Observational research continues to confirm the prevalence of tropopause-level perturbations preceding surface cyclogenesis. The observations also suggest that the growing disturbances have time-varying vertical structures. Relating these observations to the classical theory of baroclinic instability is not immediately obvious, since in the latter, the precise form of the initial condition is not important, and the theory predicts cyclogenesis with a fixed-in-time vertical structure. For these reasons and others, the relevance of baroclinic instability to cyclogenesis has been called into question. To explain these observations, it has been suggested that real cases of development might be better understood by looking at more general initial-value problems where the near-term solution would generally involve various compositions of a system's normal modes, rather than the single exponentially unstable normal mode of the textbook variety. It has even been suggested that real systems are stable (i.e., there are no exponentially unstable normal modes) and, therefore, that any growth that occurs must be understood by composition effects of modes that are exponentially stable. Investigating these issues, Richard Rotunno and Jian-Wen Bao (ASP postdoctoral visitor) showed that a simple composition of modes is needed to describe the time-evolving vertical structure and growth rate of a real system; however, one of the modes in the composition is the exponentially unstable normal mode, and is crucial to explaining the observed development.
In addition to the synoptic-scale perturbations evident prior to surface cyclogenesis (as in the study outlined above), forecasters and synopticians often focus their attention on smaller-scale disturbances, such as jet streaks and short waves, embedded within the synoptic-scale wave pattern. Since the origin and dynamics of these disturbances are unclear, Christopher Snyder, in collaboration with Alain Joly (Centre National de Recherche Meteorologique, CNRM, Meteo-France), has performed experiments with a simple quasi-geostrophic channel model to understand whether and how perturbations may grow on a pre-existing baroclinic wave. They calculated, via adjoint techniques, the perturbations that grew most rapidly (in total energy) over various two-day intervals during the evolution of the parent baroclinic wave. These optimal perturbations grow somewhat faster than the parent wave and produce structures with short meridional scales, located at the edges of the upper-level fronts. The accompanying streamwise jets in the perturbation velocity, and the propagation of these features relative to the basic-state wave, are suggestive of atmospheric jet streaks.
Snyder and Daniel Keyser (State University of New York, Albany) have continued their study of how the planetary boundary layer modifies the structure and dynamics of surface fronts. Using a two-dimensional nonhydrostatic numerical model, they considered the "spin-down" of a frontal zone, initially in thermal-wind balance, that is subjected to parameterized surface drag and eddy mixing. To simplify the problem, neither surface heat flux nor large-scale forcing of frontogenesis are allowed. For typical mesoscale fronts, they find that convergence of the frictionally-driven flow overwhelms the frontolytic effects of mixing within the planetary boundary layer, resulting in robust frontogenesis. Thus, it seems unlikely that boundary-layer mixing can limit the scale of atmospheric fronts, except perhaps when the front reaches a scale comparable to typical boundary-layer eddies, which of course are not explicitly represented in these calculations. Snyder and Keyser also found that, as the frontogenesis proceeds (and the model grid spacing is decreased below a few hundred meters in the horizontal), the leading edge of the surface front acquires characteristics of a density current, with a balance between the cross-front acceleration and pressure gradient and a frontal head that breaks down into Kelvin-Helmholtz billows.
The accuracy of forecasts for the western United States is often limited by the lack of upstream data over the Pacific Ocean. Adaptive observing strategies are an intriguing potential remedy for this problem in which movable observing platforms (such as manned or unmanned aircraft) provide additional observations in those specific regions where, depending on the character of the flow at a given time, analysis errors are likely to be large and to grow rapidly. Together with Kerry Emanuel (Massachusetts Institute of Technology, MIT), Snyder is developing an effort to test adaptive strategies during the Fronts and Atlantic Storm Track Experiment (FASTEX) in January and February 1997. This effort is currently focused on developing a quasigeostrophic model as a testbed for adaptive strategies; tangent linear and adjoint versions of that model, and a simple data-assimilation scheme, are required.
Christopher Davis (joint appointment with RAP) has applied potential vorticity concepts to the problem of cyclogenesis in the lee of the Rocky Mountains. Two observed cases have been examined, and C. Davis has shown how the vertical tilt of a baroclinic wave is increased as the mountain anticyclone and lee trough form. The increased tilt results in a partial cancellation between upper and lower Potential Vorticity (PV) anomalies while allowing the appearance of a distinct circulation in the lee. This lee circulation is subsequently weakened as the mountain-induced perturbation, featuring strong northerly winds in the lee, advects cold air southward. The basic mountain effects observed in these cases are captured in an idealized, quasigeostrophic model. Baroclinicity, while reducing the overall mountain effect, allows the northerlies in the lee to extend further downstream. Overall results suggest that the effect of the Rocky Mountains is generally cyclolytic, with brief cyclonic spinup in the lee occurring at the expense of the large-scale baroclinic wave.
Mesoscale gravity waves have been the focus of a study concluded by Jordan Powers (ASP postdoctoral visitor). Investigating these waves using a mesoscale model, he found that the NCAR/Pennsylvania State University (PSU) Mesoscale Model (MM5) can produce them in simulations of different events, and that the mechanism forcing them is convection. His work has resolved a number of issues related to sensitivities of wave simulations, including those involving moist process schemes and horizontal-grid resolution. The investigations have also illuminated the spectral characteristics of observed and simulated events and the potential role of interference on wave scale.
James Bresch (visitor, University of Washington) conducted high-resolution real-data simulations of polar lows over the Bering Sea, Hudson Bay, and Mediterranean Sea using the MM5 model. Results show that both surface heat fluxes and latent heating of condensation contribute about equally in the development of the storms. In each case, an upper-level potential vorticity anomaly was associated with the development.
