet and dry spells 5.3.3 Diurnal variation of winds at the surface and aloft. 5.4 Spatial and diurnal variation of LLJ and TEJ in wet and dry spells 5.5 Summary and Conclusions 6. Rainfall characteristics during wet and dry spells 110 6.1 Introduction 6.2 Data description 6.3 Characteristics of rainfall in wet and dry spells of the monsoon 6.3.1 Spatial structure of rainfall during wet and dry spells vii

Contents 6.3.2 Occurrence percentage of different types of rainfall in wet and dry spells. 6.4 Discussion on rainfall causative/modifying mechanisms in different spells. 6.5 Conclusions 7. Summary and future scope 134 7.1 Summary 7.2 Future scope References 142 Appendix -I: Computation methodology of CAPE and CINE Appendix -II: List of Publications viii

List of figures List of Figures S.No Description Page 1.1 Geographical extent of the global surface monsoons. The red, green 3 and blue areas indicate the tropical, subtropical and temperate-frigid monsoons respectively. The red and blue thick lines represent the ITCZ in summer and winter, respectively. (Li and Zeng 2003) 1.2 Spatial distribution of mean seasonal summer monsoon rainfall. 5 1.3 Amplitude spectrum of rainfall over the monsoon zone (15-25 N, 75-85 E) for the year 2007 constructed using daily rainfall generated by India Meteorological Department (IMD) (see chapter 2 for more details). The dashed line indicates the 7 1.4 Spatial distribution of rainfall fraction (%) in all-india break periods defined by several authors. (courtasy Rao et al. 2009). 17 2.1 The Automatic Weather Station at Gadanki. 23 2.2 Major components of ORG-815. 25 2.3 GPS balloon ascent (left panel), radiosonde, antenna/ground receiving system and data acquisition system (right panel from top to bottom). 27 2.4 Schematic diagram of index of variation ( n) for different scattering mediums (courtesy RÖttger and Larsen, 1989). 30 2.5 SODAR system at Gadanki. 32 2.6 LAWP system at Gadanki. 33 2.7 Antenna array of the state-of-art Indian MST radar. 35 2.8 Schematic diagram of TRMM microwave sensor s observation geometry. 2.9 Location of 1803 rain gauge stations (courtesy Rajeevan et al., 2006) 42 2.10 Spatial distribution of correlation coefficient between the rainfall at each grid point and area integrated rainfall (integration is over the shaded area). Daily rainfall data, during 1951-2009, prepared by IMD in the summer monsoon are used for this purpose 38 43 2.11 Time series of area averaged daily rainfall in 2007 (histograms) superimposed on area averaged climatological daily rainfall. The climatological daily rainfall is estimated from 58 years of high 43 ix

List of figures resolution rainfall data generated by IMD. The solid line represents the climatological area averaged rainfall and the dashed lines represent climatological rainfall ± standard deviation. The hatched and cross hatched bars represent the dry and wet spells, respectively. 2.12 Periods of wet and dry spells identified for 14 years (1995 2009) using high resolution 1º x 1º gridded rainfall data. 2.13 Rainfall percentage contribution during (a) dry and (b) wet spells of the southwest monsoon from 1995-2009. 3.1 (a) Spatial distribution of the climatalogical seasonal (June- September) mean rainfall retrieved from high-resolution (1 x1 ) rainfall maps prepared by India Meteorological Department (IMD) (Rajeevan et al. 2006). Also shown are vector winds at 850 hpa from NCEP (Kalney et al. 1996). Solid circle indicates the location of Gadanki. (b) Variation of monthly rainfall at Gadanki, retrieved from 9 years of ORG measurements, showing the annual cycle of rainfall. 3.2 Frequency distributions (in terms of % occurrence) for surface (a) temperature and (b) humidity during wet and dry spells. The means of the distribution are also shown in the figure. 3.3 Vertical profiles of mean temperature, humidity and equivalent potential temperature differences between wet and dry spells ( T = T d -T w, q = q d -q w and e ed - ew). Suffixes d and w denote dry and wet spells, respectively. 3.4 Same as figure 3.2, but for the altitudes of (a) ABL, (b) 0 C isotherm, (c) LRT and (d) CPT. 3.5 Same as figure 3.4, but for (a) LCL, (b) LNB, (c) CAPE and (d) CINE. 3.6 CAPE variation with reference to the start time of rain in (a) dry and (b) wet spells. The color indicates the amount of rain occurred during that rain event. 3.7 Positive thermal buoyancy (in K) profiles during dry (top panel) and wet spells (bottom panel). 3.8 Vertical distribution of % occurrence of stable layers defined by different temperature lapse rates during (a) dry and (b) wet spells. (c) Frequency distribution of temperature lapse rates during wet and dry spells in two height regions (2-3.5 km and 4.5-6 km). D and W denote dry and wet spells, respectively. 3.9 Time-height distribution of vertical air velocity (w) on 24 July 2008 (representing wet spell) and on 07 June 2008 (representing dry spell) during the passage of convection. The white area on 07 June 2008 44 45 49 50 52 54 56 59 61 62 63 x

List of figures represents the gap in the data due to technical problems. 3.10 Vertical profiles of (a) composite w and (b) buoyancy during the passage of convection in wet and dry spells. 4.1 Comparison of LCL, LFC, EL and CAPE at Gadanki as estimated by GPS sonde, ECMWF-interim, NCEP and WRF data. 122 data points correspond to one monsoon season (June September). Total data are for 4 years (2006-2009). 4.2 A statistical comparison (in terms of histograms) of LCL, LFC, EL and CAPE at Gadanki, as estimated by GPS sonde, ECMWFinterim, NCEP and WRF data. Since, total number of soundings by sondes and models are not same, occurrence percentage is plotted instead of count/frequency. 4.3 Scatter plot between CAPE and parcel s surface equivalent potential temperature (left panel) and also specific humidity (right panel), showing the dependency of CAPE on lifting air parcel s thermal properties. The correlation coefficients are also shown in the figure 4.4 Histograms of CAPE (at Gadanki) in (a) dry and (b) wet spells, respectively, as estimated by GPS sonde measurements and model outputs, illustrating the differences in CAPE between the spells. 4.5 Contribution of CAPE in different layers (L1: 700-500 hpa, L2: 500-300 hpa, L3: 300-200 hpa) to total CAPE in wet (left panel) and dry (right panel) spells at Gadanki. GPS sonde measurements and NCEP, ECMWF-interim and WRF model outputs are used in this analysis. 4.6 Spatial distribution of mean CAPE in (a) dry and (c) wet spells, respectively, over the southern peninsular India. The spatial map of standard deviation of the mean CAPE in (b) dry and (d) wet spells, respectively, are also shown in the. The dot denotes the location of Gadanki. 5.1 Number of diurnal cycle days in wet and dry spells obtained by different instruments (a) in each year and (b) from all years. Note that the y-axis in (a) is shown in logarithmic scale. 5.2 Vertical profiles of zonal (left panel) and meridional (right panel) winds for wet and dry spells obtained from different instruments ((a)-mst radar, (b)-lawp, (c)-sodar and AWS). The average wind shown at 0 km (or surface) in figure (c) is obtained from AWS. Vertical profiles shown in (a)-(c) are daily averages. The standard deviation is represented by error bars. 5.3 Mean zonal wind on 850 hpa level for (a) dry and (b) wet spells showing the spatial variation of LLJ. The black solid line in figures (a) and (b) represents 10 m s -1 contour. (c) The difference in LLJ 64 70 71 73 76 78 80 90 91 94 xi

List of figures between spells (dry-wet). The dot denotes the location of Gadanki. (d) and (e) shows the spatial distribution of standard deviation of mean values for dry and wet spells, respectively. 5.4 Same as figure 5.3, but for zonal winds at 100 hpa level, showing the TEJ variation. The black thin and thick solid lines in figures (a) and (b) represent 24 m s -1 and 28 m s -1 contours, respectively. 5.5 Diurnal variation of hourly mean zonal wind from the surface to lower stratosphere in dry (left panel) and wet (right panel) spells obtained from several instruments. (a) and (b) shows the diurnal variation in the height range of 3.6-19 km with IMSTR. The solid, black-dashed and white-dashed lines on the figure represent 8 m s -1, 0 m s -1 (height of wind reversal) and -30 m s -1 contours. (c) and (d) same as (a) and (b) but for the height range of 600 m 4 km obtained with LAWP. The solid line represents the 15 m s -1 contour. (e) and (f) same as (a) and (b) but for the height region of 60 m 1 km obtained with SODAR. The solid line represents 3 m s -1 contour. The surface wind variation obtained from AWS (denoted as sfc ) is also included in the figure. Note the change in color scale between different plots. 96 97 5.6 Same as figure 5.5, but for meridional winds. 98 5.7 Hodograph between mean zonal and meridional winds, derived from SODAR, in (a) dry and (b) wet spells. The mean is taken over the entire spell and also in the height region of 300-500 m. 99 5.8 Comparison of zonal wind anomalies in the overlapping region of different instruments (1 km for SODAR and LAWP and 3.6 km for LAWP and MST) in (a) dry and (b) wet spells. SODAR, IMSTR and LAWP measurements are represented with solid line with symbols and the ECMWF interim data are represented with a symbol. 5.9 The spatial variation of zonal wind anomaly for dry spell on 850 hpa level, showing the diurnal variation of LLJ. The ECMWF interim data at (a) 05:30, (b) 11:30, (c) 17:30 and (d) 23:30 LT are used to obtain the wind anomaly at the above synoptic hours. 103 104 5.10 Same as figure 5.9, but for wet spell. 105 5.11 Same as figure 5.9, but for TEJ (on 100 hpa level). 106 5.12 Same as figure 5.11, but for wet spell. 106 6.1 Spatial variation of total rain amount and rain fraction during dry (a and c) and wet (b and d) spells, respectively. Location of Gadanki is indicated with a star. Rain fraction of any spell indicates the fraction of seasonal rainfall in that spell (see text for more details). 116 xii

List of figures 6.2 Total number of rain pixels in dry and wet spells available over each grid point (grid resolution is 0.25º x 0.25º). TRMM PR measurements during 1998-2009 are used for this study 6.3 The percentage occurrence of stratiform, convection and other types of rain during dry (top panel) and wet (bottom panel) spells, depicting the spatial variation in the occurrence of different types of rainfall. 6.4 Spatial variation of the occurrence percentage (shown in color) of convection (top panel) and stratiform (bottom panel) rain at different synoptic hours during dry spell, depicting the diurnal variation of the occurrence percentage of different types of rain. The numbers on each panel indicate the time of measurement (3 hourly bins). 119 120 121 6.5 Same as figure 6.4, but for wet spell 122 6.6 Spatial variation of surface sea level pressure in (a) wet and (b) dry spells. The wind vectors on 850 hpa level are overlaid to depict circulation features. Both pressure and winds are obtained from ECMWF interim analysis. 124 6.7 Diurnal variation of TRMM 3B42 rainfall (3-hourly) (in cm) during dry (top panel) and wet (bottom panel) spells. 6.8 Same as figure 6.7, but for 2 years (2001- top panel and 2008-bottom panel) during wet spell. 6.9 Vertical wind shear estimated using surface and 850 hpa winds averaged over spells 125 126 128 6.10 Wind reversal altitude during wet and dry spells 129 xiii

List of tables List of Tables S.No Description Page 2.1 Types of sensors and accuracies of surface weather parameters. 26 2.2 Types of the sensors and accuracies of Radiosondes. 28 2.3 Important specifications and parameters of SODAR, LAWP, IMSTR and TRMM PR used in the present thesis. 3.1 Mean CAPE and e (and standard deviation) for 4 categories (no rain, before rain, during rain and after rain) during wet and dry spells. 4.1 The average LCL, LFC, EL and CAPE values along with standard deviation (SD) estimated by radiosonde and different models. The value in the parenthesis represents the number of vertical profiles used for the estimation of mean values. 4.2 Correlation analysis between CAPE and e (and also q) using different model outputs and radiosonde data. The correlation coefficient (significant at 99% level), RMSE (estimated from the linear fit and data points), and the slope of linear fit are also included. 4.3 Comparison between models and observations on their dependency of CAPE on air parcel s lifting height (pressure level). 4.4 Mean CAPE and standard deviation for wet and dry spells as estimated by GPS sonde and different models. The value in the parenthesis indicates the number of CAPE values averaged to obtain the mean value. 4.5 Percentage occurrence of single- and double-peaked buoyancy profiles in wet and dry spells as estimated by GPS sonde and model data. 4.6 The percentage occurrence of single and double peaks in buoyancy profiles during wet and dry spells at different grid points. The analysis is performed using ECMWF interim data. 37 60 72 74 75 77 80 81 xiv

Acronyms List of Acronyms ABL ARMEX AWS BOBMEX CAPE CBL CDC CINE CMAP CPC CPT DBS DSR ECMWF EL ENSO EqIO ERA GPI GPS HWSI IAV IMD IMSTR INSAT ISMR IO ISO ITCZ JASMINE LAWP LCL LFC LHF LLJ LRT METEOSAT MJO MOSDAC NBL NCEP Atmospheric Boundary Layer Arabian sea Monsoon Experiment Automatic Weather Station Bay of Bengal Monsoon Experiment Convective Available Potential Energy Convective Boundary Layer Climate Data Centre Convective Inhibition Energy Climate Prediction Center Merged Analysis of precipitation Climate Prediction Centre Cold Point Tropopause Doppler Beam Swinging Downward Shortwave Radiation European Centre for Medium Range Weather Forecasting Equilibrium level El Nino Southern Oscillation Equatorial Indian Ocean European Centre for Medium Range Weather Forecasting Reanalysis Global Precipitation Index Global Positioning System Horizontal Wind Shear Index Interannual Variability India Meteorological Department Indian Mesosphere Stratosphere Troposphere Radar Indian Satellite Indian Summer Monsoon Rainfall Indian Ocean Intra Seasonal Oscillation Inter Tropical Convergence Zone Joint Air Sea Monsoon Interaction Experiment Lower Atmospheric Wind Profiler Lifting Condensation Level Level of Free Convection Latent Heat Flux Low Level Jet Lapse Rate Tropopause Meteorological Satellite Madden Julian Oscillation Meteorological and Oceanographic Satellite Data Archival Centre Nocturnal Boundary Layer National Centre for Environmental Prediction xv

Acronyms NLLJ NOAA NPISO OLR ORG PR SAD SHF SODAR SST TEJ TRMM TTL UHF VHF VIRS WMO WRF Nocturnal Low Level Jet National Oceanic Administration Agency Northward Propagating Intraseasonal Oscillation Outgoing Longwave Radiation Optical Rain Gauge Precipitation Radar Spaced Antenna Drift Sensible Heat Flux Sound Detection and Ranging Sea Surface Temperature Tropical Easterly Jet Tropical Rainfall Measuring Machine Tropical Tropopause Layer Ultra High Frequency Very High Frequency Visible and Infra red Scanner World Meteorological Organisation Weather Research and Forecast xvi

Chapter 1 Introduction.

Chapter 1 Introduction 1. Introduction: the term Monsoon is an Arabic word, which means seasonal reversal in the wind direction. Prevailing wind direction between winter and summer seasons are the basis for the delineation of the monsoon regions around the world. According to Ramage (1971), the monsoonal region in the tropics extends between 25º S - 35º N and 30º W - 170º E. On the other hand, Webster et al. (1998) argued that distinct variation in monsoon annual cycle occurs in Asia, Australia, Africa and America (Webster et al. 1998). The global scale persistent reversal of winds varies with time of the year (Trenberth et al. 2000). In high latitudes, the monsoon is defined as those sub regions, where surface cyclones and anticyclones alternate infrequently in summer and winter (Ramage 1971). From the past 350 years, these monsoon systems acquired fervent attention by scientific community and common people alike. These are globally distributed atmospheric phenomena (Figure 1.1) giving surplus amount of water to mankind. It is estimated that the monsoon rains all over the world provides 60% of global water supply. Some of the global monsoon systems are briefly described below: 1.) African Monsoon: The seasonal shifts in the Inter Tropical Convergence Zone (ITCZ) and the difference in temperature and moisture content of the atmosphere between Saharan Africa and equatorial Atlantic Ocean are the main sources for this monsoon circulation. The northward movement of ITCZ starts from equatorial Atlantic in February and reaches western Africa around last week of June. It retreats to its normal position in October. The semi arid regions of Sahel and Sudan receive abundant rainfall during African monsoon season. 2.) North American Monsoon: It is also known as southwest or Mexican monsoon. The onset of this monsoon system occurs during the last week of June or early July and persists up to September. It completely occupies southwest United States of America (USA) by mid- July. 3.) Australian Monsoon: The Australian monsoon system acts as a major energy source for the Hadley circulation of the southern hemisphere. In 2

Chapter 1 Introduction general, the rainy season starts from September and ends in February. Over ¾ of the annual rainfall occurs during this season. The Siberian antihigh pushes the low pressure towards southern hemisphere creating cyclonic circulation over Australia, which interacts with the cold air from high latitudes causing significant weather phenomena. 4.) European Monsoon: This monsoon is also known as the return of westerlies. In general the westerly winds are a common feature during European winter. Strength of these westerlies decreases during March- April- May and picks up again its strength in June, which is why it is called return of westerlies. The European monsoon is not a monsoon in the traditional sense. It doesn't meet all the requirements to be classified as monsoon. Return of the westerlies is more regarded as a conveyor belt that delivers a series of low pressure systems to Western Europe where they create weather disturbances. Figure 1.1: Geographical extent of the global surface monsoons. The red, green and blue areas indicate the tropical, subtropical and temperate-frigid monsoons, respectively. The red and blue thick lines indicate the position of ITCZ in summer and winter, respectively. (Courtesy: Li and Zeng 2003). 5.) Asian Monsoon: This is the largest planetary scale system giving good amount of rainfall to Asian countries. This giant monsoon system occupies entire tropical and subtropical northeastern hemisphere. Asian regions are 3

Chapter 1 Introduction critically influenced by the evolution and variability of the Asian monsoon. It interacts with El Nino/Southern Oscillation (ENSO) and extratropical circulations and thereby controls global circulations through teleconnections. It can be classified into two types: 1) South Asian monsoon or Indian monsoon, which effects Indian subcontinent and surroundings regions and 2) East Asian monsoon, which affects countries like China (southern part), Korea and Japan. This thesis focuses on regional differences in the South Asian monsoon system. 1.1 Indian Monsoon: Indian subcontinent is situated at the vicinity of the monsoonal region defined by Ramage (1971). Copious amount of rainfall occurs over the Indian sector during June-September, also known as summer (or southwest) monsoon season. It accounts for ~75-80% of the annual rainfall over major parts of the Indian subcontinent. Agrarian countries like India heavily depend on this rainfall. Any deficit in the seasonal rainfall will have an adverse effect on the agriculture and economy of the country. Onset of the monsoon brings cross equatorial flow and moisture towards the Indian landmass and causes heavy rainfall. The commencement of rainy season over Indian subcontinent is distinguished by the wide spread rainfall over Kerala coast in late May/early June. In general monsoon sets over Kerala on June 1 with a standard deviation of 8 days (Pai and Rajeevan 2009). Among the two branches of monsoon system, the Arabian Sea branch first appears between 22 and 25 May and later, after one week, strikes Trivandrum (8.51 N, 76.5 E). Monsoon arrival to the central India takes 10-15 days after the onset, and completely occupies the subcontinent by mid- July. The withdrawal of monsoon takes place in September from northwest India and by 20th October retreat of the southwest monsoon completes and another monsoon, called northeast monsoon, sets in over southeast India. Mean seasonal summer monsoon rainfall is not homogeneous (see figure 1.2); it varies spatially (within India) and temporally (within the season) over the Indian region. The rainfall occurrence is high over the West coast and Northeastern India 4

Chapter 1 Introduction (~100-300 cm), moderate to heavy over central India (100-150 cm) and low over northwestern and southeastern peninsular India (~50 cm). Figure 1.2: Spatial distribution of mean seasonal summer monsoon rainfall. The high resolution rainfall data generated by IMD are used for this plot (Details of high resolution gridded rainfall data are given in chapter 2). Southeast peninsular India receives considerable amount of rainfall during the northeast monsoon season, i.e., from October through December. After the onset of the southwest monsoon over Kerala coast, monsoon rainfall and ITCZ moves northward. However, the northward propagation or advance of the monsoon is not always smooth, rather it takes place in pulses or epochs, in accordance with the convective activity (Gadgil and Kumar 2006). During the southwest monsoon season the large scale distribution of precipitation is mostly over two regimes: North Bay of Bengal and south of the equator in Indian Ocean. Winds converge in these regions and these regions become most favored zones for convection and the formation of Tropical Convergence Zone (TCZ). Mean position of the TCZ is not constant instead it oscillates with different periodicities and scales. Oscillations of the TCZ are synchronized with enhanced and suppressed convective activity known as active and break phases of the monsoon. Thus, the monsoon is a manifestation of the seasonal migration of the ITCZ and monsoon variability is associated with the space-time 5

Chapter 1 Introduction variation of the ITCZ (Sikka and Gadgil 1980). As discussed earlier the seasonal monsoon rainfall varies spatially and temporally and the variation of rainfall depends mainly on the duration and the time of occurrence of active and breaks phases. Natural disasters caused by the extreme hydro-meteorological events are manifestation of intraseasonal variability (Goswami and Ajay Mohan 2001, Goswami et al. 2003, Jones et al. 2004a). In the following sections, intraseasonal oscillations (ISO) of the monsoon rainfall are discussed briefly. 1.1.1 Intraseasonal Oscillations: Indian Summer Monsoon Rainfall (ISMR) exhibits quasi-periodic oscillations over India and neighbouring seas. Intraseasonal oscillations are the embedded parts of the ISMR with different temporal and spatial scales. These oscillations are mainly characterized by different periodicities varies from, (1) synoptic scale fluctuations (3-7 days), (2) quasi-biweekly oscillations (10-20 days) (Krishnamurty and Bhalme, 1976; Chen and Chen, 1993; Chatterjee and Goswami 2004). (3) low- frequency or Madden-Julian oscillations (MJO) (30-60 days) (Yasunari, 1979; Sikka and Gadgil, 1980; Goswami, 2004a, Kulkarni et al 2009). Since 1970s, several studies were made to understand the relation between MJO and ISO of the summer monsoonal rainfall (Murakami, 1976, 1977; Yasunari, 1979, 1980; Dakshinamurti and Keshavmurty,1976; Chen and Murakami 1988; Goswami et al. 1998; Annamalai et al. 1999; Annamalai and Slingo 2001; Lawerence and Webster 2002; Kempball-cook and Wang, 2001; Hsu et al 2004; Kulkarni et al. 2009, 2011). MJO is the dominant component in the tropical atmosphere. It consists of large-scale coupled patterns in atmospheric circulation and deep convection propagating eastward from the warm sea surfaces. The convection maxima over the equatorial Indian Ocean (EqIO) (Hartman et al. 1992; Annamalai and Slingo, 2001) propagate northward from EqIO into Bay of Bengal and over India with a speed of 1-2 m s -1, or about 1-2 latitude day - 1 (Klingaman et al. 2008). To obtain the dominant periodicities in rainfall, the daily rainfall data during southwest monsoon season of 2007 are subjected to the spectral analysis. Figure 1.3 shows a typical amplitude spectrum of rainfall over the monsoon zone (central and 6

Chapter 1 Introduction northern India). Clearly, figure 1.3 depicts different scales of oscillations embedded in the southwest monsoon rainfall. Among them three modes are found to be significant (amplitude is larger than the 1 level, where is the standard deviation). They are 32 days (corresponding to MJO), 16 days (corresponding to 10-20 days) and 8 days (corresponding to 3-7 days). Figure 1.3: Amplitude spectrum of rainfall over the monsoon zone [central and northern India (15-25 N, 75-85 E)] for the year 2007, constructed using daily rainfall generated by India Meteorological Department (IMD) (see chapter 2 for more details). The dashed line indic The day-to-day variability in the rainfall plays an important role in deciding the seasonal rainfall. Seasonal mean monsoon rainfall is affected by the occurrence and strength of the active and break spells (Goswami and Ajaymohan, 2001). It is well known from the earlier studies that the modes of 30-60 days and 10-20 days are most important in controlling the ISMR (Sikka and Gadgil 1980; Singh and Kripalani 1985, 1986, 1990; Kripalani et al. 1991, 1999). Kulkarni et al. (2009) studied the spatial variability of ISO s in deficit and excess monsoon years. They observed the dominance of 30-60 days over west coast and southeast region during the deficit monsoon year, while excess monsoon years are characterized by high frequency oscillations. Another important result obtained by them is the weakening of 30-60 7

Chapter 1 Introduction days oscillation and strengthening of 10-20 day oscillation over central India and some parts of west coast during the last two decades. 1.1.2 Characteristics of ISO: The synoptic activity or conditions during the Indian summer monsoon season mainly consists of low pressures and depressions that are formed in Bay of Bengal. A substantial fraction of the large-scale rainfall over the Indian monsoon zone occurs in association with the propagation of synoptic scale systems generated over the Bay of Bengal on to Indian monsoon zone. Recent studies, during the last decade or so, observed an increase in the number of low-pressure systems and a decrease in the number of depressions, keeping the total sum unchanged during the monsoon season (Kumar and Dash 2001, Goswami et al. 2003). They observed that conditions for the genesis of low-pressure systems are more conducive in active phase than in break phase. Their study also reported that low pressure systems are spatially and strongly clustered along the monsoon trough region during active phases of the monsoon. The background mean flow associated with tropical interdecadal variability is the main responsible factor for the variability of lows and depressions (Krishnamurthy and Goswami 2000, Goswami 2004a, Dash et al. 2004). As discussed above, the active monsoon is associated with low pressure systems over the monsoon trough region (a semi-permanent low pressure zone during southwest monsoon season extending from the head Bay of Bengal to heat low over northwestern parts), cyclonic vorticity patterns and strong low-level westerly jet, known as low-level jet (LLJ), over the southern peninsular India. Low-level monsoon winds, associated with the ISOs, with large meridional shear and cyclonic vorticity on 850 hpa level enhance the cyclogenesis during active spells. On the other hand, during break monsoon, the LLJ moves southward from its mean position of ~15º N and settles between Sri Lanka and the equator (Joseph and Sijikumar 2004). Also an anticyclonic circulation pattern is observed over the monsoon zone at the boundary layer heights (Ramamurty 1969, Sikka and Gadgil 1978 and Kusuma et al. 1991). Circulation patterns during break spells are in general weak compared to those during active spells. The important characteristic features observed over the monsoon trough region during break spells are the presence of heat-low type of circulation, subsidence 8

