II. MMM Core Program
1. Prediction of Precipitating Weather Systems (PPWS) Program
One of the two primary scientific programs in the division is the Prediction of Precipitating Weather Systems (PPWS) program. Its goal is to advance the understanding and prediction of significant precipitation events in order to substantially reduce forecast errors toward the limits of predictability. The accurate prediction of precipitating weather systems is an important topic in the U. S. Weather Research Program (USWRP). There is good potential for advancements in this area because of emerging operational observing systems, high-resolution non-hydrostatic forecast and data assimilation systems, and the continued rapid growth in computer power. Within MMM, there is broad interest and expertise in the observation, analysis, and prediction of precipitation systems, and opportunities to leverage division resources through collaboration with other NCAR divisions, government laboratories, and the university community. The research within MMM focuses on specific areas where MMM's expertise is best suited to advance the science. These areas include mesoscale predictability, the life cycle of mesoscale precipitating weather systems, mesoscale dynamics, mesoscale data assimilation, and high-resolution numerical weather prediction (NWP). These topics are highly interrelated and research is coordinated to contribute to the advancement of mesoscale assimilation and forecast systems.
In advancing these areas of research, the PPWS plans to conduct periodic workshops. One such workshop was held in early September 1999 at NCAR with a focus on ensemble forecasting in the short- to medium-range (0 to ~14 days). The three-and one-half-day workshop covered the topics of initial conditions and coupling of ensemble forecasting and data assimilation systems; dealing with model error; and issues related to the practical use of ensemble forecasts. The workshop announcement and preliminary agenda can be found at http://www.asp.ucar.edu/~hamill/pred_workshop.html.
a. Mesoscale Predictability and NWP
Weather forecasts, particularly those of precipitation, always contain errors. While such errors can arise in a variety of ways (such as from imperfections in the forecast model, its boundary conditions or initial conditions), the fundamental source of forecast error is the strong tendency for two initially similar states of the atmosphere to diverge with time. The time required for this divergence to become pronounced, and the mechanisms that lead to it, are poorly understood at the mesoscale.
Perturbations Techniques for Ensemble Forecasts
Thomas Hamill (ASP), Chris Snyder, and Rebecca Morss (Massachusetts Institute of Technology) completed an evaluation of three techniques for generating perturbations for ensemble forecasts. They are using a quasi-geostrophic model under perfect-model assumptions, whereby the model is used to generate a simulation taken as the truth. They are testing the breeding technique, used at the National Centers for Environmental Prediction (NCEP), the singular vector technique, used at the European Centre for Medium-Range Weather Forecasts (ECMWF), and a "perturbed observation" strategy, pioneered by Peter Houtekamer (Canadian Meteorological Center). They find that the perturbed observation strategy is superior, in that the analyses and forecasts are calibrated, indicating that probabilistic forecasts can readily be made from the ensemble forecasts. NCEP and ECMWF are very interested in this research, and the possibility of future collaboration with an operational center to test and possibly implement these ideas is being explored.
Statistics of Forecast and Analysis Errors
Hamill, Snyder, and Morss also used the quasi-geostrophic model, along with a three-dimensional variational assimilation (3D-Var) scheme, to explore the statistics of forecast and analysis errors. They found that both forecast and analysis errors reflect the influence of the dynamics: The errors have significant projection on the subspace of leading Lyapunov vectors, their time-mean vertical distribution in both energy and potential enstrophy is similar to that in the "true" state, and error variance in potential vorticity is typically confined to regions in which the true state has large gradients of potential vorticity. A consequence of this dynamical influence is that the spectrum of the covariance matrix for both forecast and analysis errors is steep and small samples (or ensembles, of a few 10's of members) can provide much information about the errors. Indeed, Hamill and Snyder showed (see Data Assimilation Research, section 1.D.) that an ensemble Kalman filter using a small sample can outperform 3D-Var.
Predictability of Moist Convection in Baroclinic Systems
Richard Rotunno, Snyder, and Wei Wang began work on numerical simulations of a baroclinic wave that will explicitly resolve embedded moist convection (through nested grids). The goal of these simulations is to understand the mechanisms that organize and control precipitation at the mesoscale and to assess its predictability through a simulation with a minimum of parameterizations. They are presently examining intensive observation period (IOP) 17 from the Fronts and Atlantic Storm-Track Experiment (FASTEX) as a suitable test case.
Real-time Experimental Numerical Weather Prediction
With the acquisition of powerful multiprocessor Compaq workstations during FY 99, it became possible for James Bresch to make significant upgrades to MMM's real-time forecast system. Beginning on 1 January, twice-daily, nationwide 48-hour MM5 forecasts using 30-km resolution were implemented. These forecasts continue to be the highest-resolution, nationwide forecasts available and permit researchers to examine model behavior over a wide range of terrain and climates. A 10-km nested domain is also run once daily, usually over the central United States, but sometimes in support of field programs (two wintertime field programs sponsored by RAP, used the MM5 forecasts for their operations planning). The real-time MM5 runs correctly pointed out the tornado threat over central Oklahoma on 3 May 1999 [See Figure], but, like the other operational models, under-predicted the strength of the upper-level winds associated with the approaching trough. Finding and correcting the reasons for this under-prediction is an area for further study.
For the 1999 hurricane season, a special 30-km MM5 domain was placed over the western Atlantic to examine MM5's skill in predicting hurricane tracks and intensity, as well as for testing various tropical cyclone bogussing schemes. The MM5 forecasts demonstrated considerable skill, especially with Hurricane Floyd, for which MM5 correctly predicted landfall in North Carolina 48 hours in advance. A completely new World Wide Web interface was developed which allows users simple access to many forecast models and side-by-side model comparisons. Numerous graphical products, animations, and even computer-generated worded forecasts are available (see http://rain.mmm.ucar.edu/mm5).
In work to design a mesoscale model-based forecasting tool for Taiwan's Civil Aeronautics Administration, Jordan Powers, Bresch, and Kevin Manning developed and implemented an operational MM5 system within the numerical weather prediction (NWP) environment of the Civil Weather Bureau (CWB) of Taiwan. The system is relocatable and can run on a variety of architectures (e.g., DEC, SGI, Fujitsu). It can ingest a broad range of input data, including fields from various large-scale models (e.g., CWB Global Model, NCEP Eta model, NCEP AVN model) and diverse observation types (soundings, ACARS, satellite-derived winds and temperatures, typhoon reports, etc.). To support this system, the MM5 graphics package RIP was expanded, ported to Linux, and better documented. The forecasting system features a WWW display allowing for easy selection of products from multiple forecast domains, vertical levels, and initialization times; it also offers a variety of window and animation viewing options. The work has yielded a powerful, user-friendly interface for the dissemination and analysis of real-time MM5 forecasts.
Process Studies and Model Validation
Evaluation of the land-surface model was continued in FY 99 by Jimy Dudhia and Fei Chen (RAP). Four 48-hour cases from the First ISLSCP Field Experiment (FIFE) 1987, and three from DOE Atmospheric Radiation Measurement (ARM) program in the summer of 1997 were chosen that represent a variety of regimes from clear-sky to heavy precipitation. The clear-sky cases show an improvement due to the soil moisture information producing a better partitioning of surface fluxes. Sensitivity studies showed that 10 percent changes in soil moisture in dry regions have a much greater impact on fluxes than similar changes in relatively moist regions. However, rainfall showed little sensitivity to the land-surface treatment in cases studied so far. This is possibly because two-day simulations are too short for a statistically significant signal in precipitation patterns to show up, and because the events were driven by moisture provided by advection rather than prior evaporation during the period of the simulation.
Dudhia and Bresch, in collaborative work with Francois Guichard (visitor, Meteo France), and David Parsons (ATD), carried out an evaluation of the 30-km grid forecasts in May 1998 as NCAR's contribution to the Storm and Mesoscale Ensemble Experiment (SAMEX). The forecasts were run daily for more than a month and the domain was centered on Oklahoma. Evaluation was carried out against surface, upper air, radiation, and cloud radar data at the ARM cloud and radiation testbed (CART) site in Oklahoma. The model cloud and rain verified very well against radar data. The evaluation revealed small temperature and moisture biases, but the primary area of model deficiency was in the surface downwelling long-wave radiation from the cloud-radiation scheme, which was over-estimated particularly at night and in moister clear-sky conditions. This led to testing of the NCAR Community Climate Model 2 (CCM2) and new Rapid Radiative Transfer Model (RRTM) long-wave schemes in a chosen case from the SAMEX period, and in one-dimensional off-line tests. These more sophisticated schemes were better able to capture the nocturnal surface long-wave flux with less cooling in the surface atmospheric layers, and more cooling of the ground surface, leading to better surface air temperatures in the model. An effort is now underway to incorporate the RRTM scheme as an option in the released version of MM5.
This collaborative work is now focusing on a Tropical Ocean and Global Atmosphere Program Coupled Ocean-Atmosphere Response Experiment (TOGA-COARE) tropical Pacific case. Initial testing with low-resolution (45 km) 10-day simulations and various physics options is revealing a possible moist bias in the boundary layer that will be further investigated. The convective morphology over the period is also quite sensitive to the cumulus parameterization, and this is being evaluated against satellite and radar data in the Intensive Flux Array (IFA) region.
Additional work related to the prediction of boundary-layer processes was performed by Manning, who continued his modeling of coastal fog performing trajectory analysis on a simulation of a fog event off the coast of Scotland. One example shows the potential for simulating and forecasting coastal fog in numerical models: In an area where fog moves over land, up a lower area between ridges, air originating at low-levels off the coast moves more-or-less up the valley, becoming saturated as it loses heat to the ground. Meanwhile, air at higher levels (which originated near the surface off the coast) passes over the ridge near the surface and above the saturated air in the valley, ending up subsaturated atop a fairly thin saturated layer.
In a collaboration with Alexis Lau (Hong Kong University of Science and Technology), Simon Low-Nam, and Powers developed software to verify MM5 forecasts. The software performs analysis verification on MM5 simulations and yields skill scores for the model output such as bias and rms error. Initially being developed to assess previous simulations from MMM's real-time MM5, it is to be adapted to operate in real-time to provide objective error analyses for MM5-based forecasting systems. Results from applications of the software to the real-time MM5 forecasts performed in 1998 indicate consistent moist biases in the troposphere for 12-36 hour forecasts, in agreement with previous MM5 case-study verification work. MM5 forecast temperature biases appear to be negative in the lower- to mid-troposphere, and generally positive in the upper-troposphere. As might be expected, it is confirmed that for the suite of analyzed variables (e.g., T, RH, u, v), model rms error does increase with forecast hour for all tropospheric levels.
Retrieval and forecast experiments of convective system from CASES
N. Andrew Crook (joint appointment with RAP) commenced a retrieval and forecast study of 26 May 1999 convective system from CASES-97. This is a long-lived supercell storm that was observed by the Wichita WSR-88D radar and at low levels by NCAR's S-Pol polarimetric radar. The WSR-88D data was interpolated to Cartesian coordinates and assimilated into a cloud model using the adjoint system developed by Juanzhen Sun (joint appointment with RAP). Forecast experiments will be performed of this case to test the sensitivity to the low-level moisture and temperature fields.
b. Life Cycle of Precipitating Weather Systems
Increasing our understanding of how precipitation systems initiate, mature, and decay is a fundamental problem in atmospheric science and central to quantifying the intrinsic predictability of such systems and improving methods to forecast them. The principal type of system considered is that in which deep moist convection is organized, long-lived, and exhibits upscale growth. A second major topic involves the dynamics of systems in which precipitation is strongly localized by frontal or orographic circulations and may involve frozen precipitation. These two subsets of precipitating systems probably represent the greatest challenge for PPWS.
Morphology of Continental Organized Convection
Using operational data including GOES, NEXRAD, NOAA profiler and the NWS Family of Services analyses, Richard Carbone, David Ahijevych, and John Tuttle conducted a regional and seasonal investigation of warm season heavy precipitation episodes, building on the radar compositing efforts of L. Jay Miller, Sherrie Fredrick, and Richard Oye (ATD). [See Figure] An "episode" is defined as a clustering in time and/or space of multiple mesoscale convective systems that can lead to flooding or other societal disruption.
