iii WRF SIMULATIONS OF A SEVERE SQUALL LINE: COMPARISONS AGAINST HIGHRESOLUTION DUAL- AND QUAD-DOPPLER RADAR MEASUREMENTS FROM BAMEX BY BRYAN ANDREW GUARENTE B.S., University of Northern Colorado, 2003 THESIS Submitted in partial fulfillment of the requirements for the degree of Master of Science in Atmospheric Sciences in the Graduate College of the University of Illinois at Urbana-Champaign, 2007 Urbana, Illinois iv ABSTRACT Historically, quantitative comparisons between modeled mesoscale convective systems (MCSs) and observations were limited by spatial and temporal variations between the datasets. Prior studies often compared specific values that characterized the MCSs, such as maximum rear-inflow jet wind speed, cold pool strength, or average storm motions. Using high-resolution multi-sensor data collected during the Bow Echo and Mesoscale Convective Vortex Experiment (BAMEX 2003), we now have an excellent opportunity to compare modeled versus observed MCS structures. In this thesis, we compare the statistical distribution of the radar reflectivity and wind fields within a modeled MCS to those from BAMEX observations. Specifically, we compared airborne dual- and quad-Doppler observations of the June 10 2003 BAMEX MCS against high-resolution (3 km grid spacing and 54 vertical levels) simulations made with the Weather Research and Forecasting (WRF) model using contoured-frequency-by-altitude diagrams (CFADs) and a new method, contoured-frequency-by-distance diagrams (CFDDs). These diagrams yielded bulk statistical comparisons of how the frequency distributions of the observed and modeled systems vary with height and distance, respectively. Comparisons using CFADs and CFDDs of modeled reflectivity and kinematics to those from airborne dual- and quad-Doppler radar syntheses were used to quantify the simulated squall line morphology, rear-inflow jet evolution, and microphysics in this storm system. CFADs of reflectivity showed that the model over-predicted the frequency of > 35 dBZ reflectivities near the surface with this microphysics package for this specific storm. A “hole” in only the modeled reflectivity CFADs was noted below the melting level between 5 and 30 dBZ, where there were no areas with frequencies greater than 1%. CFADs of RIJ-parallel squall-line-relative winds suggested an overprediction of modeled wind speeds above the melting level. The distance behind the convective line where the interface between front-to-rear and rear-to-front flow v occurred was easily identified on CFDDs from the RIJ-parallel squall-line-relative winds. With average altitude per bin per distance diagrams, the height of this same interface was readily demonstrable with time and distance from the convective line, presenting a chance to quantify the typical slope of the interface over the trailing stratiform region. Using all of the statistical methods included for comparisons between modeled and observed systems could help to identify where model parameterizations are lacking a physical understanding of the atmosphere or where we may need higher resolution observations to continue to test our understanding. ACKNOWLEDGEMENTS I would like to acknowledge the advising of Drs. Brian Jewett, Greg McFarquhar, and Robert Rauber. Without their constant leadership and prodding this project would not have happened. When I finally was able to take the project into my own hands, their ideas lead to many of the discoveries involved in this research. My greatest debt of gratitude goes out to them for letting me make my own mistakes. Special thanks go to Dr. Brian Jewett for his open-door policy and quick responses to my e-mails. All questions computer-related were answered to the best of his ability, often going two steps beyond where my problem actually lay. His quiet humor and mumbles also made BAMEX meetings run smoother, in my honest opinion. To Dr. David Jorgensen, I thank you for supplying the radar observations and the code to convert them to ASCII format. To the other members of the BAMEX research group, Mrs. Andrea Smith-Guarente and Mr. Joseph Grim, I would like to thank you both for keeping our advisors in check, and reviewing all of my work that you possibly could with all the work you had of your own to complete. I would like to thank Mr. Grim for his aid in making some of the dual- and quad- vi Doppler radar images from this BAMEX case. Andrea, I thank you for talking about research so often, even when there were other things more pressing in your life, and even when it was 3 or 4 a.m. Your help proofreading every draft of my chapters was unmatched. I would like to take this opportunity to thank you, my wife, for dealing with me finishing my thesis, even when the stress was getting to you and our unborn son. Thanks for holding him in. Lastly, I would like to thank my unborn son, for making me complete this project in a timely manner. Without your “gentle nagging,” I would have been floundering my way to a May 2008 graduation. I’ll have time to hang out with you more often after this thesis is submitted. The computer time provided by the National Center for Supercomputing Applications (NCSA) and the grant money provided by the National Science Foundation are obviously important contributors to my work. Completion of this work was dependent on these grants. Any opinions, findings, recommendations, or conclusions expressed in the material are those of the author and do not necessarily reflect the views of the sponsors. TABLE OF CONTENTS CHAPTER PAGE 1. INTRODUCTION .........................................................................................................1 1.1. Introduction ...........................................................................................................1 2. METHODOLOGY ........................................................................................................6 2.1. Model Description ................................................................................................6 2.2. Observations Description ......................................................................................8 2.3. CFADs...................................................................................................................9 2.4. CFDDs ................................................................................................................12 vii 3. COMPARISONS USING CFADS .................................................................................20 3.1. Masked CFADs ...................................................................................................20 3.2. RIJ-centered CFADs of Reflectivity ...................................................................28 3.3. RIJ-centered CFADs of Y-axis-parallel Squall-line-relative Wind Speeds.........35 4. COMPARISONS USING CFDDS ................................................................................45 4.1. Contoured-Frequency-by-Distance Diagrams ....................................................45 4.2. Average Altitude per Bin per Distance Diagrams ...............................................54 5. CONCLUSIONS............................................................................................................67 5.1. Conclusions .........................................................................................................67 REFERENCES ..................................................................................................................74 1 CHAPTER 1 INTRODUCTION 1.1. Introduction Mesoscale convective systems (MCSs) inundate the Great Plains with precipitation during summer, accounting for 30-90% of total rainfall (Fritsch et al., 1986). These systems are also known to frequently generate damaging winds, small hail, and occasionally, weak tornadoes. A theory on the maintenance of MCSs (RKW Theory) exists from numerical model simulations (Weisman and Rotunno 2004), plus an expansion describing the maintenance of bow-echoes and the role of convectively generated rear-inflow jets (RIJs) (Weisman 1992). A substantial amount of work has gone into discovering the mechanisms for the development of RIJs (e.g., Smull and Houze 1985, 1987a, b, Rutledge et al., 1988, Johnson and Hamilton 1988, Houze et al., 1989), and other studies have explored the mechanisms (dynamical and microphysical) by which the RIJ descends to the surface (e.g., Biggerstaff and Houze 1991, 1993, Braun and Houze 1994, 1996, Gallus 1996), occasionally causing strong winds and severe damage (Funk et al., 1999, Atkins et al., 2005, Wakimoto et al., 2006a, b). A number of studies examine this subject from one of two frameworks: modeling or observations. Each framework has its advantages and downfalls. A lack of modern high-resolution (temporal and spatial) MCS observations has made discoveries from observations lag behind discoveries from modeling studies of MCSs. With rare exception, observational studies lack high-resolution and/or the number of samples necessary to make statistically significant statements about the overall structure of a typical MCS or deduce what controls this structure. Although conducting model simulations at high-resolution overcomes some of these weaknesses, limited evaluation of the results of simulations against 2 observations has led to questions about the robustness of both the simulations and parameterization schemes. Before 2003, the most recent field campaign initiated to study midlatitude continental MCSs was held in 1985. In May-July 2003, the Bow Echo and Mesoscale Convective Vortex Experiment (BAMEX, Davis et al., 2004) was held in an attempt to obtain high-resolution observational datasets of MCSs, bow-echoes, and mesoscale convective vortices through in-situ microphysical, thermodynamic, and dual- and quad-Doppler radar observations. With these data, a study using both modeling and observational frameworks with high resolution was possible. Quantitative comparisons of MCS characteristics have been made in the past using models, observations, or a combination of the two. Variables like vertical and horizontal velocity, reflectivity, and rainfall rates have all been used to characterize MCSs (e.g., Smull and Houze 1987a, b, LeMone and Moncrieff 1994, Caniaux et al., 1994, Grady and Verlinde 1997, Montmerle et al., 2000, Dawson and Xue 2006, Grams et al., 2006, Pasken and Martinelli 2006, Wheatley et al., 2006, Storm et al., 2007). While many of these studies examined highresolution radar data, none of these studies employed statistical approaches to show the characteristics of the given variable over certain regions of the MCS, including the convective line and trailing stratiform region. For example, stratiform horizontal velocities have been compared from several different storms, with vertical profiles of horizontal velocities either obtained by averaging over the entire stratiform region or from a single sounding through the stratiform region (Smull and Houze 1987b). The problem here is that this method may not portray what is occurring over the entirety of a given MCS region. By averaging over the entire stratiform region, for instance, one eliminates the inherent variability that may be important to MCS structure (e.g., the RIJ). Thus, other statistical methods would be appropriate here. 3 Methods similar to Smith et al., 2008, where distributions of kinematic and microphysical variables are coordinated with specific regions of MCSs show how ill suited averaging over an entire inhomogeneous region can be. A distribution of each variable over each region is better suited to create an understanding of key MCS variables. Yuter and Houze (1995), henceforth YH95, introduced contoured-frequency-by-altitude diagrams (CFADs) for examining statistical changes in vertical distributions of vertical velocities, reflectivity, and differential reflectivity at small altitude increments obtained at highresolution with dual-Doppler radar for an evolving field of cumulonimbus clouds from the Convection and Precipitation/Electrification Experiment (CaPE). Their statistical method has been used to quantify vertical variations of numerous variables in many different types of systems by multiple authors (e.g., Steiner et al., 1995, Smith et al., 1999, Cifelli et al., 2000, Geerts and Heymsfield 2000, Yuter et al.,, 2005, Mori et al., 2006, Swann et al., 2006, Kingsmill et al., 2006). However, relatively few authors have applied this method to examine model fields (e.g., Lang et al., 2003, Rogers et al., 2006). Also, few authors have used this method to compare observations versus model output (e.g., Rogers et al., 2004, Smedsmo et al., 2005). Here, we shall show that CFADs are an ideal tool for making comparisons where temporal and spatial co-location is poor, as often occurs between observations and some model datasets. In this thesis, we focus on statistical comparisons between high-resolution dual- and quad-Doppler radar observations from the National Oceanic and Atmospheric Administration (NOAA) and Naval Research Laboratories (NRL) P-3 airborne radar systems of the 10 June 2003 MCS from the BAMEX field campaign and high-resolution Weather Research and Forecasting (WRF) model simulations of the 10 June case. CFADs are used in this study to 4 quantify the vertical distributions of reflectivity and horizontal winds from both observations and model simulations. While CFADs are useful to explain vertical distributions of observed and modeled fields, they inherently lack the ability to portray horizontal variations. Some meteorological phenomena exhibit the largest variability/gradients in the horizontal (e.g., mesoscale gravity waves, jets, and fronts). In this thesis, we introduce a new statistical tool for characterizing the variability in this direction. The new method, a contoured-frequency-by-distance diagram (CFDD), presented herein, extends the concept of YH95's CFADs to a new dimension. CFDDs were designed to ignore vertical variability. Instead of using increments of altitude to define a histogram volume, increments of horizontal distance from a given plane are used to define the histogram volume. This technique is helpful for comparisons of the June 10 2003 system, or any bow echo, because of their strong horizontal variations (e.g., winds associated with RIJ) behind the leading convective line. CFADs summarized the vertical distribution of meteorological fields while ignoring horizontal variability, while CFDDs ignore vertical variability while statistically summarizing the horizontal structures. A new methodology for thorough comparison between observations and model simulations can lead to further studies of MCS structures using numerical models, with an improved understanding of how well the model replicates certain characteristics of the MCS. Studies using these statistical techniques can quantify the errors originating from physical representations within the numerical model. The remaining sections of this thesis are arranged as follows: Chapter 2 presents the methods by which CFADs, CFDDs, and