Chapter 1 Introduction in the entire troposphere, and a prominent trough on 700 hpa level (Raghavan 1973). Occurrence of the heat-low type of circulation during peak monsoon months July August) indicates the break phase, and the situation is revived by the establishment of ITCZ. The question of how the TCZ and heat-low type circulations replace each other in accordance with the active and break conditions still remains unanswered and it is a challenging task. The influence of ISOs on ISMR by using station/gridded rainfall data was carried out by several authors (Singh and Kripalani, 1985; Chowdary et al. 1988; Hartmann and Michelson 1989; Singh and Kripalani, 1990, 1991; Gadgil and Asha 1992; Singh et al. 1992; Krishnamurthy and Shukla, 2002; Bhanukumar et al. 2006). Recently, a diagnostic study, by Bhanukumar et al. (2010a), investigated the impact of ISO in terms of number, duration and intensity on rainfall during June through September during floods and drought conditions. They observed 4 to 6 oscillations during normal and flood years, whereas only 2 to 3 oscillations during the drought years. Further, their study correlates the rainfall with the ISO oscillations. They observed a correlation of 0.56 between number of oscillations and rainfall. The strength of ISO interms of different indices (MJO index, Monsoon Shear Index) is also investigated. In a comprehensive study Bhanukumar et al. (2010b) investigated the interannual variability of the ISO, which is a manifestation of changes in the atmospheric circulations and extreme weather events. They used OLR and winds at 850 hpa and 200 hpa level and observed a relation between rainfall and indices of MJO. They observed that the interannual variability of ISO is partly linked to ENSO cycle. They concluded that the strong ISO activity is often observed with LaNina. 1.1.3 Causative mechanisms of ISO: The satellite (e.g. National Oceanic and Atmospheric Administration (NOAA), Tropical Rainfall Measuring Mission (TRMM), Quick Scatterometer (Quick SCAT), Meteorological Satellite (METEOSAT) and Indian Satellite (INSAT)) and reanalysis (National Centres for Environmental Prediction/National Center for 9

Chapter 1 Introduction Atmospheric Research (NCEP/NCAR) and European Center for Medium Range Weather Forecasting Reanalysis (ERA)) products were very useful in studying and understanding the spatial and temporal characteristics of monsoon intraseasonal oscillations. For example, using the daily satellite imagery, Sikka and Gadgil (1980) observed two cloud bands over the Indian sector, one band is situated over the heated subcontinent and the other is over the warm waters of the Equatorial Indian Ocean. These convergence zones or cloud bands are known as Continental TCZ (or monsoon trough) and Oceanic TCZ, respectively. The space-time variability of monsoon rainfall is a sign of seasonal migration of continental TCZ (Gadgil 2003). Vertical instability observed during the monsoon season and mid-tropospheric heating by deep convection over the monsoon region are the feedback processes for the activation of Continental as well as Oceanic TCZ. Apart from these, land-surface processes also contribute for Continental TCZ (Gadgil et al. 2007). The superposition of two dominant modes, 10-20 days and 30-60 days present in the monsoon rainfall, are also responsible for the movement of TCZ (Yasunari 1979, Sikka and Gadgil 1980). Mechanisms governing northward propagation of 30-60 days oscillation are confined to Indian and western pacific latitudes (Sikka and Gadgil 1980; Yasunari 1980). These mechanisms are classified as those invoking dynamics (Wang and Xie 1997) and thermodynamics (Fu and Wang 2004; Zheng et al. 2004; Webster 1983). These 10-20 day or bi-weekly oscillations are associated with the westward/north westward moving synoptic systems, generated over warm waters of Bay of Bengal, and produce substantial amount of rainfall. On the other hand, 30-60 days or MJO oscillations, generated over Equatorial Indian and Pacific Ocean, propagate northward/north eastward (Singh and Kriplani, 1985, 1986, 1990; Kriplani et al, 1991). Bi-weekly oscillations contribute ~25% to the total sub seasonal variability of monsoon rainfall, whereas MJO oscillations contribute ~67% (Annamalai and Slingo 2001). Interaction between upper-middle tropospheric westerly trough from midlatitudes and western disturbance in the lower troposphere influences the monsoon trough movement. Western disturbances can shift the monsoon trough towards the 10

Chapter 1 Introduction foot hills of Himalayas and can cause draught conditions over central and northern India (Ding 2007). Over the tropical latitudes, the nonlinear interaction between convection and dynamics i.e., the interactions between the convective instabilities and large scale circulations plays a major role in the generation of ISO s (Goswami and Shukla 1984; Neena and Goswami 2010). 1.1.4 Predictability of ISO: Prediction of ISOs two to three weeks in advance could be of immense help in agricultural planning and water management. The predictability of day-to-day (high frequency) fluctuations governed by synoptic systems is limited only to 2-3 days, whereas low-frequency ISO s can be predicted up to 2-3 weeks in advance (Ramasastry et al. 1986; Singh and Kriplani 1990; Krishnamurti and Ardunay 1980; Cadet and Daniel 1988). Further, quasi-periodic nature of ISO s indicates certain potential predictability. By using weekly rainfall data during 1979-1983 over four zonal belts, from the southern tip to the northern parts of India, Ramasastry et al.(1986) observed that the year to year variability of the 40-day mode is quite large and also it fluctuates from one area to another in the same year. They, therefore, concluded that the potential for long-range prediction of rainfall is rather small. In another study, Singh and Kriplani (1990) examined interannual variability of 30-40 days oscillation of rainfall, utilizing daily rainfall data of 290 stations over the Indian continent for a period of 80 years. They noticed significant interannual variability in the 30-40 days oscillation and, therefore, concluded that the potential for its prediction is limited to 10 days. Recently, Goswami and Xavier (2003) proposed an empirical model for the prediction of monsoon ISO s (active and break phases) with the help of rainfall and circulation data of 23 years (1979-2001). They defined an index of ISO by filtered precipitation anomaly over the monsoon trough region (70-90º E, 15-25º N). They concluded that the monsoon break phases are inherently more predictable than active conditions since the transition from break to active is more chaotic than active to break phase. Their results indicate that useful prediction of monsoon breaks could 11

Chapter 1 Introduction possibly be made up to about 18 days in advance while those for active conditions is likely to be limited to a lead time of about 10 days. Earlier studies have shown that ISO s control the Interannual variability of rainfall and therefore they also dictate the predictability of seasonal monsoon rainfall. The predictability of Indian summer monsoon depends mainly by the relative contribution of two factors. The highly varying low frequency internal forcing by the monsoon ISOs (Charney and Shukla 1981; Goswami and Ajay Mohan 2001) and the slowly varying external forcing by the background flow. Xavier and Goswami (2007) proposed that the accurate prediction of ISO s will improve the predictions of seasonal rainfall. On the other hand, recent studies observed a decrease in the ISO activity during last two decades (Goswami 2004; Kulkarni et al. 2009). Thus, the internal forcing is decreased causing the enhancement in the predictability of the seasonal rainfall. 1.2 Active and Break spells: The active and break spells of Indian monsoon associated with ISO s need to be understood properly as they control the seasonal rainfall. It is noted by several researchers that excess monsoon rainfall years generally have more active spells and deficit monsoon rainfall years will have prolonged and/or more break spells (Webster et al. 1998, Gadgil 2003, Rajeevan et al. 2006). A modest decrease in the monsoon rainfall (e.g., 10% of the long-term mean) significantly affects the food productivity in countries like India (Swaminathan 1987, Parthasarathy et al. 1988, Webster et al.1998, Gadgil et al. 1999a; Abrol and Gadgil 1996). The knowledge of active and break spells on a regional scale is more crucial and important than all-india integrated active/break spells for agriculture and water management sectors. Therefore, the prediction of the time of active/break spell occurrence and their initiation date is much more important for the management of sowing and harvesting than the seasonal mean rainfall. 1.2.1 Identification of Active and Break Spells: Several research works have been carried out to understand and identify the active and break spells over the Indian subcontinent using various parameters. In the 12

Chapter 1 Introduction present section an overview of the earlier works on the identification of the active and break spells is briefly given: a) The characteristics of active and break spells were first described by Blandford in1886. According to Blandford, break or drought conditions exists, when northwesterly and westerly winds interrupt the monsoon trough movement in the Northwestern and Central India. The monsoon trough was further pushed to the foothills of the Himalayas. On the other hand, during active or height of the rains, cyclonic vortices, with closed isobars are formed in the immediate neighbourhood of the trough. b) In an extensive study Ramamurthy (1969), using 80 years (1888-1967) of all India rainfall anomalies, noticed the following patterns during the break period: the trough (TCZ) of low pressure moved north from its mean position and easterlies on 850 hpa were absent. Also during break phases, the surface pressure in the monsoon zone is nearly 4 hpa larger than the surface pressure observed when the trough is at its normal position. He could observe that break periods are more frequent in the month of August with a duration of about 3-5 days. Whereas, during active monsoon, mean position of the monsoon trough/tcz is situated south of its normal position. c) Magna and Webster (1996) used 12 years (1980-1991) of wind data (both zonal and meridional components) on 850 hpa level and outgoing longwave radiation (OLR) to define active and break periods in the monsoon rainfall. The criteria that they have adopted is as follows: active (break) monsoon conditions subsist when the zonal wind on 850 hpa level and OLR averaged over 65º - 95º E and 10º - 20º N exceeds 3 (-3) m s -1 and less (greater) than 10 Wm -2, respectively. Also, the meridional winds on 850 hpa level at 45º E, 0º N should exceed 3 m s -1. d) Later, Goswami and Ajaymohan (2001) also used the strength of wind on 850 hpa level to identify active/break spells. They, however, utilized the filtered zonal winds at a single grid point (15º N, 90º E) and used a different wind thr 13

Chapter 1 Introduction e) In a classic study, De et al. (1998) identified the break spell following India Meteorological Department (IMD) s definition, first suggested by Rao et al. (1976). i.e., There are periods when the monsoon trough (TCZ) moves very close to foothills of Himalayas, as a result there is an abrupt decrease of rainfall over the monsoon zone or central India and an increase in rainfall near the foothills of Himalayas, northeastern and southern peninsular India. f) Krishnan et al. (2000) used 17 years (1979-1995) of NCEP-derived OLR to understand the evolution of break phases. The study was carried out over a large area, covering 18º - 28º N and 73º - 82º E, and longer period in each season, i.e., 15 June - 15 September. The longer time period avoids ambiguity, if any, arising because of the early onset and delayed withdrawal of monsoon. The objective criterion to identify the break period uses OLR anomaly. The period is considered as break if the area averaged OLR anomaly exceeds +10 Wm -2 for four consecutive days. g) Annamalai and Slingo (2001) utilized the ISMR data during 1970-1995 to identify active and break spells. To eliminate small scale fluctuations, 5-day running means of ISMR (rainfall over the entire India), expressed in percentage departures from climatological means, are computed. The active phase is defined, when the ISMR is > 20 % of the normal for three consecutive days and the break phase when the ISMR < 20 % of the normal for three consecutive days. h) Vechhi and Harison (2002) developed a monsoon break index based on the composite monsoon break structure of OLR, following Krishnan et al. (2000). The new index was computed from the difference between normalized 7-day boxcar-smoothed OLR anomalies averaged (10 30 N, 65 85 E) and (10 S 5 N, 75 95 E), minus its 50-day centered mean. The index is positive (negative) for monsoon breaks (active periods). 14

Chapter 1 Introduction i) In an extensive study on breaks in the monsoon rainfall over India, Gadgil and Joseph (2003) used rainfall distribution over the monsoon zone to identify the active and break epochs. The criterion is taken in such a way that there should be maximum overlap with the traditionally defined breaks and active spells (Ramamurthy 1969 and De et al 1998). They used 89 years (1901 1989) of daily rainfall data during peak monsoon months, i.e., July-August, from 273 stations distributed evenly over the monsoon zone. The period is considered as active (break) if the area averaged rainfall rate is larger (smaller) than the rainfall threshold. They noted that the rainfall rate within the monsoon zone is not same rather it is found to be large (small) in the eastern (western) part of monsoon zone. Therefore, they suggested the usage of different rainfall thresholds for eastern and western parts. The stratification of eastern and western parts is done based on correlation analysis. The thresholds for the identification of active (break) spells in western and eastern parts of the monsoon zone are 8 mm day -1 and 12.5 mm day -1 (2.5 mm day -1 and 7.5 mm day -1. j) Using all India daily rainfall during 1901-2002, rather than traditional surface pressure and circulation patterns, Ramesh Kumar et al. (2004) identified break periods during monsoon. They referred break, when the daily rainfall is less than the threshold rainfall (9 mm day -1 ) for continuously 3 days. k) Rajeevan et al. (2006) by using rain gauge measurements, made over 1803 stations during 1951-2003, developed a high resolution (1º x 1º) gridded rainfall data set. Further, by means of standardized rainfall anomaly averaged over the central India (21º - 27º N and 72º - 85º E) they identified active and break spells. The procedure is as follows; the area averaged climatological mean and standard deviation of rainfall for each calendar day was prepared, then the standardized rainfall anomaly was computed by subtracting the daily rainfall data from climatological mean and dividing with daily rainfall standard deviation. If the standardized rainfall anomaly, is less (greater) than 15

Chapter 1 Introduction -1.0 (+1.0) for three consecutive days, then that period is considered as break (active) spell. l) Further, Prasad and Hayashi (2007) proposed a new method for the identification of active, weak and break spells of the monsoon by using ERA- 40 reanalysis and 850 hpa level wind shear data during 1958-2002. They used Horizontal Wind Shear Index (HWSI) defined by Wang et al (2001). It is the difference in 850 hpa winds between southern, Zone 1 (5-15 N, 40-80 E) and northern, Zone 2 (20-30 N, 70-90 E) zones. Days with HWSI values larger than the mean plus standard deviation are considered as active periods, similarly the days with small values of HWSI and weak zonal winds than the mean minus standard deviation in Zone 2 are considered as weak days. Break days are identified as days with small values of HWSI and relatively strong westerly winds in Zone 2. m) A simple index was defined by Klingaman et al. (2008) based on lag correlations between OLR over India and over equatorial Indian Ocean for diagnosing the northward propagating (30-40 day periodicities) ISO s. The daily OLR data is obtained from NOAA Cooperative Institute for Research in Environmental Sciences Climate Diagnostics Center (CIRES CDC). The period is considered as active (break), if the index is greater (less) than 1 (-1) standard deviation for at least 5 consecutive days. Since, the definitions and techniques used by several authors for the identification of active and break spells are different, the duration and occurrence of the spells in any monsoon season can be different. For instance, Gadgil and Joseph (2003) reported that there is hardly any overlap between the spells observed by them and those observed by Webster et al. (1998). There were three or four active break sequences observed by Webster et al. (1998) in each summer monsoon season. They could observe that the duration of break spell days is between 1-7 days with 90% of the spells are having 3-5 days duration. There was no break days observed in 10 years by Ramamurty (1969). However, the study reported that the duration of break days varies from 3 to 15 days with 30 % of the spells are longer than or equal to 7 days. 16

Chapter 1 Introduction 1.2.2 Spatial variability of Active and Break spells: As mentioned above, the knowledge and prediction of active and break spells on a regional scale is important as different regions grow different crops and follow different practices. Earlier studies have shown that intraseasonal variations in rainfall are not coherent over the Indian region and the active/break spells of some subdivisions are in opposite phase with each other (Krishnamurthy and Shukla 2000; Gadgil 2003; Rajeevan et al. 2010 and references therein). For instance, the spatial structures of the active/break spells are organized in such a way that, southeast peninsular and northeastern parts of India exhibit an out of phase relation with the monsoon zone or central India. Also, rainfall increases near the foothills of Himalayas, following the northward movement of monsoon trough, during break spells for the monsoon zone. Recently, Rao et al. (2009) demonstrated that there exists large spatial variability in active and break spells using 58 years (1951-2007) of high resolution (1 x 1 ) rainfall data generated by IMD (Rajeevan et al. 2006). They generated spatial maps for rainfall fraction (in terms of %) corresponding to break periods as defined by different criteria, discussed in the earlier section (Ramamurty 1969, De et al. 1998, Gadgil and Joseph 2003, Rajeevan et al 2006). Figure 1.4: Spatial distribution of rainfall fraction (%) in all-india break periods defined by several authors. (courtesy: Rao et al. 2009). Figure 1.4 shows that large rainfall fraction of ~20-40% occurs over southeast India; in particular over Gadanki during all India break (Rao et al. 2009). The clear discrepancy in the rainfall distribution between these regions indicates the need for identifying active and break spells separately for the southeastern region of India. In chapter 2, a different algorithm for the identification of active and break spells is 17

Chapter 1 Introduction described. As mentioned above, the spatial patterns of ISO s and associated active/break spells are found to be different in excess and deficit monsoon years (Kulkarni et al. 2009). For ex., 30-60 day period is dominant during deficient monsoons over west coast and south-east regions. 1.3 Scope of the Thesis: It is clear from the above discussion that ISO s substantially influence the seasonal mean rainfall and active/break spells exhibit large spatial variability within India. The intraseasonal variability of the Indian summer monsoon and its association with interannual variability has been studied by several researchers (Singh et al. 1992; Webster et al. 1998, Krishnamurthy and Shukla 2000; Goswami and Mohan 2001; Lawrence and Webster 2001, Gadgil et al. 2003, Goswami 2005 and references therein). Most of the studies on ISO s confined to the monsoon zone for obvious reasons. It is mainly because the rainfall variations over monsoon zone (both seasonal and intraseasonal) are highly correlated with all-india rainfall variations. Therefore, often, active/break spells defined using the rainfall over monsoon zone is considered as all-india active/break spells. This, probably, tempted researchers to concentrate more on rainfall variability over the monsoon zone. Only a few reports available in the literature on ISO are in other parts of India. In particular, over the southeast peninsular India, the region where active/break spells are in opposite phase to those observed over the monsoon zone. Further, most of the studies listed above focused on rainfall variability on ISO scale and changes in large-scale circulations without detailing the vertical structure of the atmosphere. A few studies attempted to study the variations in the vertical structure of the atmosphere (in terms of temperature, humidity, boundary layer height, winds, Convective Available Potential Energy (CAPE), etc.) in different spells of the monsoon. But, these studies mainly utilized the campaign mode observations made over the Ocean (Bay of Bengal Monsoon Experiment (BOBMEX) (Bhat et al. 2001), Arabian Sea Monsoon Experiment (ARMEX) (Murthy et al. 2006), and Joint Air-Sea Monsoon Interaction Experiment (JASMINE) (Webster et al. 2002)) and over the monsoon zone (Parasnis 1991; Kusuma et al. 1991). None of the above studies used 18

Chapter 1 Introduction long-term observations to study the vertical structure of the atmosphere in active and break spells of the monsoon. A Recent study by Rao et al. (2009) over Gadanki has reported intriguing differences in the occurrence of draft cores and also in the vertical structure of draft cores between wet and dry spells. The shallow draft cores are predominantly seen in dry spell (~88% of total draft cores). Also the vertical profile of mean vertical velocity, obtained by Indian Mesosphere Stratosphere Troposphere radar (IMSTR), for deep convection showed two peaks during dry spells, while only a single peak is seen during wet spells. Nevertheless, the causative mechanisms for those interesting observations are not discussed in detail. In the present study, for the first time, an attempt has been made to document the differences in thermal and dynamical characteristics of the atmosphere between wet and dry spells and their vertical structure in these spells. How the differences in the background thermodynamics between spells modify the spatial rainfall patterns is also studied. The observed differences are used to address unresolved and interesting observations reported by Rao et al. (2009) on preferential occurrence of draft cores in dry spell and the differences in vertical structure of draft cores. Chapterization of the thesis is briefed below. Chapter 2 describes different in-situ and remote sensing (both ground-based and space-borne) instruments used in the present study. Different model datasets (ECMWF, NCEP and Weather Research and Forecast (WRF)) along with their temporal and vertical resolutions are briefly discussed. Spatial rainfall data products (2A25, 3B42) generated by TRMM satellite are also described. The methodology for identification of wet and dry spells over southeast peninsular India is discussed in detail. In this thesis the terms wet and dry are used instead of active and break. The active and break spells are used to refer all India (or monsoon zone) ISO s. In chapter 3, surface and upper-air observations of meteorological parameters at Gadanki are utilized to study differences in the thermal structure of atmosphere between wet and dry spells and its implication on the occurrence of draft cores. To 19

Chapter 1 Introduction examine whether the observed differences in thermal parameters over Gadanki are localized or exist over the entire southeast India, reanalysis outputs were sought. If reanalysis data sets are to be used for such purposes, it is necessary to examine how far model CAPE estimates match with those estimated by in-situ radiosoundings. In chapter 4 several key questions related to CAPE are addressed: 1) Can model CAPE estimates mimic CAPE variations observed by GPS balloon measurements over different temporal scales? 2) How does CAPE vary between different spells of the Indian summer monsoon (i.e., from wet to dry spell)? 3) Does differences in CAPE and in the vertical structure of buoyancy between spells are localized features over Gadanki or observed all over the southeastern peninsular region? 4) What physical/dynamical processes are responsible for the differences in CAPE between spells and how do they effect the convection growth in the dry spell? Since the forcing (thermal, CAPE, etc.) in different spells is different, one would expect differences in wind patterns. Therefore, chapter 5 focuses on differences in the vertical structure of mean wind and its diurnal variation between wet and dry spells, using a unique dataset consisting of measurements from ground based remote sensing (wind profilers) and in-situ devices augmented by ECMWF reanalysis (interim) datasets. Special emphasis was given to study the variation of LLJ and tropical easterly jet (TEJ), two most conspicuous features of the summer monsoon circulation. In chapter 6 rainfall characteristics during wet and dry spells of the Indian summer monsoon are briefly discussed. Several key issues related to the rainfall are addressed in this study: 1) How much seasonal rainfall occurs in each spell of the monsoon? 2) What kind of rainfall occurs (convective or stratiform) in different spells of the monsoon? 3) How different types of rain vary diurnally in different spells of the monsoon?. 4) What are causative mechanisms of rainfall over southeast India in different spells of the monsoon? In particular, we focus on why rain producing systems that produce heavy rain along the west coast of India are not propagating to east during dry spell? 20

Chapter 2 Data analysis and Instrumentation.