The emphasis of recent work was to reduce the inherent four-dimensionality of convective episodes into two-dimensional representations so that the linkages between convective systems can be better visualized. One of the diagnostic approaches was to examine the coherent patterns of heavy rainfall production in time and one dimension of space (commonly known as a Hovmoeller diagram). An example constructed from NEXRAD Information Dissemination Service (NIDS) data shows meridionally averaged rainfall rates for the period 1 June through 25 July 1998 [See Figure]. As expected, the figure is dominated by eastward propagating systems in westerly flow. Embedded in this figure, however, are structures that represent convective system mergers, amplifications, dissipations, regenerations, and retrograde motions, as well as diurnal and topographically induced variations. These features are being examined in light of known dynamical mechanisms and environmental dynamical stratification. One result of this effort (after a suitable time series is acquired) will be a dynamically based mesoscale climatology of heavy precipitation.
Two particular episodes were examined in detail. The first, 25-29 May 1998, began as a large leading-line, trailing stratiform system governed by cold pool dynamics. This rapidly moving system spawned a mesoscale convective vortex (MCV) which subsequently controlled the regeneration of slowly moving, non-squall type convective systems beginning at 0000 on 28 May. Another episode (14-15 July) was an example of a long-lived shear perpendicular system, which experienced significant changes in shear direction and thermodynamic stability over its lifetime [See Figure]. Over Montana, the shear was westerly and the line acquired a north-south elongation. Later, over Minnesota, the low level shear vector assumed a N-S orientation and the boundary-layer moisture increased. The system reoriented itself E-W and this line propagated south-south westward across Iowa and Kansas before dissipating over western Oklahoma, presumably as it moved into dryer air.
The collective behavior of all detectable MCVs during the 1998 warm season over the Central U.S. (17 in all) was examined by Stanley Trier, Christopher Davis (joint appointment with RAP), and Tuttle. The intensity and horizontal scale of MCVs were quantified using a cloud-tracking correlation algorithm [See Figure]. This analysis advanced knowledge of MCV structure since the diameter of these circulations (D = 50 - 300 km) is typically too small to be adequately resolved by the existing radiosonde and profiler networks. The longest-lived MCVs typically had horizontal scales D > 150 km and occurred in environments of weak vertical shear. These MCVs were associated with subsequent MCS development in conditionally unstable environments when convective inhibition was small. The total of 17 MCV events over the central United States in 1998, detected in this study, along with an even greater MCV frequency in preliminary analyses of MCV occurrence during the 1999 warm season, underscores the potentially important contribution of these circulations in the seasonal production of convective precipitation over the central United States.
Idealized simulations of mesoscale vortices in vertically sheared flows by Trier, Davis, and William Skamarock helped to confirm the dependence of quasi-balanced lifting on MCV intensity, horizontal scale, and vertical structure. This lifting can provide the necessary thermodynamic destabilization for subsequent mesoscale convective system (MCS) development.
Severe convection
Jeffrey Trapp (visitor, National Severe Storms Laboratory) studied the climatological distribution of tornadoes within quasi-linear convective systems (QLCS's). The horizontal extent of viable tornado-breeding sites is an order of magnitude larger in QLCS's than in individual supercells. QLCS tornadoes can be strong and produce extensive damage despite "conventional wisdom" that suggests otherwise. Also, QLCS tornadogenesis appears to occur, on average, more rapidly than does supercell tornadogenesis from the perspective of Doppler radar. The geographical, seasonal, and diurnal distributions of QLCS tornadoes are unknown.
Trapp also initiated a complementary two-part study of QLCSs. In phase 1 of the project, the annual number of U.S. tornadoes associated with QLCSs will be estimated using existing radar and verification data. QLCS tornado attributes such as average duration and damage-based intensity will then be determined, and possible geographical, seasonal, and diurnal dependencies will be explored. Some of this work was accomplished recently by Sarah Tessendorf (Significant Opportunities in Atmospheric Research and Science (SOARS) program, University of Nebraska). In phase 2 of the project, hypothesized mechanisms of QLCS tornadogenesis will be investigated numerically and theoretically, as will limitations on subsequent tornado intensity and duration. In collaboration with Morris Weisman and Nolan Atkins (Lyndon State College), some of the model experimentation and analysis was begun.
Weisman completed collaborations with Howard Bluestein (University of Oklahoma) on the study of supercell interactions within lines of convective storms. Through the use of idealized simulations, they demonstrated that the tendency to generate and sustain supercell storms within convective lines depends strongly on the orientation of the vertical wind shear profile relative to the line. This is due to the tendency for convective storms within strongly sheared environments to split, producing diverging cell motions and complicated cell collisions and interactions within a line that can interfere with the supercell mechanisms. For instance, for vertical wind shear parallel to a north-south oriented line, only the most northern cells (for southerly shear) are able to develop supercell characteristics, as other cells become engulfed by the system-scale cold pool. However, for shear profiles oriented at a 45º angle to the line (e.g., southwesterly shear), right-moving supercells are sustained along the leading edge of the gust front, while left-moving cells move into the cold pool and dissipate. An unexpectedly strong response, though, is produced when the shear is oriented at a 135º angle to the line (northwesterly shear), whereby a single, strong supercell is generated and maintained at the southern end of the system. These simulations are being compared to observed supercell line cases to determine the applicability of these results to forecasting real supercell behavior in such situations.
Further work related to the behavior of supercell storms was completed by Weisman, in collaboration with Matthew Bunker, Brian Klimowski, John Zeitler (all of National Weather Service, Rapid City, South Dakota), and Richard Thompson (NOAA Storm Prediction Center, Norman, Oklahoma), on developing and testing a technique for predicting supercell motion, as previously proposed by Weisman in the COMET CBL module "Anticipating Convective Storm Structure and Evolution." This technique, referred to as the Internal Dynamics (ID) method, is based on the theoretical work of Rotunno and Joseph Klemp, and emphasizes the shear-dependence of the supercell propagation mechanism, as opposed to the ground-relative wind dependence applied in most NWS storm-prediction algorithms. Results confirm the superiority of the ID method as compared to the currently used techniques, and are being incorporated into updated NWS forecast procedures.
Orographic Precipitation
A recent numerical study by Rossella Ferretti (visitor, University of L'Aquila, Italy), Low-Nam, and Rotunno [See Figure] of the 1994 Piedmont flood suggests that a convectively generated cyclonic vortex located over the Po Valley played a major role in determining the flood-producing orographic flow. This year Rotunno and Ferretti repeated these numerical simulations with progressively simplified orography. They found that the simplest shape that produces the same qualitative behavior as the simulation with the full topography is that of an ellipse oriented SW-NE located in the position of the Alpine massif. This idealized simulation suggests that as the front approaches from the west, the pre-frontal southerly flow induces a mountain anticyclone and hence frontogenesis on the south-west corner of the ellipse (where Piedmont would be); as usual, the frontogenesis is accompanied by rising motion and, with unstable pre-frontal air, intense precipitation.
NSSL Contributions to FASTEX Analysis
David Jorgensen and Diana Bartels (both NSSL) collaborated with Robert Gall and Melvin Shapiro (visitor, NOAA) on the analysis of aircraft data collected during the Fronts and Atlantic Storm Tracks Experiment (FASTEX) during the field phase in January-February 1997. The primary objectives of FASTEX were to improve the forecasts of end-of-the-storm-track cyclogenesis (primarily in the eastern Atlantic but with applicability to the Pacific) in the range of 24 to 72 hours, to enable the testing of theoretical ideas on cyclone formation and development, and to document the vertical and mesoscale structure of cloud systems in mature cyclones and their relation to the dynamics. Doppler and dropsonde analyses were completed for a strong cyclone that was observed by the P-3 aircraft on 23 February 1997 (IOP 18). Vertical velocity was calculated by upward integration of the horizontal divergence using a lower boundary condition of no vertical velocity at the sea surface. Results indicate that the strongest upward motions occurred in the "Cloud Head" Region north of the cyclone center. Upward motions of 10-15 cm/s were noted above 2.5 km MSL. A small downdraft, probably associated with precipitation was seen below 2.5 km. Extensive downward motions of up to 10 cm/s were seen within the "eye" or cyclone center. The strong descent with the cyclone center is consistent with pronounced drying seen in dropsonde cross sections taken as the aircraft flew from south to north. Relative humidities of <20 percent were seen in the area south of the cloud head band. Relative humidity in the rain region of the cloud head at low levels rarely exceeded 80 percent, which may be indicative of dropsonde humidity sensor bias.
More information can be obtained at the NSSL/MRB accomplishments WWW site (see http://mrd3.nssl.ucar.edu/accomplishments/FY99/FASTEX.html).
Tropical Cyclones
Davis and Lance Bosart (NCAR Affiliate Scientist, State University of New York at Albany) began an investigation of the formation of tropical cyclone Diana (1984). In this case there was a transformation of a weak, low-latitude, baroclinic cyclone into a tropical cyclone. Using the PSU/NCAR model (MM5) they were able to simulate the entire transformation process, including the development of a warm core, eye wall and spiral rainbands using horizontal resolution as fine as 3 km [See Figure]. The development consisted of two highly distinct phases. In the first phase, quasi-balanced mesoscale lifting initiated widespread convection. The more organized regions of convection produced cyclonic potential vorticity (PV) anomalies, which congealed into a coherent vortex. After becoming a marginal tropical storm, the convection intensity waned for 12-18 hours and the circulation intensity remained nearly steady. The core of the disturbance gradually became saturated below about 3-km mean sea level (MSL) during this period. New convection initiated near the radius of maximum wind as the second phase of deepening commenced, a phase apparently driven by air-sea interaction instability.
c. Mesoscale Dynamics
Fundamental studies of fluid dynamics are a key element behind major advances in understanding and prediction of atmospheric flows. Topics such as mesoscale flow over complex terrain, the dynamics of fronts and baroclinic waves, gravity waves and gravity currents are heavily emphasized. Considerable effort is also applied to the search for simplifications of the full primitive equations that are valid on the mesoscale and illustrate the underlying dynamics of mesoscale flows.
Catalina Eddy
Davis and Low-Nam, in collaboration with Clifford Mass (University of Washington), used the MM5 to examine the dynamics of a Catalina Eddy event that formed during the period 26-30 June 1988 off the coast of southern California. Their results document the importance of the diurnal cycle in modulating the interaction of coastal northerlies with coastal terrain. Only during a narrow temporal window in the late afternoon is the Froude number sufficiently large for the air to traverse the bulk of the coastal mountains and depress the marine inversion over the bight. The associated warming creates low pressure, low-level convergence and a spinup of cyclonic vorticity on a scale of 100 km. During the remainder of the day, flow is blocked (low Froude number) and the flow over the bight resembles a wake, with elongated filaments of vorticity. This diurnal variation is superposed on the synoptic scale deceleration of northerlies, which, after about three diurnal cycles, allows vorticity generated in the late afternoon and evening to persist in the bight during the following day as a mature Catalina Eddy.
These results extend the results of Skamarock, Klemp, and Rotunno who showed that realistic Catalina eddies can be obtained in flows without heating and with relatively simple coastal terrain. Thus, it appears that while planetary boundary layer (PBL) considerations appear crucial in determining the timing and structure of Catalina eddies, it is likely that the synoptic-scale flow and the gross aspects of the coastal terrain determine whether an eddy will form in the first place.
NSSL Contributions to the Mesoscale Alpine Project
David Jorgensen (NSSL) collaborated with Rotunno on the design and execution of the Mesoscale Alpine Project (MAP). During the field phase of MAP, 15 September to 15 November 1999, 19 Intensive Observing Periods (IOPs) were conducted. These IOPs involved investigation of heavy precipitation events as well as "dry" events involving flow through mountain gaps, gravity wave breaking, and investigations of flow disturbed by the Alpine mountain range. Subsequent analysis will focus on airborne Doppler radar observations of the evolution and structure of mesoscale convective systems and how complex topography interacts with the circulations to produce heavy precipitation.