other statistical tools are constructed, and provides background information for the model simulations. Chapter 3 includes information about 5 comparisons of modeled fields to observed fields using CFADs, while Chapter 4 contains comparisons using CFDDs. Chapter 5 expands on key results and conclusions from this work. 6 CHAPTER 2 METHODOLOGY 2.1. Model Description We carried out a 36-hour simulation of the 10 June 2003 BAMEX MCS using the fully compressible non-hydrostatic ARW (Advanced Research WRF) core of the WRF model version 2.1.2 (Skamarock et al., 2005). We selected this case because it had the fastest flight-level winds recorded during BAMEX (80 kts), yet had few surface wind damage reports, suggesting that a strong RIJ never impinged on the surface. We ran multiple simulations with differing initializations, grid layouts, and parameterizations, but selected the simulation whose evolution was qualitatively the most similar to the observed MCS using timing for initiation and dissipation of certain squall line features (e.g., convective line, trailing stratiform, and northern bookend precipitation) and areal coverage of stratiform and convective regions. Times used for analysis purposes include 0412, 0430, 0448, 0515, 0542, 0600, 0618, 0636, 0654, 0712, 0748, and 0815 UTC. These analysis times were selected as they maintained the most consistent RIJ locations and orientations based on automated criterion of the fastest ground-relative wind speed below five kilometers. This simulation used two-way nested grids with three-tiered 27 – 9 - 3 km grid spacing (Fig. 2.1), and a mass-based, terrain-following, stretched 54 layer vertical grid. Parameterizations used in this work included the Thompson et al., (2004) microphysics, the RRTM longwave radiation scheme (Mlawer et al.,, 1997), the Dudhia shortwave radiation scheme (1989), the Monin-Obukhov surface layer scheme (Janjic, 1996), the Noah land surface model (Chen and Dudhia, 2001), the Yonsei University planetary boundary layer scheme (Hong and Pan, 1996 and Hong et al., 2006), and the Betts-Miller-Janjic cumulus parameterization 7 Fig. 2.1. Location of model simulation domains. Domain 1 (D1), domain 2 (D2), domain 3 (D3) have grid spacing of 27, 9, and 3 km respectively. 8 (Janjic, 1994). Convection was parameterized on the 27 and 9 km grids, while it was explicitly resolved on the 3 km grid. The coarse grid time step was 60 seconds. Initialization and lateral boundary condition data came from the NCEP Eta model with boundary conditions updated trihourly. The simulation began at 0000 UTC 9 June 2003 with all domains run for 36 hours. Model data was saved every 9 minutes. Henceforth, we only present data from the innermost grid using the model horizontal grid spacing and sampling the data at 0.5 km vertical resolution, which is not always consistent with the vertical grid spacing due to the stretched vertical coordinate, but matches the vertical grid spacing of the observational dataset. Computations were carried out on the National Center for Supercomputing Applications machines at the University of Illinois. 2.2. Observations Description The observational dataset came from Doppler radars mounted on the NOAA and NRL P- 3 aircraft that were flown on either side of the convective line of the MCS (Jorgensen et al., 2005). Measurements were taken at the following times on 10 June: 0422-0437, 0439-0458, 0500-0525, 0529-0558, 0600-0617, 0617-0636, 0636-0657, and 0740-0752 UTC. The lack of radar observations between 0715 and 0740 UTC was due to the NOAA P-3 performing a microphysical spiral descent where a quad-Doppler synthesis of the wind field was impossible due to the heading of the NOAA P-3 changing so rapidly. Each radar has a range of 46.2 km, but this range is rarely used to its fullest, as the objective was to retrieve dual- and quad-Doppler analyses, which required the sampled areas to overlap at more than just one location. This range limits the sampled area over which the observed and modeled fields can be compared. Radar reflectivity and derived horizontal velocity fields were transferred to a Cartesian coordinate 9 system for analysis following techniques from Jorgensen et al., (1996). The grid spacing of the radar output is 1.5 km in the horizontal and 0.5 km in the vertical. In the observations below 0.5 km, data were not retrievable. A signal may not be accurate in the lowest kilometer of the atmosphere due to interference by the surface clutter (Jorgensen et al., 1984). Because of these stipulations, observed radar reflectivity values at altitudes below 0.5 km, and occasionally up to 1 km, were not available. In addition, near the edges of the radar volume, over-estimation of the derived horizontal wind fields occurred due to a lack of sample points to derive the wind field properly. These areas have been removed from the analyses, which were provided by Dr. David Jorgensen. 2.3. CFADs A key tool employed to evaluate the simulations was the contoured-frequency-by-altitude diagram (CFAD, YH95). Because the simulated and observed storms were not collocated in space or time, CFADs provided a quantitative measure for comparing the vertical structure of the observations and the model. A CFAD is a two dimensional depiction of a collection of histograms, or frequency distributions, of a particular variable at evenly spaced altitudes (0.5 km apart in our work). In this study, we used CFADs to compare observed radar reflectivity and derived horizontal velocity fields obtained by the NOAA and NRL P-3 airborne radar systems with simulated reflectivity calculated following Stoelinga (2005) and horizontal wind fields from the model. The simulated reflectivity calculations are consistent with the microphysics package employed in the model (personal communication with Dr. Greg Thompson). Radar reflectivity (simulated and observed) and horizontal velocity (simulated and dual- or quad-Doppler derived, 10 where available) data were binned at each altitude within a predefined area using bin sizes of Z = 5 dBZ and V = 1 m/s, respectively. A CFAD is constructed by compiling histograms of a particular variable at each altitude into a single contour plot. Fig. 2.2a is an example of a simulated reflectivity histogram at 10.5 km above ground level (AGL) showing the frequency of occurrence of different reflectivity values at that altitude within an area enclosed within a two-dimensional horizontal mask. A mask was necessary to limit the CFAD analysis area to only the MCS. The mask applied to all CFADs was defined as the horizontal area in which the maximum reflectivity in each grid column is greater than, or equal to 0 dBZ, regardless of the variable plotted on the CFAD. The histograms included in each CFAD only used data from within the masked area. Use of this twodimensional mask simplified CFAD interpretation since the analysis area (number of points per histogram) remains constant with height. When masked with this threshold, the CFAD included data from the trailing stratiform region, convective region, as well as the leading anvil parts of the storm at all altitudes. Some extraneous information was included in this masked area, including other convection occurring in the 3 km domain. Another method, discussed later, was applied to limit the inclusion of these extraneous data points. Fig. 2.2b shows histograms as a function of altitude plotted as a single three-dimensional diagram. The CFAD (Fig. 2.2c) represents a contour plot of the frequencies constructed by looking down the z-axis on the three dimensional diagram and contouring the data as if it were a topographic map. CFADs portray the bulk characteristics of the vertical profile of the data. For a given CFAD, each point provides the frequency of occurrence of the data in that bin at a specific altitude. Each altitude on a CFAD has frequencies that should add up to 100%; those altitudes with frequencies that do not add up to 100% have horizontal areas that lack applicable 11 Fig. 2.2. Conceptual model of how to build a contoured-frequency-by-altitude diagram (CFAD). All sub-diagrams are from 0712Z in the simulation. a) Histogram of simulated radar reflectivity at 10.5km above ground level. Reflectivity binned using 5 dBZ intervals. b) Isosurface of frequencies of simulated radar reflectivity with altitude increasing into the page. c) CFAD of simulated radar reflectivity within masked area. Frequency interval is 2% beginning at 1%. d) Plan view of maximum reflectivity in a column mask. Northern portion of reflectivity pattern cut off by edge of domain. 12 values (areas that are contained within the 2-D mask but with point values below the 0 dBZ threshold or did not contain retrievable values of reflectivity). These null value areas are not plotted on the CFADs, but need to be included when adding up frequencies at a given height to achieve a cumulative frequency of one hundred percent. In the masked CFADs in this thesis, such as Fig. 2.2c, the contour interval was two percent beginning at one percent, meaning all areas that do not constitute at least one percent of the masked area did not appear on the CFADs. To understand how to interpret the diagrams, consider, for example, the overall upper-level precipitation mode. As the convection initiates, there would be very little stratiform precipitation, but a large amount of convective precipitation. On a CFAD, higher percentages would be found in the higher reflectivity bins (particularly at lower altitudes) indicative of convective cells, while lower percentages would be found in the lower reflectivity bins indicative of stratiform precipitation. As the MCS evolves, a stratiform region would begin to develop. Now on a CFAD, nearly equal percentages would be found in the higher reflectivity bins and the lower reflectivity bins. Eventually, the MCS outflow would surge away from the convective line, cutting off the convective updraft, leaving an orphaned stratiform precipitation region. This CFAD would show lower percentages of higher reflectivity bins with higher percentages (particularly at higher altitudes) of lower reflectivity bins. With time, a line through the maximum percentage at each altitude in these CFADs would change from vertically oriented to negatively tilted with a shift toward lower reflectivity values. 2.4. CFDDs Contoured-frequency-by-distance diagrams (CFDDs), developed for this study, are an extension of the YH95 CFAD method. To create CFDDs, we used a rotated coordinate system. 13 Where the y-axis (distance of CFDD) was oriented parallel to the maximum RIJ horizontal wind vector, the x-axis was oriented perpendicular to the y-axis, and z continued to represent height AGL. A CFDD is a collection of histograms of a particular variable at each distance (y-axis point) rearward from the leading edge of the anvil (defined by the mask) compiled into a single contour plot. The data points composing a single histogram are all points within the mask region lying in the x-z plane within a small range of y. To visualize the CFDD, we can think of the analysis volume as a stick of butter, where the y-axis is along the long axis of the butter, and the x-z plane is a pat of butter containing the data of one histogram (Fig. 2.3). To construct a CFDD, we built histograms of a specific variable for each x-z slab. The y-width of the x-z slab (along the y-axis) was chosen in this study to be 3 km for the model simulations, the minimum possible considering the model grid spacing, and 1 km for the observations. To create the CFDD, the histograms were assembled as a function of distance in a three-dimensional diagram. As with the CFAD, looking down the z-axis and contouring the data as if it were a topographic map produces a CFDD (Fig. 2.4). In the diagrams in this thesis, such as Fig. 2.4, the contour interval was five percent beginning at one percent, unless otherwise noted. To restrict our analysis to the RIJ and exclude higher wind speeds characteristic of the upper atmosphere, the CFDDs presented herein used data only at and below 7 km (which remained below the strongest front-to-rear squall-line-relative flow in the observed and simulated MCS). Horizontally, CFDDs only included data within a swath 57 km wide (typically the width of our model RIJ) centered on the maximum RIJ wind vector. Binning of data was the same as with CFADs (Z = 5 dBZ and V = 1 m/s). Frequencies at a given distance add up to one hundred percent as was the case with CFADs when accounting for the null values. CFDD fields included radar reflectivity (observed and simulated), absolute wind speed, squall-line- 14 Fig. 2.3. Conceptual model of how to build a contoured-frequency-by-distance diagram (CFDD). The volume defined by the larger rectangular prism was the total volume encompassed by the CFDD. The long axis of the rectangular prism was defined parallel to the maximum RIJ wind vector. Note the rotation of the Cartesian coordinate system. The x-dimension of the CFDD volume was chosen to be 57 km. The z-dimension of the CFDD volume was chosen to be 7 km. The y-dimension of the CFDD volume changed depending on the size of the masked area. Each x-z slab (smaller rectangular prism) was 3 km along the y-dimension, consistent with model grid spacing. 15 Fig. 2.4. Contoured-frequency-by-distance diagram (CFDD) of squall-line-relative velocity parallel to the y-axis (m/s) from model simulation at 0712Z June 10. Y-axis begins ahead of the leading convective line and ends at the back edge of the stratiform region. Convection is located where the density of positive wind speeds drastically decreases. Frequency interval is 5% beginning at 1%. 