Chapter 2 Data analysis and instrumentation 2.1 Introduction: The present study uses variety of datasets, both observational and models, to understand intriguing differences in thermodynamical characteristics between wet and dry spells of the Indian summer monsoon over southeastern peninsular India. The basic surface meteorological parameters are obtained from an Automatic Weather Station (AWS). Surface rainfall measurements are obtained from an Optical Rain Gauge (ORG). For continuous profiling of winds from the surface to lower stratosphere (~20 km), SOund Detection And Ranging (SODAR), Lower Atmospheric Wind Profiler (LAWP), Indian Mesosphere-Stratosphere-Troposphere (MST) radar are employed. Vertical profiles of temperature and humidity along with winds are acquired from radiosonde observations. All the above measurements are made at Gadanki. To study the spatial variability of thermodynamical parameters over the Indian domain, various meteorological field variables obtained from NCEP/NCAR, ECMWF and WRF models are used. Various satellite-derived geophysical products from TRMM Precipitation Radar (PR) are also utilized. In the following sections, a brief description of the above instruments, working principles, important specifications, and accuracy of sensors and geophysical parameters is given. 2.2 Surface observations 2.2.1 Automatic weather station (AWS): The AWS at Gadanki measures basic surface meteorological parameters (Temperature, pressure, humidity, rainfall, wind speed and direction) in real-time with good accuracy. In addition to basic meteorological parameters, AWS also contains a sensor for sunshine hours. This system contains sensors for meteorological parameters, data logger, solar panel, rechargeable battery, GPS antenna and electronics for data transfer through satellite link. It consists of 6 sensors to measure the above atmospheric parameters and a tipping bucket rain gauge to measure rainfall. The sensor used to measure the temperature is a thermistor (PT 1000). The output of the thermistor is resistance, which is proportional to the ambient air temperature. The measurement of relative humidity of the air is done with a thin-film 22

Chapter 2 Data analysis and instrumentation Figure 2.1: The Automatic Weather Station at Gadanki capacitor type hygrometer. A dielectric polymer absorbs water molecules that are present in the air through a thin metal electrodes causing capacitance change proportional to humidity. The output voltage varies linearly with relative humidity. These sensors are encased in a stainless steel probe about 4 to 6 inches long. Both temperature and humidity sensors are kept in Stevenson screen to minimize the solar radiation effects. A special integral diaphragm type transducer is used to measure the absolute pressure in the atmosphere. Basic sensors are strain gauges bonded on the integral diaphragm. The overall output signal is proportional to the input resistance. To measure the wind speed and direction, cup anemometer and wind vane are used. A 3-cup anemometer, made of dark colored plastic material to prevent corrosion and to withstand winds up to 100 mi/h, is a fast-response, low-threshold optoelectronic device. When rotated by wind, chopper on the anemometer shaft interrupts an infrared light source, generating pulses from a photo transistor. Signal is amplified and fed to a line driver. The frequency of pulses is proportional to the wind speed. Wind direction sensor is an aluminum vane mounted on a low-torque, precision potentiometer. Output signal produced in the potentiometer is proportional to the wind direction. 23

Chapter 2 Data analysis and instrumentation The measurements of temperature, pressure, humidity, wind speed and wind direction were made for 3 minutes in each hour (29 th, 30 th and 31 st minute of each hour). The average of these 3 measurements is considered as the representative of that hour. These hourly values of the above parameters are stored in the data logger. The dynamometer measures radiation received on a horizontal surface from both the sun and sky. When exposed to radiation, the temperature of the blackened horizontal surface rises. Heat is lost from the blackened surface by conduction, convection and radiation. The equilibrium temperature reached is a measure of the radiation. This temperature is measured by a thermopile. A thin metallic film blackened with a special paint (which absorbs energy completely in the range of 0.3 contact with this thin metal film. Alternate junctions of this thermopile are in thermal contact with the massive body of the instrument at ambient temperature which serves as the cold junction. This way a milli volt output proportional to the radiation received (about 4mV/kW/m 2 ) develops across the thermopile. The instrument has a time constant less than 22 seconds. Tipping bucket rain gauge is used to measure the rainfall. The body and the funnel of the instrument are made up of Fiber glass Reinforced Plastic (FRP), and the rim is made with gun material. All parts having contact with water are made of stainless steel. When incremental amount of precipitation (0.5 mm in this case) is collected, bucket assembly tips and activates the magnetic reed switch. Each tip of the bucket produces on-off output. Number of tips or amount of rainfall in each hour is then calculated and is stored in the data logger. All the sensors are mounted on a 5-m tower at various heights, pressure sensor is mounted at 1.5 m, temperature and humidity sensors are placed at 2 m, radiation sensor at 1.90 m, solar panel is mounted at 2.5 m and wind speed and direction sensors are mounted at 5 m. Rain gauge is installed on the surface, as per requirements (see figure 2.1). The AWS system is designed to operate in remote locations. Therefore, it is powered with a rechargeable battery connected to a solar panel. The stored meteorological data in the data logger are transmitted to Meteorological and 24

Chapter 2 Data analysis and instrumentation Oceanographic Satellite Data Archival Center (MOSDAC) through satellite link. Also there is a provision to read/copy data from the data logger. The data logger is having sufficient memory to store data of 1 year. Accuracies of AWS sensors are given in Table 2.1. 2.2.2 Optical Rain gauge (ORG-815): The ORG-815 provides rainfall in all weather conditions with high temporal resolution (1 minute). ORG provides basic rainfall parameters like, precipitation rate, type and water equivalent accumulation. Figure 2.2: Major components of ORG-815 It is immensely superior to the traditional rainfall measuring (such as tipping bucket, siphon, weighing and electrical grid) sensors. ORG consists of 3 sensors: transmitter, receiver and electronics. Transmitter contains an IRED diode and a lens with disk heater. Receiver is equipped with photo diode, lens with aperture slit, disk heater, thermistor probe and connector for power cable. Transmitter and receiver are together mounted on a pipe (figure 2.2). When the partially coherent infrared light beam, from the transmitter encounters the precipitation particles, irregularities in the optical properties of the beam, known as scintillations, are received in the receiver. These scintillations have characteristic patterns and are converted to precipitation rate. This system can measure rain accumulations from 0.001 upto 999.999 mm with an accuracy of ~5% of accumulation. The dynamic range of the system is 0.1 500 mm hr -1. 25

Chapter 2 Data analysis and instrumentation Table 2.1: Types of sensors and accuracies of surface weather parameters. S. No Weather parameter Type of the Sensor 1. Air Temperature Thermistor (PT 1000) 2. Relative humidity Thin-film Capacitor 3. Atmospheric Integral diaphragm pressure with built in amplifier 4. Wind speed 3-cup rotor and IR emitter/detector circuit 5. Wind direction Linear potentiometer 6. Rainfall Tipping bucket rain gauge 7. Sunshine hours Thermopile (72 element) Operating range -40 C to 60 C 0 100 % ±3 % Measurement accuracy ±0.1 C 600-1100 hpa 0.1 hpa 0 60 ms -1 + 2% of full scale 0-359 degrees from North 0-1500 Wm -2 + 3 degrees Better than 1mm with resolution of 0.5 mm 2.2 Upper-air observations: 2.2.1 Radiosonde: Radiosonde is a balloon-borne instrument for measuring atmospheric parameters (temperature, pressure, humidity, wind speed and direction) at various altitudes. For the present study, vertical soundings of above meteorological parameters are obtained from Vaisala/Meisei balloon ascents from Gadanki at ~12 GMT (17:30 Local time (LT)). Since the present study mainly focus on the wet and dry spells of the southwest monsoon, upper-air measurements from Vaisala (Meisei) GPS sondes during 2006 (2007-2009) southwest monsoon season are only utilised. On few days, sondes were either not launched or not reached to higher altitudes. As a result, useful number of vertical soundings from these four years is 396. Radiosonde system consists of three subsystems (figure 2.3), namely 1. On-board system (Sensor package or radiosonde) 2. Ground receiving system 26

Chapter 2 Data analysis and instrumentation 3. Data acquisition system Hydrogen filled meteorological balloons made up of high-quality neoprene rubber are employed for the ascent of radiosonde. The sensors used to measure meteorological parameters are kept in a package, called radiosonde, and is attached to the balloon. The sensors measure the required meteorological parameters at different altitudes as the balloon ascents. Measurements of atmospheric parameters and sensors information is transmitted back to the ground receiving station by using a radio transmitter. A white ceramic covered resistance thermistor is used as a temperature sensor, whose resistance varies proportionately with temperature. Thermistor is painted with white to minimize the heating by sunlight. The humidity sensor, hygristor, is a thin film capacitor with a polymer dielectric. Sensor capacitance is dependent on the water absorption on the dielectrical material. The electrical resistance of the dielectrical material changes with atmospheric humidity. Figure 2.3: GPS balloon ascent (left panel), radiosonde, antenna/ground receiving system and data acquisition system (right panel from top to bottom) Humidity sensor is placed in such a way that the outside air passes the hygristor. Capacitance-based aneroid barometer is used to measure the pressure in Vaisala soundings. Meisei radiosondes do not use any pressure sensor, instead the pressure levels are computed from the hydrostatic equation from the height information obtained with GPS system. Upper-air winds (horizontal wind speed and 27

Chapter 2 Data analysis and instrumentation direction) are determined by measuring the radiosonde position (obtained with GPS) relative to the balloon launch location. Information regarding the azimuth and elevations of the balloon are converted to wind speed and direction at various altitudes. The electronic subsystem samples each sensor at regular intervals (1 sec. for Meisei sondes and 2 sec. for Vaisala sondes) and transmits the data to ground-based receiver and data acquisition system. The data acquisition system reduces the data in near real-time and stores in a computer. The type of sensors used by Vaisala and Meisei sondes and their accuracies are given table 2.2. Table 2.2: Type of sensors used in Vaisala and Meisei radiosondes and their accuracies. Parameter Sensor Range Accuracy Resolution Vaisala (RS 92) Temperature Thermo cap Capacitive Bead -90 C to 60 C 0.2 C to 0.3 C Relative Humidity Thin film capacitor 0-100% 2% 1% 0.1 C Pressure Aneroid barometer 3-1080 hpa 1 hpa 0.1 hpa Wind speed and GPS technique 0.15 m/s, 2 0.1 m/s direction Meisei (RS01GII) Temperature Thermistor -90 C to 40 C 0.2 C to 0.5 C 0.1 C Relative Humidity Carbon humidity sensor 0-100% 2% 1% Pressure Wind speed and direction Calculated from hydrostatic equation using height determined from GPS GPS technique 0.2 m/s 0.1 m/s 28

Chapter 2 Data analysis and instrumentation 2.4 Radar observations: Radar is the acronym for radio detection and ranging and is defined as the art of detecting by means of radio echoes, the presence of targets, determining their direction and range, recognising their character and employing the data thus obtained (Battan 1973). Radar is a powerful tool to study the atmosphere. The radars used for lower atmospheric studies are classified as clear-air radars and weather radars depending upon the echoing mechanism and probing frequency. Weather radars are normally operated at frequencies in the range of 3-14 GHz, while clear-air radars are operated in VHF/UHF bands mostly at frequencies ~ 50, 400 and 1000 MHz. The radars operating at VHF/UHF frequencies are designed primarily for wind profiling but have been effectively used to explore both clear air and hydrometeor structures (Gage et al. 1994; Ralph et al. 1995). These radars can be operated continuously with good temporal and spatial resolutions. The principal targets of the radar are the refractive index irregularities and distributed particles such as hydrometeors, present in the earth s atmosphere. 2.4.1 Scattering mechanisms: The scattering/reflection mechanisms responsible for radar returns at VHF/UHF are well documented in Gage and Balsley (1980), Röttger and Liu (1978) and Röttger (1980). These scattering/reflection mechanisms are classified as Bragg scattering, Fresnel (also known as partial) reflection/scattering, Rayleigh scattering and Thomson scattering. Bragg scattering, Fresnel reflection/scattering and Rayleigh scattering are the dominant mechanisms for UHF/VHF radar returns in the troposphere. The radio refractivity or refractive index (n) in non-ionized medium can be approximately expressed as (Balsley and Gage 1980), N 5 6 3.73 10 ( n 1) 10 2 T e 77.6P T (2.1) where P is atmospheric pressure (hpa), e is the partial pressure of water vapor (hpa), T is the absolute temperature (K). The wet term (e/t2) dominates in the lower 29

Chapter 2 Data analysis and instrumentation troposphere, but it is relatively negligible in the upper troposphere. The second term is the dry term, which is important above mid troposphere. Macroscopic spatial changes of n cause refraction or reflection, and microscopic spatial changes cause scattering. In the troposphere and the stratosphere, the fluctuation in refractive index, n, depends mainly on variations of e, T and P due to atmospheric turbulence. 2.4.1.1 Bragg Scattering: Figure 2.4: schematic diagram of index of variation ( n) for different scattering mediums (courtesy RÖttger and Larsen, 1989) Bragg or turbulent scattering is the primary source for VHF/UHF radar returns in clear-air. According to the theory of radio wave scattering from turbulent fluctuations of refractive index (Booker and Gordon 1950), the backscattered signal arises from the spatial Fourier component, whose wavelength is equal to one half of the radar wavelength,. Hence in order to have coherent backscatter, the condition to be satisfied is min < /2 < max, where min and max are related to the inner (l 0 ) and outer (L 0 ) scale sizes of the turbulence. The height distribution of l 0 determines the maximum altitude from which radar echoes at a given wavelength can be detected, provided the systems are sufficiently sensitive. The expression for volume reflectivity due to turbulence (Gage and Balsley 1980) is: 0.38C 2 n 1/ 3 (2.2) Where 2 C n is the refractive index structure parameter and is expressed, 2 C n 5.25( n) 2 L 2 / 3 0 (2.3) 30

Chapter 2 Data analysis and instrumentation and the radar equation is, P r P A L t e 32ln 2r r 2 (2.4) where Pr is the received power; Pt is the transmitted power; Ae is the effective antenna aperture; L (= r t) is loss factor due transmission ( t) and reception ( r); r is the range resolution; and r is the range. 2.4.1.2 Rayleigh scattering: In 1890 s Lord Rayleigh showed that, when the size of the scatterer is small compared to the wavelength of the incident radiation a dipole is induced within the scatterer. Rayleigh scattering theory can be used to approximate the characteristics of signals reflected back to the radar from precipitation particles (rain, snow, graupel and hail), when the diameter of the scatterer is small compared to the radar wavelength. For Rayleigh scattering from spherical particles, the backscattering cross-section of a single particle is: 5 4 K 2 D 6 (2.5) 5 4 K 2 Z e (2.6) Where Z 6 3 6 e ( mm m ) N i Di is radar reflectivity factor and is often expressed by radar meteorologists in the form of dbz e equal to 10 log (Z e ). K 2 is a function of the complex index of refraction of the target and is equal to 0.93 for liquid water and 0.18 for ice particles. The value of K 2 depends on the temperature and the physical constitution of the scatterers. The radar equation (Probert-Jones 1962) for Rayleigh scattering is: P r P hg t 2 512(2ln 2) 2 2 where, G is antenna gain, is beam width, h = C Substituting the equation (2.6) in equation (2.7): 2 r 2 (2.7) P r P hg t 2 3 2 1024(ln 2) K 2 r 2 2 Z e (2.8) 31

Chapter 2 Data analysis and instrumentation In the above equation, contributions of loss factors are not included. After incorporating the loss factors, this equation can be used to derive Ze. 2.4.2 SODAR: Figure 2.5: SODAR system at Gadanki Sodars are potential tools and they are used in various applications, for example, ABL characteristics and dynamics, formation of the nocturnal low level jet (NLLJ), inversion layers, and wind climatology in the boundary layer. Atmospheric scatters and absorbs sound waves much more strongly than the electromagnetic waves. This limits the maximum altitude coverage of the sodar to 1.5 km or less. Sodar systems and sodar techniques are used to profile the lower atmosphere, especially in the boundary layer heights. The National Atmospheric Research Laboratory (NARL) has established a multi frequency phased-array Doppler Sodar system. This sodar system consists of 8 X 8 antenna elements made up of piezoelectric transducers (CTS model KSN1165) operates in the frequency range 1600-2000Hz. The transducer generates 100 W acoustic powers. In order to make a circular array pattern, with a side-lobe suppression of 17 db, three elements from the corners are removed. The orientation of the 52-element antenna is tilted to 70 and the reflector to 35 to make the transmit beam vertical to the horizontal plane. Beams can be pointed towards East, North and vertical directions with a tilt angle of 22. Doppler spectrum at each range bin is computed from received echoes, using the fast Fourier Transform (FFT). Three lower order moments: signal strength, weighted mean Doppler shift, and half-power spectral 32

Chapter 2 Data analysis and instrumentation width of the power spectrum, are estimated by using the formulae given by Woodman (1985). Backscattered echoes with Signal to Noise Ratio (SNR) less than -15 db are not considered for wind vector computation. The computation of the wind vector (Zonal meridional and vertical) is done using standard expressions given by Kouznetsov et al (2004). Complete system description and data validation can be found in Anandan et al., (2008). 2.4.3 LAWP: Figure 2.6: LAWP system at Gadanki. The LAWP installed at Gadanki, is a coherent, phased array, Doppler radar operating at 1357.5 MHz with a peak power aperture product of 10 4 W m 2. The phased antenna array consists of 576 circular micro strip patch antenna elements arranged in a 24 x 24 matrix over an area of 3.8 m x 3.8 m. Figure 2.4 shows the house of LAWP antenna array and transmitter. The total array is organized into four quadrants. A total peak power of 1000 W is delivered to the antenna array by a parallel array of four outputs from power amplifier, each feeding 250 W to one quadrant (12 x 12 elements) of the array. The transmitter (Tx) unit, preceded by power amplifier, generates an output power of 175 W, which is sufficient to drive the power amplifier. The power amplifier generates the required final output power by a division-amplification-combining technique. The output power is fed via the beam changer switch and hybrid circulator. The power distribution across the array is tapered to obtain better side lobe suppression. The array produces a pattern having a beam width of 4º with a directive gain of 33 db. However, due to the ohmic loss of 4 33

Chapter 2 Data analysis and instrumentation db, the effective gain of the array remains 29 db. The beam can be steered electronically in three principal directions zenith, east and north. In the two principal planes (north and east) the beam is tilted to 15 from the zenith. The echo received by the antenna array from the atmosphere is delivered to the receiver via circulators. The receiver is a phase coherent heterodyne type having a quadrature detector at the final output, and delivers the video outputs to the signal processor. The receiver has an overall gain of 50-120 db depending upon the gain setting of automatic gain controller amplifier. The dynamic range of the receiver is about 66 db. The quadrature (I and Q) outputs of the receiver are limited to a peak to peak voltage of 10 V and given to the signal processor unit. The signal processor unit consists of an ADC and a coherent accumulator. The ADC has 12-bit resolution and samples the analog input with an interval set at the data processing unit. The signal processor unit performs the coherent accumulation on the ADC output data. The constituted coherent data are then transferred to the data processing unit for further processing. The data processing unit performs FFT on the received coherent data. The data are further processed to compute moments before being transferred to the off-line computer via ethernet for archival. The specifications of the LAWP are given in table 2.2.complete description of the system and data analysis methods are given in Rao et al. (2001) and Reddy et al. (2001). 2.4.4 IMSTR: IMSTR was built at Gadanki to probe the tropical atmosphere, as a part of the middle atmosphere program. Apart from the middle and lower atmosphere probing, it has been effectively used for studying ionospheric irregularities as well. The IMST radar is a highly sensitive, high resolution, pulse coded, phase coherent radar operating at 53 MHz (Rao et al. 1995). The panoramic view of the antenna array is shown in figure 2.7. IMSTR phased antenna array consists of two orthogonal sets, one for each polarization, of 1024 three element Yagi-Uda antennas arranged in a 32 x 32 matrix covering an area of 130 m x 130 m. 34

Chapter 2 Data analysis and instrumentation Figure 2.7: Antenna array of the state-of-art Indian MST radar The inter-element spacing in the matrix grid is 0.71 lobe free beam up to 20 steering from the zenith. The RF power from a transmitter is fed to a 3-dB in-phase power divider (combiner for reception) and distributed along the sub array through appropriate couplers of the feeder line. A modified Taylor distribution is adopted to achieve the desired first side lobe level of 20 db. The antenna pattern has been characterized in the receive mode by recording the radio source Virgo-A (3C 274) using the phase switching interferometer technique of Ryle (1952). Based on the measurements taken on several passes of the radio source during May-June 1993, it has been found that the beam pointing accuracy is better than 0.2 and 3 db beam width is in the range of 2.8 to 3 (Rao et al. 1995). The radar beam can, in principle, be positioned at any look angle within ±20 from zenith in the eastwest and north-south planes with a resolution of 1 by using 8-bit phase shifters. A peak power of 2.5 MW is generated by 32 transmitters, whose output power ranges from 15 kw to 120 kw, each feeds a sub array of 32 Yagis. The input to the transmitter is a low-level (1 mw) pulse modulated (coded/uncoded) signal at 53 MHz generated by a mixer which receives 5 MHz pulse modulated signal and an appropriately phase shifted 48 MHz local oscillator signal as an input. The transmitters can operate up to a duty cycle of 2.5%, limiting the total average power to about 60 kw. It is possible to transmit both coded and uncoded pulses with a pulse repetition frequency up to 8 khz, by keeping the duty cycle within the specified limits. The duplexer, which serves as a switch connecting antenna array to the transmitter and receiver channels during transmission and reception, respectively, is realized by means of distributed and lumped couplers, and PIN diodes. The Doppler shifted back scattered echo is delivered to a high-sensitive phase coherent receiver. The receiver is a super heterodyne type with a quadrature detector. 35

Chapter 2 Data analysis and instrumentation The received RF signals are passed through distributed low noise amplifiers, down converted to 5 MHz IF and then amplified before detection. The phase coherent receiver with quadrature channels (I and Q) has an overall gain of 120 db, a dynamic range of 70 db, and a bandwidth matched to the baud length of the coded pulse, detects the received weak back-scattered signals. In IF section, the amplifier chain gain is about 90 db and the bandwidth is 1.7 MHz. The IF signal is then split into and applied to a pair of quadrature mixers which mix them with 5 MHz local oscillator signals having quadrature phase of 0 and 90. The quadrature signals from the mixers are fed to two identical channels of low pass filters and video amplifier to obtain the two bipolar video signals of Acos (I) and (Q) at the output. The quadrature (I and Q) outputs of the receiver are fed to two identical 14-bit analog to digital converters (ADCs). After that, these signals go through a sequence of sampling, digitization, decoding and coherent averaging. The fast Fourier transform (FFT) routine is implemented to get the Doppler spectra. The specifications of the Indian MST radar are given in table 2.3. 2.4.5 Wind vector computations: The radial velocities of the targets are obtained from the Doppler spectra at specified height, for different radar beam positions. Velocity of the targets measured by the radar with the Doppler technique is called line-of-sight velocity, which is the projection of the velocity vector to the radial direction. Distinct techniques are used to measure the 3-components of the wind vector: the Doppler-Beam-Swing (DBS) and Spaced-Antenna-Drift (SAD). Here we need to make an assumption that the velocity field is uniform in space over the radar volume, contains range cells for computation of wind vector. For each height the radial velocities from the three beams are combined to compute the east-west (zonal), north-south (meridional) and vertical components of the wind respectively. At least 3 non-coplanar radial wind velocities are required to derive three components of wind. In the following section brief derivation of three components of wind vector is given; Winds from 5-beams or more: Least squares method is deployed for the computation three wind components, if the beams are 5 or more (Sato 1989). The line-of-sight wind velocity V d at a particular range and height is expressed as, 36

Chapter 2 Data analysis and instrumentation V d V. i Vx cos x Vy cos y Vz cos z (2.5) Where, i is the unit vector along the radar beam. East-west, North-south and Zenith directions are denoted by X, Y and Z respectively. By applying least squares method (Sato 1989)to the above equation, 2 V x cos x V y cos y V z cos z V Dj 2 (2.6) V Dj f Dj 2 (2.7) where, j represents the beam number. The condition for minimum residual is 0 (2.8) 2 V k V V V x y z cos cos cos x1 x2 x3 cos cos cos y1 y2 y3 cos z1 cos cos z 2 z3 1 V V V d1 d 2 d 3 (2.9) Solving the above matrix by using carmer s rule one can obtain V x, V y, and V z. Obtained winds corresponds to zonal meridional and vertical wind velocities. Table 2.3: Important specifications and parameters of SODAR, LAWP, IMSTR and TRMM PR used in the present thesis. Parameter SODAR LAWP MST Radar TRMM PR Frequency 1.8 khz 1.3575 GHz 53 MHz 13.8 GHz Beam width 5 3 3 0.71 Pulse width 180 ms 1 s 1.57 s FFT points 4096 128 256-512 Range 30 m 150 m 150 m 250 m resolution Beams Zenith, North, East Zenith, North, East Zenith-x, Zenith-y, East, West, North, South Vertically downward IPP 9 s 80 s 1000 s 3600 s Peak power 100 W 1 kw 2.5 MW 224 W 37

Chapter 2 Data analysis and instrumentation 2.5 Satellite observations: 2.5.1 TRMM: The TRMM is a joint project between NASA (National Aeronautical Space Agency) and Japan s JAXA (Japan Aerospace Exploration Agency). It is launched on 27 November 1997. TRMM is a low inclination near sun-synchronous orbital satellite with a suite of multi-frequency microwave and IR sensors (PR, TMI (TRMM Microwave Imager), VIRS (Visible and Infrared Scanner), LIS (Lighting and Imaging Sensor) and CERES (Clouds and the Earth's Radiant Energy System)). The presence of multi-frequency sensors has a greater advantage in obtaining atmospheric parameters from space. The primary scientific objective of the TRMM is to obtain tropical precipitation and to better understand the hydrological cycle and the role of latent heat in driving atmospheric circulations. Figure 2.8: Schematic diagram of TRMM microwave sensor s observation geometry. The first space-borne precipitation radar, onboard TRMM, measures 3-D distribution of rainfall over the land and ocean. It also provides the rain height information, which is used in radiometer-based rain rate retrieval algorithms. The small foot print of the PR enables a better estimate of rainfall parameters, when compared with those derived from other coarse microwave sensors. PR is an active phased array radar operates at 13.8 GHz frequency. PR antenna beam scans the scene in the cross-track direction over ±17, which results a swath width of 220 km. Antenna beam width is 0.71 and there are 49 range bins within the scan angle. The 38

Chapter 2 Data analysis and instrumentation horizontal resolution of the PR at nadir is 4.3 km before boosting the altitude of TRMM. At present, the horizontal resolution is ~5 km at nadir.. PR echo sampling is performed over 80 range gates above the surface with a range resolution of 250 m. TMI is a multi-frequency microwave imager with 5 channels. Out of 5 frequencies, 4 are dual polarisaton channels. The 5 frequencies are 10.65 GHz, 19.35 GHz, 22.235 GHz (single polarization), 37.0 GHz and 85.5 GHz. The horizontal resolution of the TMI will range from 5 km at 85.5 GHz to 45 km at 10.65 GHz. Scan angle of the sensor is 65 with a wider swath of 760 km. VIRS is a 5 channel crosstrack scanning radiometer operating at 0.63, 1.6, 3.75, 10.80, and 12.0 microns. Swath width of the instrument is 720 km. It provides cloud distribution by type and height and estimates rain from brightness temperatures at a horizontal resolution of 2.1 km at nadir. A calibrated optical Lightning sensor (LIS) onboard estimates the global incidence of lightning and its relation to the global electric circuit. The operating frequency of the sensor is 0.7774 microns with a swath width of 790 km. 2.5.1.2 TRMM rainfall products (TRMM 2A25): PR data products are available in three levels: Level 1 (1B21, 1C21), Level 2 (2A21, 2A23, 2A25) - IFOV (instantaneous field of view) data products and Level 3 - monthly statistical values of rain parameters mainly in 5 x 5 grid boxes. 2A25 algorithm objectives are to correct the measured reflectivity for rain attenuation and to estimate instantaneous 3-D distribution of rain from the TRMM PR data. Retrieved vertical profiles of attenuation-corrected radar reflectivity factor and rainfall rate are given at each resolution cell of the PR. The estimated rainfall rate at the actual surface height and the average rainfall rate between the two predefined altitudes (2 and 4 km) are also calculated for each beam position. 2A25 basically uses a hybrid of the Hitschfeld-Bordan method and the surface reference method to retrieve vertical profiles of attenuation-corrected effective radar reflectivity factor (Ze), (Iguchi and Meneghini (1994). The vertical rain profile is then calculated from the estimated Ze profile by using an appropriate Ze-R relationship. It also retains rain type information, primarily, derived by 2A23 algorithm. In the present study, 2A25-derived rain type information is effectively utilized. 39