Stanley Trier, William Skamarock, and Margaret LeMone are investigating the dynamics governing the evolution of tropical squall lines, comparing adaptive simulations produced using the automated adaptive COllaborative Model for Multiscale Atmospheric Simulation (COMMAS), with observations of squall lines from the Tropical Ocean Global Atmosphere Coupled Ocean-Atmosphere Response Experiment (TOGA COARE). They have completed simulations of a shear-perpendicular squall line that compare well with the 22 February 1993 TOGA COARE squall line. Efforts to simulate shear-parallel lines have begun. They have found that, as expected, tropical convection is more difficult to simulate than midlatitude convection, generally due to weaker convective instability and weaker shear. Higher horizontal and vertical resolution, along with different convective initiation procedures, are found to be necessary for successful simulation of tropical squall lines in idealized model environments.
Eugene McCaul (NASA) and Morrris Weisman have completed investigations into a type of severe convective cell that may evolve in hurricane-landfall environments. Such environments are characterized by minimal amounts of convective available potential energy (CAPE), yet are often responsible for significant outbreaks of tornadic storms. One of the factors found responsible for the severity of these convective outbreaks is the very strong low-level vertical wind shear that is also observed in such environments. A series of nonhydrostatic simulations confirms that supercell structures can be generated within such environments, despite the lack of significant CAPE, as the storm updraft interacts with strong low-level vertical wind shear to produce updraft rotation and an associated dynamically-induced upward-directed vertical pressure gradient. This parallels the results for more classic supercells that are generated in high-CAPE, midlatitude environments, with the exception that the storm circulations in the hurricane environments are much shallower. The existence of supercell-type structures in such environments helps to explain the frequent occurrence of tornado outbreaks as hurricanes come ashore, as well as tornado outbreaks in other non-classic environments which display CAPE and vertical wind shear profiles similar to the hurricane cases. The existence of such shallow supercells in a variety of environments is just becoming well recognized by the forecast community, with forecasting techniques being modified to account for these kinds of storms.
McCaul and Weisman are currently expanding this work to more carefully document the role of the joint stratification of buoyancy and vertical wind shear in controlling supercell storm intensity. By varying the depth of the shear layer and height of maximum buoyancy over a range of values for a given net magnitude of CAPE and shear, they have been able to demonstrate that the strongest storms are realized when the CAPE and shear are both concentrated at low levels in the sounding. These results highlight the importance of considering the vertical stratification of these forecast parameters when trying to anticipate convective storm intensity.
John Tuttle and Robert Gall have concluded an in-depth analysis of radar data taken during the landfall of Hurricanes Andrew and Hugo. Both hurricanes passed over or very close to the WSR-57 radars in Miami, Florida and Charleston, South Carolina, respectively, and both radars survived at least until the bulk of the hurricane circulation was over land. Two powerful, relatively new analysis techniques were applied to the data: Tracking Reflectivity Echos by Correlation (TREC), and wavelet analysis. The TREC technique, which provides direct measurement of echo motion, was applied to search for small- scale structures nearby just inside the eyewall. None were found using this technique, but using simultaneous aircraft observations within the hurricane, it was shown that TREC provides reasonable estimates of windspeeds in the lower portion of the hurricane. The wavelet analysis, on the other hand, provided very strong evidence of the existence in these hurricanes (which at the time were very intense and symmetric), of regular spiral structures from just outside the eyewall to a radius of 100-150 km, that are spaced about 10 km apart and extend in a spiral fashion around the eye for several hundred kilometers. These spiral structures are of relatively low contrast in reflectivity (about 10 DBz), extend from the surface to about 5 km, are of smaller radial wavelength than the more familiar spiral rainbands (though they often coincide with rainbands), move with the local windspeed, and spiral outward against the wind (and so appear to move radially outward). A few NOAA P-3 passes through these bands in Hugo suggest that they are warm compared to their immediate surroundings. It is suggested that these bands have the structure of deep boundary layer rolls, but there are other possible explanations.
Weisman, Joseph Klemp, and Skamarock have completed preliminary investigations into the resolution dependence of explicitly modeled convective systems within a nonhydrostatic cloud model by considering the evolution of both two-dimensional and quasi-three-dimensional periodic squall lines for varying resolutions and under varying environmental conditions. The goal of this study was to determine the minimum grid resolution necessary to explicitly simulate convective systems as well as to clarify the physical factors that limit the ability to simulate such systems as the resolution is decreased. By varying the resolution between 1 and 20 km, they documented that resolutions as coarse as 4 km are sufficient to replicate most aspects of the system-scale structure for simulations of 6 h or greater, including the development of an upshear-tilted circulation and associated mesoscale temperature perturbations. Certain overall aspects of the evolution are also replicated for resolutions as coarse as 8 to 12 km, although the evolution is much slower. The physical factors most responsible for the deteriorating solutions for coarser resolutions include both the delayed strengthening of the convectively generated cold pool and the overprediction of the upward mass transport generated by the convective updrafts. This deterioration arises largely due to the inability to properly represent the critical non-hydrostatic processes within the simulations when resolutions are decreased to 6 km or greater. These results more clearly establish the limits of resolution beyond which convective parameterization schemes need to be implemented within mesoscale models, and help clarify which physical processes are most crucial to address in formulating parameterization schemes that are appropriate for mesoscale models with resolutions between 5 and 20 km.