16 relative wind speed, and the y-axis parallel component of both wind speed and squall-linerelative wind speed. To construct CFDDs of modeled fields, the local maximum RIJ wind vector was selected from each horizontal level, and then the absolute maximum was picked from those local maxima to define the y-axis of the CFDD volume. In the observational dataset, despite quality control, there remained some false velocities on the edges of the sampled volume. Because these erroneous data points could have been the maximum velocity from each altitude, selection of the observed maximum RIJ wind vector was somewhat subjective. To exclude precipitation areas from the CFADS within the masked area but outside the MCS, a further step was necessary. Not only was it necessary to exclude extraneous precipitation areas, but it was necessary to compare CFADs of observed and modeled fields on a similarly sized domain. The observational domain was smaller than the model domain due to constraints imposed by the quad-Doppler flight tracks. Because the model domain was larger than the observational domain, differences in percentages from smaller phenomena may have occurred in the larger model domain compared to the smaller observational domain where the smaller phenomena may have been missing. Moreover, different fractions of each domain may have been occupied by different regions of the MCS (i.e., trailing stratiform or convective line) when the sizes of the domains were mismatched. When looking at reflectivity images from the model compared to observations, the model captures a larger scale southerly transport of stratiform precipitation particles which the observations did not see due to its smaller domain size. This bias caused interpretation difficulties between the observed and modeled datasets. By using the “stick of butter” domain taken from the CFDDs when computing a CFAD, data points from other precipitating systems were excluded so that the volume of the MCS with an active 17 RIJ was isolated. This also makes interpretation of the CFADs consistent across the datasets and even accounted for orientation differences between the observations and model. All previously defined variables used for CFADs and CFDDs have been plotted using this method. Shading of frequencies on these diagrams was by five percent starting at one percent as with CFADs. An example of this type of diagram is shown in Fig. 2.5. After making CFDDs, it was insightful to plot the average contributing altitude (with an interval of z, where z = 1 km in this study) where a particular velocity value (binned as V) occurred as a function of distance from the leading edge of the anvil. These diagrams are called average altitude per bin per distance diagrams and were constructed as follows. Consider a histogram showing the frequency of velocities occurring within the x-z slab on Fig. 2.3. Within an x-z slab, there may be a number of points occurring within a given velocity bin V that originate from different altitudes within the slab, making interpretation of the data somewhat ambiguous. There may also be data points in one bin (V) on the histogram composed primarily of data from a low altitude while most data points in an adjacent bin (V 1) may originate from a high altitude. A plot of the average altitude for each bin of the histogram allows one to determine the altitude range that contributes most to the bin. This method is extended to the entire CFDD by plotting the average altitudes from all x-z slabs on a single diagram (Fig. 2.6). 18 Fig. 2.5. Contoured-frequency-by-altitude diagram (CFAD) of simulated radar reflectivity within contoured-frequency-by-distance diagram (CFDD) area. Frequency interval is 5% beginning at 1%. Note the difference in vertical extent of this CFAD versus the masked CFAD. 19 Fig. 2.6. Same as Fig. 2.4. except average altitude per bin per distance diagram. 20 CHAPTER 3 COMPARISONS USING CFADS 3.1. Masked CFADs CFADs were produced for the model simulations and the observations just prior to, during, and after bowing segments occurred. These diagrams compare statistically the distributions of modeled and observed fields. Model CFADs were constructed for 12 times: 0412, 0430, 0448, 0515, 0542, 0600, 0618, 0636, 0654, 0712, 0748, and 0815Z. These times were selected because a consistent RIJ region was maintained on the CFDDs (i.e., maximum RIJ wind remained quasi-stationary with respect to the convective line, with only slight orientation changes) and nearly steady state CFADs and CFDDs were seen. Times processed from the observations included 0430, 0450, 0510, 0540, 0610, 0630, 0650, and 0750Z. Observation time selection depended on the proximity of the two airborne radar systems and sufficiently straight flight legs for quad-Doppler synthesis. When the NOAA P-3 performed microphysical spirals, data were not available for analysis as the plane’s heading changed so rapidly. In this chapter, comparisons of radar reflectivity and wind speeds from the model and observations used masked CFADs, with the goal of identifying similarities and differences in the systems’ reflectivity/precipitation structures. The observed data were only available on a small sub-section of the MCS as sampling of the system was performed with airborne radars that could not view the entire system at one time. The model, comparatively, contained the whole MCS. Comparisons of model and observations using the frequencies on a masked CFAD were difficult to interpret, since the domain sizes were different. For this purpose, we therefore focused on RIJ-centered CFADs for statistical 21 comparisons of distributions for reflectivity and RIJ-parallel, squall-line-relative wind speeds. In this way, we can compare CFADS from each dataset. We focus first on masked CFADs of simulated reflectivity (Fig. 3.1) over the entire model domain within the maximum reflectivity mask (Fig. 3.2), at the model analysis times listed in Chapter 2. In Fig. 3.1, we see that the maximum frequency of simulated radar reflectivity at 1 km remained in the same radar reflectivity bin (35-40 dBZ) throughout the analysis period. Tokay and Short (1996) showed that almost all of the precipitation in tropical MCSs observed during TOGA-COARE (Tropical Oceans Global Atmosphere – Coupled Ocean Atmosphere Response Experiment) with reflectivity values above 40 dBZ fell from convective clouds. In addition, five percent of the stratiform precipitation from Tokay and Short’s TOGACOARE cases had reflectivity values above 36 dBZ. Because of the small percentage of stratiform precipitation particles with reflectivities above 36 dBZ, we will consider all reflectivity values above 35 dBZ to be convective. The radar reflectivity of an intercepted volume of precipitation is given by 6 ZN (D )D dD (1.1) where N(D) is the number distribution function of precipitation particles with maximum dimension D. In Fig. 3.1, as one travels vertically upward from the surface, the frequency of convective-dominant precipitation areas, defined as > 35 dBZ radar reflectivity, decreases to below one percent above approximately 9.5 - 10 km at all analysis times, showing a vertical transition to mixed-type (stratiform and convective) precipitation. The manifestation of this transition from convective-dominant precipitation particles to mixed-type precipitation particles in a CFAD is a negative tilting of the maximum frequency axis at higher altitudes. Over time, the maximum frequency axis can change showing a temporal transition of the precipitation mode 22 Fig. 3.1. Temporal evolution of masked CFADs of simulated radar reflectivity binned in 5 dBZ intervals. Frequency interval is 2% beginning at 1%. 23 Fig. 3.2. Temporal evolution of model RIJ-centered CFAD locations (thin black box) overlaid on reflectivity at 3.5km. Thin black outline around reflectivity pattern indicates maximum reflectivity in a column mask. The short dimension of the RIJ-centered box is always 57km, with the longer dimension varying with time, depending on mask outline. (from convective-dominant to mixed-type). Careful comparison of the panels in Fig. 3.1 shows that the whole system 24 became more stratiform as the frequency distributions shifted toward smaller radar reflectivity values (to the left on progressive CFADs in Fig. 3.1), and the maximum frequency axis tilted more negatively with time (representing a reduction in the height to which the convective region penetrated. Since the model CFAD domain was larger than the observational domain (compare Figs. 3.2 and 3.3, noting the scales on the axes) and, in terms of the MCS size, included the observational domain, it was only possible to compare the general trend or slope of the frequency maxima in the CFADs, but not the actual frequencies. The maximum frequency at 1 km in the observed reflectivity CFADs (Fig. 3.4) varied temporally between 20-25 dBZ and 35-40 dBZ, so that the observed reflectivity frequency maxima were less than the simulated reflectivity frequency maxima at all times. Over-prediction of simulated reflectivities consistently occurred in the model fields as has been noted by other authors simulating tropical cyclones (McFarquhar et al., 2006 and Rogers et al., 2007). The observed > 35 dBZ reflectivity values (convectivedominant precipitation) at all analysis times reached higher altitudes (14.0 km) than in the simulations (11.5 km). Higher reflectivity values at higher altitudes may be attributed to stronger vertical velocities in the real atmosphere, insufficient vertical resolution in the upper reaches of the model domain, updraft intensity dependence on horizontal resolution (Weisman et al., 1997 and Bryan et al., 2003), and/or an upshear tilt of the simulated convection which decreased raindrop collection efficiencies, as noted by Ferrier et al., (1996). The slope of the frequency maxima in reflectivity was less negative (more vertical) in the observations (Fig. 3.4) than in the simulations (Fig. 3.1) at low and middle altitudes (< 10 km), but changed to more negative (more horizontal) in the high altitudes (≥ 10 km). This makes the 25 Fig. 3.3. Temporal evolution of observed RIJ-centered CFAD locations (thin black box) overlaid on reflectivity at 3.5km altitude with ground-relative wind barbs (kts). The short dimension (generally oriented SW to NE) of the RIJ-centered box is always 57km, with the longer dimension varying with time, depending on mask outline. 26 Fig. 3.4. Temporal evolution of masked CFAD of observed radar reflectivity binned in 5 dBZ intervals. Frequency interval is 2% beginning at 1%.average slope approximately the same with a slight negative bias in the simulations because of the under-predicted heights of the convective reflectivity values. However, the instantaneous slope of each can vary dramatically from altitude 27 to altitude. Because of the negative bias in the simulations, these slope differences show the observations as more convective. This however may be due to the domain differences. The cumulative frequencies of simulated reflectivity greater than 0 dBZ at 1 km altitude only accounted for 11-21% of the masked area at their maximum (0600 and 0636Z on Fig. 3.1), excluding frequencies below 1% of the masked region as they are not plotted on CFADs (note starting value of legend). This meant that between 11 and 21% of the masked region of the MCS likely produced surface rainfall in the model. In the observations, the cumulative frequencies of radar reflectivity greater than 0 dBZ at 1 km accounted for 65-87% of the masked area at their maximum (0650Z on Fig. 3.4). This disparity may be partially due to domain size differences and/or lack of color fill of areas smaller than 1% on masked CFADs, especially in the larger domain of the model. Below the melting level (located at ~3.5 – 4.0 km in the model and ~3.0 - 3.5 km in observations), there were few bins with reflectivities less than 35 dBZ at 1 km that exceeded the one percent threshold on the simulated reflectivity CFADs (Fig. 3.1). The time of greatest cumulative frequency less than 35 dBZ at 1 km (0815Z) only covered 5-11% of the masked area. None of the reflectivity bins between 5 and 30 dBZ at 1 km contained larger than 1% frequencies. The time of greatest cumulative frequency at 1 km of observed reflectivity values less than 35 dBZ (0610Z) accounted for 44-56% of the precipitation (Fig. 3.4). The minimum in the simulated reflectivity CFADs (Fig. 3.1) below 3 km between 5 and 30 dBZ may be due to conversion of particles from sub-freezing to above freezing temperatures in the microphysical parameterization, as the minimum began just below the melting level. Other studies have suggested that microphysical parameterization may not be the cause of the low-level simulated reflectivity deficiencies, but rather the boundary layer scheme (Smedsmo et 28 al.,, 2005). It is uncertain in this case, which of these schemes, if any, was the cause of the deficiencies. Use of RIJ-centered CFADs will clarify the differences noted here, as the domain sizes were not the same in this comparison. Comparing Figs. 3.1 and 3.4, the width of the contoured region at each altitude above the melting level greater than 0 dBZ reflectivity was consistently wider in the simulated MCS compared to the observations. Since we cannot compare frequencies due to the domain size mismatch, it was impossible to compare the amplitudes of these histograms until we looked at RIJ-centered CFADs. Below the melting level, the model reflectivity distributions became bimodal as the reflectivities split into a convective-dominant precipitation mode and mixed-type mode. Although the observed histograms widened, they never became bi-modal in the lowest 3 km. 3.2. RIJ-centered CFADs of Reflectivity With RIJ-centered CFADs (Figs. 3.5 and 3.6), it is now possible to compare observed and modeled frequencies, distribution widths, and the slopes of the frequency maxima because of comparable domain sizes. On a RIJ-centered CFAD, the frequency interval was 5% beginning at 1%. The highest frequencies present on these CFADs in all times at 1 km occurred between 25 and 45 dBZ for the model (Fig. 3.5) and between 10 and 45 dBZ for the observations (Fig. 3.6). The shape of the histograms at 1 km altitude from the model showed a long low-frequency tail in the lower reflectivity values (< 30 dBZ reflectivity), a relatively high frequency in the higher 29 Fig. 3.5. Temporal evolution of RIJ-centered CFAD of simulated radar reflectivity binned at 5 dBZ intervals. Frequency interval is 5% beginning at 1%. Area encompassed in this CFAD is located on corresponding panels on Fig. 3.2. 30 Fig. 3.6. Temporal evolution of RIJ-centered CFAD of observed radar reflectivity binned at 5 dBZ intervals. Frequency interval is 5% beginning at 1%. Area encompassed in this CFAD is located on corresponding panels on Fig. 3.3. 