More information can be obtained at the NSSL/MRB accomplishments WWW site (see http://mrd3.nssl.ucar.edu/accomplishments/FY99/MAP.html).
Corrections to Quasi-geostrophy
Understanding the relation between meso- and synoptic-scale flows involves, in part, understanding how atmospheric dynamics change as the Rossby number increases. Quasi-geostrophic theory represents a leading-order theory in the sense that it is derivable from the full primitive equations in the asymptotic limit of zero Rossby number. At small Rossby number, virtually all dynamical theories rest upon the foundation of quasigeostrophy (QG), which is the leading-order theory in Rossby number. David Muraki (Courant Institute, New York University), Snyder, and Rotunno (Muraki et al.1999) introduced a convenient technique for extending QG to an additional order in Rossby number; they call this extended theory "QG+1." Several studies employing QG+1 are underway. Rotunno et al. (1999) applied QG+1 to idealized baroclinic waves [See Figure]; they illustrated how familiar synoptic diagnostics such as Q-vectors fit within the QG+1 framework. They also showed that the characteristic cyclonic bias in primitive-equation simulations of baroclinic waves arises from a next-order correction in the inversion of potential vorticity. This represents the first mathematically complete and relatively simple explanation for this asymmetry.
In other related work, Gregory Hakim (University of Washington) and Muraki (1999) studied waves on the tropopause and showed that QG+1 captures the characteristic asymmetries of the tropopause displacement in observed disturbances. Muraki, Rotunno, and Snyder are considering flow over topography with QG+1 and examining how gravity waves and fore-aft asymmetry enter the topographic flow problem at small Rossby number. Finally, Hakim, Muraki, and Snyder are applying QG+1 to simulations of quasi-two-dimensional, balanced, decaying turbulence to understand the observed preference for cyclonic vortices on the tropopause at subsynoptic scales.
Traditionally, synoptic-scale observations are grossly inadequate to initialize high-resolution cloud and mesoscale models. An effective assimilation of non-traditional observations (i.e., radar, satellite, GPS data, etc.) for model initialization is one of the most challenging research problems in numerical weather prediction. Our research over the past year focuses on three major areas. The first one is related to the development of three-dimensional variational (3D-Var) data assimilation systems suitable for cloud and mesoscale model applications. The second topic focuses on the assimilation of observed Doppler radar data for cloud-scale model initialization and prediction. The third subject addresses the applications of the MM5 3D-Var/4D-Var system for the assimilation of remote sensing observations and for mesoscale dynamic studies.
Development of a 3D-Var system for radar data assimilation
Juanzhen Sun (joint appointment with RAP) developed a 3D-Var radar data assimilation system and tested the system using simulated data of a collapsing cold pool. The 3D-Var system is able to retrieve most of the 3-D wind structure. A Gal-Chen thermodynamic retrieval algorithm is then used to obtain the temperature information. The system was compared with the 4D-Var system in terms of the retrieval accuracy and the performance of the subsequent forecast using the simulated data. Further research using real data will be conducted in the near future.
Development of a Community MM5 3D-Var System
Francois Vandenberghe, Wei Huang, Dale Barker, and Ying-Hwa Kuo continued the development a 3D-Var data assimilation capability for MM5. This procedure assimilates real-time conventional observations (synoptic, upper air, buoy and radiosonde) in a univariate statistically optimal scheme. System evaluations against the current Cressman scheme (RAWINS program) showed comparable performance. This is encouraging given the simplicity of the current system. The code was upgraded to work with the latest version (3.2) of MM5 and to be Y2K compliant. Efforts now focus on the addition of a balance constraint in the cost function that will render the system multivariate and alleviate the problem of initial imbalances associated with model initialization. For that purpose, statistics on MM5 forecast errors were collected on a daily basis over selected regions (Continental U. S. and South Asia) and compiled into monthly means. These statistics are expected to provide useful information on the geostrophic/ageostrophic parts of the model-forecast error, and will be incorporated into a background error co-variance term.
Development of a hybrid ensemble Kalman filter/3D-Var scheme
Hamill and Snyder developed a hybrid ensemble Kalman filter/3D-Var scheme and evaluated this scheme using a quasi-geostrophic model under perfect-model assumptions. Four networks with differing observational densities were tested, including one network with a data void. The hybrid scheme operates by computing a set of parallel data assimilation cycles, with each member of the set receiving unique perturbed observations. Background error covariances for the data assimilation are estimated from a linear combination of time-invariant 3D-Var covariances and flow-dependent covariances developed from the ensemble of short-range forecasts. The hybrid scheme allows the user to weight the relative contributions of the 3D-Var and ensemble-based background covariances.
The analysis scheme was cycled for 90 days, with new observations assimilated every 12 hours. Generally, it was found that the analysis performs best when background error covariances are estimated almost fully from the ensemble, especially when the ensemble size was large. When small-sized ensembles are used, some lessened weighting of ensemble-based covariances is desirable to reduce the impact of spurious analysis corrections over long distances. The relative improvement over 3D-Var analyses was dependent upon the observational data density; generally, there is less improvement for data-rich networks than for data poor networks, with the largest improvement for the network with a data void. As expected, errors depend on the size of the ensemble, with errors decreasing as more ensemble members are added. The sets of initial conditions generated from the hybrid are generally well calibrated and provide an improved set of initial conditions for ensemble forecasts.
Assimilation of surface mesonet data
Davis and Low-Nam implemented an algorithm to combine background information from an MM5 forecast with surface mesonet data into an analysis suitable for grid nudging on a domain with 3.3 km horizontal resolution. The increments were assumed to have a Gaussian weighting function in the horizontal with the decay scale determined by station density. The scheme also reduced the influence of observations in proportion to the variance of terrain between grid points and observation locations. Increments of wind and temperature were analyzed at 10 m and 2 m, respectively, and then mapped back to the lowest model level using similarity theory consistent with the MM5 boundary layer scheme.
Assimilation of GPS/MET radio-occultation
Vandenberghe and Kuo, in collaboration with John Derber (NCEP) and Xiaolei Zou (Florida State University), completed the 3D-Var assimilation of GPS/MET radio-occultation measurements. Sixty-two soundings of refractivity profiles and refraction angle distributions available within a 12-hour time window (12:00-24:00 UTC 10/11/99) were assimilated into the NCEP global analysis. Results showed that assimilation of refractivity data is computationally cheaper and worked well in most cases. However, refraction angle must be used in presence of strong humidity gradients. Sensitivity analysis using the adjoint of the ray-tracing model showed that the integrated information of the radio-occultation measurement impacted a very large area (up to 600km x 600km for a single sounding).
Background error covariance for a real-time radar data analysis system
Sun, in collaboration with Doug Nychika (CGD), developed an algorithm to model the forecast background error covariance in a real-time 4D-Var radar data analysis system. This algorithm employs a homogeneous correlation function with an exponential distribution. This exponential function is separable in x and y and so can be easily inverted. The inverted matrix is then truncated to create a filter which, when applied repetitively, behaves similarly to a Gaussian filter. The background error variance is adaptive and estimated by comparing the radial velocity from the analysis and that from the observations. The algorithm was tested on both simulated data of a collapsing cold pool and WSR-88D radar data. It was demonstrated that the inclusion of the background term with the specified error statistics improves the subsequent forecast.
Assimilation of heterogeneous mesoscale observations for a mesoscale convective system
Yong-Run Guo, Kuo, Dudhia, Parsons, and Christian Rocken (GST) completed a study on the assimilation of heterogeneous data for a mesoscale convective system over Oklahoma that occurred in September 1996. The objective of the study is to assess the relative importance of various types of mesoscale observations on the results of four-dimensional variational assimilation. The data included wind-profiler data, surface mesonet dew point, GPS receiver precipitable water vapor, and surface rainfall data. Six-hour assimilation experiments were carried out with the MM5 4D-Var system at 20-km grid resolution. The MM5 4D-Var system was able to reproduce the observed rainfall in terms of precipitation pattern and amount, and substantially reduced the model errors when verified against independent observations. Wind profiler data, the only three-dimensional data set used, was found to be vital to retrieving the vertical structure of moisture and temperature of the convective system. Assimilation of surface rainfall data also appeared to help the retrieval of temperature fields. Further studies showed that, in general, more sophisticated cumulus and microphysical schemes in the adjoint model lead to better assimilation of the observations by the 4D-Var procedure. The primary limitation of this method appears to be that with a relatively small domain and without having the boundary conditions as control variables, the benefit of 4D-Var assimilation is restricted to a small area immediately downstream of the bulk of the observations. Adding boundary conditions to the control variables and/or expanding the domain would alleviate this problem. Work continues to evaluate the use of a penalty term in the cost function to expedite the minimization process and produce better convergence.
Impact of moist physics on mesoscale data assimilation
Guo, Dudhia, and Kuo examined the effects of different combinations of moist physics used in the MM5 4D-Var system on the results of the data assimilation for the MCS case described above. The physical parameterization schemes examined include: (i) the Anthes-Kuo scheme, (ii) the Grell cumulus parameterization schemes, (iii) the Dudhia explicit moisture scheme, and (iv) removal of the large-scale supersaturation. They found that the use of the explicit moisture scheme with ice effects and no cumulus parameterization in the MM5 4D-Var system provides the best fit to the observations. The inclusion of cumulus parameterization degraded the convergence to a minimization of the cost function. The Anthes-Kuo scheme is even worse than the Grell scheme. It is not clear if this is a general conclusion for the assimilation of mesoscale convective systems. Additional diagnosis and assimilation experiments using different cases will be performed in the future.
Assimilation of ground-based GPS total zenith delay
Manuel De Pondeca (visitor, Florida State University), in collaboration with Jean Dickey (Jet Propulsion Laboratory), Richard Lind (Naval Postgraduate School), and Zou performed high resolution (6 km) 4D-Var assimilation experiments of GPS/MET total zenith delay over South California using data from the Jet Propulsion Laboratory. Results indicate that assimilation of GPS total zenith delay measurement alone increases the accuracy of 6 hour accumulated rainfall forecasts only marginally (~23%). However, this improvement becomes significant (~78%) when wind profiler observations are added and are further improved (83%) with the addition of radio acoustic sounding system (RASS)-virtual temperature observations.
Targeted observations
In order to identify a target area for enhanced observations (with a goal to improve forecast accuracy), adjoint model sensitivity vectors are often used as guidance. A key question is: Will two different adjoint models produce similar sensitivity vector for a given synoptic flow regime? To answer this question, Guo, Jian-Wen Bao (NOAA/Environmental Technology Laboratory), and Kuo compared the sensitivity analysis from two different adjoint models based on the same response function. They calculated the sensitivity fields by integrating the dry version of the Multispectral Atmospheric Mapping Sensor (MAMS, developed by Ronald Errico, CGD) and MM5 4D-Var systems backward for 36 hours for an extratropical cyclone case, the Experiment on Rapidly Intensifying Cyclones in the Atlantic (ERICA) IOP2. They found that the sensitivity vector from the different modeling systems, are transferable, when wind and temperature fields at a model grid point are used as a response function. This has practical implications in the planning of targeted observations. They plan to explore the feasibility of using an ensemble of sensitivity vectors from several modeling systems to define a strategy for identifying targeted areas for enhanced observations.
NSSL Contributions to the Winter Storms Project (NORPEX)
Diana Bartels (NSSL) continued collaborations with Shapiro and Gall on preliminary analyses of dropsonde data collected with the NOAA Gulfstream aircraft during January and February 1999 over the North Pacific. The data were collected as part of the NCEP Winter Storms Reconnaissance Program (WSRP). High priority cases for analyses include several clear air turbulence events.
More information can be obtained at the NSSL/MRB accomplishments WWW site (see http://mrd3.nssl.ucar.edu/accomplishments/FY99/Bartels2.html).