Bolide impacts and large-scale volcanic eruptions have been proposed as possible causes of the massive extinction of life that has occurred episodically in earth's history. In spite of the catastrophic disruption of the local environment that accompanies bolide impacts and volcanic eruptions, it has been difficult to explain why these events sometimes lead to global extinction of species. Emanuel, Kevin Speer (Institut Francais Recherche pour Exploitation de la Mer, IFREMER), Rotunno, Ramesh Srivastava (University of Chicago), and Mario Molina (MIT) have proposed that in some cases the missing link may be provided by hypercanes, runaway hurricanes that are capable of injecting massive amounts of water and aerosols into the middle and upper stratosphere, where they may have profound effects on atmospheric chemistry and radiative transfer. Hypercanes are theorized to occur when the sea surface temperature exceeds a critical threshold, which may occur when sea water is locally heated by bolide impact, shallow-sea volcanism or, possibly, by overturning of superheated brine pools formed by underwater volcanic activity. Simulations using a convection-resolving nonhydrostatic, axisymmetric numerical model show that hypercanes can indeed develop when the sea surface temperature is high, and that they inject large amounts of mass into the stratosphere.
Over the past several years, much confusion has arisen in comparing the traditional views of supercell dynamics that are based on the role of vertical wind shear to relatively newer theories based on the use of storm-relative environmental helicity. This has become especially evident in the forecast community, often leading to incorrect applications of these concepts, particularly when interpreting the convective potential for straight versus curved hodograph environments. In an attempt to remedy this confusion, Weisman has been reexamining the basis for these theories to clarify how these concepts should be correctly applied toward forecasting convective storms.
Tropical Ocean Global Atmosphere Coupled Ocean-Atmosphere Response Experiment (TOGA COARE) work continues in three areas: (a) definition of the environment of COARE mesoscale systems sampled by the turboprop aircraft, (b) case studies of representatives of the major types of convective systems observed in COARE, and (c) numerical simulation of these systems. All three efforts are contributing to a better understanding of how mesoscale structure and surface fluxes relate to the environment.
The definition of the environment of COARE mesoscale systems is joint work that involves LeMone, Trier, Michael Dey, Edward Zipser and students (Texas A&M University), and David Jorgensen, Sharon Lewis, and Bradley Smull (NOAA, NSSL-Mesoscale Research and Application Division, MRAD). The nearly-completed set of soundings, constructed mainly from aircraft and radiosonde/ISS data, confirms the association of squall-like systems with low-level jet wind profiles, and slower-moving convective bands parallel to the vertical shear with considerable tropospheric shear but little or no shear in the boundary layer. The slower-moving bands seem to modify their immediate environment to draw in warm, moist boundary-layer air. CAPEs for these systems vary; typical values are between around 1200 and 1600 J/kg in the least modified environment.
Case studies of the squall-type system of 22 February 1993, a slower-moving band parallel to the shear on 9 February, and a complex system that seemed to evolve out of a squall line on 20 February, are nearing completion. Jorgensen has computed momentum-flux budgets for all three cases, plus the convective system of 6 February; these will be compared to the predictions of the Moncrieff archetypal momentum-flux model.
Trier, Skamarock, and LeMone are investigating the dynamics governing the evolution of these systems through comparisons of simulations using COMMAS with the observations. The model includes ice physics and surface fluxes. They have found, as expected, that tropical convection is more difficult to simulate than midlatitude convection, generally due to the weaker convective instability and shear. Higher horizontal and vertical resolution and different initiation procedures are needed for successful simulation in idealized model environments. Skamarock has just begun simulating a slow-moving convective band using the 9 February sounding.
For the 22 Feburary simulation, Trier set up the model so that only surface fluxes and stresses that differed from the undisturbed environment were included in the physics since the model would otherwise have produced an unrealistically warm and moist environmental boundary layer and hence artificially high environmental CAPEs. The squall-line-induced fluxes influenced the strength and, more dramatically, the areal extent of the subcloud cold pool, where fluxes are altered. However, the impact of squall-line-induced fluxes on the deep convection at the leading edge of the cold pool was significant only in the later stages of the simulation and only where the ground-relative winds were strong (strong fluxes) and the cold pool was weak and shallow. Additional tests revealed that the strength of the cold pool and updraft structure were sensitive to precipitation physics.
The model successfully reproduces the important observed features: the transition of the leading convective band from a linear leading edge of deep convection to a pronounced bow morphology accompanied by development of mesoscale vortices at the line ends, a double-peaked, multicellular convective updraft structure with a shallow but intense leading-edge updraft in the lowest few kilometers and a midtropospheric maximum about 20 km to the rear, and the development of elongated precipitation bands rearward and transverse to the leading-edge convective band. Analysis of these model-produced features indicates that their development was crucially influenced by the mesoscale wind and pressure fields, which themselves were clearly impacted by early interaction of the system with the environmental wind shear.
The strong definition of the vortices in the numerical simulations led LeMone to reexamine the in situ low-level wind field. The northern vortex and associated low-pressure center were well-defined at 940 m, and the counter-clockwise circulation was still apparent 150 m above the surface, which appears to be within the mixed layer. Other features in the pressure field that resembled the model were a broad mesolow at low levels in the vicinity of the trailing transverse band, and a narrow low-level mesolow further north.
Richard Carbone, Mitchell Moncrieff, James Wilson (ATD), Steven Rutledge (Colorado State University, CSU), Thomas Keenan (Bureau of Meteorology Research Center, BMRC, Australia), Gregory Holland (BMRC), and Joanne Simpson (NASA) completed the design of the Maritime Continent Thunderstorm Experiment (MCTEX) to investigate various aspects of convection over the Tiwi Islands north of the Australian continent. Carbone and Wilson will study the initiation of convection in an island sea-breeze regime, exploring mechanisms such as colliding sea breezes and the effects of small topography. Carbone, Keenan, and Simpson will examine the further evolution of convection from zonally-oriented island initiation, through convective-scale cloud mergers, to the development of meridionally-oriented mesoscale convective systems. This evolution takes place in the presence of the tropical easterly jet at 700 mb and the attendant easterly shear vector below. Moncrieff will use the MCTEX dataset and analyses thereof as a basis for validating upscale effects in mesoscale convective system (MCS) simulations as part of the Global Energy and Water-Cycle Experiment (GEWEX) Cloud System Study (GCSS)/Role of Clouds, Energy, and Water (ROCEW) program.