31 reflectivities (> 30 dBZ reflectivity), and a sudden drop off to the right of the convective frequency peak. This, along with the highest frequencies of 1 km reflectivities from each dataset, suggested that the model was predicting a lower fraction of stratiform precipitation on this smaller scale. In the observations (in which quality control often removed below 0 dBZ reflectivity values), the frequencies climb more gradually to their peak and often have a slower drop-off in the higher reflectivities making for a longer high reflectivity tail. Despite generally showing the same maximum reflectivity on these CFADs, the peak frequencies in the observations often have a lower frequency than the simulations as the longer length (thus greater horizontal area) of the RIJ-centered box in the CFADs showed lower frequencies. The convective line would have to take up a much greater area of the RIJ-centered box to account for the greater frequencies. A change in the orientation of the RIJ-centered box or a bowing in the convective line can alter the frequencies of convective reflectivities due to non-normal orientation to the convective line. Verifying this is easy with a check of the box orientation to the convective line (Figs. 3.2 and 3.3). The simulations consistently over-predicted the frequency of higher reflectivity values below the melting level altitude, possibly due to assumptions about conversion from ice crystals to liquid droplets in the model microphysics. At the surface, these reflectivity values and frequencies may be shifted due to evaporation, collisions, coalescence, and breakup of precipitation-sized particles. The changes in the frequency of > 35 dBZ reflectivity values with height discussed with masked CFADs were not evident on RIJ-centered CFADs, since a vertical limit was imposed to exclude altitudes that may alter the wind statistics by including synoptic scale flows instead of just the mesoscale flow patterns associated with the RIJ. 32 In the observed reflectivity RIJ-centered CFADs (Fig. 3.6), the temporal fluctuations of the maximum frequency axis below the melting level implied a rapid evolution of the MCS and a change in the amount or shape of the convection. This became most evident with close comparison of observed reflectivities on Fig. 3.3, where a bowing segment formed around 0510Z, after which the gust front pushed out from below the main convective line between 0630 and 0650Z. The maximum frequency axis of simulated reflectivity CFADs (Fig. 3.5) remained nearly steady state at 35-40 dBZ throughout the analysis times, suggesting little transition to post-bowing stages, consistent with Fig. 3.2, which showed consistent reflectivity structure in the RIJ-centered box, particularly after 0515Z. The model simulation showed a nearly steady state, with relatively few bowing events, and certainly no gust front surging out ahead of the line as in the observations. The maximum frequency on each RIJ-centered CFAD of reflectivity, in either dataset, seemed to cycle with height and frequency over time giving the perception of multiple convective updraft pulses. The maximum frequency in all RIJ-centered CFADs (model and observations) was above the melting level, between 0430Z and 0750Z. In the model (Fig. 3.5), there were three convective upward pulses, with some of the frequency change between 0636 to 0654Z occurring because the orientation of the RIJ-centered box became less normal to the convective line, thus including more of the convective region in the CFAD. The peaks of the convective pulses appeared on RIJ-centered CFADs at 0515, 0654, and 0748Z (Fig. 3.5), which resulted in an average period of 71 minutes. Observations from 0430 and 0450Z covered small areas as seen in Fig. 3.3, accounting for a smaller subsection of the MCS than the other observational analyses. This could have artificially increased the maximum frequency as the percentage of the area containing 25-30 dBZ 33 reflectivity increased, while the physical area containing 25-30 dBZ reflectivities did not change significantly. This could have contributed to the appearance of convective pulses simply by having a smaller analysis area. In the observations (Fig. 3.6), there were likely two convective pulses one at 0450 and the other at 0610Z. The period of these pulses was 80 minutes. With the large time interval (> 20 min) between each analysis time, it was possible that documentation of convective pulses from either dataset was incomplete, making pulses appear to have a longer period. Without a larger sample population, convective updraft periods presented from the model and observations are the best guesses possible from the data. The maximum frequency (51-56%) on the model RIJ-centered CFADs (Fig. 3.5) occurred in the 35-40 dBZ simulated reflectivity bin. This occurred in two consecutive analysis times: 0748 and 0815Z. In the observations (Fig. 3.6), the maximum frequency is 56-61% and falls in the 25-30 dBZ bin at 0450Z. However, because of the aforementioned artificial increases in frequency from the size of the observed area, it is possible 0610Z could have been the time of maximum frequency with a peak of 51-56% within the 20-25 dBZ reflectivity bin. The peak frequencies, for analysis purposes, are considered the same between datasets. The difference in the reflectivity bin of the peak frequency shows that even on a similar domain, the most frequently occurring reflectivity value was over-predicted in the model compared to the observations. Comparing either of the times where the model showed its peak frequency, the model predicted a peak frequency at a lower altitude (5.5 to 6.0 km) than the observed peak (7.0 to 7.5 km). This may have been due to stronger updrafts in the observations or comparing different times within the convective pulses. The trend in the maximum frequency of the observed reflectivity below the melting level (Fig. 3.6, below 3.0 – 3.5 km) showed a vertically oriented maximum frequency axis or a slight 34 negative slope. As one approaches the melting level from below, the maximum frequency axis becomes positively sloped within 0.5 – 1.0 km. This is one of few areas on observed RIJcentered reflectivity CFADs that had a positive slope. In the positively sloped area, this suggested acceleration of the terminal velocity over a 0.5 – 1.0 km depth (e.g., Fig. 3.6 at 0610 and 0630Z). Above the melting level, the slope of the maximum frequency axis changed to negative once again for the remainder of the diagram, initially consistent with a change in the dielectric constant between ice crystals and liquid water droplets as the reflectivity values only decreased by 5-10 dBZ in all cases through the bright band. This is consistent with the expected 7 dBZ shift from a change in the dielectric constant (Smith 1986). In the observations, the bright band is visible on radar scans, but it is difficult to locate on RIJ-centered CFADs, as the reflectivity change is often smaller than the bin size chosen for our analysis. Farther aloft, the negative slope of the maximum frequency was due to the height of convective penetration and the decreased size of the precipitation particles with height. In the model output, the maximum frequency axis of reflectivity below the melting level (3.5 – 4.0 km) remained constant with height, except for 0515, 0712, and 0815Z where maximum frequencies shift ± 5 dBZ over 0.5 km. Half a kilometer below the modeled melting level, the slope of the maximum frequency axis became positive over 0.5 km. Above the melting level, all the simulated reflectivity RIJ-centered CFADs indicate a vertical maximum frequency axis or a vertical axis changing to a slight negative slope in the highest altitude plotted (7 km). In the simulated reflectivity calculation from the plotting package, there is a correction for ice particles scattering radiation as if they had a liquid water skin (i.e., melting ice) near the melting level, so there should be a bright band in the reflectivity pattern. The simulations did not seem to show a difference in reflectivity based on the scattering differences, but the aforementioned over- 35 prediction of reflectivity values may have over-shadowed this phenomenon, as it is not present in simulated reflectivity images or clearly defined in reflectivity CFADs. At the surface, the cumulative percentage of reflectivity values below 35 dBZ varied dramatically between the model and observations. As was the case with masked CFADs (Figs. 3.1 and 3.4), even at its maximum cumulative percentage, the model RIJ-centered CFAD contained a smaller percentage of area with reflectivity below 35 dBZ at 1 km (31-61% at 0815Z; Fig. 3.5) than the observed RIJ-centered CFAD at 1 km (81-100% at 0630Z; Fig. 3.6). More importantly from the same CFADs, 27-37% of the model domain contained convectivedominant precipitation, while the observations only yielded 7-17% of points containing convective-dominant precipitation. The model RIJ-centered box had a longer along-RIJ axis than the observations, thus an even larger area covered with > 35 dBZ reflectivity to account for such high percentages, yet the RIJ-centered CFADs suggested once again that the model is overpredicting the percentage of area with higher reflectivity values. 3.3. RIJ-centered CFADs of Y-axis-parallel Squall-line-relative Wind Speeds To compare the location and intensity of the RIJs, it was insightful to plot y-axis-parallel squall-line-relative winds from both datasets. These diagrams portray vertical variation in wind speeds. In Fig. 3.7 or Fig. 3.8, it is important to note that positive wind speeds were from the rear of the system to the front of the system (rear-to-front flow) while negative wind speeds were from the front of the system to the back of the system (front-to-rear flow). To understand how a typical RIJ appears on a CFAD, we have produced an idealized east-west oriented wind field. This field had a RIJ core (+30 ms-1) descending toward the surface from the rear of the system to the front. Above the RIJ, there was a similar front-to-rear 36 Fig. 3.7. Temporal evolution of modeled RIJ-centered CFAD of y-axis-parallel line-relative wind speed binned at 1 ms-1. Frequency interval is 5% beginning at 1%. Area encompassed in this CFAD is located on corresponding panels on Fig. 3.2. 37 Fig. 3.8. Temporal evolution of observed RIJ-centered CFAD of y-axis-parallel line-relative wind speed binned at 1 ms-1. Frequency interval is 5% beginning at 1%. Area encompassed in this CFAD is located on corresponding panels on Fig. 3.3.jet core (-30 ms-1) angled toward the rear of the system. A vertical cross section of the idealized RIJ-parallel line-relative wind speed is shown in Fig. 3.9 with the corresponding RIJ-centered CFAD in Fig. 3.10. In this idealized 38 situation, one expects to find rear-to-front flow near the surface, as was seen in Fig. 3.10. As one increases in altitude, the area of the rear-to-front flow slowly shrinks, filling in with front-to-rear flow, yielding a mix between front-to-rear and rear-to-front flow below 5 km, leaving solely front-to-rear flow above 5 km. The crossover region from 2.5 – 4.0 km contained contributions from both jets, resulting in the increased spread of frequencies in that layer in Fig. 3.10. The small gaps (frequencies under 1%) in the frequencies were due to sampling and the small V = 1 ms-1 bin size for winds. Low-level positive wind speeds (e.g., below 3 km) in either the model or observations CFADs generally indicated a contribution from the low-level RIJ or from winds behind the outflow boundary. It is impossible to separate these phenomena from each other on a CFAD. The presence of an outflow boundary ahead of the convective line on CFADs is dependent on the location of the RIJ-centered box. In the observations, the box was only far enough ahead of the line to see this feature at 0540, 0610, and 0650Z. These times were the only times that showed significant front-to-rear flow patterns at higher altitudes as the RIJ-centered box will be filled more frequently by front-to-rear flow (mid-layer inflow) ahead of the convective line. Winds in the lowest 1 km of observations were irretrievable, making a comparison of surface winds impossible. From these observations alone, it was impossible to determine the surface wind field, although the RIJ and gust front boundary likely influenced the wind field near the surface. Thus, a comparison of the 1 km altitude wind data was conducted. CFADs of y-axis-parallel squall-line-relative wind speed from the model (Fig. 3.7) contained a temporal progression at 1 km from percentages less than 1% rear-to-front flow to 39 Fig. 3.9. Vertical cross-section of idealized RIJ-parallel line-relative wind speeds. Wind speeds are in ms-1 with solid contours being positive (rear-to-front) and dashed contours being negative (front-to-rear). The convective line of the system would be off the right side of the diagram. 40 Fig. 3.10. Idealized RIJ-centered CFAD of RIJ-parallel line-relative wind speed binned at 1 ms-1. Frequency interval is 5% beginning at 1%. Area encompassed in this CFAD is located on Fig. 3.9. 41 increasing percentages of rear-to-front flow until 0636Z when the cumulative percentages decreased to 2-12% at 1 km. Stated differently, the area that was rear-to-front flow only accounted for 2-12% of the area at 1 km. After 0654Z, the positive trend in cumulative percentages at 1 km resumed. On the contrary, the observed system (Fig. 3.8) maintained 1 km altitude rear-to-front flow with a positive trend in cumulative frequencies up to 0630Z where the cumulative frequencies declined slightly. The cumulative percentages artificially declined at this time as the RIJ-centered box did not extend ahead of the convective line where rear-to-front flow may have occurred behind the outflow boundary. At 0650Z, the cumulative percentages of rearto-front flow reached their maximum at 1 km. After this time, another decline occurred as the convective line was not within the RIJ-centered box. The rear-to-front flow varied from being 212% of the RIJ-centered box area to 9-54% at its peak cumulative percentage, and never dropped below 1%. The cumulative percentage of rear-to-front flow in the model at 1 km rarely exceeded the cumulative percentage of front-to-rear flow. The same was true in the observations, except the cumulative percentage of rear-to-front flow never exceeded its front-to-rear counterpart. The most significant finding from these diagrams is associated with the low-level rear-tofront flow. In the observations, the wind field CFADs suggested three times in which an outflow boundary or relatively strong RIJ reached down to 1 km above the surface (0540, 0610, and 0650Z). A plan-view of the 1 km observed reflectivity (contoured) with squall-line-relative wind barbs is provided for 0610Z in Fig. 3.11 to show the rear-to-front flow. At all other times, the RIJ-centered box did not extend ahead of the leading convective line, making y-axis-parallel squall-line-relative wind speed contributions from an outflow boundary to the CFAD impossible. Since the prominent rear-to-front ‘nose’ in the observation CFADs at 1 km was absent at these other times, we interpret this to mean that the RIJ had not descended to lower levels at those 42 Fig. 3.11. Plan view of observed reflectivity (contoured; dBZ) and squall-line-relative wind barbs (kts) at 1 km from 0610Z. Included is the RIJ-centered box (thin black outline). 