Chapter 2 Description of instruments data and methodology 2.5.1.2 TRMM rainfall products (TRMM 3B42): The TRMM and other data precipitation data set generated by TRMM Multi-Satellite Precipitatioin Algorithm (TMPA) contains two products; 1) 3-hrly combined microwave-ir estimates (with gauge adjustment) (3B42). 2) monthly combined microwave-ir-gauge estimates of precipitation (3B43), computed on quasiglobal grids about two weeks after the end of each month starting in January 1998. The TRMM 3B42 is a calibrated infrared (IR) rain product. TRMM combined instrument algorithm (TCA), 2B31 and VIRS (1B01) data are used to calibrate the geosynchronous satellite infrared data. The entire avaliable archive of AMSR-E and AMSU-B rain estimates are incorporated in 3B42 data processing algorithm. Global rainfall estimates are made by adjusting the global precipitation index (GPI) to TRMM estimates. The main purpose of the 3B42 algorithm is to produce a merged high quality IR precipitation product with high temporal (3-hourly) and spatial resolution, 0.25 x 0.25. In the period 1 January 1998 to 6 February 2000, each grid box s histograms in the 1 x1 3-hourly GPCP IR histograms is zenith-angle corrected, averaged to a single T b value for the grid box, and plane-fit interpolated to the 0.25 grid. From 7 February 2000 onwards, the CPC Merged IR brightness temperature data at 4 km spatial resolution (over a latitude band 60 N-S, with a total grid size of 9896 x 3298), is taken as input for the processing algorithm. The amount of imagery delivered to CPC varies by satellite operator, but international agreements mandate that full coverage is provided for the 3-hourly synoptic times (00Z, 03Z,, 21Z). 2.6 Model observations: 2.6.1 NCEP/NCAR reanalysis data: The NCEP and NCAR are two organizations operating together to produce atmospheric variables on a global scale to support the research and climate monitoring communities. The daily data products are available at a temporal resolution of 6 hours times (i.e. 00, 06, 12 and 18 UTC) and with a spatial resolution of 2.5º x 2.5º. The model data has 17 vertical levels (12 of which are below 100 hpa) starting from 40

Chapter 2 Description of instruments data and methodology surface to 10 hpa. The complete description of the NCEP/NCAR data was given in (Kalney 2006). 2.6.2 ECMWF Interim data: ECMWF Re-analysis (ERA) interim data sets provide access to various atmospheric parameters. Interim data sets are available from 1 January 1989 with spatial resolution of 1.5 x 1.5. These data sets are more extensive than ERA-40 model data sets, e.g. the number of pressure levels is increased from ERA-40 s 23 to 37 levels, starting from 1000 hpa to 1 hpa, and additional cloud parameters are included. Among the 37 vertical pressure levels (27 of which are below 100 hpa). ERA-Interm daily data are available for surface, pressure and model levels. Analyses of the field variables are computed at 00, 06, 12 and 18 UTC every day. Complete descriptions of the data products are extensively given in ECMWF technical newsletters no. 110, 111 and 115. 2.6.3WRF data: A Real Time Weather Research and Forecast System (RTWRFS) was operational at NARL. The system uses WRF Version 3.1.1 model which runs on everyday at 00 UTC and provides 72-hr forecast over Indian region. The model runs with two nested domains with spatial resolution of 27 km and 9 km. For model initialization, NCEP GFS data at 0.5 resolution is employed. In addition, about 500 surface meteorological parameters derived from Automatic Weather stations spread all over India and satellite derived winds (Kalpana) are assimilated in the model by using nudging method. Data has 38 vertical pressure levels. The model can construct the thermal and dynamical meteorological parameters with high resolution (say 1 hr) between radiosonde ascents. 2.7 IMD high resolution daily gridded rainfall data: High resolution 1 x 1 daily gridded rainfall data set was prepared by Rajeevan et al., (2006) by using the daily rainfall data archived at National Data Center (NDC), IMD, Pune. IMD has rainfall records of 6329 stations with varying periods. Out of these, 537 are IMD observatories, which measures and report rainfall 41

Chapter 2 Description of instruments data and methodology that has occurred in the past 24 hours ending 0830 hours IST (0300 UTC), 522 are under the Hydro-meteorology program and 70 are Agromet stations. The remaining are rainfall reporting stations maintained by state governments. However, only 1803 out of 6329 stations had a minimum 90% data availability. The network of stations used in the analysis is given in fig 2.8. In order to interpolate the irregularly distributed data to a regular N-dimensional array Shepard numerical interpolation method is used. A comprehensive description of the procedure and data discussed above is given in Rajeevan et al. (2006). Figure 2.9: Location of 1803 rain gauge stations (courtesy Rajeevan et al., 2006). 2.8 Identification of wet and dry spells: In the present thesis, the terms wet and dry spells, are used instead of active and break spells, in order to avoid confusion with all India active and break spells of the summer monsoon rainfall. The method of analysis for the identification of wet and dry spells is discussed below. Wet and dry spells are identified using area integrated surface rainfall measurements, following Rajeevan et al. (2006). 42

Chapter 2 Description of instruments data and methodology Figure 2.10: Spatial distribution of correlation coefficient between the rainfall at each grid point and the area integrated rainfall (integration is over the shaded area). Daily rainfall data, during 1951-2009, prepared by IMD in the summer monsoon are used for this purpose. The area over which the surface rainfall is integrated (9.5-15.5 N and 77.5-81.5 E) is selected based on the correlation analysis. First, climatological time series rainfall at each grid point in the area is correlated with the areal average rainfall. The grid points are excluded if the correlation is not significant. The exercise is repeated till the correlation between time series of rainfall at each grid point with areal rainfall is good (correlation coefficient > 0.5). Spatial map of the correlation coefficient at each grid point obtained following the above procedure is shown in figure 2.10. Figure 2.11: Time series of area averaged daily rainfall in 2007 (histograms) superimposed on area averaged climatological daily rainfall. The climatological daily rainfall is estimated from 58 years of high resolution rainfall data generated by IMD. The solid line represents the climatological area averaged rainfall and the dashed lines represent climatological rainfall ± standard deviation. The hatched and cross hatched bars represent dry and wet spells, respectively. 43

Chapter 2 Description of instruments data and methodology The period is considered as wet (dry), if the area-averaged rainfall is more (less) than the rainfall threshold (climatalogical mean rainfall + (-) standard deviation) for the same calendar day for at least 3 continuous days. (The climatological mean rainfall is obtained from 58 years of high resolution gridded data prepared by IMD (Rajeevan et al 2006)). Note that, the condition of areal averaged rainfall exceeding the climatological mean continuously for 3 (or more) days excludes the possibility that wet spells may arise due to isolated convective showers. A typical picture showing the identification of wet and dry spells for the year 2007 is shown in Figure 2.11. Figure 2.12: Periods of wet and dry spells identified for 14 years (1995 2009) using high resolution 1º x 1º gridded rainfall data. Following the procedure described in the above section, a total of 943 from 72 dry spells, and 391 from 54 wet spells are identified from the fifteen years (1995-2009). The spells are ranging from 5 50 days during dry, and from 3 31 days during wet. Average time span during wet spell is ~7, where as the span is ~13 days during the dry spell. Among dry (wet) spells, ~34 (48) % of spells have time span longer than the average length of dry (wet) spell. Prolonged dry and wet spells are observed in 2002 and 2005 years respectively. Seasonal monsoon rainfall during these extreme years is 384.4mm and 471.9 mm respectively. 44

Chapter 2 Description of instruments data and methodology Figure 2.13: Rainfall percentage contribution during (a) dry and (b) wet spells of the southwest monsoon from 1995-2009. It is general belief that the rainfall in the rain-shadow region of southeastern peninsular India occurs in isolated convective storms or along the coast mainly due to sea-breeze intrusions. To examine how much rainfall is due to large scale systems (in wet spell) and how much is due to isolated storms (in dry spell), the rain amount contribution by each spell to the seasonal rainfall is estimated (figure 2.13). As also mentioned in chapter 1, the wet and dry spells are in opposite phase in the monsoon zone and southeast peninsular India. i.e., during wet spell, southeastern peninsular India gets good amount of rainfall, nevertheless the monsoon zone seldom gets rainfall and vice-versa. Though the wet spell persists only 22% of time in the southwest monsoon, 50-60 % of the seasonal rainfall occurs in that spell. On the other hand, only 5-20 % of seasonal rainfall occurs in the dry spell. One can see clearly that good amount of rainfall (figure 2.12b) occurs along the west coast in dry spell and almost no rain in the eastern side of the Western Ghats in the southern peninsula (south of 15 N) in wet spell. Then a logical question arises, why the rainfall systems, which produce good amount of rainfall in the west coast, are not propagating and/or translating to east in dry spell? How they migrate to east in wet spell? To understand this, differences in thermal and dynamical characteristics and also energetics between wet and dry spells are studied in detail in the following chapters. 45

Chapter 3 Thermal structure during wet and dry spells.

Chapter 3 Thermal structure during wet and dry spells 3.1 Introduction: During the course of Indian summer monsoon, there are periods of enhanced and reduced rainfall and convective activity, referred to commonly as active and break spells of the monsoon. Extensive studies have been made on the basic characteristics and the genesis of these ISOs (reviews on ISOs by Webster et al. 1998; Gadgil 2003; Goswami 2005; see also Chapter 1). These studies advanced our understanding of surface rainfall patterns and excursions of the convective zone during all India active and break spells. Nevertheless, there are only a few studies focused on variations of the vertical structure of the atmosphere from active to break spells. The studies that dealt the vertical structure were based on field campaigns over Bay of Bengal and Arabian sea BOBMEX (Bhat et al. 2001), JASMINE (Webster et al. 2002), ARMEX (Murthy and Sivaramakrishnan 2006). A few studies focused on boundary layer variations in the monsoon zone (Parasnis 1991; Kusuma et al. 1991). Further, all above mentioned studies are based on observations over the ocean (Bhat et al. 2001; Webster et al. 2002; and Murthy and Sivaramakrishnan 2006), or near the monsoon zone (Parasnis 1991; Kusuma et al. 1991). Unfortunately, there is no comprehensive documentation available in the literature dealing with variations in the vertical thermal structure and stability of the atmosphere in wet and dry spells in these regions, in particular in the southeastern region. Such studies are warranted in this region, because a recent study has shown intriguing differences in the draft cores in wet and dry spells at Gadanki (Rao et al. 2009). They observed a bi-modal vertical draft structure in dry spell with one peak appearing at ~5 km and the other between 11 and 13 km. On the other hand, the mean vertical velocity in draft cores during wet spell shows only a single peak in the upper troposphere. The other interesting observation noted by Rao et al. (2009) is that the shallow cores are preponderant in dry spell (~88% of total shallow cores are observed in dry spell). To better understand these observations, it is essential to study the thermal structure (stability, CAPE, Convective Inhibition Energy - CINE, etc.) of the atmosphere in wet and dry spells using vertical soundings of temperature and humidity. In this chapter, differences in thermal characteristics (temperature, humidity, stability, CAPE, etc.) of the atmosphere near the surface and aloft in wet and dry spells of the monsoon over Gadanki are studied. These observations are used 47

Chapter 3 Thermal structure during wet and dry spells to address unresolved issues related to the vertical structure of draft cores. For instance, why shallow cores are predominant in dry spell? Why vertical profiles of mean vertical air velocity corresponding to deep cores show two peaks in dry spell, while a single peak in wet spell? The chapter is organized as follows. Section 3.2 describes climatological rainfall features and the wind pattern at and around Gadanki. A brief discussion on the data used for the present study is included in this section. Section 3.3 focuses on variations of the thermal (temperature and humidity) and stability parameters (CAPE, CINE) and also the significant levels, which are interesting to the meteorological community (altitudes of the boundary layer, the 0 C isotherm and the tropopause) from wet to dry spell. These observations are discussed in light of differences in the draft core statistics in section 3.4. Important conclusions drawn from the present study are summarized in section 3.5. 3.2. Data used and meteorological background: The data used in the present chapter includes temperature (T) and specific humidity (q) fields at the surface and aloft during 4 summer monsoon seasons (2006-2009), high-resolution (1 x 1 ) rainfall data over India, generated by IMD (Rajeevan et al., 2006), vertical air velocity (w) measurements made with IMSTR during the passage of draft cores, and the surface rainfall measured by the ORG at Gadanki. Surface measurements are made with an AWS located at Gadanki, which provides all meteorological parameters at hourly intervals (chapter 2). Vertical soundings of T and q are obtained from Vaisala/Meisei balloon ascents launched at around 12 GMT (17:30 Local time (LT)) from Gadanki. Detailed discussion of the above instruments is included in Chapter 2. The observational area, Gadanki, is a semiarid region in the southeastern part of India (see figure 3.1a), far south of the mean position of the monsoon trough. The high-resolution (1 x 1 ) daily rainfall maps, generated using the surface rainfall from 1803 stations spread all over India (Rajeevan et al. 2006), are used to produce Figure 3.1a. Also shown in the figure are the seasonal mean wind vectors at 850 hpa over India, derived from NCEP reanalysis (Kalney et al. 1996). Strong westerlies/northwesterlies are predominantly seen over the southern peninsula. 48

Chapter 3 Thermal structure during wet and dry spells Figure 3.1: (a) Spatial distribution of the climatalogical seasonal (June-September) mean rainfall, retrieved from high-resolution (1 x1 ) rainfall maps prepared by IMD (Rajeevan et al. 2006). Also shown are vector winds at 850 hpa from NCEP (Kalney et al. 1996). Solid sphere indicates the location of Gadanki. (b) Variation of monthly rainfall at Gadanki, retrieved from 9 years of ORG measurements, showing the annual cycle of rainfall. It is apparent from the figure that the summer monsoon seasonal rainfall is abundant over the west coast and northeastern parts of India and to some extent in north and central India. The rainfall in these regions constitutes most of the ISMR and also variations of rainfall over this zone are highly correlated with the variations of ISMR (Gadgil 2003). On the other hand, rainfall is scant over northwest and southeastern parts of India. The annual rainfall at Gadanki is about 790 mm, of which 53% falls in the southwest monsoon (June-September) and 33% in northeast (NE) monsoon (October-December) (Figure 3.1b). Gadanki receives maximum rainfall in the transition period from the southwest to the northeast monsoon (September- October). The diurnal variation of rainfall is significant in the southwest monsoon with most (~80%) of the rain occurring during evening to midnight period (during 16-02 LT) (Rao et al. 2004; Kozu et al. 2006). Most of the observed rain in the southwest monsoon is associated with isolated convection and mesoscale convective 49

Chapter 3 Thermal structure during wet and dry spells systems (Rao et al. 2008). A detailed description of method of the identification of wet and dry spells during the southwest monsoon season over southeast India is described in Chapter 2. A total of 92 (201) wet (dry) days are identified from 15 (18) wet (dry) spells, during 2006-2009. The radiosondes were launched on 179 (70) days out of 201 (92) days of dry (wet) spell. Most of the sondes reached high altitudes in both spells with about 96% of sondes burst above the tropopause altitude. 3.3. Results: 3.3.1. Variation of temperature and humidity on the surface and aloft: The variation of surface T and q between wet and dry spells are depicted in the form of histograms (in terms of % occurrence) in figure 3.2. The mean and standard deviations of the distributions for wet and dry spells are also shown in the figure. It is apparent from the figure that the wet spells are relatively moist and cooler than dry spells. Figure 3.2: Frequency distributions (in terms of % occurrence) for the surface (a) temperature and (b) humidity during wet and dry spells. The means of the distribution are also shown in the figure. Although, the surface T and q distribution in these spells show some overlapping, the mean (mode) values differ by 1.6 K (~2 K) in T and ~2 g kg -1 in q. In general, temperature (humidity) is higher (lower) during June-July than during August-September months. To verify whether the mean values are biased by these variations within the observational period, monthly means of T and q are estimated for 50

Chapter 3 Thermal structure during wet and dry spells wet and dry spells for each month. The monthly mean values (both the T and q) corresponding to wet and dry spells in each month also show large differences indicating that differences in T and q between spells are caused primarily by active/subdued convection in respective spells. To understand how the convection in the wet phase of the monsoon changes the vertical structure of thermodynamic variables through complex feedbacks (latent heating, etc.), the difference between wet and dry spells of composites of T ( T = T d - T w ), q ( q = q d - q w ) and equivalent potential temperature, e,( e ed - ew) are estimated (Figure 3.3). Suffixes d and w denote dry and wet spells, respectively. Significant differences in composites of T, q and e from wet to dry spell and also with height are observed. At most of the altitudes, atmosphere is warmer in dry spell than in wet spell. Maximum temperature differences are seen in the lower and upper troposphere and are of the order of 1-2 K. The lower troposphere (below ~2 km) and upper troposphere (11-16 km) are warmer in dry spell than in wet spell. Between 2 and 11 km, the temperature difference between spells is small ( the temperature is higher in wet spell than in dry spell in two height regions, 2-5 km and 9-10.7 km. Earlier studies have shown cooling in the lower troposphere and heating in the middle troposphere during the wet spell of the Australian and Indian monsoon systems (over oceans) (McBride and Frank 1999; Bhat et al. 2002; Webster et al. 2002). The present observations do not show a clear heating in the middle troposphere during the wet spell, except for those two height regions (2-5 and 9-10.7 km) mentioned above. The difference between wet and dry composites of q is relatively large below 7 km (> 0.5 g kg -1 ), except for a height region of 2 km centered on ~3 km. Later, it will be shown that strong inversion layers exist in this height region during dry spell. These stable layers prohibit the growth of convection and modify the distribution of moisture. In other words, the stable layers will enhance the detrainment and thereby moisture near that level. Recent draft core statistics reported by Rao et al. (2009) also supports this view. They observed that the shallow cores are more prevalent in dry spell with the maximum percentage occurrence of core tops in the height region of 3-5 km. The small q in the height region of 2-4 km is, therefore, is associated with the enhanced detrainment and humidity in this height region in dry spell. Except for this 51

Chapter 3 Thermal structure during wet and dry spells narrow layer, our results are consistent with those obtained in Australia and India, where a moist environment was observed over greater depths during the active phase of monsoon (McBride and Frank 1999; Bhat et al. 2002; Webster et al. 2002). Atmospheric humidity will change primarily through horizontal advection, evaporation and subsidence of drier air from higher altitudes. At Gadanki, the wind remained southwesterly to northwesterly throughout the observational period, irrespective of the phase of the monsoon. Nevertheless, the magnitude of the wind is different in different spells (stronger winds in dry spell than in wet spell) (chapter 5). Therefore, the role of advection in drying the troposphere during dry spells is not clear. Evaporation of surface moisture can explain the observed humidity differences between spells. Though there are no large local water bodies nearby, evaporation of surface moisture (in general is more during wet spell because of more rain) can enhance humidity in the lower troposphere, as observed in the figure, during wet spell. The other possible candidate for the dryness of the atmosphere during dry spells is large scale subsidence from higher altitudes. Figure 3.3: Vertical profiles of mean temperature, humidity and equivalent potential temperature differences between wet and dry spells ( T = T d -T w, q = q d -q w and e ed - ew ). Suffixes d and w denote dry and wet spells, respectively. Vertical variation of e is a measure of conditional instability of the atmosphere. The negative gradient of e indicates conditionally unstable atmosphere. Following Bhowmik et al. (2008), we have examined the vertical variation of e in wet 52

Chapter 3 Thermal structure during wet and dry spells and dry spells (not shown here). There is not much variation in the depth of negative gradient in e between spells. However, large variation in the magnitude of e is found between spells in the lower troposphere (below 2 km) (figure 3.3 b). Larger e values are observed during wet spell indicating that the atmosphere is moist in that spell. Bhowmik et al. (2008) studied differences in e between rainy and non-rainy days at 4 locations in India. They observed large variation in e over Delhi in pre-monsoon season and attributed this variation to the advection of moist air during rainy days. Nevertheless, at Gadanki, as discussed above, the role of advection is unclear. However, the evaporation of surface moisture (latent heat flux) during wet spell seems to be mainly responsible for the observed e variation in the lower troposphere. 3.3.2. Variation of significant levels: In this section, variation of meteorologically important levels, like the atmospheric boundary layer (ABL), 0 C isotherm level, lapse rate tropopause (LRT), and cold point tropopause (CPT), from wet to dry spell are studied. The ABL height is estimated from virtual potential temperature ( v ) profiles, following Stull (1988) and Kusuma et al. (1991). Within the boundary layer, the v remains nearly constant, but increases rapidly at the top of the boundary layer (typically larger than 1 K per 100 m). This property is used to identify the ABL top. The frequency distribution (in terms of % occurrence) of the ABL height for wet and dry spells is shown in figure 3.4a. Although there is an overlap of distributions for wet and dry spells, the mean and mode of the distributions are different from each other. The mean ABL altitudes in wet and dry spells are, respectively, 1.9 km and 2.3 km. The ABL altitudes derived using wind profiler measurements at Gadanki are also in the same range as measured by sondes in the present study (Reddy et al. 2002; Krishnan et al. 2003; Kumar and Jain 2006). It is evident from the figure that the ABL, in general, is deeper in dry spell than in wet spell. Deep (Shallow) boundary layers during the break (active) periods also have been observed over land (at Pune, Parasnis (1991) and in the monsoon zone, Rao et al. (1991)) and over oceans (Bhat 2001), although the mean height of the ABL and its difference between active and break spells varies from region to region. The ABL height difference between active and break spells is maximum (1-2 km) in the north western region (near the desert) and minimum near the coast (~100-500 m) (Rao et al. 1991). The observed difference at Gadanki is in between these two 53

Chapter 3 Thermal structure during wet and dry spells extremes. The height of the ABL primarily depends on solar radiation, soil condition, and convective forcing. Persistent cloud cover, coupled with wet soil, in the wet spell reduces the surface temperature (Figure 3.2a) as well as the boundary layer height. On the other hand, absence of deep clouds and dry soil increases the sensible heat flux during dry spells (Kusuma 1986) and thereby the surface temperature and the ABL height. Figure 3.4: Same as figure 3.2, but for the altitudes of (a) ABL, (b) 0 C isotherm, (c) LRT and (d) CPT. The 0 C isotherm level is an important level in meteorology, because several microphysical processes occur in the vicinity of this level, for example, melting/freezing of hydrometeors below/above this level. The latent heating associated with these processes drives circulations at several scales. Although, the 0 C isotherm level exhibits large height variations in the mid and high latitudes (both on diurnal and seasonal scales), the variation in tropics is not much. Rao et al. (2008b) observed a variation of about 500 m at Chennai (~120 km from Gadanki) during the course of the annual cycle. Figure 3.4b shows that variability of the 0 C isotherm level from wet to dry spell is also not much at Gadanki. The mean height of the 0 C isotherm is equal in wet and dry spells (5.1 km). On the other hand, the day-to-day variability of 0 C isotherm level is considerable in all years (not shown). The range 54

Chapter 3 Thermal structure during wet and dry spells of variation is nearly same in wet and dry spells with >90% of population is in the height range of 4.8-5.6 km The tropical tropopause layer (TTL) is a demarcation layer separating the convectively buoyant troposphere from the stratosphere and is also important for its role in the exchange between the troposphere and stratosphere (Highwood and Hoskins 1998). Several tropopause definitions and their identification procedures exist in the literature. In the present study, only the traditional lapse rate tropopause (LRT) and cold point tropopause (CPT) levels are considered. The CPT is the height at which the coldest temperature is observed in the height region of 14-19 km. The LRT is defined, following the WMO definition, as the lowest height at which the temperature lapse rate reduces to 2 K km -1 and continues to be smaller than 2 K km -1 at least for the next 2 km. The frequency distributions for LRT and CPT in wet and dry spells are shown in figures 3.4 c and 3.4 d, respectively. The distribution plots and their mean values indicate that the CPT is ~600 m above the LRT in both spells. On the other hand, the mean heights of CPT (and also LRT) for wet and dry spells are nearly equal, i.e.16.8 km (16.1 km), i.e., there is no appreciable difference in the heights of CPT or LRT between the spells. Contrasting the significant levels (figure 3.4) and temperature and humidity fields at the surface and aloft (figures 3.2 and 3.3) during wet and dry spells indicates that the difference in the thermal structure between spells is large below the ABL height. 3.3.3 Variation of stability and instability indices: Development of convection depends on favorable environmental conditions. Several potential instability indices (Stability Index, CAPE, etc.) have been developed to measure the susceptibility of a given temperature and moisture profile to the occurrence of deep convection. Perhaps the most popular and widely used parcel instability parameter is CAPE. The computation methodology for CAPE and CINE is given in Appendix-I. The variabilites of the different parcel instability parameters (CAPE, CINE, and stable layers) from wet to dry spells are primarily discussed in this section. 55