Use of sensitivity fields to improve model initial conditions
Kuo, Guo, and Wei Huang, in collaboration with Richard Reed (University of Washington), completed a study of a mesoscale cyclone over the Mediterranean Sea. The storm formed over the warm waters between Sicily and Libya on 23 January 1982 in a region of weak flow ahead of a cold, upper-level trough. During the ensuing five days it moved erratically, performing two loops, one near Sicily and a second over the Ionian Sea, before dissipating off the south coast of Turkey. During the latter part of its life, the storm diminished in size, assuming a hurricane-like appearance in satellite imagery. Ships near the vortex center reported near-hurricane force winds at this time. An attempt to simulate the storm development with MM5 met with mixed success. The initial formation and movement were correctly predicted, but the rapidity of the development and the depth of the system fell short of the prediction. More serious errors occurred later when the predicted track departed substantially from the observed and the contraction of the storm to mesoscale dimensions was not predicted. A number of forward- and adjoint-sensitivity experiments were conducted to identify factors influencing the development and to explore ways of improving the prediction. The most realistic prediction was achieved by implanting a vortex, in the manner commonly done in tropical cyclone prediction, at an early stage in its history and, in addition, by using adjoint sensitivity to create revised initial conditions at that stage. Specifically, perturbations were created based on the results of 36-hour adjoint sensitivity fields, and then added to the initial condition. The revised initial condition based on the resulting adjoint sensitivity fields led to improved prediction of the cyclone track.
e. Weather Research and Forecast (WRF) Model Development
The overall goal of the WRF Model project is to develop a next-generation mesoscale forecast model and assimilation system that will advance both the understanding and prediction of important mesoscale precipitation systems, and promote closer ties between the research and operational forecasting communities. The model is intended to improve the forecast accuracy of significant weather features across scales ranging from cloud to synoptic, with priority emphasis on horizontal grid resolutions of 1-10 kilometers. The model will incorporate advanced numerics and data assimilation techniques, multiple relocatable nesting capability, and improved physics, particularly for treatment of convection and mesoscale precipitation. It is intended to be well suited for a range of applications, from idealized research to operational forecasting, and have flexibility to accommodate future enhancements. As soon as the WRF Model becomes functional, it will be maintained and supported by NCAR as a community model and freely distributed.
The WRF project began as a collaborative effort among NCAR/MMM, NOAA/NCEP, NOAA/FSL, and OU/CAPS, together with the participation of a number of university scientists. During the past year, these participants conducted a comprehensive planning process to identify the specific research components of the overall effort, and submitted a joint proposal to the NSF/NOAA USWRP program that received approval in August. Recently, participation in the WRF effort has broadened through collaborations with other organizations, including the USAF Air Force Weather Agency (AFWA), NOAA Geophysical Fluid Dynamics Laboratory, NRL Marine Meteorology Division, NASA/GODDARD Atmospheric Sciences Division, and EPA Atmospheric Modeling Division. To accommodate the growing involvement in WRF, a formal management plan has been formulated that facilitates broad participation in the development of the system components, advisory involvement of potential model users, and oversight by the principal funding agencies. This plan is presently under review by the USWRP Interagency Working Group and participating agencies.
Model Prototypes for Integrating the Dynamical Equations
During the past year there was significant progress in developing three prototypes for the nonhydrostatic WRF model numerics that can be used for comparative evaluation in selecting the final numerical design. Two of these prototypes are split-explicit Eulerian models, while the third is a semi-implicit semi-Lagrangian formulation.
The Eulerian prototypes were designed by Klemp, Skamarock, and Dudhia following the philosophy that prognostic variables should be quantities that have conservative properties, and that variables that have no conservation characteristics be computed from appropriate diagnostic relations. In this manner, equations for momentum and specific entropy are integrated in flux form, while pressure is diagnosed from the gas law. A new third order accurate two-time-level split-explicit integration scheme is employed (discussed below) together with high order upwind or centered differencing for advection. One Eulerian prototype is formulated using a terrain-following height coordinate while the second employs hydrostatic pressure (or mass) as the vertical coordinate. While the height-based coordinate was used extensively in nonhydrostatic applications there exists little experience with the nonhydrostatic mass coordinate. Both prototypes appear to be robust in simulations over a broad range of scales, and testing began for a variety of idealized flows, including moist convection with simple cloud microphysics [See Figure].
James Purser (visitor, NOAA/NCEP) is leading the development of the semi-implicit semi-Lagrangian prototype. To attain a high formal order of accuracy for the spatial operations of differentiation and quadrature, Purser developed a package of efficient "compact" or "Pade" schemes that is well suited for either serial or parallel computers. These methods form an integral part of another package of high-order conserving "cascade" interpolations, which are designed for use in the grid-to-grid interpolations needed for the semi-Lagrangian calculations. This prototype is presently under development and will be evaluated in comparison with the other prototypes. The code is being written so that the interpolations will eventually be able to accommodate both conventional (hydrostatic sigma) and hybrid vertical coordinates, and a barotropic capping layer above the top of the model-proper will be explored to enable an accurate accounting for mass and tracer budgets.
Software Architecture Prototype
Led by John Michalakes (visitor, Argonne National Research Laboratory), a demonstration prototype code for the WRF software architecture was designed and implemented. Within this prototype, a single version of the code may be configured for efficient execution on a variety of platforms covering the range of current and foreseen high performance computing hardware (such as shared-memory multiprocessors (SMPs), distributed- memory multiprocessors, and distributed-memory clusters of SMPs). Model domains are decomposed for parallelism on two levels. For distributed memory, the full domain is divided into rectangular "patches" of arbitrary size and shape that are assigned to each node. For shared-memory processors within a node, patches may be further subdivided into "tiles." The WRF software architecture abstracts and encapsulates hardware and application specific concerns within appropriate layers of a software hierarchy, allowing WRF to adapt to diverse hardware platforms without change to the model layer of the code. The WRF software prototype is presently configured with the Eulerian height-coordinate dynamics solver, and implementation using the other two dynamics modules is under development. Benchmarking is underway to test the efficiency of the prototype on distributed-memory, shared-memory, and hybrid cluster-of-SMP parallel configurations, and to evaluate the tradeoffs between communications and computations in the split-explicit formulations. The design is also flexible in adapting to vector processor and microprocessor cache-based machines. Development of a portable efficient I/O interface is also proceeding.
Numerical Methods Development
Louis Wicker (NOAA/National Severe Storms Laboratory) and Skamarock continued the development and testing of alternative approaches for time integration in the split-explicit flux-form WRF model prototypes. Two different two time level schemes were adapted to a time-split application and are being evaluated together with the traditional leapfrog time-discretization: a 2nd order Runge Kutta (RK2) time discretization and a 3rd order Adams-Bashforth-Moulton (ABM) time integration method. The RK2 integration technique allows the use of 3rd-order upwind advection operators and does not suffer from the large dispersion errors found in the leapfrog schemes. But, stability constraints are severe for higher order upwind advection operators and the 3rd order upwind scheme does result in significant computational damping; centered advection operators are unstable. The 3rd order ABM scheme has very low dispersion errors and is not severely constrained when used with either odd or even higher order advection operators. All these schemes appear robust, and, as anticipated, the higher order schemes give superior solutions at marginal resolution.
Stephen Thomas (SCD), Skamarock, Klemp, and Michalakes examined semi-implicit formulations of the Eulerian height-coordinate flux-form models and mass (hydrostatic pressure) coordinate model. For the height coordinate models, the semi-implicit formulations are straight-forward extensions of previously outlined approaches. For the mass coordinate models, extensions to a semi-implicit formulation are not obvious, and existing approaches have some stability problems. Multi-dimensional Helmholtz/elliptic equations arise in semi-implicit formulations, and parallelization of conjugate gradient based Helmholtz equation solvers, along with a search for robust and efficient preconditioners, is the another focus of this research effort.
As part of the WRF model development project, Skamarock, Stanley Benjamin (NOAA Forecast Systems Laboratory), Reiner Bleck and Zuwen He (both University of Miami) continued to examine and refine hybrid coordinate model formulations for the nonhydrostatic compressible equations. The hybrid coordinate takes the form of a terrain-following sigma-like coordinate near the surface and relaxes to an isentropic coordinate (or any other specified coordinate) aloft. Using the split-explicit solution technique, the integration of the acoustic modes is split from the coordinate-surface movement. This approach may be generally more robust than semi-implicit techniques requiring 3D Helmholtz solvers, but other problems are being examined, including possible detrimental effects associated with splitting, in a variety of idealized model simulations.
WRF Model Data Assimilation
To speed progress on the development of assimilation techniques for the WRF model, a package of 3D variational analysis tools and some efficient grid-to-observation interpolators were assembled by David Parrish (NOAA/NCEP), from the operational NCEP code, for distribution to the WRF development community. In addition, the filtering methods that lie at the heart of the proposed analysis techniques underwent some significant enhancements by Purser during the year, including improved conformity of the basic "building blocks" to the intended Gaussian function shape, and improved numerical robustness and efficiency brought about through the adoption of "multigrid" techniques.
2. Cloud and Surface Processes & Parameterization (CaSPP) Program
One of the two primary scientific programs in the division is the Cloud and Surface Processes Parameterizations (CaSPP) program. Its goal is to study the impacts and parameterizations of mesoscale and microscale processes in large-scale models. CaSPP intends to achieve this approach in a systematic way, by integrating the NCAR Clouds in Climate Program (CCP), which is mostly devoted to tropical convection, with ongoing research in boundary layer clouds and small-scale surface processes. To collectively carry out the division's Scientific Strategic Plan, workshops are planned on an annual basis to demonstrate progress and to prioritize key issues in the treatment of phenomena in large-scale forecast models. The first workshop was held in September 1999 in Estes Park, Colorado. In addition to 13 MMM scientists involved in the CaSPP program, 11 collaborators from other NCAR Divisions (primarily CGD scientists) and universities were present.
The three primary research thrusts, namely tropical cloud systems, marine stratocumulus, and land-surface interactions were selected for discussion at the Workshop because of the scientific expertise and breadth within the MMM Division. Manifested initially as process studies, this research now extends to interactions between processes and the environment -- building blocks for physically based parameterizations. The Workshop participants highlighted progress that has been demonstrated in impact studies. Examples included effects of convective cloud systems on the atmospheric energy balance; effects of convective momentum transport on the global circulation of the atmosphere; clouds and radiation; and impact of surface processes on regional weather. Several action items and goals that have direct relevance to parameterization issues were discussed and proposed by the Workshop participants.
a. Tropical Cloud Systems
At spatial resolutions of climate simulations, the fundamental problem is how to deal with the disparity between the resolved scales and processes occurring at much shorter time and space scales that must be parameterized. The primary issues concern phase changes of water occurring in various kinds of cloud systems and the effect these systems have on radiative transfer and the atmospheric circulation at large. A poorly understood aspect is the role of cloud systems organized on various scales. Quantifying the collective effects of such systems is the key challenge. This requires an understanding of how cloud microphysics, turbulence, and radiative transfer are coupled by means of cloud-scale and mesoscale dynamics and how feedback with larger scales operates on various time scales.
Nonhydrostatic Global Model Development
Piotr Smolarkiewicz, Wojciech Grabowski, Len Margolin (Los Alamos National Laboratory), Andrzej Wyszogrodzki and Miroslaw Andrejczuk (both University of Warsaw, Poland) continued the development of their nonhydrostatic global model based on nonoscillatory forward-in-time (NFT) numerical methods. They extended the model to include natural orography and moist precipitating convection, enabling more realistic simulations of the Earth's climate and weather. They also added an extensive statistical analysis package, and special purpose programs for global mesh refinement and/or reduction to increase the versatility and utility of the model. While visiting the European Centre for Medium-Range Weather Forecasts (ECMWF), Smolarkiewicz implemented the implicit and explicit (with respect to internal gravity waves) schemes in the small- and global-scale models as selectable options. This makes the global and small-scale models fully compatible, and prepares the ground for a future unified all-scale nonhydrostatic model.