Changhai Liu (visitor, Texas A & M University) and Moncrieff have begun numerical studies of the sea-breeze produced tropical thunderstorms in MCTEX. Preliminary results demonstrated that the sea breeze alone is able to trigger vigorous convection under certain circumstances. In particular, the classical squall-type organized convection with an extensive trailing cloud deck can be reproduced using the wind profile similar to monsoonal and break-season flow documented in Keenan and Carbone (1992). They continue to investigate the detailed initiation processes and structures of maritime continental thunderstorms as well as the interaction between the convective clouds and radiation.
Carbone and Roger Wakimoto (University of California, Los Angeles, UCLA) obtained data for a dual case study on supercell tornadogenesis in the Verification of the Origins of Rotation in Tornadoes Experiment (VORTEX). Carbone is concentrating on the 31 May Sweetwater, Texas case dataset, which documents the genesis and dissipation of two major tornadoes estimated to be of F-3 and F-4 intensity. During Wakimoto's visit to MMM, preliminary examination of ELectra DOppler RAdar (ELDORA) data confirmed excellent kinematic representations of this storm to form the basis for comparison with a Kansas case being examined by Wakimoto. Among the common features observed is the sudden collapse of the mesocyclone through nearly the entire depth of the troposphere to form a deep tornadic-scale vortex. This is distinct from non-supercell tornadoes, which often have their origin in planetary boundary layer (PBL) circulations along shear lines when coupled to convective updrafts.
Weisman continued to serve as a principal investigator for VORTEX, which completed its second season during the spring, 1995, in Oklahoma, southern Kansas, and northern Texas. The main objective of VORTEX is to gain a better understanding of the process of tornadogenesis by using mobile observing systems to observe the evolution of the sub-cloud region within and around evolving tornadic storms with much higher resolution than previously available. Weisman's specific contribution to the experiment was to take serial soundings in the inflow sector of a given storm using an ATD mobile Cross-Chain Loran Atmospheric Sounding System (CLASS) to better establish the association between the environmental conditions and the observed storm structures that lead to tornadogenesis, as well as to establish the degree to which a storm can alter its near environment. Many good case studies were obtained during this second field season, with analyses of this data just beginning.
Klemp, Rotunno, and Skamarock have continued work on an Office of Naval Research (ONR) project in which the goal is to understand the origin and nature of the sudden transition in the local weather from clear skies to dense stratus clouds that often occurs during the spring and summer along the west coast of the U.S. Their working hypothesis is that the basic fluid-dynamics problem is one of the release of a reservoir of relatively dense air (the marine layer), bounded by an insurmountable wall on one side (the coastal mountains of the U.S. west coast), under the influence of the earth's rotation. Furthermore, the above-described sudden transitions are due to a change in the equilibrium of the marine layer to the south that requires an adjustment under gravity to occur that affects the marine layer to the north. Treating the marine layer and the free atmosphere as two homogeneous layers, they used the reduced-gravity shallow-water equations to study the evolution of the marine layer given a synoptic-scale evolution of the pressure in the upper layer. They found that, for the typical time and space scales of the synoptic-scale forcing, a Kelvin wave is produced in the marine layer. Since the Kelvin wave is basically a hydrostatic long wave, the elevated part of the wave eventually steepens to a discontinuity, implying an internal bore in the real fluid. Since the existing literature on internal bores has several disturbing aspects (e.g., theory predicts infinite propagation speed when the thickness of the layer ahead of the bore goes to zero), they completed a general study on the propagation of internal bores, and have proposed a resolution.
To investigate how wave motions and gravity currents behave as they try to propagate along a curved coastline, Skamarock, Klemp, and Rotunno constructed a curvilinear-coordinate numerical model for solving the nonlinear shallow-water equations, which allows examination of wave propagation around irregularities in the bounding wall. They have found that trapped propagating disturbances, such as Kelvin waves, bores and gravity currents will, in general, propagate robustly around coastal bends. The disturbances are not greatly affected by linear and nonlinear diffraction effects which had previously been suspected of producing wave cessation at convex bends of the bounding wall.
Joseph Prusa (Iowa State University), Piotr Smolarkiewicz, and Rolando Garcia (ACD) continued their collaborative study of gravity-wave activity in the upper mesosphere and lower thermosphere. Based on linear analysis and two-dimensional numerical simulations, they have found that the frequency and wavenumbers of breaking waves at mesopause altitudes may bear little resemblance to forcing sources in the lower atmosphere. As a direct result, it appears not to be possible to compute physically realistic simulations of wavebreaking using domains of relatively small vertical extent. The wavebreaking region initially has horizontal and vertical extents of the order of the vertical and horizontal wavelengths of the pre-breaking wave. This region rapidly grows windward as longer, slower waves reach breaking altitude. Downwind growth occurs as higher-frequency, almost evanescent waves reach breaking altitude. The resulting region of turbulence increases in altitude until the turbulent dissipation, which is approximately constant throughout the breaking region, is matched by molecular dissipation. Downward growth occurs until the longer, slower waves that are present there have insufficient spectral power (set by the nature of the forcing source) to break. A WKB analysis has been developed that predicts the horizontal wavelength and time of breaking waves given a specified Gaussian forcing. Wave activity arguments have been used to predict the value of turbulent dissipation, as well as the height of the turbopause. These results agree well with the numerical computations, as well as with available observations (noctilucent clouds, airglow, meteor trails, radar, and rocket soundings). An interesting prediction is that the turbopause should be approximately 20 km higher in winter than during summer. This work culminated in a manuscript submitted to the Journal of the Atmospheric Sciences in August 1995.