Chapter 3 Thermal structure during wet and dry spells Figure 3.5: Same as figure 4, but for (a) LCL, (b) EL, (c) CAPE and (d) CINE. Figures 3.5a-d depict the frequency distributions of different thermodynamic parameters important for convection, such as Lifting Condensation level (LCL), Equilibrium Level (EL), CAPE and CINE, for wet and dry spells. Typically LCL represents the altitude of cloud base and EL the cloud top (or convective outflow level). In general, the LCL (or cloud base) appears at low levels during the wet spell in comparison with that of in dry spell. Quantitatively, about 71% (only 25%) of the LCL population is larger than 800 hpa during wet (dry) spell. The mean values of LCL for wet and dry spells are, respectively, 827 and 772 hpa. Contrary, the EL distributions for wet and dry spells show an opposite trend with the clouds reaching higher altitudes in wet spells than in dry spells, corroborating Rao et al. (2009) s observations of draft cores in wet and dry spells. About 42% of the EL population during wet spells shows values smaller than 200 hpa (in other words, crossed the altitude of 200 hpa), but only 11% of population in the dry spells reached that altitude. The mean altitudes of EL for wet and dry spells are, respectively, 227 and 314 hpa. It is clearly evident from the distributions of LCL and EL that the clouds are deeper in wet spell than in dry spell. Figure 3.5c shows that the CAPE distribution and mean CAPE values are different in different spells. Large values of CAPE are seen more frequently in wet 56

Chapter 3 Thermal structure during wet and dry spells spell. For instance, ~32% of the CAPE population in wet spells shows values larger than 2000 J kg -1, but only 6% of the population in dry spells exceeds that threshold. The mean CAPE value for wet spells (1627±1411 J kg -1 ) is larger than for dry spells (608±823 J kg -1 ). The standard deviation of CAPE (indicates the combined variability of CAPE within the spell and between spells) is also large in both spells. Note that the CAPE is estimated from the sounding data available at a fixed time, i.e, ~17:00 LT. Some of the soundings, therefore, represent conditions before active convection and some after convection and others at the time of convection. This is perhaps the main reason why we see significant variation in CAPE values within the spell. The distributions for CINE look similar in both spells (figure 3.5d). The mean CINE for wet and dry spells are also nearly equal (53.3±41.2 and 44±41 J kg -1 for wet and dry spells, respectively). Like in the case of CAPE, in both spells, the variability within the spell is comparable to the mean value of CINE for that spell. The range of CINE observed at Gadanki is comparable to that observed over other tropical continental stations (Singapore, Yi and Lim, 2004) but smaller than that observed over oceans (Bhat et al. 2001). Although the mean values of CINE are small, they can be considerable on individual days. The CINE acts as a lid to the growth of convection and it must be overpowered either by forced lifting or by the buildup of latent instability (by heating the surface layer). Note that the present observations were taken during the dusk hours and therefore the second possibility can be ruled out. If we don t consider any forced lifting, an updraft of ~9 m s -1 is required to overcome CINE of ~40 J kg -1. Existence of such strong vertical velocities in non-convective periods is rare. Therefore, in both spells, some external forcing is required to overcome this inhibition energy and to trigger convection. Contrasting our results with other studies that dealt with CAPE differences in active and break spells reveals interesting facts. The CAPE values estimated over tropical oceans (Bay of Bengal and western Pacific) are in contrast to the present observations, with large values in break phase (before convection) and small in active phase (Bhat et al. 2001; 2003; Mapes and Houze 1992; Williams and Renno 1993; McBride and Frank 1999). The regions influenced by maritime and continental flows also show similar features with larger values of CAPE in mesoscale convective systems associated with the break phase than those with active phase, for example 57

Chapter 3 Thermal structure during wet and dry spells over Darwin in northern Australia (Cifelli et al. 1998) and over Brazil (Halverson et al. 2002). The main physical reason for the lower values of CAPE in active spells is associated with the decrease of equivalent potential temperature ( e ) due to the overturning of the atmosphere during convection (McBride and Frank 1999). The convective downdrafts induced by precipitation loading, melting and evaporation bring the relatively dry mid-tropospheric air into the boundary layer and reduces e near the surface. In addition, the latent heat release in deep clouds warms the middle and upper troposphere and reduces the instability of the atmosphere. All these factors reduce the CAPE during active phase of the monsoon or during the convectively active period. However, at Gadanki, variation in thermodynamical parameters from wet to dry spells is greater near the surface and in the boundary layer. Above the boundary layer and in the middle troposphere (~2.5-11.5 km), the temperature difference between the composites for wet and dry spells is within 0.5 K (Figure 3.3). Also, in contrast to the expected smaller e in wet spell, larger e values are observed during wet spell in the lower troposphere. Since the diurnal variation of rainfall is pronounced at Gadanki (Rao et al. 2004), it is possible that the rainfall can occur after the balloon ascent and the environment that the balloon is sampling could be pre-convective. To examine whether this is responsible for large CAPE in wet spell, the variation of CAPE is plotted with respect to the rain start time for wet and dry spells (figure 3.6). Also shown in the figure (in color) is the amount of rainfall (in mm) occurred in each rain event. Interestingly, large CAPE values are observed when the rain occurred before the balloon launch time in both wet and dry spells. Contrary, when the balloon samples pre-convective environment (i.e., balloon ascent is before the rain start time), CAPE is found to be small in both spells. It is true irrespective of the rain amount. For example, largest values of CAPE are observed when rain occurred in the afternoon-early evening period (14:00-16:00 LT). In some of those events, considerable rainfall occurred (>10 mm). To study quantitatively the CAPE differences between spells, data are grouped based on balloon ascent time with reference to the rain start time into 4 categories (no rain, before rain rain occurred before the balloon ascent, during rain balloon ascent is during the rain, and after rain rain occurred after the balloon ascent). 58

Chapter 3 Thermal structure during wet and dry spells Figure 3.6: CAPE variation with reference to the start time of rain in (a) dry and (b) wet spells. The color indicates the amount of rain occurred during that rain event. Table 3.1 shows the mean and standard deviation of CAPE and e for the above 4 categories for wet and dry spells. It can be noticed from the table that mean CAPE values are larger in wet spell than in dry spell irrespective of whether the rain exists or not and also whether the balloon ascent is before/during/after the balloon launch period. Also, the e values are larger in wet spell than in dry spell in all 4 categories, indicating that the atmosphere is relatively moist and cool during wet spell irrespective of the rain occurrence. 59

Chapter 3 Thermal structure during wet and dry spells Table 3.1: Mean CAPE and e (and standard deviation) for 4 categories (no rain, before rain, during rain and after rain) during wet and dry spells. CAPE (J kg -1 ) e (K) Wet Dry Wet Dry No rain 1874.3 ± 1259.9 545.5 ± 812.2 346.3 ± 4.9 343.3 ± 3.8 Before rain 2166.7 ± 1707.5 873.7 ± 710.4 348.7 ± 3.2 344.2 ± 3.2 During rain 1178.8 ± 1176.5 283.6 ± 207.9 347.9 ± 4.9 343.8 ± 2.1 After rain 1127.8 ± 817.0 518.8 ± 496.6 347.7 ± 2.8 343.8 ± 3.6 To examine the vertical extent and magnitude of the buoyancy during wet and dry spells, vertical profiles of parcel positive buoyancy (based on air-parcel lifted from the surface) for all days (in each spell) are plotted in figure 3.7. The figure clearly depicts major differences and similarities in buoyancy profiles between wet and dry spells: 1) The vertical extent of positive buoyancy profiles is deep during wet spells, while most of the buoyancy profiles during dry spells are limited in vertical extent, 2) The magnitude of the positive buoyancy generally increases as the vertical extent increases, 3) The negative buoyancy (gap areas in positive buoyancy profiles) areas are limited during wet spell, while they are seen in most of the profiles during dry spell. The negative buoyancy areas are mostly centered on two height regions, ~700 and ~500 hpa during the dry spell. 4) In both spells, maximum virtual temperature excess is found in the middle and upper troposphere (above 600 hpa), consistent with McBride and Frank (1999). Figure 3.7 shows several negative buoyancy areas during dry spell and most of them are seen at 700 and 500 hpa. Stable layers can produce such negative areas in buoyancy profiles. Although strong convection penetrates the stable layers, sometimes these layers force weak-moderate convection to detrain mass and prevent further development. 60

Chapter 3 Thermal structure during wet and dry spells Figure 3.7: Positive thermal buoyancy (in K) profiles during dry (top panel) and wet spells (bottom panel). Following Johnson et al. (1996), the vertical distribution of % occurrence of temperature lapse rates exceeding certain thresholds (3, 4, and 5 K km -1, all are smaller than the moist atmospheric lapse rate in the troposphere, ~6 K km -1 ) during dry and wet spells are shown in Figures 3.8a & 3.8b. The % occurrence at each altitude is estimated from the ratio between the number of occurrences when the lapse rate exceeding a threshold and the total number of lapse rate data points at that altitude. The statically stable layers are predominantly seen in the lower and middle troposphere (below 8 km). The % occurrence of stable layer distribution is somewhat similar in both the spells. The % occurrence distribution for 5 K km -1 temperature lapse rate shows a broad distribution between 2 and 7 km with two small peaks centered on 2-3 km and 5 km. While the distributions for other two temperature gradients show a single peak centered around 4.5 km in dry spell and slightly at a lower altitude in wet spell. Nevertheless, the magnitude of % occurrence is different in both spells. The peak in the height region of 2-3 km corresponds very well with the altitude of the ABL (Figure 3.4a), and the second peak with the 0 C isotherm level (figure 3.4b). 61

Chapter 3 Thermal structure during wet and dry spells Figure 3.8: Vertical distribution of % occurrence of stable layers defined by different temperature lapse rates during (a) dry and (b) wet spells. (c) Frequency distribution of temperature lapse rates during wet and dry spells in two height regions (2-3.5 km and 4.5-6 km). D and W denote dry and wet spells, respectively. We now examine the magnitude of stable layers in the height region of 2-3.5 km and 4.5-6 km (the height regions in which the % occurrence of stable layers is relatively more) during dry and wet spells (Figure 3.8c). The figure shows cumulative distribution of temperature lapse rates in the height regions mentioned above. In both height regions, the strongest stable layers (smaller lapse rates) exist in dry spell. Lapse rates of 2 K km -1 are also observed in dry spells. Contrasting the lapse rate distributions in these two height regions indicates that gradients near the freezing level are stronger than their counter parts at low-levels. This is true in both spells of the monsoon. 3.4. Discussion: In this section, the thermal structure of the atmosphere in wet and dry spells, discussed above, is utilized to understand the differences in observed draft core statistics in wet and dry spells reported by Rao et al. (2009). As seen in section 3.3 that the background conditions, (large CAPE values, humid weather and weak stable 62

Chapter 3 Thermal structure during wet and dry spells layers) are ideal during wet spells, therefore, convection once triggered can grow into a deep system during this spell. On the other hand, the presence of strong stable layers in two height regions below 6 km, lack of sufficient humidity above the ABL, and weak CAPE all retard the growth of convection in dry spells. These clouds therefore develop as non-precipitating shallow cumulus (fair weather cumulus) or precipitating shallow convection clouds. These observations are consistent with the results of Rao et al. (2009). They observed that the shallow convective cores are preponderant in dry spells (22 out of 25 are seen in dry spell). Figure 3.9: Time-height distribution of vertical air velocity (w) on 24 July 2008 (representing wet spell) and on 07 June 2008 (representing dry spell) during the passage of convection. The white area on 07 June 2008 represents the gap in the data due to technical problems. The other intriguing observation by Rao et al. (2009) is the difference in the vertical draft structure in wet and dry spells. The composite w profile for majority of the draft cores in wet spell show a single peak, while two-peaks in dry spell. Typical examples of deep cores in both spells are shown in figure 3.9. It shows time-height sections of w on 24 July 2008, representing a wet spell, and on 7 June 2008, representing a dry spell. It is evident from the figure that the updraft core exists in the upper troposphere-lower stratosphere during 22:30-22:50 LT on 24 July 2008, while 63

Chapter 3 Thermal structure during wet and dry spells is present in the entire troposphere during 20:50-21:50 LT on 7 June 2008. The radar data were not available during 21:50-22:25 LT due to technical problems. Figure 3.10: Vertical profiles of (a) composite w and (b) buoyancy during the passage of convection in wet and dry spells. The composite w profiles for wet and dry spells are shown in figure 3.10a. They look similar to the profiles reported in Rao et al. (2009) with a bi-modal distribution (one peak ~5 km and the other peak at 11-12 km) in dry spell. The lowlevel peak is absent in the composite profile for wet spell. The buoyancy profiles on 24 July 2008 and 07 June 2008 are consistent with corresponding w profiles with bimodal distribution in dry spell (not as clear as in w profile) and single peak in wet spell (figure 3.10b). Most of the buoyancy profiles in dry spell (figure 3.7) also show bi-modal distribution either with reduced buoyancy or with negative buoyancy areas centered at two heights, ~700 and ~500 hpa. In other words, the negative area/reduced buoyancy at ~500 hpa separates the positive buoyancy profile and is responsible for the observed bi-modal distribution in composite w profile. This negative buoyancy is due to the presence of stable layers at this level (see figure 3.8c, maximum % occurrence is at this altitude). However, the sources for the negative area or for the existence of stable layers (among others, advection, radiation, large-scale subsidence, internal gravity waves, etc.) need to be studied further. 64

Chapter 3 Thermal structure during wet and dry spells 3.5. Summary and Conclusions: The variability in the thermal structure of the atmosphere from wet to dry spell over Gadanki, a station located in a region which shows contrasting rainfall features when compared to all-india rainfall, is studied using 4 years of high-resolution upperair measurements from GPS sondes and surface observations from AWS. The variation of meteorological variables at the surface and aloft (T, q and e), meteorologically significant levels (ABL, 0 C isotherm level, LRT, and CPT), and stability and instability parameters for convection (CAPE, CINE and stable layers), from wet to dry spells are studied to better understand the differences in draft core statistics reported by Rao et al. (2009). The variation of near surface T and q from wet to dry spells is quite pronounced with the mean values in these spells differing by ~1.6 K in T and 2 g kg -1 in q. These large variations in T and q are seen in the entire boundary layer. The atmosphere is more humid and cool in the wet spell than in dry spell, except at a T q e values are small. The presence of stable layers associated with the ABL top is found to be responsible for this distribution. T q e shows large values below 2 km, indicating that evaporation of surface moisture may be primarily responsible for those large values. Nevertheless, the role of advection (which is unclear at present) and subsidence cannot be ruled out. Differences in significant levels between spells are also consistent with the differences in temperature and humidity discussed above, i.e., large differences in those parameters are seen only in the lower troposphere. For example, the ABL altitude show significant variations from wet to dry spell with a mean ABL height difference of 400 m. Persistent cloud cover and wet soil (as can be evidenced from continuous rainfall) during wet spell reduce the surface temperature and also height of the ABL. On the other hand, not much variation is observed in mean values of the melting level, LRT and CPT from wet spell to dry spell. Distributions of CAPE for wet and dry spells show large overlap, but the mean values differ by ~1000 J kg -1 (1627 and 608 J kg -1 in wet and dry spells, respectively). Although the range of CAPE values observed at Gadanki are comparable to those 65

Chapter 3 Thermal structure during wet and dry spells observed over warm oceans (Kingsmill and Houze 1999; McBride and Frank 1999; Bhat et al. 2003) and land stations (Cifelli et al. 1998; Halverson et al. 2002; Yi and Lim 2004), the variation of CAPE from wet to dry spells is not consistent with earlier studies. The CAPE values estimated over the Bay of Bengal and western Pacific are larger in break phases than in active phases (Mapes and Houze 1992; Cifelli et al. 1998; McBride and Frank 1999; Bhat et al. 2003), in contrast to the present results. The CAPE variation with respect to the start time of rain is studied in wet and dry spells. The results clearly show that CAPE is larger in wet spell than in dry spell, irrespective of whether the rain occurred before/after/during the balloon launch period. Surprisingly, large CAPE values are observed after the rain occurrence in both wet and dry spells. Reduction of e following the convection (because of downdrafts) is not observed in both spells. In fact, larger values e are observed in before-rain category (i.e., when rain occurred before the balloon launch) than in other categories. During dry spells, several negative buoyancy areas are seen above the LFC. These areas are concentrated primarily in two height regions, ~2.5 and 5 km, and are found to be associated with stable layers. Stable layers near the 0 C isotherm level (at 5 km) appear to be a common phenomenon at Gadanki, as they are seen in nearly 40% of data irrespective of the phase of monsoon. The strongest stable layers are, however, observed during dry spell. The convection growth is limited in dry spells due to the presence of strong stable layers, weak CAPE and a relatively dry environment. This is perhaps, the main reason, for the frequent occurrence of shallow draft cores in dry spell at Gadanki (Rao et al. 2009). Even the deep cores show differences in vertical draft structure between spells (figures 3.9 and 3.10). The single peak in the vertical profile of composite w during wet spell and two peaks in the dry spell are consistent with the buoyancy distribution in different spells. This chapter clearly shows that the difference in the thermal structure between wet and dry spells is significant in the lower troposphere (< 2-3 km). The vertical buoyancy profiles are also different in different spells of the Indian summer monsoon. Nevertheless, these studies are based on radiosondes observations over Gadanki. Therefore, it is not clear whether the observed variations are localized features or prevalent over the entire southeastern India? This issue is addressed in the next chapter. 66

Chapter 4 Energitics during wet and dry spells.

Chapter 4 Energetics during wet and dry spells 4.1 Introduction: Convective Available Potential Energy (CAPE) is often regarded as a potential indicator for convective instability of the atmosphere and also for storms intensity (Emanuel 1994). In addition to assess how intense the storm is, the CAPE is widely used as an indicator for global climate change (Ye et al. 1998, Gettleman et al. 2002, DeMott and Randall 2004), for studying the trends in the instability of the atmosphere (Goswami et al. 2006; Mani et al. 2009; Goswami et al. 2010), for predicting lightning intensity (Williams et al. 2002) and for estimating precipitation (McBride and Frank 1999). Given its wide utility in modeling and experimental studies, it is being studied extensively. A few studies exist over the Indian region discussing the variability of CAPE over different temporal scales and relating this variability to precipitation (Bhat et al. 2001, Murthy and Sivaramakrishnan 2006, Goswami et al. 2006, Bhowmik et al. 2008, Alapattu and Kunhikrishnan 2009, Goswami et al. 2010). These studies advanced our understanding of CAPE-convection relations. Most of the studies mentioned above, on CAPE focused on its variation over seasonal, annual and interannual scales over the Indian land mass and adjoining oceans. None of the studies dealt with the intraseasonal variability (associated with active and break spells) of CAPE over southeastern India. For the first time a comprehensive study on the intraseasonal variability (associated with wet and dry spells) of CAPE, over Gadanki has been done. It is observed that large variability and intriguing differences in the vertical structure of buoyancy in wet and dry spells of the monsoon. However, it is not clear from the above study whether the observed variability of CAPE between spells at Gadanki is localized or seen over the whole southeast peninsula. In the present study the study region is extended from Gadanki to southeast India by using the Reanalysis/model data sets. Several models can be used for this purpose, be it reanalysis products - ECMWF and NCEP or high-resolution regional forecast models WRF model. However, identification of the most appropriate model to represent the observed CAPE at Gadanki is essential as the same model output is used to study the CAPE variation over a large domain. Therefore, the main purpose of this article is to: 1) find out which model estimated CAPE (ECMWF and NCEP reanalysis and WRF) approximates the sonde estimated CAPE and shows similar 68

Chapter 4 Energetics during wet and dry spells variation over different temporal scales to that observed by sonde, 2) extend the study region from Gadanki to the southeast peninsular India to study the variability of CAPE between spells, 3) examine whether the differences in the vertical structure of buoyancy in wet and dry spells are confined to Gadanki region or spread over the whole southeast India? 4) What physical/dynamical mechanisms explain CAPE differences between spells and how these differences affect the convection growth in different spells? The chapter is organized as follows. In Section 4.2 a brief description of data utilized in the present study is given. Section 4.3 discusses the comparison of model estimated moist instability parameters, over different temporal scales, with those estimated by radiosonde at Gadanki. It also compares CAPE dependence on lifting air parcel s condition as seen by models and GPS sonde observations. The fractional energy contribution of different layers in the atmosphere to the CAPE as measured by models and sonde is also compared. Section 4.4 highlight differences in CAPE and vertical structure of buoyancy between spells over southeast India and discusses possible causative mechanisms. The results are summarized in section 4.5. 4.2 Data base: In the present study, GPS radiosonde (Vaisala/Meisei) measurements, ECMWF and NCEP reanalysis and WRF model data are utilized to estimate the buoyancy and CAPE. A complete discussion about the GPS radiosonde, reanalysis and forecast data are given Chapter 2. The time of balloon launch coincides with the convection occurrence period (i.e., afternoon to midnight) at Gadanki (Rao et al. 2004). Therefore, the estimated CAPE value represents pre-storm environment for some days and post-storm environment on other days. The present study is not intended to evaluate the performance of different models. To do that, we need to match resolution (particularly vertical resolution) of models and observations. This study, rather, tries to see which model mimics the observations reasonably, so that the model data can be utilized to extend the study region from Gadanki to southeast India. The computational methodology of various stability indices is given in Appendix -I 69

Chapter 4 Energetics during wet and dry spells 4.3. Comparison of CAPE estimated by radiosonde and models: It is always difficult to compare observations from a single point with regional averaged and smoothed datasets generated by models. Therefore, the comparison of CAPE estimations is made in a statistical sense in addition to one-to-one comparison for individual days. Different methods and indices are employed for comparison of CAPE estimates which include day-to-day variations, correlations, histograms, differences between spells, CAPE dependency on air parcel s thermal properties and level of origin, contribution of different layers to total CAPE. 4.3.1 Time series and statistical comparisons: Figure 4.1: Comparison of LCL, LFC, EL and CAPE at Gadanki as estimated by GPS sonde, ECMWF-interim, NCEP and WRF data. 122 data points correspond to one monsoon season (June September). Total data are for 4 years (2006-2009). Time series plots of LCL, LFC, EL and CAPE obtained by radiosonde and different models show large day-to-day variability during 4 monsoon seasons considered in the study. None of these estimates show any increasing or decreasing trend with time as monsoon evolves in any of the 4 monsoon years considered here. Rather, Figure 4.1 shows periods (3-10 days) of enhanced and reduced CAPE values. 70

Chapter 4 Energetics during wet and dry spells Qualitatively, model estimates of LCL and LFC seem to vary in tandem with those estimated by radiosonde. However, all models seem to overestimate the EL height with reference to radiosonde estimated EL (seen more clearly in 2007-2008). As a result, models overestimate CAPE on many days. It is interesting to note some of these differences between models and sonde estimated CAPE values. Models predict relatively large (>1000 J kg -1 ) values of CAPE when sonde estimated CAPE is small (< 200 J kg -1 ) on many days in 2007 and 2008. Only on a few days, sonde estimates are larger than model estimates. However, when we consider periods (few days together) of large CAPE and small CAPE, then the agreement between models and sonde estimates is better. For instance, both models and sonde estimated large values of CAPE during middle of June 2006, September 2006, July-August 2008, etc. Also, both ECMWF and NCEP show small-to-moderate correlations with sonde estimated CAPE values with a correlation coefficient of 0.37 and 0.34, respectively. While, the correlation between WRF and radiosonde CAPE is much smaller (0.21). Figure 4.2: A statistical comparison (in terms of histograms) of LCL, LFC, EL and CAPE at Gadanki, as estimated by GPS sonde, ECMWF-interim, NCEP and WRF data. Since, total number of soundings by sondes and models are not same, occurrence percentage is plotted instead of count/frequency. 71

Chapter 4 Energetics during wet and dry spells To better understand the statistical differences in the above parameters estimated by models and sonde, histograms of above parameters are plotted in figure 4.2. The WRF model and sonde estimated LCL and LFC match well, with a peak in their distributions at 790 and 775 hpa for LCL and LFC, respectively. The ECMWF and NCEP models predict the height of LCL and LFC at a lower altitude than WRF and sonde by 50-75 hpa. Nevertheless, considerable population in the distribution of LCL and LFC for sonde show higher cloud base and LFC than model estimates. On the other hand, the WRF model overestimates EL by 50 hpa relative to other estimates. Though the peaks in the distribution of EL for other models and sonde match, considerable population for sonde show cloud top at a lower altitude. Lower cloud base and higher cloud top estimated by models in comparison with sonde estimates results larger CAPE values by models (relative to sonde values). It is seen in the distributions of CAPE for models and sonde. The mean values shown in table 4.1 also confirm that, on average, models overestimates CAPE. It is pronounced in the case of WRF CAPE estimations, even though its LCL and LFC estimations closely match with sonde estimates. Nevertheless, the standard deviation for CAPE (represents the combined intra- and inter-seasonal variability) estimated by radiosonde is larger than that estimated by models. NCEP (ECMWF and WRF) estimated mean (standard deviation) CAPE is nearer to radiosonde estimated CAPE (standard deviation). Table 4.1: The average LCL, LFC, EL and CAPE values along with standard deviation (SD) estimated by radiosonde and different models. The value in the parenthesis represents the number of vertical profiles used for the estimation of mean values. Parameter Radiosonde ECMWF NCEP WRF (396) (482) (487) (112) LCL 796.4 ± 51.3 835.2 ± 50.2 847.6 ± 43.3 807.4 ± 37.7 LFC 753.5 ± 65.5 792.6 ± 60.6 815.3 ± 44.1 783.2 ± 33.6 EL 279.8 ± 107.1 225.1 ± 63.5 235.2 ± 48.4 195.9 ± 38.6 CAPE 971.2 ± 1111.1 1352.4 ± 823.0 1182.4 ± 652.5 1638.7 ± 859.2 72