Grabowski and Smolarkiewicz implemented moist precipitating thermodynamics in their nonhydrostatic anelastic global model. The model is a two-time-level nonoscillatory forward-in-time (NFT) Navier-Stokes solver suitable for modeling a broad range of natural and idealized atmospheric flows. The implementation of the moist scheme follows the strategy developed for cloud models and leads to a new dynamical/thermodynamical framework for future cloud-resolving, climate/numerical weather prediction models. The approach not only realizes stable integrations when cloud processes are poorly resolved but also converges to the standard formulation used in cloud models as the resolution increases. Because over-parametrization at high resolution is avoided, this is convenient for mesh refinement schemes and the new Cloud Resolving Convection Parameterization (CRCP) approach, defined in section 2.E.
TROPICAL CLIMATE AND CLOUD MICROPHYSICS
Grabowski performed a series of idealized numerical simulations to investigate the role of cloud microphysics in tropical climate, assuming a radiative-convective-dynamical equilibrium paradigm for the tropical atmosphere. Simulations applying dramatically different parameters in the representation of cloud microphysics were compared. The results show that the mean temperature and moisture profiles in the tropics are insensitive to the cloud microphysics, in agreement with previously reported work. Consequently, the collective role of cloud microphysics is not as crucial as studies applying microphysical parameterizations to simulations of single clouds might imply. The key point is that the main impact of cloud microphysics is not so much on atmospheric processes and dynamics, but rather on the ocean surface. If this proves to be realistic, effects of cloud microphysics will be difficult to quantify in the coupled ocean-atmosphere system because of the long time scales associated with ocean dynamics.
CLOUD-RESOLVING MODELING OF LARGE-SCALE TROPICAL CIRCULATIONS
Grabowski, in collaboration with Mitchell Moncrieff and Jun-Ichi Yano (Climate Research Centre, Monash University, Australia), continued to study effects of atmospheric moist convection on large-scale tropical dynamics using cloud-resolving models. Two-dimensional simulations reported in previous years (4,000-km domain) and a large-scale sea-surface temperature (SST) gradient were analyzed. A vertical normal mode analysis showed that quasi-two-day oscillations are associated with the gravity waves radiated from the convective region over high SSTs into the descending branch over the low SSTs. Another set of numerical simulations assumed a periodic global-scale horizontal domain (20,000 km) and homogeneous SST. Spontaneous organization of convection into eastward-propagating superclusters occurs in some simulations [See Figure]. These simulations are also being used to evaluate the CRCP technique by comparing with the cloud-resolving simulations.
HIERARCHICAL MODELING OF CLOUD SYSTEMS OBSERVED IN GATE
Changhai Liu, Moncrieff, and Grabowski simulated cloud systems observed during 1-7 September 1974 in Phase III of the Global Atmospheric Research Program (GARP) Atlantic Tropical Experiment (GATE) using a hierarchical approach. This approach consists of: (i) two-dimensional cloud-resolving modeling that explicitly treats cloud-scale and mesoscale dynamics (2-km resolution); (ii) two and three-dimensional modeling applying the Kain-Fritsch convective parameterization (10-, 15- and 25-km resolution); and (iii) coarse-grid modeling without any convective parameterization whatsoever. Both (i) and (ii) realize three distinctive cloud systems (squall line, non-squall convection, and scattered convection) and transitions among these systems as the (specified) large-scale conditions evolve. For the first time in a three-dimensional experiment with parameterized convection, a fast-moving squall system observed on 4 September was realized [See Figure]. Method (iii) fails to produce the squall system (moderately disturbed conditions), but successfully represents the non-squall convection in highly disturbed conditions. This begins to quantify the conditions under which convective parameterization is necessary at 10-km resolution (i.e., next generation general circulation models). The GATE cloud-resolving three-dimensional simulations referred to here and elsewhere may be viewed at: http://www.scd.ucar.edu/vg/GATE/GATE.html.
WAVELET ANALYSIS OF SIMULATED TROPICAL CONVECTIVE CLOUD SYSTEMS
Yano, Moncrieff, Xiaoqing Wu (joint appointment with CGD), and Michio Yamada (University of Tokyo, Japan) completed their wavelet analysis of the aforementioned three-dimensional 7-day cloud-resolving simulation of convective cloud systems in GATE Phase III. Three distinct regimes (squall-line, nonsquall cloud cluster, scattered convection) were analyzed using complete discrete Meyer wavelets. The wavelet spectra characterize the spatial localization of each physical variable and decompose cloud systems into convective and stratiform components. The wavelet spectrum also represents a preferred spatial orientation. This proves the utility of discrete Meyer wavelets to characterize and classify organized convective cloud systems in an objective way.
NUMERICAL SIMULATIONS OF TROPICAL ISLAND THUNDERSTORMS
N. Andrew Crook (joint appointment with RAP) continued his numerical study of convection initiation over the Tiwi Islands, north of Darwin, Australia. A linear model of flow over an isolated heat source was developed and compared with nonlinear simulations. Examinations included the sensitivity of the convective strength to a number of parameters such as flow speed and direction, Bowen ratio and various microphysical parameters. Snapshots of surface rainwater and velocity [See Figure] show a large thunderstorm that was simulated. The sea breezes from the north and south coastline converged to within 10 km of each other before strong convection developed. Observations from the Maritime Continent-Thunderstorm Experiment (MCTEX) showed that strong convection can develop when the two sea breezes are some distance apart; however, the simulations indicate that the sea breezes from both coastlines are still important for strong initiation.
PRECIPITATION DEVELOPMENT IN SMALL MARITIME CUMULUS
Charles Knight continued analysis of the Small Cumulus Microphysics Study (SCMS) of early cloud echo development in Florida. He completed time-height diagrams for all 35 cases with radar coverage that started early (less than 0 dBZ at X-band) and developed precipitation detectable by radar. The primary message is the extreme variability of the echo growth, but correlation with the aerosol measurements is being sought. Sonia Lasher-Trapp (ASP), working with Knight, developed a model study of the possible role of giant aerosol on first-echo development in cloud observed in the SCMS. It is possible that the first radar precipitation echo is due to drops that grew by direct collection of cloud droplets on giant aerosol particles or giant haze droplets.
Knight started analysis of S-pol data from Florida in August 1998 (from the Precipitation [PRECIP] 98/Tropical Experiment), comparing time-height diagrams of ZDR with those of dBZ in first-precipitation-echo development. As in the SCMS data, variability is the rule, although very large ZDR values (+2 dB or greater) are frequently seen very early in the echo growth–even at 0 dBZ.
Participation in the Indian Ocean Experiment (INDOEX)
Andrew Heymsfield and Gregory McFarquhar participated in the Indian Ocean Field Experiment (INDOEX), an experiment with the objectives of measuring the direct and indirect (cloud) effect of aerosols and greenhouse gases emitted from India on the radiative budget of the Indian Ocean downwind of India. This experiment was organized in the U.S. by the Center for Clouds, Chemistry and Climate (C4) at Scripps Institute for Oceanography, and included a large NCAR component involving the use of the NCAR C-130 for the radiative and microphysical measurements (ATD) and a modeling effort (CGD). Using the microphysical data from the C-130, especially data from the "gradient" flights whereby the C-130 flew from "polluted" regions off the coast of India to "clean" conditions to the south of the intertropical convergence zone, Heymsfield and McFarquhar examined the relationship of the cloud microphysical properties to the degree of pollution. Earlier studies showed that condensation nucleus (CN) concentration is a measure of anthropogenic emissions, in that CN <500 cm-3 represent "clean" conditions, 500 to 1500 cm-3 relatively polluted conditions, and >1500 cm-3 highly polluted conditions. They found that the bulk cloud properties, e.g., liquid water content, cloud width, and cloud top height, were about the same, on average, in all three regimes. However, the droplet concentrations in the polluted clouds were three times larger than in the clean clouds, whereas the mean diameters were on average 25 percent larger in the clean versus polluted clouds. Furthermore, there was a prevalence of large drops in the polluted clouds. These results have implications for cloud lifetimes, in that the polluted cloud lifetimes are not constrained by the fallout of precipitation through the drizzle process, as well as for cloud albedos, as cloud albedo for a given liquid water content is inversely related to drop size.
Boundary layer clouds are important in climate modeling because they strongly affect shortwave radiative transfer and surface-atmosphere exchange. Key aspects include understanding the entrainment mechanism and how to properly parameterize the entrainment rate and the cloud amount.
ENTRAINMENT RATE, CLOUD FRACTION AND LIQUID-WATER PATH IN STRATOCUMULUS
With the goal of developing simple schemes to parameterize key parameters in
climate models, Chin-Hoh Moeng simulated a number of STBLs (stratocumulus-topped
planetary boundary layers (PBLs)). Large Eddy Simulation (LES) was used to cover
a range of large-scale conditions and forcing. This synthetic database was used
to investigate mechanisms that determine the entrainment rate, cloud fraction
and liquid water path. The resulting entrainment-rate formula is shown [See
Figure] to give a good estimate of the entrainment
rate of the simulated STBL, and two key factors, 
,
a measure of cloud-top entrainment instability and, zf,
the cloud-top height that determine the simulated cloud fraction and liquid
water path, are identified [See Figure].
In an effort to gain more understanding of the scales and structure of the entrainment zone at the top of marine stratocumulus, Donald Lenschow (joint appointment with ATD), in collaboration with Mingyu Zhou (National Research Center for Marine Environmental Forecasts, Beijing, China), analyzed data from the NCAR Electra aircraft collected during the Dynamics and Chemistry of the Marine Stratocumulus Experiment (DYCOMS) off the California coast. They found a near discontinuity in the thermodynamic structure across cloud-top, but that PBL air leaks across this interface so that the layer just above cloud top is a mixture of PBL and free-tropospheric air. They also found that formulations developed for refractive index fluctuations at the clear-air PBL top were applicable to the stratocumulus-topped PBL. The results of this work are being used by Lenschow and Bjorn Stevens (University of California, Los Angeles) to help plan a new experiment that was proposed for the same stratus regime using the NCAR C-130 aircraft. DYCOMS-2 will focus on obtaining more accurate entrainment-rate estimates than previously possible using new techniques for measuring scalar fluctuations, mean flow divergence and PBL height.
REMOTE SENSING OF THE PLANETARY BOUNDARY LAYER STRUCTURE
Lenschow, Volker Wulfmeyer (ATD), and Christoph Senff (Cooperative Institute for Research in Environmental Sciences) developed a technique for removing random uncorrelated noise from measurements of third- and fourth-order moments of atmospheric variables. They applied the technique to lidar measurements of vertical velocity and humidity to obtain vertical profiles of these moments throughout the PBL from ground-based measurements.
Lenschow and Shelby Frisch (NOAA/Environmental Technology Laboratory) studied the evolution of vertical velocity statistics throughout the daytime PBL using the NOAA 35 GHz Doppler radar which was deployed in the Canadian boreal forest during the Boreal Ecosystem-Atmosphere Study (BOREAS) in the summer of 1994. Backscatter targets in clear air are mostly insects, while in clouds the radar detects cloud droplets. They obtained measurements of vertical velocity variance, skewness, and spectra, as well as coherence and phase angle as a function of vertical separation throughout the PBL. These measurements were used to document the development of the clear air convective PBL throughout the morning, the subsequent formation of fair-weather cumulus in the early afternoon, and their dissipation in late afternoon. They found, for example, that as clouds form, the variance increases with height throughout the PBL, in contrast to the cloud-free PBL where the maximum variance occurs at about 40 percent of the PBL height.
c. Land-Atmosphere Coupling at Small Scales
Scientific challenges in land-surface coupling involve developing effective sampling strategies and accurate parameterization of local land-atmosphere interactions, integrating the relevant fluxes across the heterogeneous interface, representing these fluxes on a scale usable by a numerical model and, finally, developing parameterizations that can work throughout the diurnal cycle. The local problem is being dealt with in a limited way through testing and development of land-surface schemes. The integration or 'aggregation' problem is being addressed with a combination of in situ and remote sensing data, land-surface models, and large-eddy simulations.