Roelof Bruintjes (joint appointment with RAP) and Smolarkiewicz studied the effects of strongly stratified flows past valleys. A series of numerical experiments simulating such flows past idealized valleys has been analyzed against relevant laboratory flows and observations collected in the region of the Verde Valley (Arizona). This analysis revealed an interesting transient effect that may be important for understanding and predicting a class of natural valley flows. For a potential flow initial condition (relevant to certain natural scenarios), both computational and laboratory flows remain attached to the valley over many advective time scales. The morphology of these flows is consistent with that of strongly stratified flows past isolated hills. After this time, there is a sudden transition from a "mountain-flow" regime to a "cavity-flow" regime where the valley flow separates from that aloft. This critical transition has a dramatic impact not only on the local valley flow, but also on the gravity-wave field aloft. The original explanation is offered in terms of the columnar-mode arguments and upwind blocking of the ascending lee slope (of the valley). Such arguments provide quantitative estimates of the transition occurrence and time, as a function of environmental conditions and geometry of the valley. These theoretical estimates are in good agreement with numerical predictions, and they agree well with relevant laboratory results.
The ongoing collaboration between Smolarkiewicz and Balu Nadiga and Len Margolin (Los Alamos National Laboratory) investigates the applicability of nonhydrostatic but vertically-averaged fluid-dynamic equations to geophysical flows. During the past year, they have synthesized their analysis of the solutions of the hydrostatic shallow-water equations, two different sets of nonhydrostatic (the generalized-Boussinesq and the Green-Naghdi) equations, and the full Euler equations, in the context of the flow of a homogeneous fluid over a two-dimensional ridge. The parameters consisted of the flow speed and the height and width of the ridge. This analysis revealed the inadequacy of the hydrostatic approximation either when the flow speeds are transcritical or when the width of the ridge is not small compared to the depth of the fluid layer, and showed the importance of the nonhydrostatic terms. More importantly, they have also verified that the (fully nonlinear) Green-Naghdi system is able to predict vertically-averaged features of the flow better than the (weakly nonlinear) generalized-Boussinesq system in almost all regions of the parameter space.
In collaboration with the Hong Kong University Institute of Science and Technology (HKUST) and the University of Wyoming, over the last year Terry Clark, Hsaio-Ming Hsu (visitor, Woods Hole, MA, joint appointment with RAP), Teddie Keller (joint appointment with RAP), and Janice Coen continued their studies of flows over Lantau Island as part of understanding terrain-induced turbulence at the new Hong Kong airport. They continued their work on idealized critical-level flows and on two case studies from the Lantau Experiment (LANTEX). Comparisons were made between the model simulations and observations from aircraft, automatic weather stations, and lidar.
The range of cases was extended for the idealized critical-level flows with variations in low-level flow speed and direction, static stability, and critical-level height. The solutions showed no evidence of strong resonances as previous studies found for more hydrostatic flows. The typical flow over Lantau was so near critical nonhydrostaticity (i.e., N/kU ~ 1) that the vertical scale of the waves/eddies was determined almost solely by the height of the critical level. For this parameter space, linear dispersion requires transient solutions. As a result, the typical critical-level flow results in nonlinear propagating eddies even in the absence of significant resonant amplification. This transience should be an important consideration for operations at the new airport.
Moderate turbulence was observed on 20 May 1994, in a flow regime characterized by an ambient critical level. Comparisons between the lidar scans and model-simulated scans were surprisingly good. The aircraft comparisons were also good but more difficult to interpret. The simulation compared well with the three mountaintop automatic weather station (AWS) sensors, but poorly at the low-elevation stations. This discrepancy needs further investigation.
The most severe turbulence event observed occurred on 7 June 1994, associated with the passage of tropical storm Russ, in which the flow was characterized by a deep region of uniform flow. The turbulence was dominated by mechanical generation rather than by gravity-wave dynamics as in the critical-level cases. Horizontal resolutions as high as 62.5 m were used to capture the surface-layer physics and the resulting flow separations. These results have yet to be fully analyzed but are extremely encouraging. A series of simulations using idealized deep uniform flow profiles shows that the maximum variance of the wind increases substantially for ambient wind speeds greater than 12 m/s.
In collaboration with Robert Banta (NOAA, ETL), using the large-scale initialization software, Clark and William Hall continued their study of the 1 December 1992 incident over Evergreen, Colorado, where a DC-8 lost an engine in clear-air turbulence. Simulations of both the onset and demise of gravity-wave breaking as well as the details of the turbulence in the wave-breaking region have been completed using low resolution. Higher-resolution simulations will be performed in the future to further our understanding of the severe turbulent gusts affecting both in-flight and near-airport aircraft operations.
Carbone, William Cooper (joint appointment with ATD), and Wen-Chau Lee (ATD) completed a study on the origin of flow reversal along the windward slopes of Hawaii. Through analyses of Hawaiian Rainband Project (HaRP) data and comparisons to earlier Clark model simulations by Smolarkiewicz and collaborators, they determined that evaporation from diurnally-forced orographic rainfall provides the principal initiation mechanism for flow reversal over the island. An evaporatively-driven cold pool descends from the mountainside, flows offshore, and is strengthened at night by radiative cooling. In a low-Froude-number regime, island blocking contributes to maintenance of the reversed flow further off-shore.
Gary Austin and Robert Rauber (University of Illinois), Harry Ochs (Illinois State Water Survey), and L. Jay Miller completed a study of the sources of rainbands off the windward coast of the big island of Hawaii, using available radar and satellite data from HaRP conducted in the summer of 1990. This study revealed that radar-observed rainbands and cells affecting the windward coastline were preferentially associated with cloud clusters and cloud lines that originated in the large, shallow stratocumulus cloud mass which covered much of the ocean between Hawaii and California. When the cloud lines or cells approached the island, they were stretched along the flow separation line as the trade winds deflected around the island. Though cloud lines were frequently observed to form along the flow separation line, analyses of upstream aircraft thermodynamic soundings suggested that lifting predicted from dry simulations was typically not enough to trigger free convection and rainbands in the absence of a density discontinuity.