43 times and the only contributions made during 0540, 0610, and 0650Z may have come from preconvective line activity. This however, is only speculative. The location of these stronger rearto-front winds was indeterminable from a CFAD, but can be determined from CFDDs as shown in Chapter 4 when the convective-line-relative location of these maxima is derived. By comparison, the model CFADs never showed the occurrence of faster rear-to-front wind speeds at low-levels. This suggested that the model may have a weaker cold pool, stronger vertical wind shear keeping the outflow boundary vertically aligned with the convective towers, or a RIJ that was not as low in altitude or as fast in y-axis-parallel squall-line-relative wind speed as the observations, suggesting a possible problem in the treatment of the planetary boundary layer. Below the melting level, aside from the aforementioned missing low-level rear-to-front flow in the model, comparisons of the model and observation CFADs show good quantitative correspondence. From the RIJ-centered CFADs where frequencies are greater than 1%, the maximum and minimum absolute values of y-axis-parallel squall-line-relative winds at each altitude are similar in magnitude, not varying more than 3 or 4 ms-1 between datasets. Meanwhile, above the melting level, the datasets’ absolute maxima and minima diverge. Using only the locations on RIJ-centered CFADs where the frequency of occurrence was greater than 1%, the model over-predicted the absolute maximum wind speeds at most levels, with the largest disagreement at or near the melting level (as much as 10 ms -1 greater in the model fields). However, the absolute minimum was predicted well (within 3 or 4 ms -1) near the melting level, with the greatest discrepancies 2.0 – 3.0 km above the melting level (9 to 10 ms-1) again biased toward model over-prediction of the absolute wind speed. It is not clear from a CFAD whether it was the magnitude or the direction of the winds that caused the absolute magnitude differences. The model may have artificially enlarged the mesoscale phenomena due 44 to too large (relative to the radar) grid spacing to properly resolve the smaller phenomena. The model RIJ-centered box was longer in the y-dimension (along RIJ) making the populations at each altitude larger, which would enlarge the area required to be represented on a CFAD. This means that there would need to be a larger number of points in any given V bin to produce a statistical representation on a CFAD. As there are larger absolute maxima, there are also larger samples of those faster absolute wind speeds since their representation appeared on CFADs. Another possible reason for the over-prediction of wind speeds was an improper squall line motion vector from the model, causing the storm motion calculations to contain a bias solely from “poor” selection of a squall line motion vector. The manifestation of this in the model CFADs is the orientation of the RIJ-centered box, which determines the y-axis-parallel wind speeds. CFADs have provided useful statistics for comparison of modeled and observed fields. CFADs showed that the model over-predicted radar reflectivity below the melting level and yaxis-parallel squall-line-relative wind speeds above the melting level. It was also noted that the model did not produce an outflow boundary ahead of the convective line or a strong RIJ at the surface. Although CFADs hold great value in summarizing the vertical profiles of variables, they were not designed to study phenomena that show strong gradients in their horizontal dimension. Another statistical method is required for analysis in this direction: contoured-frequency-bydistance diagrams. In the next chapter, we discuss our use of CFDDs. 45 CHAPTER 4 COMPARISONS USING CFDDS 4.1. Contoured-Frequency-by-Distance Diagrams CFDDs were produced for the model simulations and the observations at similar stages (just prior to, during, and after bowing segments occurred) of the MCS’s evolution. The purpose of these diagrams was to compare statistically the horizontal distributions of modeled and observed fields. Model CFDDs were constructed for 12 times: 0412, 0430, 0448, 0515, 0542, 0600, 0618, 0636, 0654, 0712, 0748, and 0815Z. Model times selected maintained a consistent RIJ region from the maximum RIJ wind below 5km, which produced nearly steady state CFDDs over time. Times processed from the observations included 0430, 0450, 0510, 0540, 0610, 0630, 0650, and 0750Z. Observation time selection depended on the proximity of the two airborne radar systems and sufficiently straight flight legs for quad-Doppler synthesis. When the NOAA P-3 performed microphysical spirals, data were not available for analysis as the plane’s heading was changing so rapidly. Herein, comparisons of y-axis-parallel squall-line-relative wind speed from the model and observations used CFDDs made to characterize similarities and differences in the systems’ kinematic structures specifically with respect to the RIJ. Care must be used when interpreting CFDDs presented in this chapter as the distance covered along the y-axis varies, sometimes dramatically, between the model and the observations. On a CFDD, be it model or observations, a dot-dashed black horizontal line indicates the location of the convective-line reflectivity maximum, as in Figs. 4.1 and 4.2. If there is no black line on the diagram, the convective line reflectivity maximum is off the bottom of the diagram, as in observational analyses from 0430 and 0450Z. Also indicated is a standard 46 Fig. 4.1. Contoured-frequency-by-distance diagram (CFDD) of y-axis-parallel squall-linerelative wind speeds from the model simulation. Frequencies colored according to the legend. Time of each panel indicated in top right corner of each panel. Distance defined in kilometers. Wind speeds are ms-1, with 0 ms-1 indicated with the vertical black dashed line. Note the change in y-axis distances when viewing each panel. The dot-dashed horizontal black line indicates the location of the convective line (identified by the maximum reflectivity). 47 Fig. 4.2. Same as Fig. 4.1, except data are from observations. Where no dot-dashed horizontal black line occurs, the convective line is not located within the RIJ-centered box and occurs below zero kilometers distance on the diagram. 48 horizontal distance of twenty-five kilometers to depict similar length scales on each diagram. CFDDs from the model spanned the entire width of the MCS, while observation CFDDs spanned the entirety of the observations at most, with CFDDs for most analysis times not reaching the back edge of the stratiform region. Thus, the region best suited for comparing observed and modeled CFDDs was adjacent to and within the convective line, where the most consistent retrieval of observations occurred. For discussion purposes, the volume was split into regions ahead of and behind the convective line. CFDDs show different phenomena that we will focus on. We will discuss the significance of these diagrams referring to the clustering of points, histogram ranges at a given distance, absolute maximum wind speeds, discrete “streaks” of consistent wind speeds, and their temporal evolution. For reference, the CFDD from the idealized rear-to-front/front-to-rear flow example (Fig. 3.8) from the previous chapter has been included here as Fig. 4.3. On Fig. 4.3, positive wind speeds are rear-to-front and negative wind speeds are front-to-rear flow. On CFDDs of y-axis-parallel squall-line-relative wind speed, the density or clustering of positive or negative velocities at any given distance revealed similar y-axis-parallel squall-linerelative wind speeds with little variation in winds. It is uncertain from these diagrams whether wide distributions were an effect of friction near the surface, a vertically deep, contiguous lowlevel jet, a stable atmosphere conducive of minimal mixing, or another mesoscale phenomenon. A loose grouping (low density of points) represents a distance including varying y-axis-parallel squall-line-relative wind speeds that have variable winds. Ahead of the convective line in the simulated field (Fig. 4.1), the CFDDs for all times showed a distinct clustering of rear-to-front flow (V ranges between 1 and 21 ms-1), which widened approaching the convective line. With time, the widening lessened with approach to the convective line, suggesting the rear-to-front 49 Fig. 4.3. CFDD of idealized rear-to-front/front-to-rear flow pattern shown in Fig. 3.8. 50 flow ahead of the line became more organized. The front-to-rear flow ahead of the convective line consistently showed a varying wind field (V ranges between -19 and -26 ms-1). In the observations ahead of the convective line (Fig. 4.2), CFDDs contained a relatively wide distribution of rear-to-front flow (V ranges between 8 and 21 ms-1) as well as a generally clustered distribution (V ranges between 1 and 25 ms-1) of front-to-rear flow, likely due to a small population size per distance. The density pattern may be reversed between the model and observations because the model grid spacing may not be small enough to resolve some of the smaller phenomena causing variations in the wind field. In addition, the planetary boundary layer parameterization may be over-mixing the atmosphere leading to homogeneity. Modeled front-to-rear flow did not show large changes in histogram ranges from ahead of to behind the convective line (1-26 ms-1 and 1-31 ms-1, respectively), meaning there was some consistency between winds ahead of and behind the convective line. This, however, may have depicted winds from different altitudes that coincidentally have similar histogram widths and were centered in approximately the same velocity bin. Use of average altitude per bin per distance diagrams aid in the interpretation of this change from ahead of to behind the convective line. Significant widening of the rear-to-front flow distributions occurred across the convective line in both the model and observations as shown by visual inspection of CFDDs. Behind the convective line in the simulations, all CFDDs showed generally wide distributions with histogram ranges from 4 to 31 ms-1 for rear-to-front and 1 to 31 ms-1 for front-to-rear flows. The changes from ahead of to behind the convective line were dramatic. Although histogram ranges are relatively consistent, the mean velocity with distance behind the convective line was approximately 12 ms-1 and approximately 22 ms-1 ahead of the convective line in the 51 model. This shift in mean velocity was also present in the observations, where there were retrieved data points ahead of the convective line, but with a smaller shift from front to rear (-19 ms-1 to -11 ms-1, respectively). What may have occurred though was that the winds ahead of the line were from low altitudes and the winds behind the line were high altitudes. This could easily have accounted for a change horizontally through a rear-to-front/front-to-rear couplet. Yet, without the aid of average altitude per bin per distance diagrams, it was difficult to evaluate the height from which each velocity bin got its frequencies. Thus, one could not tell whether the rear-to-front flow ahead of the convective line was at the surface or farther aloft, making interpretation more complex. A number of physical phenomena might have caused the changes from ahead of to behind the convective line. The model may not have resolved the smaller scale variations present, it may be over-mixing the atmosphere, or the winds may have been influenced by the presence of the convection causing divergence from the convective line even in the midlevels. Moving toward the rear edge of the stratiform region, the winds adjusted to consistent frequency clusters or single “streaks” of consistent wind speeds. In the model, these lines start occurring just behind the stratiform precipitation area at each given level extending to the back of the mask. In the observations, there is only a hint of this at 0540, 0610, 0630, and 650Z. The distance covered before the lines orient toward the back of the system changes between each observation time, but is approximately 30 km behind the convective line. In the model, the streaks of consistent wind speeds do not form until much farther back, around 80 km from the convective line. This difference of almost three fold suggested resolution problems in the model because of the 3 km grid spacing. This could however have been due to a stronger component of the winds rearward in the model, which transported particles farther backward creating a larger 52 stratiform rain region, because the front-to-rear flow in and behind the convective line of the simulations was approximately twice as fast. Yet, the highest altitude included in a CFDD is 7km, well below where the greatest transport of ice crystals would occur. It may also be possible that the “streaks” of consistent wind speeds were not yet discrete in the observations. One final possibility is that the RIJ-centered box was oriented more oblique to the convective line, making the stratiform region appear wider, which was quite evident when looking at the RIJ-centered boxes in Figs. 3.2 and 3.3, especially comparing 0654Z from the model (Fig. 3.2) with 0650Z from the observations (Fig. 3.3). Since the stratiform precipitation area sloped rearward with height, the streaks that became discrete closer to the convective line are closer to the surface. Confirmation of this comes with average altitude per bin per distance diagrams. The speeds associated with the absolute maxima of each wind regime from each dataset can be compared. In the model, the maximum rear-to-front wind speeds were behind the convective line at 35 ms-1 compared to a maximum of 29 ms-1 in the observations. The maximum front-to-rear wind speed behind the convective line in the model was -32 ms-1 while their observational counterpart was -26 ms-1. As was the case with reflectivity, the absolute maxima of y-axis-parallel RIJ-centered winds speeds in the model seemed to be over-predicted. However, when comparing only similar widths of trailing stratiform, the observations often had faster rear-to-front flows as the fastest rear-to-front flows in the model generally occurred at a distance farther back in the stratiform region than the observations retrieved. On the other hand, the front-to-rear flow in the model was greater in general, as the observations show only limited amounts of front-to-rear flow behind the convective lines. Without a larger observational sample size, comparisons of front-to-rear flow were not viable. 53 A combination of the last two analyses produced an interesting result from the simulations. The maximum rear-to-front flow behind the convective line at all times was rearward from the first indications of discrete “streaks” of consistent wind speeds. These two phenomena were not likely present at the same altitude, as the first indications of discrete “streaks” are found in the front-to-rear flow, but the altitude was indeterminable from CFDDs. There was no clear physical explanation for why this was occurring. Average altitude per bin per distance diagrams make conclusions about these phenomena possible. In the observational dataset, no definitive statements could be made about these phenomena aside from their location with respect to each other, as the radar data did not extend far enough back in the system to indicate the fastest absolute wind speeds. From the observations present, the locations of the streaks and the maximum RIJ varied wildly with respect to each other. Ahead of the convective line, the maximum rear-to-front flow in the model (35 ms-1) was greater than the maximum rear-to-front flow in the observations (29 ms-1). In this case, the rearto-front flow in the model was over-predicted as the observational and model dataset maxima were both within 20 km of the convective line. The front-to-rear flow maxima for the model and observations are 33 ms-1 at 39 km ahead of the convective line and 27 ms-1 at 31 km ahead of the convective line respectively. Coincidentally, the model over-predicted all the observed maxima by 6 ms-1. This may be due to slightly different wind directions ahead of the convective line in comparison to the observations. A vector more normal to the line may have been present. It is interesting to note that the location of the rear-to-front flow maxima behind the convective line in the model did not vary cyclically over time. This may have been an effect of the RIJ-centered box changing orientation over time. Instead of a forward migration of the RIJ maximum wind speed, the local maxima just behind the convective line increased in speed with 54 time, indicating a local acceleration of winds with time in both datasets, with the caveat that the absolute maximum RIJ wind speed in the observations was likely not sampled on a CFDD due to sampling constraints. Again, it was unclear whether these winds were near the surface or aloft. We will revisit this when talking about average altitude per bin per distance diagrams. 