AREA-AVERAGED TURBULENT FLUXES IN BOREAS
Jielun Sun, along with Larry Mahrt (Oregon State University), Forrest Hall and Venkataraman Lakshmi (both NASA/Goddard Space Flight Center), Jing Chen (Canada Centre for Remote Sensing), Valentijn Pawels (Princeton University), and Harry McCaughey (Queens University, Canada), used BOREAS data to map surface fluxes. The tower-based data span three seasons and different semi-homogeneous surface types, albeit mostly high and dry ground. Daytime aircraft measurements cover a variety of surfaces including types that the tower data do not cover. In order to provide area-averaged surface turbulence fluxes for numerical simulations, surface flux maps were constructed using not only tower and aircraft data but also high-resolution satellite land classifications. The aircraft-derived fluxes over surfaces where towers are absent are related to the tower fluxes to estimate the diurnal variation of the surface fluxes over all the surface classes. The area-averaged surface flux is based on this flux map.
Using two distinct approaches, two groups collaborating with Margaret LeMone and Robert Grossman (University of Colorado), produced flux maps for the 21 April-21 May 1997 period of the Cooperative Atmosphere Surface Exchange Study (CASES)-97, during which the primary surface vegetation changes from a mix of dormant grass and rapidly growing winter wheat to more uniform green vegetation. Both sets of flux maps are consistent with observations and will be used to drive mesoscale simulations. Jie Song (Northern Illinois University) and Marvin Wesely (Argonne National Laboratory) used satellite data and meteorological data from CASES-97 in the Parameterization of Subgrid-scale Surface fluxes (PASS) model to produce flux maps. Model fluxes agree reasonably with those from the eight NCAR stations and from low-level aircraft legs.
David Yates and Fei Chen (both RAP) applied interpolation techniques to CASES-97 surface-station data and collected satellite, gauge-corrected radar-estimated rainfall from both S-Pol (Edward Brandes, RAP) and NEXRAD land-use and soil data to produce a time-varying, multi-scale (with spatial resolution of 1, 5, and 10 km) gridded dataset [See Figure]. Careful examination of the data during this project led to improvements of the CASES-97 surface dataset, implemented by Steven Oncley (ATD). Comparisons of the interpolated fluxes with stations not used in producing the data are encouraging. Harayasu Nagai (visitor, Japan Atomic Energy Research Institute), Yates, and Chen utilized three land-surface models along with the above gridded dataset to generate multi-scale surface heat flux maps. These maps are being evaluated against in situ measurements. Preliminary comparisons are promising.
LeMone and Grossman refined the late-morning heat and moisture budgets for two contrasting days in CASES-97. Early budgets suggested that direct heating of the PBL by solar radiation was significant on one of the days. Net-radiation data from two aircraft are being supplemented with radiative flux divergences computed from radiosonde data by Kuo-Nan Liou (University of California, Los Angeles) and aerosol impacts based on a simple model of Stuart McKeen (NOAA Aeronomy Laboratory).
NOCTURNAL BOUNDARY LAYER (NBL)
Nocturnal boundary layer data from CASES-97 suggests a simple behavior for the temperature at 2 m over mildly undulating terrain. In an analysis of half-hour-averaged data, LeMone found that the 2-m temperature varied linearly with elevation. The temperature increase with elevation is a fraction of the static stability measured by radiosondes; if the wind is strong or the static stability weak, the potential temperature is constant. This suggests that the air flows over, rather than around, the terrain. Weak downslope flows are observed in strong static stability and weak winds. In weaker static stability or strong winds, wind speed increases linearly with station elevation.
MEASUREMENTS OF THE NBL, MORNING, AND EVENING TRANSITIONS IN CASES-99
During October 1999, Jielun Sun, Sean Burns, Lenschow, and LeMone participated in CASES-99, located in SE Kansas at a relatively flat 3 x 5 km site. The main objectives were to document and understand (i) Kelvin-Helmholtz instabilities, internal gravity waves and turbulence events, and their role in transport; (ii) creation of and transport by drainage currents; (iii) departure from surface-layer similarity theory for weakly stable (0.25<Ri<1) and very stable (Ri >>1) conditions; and (iv) evening and morning transitions. The centerpiece was a 55-meter tower (wind, humidity, temperature, radiative flux divergence and turbulence flux measurements), surrounded by three lidars, a Frequency Modulated Continuous Wave (FMCW) radar, a turbulence eddy profiler, a tethered balloon, and a tethered balloon/kite. Providing the meteorological context were the Atmospheric Boundary Layer Experiment's (ABLE) 915-MHz/Doppler SODAR wind profiling systems, two ISS 915-MHz profiling systems, and several surface towers, all concentrated near the main tower. The University of Wyoming King Air and the NOAA Long EZ aircraft flew patterns, and radiosondes were released frequently from one to four locations during nocturnal intensive observation periods (IOPs). The 11 IOPs sampled a variety of wind/stratification conditions, promising a rich dataset for analysis and modeling studies. Information on CASES-99 is on the WWW (see http://www.colorado-research.com/cases/CASES-99.html).
INTERNAL GRAVITY WAVES AND TURBULENT TRANSPORT IN NOCTURNAL BOUNDARY LAYERS IN A FOREST
Jielun Sun, Niels Otto Jensen (Riso National Laboratory, Denmark) and Mahrt analyzed nighttime data from the EUROFLUX tower located in a Denmark beech forest. These data frequently exhibit intermittent turbulence at night, in terms of temperature fluctuations possibly caused by gravity waves. In the early morning of 11 June 1996, when the air within and above the canopy was still stably stratified, the flow over the beech forest showed more than a dozen oscillations before yielding to significant turbulent mixing. The oscillations triggered turbulent mixing immediately above the top of the canopy, where the Richardson number was smallest. In this case, the internal gravity waves were likely generated by shear instability associated with an inflection point above the forest. This strong turbulent mixing led to large carbon dioxide and water vapor transport, which may be an example of the "chimney effect" proposed in previous work to account for "missing carbon dioxide."
ATMOSPHERIC RESPONSE TO SPATIAL VARIATION OF SOIL MOISTURE
Jielun Sun, jointly with Mahrt, Ian MacPherson (National Research Council, Canada), Ron Dobosy (NOAA/Atmospheric Turbulence and Diffusion Division), James Famiglietti (University of Texas at Austin), Thomas Jackson, and William Kustas (both USDA Agricultural Research Service Hydrology Laboratory, MD), used the ARM Southern Great Plains (SGP) data to study the atmospheric response to the spatial variation of soil moisture. Preliminary data analysis indicates that the spatial variation of the surface fluxes is very small when the ground surface is wet. As the soil moisture dries, the sensible heat flux responds to the spatial variation of the soil moisture much faster than the moisture flux. As the convective boundary layer develops during the day, large turbulent eddies tend to eliminate the spatial variation of the surface fluxes by strong horizontal mixing.
HYDROLOGIC COUPLING BETWEEN LAND-SURFACE AND THE CLOUD-FREE PBL
Peter Sullivan, Edward Patton (visitor, University of Minnesota), and Moeng continued their investigations of hydrological coupling between land surfaces and the cloud-free PBL using a coupled land-surface large-eddy simulation (LES) model. High resolution is used to simultaneously capture fine-scale turbulence and the PBL's response to kilometer-scale heterogeneous forcing. The research objectives are (i) prediction of surface temperature and moisture, their fluxes, and the land-surface atmosphere moisture budget at scales varying from less than 50 meters to more than 30 kilometers; (ii) use of remote sensing and in situ observations as initial conditions and as validation for fully coupled numerical simulations of typical field campaigns; (iii) elucidation of the temporal and spatial scales at which land surface heterogeneity influences the surface fluxes of heat and moisture, boundary layer depth, and turbulent motions; (iv) exploration of the effect of soil-vegetation properties on the diurnal cycle of the PBL; and (v) development of better parameterizations for surface fluxes and boundary layer depth in the presence of heterogeneous land surfaces for use in larger scale models and remote sensing.
FORM DRAG IN FLOW OVER PERIODIC TOPOGRAPHY
Wendell Welch (Yale University), Piotr Smolarkiewicz, Richard Rotunno, and Byron Boville (CGD) accomplished their study of stratified flows past 2D sinusoidal topography using simulations from a nonhydrostatic anelastic numerical model. Analytic formulae, intended for application with GCMs, were derived for the amount of form drag as a function of mountain height, mountain horizontal lengthscale, flow velocity, and Brunt-Vaisala frequency. Predictions from such formulae were shown to agree well with model results over a wide range of parameter space. The flow is divided into two regimes: a "linear" regime for small mountain heights, and a "blocked" regime for taller mountains. In the first, the form drag is equal to that calculated from linear theory. In the second, a layer of stagnant flow exists in the valleys and the form drag does not match the linear prediction. The cut-off between these two regimes is argued theoretically, varying only slightly with horizontal-mountain lengthscale.
d. Ocean-Atmosphere Coupling at Small Scales
Exchange of heat and momentum, as well as water and trace constituents, between the oceans and the atmosphere are key aspects of the climate system. For example, tropical atmosphere is constantly destabilized by radiative processes. It is brought back to the equilibrium by convective heating associated with phase changes of water substance.
However, the sensible and latent energy for the moist convection ultimately comes from the ocean surface. Oceanic circulations, on the other hand, are driven mostly by stress due to surface winds. As a result, the ocean-atmosphere coupling is of critical importance. Because the coupling involves small-scale and micro-scale processes occurring in coupled oceanic and atmospheric boundary layers, a need for physically based parameterizations of these processes in large-scale and climate models is apparent.
NUMERICAL SIMULATION OF TURBULENT FLOW OVER A WAVY SURFACE
Sullivan, James McWilliams (University of California, Los Angeles), and Moeng continued to investigate air-sea interaction using numerical simulations. Recently, they examined turbulent flow over idealized water waves with varying waveslope ak and wave age c/u* using direct numerical simulations at a bulk Reynolds number Re=8000. The imposed waves significantly influence the mean flow, vertical momentum fluxes [See Figure], velocity variances, pressure, and form stress [See Figure]. Compared to a stationary wave, slow (fast) moving waves increase (decrease) the form stress. At small c/u*, waves effectively increase surface roughness zo resulting in mean vertical velocity profiles with shorter buffer and longer logarithmic regions [See Figure]. With increasing wave age, zo decreases so that the wavy lower surface is nearly as smooth as a flat lower boundary. Vertical profiles of turbulence statistics show that the wave effects depend on wave age and waveslope but are confined to a region kz < 1 (where k is the wavenumber of the surface undulation and z is the vertical coordinate). The turbulent momentum flux can be altered by as much as 40% by the waves. A region of closed streamlines (cat's-eye pattern) centered about the critical layer height was found to be dynamically important at low to moderate values of c/u* [See Figure].
RELATIONSHIP BETWEEN ATMOSPHERIC MOMENTUM TRANSPORT AND OCEAN WAVES
Jielun Sun, Timothy Crawford and Jerry Crescenti (both NOAA Air Resources Laboratory, ID), Edward Dumas and Chris Vogel (NOAA/ATDD, TN), Douglas Vandemark (NASA/Goddard Space Flight Center), Pierre Mourad (University of Washington), and Mahrt investigated air-sea momentum transfer off the coast of Duck, North Carolina, using atmospheric observations over land and the NOAA LongEZ aircraft equipped to measure atmospheric turbulence and ocean-surface roughness. For the first time, both atmospheric and oceanic measurements were obtained as functions of offshore distances. With on-shore flow, the spatial variation of the stress in the coastal zone is small, and is influenced partly by the spatial variation of the atmospheric stability and partly by the sea surface roughness caused by shoaling. The roughness measured by the downward looking Ka-band scatterometer increases with increasing friction velocity and decreasing wave age. With offshore flow, the observed momentum flux significantly decreases with offshore distance, independent of the atmospheric stability, and the sea surface roughness observed by the on-board laser altimeters and the radar. The momentum flux over the coastal zone is strongly influenced by the advection of large momentum flux from the upstream land surface.