A well-documented diurnal variation in rainfall occurs on the windward coast of Hawaii, with rainfall occurring most frequently during the night and very early morning hours. Radar and satellite data from HaRP showed a strong diurnal oscillation in rainband frequency upwind of the island, which appeared to be associated with diurnal variations in cloud cluster structure and coverage over the open ocean. This diurnal behavior in cloud clusters may be related to the diurnal variation in radiational forcing that occurs near the top of the tradewind layer. Tropical cyclones originating in the Intertropical Convergence Zone sometimes moved north through the stratocumulus cloud mass, producing, in their wake, cloud-free regions that advected toward the Hawaiian Islands. Passage of these cloud-free regions every few days apparently disrupted the normal diurnal patterns of clouds, rainbands, and rainfall near and on the Island.
Several NCAR investigators are involved in a joint modeling and observational study to determine the effects of moist processes on the airflow around the island of Hawaii and the factors affecting the morphology of Hawaiian rainbands.
Vanda Grubisic (ASP), Smolarkiewicz, Carbone, Roy Rasmussen (joint appointment with RAP), and William Anderson will focus on numerical simulations. While building upon the experience gained in earlier studies of rainbands, this effort examines the rainband cases observed during HaRP. Flow past Hawaii is an example of a low-Froude-number flow past a three-dimensional obstacle, whose salient features include upstream flow separation and reversal. In the earlier modeling studies which did not fully represent moist processes, the origin of the arc-shaped rainbands upstream of the island was linked to uplift at the convergence zone associated with the flow-separation line. However, more recent observational studies from HaRP indicate that the predicted lifting typically is not enough to initiate a convective band. Thus, there is a need to quantify the sensible and latent thermal effects (such as surface boundary-layer forcing, precipitation evaporation, radiative cooling, etc.) which control the morphology of the local flow and alter simple low-Froude-number predictions for the strength and location of the rainbands. The high-resolution simulations will be performed using a massively parallelized version of the semi-Lagranagian/Eulerian nonhydrostatic model for fluids.
In parallel, Carbone, Tuttle, Cooper, and Lee have initiated an observational study on the morphology of Hawaiian rainbands and the sensitivity of their dynamical organization to subtle variations in the properties of the undisturbed tradewind layer. Potential controlling factors being examined include island Froude number, conditional instability, horizontal vorticity, Richardson number, and susceptibility to internal gravity-wave propagation and Kelvin-Helmholtz instabilities. Consistent with Austin et al., preliminary results indicate that most rainbands initiate upstream of any flow separation line while under the influence of island topography, that amplification of rainbands commonly occurs along the convergence line where off-shore flow meets the incoming trades, and that rapid dissipation occurs when rainbands propagate toward land after leaving the flow-separation line. Rain in Hilo is commonly associated with "landbands," rainbands that form over land in the diurnal transition between afternoon orographic rains and morning offshore flow. Multiple rainbands are formed by a discrete propagation process, the mechanisms of which are presently under study. The intensity of rainband activity is correlated with the depth of the tradewind layer and conditional instability.
There is a phenomenon in the Gulf of California associated with the Mexican Monsoon known as the "Gulf Surge." It has been described by forecasters in the southwestern part of the U.S. and is assumed to be responsible for bringing substantial amounts of water vapor into the Arizona region during the summer months (supplying moisture to summer thunderstorms there) and also for triggering the thunderstorms themselves. The surge is described as having a sharp leading edge where the flow changes from light northerly or westerly to strong southeasterly as the surge passes. After the passage of the leading edge, a strong low-level southeasterly flow can persist for a day to several days. During the Southwest Area Monsoon Project (SWAMP) held in 1990, the NOAA P-3 obtained what is perhaps the best cross-section through a surge event. Using MM4, David Stensrud (NOAA, NSSL) and Gall have produced a simulation of this event that reproduces the P-3 cross-section in all its major features. By examining the history of this event in the simulation, as well as others, during the one month of SWAMP, they noted that the surge in all cases was triggered by a thunderstorm somewhere over the Gulf, occurred when an easterly wave forced a reversal of the pressure gradient along the gulf (to lower pressure at the northern end), and was strongest when the easterly wave had a proper phase with respect to westerly disturbances passing the northern end of the gulf. In its later stages it looks much like a density current, but shortly after its initiation it appears to be a Kelvin wave initiated by the thunderstorm outflow in the changing pressure field.
In support of the Department of Energy (DOE) Atmospheric Radiation Measurement (ARM) Program, Wojciech Grabowski and Moncrieff, in collaboration with James Hack and Jeffrey Kiehl (both of CGD) and Gregory Tropoli (University of Wisconsin, Madison), have completed their study of the iteraction among cloud dynamics, cloud microphysics, radiation and large-scale forcing reported last year. This 24-day, two-dimensional study, which had a horizontal domain of 900 km and used Tripoli's model, has been accepted for publication in the Quarterly Journal of the Royal Meteorological Society. A major finding was the warming effect of cirrus anvils in the surface and tropospheric energy budgets broadly in accordance with the "thermostat" hypothesis of Ramanathan and Collins (1991). These anvils were a part of the squall-line-like deep convection regime organized by the environmental shear. However, this warm and moist "climate state" strongly contrasted with the cool and dry climate of the Sui et al. (1994) simulation. This conundrum is being studied in terms of model design as well as the parameterization of physical processes in cloud-resolving models (e.g., radiative-microphysical-dynamical interactions).