Chapter 4 Energetics during wet and dry spells 4.3.2. Comparison of models sensitivity to lifting parcel properties: To evaluate the ability of models in simulating relationship between the CAPE and air parcel s thermal properties at the surface, correlation analysis was done between them using model and radiosonde data. Figure 4.3 shows scatter plots between the CAPE and air parcel s surface equivalent potential temperature ( e ) and specific humidity (q), estimated by model and radiosonde data. The quasi-linear variation of CAPE with e and q by both models and radiosonde reiterates earlier finding that CAPE depends strongly on lifting air parcel s thermal properties (Mapes 1993; Williams and Renno 1993, Ye et al 1998, Alapattu and Kunhikrishnan 2009). The correlations between the CAPE and e (and q) are significant (at 99% level). Models provide better correlations than the radiosonde with large correlation coefficients and small standard deviations (see Table 4.2). Among different models, the linear relationship between CAPE and parcel s e and q is weak for WRF and strong for ECMWF. The slope, which indicates mean variation of the CAPE with 1 C (or 1 g/kg) change in e (q), is also included in Table 4.2. Figure 4.3: Scatter plot between CAPE and parcel s surface equivalent potential temperature (left panel) and also specific humidity (right panel), showing the dependency of CAPE on lifting air parcel s thermal properties. The correlation coefficients are also shown in the figure. 73

Chapter 4 Energetics during wet and dry spells The slope, which indicates mean variation of the CAPE with 1 C (or 1 g/kg) change in e (q), is also included in Table 4.2. The slope of the regression line between the CAPE and e (and q) estimated by different models is found to be within 80 J kg -1 K -1 (90 J kg -1 /g kg -1 ) from that obtained by radiosonde data. On average, the CAPE is found to vary 100-200 (200-400) J kg -1 for a 1 C change in e (q) consistent with earlier findings in tropics (Williams and Renno 1993, Alapattu and Kunhikrishnan 2009). The scatter plot between radiosonde-derived CAPE and e (and q) shows wide scatter around the linear fit compared to model plots. Because of that the RMSE between the linear fit and data is found to be larger than that obtained by models. e (and also q) using different model outputs and radiosonde data. The correlation coefficient (significant at 99% level), RMSE (estimated from the linear fit and data points), and the slope of linear fit are also included. CAPE vs. e r RMSE Slope (J kg -1 /K) CAPE vs. q r RMSE Slope (J kg -1 /K) NCEP 0.80 390 122 ± 4 0.68 477 235 ± 11 ECMWF 0.85 443 180 ± 5 0.82 487 351 ± 11 WRF 0.70 614 193 ± 19 0.69 617 399 ± 39 GPS 0.45 998 116 ± 12 0.57 916 326 ± 24 The strong dependence of CAPE on surface e (and q) means that the height from which the parcel rises also becomes important, because the thermal properties change with altitude. In earlier studies, the thermal parameters of lifting air parcel for CAPE estimation are taken either at the surface (in this case the parcel is assumed to be lifted from the surface) or averaged over the mixing depth. Though a well mixed boundary layer is supposed to have nearly same thermal characteristics within the boundary layer, often thermal profiles show small gradients. The vertical variation in thermal parameters is more pronounced between the surface layer and rest of the mixed layer. Also now it is known that air parcels rose from above 900 hpa level rarely become buoyant in the tropics (Holton, 1992; McBride and Frank, 1999). 74

Chapter 4 Energetics during wet and dry spells Therefore, to gain better insight into the relationship between the CAPE and parcel s level of origin and also to examine how this relationship is reproduced by models, CAPE is estimated by models and radiosonde data by considering the air parcel s level of origin at different pressure levels (surface, 900 and 850 hpa) (Table 4.3). Table 4.3: Comparison between models and observations on their dependency of CAPE on air parcel s lifting height (pressure level). GPS ECMWF NCEP WRF 960 900 850 960 900 850 960 900 850 960 900 850 LCL 796 778 767 826 806 783 848 817 787 807 783 768 LFC 753 727 715 784 757 730 815 778 741 783 747 724 EL 280 298 312 223 239 257 235 290 334 196 230 252 CAPE 971 646 547 1338 992 694 1182 643 354 1639 964 695 It is evident from Table 4.3 that both models and observations (radiosonde) show reduction in CAPE with increase in the height of air parcel s origin. The heights of LCL and LFC (EL) increased (decreased) with the air parcel s lifting height. In other words, the cloud vertical extent (and also the instability) reduces if the parcel is lifted from a higher altitude. Both models and GPS sonde measurements show this feature, though the differences in instability parameters, obtained by varying the air parcel origin, are more pronounced in the model data. For instance, all models show a reduction of about 300-400 J kg -1 as the parcel s lifting level changed from 960 hpa to 900 hpa and to 850 hpa. In terms of percentage variation, CAPE varies by about 30-60% as the level of air parcel s origin changes from 960 hpa to 850 hpa. On the other hand, sonde-derived CAPE varies by ~16-40%. One interesting observation is that the % variation in GPS-derived CAPE is larger when the parcel origin is varied from 960 to 900 than from 900 to 850. While such a distinct variation in CAPE with parcel origin is not seen in any model data. It implies that the mixing layer depth and the mixing with in it, mainly forced by surface heating, may not be similar in models and observations (GPS). The other notable observation is, as also seen in figures 4.1 and 4.2, that the WRF-derived LCL and LFC are close to their 75

Chapter 4 Energetics during wet and dry spells GPS counterparts, no matter what ever may be the height of parcel origin. In the case of EL, NCEP is close to the observed value. 4.3.3. Comparison of Intraseasonal variability and vertical structure of buoyancy and CAPE: The average CAPE for wet and dry spells, estimated with radiosonde data at Gadanki, is different for different spells (Chapter 3). It is therefore important to check whether or not models reproduce this feature. Figure 4.4 shows the comparison of histograms of CAPE, separately (grouped using the methodology described in section 4. 2), for wet and dry spells estimated with radiosonde and model data Figure 4.4: Histograms of CAPE (at Gadanki) in (a) dry and (b) wet spells, respectively, as estimated by GPS sonde measurements and model outputs, illustrating the differences in CAPE between spells. Overestimation of CAPE by models in comparison with observations is observed in both wet and dry spells. The overestimation is more in the case of WRFestimated CAPE, as also seen in figure 4.2. Though models overestimate the CAPE, but they reproduce differences in the CAPE between spells accurately. It is clearly evident from figure 4.4 that the mean (or mode) of CAPE distribution is larger in wet spell than in dry spell (see Table 4.4 for mean CAPE values). All models show this feature, albeit with different magnitudes of CAPE variation between spells. For instance, the CAPE difference between wet and dry spells is nearly equal (~950 J kg - 1 ), when estimated with GPS and ECMWF data. The NCEP- and WRF-estimated mean CAPE values show smaller difference between spells (~650 J kg -1 -NCEP and 76

Chapter 4 Energetics during wet and dry spells ~770 J kg -1 -WRF). Other notable feature is large variation of CAPE within the spell (in terms of standard deviation) by both models and observations. Nevertheless, the standard deviation is small when estimated with model data. In other words, models underestimate the CAPE variation within the spell in comparison with GPS observations. Table 4.4: Mean CAPE and standard deviation for wet and dry spells as estimated by GPS sonde and different models. The value in the parenthesis indicates number of CAPE values averaged to obtain the mean value. Dry Wet GPS 609.0 ± 822.6 (168) 1592.9 ± 1410.3 (78) ECMWF 954.4 ± 717.0 (201) 1919.9 ± 702.0 (92) NCEP 938.1 ± 579.1 (201) 1583.7 ± 661.5 (92) WRF 1180.5 ± 809.6 (64) 1950.5 ± 948.9 (24) McBride and Frank (1999) have shown that while CAPE is strongly controlled by the properties of the boundary layer air, large positive buoyancy and realization of CAPE, however, occurs above 600 hpa. To check how far models reproduce this feature, the positive buoyancy area is divided into 3 layers (L1:700-500 hpa, L2:500-300 hpa and L3:300-200 hpa) and CAPE is estimated for each layer (hereafter called layered CAPE). This exercise also helps us in understanding which layer is contributing more to total CAPE in wet and dry spells. First, CAPE for each layer is estimated from each sounding and the contribution of each layer to total CAPE is estimated. This data are then grouped based on wet and dry spells and the % occurrence for the layer contribution to CAPE is estimated for each layer separately (Figure 4.5). Figure 4.5 shows some intriguing similarities and differences in CAPE between spells, between layers and also between models and observations. In general, both models and observations show similar distribution of CAPE contributions by different layers. Among layers, contribution of middle layer (L2: 500-300 hpa) to CAPE is more (~50%) in majority of the cases in both spells, consistent with McBride and Frank (1999). The contribution of L1 to CAPE is about 77

Chapter 4 Energetics during wet and dry spells 20% (20-30%) and remaining is by L3 in the wet (dry) spell. It is also noted that the distribution in L1 is broader during dry spell than in wet spell. Figure 4.5: Contribution of CAPE in different layers (L1: 700-500 hpa, L2: 500-300 hpa, L3: 300-200 hpa) to total CAPE in wet (left panel) and dry (right panel) spells at Gadanki. GPS sonde measurements and NCEP, ECMWF-interim and WRF model outputs are used in this analysis. Also, the contribution of L1 (and to some extent L2) to CAPE is found to be as high as 100% in dry spell, albeit the occurrence percentage of such cases is less. It means that the vertical extent of clouds is limited in dry spell. This is consistent with reported draft core statistics over Gadanki (i.e., 85% of total shallow draft cores are in dry spell) (Rao et al. 2009). All models reproduced above features, nevertheless, contributions of different layers to total CAPE are observed to be different by different models. For example, the contribution of L1 (L3) to CAPE is relatively small (large) by WRF. Since WRF overestimates EL (figures 4.1 and 4.2), the % occurrence and contribution of L3 to CAPE is more in comparison with the contribution of L3 estimated by other models and observations. It is clear from the above analysis that models reproduce similar CAPE variation between spells as obtained by radiosonde measurements. To examine how 78

Chapter 4 Energetics during wet and dry spells well models replicate differences in the vertical buoyancy structure between spells, the percentage occurrence of single- and double-peaked profiles are estimated using GPS sonde measurements and model outputs. To avoid small fluctuations in buoyancy profiles to be called as peaks, 7 th order polynomial is fitted to buoyancy profiles. Then, peaks in the buoyancy profile are identified from polynomial fits using the following criterion. A peak is considered, if the buoyancy at any height is higher than that observed above and below (± 20 hpa) that height. Table 4.5 shows the percentage occurrence of single and double peaks in wet and dry spells as obtained by the radiosonde and model data. Clearly, both observations and models noted the double-peaked structure in the buoyancy predominantly in dry spell. While majority of buoyancy profiles in the wet spell have only a single peak. It is observed from the data (not shown here) that the single peak in the wet spell is observed mostly in the middle and upper troposphere (above 500 hpa). Among the two peaks during dry spell, the lower peak is observed below 500 hpa, while the upper peak is seen above the 500 hpa level. 4.4. Variation of CAPE and buoyancy between dry and wet spells over southeast India and associated implications: From the Section 4.3 it is clear that the agreement between model-derived CAPE and radiosonde-derived CAPE is moderate on a day-to-day scale, but fairly good at intraseasonal scale (or averaged over wet and dry spells). Also, frequent occurrence of double-peaked buoyancy profile during dry spell is predicted reasonably well by models. The above analysis, therefore, indicates that the model data can be used to extend the study region for studying CAPE (and buoyancy) variation, at least, at intraseasonal scales. Though any model data can be used for this purpose, ECMWF-interim data are used here as it shows similar CAPE variation between spells as produced by the sonde. To examine whether the bimodal (single-peak) buoyancy distribution during dry (wet) spell is a localized phenomena confined to Gadanki region or exists over the entire southeast India, individual buoyancy profiles at each grid point are subjected to the analysis described in section 4.3.3. 79

Chapter 4 Energetics during wet and dry spells Table 4.5: Percentage occurrence of single- and double-peaked buoyancy profiles in wet and dry spells as estimated by GPS sonde and model data. Dry Spell Wet spell Total Single Double Total Single Double GPS 160 42% 58% 74 70% 30% ECMWF 201 39% 61% 92 64% 36% NCEP 196 35% 65% 92 75% 25% WRF 57 42% 58% 23 74% 26% Table 4.6 shows differences in the vertical structure of buoyancy between wet and dry spells at each grid point over the southeast India. It is evident from Table 4.6 that at all grid points, except for two grid points (15 N, 79.5 E and 15 N, 81 E), majority of buoyancy profiles show two peaks during dry spell. On the other hand, single peak buoyancy profiles are observed at most of the grid points in wet spell (12 out of 20). Overall, bimodal buoyancy profiles are dominant in dry spell (68%), while single peak buoyancy profiles are seen prominently in wet spell (53%), indicating that the difference in the vertical structure of buoyancy between spells is not confined to Gadanki, rather a characteristic feature over the entire southeast India. Figure 4.6: Spatial distribution of mean CAPE in (a) dry and (c) wet spells, respectively, over the southern peninsular India. The spatial map of standard deviation of the mean CAPE in (b) dry and (d) wet spells, respectively, are also shown in the figure. The dot denotes the location of Gadanki. 80

Chapter 4 Energetics during wet and dry spells Table 4.6: The percentage occurrence of single and double peaks in buoyancy profiles during wet and dry spells at different grid points. The analysis is performed using ECMWF interim data. % Occurrence of peaks Grid Points in dry spell in wet spell (Lat, Lon) Single Double Single Double 9, 76.5 19 81 50 50 10.5, 76.5 18 82 25 75 12, 76.5 19 81 26 74 13.5, 76.5 28 72 26 74 15, 76.5 42 58 30 70 9, 78.0 23 77 59 41 10.5,78.0 35 65 79 21 12, 78.0 37 63 48 52 13.5, 78.0 26 74 35 65 15, 78.0 29 71 27 73 9, 79.5 20 80 56 44 10.5,79.5 30 70 85 15 12, 79.5 37 63 62 38 13.5, 79.5 38 62 64 36 15, 79.5 62 38 67 33 9, 81.5 13 87 56 44 10.5,81.5 23 77 64 36 12, 81.5 32 68 71 29 13.5, 81.5 38 62 64 36 15, 81.5 70 30 73 27 Figure 4.6 shows the CAPE distribution over the southeast India in different spells of the monsoon. It is apparent from the figure that CAPE is, in general, large (of the order of 1000) over this region in both spells reinforcing the common view that the tropical atmosphere (land areas and also warmer parts of the oceans) possesses large energy (Williams and Reno 1993). The standard deviation (SD) of the 81

Chapter 4 Energetics during wet and dry spells mean CAPE is large at many grid points, particularly in dry spell, indicating large variability of CAPE within the spell. The CAPE variation between spells is large over most part of the region considered in the present study, consistent with radiosonde derived CAPE variations at Gadanki (Chapter 3, and also Figure 4.4 and Table 4.4). In this region, CAPE is higher in wet spell than in dry spell by as high as 1000 J Kg -1. Generally, the destabilized atmosphere stabilizes by releasing CAPE during disturbances (for instance during convection). The lower tropospheric cooling by downdrafts and evaporation, and latent heating in deep clouds in mid- and upper- troposphere reduces the CAPE substantially. Therefore, one would expect high CAPE before convection or during dry spells and small CAPE during disturbed periods or in wet spells as reported by earlier studies (Williams and Renno, 1993; Cifelli et al. 1998; Bhat et al. 2001). Kennan and Carbon (1992), on the other hand, observed not much variation in CAPE between different spells of the Australian monsoon. In contrast, present observations, interestingly, show considerable difference between spells, with higher CAPE in wet spell than in dry spell. Therefore it is imperative to understand why CAPE is low in dry spell and how these differences in CAPE between spells affect the convection growth? Let us examine plausible mechanisms for high CAPE in the wet spell. Firstly, the soundings (by both radiosonde and models) used for CAPE estimation are made at 12 UT. Given the diurnal variation of rainfall during the southwest monsoon, which shows two peaks (evening and midnight) (Rao et al. 2004), it is possible that the sounding may be representing the pre-storm environment (if midnight peak is dominant in wet spell). In such a case, high CAPE could exists during wet spell. This issue is examined in chapter 3 at Gadanki, by grouping the soundings (and CAPE values) with reference to the time of rainfall occurrence. Higher CAPE is observed in wet spell than in dry spell, irrespective of whether the sounding is made before/during/after the rainfall. Surprisingly, CAPE in the post-storm category is found to be higher than in pre-storm category in both spells. It indicates that the timing of sounding is not the real cause for the observed CAPE variation between spells. 82

Chapter 4 Energetics during wet and dry spells Second possibility for high CAPE in wet spell (and also in post-storm cases) is associated with quick build-up of the instability. Bhat et al. (2001) have shown that the atmosphere needs at least 2 days to build-up the instability through fluxes over the ocean. However, over the land, we can expect a rapid build-up of instability through surface fluxes (moisture and heat). At Gadanki, the afternoon rain is mostly shortlived even in wet spell (not shown here). Immediately after the rainfall (storm), the air is rapidly moisturized again by the evaporation of surface moisture. In chapter 3 it is observed that the atmosphere is more humid in wet spell than in dry spell in the lower atmosphere. In the absence of any large water bodies nearby and that the wind direction remained same (westerly) in both spells, the differences in surface moisture fluxes can only explain the observed differences in humidity profiles between the spells. Therefore, the rising air parcel in wet spell (and also in the post-storm environment) may be more moist (or having large e). As seen in section 4.3.2, the cool and moist air parcel, if lifted, produce higher CAPE, as observed in the present study. Thirdly, relatively weak CAPE in dry spell could be due to small surface e. Even during convection, the ambient warm and dry atmosphere during dry spell evaporates the rain and induces large downdrafts in the lower atmosphere, and thereby lowers e (or reduces CAPE). As a result, the surface rain amount is also very small in dry spell. Rao et al. (2009) observed large downdrafts below 2-3 km in draft cores and small rainfall amount during dry spell, consistent with the above explanation. Rao et al. (2009) reported intriguing differences in draft cores between wet and dry spells, with shallow cores preferentially occurring in dry spell (for ex., 22 out of 25 are observed in dry spell). Figure 4.6 in the present study also supports core statistics reported by Rao et al. (2009). Here, we discuss how the observed difference in CAPE between spells is affecting draft cores, particularly in dry spell. Though substantial CAPE exists, on occasions, in dry spell, the deep convection (and heavy rain) is observed rarely. Even if convection breaks out, the omnipresence of persistent and strong inversion layers below 5 km (chapter 3) prohibits the growth of deep convection in dry spell. Strong inversion layers means more work is necessary to lift the near-surface air to its LFC. In addition, high temperatures and relatively dry 83

Chapter 4 Energetics during wet and dry spells atmosphere in dry spell enhances the evaporative cooling and thereby reduces the positive buoyancy acquired by the parcel at the LFC. At times, the buoyantly growing cloud becomes neutrally buoyant fossil. The other possibility for the occurrence of shallow cores in dry spell is associated with strong shears. Strong low-level winds and shears in the zonal wind are observed during dry spell (Chapter 5). The shear layer is also deep in dry spell (up to 6-8 km in dry spell against ~4 km in wet spell). Low-level shear, generally, is conducive for the longevity of the storm. However, strong shear can suppress convective growth through shearing of the weakly buoyant updrafts (Weisman and Rotunno, 2004). The above mechanisms seem to be occurring in dry spell and are perhaps the reason for the frequent occurrence of shallow cumulus clouds without substantial rainfall in dry spell. 4.5 Conclusions: In the present chapter three important key questions are addressed; 1) Whether CAPE estimations at Gadanki using model and radiosoundings data agree well enough for the former to be useful in extending the study region to the southeast India? 2) Whether the observed differences in CAPE and the vertical structure of draft cores between spells are localized or characteristic features of wet and dry spells over a large domain (i.e., southeast India)? 3) Which dynamical/microphysical process explains the observed CAPE difference between spells and how this difference affects the growth of convective clouds, particularly in dry spell? To obtain an answer to the first question, a detailed comparison of CAPE estimates obtained with model and radiosonde data is made. Though the agreement and correlation between model and sonde estimations are moderate, they improved when comparisons were made on longer temporal scales (i.e., wet and dry spells). Nevertheless, all models overestimate the CAPE on all temporal scales examined here (daily and also intraseasonal). But models reproduce similar intraseasonal variation of CAPE, similar dependency of CAPE on surface thermal parameters and similar intraseasonal variation of the vertical structure of buoyancy to that obtained by the radiosonde. The present study, therefore, suggests that the model data can be utilized 84

Chapter 4 Energetics during wet and dry spells to study CAPE variations (at intraseasonal and longer temporal scales) over a larger domain. The ECMWF-interim data are used to answer the second question. Intriguing differences are observed in CAPE estimates and also in the vertical buoyancy structure between spells. Substantial CAPE (~1000 J kg -1 ) is observed in dry spell, reinforcing the view that the tropics always possess considerable energy (Williams and Reno 1993). Interestingly, CAPE is found to be higher in wet spell than in dry spell over the southeast peninsular India. The difference is as high as 1000 J kg -1 at some grid points in the study region. Further, majority of buoyancy profiles show only one peak during wet spell, while bi-modal vertical distribution is seen predominantly in dry spell. The above analysis clearly suggests that the differences in CAPE and buoyancy vertical structure between wet and dry spells are not only confined to Gadanki (Rao et al. 2009), rather observed all over southeastern peninsular India and they are characteristic features of wet and dry spells. Several possible physical/dynamical mechanisms are invoked to explain observed CAPE differences between spells, i.e., time of sounding with respect to the rain occurrence, rapid rebuild-up of the instability, moistening of the atmosphere due to the evaporation of surface moisture in wet spell, enhanced downdrafts engendered by evaporative cooling and drop dragging in dry spell. The weak CAPE in dry spell may not be sufficient to overshoot the frequent stable layers occurring in this spell. Further, the strong (magnitude) and deep (in height) low-level wind shear observed in dry spell seems to be shearing apart the weakly buoyant updraft. The above mechanisms seem to be occurring in dry spell limiting the growing draft cores, consistent with the observations (Rao et al. 2009). This chapter clearly brought out the differences in thermal and stability parameters over the southeastern peninsular India between wet and dry spells. Since, the thermal parameters show considerable variation between spells; their forcing on the atmosphere, will also be different. In other words, the mean and the variation of winds in these spells could be different. The next chapter, therefore, discuss the vertical structure of the mean wind and its diurnal variation in different spells of the monsoon. 85

Chapter 5 Dynamics during wet and dry spells.