4.2. Average Altitude per Bin per Distance Diagrams In a given y-axis-parallel squall-line-relative velocity bin on a CFDD, winds could have been occurring at any altitude present in the volume. It was impossible to determine the altitude of these winds. For this reason, average altitude per bin per distance diagrams (hereafter, average altitude diagrams) are valuable. The same locations/patterns on a CFDD were plotted on an average altitude diagram, but the mean altitude contributing to each velocity bin at each distance from the front of the RIJ-centered box replaced frequencies. In this thesis, average altitudes were colored in 1 km intervals starting at 0.5 km from the surface, with the surface as its own color; in general, any altitude interval could be used. Since an average altitude was calculated for velocity bins that contained frequencies (as seen on CFDDs), not all areas amounting to less than 1% of the total RIJ-centered box area were assigned an average altitude. Average altitude diagrams used the same velocity conventions as CFDDs. Four patterns emerged from average altitude diagrams of wind speeds (in our case, yaxis-parallel squall-line-relative) that may not be intuitive at first glance. Fig. 4.4 demonstrates three of these possible patterns: the slope of the transition between front to rear and rear to front flow, vertical wind shear, and horizontal wind shear. Fig. 4.5 shows each of these patterns from the idealized case run in Chapter 3 (Fig. 3.8). The fifth pattern (not shown on Fig. 4.4) is an acceleration of the wind speeds at a given altitude and distance, seen by comparing diagrams 55 Fig. 4.4. Average altitude per bin per distance diagram demonstrating how to interpret information from the diagram about horizontal and vertical wind shear, and the slope of an interface. Colored asterisks indicate altitude intervals defined in the legend. All axes and dashed lines same as from CFDDs. 56 Fig. 4.5. Average altitude per bin per distance diagram of idealized rear-to-front/front-to-rear flow pattern shown in Fig. 3.8. 57 from different times. The remainder of this chapter is devoted to explanations of these patterns portrayed in modeled and observed fields. The pattern of a single altitude interval (single color of asterisk on Figs. 4.4, 4.5, 4.6, and 4.7) reveals a large amount of information. Considering temporal migration of the maximum RIJ winds, it was best to define locations as a distance relative to the convective line, since the front of the RIJ-centered box could change location and orientation to the convective line. For example, on Fig. 4.6 at 0654Z, focusing on the altitude interval from 3.5 to 4.0 km (red asterisks), moving rearward from the convective line, the flow began as front-to-rear and transitioned to rear-to-front flow 153 km back from the front of the RIJ-centered box, or 36 km behind the convective line. This change was likely due to the interface between front-torear/rear-to-front flows with height. At higher altitudes, this transition occurred farther rearward from the convective line. Between 4.5 and 5.0 km altitude (orange asterisks), the transition occurs 186 km from the front of the RIJ-centered box, 69 km rearward of the convective line. The closer together the flow interfaces are with height, the steeper the slope of the interface. In the observations (Fig. 4.7), due to the constrained area where data were retrieved, it was not possible to tell where the interface occurred at all altitudes for all times. The best example of the slope of the interface was at 0630Z. At this time, the transition between front-torear and rear-to-front flows occurred 39 - 51 km behind the convective line at 5.5 km and up. Using the convective line as the origin and positive distance defined as distance ahead of the convective line, the average slope of the observed interface was approximately -0.125 km altitude per km horizontal distance. In the simulation, at 0748Z (similar time in MCS evolution) the average slope of the front-to-rear/rear-to-front interface was -.048 km altitude per km horizontal distance and occurred between 63 and 115 km behind the convective line. The slopes 58 Fig. 4.6. Same as Fig. 4.1, except average altitude per bin per distance diagrams. 59 Fig. 4.7. Same as Fig. 4.2, except average altitude per bin per distance diagrams. 60 of the transitions were a factor of 2-3 different, suggesting possibly a grid spacing dependence to the mismatch. However, these differences may have other explanations. The descent of the RIJ was also dependent on microphysical and dynamical forcing. The forcing from the microphysics or dynamics from the model compared to the observations was different. Comparisons of observed and modeled vertical motions might shed some light on this hypothesis. We could also compare the heights of the interface between rear-to-front and front-to-rear flows between the model and the observations. In the observations from 0510 until 0630Z, the interface between the two flow regimes was 1.5 to 2.0 km above ground. After 0630Z, the interface had lowered to between 1.0 and 1.5 km above ground. In the model, the flow interface descended over time, but only within the last time of the model analysis did it reach 1.5 to 2.0 km, where the observations generally had the interface located. The rear-to-front/front-to-rear flow interface in the model was always higher than in the observations, suggesting a higher altitude or less descended RIJ. When different average altitude intervals occurred at a given distance, this inferred vertical wind shear between the two altitudes (especially if the altitudes were adjacent) as noted on Fig. 4.4 as “Vertical Shear.” There can also be vertical wind shear in a single altitude interval, and this likely occurred, but our altitude interval (1 km) masked some of these possibilities. On Fig. 4.6 (0430Z), each altitude interval showed discrete vertically aligned streaks in the rear of the system at a single distance from the convection. At 390 km or 184 km rearward from the convective line, moving from left to right across the diagram showed increasingly rear-to-front flow with an increase in altitude. This was an example of vertical wind shear. In plan views of modeled squall-line-relative wind vectors, there were few locations that showed significant horizontal wind shear in a given x-z slab (not shown). Horizontal wind shear 61 can be seen by noting a change of wind speed with distance from the convective line (along the RIJ-axis) at the same altitude. It was impossible however to tell directional from speed shear (vertical or horizontal) using average altitude diagrams. For instance, at 0712Z on Fig. 4.6, the 0.5 and 1.0 km average altitudes (green asterisks) showed discrete lines in front-to-rear velocity bins well behind the convective line (approximately 100 km to the back edge of the system). In a purely y-axis-parallel sense, there was only speed shear, as the winds at each level were of the same sign but different speeds. In a squall-line-relative sense though, there is directional and/or speed shear, as the squall-line-relative winds could have been of different directions, different speeds, or a combination of the two and maintained the same resultant velocity bins at each level on the average altitude diagram. Average altitude diagrams must be interpreted in the squallline-relative sense to gain a physical understanding of the wind field. Rotunno et al., (1988) extrapolated that a balance in the horizontal vorticity (vertical shear) of the cold pool with the pre-storm low-level horizontal vorticity (vertical shear) governed the verticality of the convective line, thus the longevity of the MCS. Here, we compared the vertical shear present (from the horizontal components of the wind only) in given layers at given distances with respect to the convective line by using an average altitude diagram. We used 15 km from the convective-line reflectivity maxima as our analysis location. This location was close to the convection, but hopefully far enough from the convection to be outside of the strongest convective downdrafts. Wind shear was calculated by subtracting the maximum and minimum wind speeds in the lowest two kilometers (excluding the surface). Wind shear associated with only the rear-to-front and front-to-rear flows were used, as contributions from the other regimes in the given locations would have been associated with phenomena not discussed in Rotunno et al., (1988). At 0712Z (Fig. 4.6), 15 km behind the convective line in the lowest 62 2.0 km, excluding the surface (for comparison purposes), there were -8 ms-1 of vertical wind shear from the rear-to-front flow. 15 km ahead of the convective line, there were 5 ms-1 of vertical wind shear from front-to-rear flow. In the observations at a similar time in the evolution of the MCS (0610Z on Fig. 4.7), the vertical shear associated with the rear-to-front flow 15 km behind the convective line was -5 ms-1. While ahead of the convective line, there were 5 ms-1 of vertical wind shear from the front-to-rear flow. In both datasets, there was significant front-torear flow behind the convective line indicative of the cold pool spreading out along the surface forward as well as rearward. By comparison, the modeled and observed vertical shears were similar, with an indication that the model cold pool or RIJ near the surface produced more vertical shear than in the observations. According to the average altitude diagrams from 15 km behind the convective line, the model under-predicted the winds at 0.5 km, and over-predicted the winds at 2.0 km. It can be inferred that the friction layer of the model was not properly modeled or the rear-to-front flow penetrated closer to the surface in the observations. Since the contributions of the vertical velocities are not included in the calculations, horizontal vorticity was not formally calculated. Horizontal wind shear could be interpreted from average altitude diagrams by noting velocity changes of a single altitude interval with distance behind the convective line (noted on Fig. 4.4 as “Horizontal Shear between slabs at same altitude”). Also, where one average altitude interval occurred in multiple velocity bins at a given distance, one could measure the difference in wind speed between these velocity bins and find the horizontal wind shear present (Fig. 4.4; “Horizontal Shear between slabs at same altitude”). This however, cannot be accurately determined in our work as our average altitude interval covered multiple heights of the system. Either of these methods for finding velocity changes could lead to misidentification of the type 63 of change, as these were only average altitude contributions to a single velocity bin, and the winds contributing to these bins could be located on opposite ends of the x-z slab. Horizontal shear as noted above could appear as deformation, divergence, or vorticity, but it is impossible to tell the difference between these three without careful review of plan-views and cross-sections of the datasets. For this reason, a limited analysis was performed with this data. To examine wind acceleration, one has to look at the evolution of the winds at similar distances from the convective line on multiple diagrams. At the same distance, acceleration is represented by increasing wind speeds at a given altitude. The reason for the acceleration cannot be deduced. Acceleration could appear as an area of faster wind speeds translated through that given area, or dynamical/microphysical changes could accelerate the winds locally. Over time on Fig. 4.6, the maximum rear-to-front flow (assumed RIJ) progressed rearward. This may be due to increased entrainment in the stratiform region while precipitation expanded rearward or another local acceleration. This is especially telling when considering the altitude at which the maximum RIJ wind speeds occurred in the model (6.5 – 7.0 km). With time though, the front to rear clustering of the average altitude intervals at their respective maximum velocities began to spread out along the distance axis, suggesting that the highest altitudes plotted here may be representative of the larger scale flow, not the mesoscale. Lower altitudes were analyzed to check for a translation of the maximum RIJ wind speeds. Even at 2.5 – 3.0 km, the maximum RIJ wind speeds showed a trend toward rearward translation. The observations mimic this action, but with a faster rearward progression, but only at altitudes of 6.0 km and below. Above 6.0 km, the shorter length scale of the observed RIJ-centered box limited the analysis of the maximum RIJ wind speeds. 64 The maximum RIJ wind speed locations at 2.5 – 3.0 km were closer to the convective line than the maximum RIJ winds at 6.5 – 7.0 km. This showed that the maximum RIJ wind speeds were tilted downward with distance toward the convective line, suggesting descent of the RIJ, albeit not to the surface. The observed average altitude diagrams (Fig. 4.7) maintained this trend in downward tilt, with only slow rear-to-front wind speeds near the surface. The most significant finding from these diagrams was the acceleration of rear-to-front flow occurring at and below 2 km altitude. In both observations and the model, the rear-to-front flow strengthened. In the model average altitude diagrams (Fig. 4.6) from 0412 until 0815Z, there was an increase of wind speeds from 10 to 17 ms-1 at 1.5 - 2.0 km, and an increase of the wind speeds between 0.5 and 1.0 km from -2 ms-1 to 13 ms-1. In the observations (Fig. 4.7) from 0510 to 0750Z, the 1.5 – 2.0 km altitudes varied from 11 to -10 ms-1 non-cyclically in rear-tofront wind speed, and at .5 to 1.0 km, wind speeds fluctuated between -5 to 12 ms-1, showing no consistent pattern of increase or decrease between analysis times. Each interval of average altitudes showed discrete bands on average altitude diagrams, we refer to these bands as altitude streaks. Where there are nearly straight altitude streaks with distance from the convective line, it was simple to find two separate streaks for the same altitude interval. These streaks are an effect of the interval containing two altitudes. Occasionally, these streaks straddle the zero velocity line, meaning the interface between front-to-rear and rear-tofront flow occurred between those two levels. This was evident on the average altitude diagrams from 0636 and 0712Z in Fig. 4.6 as well as 0540 and 0630Z in Fig. 4.7. The distance rearward from the convective line where each of these altitude streaks began was coincident with the distance to the rear of the convection where the stratiform precipitation ended at that level, found using Fig. 3.2 in conjunction with average altitude diagrams. 