Jun-Ichi Yano (visitor, Climate Research Centre, Monash University, Melbourne, Australia), Moncrieff, James McWilliams (UCLA), and Emanuel have completed their idealized study of the large-scale organization of tropical cloud systems reported last year. The results, published in the Journal of the Atmospheric Sciences, highlighted the effect of different types of convective parameterization paradigms on the large-scale organization or enhancement of convective activity in the tropics on time scales of 30-60 days.
Yano and Moncrieff have started an investigation of the role of momentum transport by organized convection in the tropical atmosphere at large scales. As a paradigm, they are using a parmeterized version of the archetypal dynamical model of Moncrieff (1992). The framework is a linear version of the shallow-water equation formulation (reported above) that has simple boundary/surface layers, a one-layer troposphere and an upper free surface at which radiation conditions are imposed. A two-layer troposphere has also been added, and the impacts of convective momentum transport on Wind Induced Surface Heat Exchange (WISHE) on barotropic and baroclinic structures, and on linear instability modes of the tropical atmosphere are being examined.
Liu and Moncrieff continued their numerical study on the effect of an ambient flow on the behavior of density currents. Of special interest was the impact of the background wind and shear on the translation and morphology of the currents. The model was initialized with a horizontally homogeneous wind profile superimposed on a cold-air reservoir in a neutrally stratified environment with a free-slip lower boundary condition. A head wind (i.e., the wind direction opposing the system movement) raises the density current head compared to calm surroundings, while a tail wind has the opposite effect. This finding is in contrast to the laboratory results of Simpson and Britter (1980). Sensitivity tests have suggested that the no-slip lower boundary condition in the laboratory experiments is responsible for the discrepancy. Ambient shear exerts a more complicated influence; a weak or moderate positive shear (positive stands for the direction of density-current movement), in general, elevates the head for the downshear-traveling system. A multi-head structure can be generated in strong shear. In comparison, the structure in the upshear-traveling system is largely insensitive to the shear. The propagation speed is linearly proportional to the ambient windspeed, and roughly reduced or enhanced by three quarters depending on the airflow direction. For an idealized uniform shear, a linear relationship provides a good approximation between the advance rate of density currents and variable shear, particularly for the upshear-moving system.
Over the last year, Clark and colleagues used a coupled mesoscale and fire model to study the dynamics of forest fires. To date, studies have identified the mechanism that produces some of the basic fire line shapes observed in the field. The experiments have shown how the "parabolic" shape develops as a result of the convergence zone from the hot-air column feeding back on the winds at the fire line. For longer fire lines, it was shown how this mechanism results in multiple protrusions or in "convective fingers." A paper was submitted to the Journal of Applied Meteorology on the above topic. This type of information is probably important to the foresters but more of a novelty compared to the basic understanding that is being gained through modeling associated with the dynamic stability of fires.
Studies also focused on mechanisms that can destabilize a fire and lead to blowups, for example, "dynamic fingering." Given the appropriate conditions, vortex tilting can convert the sheared low-level wind ventilating the fire into vertically oriented rotors that produce fingers of rapidly moving air at the fire front, causing it to leap forward. This mechanism, and perhaps other mechanisms as yet unrecognized, are thought to be responsible for transforming a fire into an uncontrollable and extremely hazardous blowup fire; some preliminary simulations have shown evidence of this effect. Once this type of mechanism is understood, an obvious application of future modeling is to show how these instabilities might be thwarted or controlled in the field.
The results of this modeling were given in an invited paper at a fire meteorology conference in Adelaide, Australia, in June 1995.
Margolin and Jon Reisner (Los Alamos National Laboratory), and Smolarkiewicz have adopted the volume of fluid (VOF) method, a specialized grid-refinement technique, to the numerical simulations of clouds. In particular, they have shown that VOF eliminates most of the well-recognized numerical difficulties (spurious oscillations and/or diffusion in the vicinity of a cloud/environment interface) associated with finite-difference Eulerian advection of cloud boundaries. In essence, VOF is a subgrid-scale advection parameterization that accounts for transport of material interfaces. In principle, VOF is an Eulerian approach, as it does not track explicitly material interfaces. Instead, it reconstructs such interfaces using auxiliary dependent variables, the partial volume fractions of immiscible materials within computational cells. A feature of VOF that is particularly important for cloud modeling is its ability to identify cells with a subgrid-scale cloud/environment interface. There, relevant parameterizations of microphysical processes can be applied consistently in "clear" and "cloudy" regions. In their study, they have documented that simulations employing VOF are not only more accurate, but are more efficient as well, requiring almost one order of magnitude lower spatial resolution for an equivalent accuracy. Their calculations demonstrate the importance of minimizing numerical diffusion at the cloud/environment interface to accurately capture small-scale flow features evolving in the vicinity of the cloud boundary.
Margolin, Reisner, and Smolarkiewicz have developed a high-performance (massively parallel) numerical model for simulating shallow-water flows on a rotating sphere. The unique feature of this model is a reduced-grid capability built around the Multidimensional Positive Definite Advection Algorithm (MPDATA). This capability increases the allowable time step based on stability requirements, and leads to significant improvements in computational efficiency. On the CM-5 architecture, at low resolution (2.8 deg), the use of the reduced grid is as accurate as the nonreduced grid with about the same computational efficiency. The principal computational advantage of the reduced grid is realized at higher resolutions, with the reduced grid being up to 35 times faster than the nonreduced grid at a resolution of 0.35 deg. Due to faster communications on machines like the Cray J90 or Cray T3D, it is anticipated that the reduced grid will lead to greater efficiency than that on the CM-5. It appears that the reduced grid framework is a useful modification for general circulation models of atmosphere and ocean based on finite difference approximations.