ONE-DIMENSIONAL OCEAN MODEL DRIVEN BY Numerically SIMULATED CLOUD-SCALE SURFACE PROPERTIES
The accurate surface energy balance obtained by cloud-resolving modeling (see section 2., E.) motivated Wu and Moncrieff to investigate how cloud-scale processes affect the upper ocean. They forced the one-dimensional upper-ocean model (William Large, CGD) by surface quantities obtained from the cloud-resolving studies of the Tropical Ocean and Global Atmosphere Program (TOGA) Coupled Ocean-Atmosphere Response Experiment (COARE) which were verified to within the accuracy of observational measurements (10 W/m2). They showed that the SST derived from the ocean model [dashed line in figure] agreed very well with observations [solid line in same figure]. On the other hand, a relatively poor prediction of SST was obtained when the ocean model was forced by surface conditions obtained from the single-column model [dotted line in same figure]. This shows that accurate atmosphere-ocean coupling requires accurate representation of mesoscale and cloud-scale processes.
The wealth of explicit synthetic results from cloud-resolving models can improve single-column models, which are basically the convective parameterizations applied in large-scale models. However, new approaches are anticipated to be needed in next generation large-scale models with ~10 km-grids where mesoscale circulations can be crudely resolved. More realistic ways to trigger convective parameterizations are needed and convective momentum transport has recently emerged as a key process. Eventually, the critical issues will focus on increasingly smaller-scale processes (turbulence, cloud microphysics, fine-scale cloud structures). This suite of problems calls for basic studies as well as improvements to existing parameterizations.
CLOUD-RESOLVING CONVECTION PARAMETERIZATION (CRCP) AND LARGE-SCALE TROPICAL DYNAMICS
Grabowski tested the CRPC approach in the context of equatorial beta-plane dynamics and global dynamics using previously reported anelastic models. CRCP represents convection by a two-dimensional cloud-resolving model in a state of radiative-convective equilibrium on a rotating aquaplanet with constant SST. Preliminary global calculations, performed at very low spatial resolution, demonstrate a spontaneous organization of tropical convection into eastward-propagating superclusters [See Figure]. The attendant large-scale disturbances resemble equatorially trapped Rossby/Rossby-gravity waves but their propagation speeds differ from classical shallow water dry analogs. This suggests that a key aspect is the coupling between deep convection and equatorial large-scale wave dynamics (explicitly captured by CRCP). These highly idealized calculations are now being extended to more realistic meridional SST distributions and higher spatial resolution.
PHYSICAL PROCESSES IN EXPLICIT AND PARAMETERIZED REALIZATIONS OF TROPICAL CLOUD SYSTEMS
Liu, Moncrieff, and Grabowski used their aforementioned hierarchical approach in a study of cloud systems observed during 19-26 December 1992 in TOGA COARE. A two-dimensional cloud-resolving (2-km resolution) simulation and observations were used to evaluate a coarse-resolution (15-km resolution) simulation that incorporates the Kain-Fritsch parameterization in a 900-km domain. With minor modifications, this mid-latitude based scheme treats tropical-precipitating convection reasonably well. Specific deficiencies are: (i) too much cirrus is generated through excessive detrainent of hydrometeors and water vapor; (ii) the cloud-top parameterized overshoot cooling is too strong, causing a cold bias near the tropopause; and (iii) the lower-level moisture is overpredicted. These discrepancies are due to the updraft mass flux being defined by a single entraining plume model, which cannot properly represent the cumulus congestus seen to occur in both cloud-resolving simulations and observations. This is a pilot study to be developed as a contribution to the Tropical Rainfall Measurement Mission (TRMM).
ATMOSPHERIC ENERGY BALANCE OBTAINED FROM EXPLICIT SIMULATIONS
Most general circulation models (GCMs) and coupled atmosphere-ocean GCMs are unable to get the energy budgets at the top of the atmosphere and the surface to agree with observations at the same time. Considerable efforts have been made to understand the cause of bias in the surface energy budget and to alleviate the climate drift problem in coupled atmosphere-ocean GCMs. Wu and Moncrieff investigated this problem from the perspective of a cloud-resolving model (CRM, which resolves cloud-scale dynamics), a single-column model of the NCAR Community Climate Model version 3 (CCM3) (SCM, which parameterizes convection and clouds), and observations during TOGA COARE. It is demonstrated that the 30-day average of numerically simulated top-of-atmosphere radiative fluxes and surface energy budgets simultaneously agree with observations to well within the 10 w/m2 observational measurement accuracy. This is in marked contrast to atmospheric general circulation models and coupled atmosphere-ocean models which have great difficulty in achieving a correct energy balance. The comparison among the CRM, SCM and observations indicated that the critical factor is the accurate realization of cloud condensate and its vertical and horizontal distributions (See Figure]. This can be achieved only if clouds are correctly realized as the large-scale conditions evolve.
USING CLOUD-RESOLVING MODEL RESULTS TO IMPROVE CUMULUS PARAMETERIZATION
Wu, Moncrieff, and James Hack (CGD) began to use results obtained by cloud-resolving simulations of tropical cloud systems forced by, and verified against, TOGA COARE observations. They used 39-day simulations of a period (December 1992-January 1993) identified by cloud systems ranging in type from shallow convection, through weakly organized convection, to cloud clusters and superclusters in a westerly wind burst. The methodology is to compare explicit results with those from the CCM3 single- column model run with exactly the same time-varying large-scale forcing.
CONTRIBUTIONS TO THE GEWEX CLOUD SYSTEM STUDY (GCSS)
Moeng continued to contribute to the GCSS Working Group (WG) 1 intercomparison study of different LESs. This year the focus is on shallow cumulus over land. The case is based on an idealization of observations made at the Southern Great Plains ARM site on 21 June 1997. The results generated from different LES groups will be compared during the next GCSS-WG1 workshop, to take place in January 2000 at NCAR.
Grabowski, Moncrieff, and Wu continued to contribute to GCSS Working Group 4, which just completed its series of TOGA COARE-based intercomparisons. They are now taking part in the next intercomparison study of deep convection over land. This study uses data sets for forcing and verification from an intensive observation period in summer of 1997 over the ARM SGP site.
GLOBAL IMPACT OF CONVECTIVE MOMENTUM TRANSPORT
Wu, Moncrieff, and Xin-Zhong Liang (Illinois State Water Survey, University of Illinois at Urbana-Champaign) performed a sensitivity study to estimate the impact of parameterized convective momentum flux on the large-scale circulation of the atmosphere using the NCAR CCM3. The parameterization included the cloud-scale horizontal pressure gradient in a mass-flux-based representation of downgradient momentum flux. Preliminary results show that the Hadley circulation was less intense but the Walker circulation was enhanced. Precipitation patterns at large scale were more realistic, for example, in the eastern Pacific. This was demonstrated by comparing the CCM3 results to direct observation and satellite-derived observations.
PARAMETERIZATION OF CONVECTIVE MOMENTUM TRANSPORT IN BAROLINIC CONDITIONS
Moncrieff and Vanda Grubisic (Desert Research Institute) completed their investigation of three-dimensional, open-cellular convection in an idealized cold-air outbreak. They quantified the effects of the in-cloud pressure field, organization and shear on momentum transport. A simple analytic model of three-dimensional blocked convective overturning gives dynamical insight into the momentum transport properties of the numerically simulated cloud ensemble. An overturning circulation having vorticity of the same sign as the ambient shear maintains the far-field momentum while the blocking effect decelerates in-cloud momentum. The overall effect is to maintain the mean flow below the environment value (i.e., down-gradient transport), in contrast to the sometimes up-gradient transport associated with two-dimensional squall lines. The Kershaw-Gregory parameterization of convective momentum transport used in the UK Meteorological Office operational model is shown to be valid for convection in cold-air outbreaks. This scheme is anticipated to be applicable to cumulonimbus convection in unidirectional shear but is not necessarily universal.
EFFECTS OF SHEAR ON HEAT AND MOISTURE SOURCES ASSOCIATED WITH ORGANIZED CONVECTION
As a contribution to TRMM, Liu and Moncrieff started to quantify the influence of environmental shear on convective heating/moistening by means of idealized cloud-resolving, two-dimensional simulations that apply the same large-scale temperature and moisture forcing but different ambient wind profiles. Preliminary analyses show that the ambient shear has significant impacts on the diabatic heating and moistening and on cloud fraction.
ROLE OF SHEAR IN TRIGGERING CONVECTION IN PARAMETERIZATION SCHEMES
Liu and Moncrieff examined the role of shear in triggering convection using a two-dimensional model initialized in horizontally uniform conditions perturbed by a density current. The main findings are as follows: (i) strong convection develops preferentially at the downshear side (upstream in unsheared flow); (ii) upper-level flow and shear are important to convective initiation even though they have little impact on density current dynamics; and (iii) the strength of the low-level convergence by density currents cannot solely determine convective initiation because the organizing influence of ambient flow and shear plays a key role. These results point the way to more realistic "trigger functions" for cumulus parameterization schemes that are linked to downdraft mass fluxes.
STATISTICAL MODELS OF CONVECTION
Moncrieff, Philippe Naveau and Doug Nychka (both CGD/GSP) began to formulate stochastic models of numerically simulated cloud systems with the purpose of understanding basic issues (e.g., how to treat small-scale variability when its distribution is non-random). The long-term objective is to provide new concepts for convective parameterization. One approach uses the statistical theory of extreme values to analyze the maximum vertical velocity in temporally non-stationary system as a combination of two generalized extreme distributions. Preliminary results show the statistical model defines a selection principle for squall-lines, non-squall clusters, and scattered convection. The statistical models are being developed in association with idealized dynamical models of convective regimes.
a. Ice Microphysics Research
The purpose of the Ice Microphysics Research Program is to highlight work on the processes of primary and secondary formation of ice in the atmosphere. Increasingly, progress in other research areas such as climate and weather is being hindered because of limited knowledge of these basic processes. The Ice Microphysics was started to try to promote additional efforts in MMM and the community in these areas Initiative (see http://www.mmm.ucar.edu/mmm/stratplan.html).
Colloquium on Ice Formation in the Atmosphere
Charles Knight and William Cooper (joint appointment with ASP) organized the annual summer ASP colloquium, which this year was on Ice Formation in the Atmosphere. The colloquium was held 14 - 25 June and attended by about 20 graduate students and recent Ph.Ds from the U.S. and abroad with 23 lecturers from 11 universities, NCAR, NOAA and NASA. The colloquium centered on the microphysical processes leading to the formation of ice crystals and ice phase precipitation in clouds including ice physics, nucleation, the roles of aerosols, types of ice clouds, airborne and remote sensing instruments, ice multiplication, hail formation, weather modification, and the links between ice formation and climate, chemical, radiative and electrical properties of clouds. In addition to providing broad based lectures to the participants, the colloquium gave a forum for the speakers and students to interact in assessing the current state of knowledge on the microphysics of ice in clouds with the hope of reinvigorating work in this area of research. The lectures and discussion are being used in the preparation of a review of the status of the field of ice microphysics.
James Dye (joint appointment with ATD) and Andrew Heymsfield participated in the collection and analysis of cloud microphysical data obtained with the University of North Dakota (UND) Citation II in the NASA Tropical Rain Measuring Missions (TRMM) in Brazil and Kwajalein, Marshall Islands. The goals of these campaigns included providing in situ data to evaluate the performance of the TRMM retrievals of rainfall rate from the TRMM satellite's precipitation radar and radiometers and to provide a data set for comparison and incorporation into mesoscale and regional scale models. They worked closely with Jeffrey Stith (now with NCAR/ATD), Cedric Anthony Grainger, and Martin Brown (all University of North Dakota).
For the first time, size distributions of ice particles and raindrops were obtained with high resolution images over the size range from 15 microns to several cm, using newly-developed probes including the Stratton Park Engineering Corporation (SPEC), Inc., Cloud Particle Imager (CPI) and High Volume Particle Sampler (HVPS). Cloud sampling was conducted in the stratiform regions of MCSs over the temperature range +10 to -50 C. In several cases the aircraft spiraled down from cloud top, through the melting level and into the rain region below. Such flight patterns revealed upper levels were characterized by a predominance of single crystals but temperatures of -15 to 0 C ice particle aggregates were often dominant. Aggregates as large as 2.5 cm were observed near the melting zone; the larger particles maintained an ice identity until they fell to the +4 C level [See Figure].