65 Additionally, the slope of the back edge of the stratiform precipitation can be interpreted from the average altitudes. As one travels upward through the system, the stratiform precipitation area should reach farther rearward from the convective line. This slope is demonstrated the same way as the interface between the front-to-rear/rear-to-front winds. In the observations, the back edge of the stratiform precipitation region was not visible due to the limited sampling by the airborne radars. In the model though, the back edge of the stratiform rain region near the surface (0.5 – 1.0 km) changes with time (42 km rearward shift), but in a decidedly non-cyclic fashion. In the 2.5 - 3.0 km altitude range, the back edge of the stratiform precipitation region varied only slightly more (46 km rearward shift), but showed a strong rearward migration from 0654 to 0815Z. Both of these altitude interval analyses suggested an expansion of the stratiform precipitation region, but only a slightly faster migration aloft, which is unexpected since this is much closer to where ice crystals would have been deposited from the divergence at the torpopause in the convective updraft. With average altitude diagrams, it became obvious that most of the front-to-rear flow behind the convective line in either dataset came from the lower altitudes. In the front-to-rear flow behind the convective line from observations (Fig. 4.7), the dominant altitude contributions were from below 1 km altitude, with the model dataset (Fig. 4.6) showing dominant altitudes below 2 km. This suggested a generally higher altitude RIJ in the model than in the observations, as the dominant altitude contributions in the front-to-rear flow were higher for the model than the observations, likely meaning a thinner cold pool in the observations. As usual, the observations may not have extended far enough behind the convective line to sample the bulk of the stratiform region, so comparisons only applied in close proximity to the convective line. 66 CFDDs and average altitude diagrams could be used for a broad range of meteorological phenomena. Using average altitude diagrams, we gleaned a diverse set of information including the slope of the rear-to-front/front-to-rear flow interface, slope of the back edge of the trailing stratiform region, expansion of the trailing stratiform region, horizontal and vertical wind shear, rearward migration of the maximum RIJ y-axis-parallel squall-line-relative wind speed, and accelerations at any given altitude over time. In several instances, CFDDs and average altitude diagrams suggested a model resolution problem, as the simulations occasionally contained distances up to 3 times greater in length than in the observed fields. With the help of vertical velocity CFDDs, a field that we did not plot, one could extract all the contributions to calculate horizontal vorticity values and compare model to observed vorticity balances, to reinforce some of the points made in RKW Theory (Rotunno et al., 1988). More information about the cold pool of a MCS and its relationship to the pre-storm environmental vertical wind shear could result from this analysis. This may be useful as calculating cold pool strength (perturbation temperatures) can be tough to quantify from an observational dataset. 67 CHAPTER 5 CONCLUSIONS 5.1. Conclusions With a high-resolution observational dataset of an MCS obtained on 10 June 2003 during the BAMEX field campaign and a similarly high-resolution WRF simulation of the same MCS, an opportunity to compare statistical distributions of similar meteorological fields. Our intention was to evaluate the model’s ability to simulate the 10 June system using previously developed statistical methods and a newly devised method. Comparisons of CFADs and CFDDs using modeled reflectivity and kinematics to those derived from airborne dual- and quad-Doppler radar syntheses were used to quantify the simulated squall line morphology, rear-inflow jet evolution, and bulk cloud properties in this storm system. Comparing these statistical diagrams one can minimize errors associated with poor spatial and temporal collocation of the model and observations. CFADs were used to describe the vertical distributions of reflectivity and horizontal velocities, ignoring their horizontal location behind the convective line of the MCS. Specific comparisons conducted using CFADs included the frequencies of reflectivity values at given altitudes, convective pulses, and the evolution of the frequencies of reflectivity values. In CFADs masked to include only locations where the reflectivity was greater than 0 dBZ, it was easily seen that the model over-predicted the frequency of reflectivities above 35 dBZ. In particular, the model over-predicted the value of the reflectivity bin where the maximum frequencies occurred. Below the melting level, the modeled reflectivity value of maximum frequency at all analysis times fell within the range of 30-40 dBZ on masked CFADs, while in 68 the observations the maximum frequency varied from 20 to 40 dBZ. Even using RIJ-centered CFADs, for which the observations and models have similar domain sizes, the model overpredicted the frequency of reflectivity values above 35 dBZ. In these RIJ-centered CFADs, the reflectivities at which the peak occurrence frequencies occurred were similar for the model and observations, but the peak frequencies were higher for the model. For example, frequencies as high as 31-36% occurred in the model at 1 km, while in the observations the frequencies peaked at 26-31%. This seemed initially like a similarity. However, closer inspection of the physical areas that these two RIJ-centered boxes covered showed that the model CFADs covered longer horizontal distances along their y-axis, meaning many more points were required to give the same normalized frequency of occurrence as plotted in the CFADs. The peak frequency below 7 km from each dataset was the same, but the extended length of the model RIJ-centered box, still indicated that there was more such reflectivity area in the model output.. Thus, not only was the model over-predicting the frequencies of reflectivities > 35 dBZ, but it was also over-predicting the area covered by them from both masked and RIJ-centered CFADs. Another key finding was that a “hole” was produced below the melting level in the masked CFADs of simulated reflectivity. This hole indicated that reflectivity frequencies fell to below 1% in all the bins between 5 and 30 dBZ, in at least the lowest two kilometers of the simulation. This suggested that there may be errors in the parameterization of microphysics or other physical processes that caused the absence of such reflectivity values in the models, but not in the observations for the 10 June 2003 squall line. The observed masked CFADs all contained significant portions of their reflectivity structure within this 5 – 30 dBZ range, and sometimes the peak observed frequencies occurred in this range. 69 The model, for all analysis times, over-predicted the value of the reflectivity bin that contained the maximum frequency. Over-prediction of reflectivity values could come from many different sources including: model microphysical parameterizations, tilting of the convective towers, and/or under-prediction of vertical velocities due to vertical or horizontal grid spacing. The vertical velocities associated with these convective updrafts were not analyzed, but the height of the convective pulses was evident on the CFADs. Convective towers reached 2.5 km higher in the observations than in the model, suggesting that the vertical velocities in the model were under-predicted. This under-prediction may have resulted from vertical or horizontal grid spacing or an over-prediction of precipitable mass in the simulated clouds. Further analysis of the vertical velocities would help identify some of these uncertainties. In the absence of any other forcing, stronger vertical velocities would be expected to produce greater amounts of condensate and hence higher reflectivities. Thus, the combination of weaker velocities and stronger reflectivity in the models compared to observations further suggests thaat the parameterizations of microphysics in the model may not be accurately portraying what occurs in nature. The reflectivity CFADs, either masked or RIJ-centered, showed a rapidly evolving system in the observations and a nearly steady state in the model, a stark contrast. The best example of this has already been stated, the value of the bin in which the maximum frequency occurred. Over time, the observations showed a strong variation (20 dBZ), while the model maximum frequency never changed over the analysis period. Although the melting layer is not visible on average altitude diagrams of y-axis-parallel squall-line-relative winds, important information can be gleaned from the location of the transition between the rear-to-front and front-to-rear flow with respect to the melting layer. In 70 the observations, the transition between rear-to-front and front-to-rear flow occurred at generally between 1.5 and 2.0 km above ground, eventually lowering to 1.0 to 1.5 km above ground. The interface between these flow regimes was located between 1.0 and 2.0 km below the melting level. In the model, the interface between the two flow regimes was located at 2.5 to 3.0 km above ground, which was consistently located 0.5 to 1.0 km below the melting level. CFADs were used to show the vertical variation of y-axis-parallel squall-line-relative winds. The winds showed the same differences in evolution, with the model showing steady state behavior and the observations more dynamic. In the observations at 1 km, the the presence of RIJ winds or the winds behind an outlfow boundary are shown in all times where the RIJcentered box reaches ahead of the convective line. The winds at 1 km in the model did increase speed over time, but at no point in the analysis times did it show a similar low-level “nose” from the RIJ or outflow as seen in the observations. This may have been due to the model planetary boundary layer scheme, a balance between the horizontal vorticity of the cold pool and the prestorm environment, a weaker cold pool, or a higher altitude RIJ in the model. Below the melting layer, the observations and the model show very similar patterns of yaxis-parallel squall-line-relative wind speeds, aside from the aforementioned RIJ or outflow boundary. Above the melting layer in the model, the absolute value of the wind speeds was overpredicted by the model at all levels. Possible explanations include a stronger RIJ at mid-levels in the model, a non-descending RIJ in the model, horizontal grid spacing too large to simulate the small scale motions occurring in the observations, and too large a pressure gradient in the model accompanied by an over-prediction of the friction layer deceleration. Over time, the below 2 km wind speeds in the model and observations increased directly behind the convective line. The model showed a larger increase in rear-to-front flow over time 71 than the observed rear-to-front flow, suggesting that the model RIJ descending, the cold pool was strengthening, or another unknown local acceleration was occurring. CFDDs took all the advantages of CFADs and turned them on their side to portray horizontal distributions instead of vertical distributions. CFDDs helped reinforce some of the points made by CFADs, plus added a few more key points. CFDDs ignore the vertical variability that CFADs were designed to capture. We designed CFDDs to show horizontal distributions of horizontally varying phenomena, such as RIJs. CFDDs of y-axis-parallel squall-line-relative winds were useful for diagnosing consistent streaks of velocities behind the convective line. The distance behind the convective line where these discrete bands of velocity formed was nearly 3 times longer in the model than the observations. This may suggest the model grid spacing was not small enough to properly resolve the winds on this scale or that the trailing stratiform region in the model was wider also possibly due to horizontal and vertical grid spacing, or that the orientation of the analysis domain (a function of the RIJ angle to the leading convective line) differed between the model and the observations and influenced the statistics. CFDDs of y-axis-parallel squall-line-relative velocities yield information about the horizontal location where the over-prediction of wind speeds above the melting layer observed in the CFADs originated. For the same behind line horizontal distance, observed CFDDs showed faster rear-to-front wind speeds than the model predicted at most times. This suggests that the fastest rear-to-front wind speeds present in the observations may be present below the melting level and close to the convective line since the over-prediction of wind speeds in the model was above the melting level. Another new analysis tool, average altitude per bin per distance diagrams aided in our look at this phenomenon. 