Chapter 5 Dynamics during wet and dry spells 5.1 Introduction: The Indian summer monsoon rainfall exhibits significant intraseasonal variation (ISV) with pulses of quasi-periodic active (vigorous convection) and break (subdued convection) cycles. In response to the active and weak convective activity, the wind pattern also changes in these spells. A few studies focussed on the intraseasonal variability of winds over the Indian region (Chen and Yen 1991; Joseph and Sijikumar 2004, hereafter JS04; Sathiyamoorthy et al. 2007, hereafter S07). These studies mainly focussed on the variability of the LLJ and TEJ, both are integral parts of the monsoon circulation. For instance, JS04 noticed that the core of LLJ (on 850 hpa level) passes from the central Arabian Sea through the southern peninsular India during all-india active spell, but migrates to south during break spell. (i.e., from 12.5º-17.5º N during active spell to 2.5º-7.5º N). The TEJ (on 200 hpa level) also shifts by ~1000 km between active and break spells in the entrance region of TEJ, i.e., east of 70º E. On the other hand not much ISV in TEJ is seen west of 70º E (S07). While ISV in the mean wind, particularly the LLJ and TEJ, was studied to some extent (Chen and Yen 1991; JS04; S07), but the diurnal variability of LLJ and TEJ during different spells of the monsoon is not well documented. The only study available in the literature on diurnal variation of TEJ is by Krishnamurthi and Kisthawal (2000). Nevertheless, their study was made with limited dataset and confined only to active spell of the monsoon. On the other hand, significant variation in the wind, on a diurnal scale, can be expected between spells. The diurnal processes forced by the diurnal cycle of solar radiation and land-sea/mountain-valley circulations may not change intraseasonally. However, the convection and its induced circulation can produce a different diurnal cycle of winds in wet spell from that of in dry spell. Moreover, the difference in soil condition between the spells induces a different diurnal cycle in surface and low-level winds (Zhong et al. 1996). Further, any change in the synoptic wind flow between the spells may cause a different interaction between the synoptic flow and local circulations, like land-sea breeze and mountain-valley winds. For instance, a weak synoptic flow allows the opposing sea-breeze to intrude deep inland, while a strong synoptic flow may not so. The first objective of this study is, therefore, to study 87

Chapter 5 Dynamics during wet and dry spells differences in the diurnal variation of winds at the surface and aloft (below 19 km) between different spells of the monsoon. The mean axis of LLJ and TEJ over the Indian land mass passes through the southern peninsular region (between 12º -16º N) (Findlater, 1971; JS04; S07). Fortunately, the National Atmospheric Research Laboratory, Gadanki, which hosts a suite of instrumentation, is located in this region. This institute boasts of having, among others, IMSTR, LAWP, SODAR and AWS for surface and upper-air meteorological observations. The ability of above instruments in providing 3D wind measurements continuously in virtually all-weather conditions allows us to study the diurnal variation of winds in different monsoon spells. Earlier a few studies utilized the IMSTR measurements to understand the differences in convection characteristics between spells (Rao et al. 2009). They noticed intriguing differences in the occurrence of shallow draft cores and in the vertical structure of deep draft cores between spells. Shallow cores are found predominantly in dry spell with nearly 88% of total cores are observed in this spell. The deep draft cores show a bi-modal distribution in dry spell with one peak at ~5 km and the other in the upper troposphere, while a single upper tropospheric peak is seen in wet spell. In Chapter 3 considerable differences in the surface and lower tropospheric (up to the boundary layer top) meteorological parameters (temperature and humidity) between wet and dry spells are observed. A significant large variation in the boundary layer height and, condensation level, and also in CAPE between spells is noticed. These observations clearly indicate that the thermal structure, available energy and their forcing are different in different spells. Since, above parameters and their forcing vary diurnally (for instance convection), one would expect a significant change in the diurnal variation of wind from one spell to other. Therefore, in the present study differences in the vertical structure of mean wind and its diurnal variation between wet and dry spells from the surface to lower stratosphere with a special emphasis on the LLJ and TEJ, using measurements from a suite of instrumentation available (mentioned above) at Gadanki and ECMWF interim data, are studied. Brief description of the data obtained from different instruments is discussed in section 5.2. The vertical variation of zonal and meridional winds in different 88

Chapter 5 Dynamics during wet and dry spells monsoon spells is discussed in section 5.3. It also discusses the spatial variability of the LLJ and TEJ in different monsoon spells. The differences in the diurnal variation of winds (both zonal and meridional) between spells are also documented in this section with a special emphasis on the LLJ and TEJ. The results are summarized in section 5.4. 5.2 Data and Instrumentation: Majority of the data were obtained by collocated instruments available at Gadanki. They include AWS for surface meteorological parameters, a Doppler SODAR for low-level winds (60 m 1 km), LAWP for winds in the height region of 600 m 4 km, IMSTR for winds in the height region of 3.6-19 km. These instruments provide wind measurements continuously and therefore provide a unique dataset for studying the diurnal variation of winds. The important specifications of the radar systems are given in Table 1. The detailed description of above systems (including sub-systems), and data processing algorithms are extensively discussed in Chapter 2. However, the above instruments were not operated simultaneously, as they were developed/installed in different years (see Figure 5.1). The IMSTR measurements were spanned over ~14 years (September 1995 September 2009). However, the radar operates in diurnal mode (24-hr operation) only for 2 days in a month. Therefore, the number of available days to study the diurnal variation is limited, though the measurements were spanned over many years. Even on those days, the radar operates for about 20 minutes in each hour. The 20-min measurements were averaged and considered as the representative of that hour. The LAWP operates continuously (virtually unattended) switching between two modes of operation (low and high) (see Rao et al. 2008 for more details). For the present study, the high mode measurements (with 150 m vertical resolution) were employed. The LAWP was operated for 2 monsoon periods (1999 and 2000). The Doppler SODAR also operates continuously with very high vertical (30 m) and temporal (~30 sec) resolutions. The wind measurements from Doppler SODAR were available for two years (2007 and 2008). The AWS provides measurements at 1 hr interval and the data were available for 4 years (2006-2009). The accuracy of measured wind speed and direction are 0.1 m s -1 and 0.1, respectively. To study the spatial variability of jets, ECMWF interim 89

Chapter 5 Dynamics during wet and dry spells data with 1.5º x 1.5º resolution (during the period 2006-2009) were employed (Uppala et al. 2008). Figure 5.1: Number of diurnal cycle days in wet and dry spells obtained by different instruments (a) in each year and (b) from all years. Note that the y-axis in (a) is shown in logarithmic scale. 5.3 Results and Discussion: 5.3.1 Differences in mean wind between wet and dry spells at the surface and aloft: The mean vertical structure of wind during dry and wet phases of the monsoon over Gadanki is depicted in figure 5.2. Since measurements by IMSTR, LAWP, SODAR and AWS were not simultaneous (fig. 5.1) and also some interannual variability may exist in winds, composite vertical profiles (by combining different datasets from different years) were not constructed. Rather, figure 5.2 shows average wind variation between spells in different altitude ranges (surface - obtained from AWS; 60 m to 1 km - obtained from SODAR; 600 m to 4 km - obtained from LAWP; and 3.6 km to 19 km - obtained from IMSTR). The average wind for any particular spell by any instrument is obtained by averaging all diurnal cycle data in that spell. In other words, the diurnal and semi-diurnal effects are removed and the average wind is a true representative of that spell. Vertical profiles of mean zonal (left panel of figures 5.2a-c) and meridional (right panel of figures 5.2a-c) winds corresponding to dry and wet phases show intriguing differences between spells. The surface mean winds (obtained from AWS) in wet spell are weaker than that of in dry spell, but the difference in both zonal and meridional winds between spells is not significant. 90

Chapter 5 Dynamics during wet and dry spells Figure 5.2: Vertical profiles of zonal (left panel) and meridional (right panel) winds for wet and dry spells obtained from different instruments ((a)-mst radar, (b)- LAWP, (c)-sodar and AWS). The average wind shown at 0 km (or surface) in figure (c) is obtained from AWS. Vertical profiles shown in figure (a)-(c) are daily averages. The standard deviation is represented by error bars. It is because of large wind variability in both spells caused, primarily, by large diurnal variation of winds (shown later). But when we examined the magnitude of winds, it is larger in dry spell than in wet spell in all hours and this variation is significant. The wind direction, however, remained the same in both spells, i.e., southwesterly. The winds in the height region of 60 m 1 km, derived from SODAR, remain northwesterly to westerly, but show large vertical variation in the magnitude, particularly the zonal component. Peak zonal winds are observed in the height region of 200-500 m in both spells. Later, it will be shown that this enhancement in zonal wind is mainly due to formation of the Nocturnal Low Level Jet (NLLJ) in this altitude range. Zonal winds are stronger in dry spell than in wet spell particularly in the height region of NLLJ; On the other hand, the meridional winds are generally weak in magnitude (~ 1 m s -1 ) in both spells and do not show any significant differences between spells. 91

Chapter 5 Dynamics during wet and dry spells The LAWP derived winds, in the height region of 600 m 4 km, continued to be in westerly to northwesterly in both spells. The LLJ is clearly seen in the zonal wind component and interestingly the height of LLJ peak is found to be different in both spells (at 2.25 km and 1.35 km in dry and wet spells, respectively). The magnitude of the LLJ is also different in different spells with stronger winds in dry spell (16.8 m s -1 ) than in wet spell (9.8 m s -1 ). The difference in the zonal wind between spells is pronounced above 1.5 km. Later, it will be shown that the weak wind in wet spell is due to migrations of the LLJ to south in that spell. The meridional winds are relatively weaker and do not show significant variation between spells. IMSTR winds reveal that the vertical structure of wind in the height region of 3.6-19 km is different in both monsoon spells. The differences, however, are pronounced in the lower and middle troposphere (below 8 km). The zonal wind profiles show typical summer monsoon circulation with low-level westerlies and strong upper tropospheric easterlies. The depth of westerlies or the height of wind reversal (i.e., from westerly to easterly), however, is different in both monsoon spells. The depth of westerlies is relatively shallow (~ 6 km) in wet spell with zonal wind reversal occurring below 6 km. On the other hand, during dry spell, the zonal wind shows westerly winds up to an altitude of 8 km and then turns to easterly due to the presence of TEJ. In the upper troposphere, the TEJ strength is found to be nearly same in both spells with an average value of ~30 m s -1. The height of TEJ maximum is also found to be nearly same in both spells (~16 km). However, the width of TEJ (at any wind speed between 15-30 m s -1 ) is larger in dry spell than in wet spell by 500-1000 m. Roja Raman et al. (2009) also found similar results on TEJ magnitude and width, though they employed a different criterion to identify the active and break spells. Note that meridional winds are generally weak (± 2 m s -1 ) in both spells; therefore, zonal winds dictate the wind vector. 5.3.2 Differences in the spatial distribution of LLJ and TEJ between wet and dry spells: The two conspicuous features of the Indian monsoon circulation are the presence of LLJ and TEJ (Krishnamurthi and Bahlme, 1976), both are seen prominently over the southern peninsular India. This section discusses differences in 92

Chapter 5 Dynamics during wet and dry spells the spatial variability of these jet streams between spells, using 4 years (2006-2009) of ECMWF (ERA-Interim) analysis data. Since the LLJ and TEJ are predominantly zonal wind streams (see figure 5.2), only zonal wind is considered for this analysis. The zonal wind data are stratified into two categories, wet and dry, following the criteria discussed in section 5.2. Composites of LLJ and TEJ for wet and dry spells and the wind anomaly (mean wind for dry spell mean wind for wet spell) are estimated. It is important to remember that the spatial variation of LLJ and TEJ in different spells is with respect to the spells over the southeast India. In other regions, the wet and dry spells need not follow the spells over the southeast India. Our idea is to study the spatial variability of LLJ and TEJ when the convection is weak or active over the southeast India. Figure 5.3 shows the mean zonal wind for dry and wet spells (a and b, respectively) and the difference in zonal wind between spells (c) on 850 hpa level, depicting spatial variation of the LLJ in those spells. The spatial distribution of standard deviation of means for dry and wet spells is shown in figures 5.3d and 5.3e, respectively. In both wet and dry spells, LLJ is prominently seen, albeit with different magnitude and spatial variation. The core of LLJ is seen over the Arabian Sea running parallel to the African coast, and its magnitude is relatively larger in dry spell than in wet spell. In dry spell (figure 5.3a), the core of LLJ splits over the Arabian Sea with one branch (say first branch) passing over the southern peninsular region centred around 16º N and the other towards southeast and veers cyclonically (second branch) and merges with the first branch in the Bay of Bengal (near the coast of Malay peninsula). The width or the latitudinal extent of the first branch of LLJ is larger than that of second branch. On the other hand, only one branch (second branch) is present in the wet spell (figure 5.3b). The splitting of LLJ in the Arabian Sea is reported earlier by many investigators and it was attributed to barotropic instability (Findlater 1971). However, JS04, while studying the LLJ characteristics in active and break spells of the monsoon, have not seen any splitting in any of the spells. They, rather, observed a strong LLJ over the southern peninsular region in active spell and southward-migrated LLJ in break spell. They attributed the migration of LLJ to the movement of convectively active zones. They, therefore, argued that the splitting is due to monthly 93

Chapter 5 Dynamics during wet and dry spells mean winds taken by Findlater (1971) for his analysis, which includes both active and break spells of the monsoon. Figure 5.3: Mean zonal wind on 850 hpa level for (a) dry and (b) wet spells showing the spatial variation of LLJ. The black solid line in figures (a) and (b) represents 10 m s -1 contour. (c) The difference in LLJ between spells (dry-wet). The dot denotes the location of Gadanki. (d) and (e) shows the spatial distribution of standard deviation of mean values for dry and wet spells, respectively. In the present study, not only the splitting of LLJ is clearly evident in dry spell (analogous to all-india active spell), but also the axis of first branch of LLJ over Indian longitudes is also close to that reported by Findlater (1971). In addition, JS04 have seen another weak branch of LLJ in northern India at 25º N, which is not seen in figure 5.3. Nevertheless, the southward migration of LLJ in break spell as reported by JS04 is also seen in our study. In fact, this branch (second branch) is present in both spells with similar magnitude, as evidenced by the small wind anomaly in that region (figure 5.3c). It is important to remember that the spell identification technique used by JS04, depends on the 15 m s -1 threshold magnitude of mean zonal wind averaged over the region 10º N - 20º N and 70º E - 80º E, is different from that used in this study. Therefore, it is not clear whether the observed differences (between different reports) are due to different techniques or different data sets employed in these studies. 94

Chapter 5 Dynamics during wet and dry spells Figure 5.3c clearly shows that large differences exist between spells in LLJ magnitude and its spatial variation. A band of large positive wind anomaly (i.e., stronger winds in dry spell than in wet spell) passes over the Arabian Sea, Peninsular India, Bay of Bengal and Malaysia with a maximum (> 6 m s -1 ) over the Southern Peninsular region. This large wind anomaly is significant and is occurring mainly due to the absence of first branch of LLJ in the wet spell. A negative anomaly in zonal wind is also observed in two regions, just south of the equator and near foot hills of the Himalayas. In general, the low-level westerly winds turn cyclonically in the North Bay of Bengal and become easterlies. These easterlies are clearly seen in the dry spell (or all India active spell) (figure 5.3a). As the monsoon trough moves northward to foot hills of the Himalayas during wet spell (or all India break spell), the easterlies seen in that region are replaced by westerly winds (figure 5.3b). In fact, the easterlies are completely absent over the Indian land mass during wet spell. Earlier studies have shown that the TEJ exhibits significant ISV, although the source of variability is highly debated (Koteswaram 1958; Chen and Yen 1991; S07). Many opined that the LLJ and TEJ are coupled through cumulus convection (Indian monsoon rainfall) occurring in north India (Koteswaram 1958; S07), while Chen and Yen (1991) have demonstrated that the ISV is controlled by eastward propagating planetary scale divergent circulation. Figure 5.4 shows the spatial variation of mean zonal wind and standard deviation on 100 hpa level corresponding to wet and dry spells along with zonal wind difference between the spells. In both spells, strong easterly winds are seen prominently between 10º and 20º N with winds as large as 30 m s -1. The longitudinal extent of TEJ in dry spell is found to be more than that of in wet spell. It can be evidenced by 28 m s -1 (thick solid line) and 24 m s -1 contours (thin solid line) of TEJ. For instance, the 24 m s -1 contour spreads over the longitudes 40º E -100º E in dry spell, while only between 50º E - 95º E in wet spell. Sathiyamoorthy et al. (2007) observed significant ISV in TEJ axis on 200 hpa level. They observed the axis of the TEJ is along 15º N in break spell, while a shift in the TEJ axis is observed in active spell, i.e., along 15º N over west of 70º E and along 5º N over east of 70º E (entrance region of jet). Such a north-south shift in TEJ axis is not observed in either of the spells in the present study. The axis is found to be aligned along ~15º N latitude in both spells. 95

Chapter 5 Dynamics during wet and dry spells Figure 5.4: Same as Figure 5.3, but for zonal winds at 100 hpa level, showing the TEJ variation. The black thin and thick solid lines in figures (a) and (b) represent 24 m s -1 and 28 m s -1 contours, respectively. 5.3.3 Diurnal variation of winds at the surface and aloft: To quantify and understand the diurnal variation of winds from the surface to lower stratosphere in different monsoon spells, hourly composites of zonal and meridional winds are estimated for wet and dry spells. Figures 5.5 and 5.6, respectively, show the diurnal variation of zonal and meridional winds at the surface (obtained from AWS) and in different height regions (60 m 1 km; 600 m 4 km and 3.6 km 19 km) in wet (right panel) and dry (left panel) spells. The diurnal variation of surface winds (denoted as sfc ) is shown in the SODAR wind plot. The wind variation within a day is considered as the magnitude of diurnal cycle in the present study. The zonal wind shows large diurnal variation at all heights (in particular below 8 km) in both spells and also between spells. The zonal wind maximum is observed at different times at different altitudes in both spells. 96

Chapter 5 Dynamics during wet and dry spells Figure 5.5: Diurnal variation of hourly mean zonal wind from the surface to lower stratosphere in dry (left panel) and wet (right panel) spells obtained from several instruments. (a) and (b) shows the diurnal variation in the height range of 3.6-19 km with IMSTR. The solid, black-dashed and white-dashed lines on the figure represent 8 m s -1, 0 m s -1 (height of wind reversal) and -30 m s -1 contours. (c) and (d) same as (a) and (b) but for the height range of 600 m 4 km obtained with LAWP. The solid line represents the 15 m s -1 contour. (e) and (f) same as (a) and (b) but for the height region of 60 m 1 km obtained with SODAR. The solid line represents 3 m s -1 contour. The surface wind variation obtained from AWS (denoted as sfc ) is also included in the figure. Note the change in color scale between different plots. Both zonal and meridional winds (5.5e-5.5f and 5.6e-5.6f, respectively) exhibit a distinct diurnal cycle near the surface with strong winds during day and weak winds during night in both spells. The strengthening of the surface winds during day is a manifestation of the intense down mixing of momentum from upper levels due to solar heating. During the night, radiative (nocturnal) cooling in the boundary layer reduces the eddy viscosity and momentum transfer from upper levels and thereby the wind speed (Arya 2001, Yu et al. 2009 and references therein). However, the time of occurrence of maximum wind is slightly different in both spells. The surface zonal wind shows a broad noon peak (centred around 12 LT (LT-Local time = UT + 5.30 hrs)) in dry spell, while it maximizes at ~10 LT in wet spell. The mean surface wind in wet spell is not only small but also shows a weak diurnal cycle when compared to that in dry spell. For instance, variation of the mean 97

Chapter 5 Dynamics during wet and dry spells zonal wind in a day is ~2 and ~4 ms -1 in wet and dry spells, respectively. Further, the difference in wind between wet and dry spells is not same throughout the day, but is large in noon hours and small during night. The meridional winds show a similar diurnal cycle in both spells with a magnitude of ~1.5 m s -1. Figure 5.6: same as figure 5.5, but for meridional winds. The noon maxima and early morning minima in zonal and meridional winds are seen up to an altitude of ~200 m in both dry and wet spells (note the data gap between the surface and 60 m), as evidenced by SODAR time-height velocity plots (figures 5.5e-5.5f and 5.6e-5.6f). Above 200 m, the diurnal cycle is significant in the zonal wind with strong westerlies in the early morning period (03-04 LT) and weak westerlies (weak easterlies are observed at higher altitudes) in the evening. Similar diurnal variation is seen in both spells, albeit with different magnitudes of diurnal cycle and its vertical variation. In dry spell, the range of wind variation within a day is found to be increasing with the altitude, while it is nearly same (decreasing to some extent) with the altitude in wet spell. For instance in dry spell, the magnitude of wind variation is ~2.5 m s -1 at 400 m and 5 m s -1 at 600 m and above, while it is in the range of 2.5-3 m s -1 in wet spell. The meridional wind (figures 5.6e and 5.6f) also shows a clear diurnal cycle with day time northerly and night time southerly winds in both spells. 98

Chapter 5 Dynamics during wet and dry spells Though the magnitude of diurnal cycle is different, but the phase or pattern of the diurnal cycle is same in both spells, i.e., strengthening of westerly wind starts at 19 LT and attains peak value during 0300 0500 LT. This kind of nocturnal low level jet (NLLJ) variation is observed at many geographical locations and is an oftstudied phenomenon (Blackadar 1957, Whiteman et al. 1997, Banta et al. 2000, Karipot et al. 2009, Baas et al. 2009 and references therein). Earlier studies attributed the occurrence of NLLJ to various atmospheric processes; baroclinicity from horizontal temperature contrast caused by land-sea and/or topographic circulations (Holton 1967, Li et al. 1983) and inertial oscillation due to frictional decoupling (Blackadar 1957). In reality, multiple processes contribute to the formation of NLLJ. Though the other processes like sea-breeze and topography (Gadanki is located in a complex terrain surrounded by hillocks of altitude 400-600 m) can form NLLJ, the diurnal variation of NLLJ resembles to that produced by the inertial oscillation. That is, following the collapse of convective boundary layer after the sunset, turbulence dies out and the remnant of the mixed layer will be decoupled from the surface because of stable stratification in the night. The winds increase in this frictionless- height region and form the NLLJ. In the absence of friction, the Coriolis force induces ageostrophic circulation in clockwise direction. Figure 5.7: Hodograph between mean zonal and meridional winds, derived from SODAR, in (a) dry and (b) wet spells. The mean is taken over the entire spell and also in the height region of 300-500 m. 99

Chapter 5 Dynamics during wet and dry spells A hodograph is plotted separately for wet and dry spells (figure 5.7). The zonal and meridional winds, obtained with SODAR, in the height region of peak NLLJ (i.e., 300-500 m) are averaged and are used for the hodograph analysis. In both spells, a clear clockwise rotation of wind is observed, suggesting that the inertial oscillation produced by the frictional decoupling could be responsible for NLLJ at Gadanki. The zonal wind at the surface and aloft (in the NLLJ region) (figures 5.5e and 5.5f) is stronger in dry spell than in wet spell. The magnitude of the diurnal cycle is also large in dry spell. Earlier modelling and observational studies also noticed weaker diurnal cycle when the soil is wet (Zhong et al. 1996). In wet spell, most part of the available energy is converted to (latent evaporation of soil moisture) and soil heat fluxes rather than sensible heat flux. As a result, the turbulence intensity (and friction) reduces in the wet spell and thereby the inertial oscillation. Further, the large cloud cover in wet spell (not shown but can be expected to be more in wet spell as the rainfall is more in this spell) will reduce the insolation and thereby the available energy. All the above processes reduce the magnitude of diurnal cycle in wet spell. The diurnal cycle of zonal winds in the height region of 600 m 4 km (figures 5.5c and 5.5d) is striking in both spells. In fact, the magnitude of diurnal cycle is largest in this height region in the entire troposphere and lower stratosphere. For instance, the variation of mean zonal wind in a day is larger than 8 m s -1 in the height region of 1.5-2 km in dry spell and 1-1.75 km in wet spell. Relatively, magnitude of the meridional wind (< 5 m s -1 ) and its variation (< 2 m s -1 ) is weak in both spells. As also noted in section 5.3.1 (figure 5.2), the height at which the LLJ is observed is different in different spells. In dry spell, the LLJ peak is observed in the early morning (04-07 LT) in the altitude region of 1.2-2.4 km and shifts gradually and systematically in time with increasing altitude (figure 5.4c). For instance, large westerly winds are seen during 8-10 LT, 14-15 LT and 17-18 LT at 2.4 km, 3 km and 3.6 km, respectively. Such a systematic shift in LLJ peak with time and altitude is also observed in wet spell. Nevertheless, it is confined to < 2 km altitude. Above 2.5 km, winds are weak during noon and evening hours in wet spell in contrast to the strong winds in that height region in dry spell. The diurnal variation is, therefore, somewhat different in both spells above 2.5 km. 100

Chapter 5 Dynamics during wet and dry spells The LAWP derived zonal winds clearly show that the LLJ peak is at 1.65 km in the morning (05 LT) in dry spell, with a magnitude of ~22 m s -1. On the other hand, the LLJ is not only weaker in wet spell (~13 m s -1 ) but is also observed at a lower altitude (1.35 km) and at a later time (07 LT) than that of in dry spell. Though the magnitude of LLJ is smaller in wet spell, but the amplitude of diurnal cycle is comparable to that of in dry spell at the height of peak LLJ and its vicinity. Further, as discussed earlier, the LLJ peak not only shows systematic variation with time and height in both spells, but also in magnitude. For instance, during dry spell, the magnitude of LLJ peak is ~22 m s -1 at 1.65 km in the morning but reduces to ~16 m s - 1 at 3.6 km in the evening. Therefore, if any study intends to obtain LLJ characteristics from observations made at a fixed time of the day (say 17:30 LT), it probably ends-up with wrong inferences. For instance, if we consider 17:30 LT measurements alone for obtaining LLJ characteristics, then it leads to the underestimation of LLJ magnitude and overestimation of LLJ height. The winds above 3.6 km, obtained from IMSTR measurements (figures 5.5a- 5.5b and 5.6a-5.6b corresponding to zonal and meridional winds, respectively), show weak diurnal variation in comparison with that seen at lower altitudes. The zonal winds are strong, particularly above 10 km, because of the presence of TEJ (Rao et al. 2000). Meridional winds are not only weak in magnitude but also shows relatively weak diurnal variation (with peak-to-peak variation rarely exceeding 4 m s -1 ) in both spells. Similar to the diurnal variation at lower altitudes, variation of winds over the diurnal cycle is larger in dry spell than in wet spell in the lower and middle troposphere (up to an altitude of ~8 km). Another conspicuous feature is the distinctly different diurnal variation of wind reversal altitude in wet and dry spells. The wind reversal altitude is nearly same during course of the day and remained at an altitude of ~ 8 km (±0.5 km) in dry spell, while it varied considerably in wet spell. It is observed at a lower altitude in the early morning but raised gradually to ~ 7 km by noon and remained there till 23 LT. This figure reiterates the importance of diurnal variation of winds and supports the earlier conclusion that the diurnal variation of winds need to be properly accounted while characterizing winds in the lower and middle troposphere. 101

Chapter 5 Dynamics during wet and dry spells In the upper troposphere, in general, near 16 km, in particular, the winds are predominantly easterly (due to the presence of TEJ) with mean winds exceeding ~30 m s -1 in both spells. Unlike LLJ, the TEJ peak has shown small diurnal variation and remained at ~16 km in both spells. The diurnal pattern is also similar in both spells with large winds in the noon-night and weak winds in the morning. The magnitude of the diurnal cycle is found to be ~2-4 m s -1 in the vicinity of tropopause. The magnitude and pattern of the observed diurnal cycle at Gadanki is comparable to that reported by Krishnamurthi and Kishtawal (2000). They observed an amplitude of 2 m s -1 in the diurnal cycle over a large region, including the Arabian Sea, Central and southern India and Bay of Bengal. Note that they have seen this diurnal variation on 200 hpa level, which they considered for TEJ study. The present study shows that the magnitude of diurnal cycle is of the same order even at 100 hpa. Krishnamurthi and Kishtawal (2000) attributed it to the diurnal strengthening and weakening of Tibetan high circulation. The TEJ lying in the southern flank of this circulation also varies in tandem and undergoes similar diurnal variation. Just like in the lower and middle troposphere, the magnitude of diurnal cycle for meridional component is small (< 4 m s -1 ) in the upper troposphere. But, the meridional wind itself shows large vertical variation, though the magnitude is small, in the upper troposphere. The meridional wind is northerly in the vicinity of TEJ (14-16 km), but is southerly below and above this height region. 5.4 Spatial and diurnal variation of LLJ and TEJ in wet and dry spells: To study the spatial variation in the diurnal cycle of LLJ and TEJ in different spells ECMWF interim data were considered. To check whether this poor temporal resolution data (4 profiles per day) is sufficient to depict the diurnal variation or not, the wind anomalies obtained with ECMWF interim data are compared with radar wind anomalies. For this, ECMWF interim wind data available at 05:30, 11:30, 17:30 and 23:30 LT are averaged and the anomalies are estimated for synoptic hours. 102