Grabowski and Smolarkiewicz introduced significant enhancements to the treatment of the moist thermodynamics in the semi-Lagrangian/Eulerian cloud model. They have developed a novel two-time-level semi-Lagrangian finite-difference solver, custom-designed for the bulk parameterization of moist thermodynamics common in anelastic cloud models. Standard semi-Lagrangian approximations appear inefficient when the precipitation enters the problem. Precipitation motions relative to air introduce some technical difficulties which may be circumvented with adequate manipulations of the governing equations, and integrating the resulting precipitation evolution equation along a modified trajectory. A heavily simplified physics of precipitation formation in the bulk parameterization justifies some special simplifying assumptions that compromise formal accuracy and computational efficiency of the method. They have demonstrated that the semi-Lagrangian approach offers highly competitive and versatile methods for modeling cloud-scale processes. Representation of the moist thermodynamic processes has been further enhanced by introducing a bulk ice parameterization and improved formulation of the saturated water-vapor mixing ratio for very cold temperatures. This allows the model to accurately simulate water-phase changes in the upper troposphere and in the stratosphere.
Anderson and Smolarkiewicz continued their work on high-performance computing strategies for atmospheric models. Last year, they ported the semi-Lagrangian/Eulerian nonhydrostatic atmospheric fluid model to the Cray T3D parallel processor using a High Performance Fortran (HPF) approach. In general, it is not necessarily obvious which of the exisiting strategies for computational fluid dynamics (CFD) model parallelization is the best for maximum performance, ease of implementation, and portability. Often, this depends on the fluid-model algorithm being implemented. The goal of this work is to assess the virtues and weaknesses of the HPF and message-passing approach for two distinct Navier-Stokes solvers. The Eulerian solver requires traditional Courant Friedrichs Lewy (CFL) stability conditions, thereby limiting local communications to the nearest neighboring points on the mesh, whereas the semi-Lagrangian solver admits Courant numbers well exceeding unity, thereby resulting in communications patterns extending over a number of grid points. So far, the model has been implemented on a Cray T3D using the HPF method. Results show that the semi-Lagrangian and Eulerian algorithms give about equally good performance. Although the message-passing version is still under development, it is anticipated that its performance will be, in general, at least twice as good as that of the HPF version, except for cases where the semi-Lagrangian algorithm takes advantages from large time steps allowed by the numerical stability of the scheme.
Cristoph Schär (Atmospheric Physics ETH, Zurich) and Smolarkiewicz have developed a conservative and synchronous flux-correction technique (FCT) which aims at a consistent transport of the coupled, density-like, dependent variables. Their technique differs from traditional FCT algorithms in two respects. First, the limiting of transportive fluxes of the primary variables does not derive from smooth estimates of the variables, but it derives from analytic constraints (implied by the Lagrangian form of the governing continuity equations) imposed on the specific counterparts of the variables. Second, the traditional FCT limiting based on sufficiency (for monotonicity) conditions, is augmented by an iterative procedure which approaches the necessity requirements. Although the approach derived is applicable to the transport of arbitrary conserved quantities, it is particularly useful for the synchronous transport of mass and momenta in elastic flows, where it assures intrinsic stability of the algorithm regardless of the magnitude of a mass-density variable. This latter property becomes especially important in fluid models with a material "vertical" coordinate (e.g., shallow waters, and isopycnic/isentropic hydrostatic models for continuously stratified fluids) where material surfaces can collapse to zero-mass layers admitting, therefore, arbitrarily large local Courant numbers.
Collaboration has been initiated between Smolarkiewicz and Randall LeVeque (visitor, SCD, University of Washington). LeVeque has been developing numerical methods for multi-dimensional hyperbolic systems of conservation laws and general software based on these algorithms (CLAWPACK). Some of this software has been incorporated into Smolarkiewicz's nonoscillatory forward-in-time (NFT) model of the anelastic fluid equations on curvilinear grids, in order to perform comparisons with the MPDATA advection algorithm developed by Smolarkiewicz. The modifications needed to perform this synthesis led to a general reformulation of LeVeque's algorithms that proves to be useful in other applications as well, and has been incorporated into CLAWPACK.
Smolarkiewicz and Igor Szczyrba (University of Northern Colorado) have developed a two-time-level finite-difference solver for Kelvin-Voigt equations. These equations describe a flow of visco-elastic fluids relevant to biomechanical modeling. Numerical solutions were obtained within an ellipse set impulsively into a spinning motion (an idealized model of a human brain response in a car-accident scenario). These results extend Ljung's classical analytic solutions available for a circular boundary. For the special case of the circular boundary, the numerical results are in excellent agreement with the analytic predictions. Further studies will address three-dimensional geometry as well as generalizations of the governing equations to those admitting propagation of shock waves. The latter may be important for understanding mechanical brain damage observed in apparent absence of external head injuries.
Skamarock, Klemp, and Smolarkiewicz are continuing to examine alternative methods for integrating various forms of the nonhydrostatic equations. Semi-implicit integration methods, incorporating both semi-Lagrangian and Eulerian approaches, are being examined for accuracy, robustness, and efficiency for cloud- and mesoscale applications. In particular, efficient Helmholtz and Poisson equation solvers are being developed for the nonconstant coefficient problem that includes cross-derivative terms. These solvers are based on the conjugate residual (CR) algorithms developed by Smolarkiewicz, and preconditioners have been identified that accelerate the algorithms such that the full models are as efficient as existing time-split models or models based on direct solvers for the constant coefficient problem. Klemp and Skamarock have also explored semi-implicit solutions to the pseudo-compressible equations and have developed an efficient technique for converging iterations using divergence damping. These studies are part of efforts to establish the best numerical framework for the next-generation version of the MM5 modeling system.