72 Average altitude per bin per distance diagrams show the average altitude which contributies to the reflectivity frequencies greater than 1% on a CFDD. The interpretation of these diagrams provides information about the horizontal and vertical shear, accelerations of wind on a given level, and transitions between flow regimes. Not all of these attributions can be made using data from the 10 June 2003 case analyzed, but because of the multiple different pieces of information that can be gained through their use, average altitude per bin per distance diagrams show great potential for future work on RIJs and other horizontally varying phenomena. Some of the fastest rear-to-front wind speeds present on average altitude diagrams are in fact close to the convective line and below the melting level. This is consistent with the fact that the observations contained a RIJ that descended closer to the surface and/or an outflow boundary that pushed under the convective line, but not ahead of the convection. When comparing the CFDDs of models and observations, it was noted that the consistent streaks of wind speeds occurred at different distances from the convective line. With average altitude diagrams, it was shown that the different streaks were actually different altitudes. Further, the distance, with respect to the convective line, between the two different altitude intervals helped derive the slope of the back edge of the stratiform precipitation. Slopes of the back edge of the stratiform region were three fold different between the models and observations, with the model having a more gradual slope. This again suggests that the horizontal resolution was not fine enough in the model. In the model as noted on CFDDs, the maximum rear-to-front winds were located behind where the consistent streaks of wind speeds occurred. Here it is noted that these two phenomena were indeed occurring at different levels. The consistent streaks closest to the convective line were from the lowest levels of the system, while the maximum RIJ 73 winds were noted at 6.5 – 7.0 km altitude. Both of these areas, where the lowest consistent altitude streaks began and where the maximum RIJ winds occurred, migrated rearward over time. The former suggests that the stratiform region of the storm was enlarging, although slowly, while the latter suggests that the maximum RIJ winds were moving rearward over time as well, with a similar rate of backward propagation. This suggests a link between the back edge of the stratiform region and the maximum RIJ winds. It has been shown that these statistical diagrams can be used to quantify the robustness of simulations against high-resolution observations even without spatial or temporal collocation. Using all of the statistical methods mentioned in this thesis for comparisons between modeled and observed systems could help to identify where model parameterizations are lacking a physical understanding of the atmosphere or where we may need higher resolution observations to continue to test our understanding. Quantitative statistical analyses performed with these diagrams represent the bulk characteristics of the system much better by comparing the distributions with height or distance. After quantifying the model’s ability to represent the features in question, modelling studies can be used to their full potential, knowing that the simulations behave in a way that is statistically representative of the physical phenomena. 74 REFERENCES Atkins, N. T., C. S. Bouchard, R. W. Przybylinski, R. J. Trapp, and G. Schmocker, 2005: Damaging surface wind mechanisms within the 10 June 2003 Saint Louis bow echo during BAMEX. Mon. Wea. Rev., 133, 2275–2296. Belcher, L. R., L. D. Carey, J. M. Davis, J. A. Kankiewicz, and T. H. Vonder Haar, 2003: Mmwave radar structure and microphysical characteristics of mixed phase altocumulus clouds. 31st International Conference on Radar Meteorology, Seattle, Amer. Meteor. Soc., 3.6. Biggerstaff, M. I., and R. A. Houze, 1991: Kinematic and precipitation structure of the 10–11 June 1985 squall line. Mon. Wea. Rev., 119, 3034–3065. __________, and __________, 1993: Kinematics and microphysics of the transition zone of the 10–11 June 1985 squall line. J. Atmos. Sci., 50, 3091–3110. Braun, S. A., and R. A. Houze, 1994: The transition zone and secondary maximum of radar reflectivity behind a midlatitude squall line: results retrieved from Doppler radar data. J. Atmos. Sci., 51, 2733–2755. Bryan, G. H., J. C. Wyngaard, and J. M. Fritsch, 2003: Resolution requirements for the simulation of deep moist convection. Mon. Wea. Rev., 131, 2394–2416. Caniaux, G., J. L. Redelsperger, and J. P. Lafore, 1994: A numerical study of the stratiform region of a fast-moving squall line. Part I: general description and water and heat budgets. J. Atmos. Sci., 51, 2046–2074. Chen, F., and J. Dudhia, 2001: Coupling an advanced land-surface/ hydrology model with the Penn State/NCAR MM5 modeling system. Part I: model description and implementation. Mon. Wea. Rev., 129, 569–585. Cifelli, R., C. R. Williams, D. K. Rajopadhyaya, S. K. Avery, K. S. Gage, and P. T. May, 2000: Drop-size distribution characteristics in tropical mesoscale convective systems. J. Appl. Meteor., 39, 760–777. Davis, C., N. Atkins, D. Bartels, L. Bosart, M. Coniglio, G. Bryan, W. Cotton, D. Dowell, B. Jewett, R. Johns, D. Jorgensen, J. Knievel, K. Knupp, W. C. Lee, G. Mcfarquhar, J. Moore, R. Przybylinski, R. Rauber, B. Smull, R. Trapp, S. Trier, R. Wakimoto, M. Weisman, and C. Ziegler, 2004: The bow echo and MCV experiment: observations and opportunities. Bull. Amer. Meteor. Soc., 85, 1075–1093. Dawson, D. T., and M. Xue, 2006: Numerical forecasts of the 15–16 June 2002 southern plains mesoscale convective system: impact of mesoscale data and cloud analysis. Mon. Wea. Rev., 134, 1607–1629. 75 Dudhia, J., 1989: Numerical study of convection observed during the winter monsoon experiment using a mesoscale two-dimensional model, J. Atmos. Sci., 46, 3077–3107. Ferrier, B. J., Simpson, and W. K. Tao, 1996: Factors Responsible for Precipitation Efficiencies in Midlatitude and Tropical Squall Simulations. Mon. Wea. Rev., 124, 2100–2125. Fritsch, J., R. Kane, and C. Chelius, 1986: The contribution of mesoscale convective weather systems to the warm-season precipitation in the United States. J. Appl. Meteor., 25, 1333– 1345. Gallus, W. A., 1996: The influence of microphysics in the formation of intense wake lows: a numerical modeling study. Mon. Wea. Rev., 124, 2267–2281. Geerts, B., and G. M. Heymsfield, 2000: High-resolution reflectivity and vertical velocity profiles in various convectively-generated precipitation systems. 24th Conference on Hurricanes and Tropical Meteorology,Ft. Lauderdale, Amer. Meteor. Soc., 16B.3. Grady, R. L., and J. Verlinde, 1997: Triple-Doppler analysis of a discretely propagating, longlived, High Plains squall line. J. Atmos. Sci., 54, 2729–2748. Grams, J. S., W. A. Gallus, S. E. Koch, L. S. Wharton, A. Loughe, and E. E. Ebert, 2006: The use of a modified Ebert–McBride technique to evaluate mesoscale model QPF as a function of convective system morphology during IHOP 2002. Wea. Forecasting, 21, 288–306. Hong, S.-Y., and H.-L. Pan, 1996: Nonlocal boundary layer vertical diffusion in a medium-range forecast model, Mon. Wea. Rev., 124, 2322–2339. __________, Y. Noh, and J. Dudhia, 2006: A new vertical diffusion package with an explicit treatment of entrainment processes. Mon. Wea. Rev., 134, 2318–2341. Houze, R. A., M. Biggerstaff, S. Rutledge, and B. Smull, 1989: Interpretation of Doppler weather Radar displays of midlatitude mesoscale convective systems. Bull. Amer. Meteor. Soc., 70, 608–619. Janjic, Z. I., 1994: The step-mountain eta coordinate model: further developments of the convection, viscous sublayer and turbulence closure schemes, Mon. Wea. Rev., 122, 927– 945. __________, 1996: The surface layer in the NCEP Eta model, Eleventh Conference on Numerical Weather Prediction, Norfolk, VA, 19–23 August; Amer. Meteor. Soc., Boston, MA, 354–355. Johnson, R. H., and P. J. Hamilton, 1988: The relationship of surface pressure features to the precipitation and airflow structure of an intense midlatitude squall line. Mon. Wea. Rev., 116, 1444–1473. 76 Jorgensen, D. P., T. Matejka, and J. D. DuGranrut, 1996: Multi-beam techniques for deriving wind fields from airborne Doppler radars. J. Meteor. Atmos. Phys, 59, 83-104. __________, M. A. LeMone, and S. B. Trier, 1997: Structure and evolution of the 22 February 1993 TOGA COARE squall line: aircraft observations of precipitation, circulation, and surface energy fluxes. J. Atmos. Sci., 54, 1961–1985. __________, T. R. Shepherd, and A. S. Goldstein, 2000: A dual-pulse repetition frequency scheme for mitigating velocity ambiguities of the NOAA P-3 airborne Doppler radar. J. Atmos. Oceanic Technol., 17, 585–594. __________, H. V. Murphey, and R. M. Wakimoto, 2005: Rear-inflow structure in severe and non-severe bow-echoes observed by airborne Doppler radar during BAMEX. Preprints, 32nd Conference on Radar Meteorology, Albuquerque, New Mexico, J5J.3. Kane, R., C. Chelius, and J. Fritsch, 1987: Precipitation characteristics of mesoscale convective weather systems. J. Appl. Meteor., 26, 1345–1357. Kingsmill, D. E., P. J. Neiman, F. M. Ralph, and A. B. White, 2006: Synoptic and topographic variability of northern California precipitation characteristics in landfalling winter storms observed during CALJET. Mon. Wea. Rev., 134, 2072–2094. Lang, S., W. K. Tao, J. Simpson, and B. Ferrier, 2003: Modeling of convective–stratiform precipitation processes: sensitivity to partitioning methods. J. Appl. Meteor., 42, 505–527. LeMone, M. A., and M. W. Moncrieff, 1994: Momentum and mass transport by convective bands: comparisons of highly idealized dynamical models to observations. J. Atmos. Sci., 51, 281–305. McFarquhar, G. M., H. Zhang, G. Heymsfield, R. Hood, J. Dudhia, J. B. Halverson, and F. Marks, 2006: Factors affecting the evolution of hurricane Erin (2001) and the distributions of hydrometeors: role of microphysical processes. J. Atmos. Sci., 63, 127–150. Mlawer, E. J., S. J. Taubman, P. D. Brown, M. J. Iacono, and S. A. Clough, 1997: Radiative transfer for inhomogeneous atmosphere: RRTM, a validated correlated-k model for the longwave. J. Geophys. Res., 102 (D14), 16663–16682. Montmerle, T., J. P. Lafore, and J. L. Redelsperger, 2000: A tropical squall line observed during TOGA COARE: extended comparisons between simulations and Doppler radar data and the role of midlevel wind shear. Mon. Wea. Rev., 128, 3709–3730. Mori, S., Hamada, J. I., Yamanaka, M. D., Kodama, Y. M., Kawashima, M., Shimomai, T., Shibagaki, Y., Hashiguchi, H., and Sribimawati, T., 2006: Vertical wind characteristics in precipitating cloud systems over West Sumatra, Indonesia, observed with Equatorial 77 Atmosphere Radar: case study of 23-24 April 2004 during the first CPEA campaign period. JMSJ, 84A, 113-131. Pasken, R. W., and Martinelli, J. T., 2006: High resolution numerical simulations of Midwestern quasi-linear mesoscale convective systems. 23rd Conference on Severe Local Storms, St. Louis, Amer. Meteor. Soc., 17.5. Rogers, R., M. Black, S. S. Chen, and R. Black, 2004: Evaluating microphysical parameterization schemes for use in hurricane environments. 26th Conference on Hurricanes and Tropical Meteorology, Miami, Amer. Meteor. Soc., 13C.6. __________, __________, F. Marks, K. Valde, and S. S. Chen, 2006: A comparison of tropical cyclone hydrometeor profiles from TRMM, airborne radar, and high-resolution Simulations. 27th Conference on Hurricanes and Tropical Meteorology, Monterey, Amer. Meteor. Soc., P5.19. __________, __________, S. S. Chen, and R. A. Black, 2007: An evaluation of microphysics fields from mesoscale model simulations of tropical cyclones. Part I: comparisons with observations. J. Atmos. Sci., 64, 1811–1834. Rotunno, R., J. B. Klemp, and M. L. Weisman, 1988: A theory for strong, long-lived squall lines. J. Atmos. Sci., 45, 463–485. Rutledge, S. A., R. A. Houze, M. I. Biggerstaff, and T. Matejka, 1988: The Oklahoma–Kansas mesoscale convective system of 10–11 June 1985: precipitation structure and singleDoppler radar analysis. Mon. Wea. Rev., 116, 1409–1430. Skamarock, W. C., J. B. Klemp, J. Dudhia, D. O. Gill, D. M. Barker, W. Wang and J. G. Powers, 2005: A description of the advanced research WRF version 2. NCAR Tech. Note. NCAR/TN-468+STR, [http://www.mmm.ucar.edu/wrf/users/docs/arw_v2.pdf.] Smedsmo, J. L., E. Foufoula-Georgiou, V. Vuruputur, F. Kong, and K. Droegemeier, 2005: On the vertical structure of modeled and observed deep convective storms: insights for precipitation retrieval and microphysical parameterization. J. Appl. Meteor., 44, 1866– 1884. Smith, A. M., R. M. Rauber, G. M. McFarquhar, B. F. Jewett, M. S. Timlin, and J. A. Grim, 2008: Explaining variations in cloud microphysics in BAMEX MCSs using high resolution radar and optical array probe measurements. Submitted to Mon. Wea. Rev. Smith, D. K., F. J. Wentz, and C. A. Mears, 1999: Comparison of TRMM TMI sea surface temperature and wind speed retrievals with measurements from selected NDBC, TAO and PIRATA buoys. 1999 American Geophysical Union Spring Meeting. Smith, P. L., 1986: On the sensitivity of weather radars. J. Atmos. Oceanic Technol., 3, 704–713. 78 Smull, B. F., and R. A. Houze, 1985: A midlatitude squall line with a trailing region of stratiform rain: radar and satellite observations. Mon. Wea. Rev., 113, 117–133. __________, and __________, 1987a: Dual-Doppler radar analysis of a midlatitude squall line with a trailing region of stratiform rain. J. Atmos. Sci., 44, 2128–2149. __________, and __________, 1987b: Rear inflow in squall lines with trailing stratiform precipitation. Mon. Wea. Rev., 115, 2869–2889. Steiner, M., R. A. Houze, and S. E. Yuter, 1995: Climatological characterization of threedimensional storm structure from operational radar and rain gauge data. J. Appl. Meteor., 34, 1978–2007. Stoelinga, M. T., 2004: An update on post-processing and visualization tools for the WRF Model. Preprints, 5th WRF / 14th MM5 Users’ Workshop, Boulder, Colorado, NCAR, 168-169. _________, 2005: Simulated equivalent reflectivity factor as currently formulated in RIP: description and possible improvements. [http://www.atmos.washington.edu/~stoeling/RIP_sim_ref.pdf.] Storm, B. A., M. D. Parker, and D. P. Jorgensen, 2007: A convective line with leading stratiform precipitation from BAMEX. Mon. Wea. Rev., 135, 1769–1785. Swann, A., A. H. Sobel, S. E. Yuter, and G. N. Kiladis, 2006: Observed radar reflectivity in convectively coupled Kelvin and mixed Rossby-gravity waves. Geophys. Res. Lett. 33, no. 10. Thompson, G., R. M. Rasmussen, and K. Manning, 2004: Explicit forecasts of winter precipitation using an improved bulk microphysics scheme. Mon. Wea. Rev., 132, 519542. Tokay, A., and D. A. Short, 1996: Evidence from tropical raindrop spectra of the origin of rain from stratiform versus convective clouds. J. Appl. Meteor., 35, 355–371. Wakimoto, R. M., H. V. Murphey, A. Nester, D. P. Jorgensen, and N. T. Atkins, 2006a: High winds generated by bow echoes. Part I: overview of the Omaha bow echo 5 July 2003 storm during BAMEX. Mon. Wea. Rev., 134, 2793–2812. __________, __________, C. A. Davis, and N. T. Atkins, 2006: High winds generated by bow echoes. Part II: the relationship between the mesovortices and damaging straight-line winds. Mon. Wea. Rev., 134, 2813–2829. Weisman, M. L., W. C. Skamarock, and J. B. Klemp, 1997: The resolution dependence of explicitly modeled convective systems. Mon. Wea. Rev., 125, 527–548. 79 __________, and R. Rotunno, 2004: “A theory for strong long-lived squall lines” revisited. J. Atmos. Sci., 61, 361–382. Wheatley, D. M., R. J. Trapp, and N. T. Atkins, 2006: Radar and damage analysis of severe bow echoes observed during BAMEX. Mon. Wea. Rev., 134, 791–806. Yang, M. H., and R. A. Houze, 1995: Sensitivity of squall-line rear inflow to ice microphysics and environmental humidity. Mon. Wea. Rev., 123, 3175–3193. Yuter, S. E., and Houze Jr., R. A., 1995: Three-dimensional kinematic and microphysical evolution of Florida cumulonimbus. Part II: frequency distributions of vertical velocity, reflectivity, and differential reflectivity. Mon. Wea. Rev., 123, 1941-1963. __________, __________, E.A. Smith, T.T. Wilheit, and E. Zipser, 2005: Physical characterization of tropical oceanic convection observed in KWAJEX. J. Appl. Meteor., 44, 385–415.