Chapter 5 Dynamics during wet and dry spells Figure 5.8: Comparison of zonal wind anomalies in the overlapping region of different instruments (1 km for SODAR and LAWP and 3.6 km for LAWP and MST) in (a) dry and (b) wet spells. SODAR, IMSTR and LAWP measurements are represented with solid line with symbols and the ECMWF interim data are represented with a symbol. Figure 5.8 shows comparison between zonal wind anomalies, during wet and dry spells, obtained by SODAR, LAWP, IMSTR and ECMWF interim data in the overlapping regions (i.e., 1 km for SODAR and LAWP and 3.6 km for LAWP and IMSTR). Also, this comparison facilitates us to check whether or not different instruments operated in different years are reproducing the similar diurnal pattern in the overlapping region. It is clear from Figure 5.9 that different datasets (SODAR, LAWP, IMSTR and models) are generating similar diurnal variation, except at 1 km in wet spell. Given that the observations are from different years and models and observations represent different spatial scales, the agreement is fairly good. It indicates that the model data perhaps can be utilized to study gross features of diurnal variation (or for qualitative studies). Figures 5.9 and 5.10 (5.11 and 5.12) show the spatial and diurnal variation of LLJ (in terms of zonal wind anomaly on 850 hpa level) (and TEJ (zonal wind anomaly on 100 hpa level)) in dry and wet spells, respectively. The standard deviation values are overlaid as contours. Though large diurnal variations are seen over many 103

Chapter 5 Dynamics during wet and dry spells geographical regions (for ex., LLJ variations over Middle East and Pakistan), we focus more in the regions, where LLJ and TEJ are prominently seen (The Arabian Sea, Southern peninsular India and Bay of Bengal). Figure 5.9: The spatial variation of zonal wind anomaly for dry spell on 850 hpa level, showing the diurnal variation of LLJ. The ECMWF interim data at (a) 05:30, (b) 11:30, (c) 17:30 and (d) 23:30 LT are used to obtain the wind anomaly at the above synoptic hours. Over the southern peninsular India, the LLJ (figures 5.9 and 5.10) shows significant diurnal variation (anomalies are more than the standard deviation) in both spells with an early morning maxima and noon minima, while a weak positive anomaly is seen in the midnight and negative anomaly in the evening. Though similar diurnal variation is observed over most of the Indian land mass, but is more prominent over the eastern side of the southern peninsula. The east-west spatial variability of the diurnal cycle is not only observed over large land masses (like India) but is also seen over a small country like Sri Lanka. 104

Chapter 5 Dynamics during wet and dry spells Figure 5.10: Same as figure 5.9, but for wet spell. It is clearly evident from the figure that the amplitude of diurnal cycle is larger in dry spell than in wet spell. The features of similar diurnal variation and larger amplitude in dry spell are also observed with LAWP (figures 5.5c-5.5d). In contrast to the diurnal cycle over the peninsular India, the LLJ is strong in the evening and weak in the midnight over the west Arabian Sea (west of 60º E along 12º -15º N latitudes) in both spells. Over this region, the amplitude of diurnal cycle is not only weak but also do not show any variation between spells. The diurnal cycle of LLJ is weak over the Bay of Bengal in contrast to that seen over the Indian land mass and Arabian Sea. The diurnal variation of TEJ (figures 5.11 and 5.12) shows large spatial variability in both spells. In dry spell (figure 5.12), the TEJ is weak (large positive anomaly) in the morning and strong in the evening over the entire India. However, the TEJ shows north-south variation at 11:30 and 23:30 LT. At 11:30 LT, the TEJ is weak 105

Chapter 5 Dynamics during wet and dry spells Figure 5.11: Same as figure 5.9, but for TEJ (on 100 hpa level). Figure 5.12: Same as figure 5.11, but for wet spell. 106

Chapter 5 Dynamics during wet and dry spells over the southern Peninsula and strong over north India, while it reversed at 23:30 LT. In wet spell, on the other hand, the diurnal cycle is opposite in the north and south India. The TEJ is weak (strong) at 05:30 and 23:30 LT over north India (southern peninsula). The results are somewhat agreeing with those obtained by Krishnamurthi and Kisthawal (2000) (their study is based on TEJ variation on 200 hpa level, but not on 100 hpa level). The weakening of TEJ in the morning and strengthening in the evening/night in our dry spell (all India active spell) is consistent with their results. The amplitude of diurnal cycle (~2 m s -1 ) is also agreeing with their study. However, they observed the largest amplitude over the Arabian Sea between 0 and 10 N. In the present study, on 100 hpa level, the largest amplitude is observed over north India, but not over the Arabian Sea. Though both jets (LLJ and TEJ) are important components of the monsoon circulation and are, primarily, generated by strong pressure gradients between the Tibetan platue and Indian Ocean, active convection modulates (both their magnitude and spatial variability) these jets (JS04, S07). Therefore, to have a better handle on the diurnal variation of jets, it is imperative to understand the diurnal variation of convection and its spatial variability. Such attempts will be made in near future by combining TRMM rainfall products (such as 3G68), ECMWF winds and OLR data. 5.5 Summary and conclusions: Variation of the mean wind and its diurnal variation between dry and wet spells of the monsoon are studied, with a special emphasis on LLJ and TEJ. Observations made at Gadanki from a suite of instruments (AWS, SODAR, LAWP and IMSTR) are used for this purpose. To study the spatial variability of LLJ and TEJ in different spells, ECMWF interim data have been utilized. The important results obtained from the study are summarized below. At Gadanki, the surface and low-level (below 4 km) mean winds are found to be stronger in dry spell than in wet spell, but the direction of wind remained the same (westerly) in both spells. The depth of westerlies is, however, deeper in dry spell than in wet spell. The height of LLJ peak and its magnitude are different in different spells (they are, respectively 2.1 km and 18 m s -1 in dry spell and 1.35 km and 10 m s -1 in 107

Chapter 5 Dynamics during wet and dry spells wet spell). On the other hand, not much variation is seen in the magnitude and height of peak TEJ between spells. The spatial variation of LLJ depicts intriguing differences between wet and dry spells. In dry spell, the LLJ core is found to be splitting over the Arabian Sea into two branches with one branch passing over the Southern Peninsula and Bay of Bengal and the second branch veering cyclonically and joins the first branch in Bay-of Bengal. On the other hand, only the second branch is apparent in the wet spell. The splitting of LLJ observed over the Arabian Sea supports many earlier reports (Findlater 1971), but is in contrast to that reported by JS04. But this splitting is observed only in dry spell, but not on all days. Nevertheless, the present study confirms that the LLJ variability between spells is quite large, as reported by JS04. The strength and the axis of TEJ, does not vary much between spells. Nevertheless, the longitudinal extent of TEJ is different in both spells (more in dry spell than in wet spell). The present study has not shown any systematic shift in TEJ axis across 70º E longitude in all India active spell, as observed by S07. The vertical variation of the diurnal cycle is complex with the wind maxima shifts in time with height in the lower troposphere (< 4 km). Noon maxima and early morning minima are observed at the surface and also between 60 and ~200 m. Above 400 m, the westerly wind maxima shifts to early morning (03-05 LT) and minima to the evening in both spells. Though the phase of diurnal cycle is similar in both spells in the height region of 400 1000 m, the height at which the maximum wind is seen and the variation of magnitude of diurnal cycle with height is different in these spells. The clockwise rotation of winds (from hodograph analysis) suggests that the inertial oscillation produced by the frictional decoupling, proposed by Blackadar (1957), could be the causative mechanism for NLLJ at Gadanki. The wet soil and high cloud cover during wet spell seem to be responsible for the weak diurnal cycle at the surface and in the lower troposphere in this spell. The magnitude of diurnal cycle is largest in the height region of LLJ. The LLJ peak also shifts systematically with time and height in both spells. But such a systematic shift in LLJ peak is observed only below 2 km in the wet spell. The LLJ 108

Chapter 5 Dynamics during wet and dry spells peak varies not only with time and height, but also in magnitude, for ex., ~22 m s -1 at 1.65 km in the morning to ~16 m s -1 at 3.6 km in the evening in dry spell. Given such a large diurnal variation in LLJ parameters (height, time of occurrence and magnitude) one should be careful in obtaining LLJ characteristics from observations made at a fixed time. IMSTR measurements clearly show that the wind reversal altitude is different in different spells with westerlies prevailing below ~8 km throughout the day in dry spell. Contrary, in the wet spell, the depth of westerlies is shallow in the morning and deep during afternoon - night. The diurnal variation of winds in the upper troposphere is relatively weak. The magnitude and phase of diurnal cycle of TEJ are consistent with those reported by Krishnamurthi and Kishtawal (2000). The diurnal variation of LLJ is significant and exhibits large spatial variability both in the amplitude and phase. The diurnal amplitude of the LLJ is large over the Southern Peninsula and small over the Arabian Sea and Bay of Bengal. As also seen with LAWP measurements, the diurnal cycle of LLJ is weak in wet spell. On the other hand, the diurnal variation of TEJ is weak in both spells. This study, documents the intriguing differences in vertical structure of the mean wind and its diurnal variation between wet and dry spells of the summer monsoon. The causative mechanisms for the observed differences need to be investigated further to better understand the diurnal variation of winds, in particular, LLJ and TEJ. 109

Chapter 6 Rainfall Characteristics during wet and dry spells.

Chapter 6 Rainfall characteristics during wet and dry spells 6.1 Introduction: Active and break spells associated with the intraseasonal variations (ISV) during the Indian summer monsoon (ISM) are not coherent throughout the India rather they exhibit large spatial variations. Earlier studies (and also discussed in chapter 1 and chapter 2) clearly demonstrated that there exists an anti correlation in the rainfall pattern between the monsoon zone and the southeast peninsular India. During the break monsoon period, the monsoon zone gets less rainfall but good amount (~20-40 % of seasonal rainfall) of rainfall occurs over southeast India (Ramamurthy 1969, De et al. 1998, Gadgil 2003, Goswami 2005, Rao et al. 2009, Rajeevan et al. 2010). Also, rainfall increases near the foothills of Himalayas, following the northward movement of monsoon trough. Good amount of rainfall is also observed over northeastern regions of India during this period. During the active monsoon period, widespread rainfall occurs along the west coast of India and over the monsoon zone. The monsoon winds transport moisture from the Indian Ocean and Arabian Sea towards the Indian land mass and enhance rainfall activity along the west coast. Also, the convergence of westerly/southwesterly monsoon flow and easterly/southeasterly flow over the monsoon zone engenders the cyclonic vorticity (along the monsoon zone) and sets conducive environment for synoptic systems. Low pressure systems and depressions that are formed in the head Bay of Bengal advect over the monsoon zone and produce copious amount of rainfall over that region. Goswami et al. (2003) studied the occurrence frequency of the low pressure systems (LPS) (including depressions and cyclones) in different spells of the monsoon and noted that the frequency is ~3.5 times higher in active spell than in break spell. They also observed that the tracks of these LPS are clustered along the monsoon trough during active spell, while such clustering of synoptic systems is not apparent in the break period. The diurnal variation of rainfall over central India is found to be different in active and break spells with stronger diurnal cycle in dry spell than in wet spell (Singh and Nakamura 2010). Using TRMM measurements, they also observed a secondary peak in rainfall in the early morning during active monsoon period. In addition to the above studies, numerous reports exist in the literature describing the variability of rainfall at intraseasonal scales, their periodicity, intensity, 111

Chapter 6 Rainfall characteristics during wet and dry spells and propagation characteristics over the monsoon zone and also on the role of ISV in producing the interannual variability (see reviews by Webster et al. 1998, Goswami 1998, Goswami 2005, Kulkarni et al. 2011 and references therein). On the other hand, the southeastern peninsular India is in a rain shadow region during the southwest monsoon. Large scale weather systems, like depressions and low pressure systems that are conspicuous features of the monsoon over the monsoon zone are generally absent over southeast India. It is generally assumed that most of the rainfall occurs due to isolated thunderstorms or occasional passage of mesoscale convective systems over interior regions and due to sea-breeze circulations near the coast. Simpson et al. (2007) examined surface meteorological parameters at 3 locations along the southeast coast of India to quantify the rainfall associated with the sea-breeze circulation. They noticed that ~70-80% of total rainfall in the Chennai region during the summer monsoon season is directly related to the sea-breeze induced convection. But, these sea-breezes are generally confined to coastal zones (i.e., 10-20 km from the coast) with an occasional deep inland penetration (Suresh 2007). Given that the magnitude of synoptic flow is different in different spells, it is not clear how deep these sea breezes and their induced convection migrate over inland in different spells of the summer monsoon. While the rainfall variability over the monsoon zone, as mentioned above, is well documented in the literature, not many reports exist on the spatial distribution, characteristics and propagation of rainfall over southeastern peninsular India during southwest monsoon. One of the objectives of this study is to characterize rainfall systems during wet and dry spells of the monsoon (note that wet and dry spells are active and break spells for southeastern peninsular India). In particular, the aim is to obtain answers to the following questions: How much (what percentage of) seasonal rainfall occurs in each spell of the monsoon? What kind of rainfall occurs (convective or stratiform) in different spells of the monsoon? How different types of rain vary diurnally in different spells of the monsoon? The first question is partly answered in chapter 2. Recall figure 2.13, where it is shown that good amount of rainfall occurs along the west coast during dry spell (and also over the monsoon zone) and almost no rain on the eastern side of Western 112

Chapter 6 Rainfall characteristics during wet and dry spells Ghats in the southern peninsula (south of 15 N). This figure raises several important questions. What kind of physical/dynamical processes are responsible for the rainfall over southeast India during wet spell; sea-breeze induced convection (Simpson et al.2007, Suresh 2007) or passage of LPSs from Bay of Bengal similar to that occur over the monsoon zone, or passage of synoptic systems with the monsoonal flow. If later is responsible, then a logical question arises, why the rainfall systems, which produce good amount of rainfall along the west coast, are not propagating and/or translating to east during dry spell? How they propagate eastward during wet spell? Further, we have seen in chapter 5 that wind speed (and therefore wind shear) in the lower troposphere (< 3 km) is much weaker in wet spell than dry spell. Also, the shear layer is much deeper in dry spell. We, therefore, have to examine the interaction of cold pool (evaporatively cooled downdrafts in thunderstorms) with ambient vertical wind shear. Earlier studies have reported that the interaction between them controls the structure and life time of squall lines by triggering new convective cells in the downshear regime (Thorpe et al. 1982, Rotunno et al. 1988, Moncrieff and Liu 1999, Weisman and Rotunno 2004). Therefore, it needs to be seen how effectively and differently the above mechanism works in the propagation of convective systems during wet and dry spells? It also needs to be checked whether the cold pool-shear interaction can explain the observed differences in rainfall distribution between spells? This chapter is organized as follows. The data used in the present study are briefed in section 6.2. The rainfall characteristics in wet and dry spells of the monsoon for southeast India are studied with the help of high-resolution rainfall data generated by IMD and TRMM (3B42 and 2A25). The rainfall fraction of seasonal rainfall in each spell and the diurnal variation of different types of rainfall are described in section 6.3. The causative physical processes for the occurrence of rainfall over southeastern peninsular India are examined in section 6.4. Special emphasis was given to cold pool-vertical wind shear interaction because the magnitude of wind speed (and shear) and the depth of shear layer are different in different spells (chapter 5). The results are summarized in section 6.5. 113

Chapter 6 Rainfall characteristics during wet and dry spells 6.2 Data description: The present study depends heavily on high-resolution gridded rainfall data generated by IMD and TRMM satellite (2A25 and 3B42). 12 years of data, during 1998-2009, from both datasets (IMD and TRMM) are used to generate spatial rainfall maps. The 2A25 provides vertical rainfall (reflectivity) profiles measured by the precipitation radar (PR), while 3B42 is a merged rainfall product. It merges the infrared radiance measurements made by geostationary satellites with microwave measurements. 2A25 is mainly utilized to examine the occurrence percentage of different types of rain and their diurnal variation during wet and dry spells of the summer monsoon. 3B42 data are used to study the propagation characteristics of rain during wet and dry spells of the summer monsoon. 3B42 data are available at 3- hourly interval with a spatial resolution of 0.25 x 0.25. The 2A25 data (available at ~5 km spatial resolution and an instantaneous measurement) are grouped on to same grids and same temporal resolution as 3B42 for easy comparison. Also the ECMWF interim data are extensively used to examine the role of background thermodynamical and circulation features in altering rainfall production and/or propagation. 6.3. Characteristics of rainfall in wet and dry spells of the summer monsoon: 6.3.1. Spatial structure of rainfall during wet and dry spells: Although chapter 2 (figure 2.13) partly describes the salient features of rainfall in wet and dry spells, a detailed discussion was not made. Figure 2.13 makes use of 15 years of high resolution IMD rainfall data (1995-2009). The TRMM data that we considered in the present study are during 1998-2009. Therefore, the figure is reproduced here with 12 years of IMD rainfall data (i.e., 1998-2009) for easy reference. It also facilitates placing all rainfall characteristics in one section. Figure 6.1 shows spatial distribution of rainfall during (a) dry spell and (b) wet spell. The rain fractions (i.e., 100 * total rainfall during all wet or all dry spells in a year/seasonal rainfall of that year) by dry and wet spells for each year are estimated at each grid point. Composite rain fraction plots for wet and dry spells are shown in figures 6.1c and 6.1d, respectively, depicting the spatial variability of rain fraction 114

Chapter 6 Rainfall characteristics during wet and dry spells during each spell. It is evident from figures 2.13 and 6.1 that there is no significant change because of reduction in the data length. Our main study region is southeastern peninsular India. It is now known that the wet and dry spells in this region maintain an out of phase relation with the spells over the monsoon zone. It will be interesting to see these contrasting rainfall characteristics in these regions. Further, the west coast of India receives abundant rainfall in the monsoon season (can be seen in figures 2.13 and 6.1). One of the aims of this study is to find out the reasons why rain systems are not propagating towards east during dry spell? Therefore, we primarily focus on these three regions, i.e., monsoon zone, southeastern peninsular India and west coast of India. The spatial structure of rainfall during dry and wet spells is distinctly different (figure 6.1). As also, noted by several researchers that the anti-correlation in the rainfall pattern between the monsoon zone and southeastern peninsular India is clearly evident. The dry spells (for southern peninsular India) are characterized by subdued rainfall over southeast India, Rajasthan and Jammu and Kashmir and well distributed rainfall along the west coast of India, over the monsoon zone, foot hills of Himalayas and northeastern states of India. As noted by Gadgil and Joseph (2003), there is a distinct variability within the monsoon zone with eastern zone getting more rain compared to the western zone. On the other hand, good amount of rainfall occurred over southeast peninsula, northeastern states and coastal regions of India and foot hills of Himalayas during the wet spell. Rainfall is conspicuously less over the monsoon zone. Since, the rainfall is not homogeneous over the whole India, comparison of the spatial distribution of rainfall (rain total) may not be appropriate. The rain fraction plots are, therefore, considered to gain better insight in rainfall characteristics in different spells. The rain fraction plots depict the spatial variation of rainfall (in terms of %) features much more clearly than total rain plots (6.1a and 6.1 b). For instance, good amount of rainfall (> 200 mm) observed in wet spell over the east and west coasts, northeastern states of India and foot hills of Himalayas, in fact, comprises < 20% of seasonal rainfall. On the other hand, nearly the same amount of rainfall seen over southeast India during wet spell comprises 40-60% of seasonal rainfall. Over southeastern peninsular India, wet spell persists for only 22 % of time during the 115

Chapter 6 Rainfall characteristics during wet and dry spells southwest monsoon, but 40-60 % of the seasonal rainfall occurs during that spell. It also brings an important point that considerable amount of the seasonal rainfall over southeast India occurs in spells of active monsoon rather than in isolated thunderstorms. Figure 6.1: Spatial variation of total rain amount and rain fraction during dry (a and c) and wet (b and d) spells, respectively. Location of Gadanki is indicated with a star. Rain fraction of any spell indicates the fraction of seasonal rainfall in that spell (see text for more details). Also, it is interesting to note that the rainfall fraction over the foot hills of Himalayas during wet spell is lesser than during dry spell. Earlier studies have shown that there is some sort of coherence in rainfall pattern over southeast and northeastern states of India and foot hills of Himalayas. They observed good amount of rainfall over these regions during the break spell for monsoon zone. The present study shows that it may not be true always, particularly when the active/break spells (or wet/dry spells) are identified based on rainfall over southeast India. It seems, therefore, 116

Chapter 6 Rainfall characteristics during wet and dry spells identification of active and break spells on a regional scale and understanding circulation features responsible for those spells will be much more appropriate than considering the rainfall over India or monsoon zone for such studies. The total rainfall and rain fraction along the west coast of India are strikingly different in wet and dry spells. This region gets good amount of rainfall because of the lifting of moist-laden marine air mass by Western Ghats. The LLJ, a prominent component of the southwest monsoon, brings this moist air on to the peninsular India from the Arabian Sea. It also generates cyclonic shear on the northern side of its mean position of 14 N and is responsible for the generation of rainfall over the monsoon zone. The strength of LLJ on 850 hpa is often related to the strength of the monsoon by several researchers (Magna and Webster 1996, Webster et al. 1998, Goswami and Ajaymohan 2001, Joseph and Sijikumar 2004). Therefore, any change in the LLJ strength alters the rainfall amount not only along the west coast of India but also over the monsoon zone. Chapter 5 clearly shows that the LLJ strength and spatial distribution is different in different spells. The LLJ is strong and shows two branches during dry spell for the southeast India, while one branch of the LLJ (over the southern peninsular India) is weak during wet spell. As mentioned above, stronger LLJ during dry spell brings more moisture on to the southern peninsular India and produces more rainfall along the west coast. Then, as discussed earlier, the question arises why this heavy rain producing systems are not propagating towards east in dry spells? We discuss this issue in section 6.4. 6.3.2. Occurrence percentage of different types of rainfall in wet and dry spells: The earlier section clearly demonstrated the spatial variability of rainfall from wet spell to dry spell. But it is not clear from the above section that what type of rainfall systems contribute to the total rainfall in different spells of the monsoon, particularly over southeastern peninsular India. On the other hand, classification of rainfall systems into convective and stratiform and studying their occurrence percentage is highly essential for obvious reasons. For example, the vertical structure of latent heating, which drives several atmospheric circulations, is different in convective rain from that of in stratiform rain. Even from ISV of Indian monsoon s 117

Chapter 6 Rainfall characteristics during wet and dry spells perspective, such classification is important. An intriguing study by Chattopadhyay et al. (2009) revealed that the northward propagation of Indian ISO, which is one of the factors responsible for the occurrence of active and break spells, is achieved mainly by the organized movement of stratiform rain. The difference in the latent heating between stratiform and convective rain is found to be primarily responsible for the northward propagation of stratiform rain systems. The other studies also highlighted the importance of stratiform rain during the monsoon season (Schumacher et al. 2003) and active spells of the monsoon (Choudhury and Krishnan 2011). The remainder of this section, therefore, concentrates on the spatial variability of the occurrence percentage of different types of rainfall in wet and dry spells. The TRMM PR is perhaps the best available instrument for such classification of rainfall systems on a global scale (in tropics). The overpass frequency of TRMM PR is relatively low (note that the PR s swath is narrower than TMI s swath), nevertheless, the measurements pooled over several years (12 in this study) may depict the gross features with reasonable accuracy. The spatial variation of number of rain pixels (with surface rain > 0.1 mm hr-1) during dry and wet spells is shown in figure 6.2. They nearly resemble the rainfall maps during dry and wet spells shown in figure 6.1 with more pixels over good rainfall regimes (i.e., along the west coast, over northeastern states of India and monsoon zone and near foot hills of Himalayas) and fewer pixels over less rainfall regimes (i.e., southeast and northwest India). But for a small region surrounding Sri Lanka during dry spell, the number of rain pixels is larger than 100 in both spells at all grid points. The TRMM PR divides the rainfall into 3 main categories as convection, stratiform and other type, based on the vertical and horizontal structure of rainfall (Awaka et al. 1998). The occurrence percentage for convection, stratiform and other type of rainfall is estimated at each grid point during wet and dry spells. The occurrence percentage of other type of rainfall is found to be very large (60-90%) over Indian region (not shown here). But, as also mentioned in TRMM Manual (Ver.6), this category mostly consists of clouds and noise with very little rain reaching the surface. We, then, imposed a condition that each profile should have some surface rainfall (> 0.1 mm hr-1) to eliminate clouds and noise entering into our analysis, and 118

Chapter 6 Rainfall characteristics during wet and dry spells recalculated percentage occurrences for all types. Note that the number of rain pixels with some surface rain is only used to generate figure 6.2. Figure 6.2: Total number of rain pixels in dry and wet spells available over each grid point (grid resolution is 0.25º x 0.25º). TRMM PR measurements during 1998-2009 are used for this study. The occurrence percentage plots for different types of rain during dry and wet spells are shown in figure 6.3. Clearly, as expected, the other type of rain hardly occurs during wet and dry spells. The stratiform rain prevails for most of time in both wet and dry spells. Quantitatively, the occurrence percentage of stratiform rain is nearly 2-6 times that of convective rain in both spells. Figure also clearly depicts the existence of some important regional differences in the occurrence of stratiform/convective rain. In both spells, considerable rain occurs in the form convection along the west coast and southeast coast and near the foot hills of Himalayas. There are also some notable differences in the occurrence of different types of rain between spells. Over the monsoon zone and head Bay of Bengal, the occurrence percentage of stratiform (convective) rain is large (small) during dry spell, where as considerable percentage of convection is present during wet spell. The higher percentage occurrence of stratiform rain during dry spell over the above regions supports recent modelling studies by Choudhury and Krishnan (2011). The opposite feature is seen over northern Indian Ocean, where the occurrence percentage of convection is higher in dry spell than in wet spell. Though not as distinct as in other regions, the occurrence percentage of rainfall systems is different over southeastern peninsular India. Relatively high occurrence percentage of convection is 119