The observations also show, not surprisingly, that in both Brazil and in Kwajalein the first precipitation forms from the coalescence process at temperatures much warmer than 0 C, but in moderate updrafts drops are carried to colder temperatures and become super-cooled. The images from the new CPI probe clearly show that the appearance of the first ice development is through the freezing of these drops and in both locations frozen drops first start to appear at -8 to -10C. These freezing drops then provide an effective means of further development of ice in the cloud. Vaughan Philips (University of Manchester Institute of Science and Technology) worked with Dye in Brazil on these results.
Over the last couple of decades the Particle Measuring Systems 2D imaging probes (the Cloud probe with size resolution of about 25 microns and the Precip probe with about 200 micron resolution) have been the standard instruments used for observing precipitation particles, especially ice, in clouds. During the past couple of years two new instruments that greatly expand our measurement capability were developed SPEC. These instruments are the Cloud Particle Imager (CPI) which produces spectacular images of particles with about 2.5 micron pixel resolution for particles up to a couple of millimeters in size and the High Volume Particle Sampler (HVPS) which has about 8 times the sample volume of the 2D-precip probe and can image particles up to almost 5 cm in size with 200-micron resolution. All these probes collect large volumes of data that need to be processed and analyzed. While vendor analysis programs and a few open source programs exist that can be used to analyze some of the data from individual probes, there are no common analysis programs that can treat and combine the data from each probe for analysis. William Hall, with Heymsfield and Dye, and Stith and Christopher Webster (ATD/RAF), are working toward developing a software base which can be used within NCAR and by the research community to combine and provide analysis products for the 2D, CPI, and HVPS probes. Hall now has an understanding of the individual codes and with Webster are working toward providing a comprehensive open source code, using the available open source analysis codes for the 2D and HVPS and pre-processed data from SPEC software for the CPI.
Modeled Radiative Properties of Tropical Anvils
Using data acquired during the Central Equatorial Pacific Experiment (CEPEX), McFarquhar and Heymsfield, in collaboration with Andreas Macke (Kiel University) and Jean Iaquinta (Ussel, France) used ice crystal sizes and derived shapes to calculate mean single-scattering properties and multi-spectral radiances using Monte Carlo radiative transfer routines. This extends previous work on this topic by using realistic geometeries of ice crystals for calculating the scattering properties. Although substantial temporal evolution in the sampled outflow anvil and uncertainties in cloud base prevented an exact comparison of the modeled radiances with those measured by radiometers above the cloud system, they performed sensitivity studies that showed that the calculated radiances were as sensitive to the assumed numbers and shapes of crystals smaller than 100 ¼m as to those of larger crystals. This shows that measurements of small crystals are needed to adequately determine multi-spectral reflectances.
Extensions to this work are underway in order to develop more realistic models of ice polycrystals or aggregates of ice crystals for the calculation of radiative properties. This should result in better parameterizations of albedo, extinction coefficient, and asymmetry parameter for use in climate models.
Improvement of upper-tropospheric relative humidity measurements
Assessing the humidity conditions that lead to formation of ice particles in cirrus clouds and aircraft contrails is very important for climate applications, yet radiosonde relative humidity (RH) measurements are widely known to be unreliable at the cold temperatures of the upper troposphere. Larry Miloshevich and Heymsfield, in collaboration with Holger Voemel and Samuel Oltmans (NOAA/Climate Monitoring Diagnostics Monitoring) and Ari Paukkunen (Vaisala Corporation), developed a correction for RH measurements for Vaisala RS80-A radiosondes, the most frequently used radiosonde in the world.
The ratio of corresponding RH measurements from the RS80-A and the NOAA hygrometer, a reference-quality instrument, demonstrate in the figure that, on average, the RS80-A measurements decrease with decreasing temperature relative to the more accurate hygrometer. The reciprocal of the mean ratio at a given temperature [See Figure] is a statistical correction factor that, when multiplied by RS80-A measurements, gives corrected RH values that are on average equal to the hygrometer measurements. The correction factor accounts for the statistical average at a given temperature of all sources of RS80-A measurement error combined. Although the correction factor is quite accurate when applied statistically to a large dataset (such as when constructing a climatology), the uncertainty is considerably larger when the correction is applied to any individual profile, as indicated by the dispersion in the data at a given temperature in Fig. 1A. The dispersion results from RS80-A measurement errors that are not simply temperature-dependent. The individual sources of RS80-A measurement error were identified and characterized, and work is in progress on an approach that corrects the individual measurement errors, promising greater accuracy for correcting individual RS80-A profiles.
Knight, in collaboration with Minoru Maruyama (Osaka University), showed that ice growing from super-cooled, air-free water does not grow with the development of prism facets and is similar to growth when air is dissolved in the water. The result refuted a claim in an earlier paper that reported that growth in air-free water did develop prism faces.
Knight, with Kevin Rider and E. Dendy Sloan (Colorado School of Mines), has nearly completed a study of the crystallization of a structure II clathrate hydrate: tetrahydrofuran hydrate, in free growth from the melt, with and without growth inhibitors and other impurities, as a function of super-cooling and of the mode of nucleation. The most interesting finding is that in many regimes -- especially at low values of super-cooling -- the growth rate is double-valued. The crystals (with cubic structure) grow either slowly with octahedral habit or much faster with plate or needle habits, depending upon how the growth is initiated. This behavior is explainable when certain impurities are present, but it also occurs in the nominally pure material, which has been hard to explain.
Wild fires are a major global problem burning 10's of millions of hectars per year causing major resource and property loss as well as human casualties. How fires propagate and interact with different atmospheric conditions, and what their orography and fuel characteristics are, are all poorly understood. A better understanding of fire behavior is extremely important to agencies around the world involved with fire management. NCAR's role in this arena is to advance our understanding of fire behavior through coupled fire-atmosphere modeling as well as through remote observations using high speed IR imaging among other sensors. This combination of observations and modeling is proving to be a powerful approach.
In July, Terry Clark and Janice Coen, along with Larry Radke (ATD), participated in FROSTFIRE (the role of fire in permafrost terrain in the boreal forest), a prescribed burn in the boreal forest outside Fairbanks, Alaska. From their positions aboard a U.S. Forest Service Piper Navajo with Robert Higgins (NASA/Ames Research Center), Clark and Radke collected airborne infrared (IR) video camera data [See Figure] of running fires [See Photo] on the slopes below. Coen, with Shankar Mahalingam, John Daily, and Yottana Khunatorn (all University of Colorado), collected IR video camera data, as well as spectral data in other ranges of these fires from a mountaintop site across the valley. Clark and Coen worked with Donald Middleton and Tim Scheitlin (both SCD) to digitize this data at 1/60th of a second, to provide a dramatic look through smoke at the dynamics of racing crown fires.
In FY 99, Clark and Coen, along with Donald Latham (U.S. Forest Service, Missoula Fire Laboratory), improved NCAR's coupled atmosphere-fire model through implementation and testing of BEHAVE, an empirically based fire spread algorithm widely used in the fire community, and BURNUP, an algorithm that describes how a mixed fuel is consumed over time. Using a new generation of the coupled atmosphere-fire model that Clark and Coen completed this year, they examined the model's sensitivity to fuel types and distributions, which is another step closer to simulations of actual fires. The point wise treatment of fuel characteristics (the tracer method) was both simplified and improved to the point of demonstrating an acceptable level of stability.
Clark, in collaboration with Radke and Craig Walther (both ATD), continued analysis of the Wildfire Experiment (WiFE) data. Algorithm development continued on the 1 and 3 mm Airborne Imaging Microwave Radiometer (AMIR) channel data. Digit roll, pitch correction, and polarization differences were successfully applied to allow mapping of fuel mass distributions using differences in horizontal and vertical polarization brightness temperatures. Work on the infrared image registration proved successful enough to next use the infrared (IR) images to extract fire winds using image flow analysis, a step that will be addressed next year.
c. Geophysical Turbulence Research
Geophysical turbulence remains one of the most challenging problems in the fields of climate and weather research. The main goal of the MMM GTP program is to gain fundamental understanding of the turbulent motions so that their effects can be properly represented in forecast models of climate and weather.
JET STREAM SHEAR ZONES
Clark, in collaboration with Robert Sharman (RAP), Melvin Shapiro (visitor, NOAA) and Radke, continued FAA-funded work on Clear-Air Turbulence (CAT) studies. Clark and Sharman concentrated on an eastern continental U. S. CAT case occurring between 11-13 December 1997 that was jet stream-related. The fine-scale model results showed that the shear zone on the cyclonic side of the jet stream easily accounted for the extended event. Subsequent analysis of large scale NMC model output was shown to prove adequate in providing predictors for the air traffic community. In the future they will redirect attention to the North Pacific Experiment (NORPEX) cases using data collected during January of 1999.
Measuring SGS motions in the atmospheric surface layer
There is an increasing demand on Large Eddy Simulation (LES) to address complicated PBL turbulence problems such as: (i) diffusion and chemical reactions in stable boundary layers where turbulence is intermittent, (ii) interactions between the boundary layer and the Earth's surface (over ocean and land), and (iii) entrainment across strongly-stratified capping inversions. All these problems deal with turbulence where small-scale turbulent motions become important or even dominant. To adequately deal with these problems with LES requires further understanding and improvement of parameterizations of subgrid-scale (SGS) motions in LES. For these reasons, Chin-Hoh Moeng, Donald Lenschow (joint appointment with ATD), and Peter Sullivan worked with Thomas Horst and Steven Oncley (both ATD) in proposing a field experiment to measure small (SGS) and large (filtered scale) turbulent eddies separately and to develop better SGS models for LES. They propose to use and extend the sonic anemometer array technique developed by Tong et al. (1999) to measure these fields at two height levels, to investigate the SGS response to stability effects, and also to test two specific existing SGS models.
Large eddy Simulation of the stably-stratified PBL
Eileen Saiki (ASP), Moeng, and Sullivan studied the characteristics of a stably stratified atmospheric boundary layer using LES. Flow visualization and evaluation of the turbulence statistics from this case reveal the development of a continuously turbulent boundary layer with small-scale structures. The stability of the boundary layer, coupled with the presence of a strong capping inversion, resulted in the development of a dominant gravity wave at the top of the stable boundary layer which appeared to be related to the most unstable wave predicted by the Taylor-Goldstein equation. As a result of the decay of turbulence aloft, a strong low level jet formed above the boundary layer, with a feature that agrees with the inertial oscillation analysis of Blackadar.
Lagrangian Modeling of Dispersion in the CBL Using LES Fields
In collaboration with Moeng and Sullivan, Jeffrey Weil (visitor, Cooperative Institute for Research in Environmental Sciences) used velocity fields from LES to model scalar dispersion in the convective boundary layer (CBL). The LES fields are coupled with a Lagrangian "particle" model for a passive scalar. Earlier calculations of the mean concentration field downstream of elevated point sources showed good agreement with the laboratory experiments of Willis and Deardorff (1976, 1978). The simulated CBL corresponded to a highly convective situation as defined by the ratio of the CBL height to the Monin-Obukhov length. Most recently, they simulated dispersion in weak convection and found the modeled concentration fields to be qualitatively similar to the earlier ones; however, the dispersion was noticeably reduced with lower peak surface concentrations. This was attributed to the greater surface shear and turbulence dissipation rate in the surface layer and immediately above it, which led to smaller turbulence time and length scales.
In addition, they simulated dispersion from a surface release in the CBL. This was a more demanding test of the Lagrangian model because of the greater reliance on the subgrid-scale dispersion, which is treated stochastically. Preliminary results showed qualitative agreement between the predicted mean plume height and results from surface-layer similarity theory, which were field-tested. Although further developments are necessary, this is a significant first step in the modeling of surface source dispersion. The Lagrangian model simulations serve an important niche for the dispersion modeling community because there are currently no benchmark experiments, simulations, or observations showing the variation of plume dispersive properties over a broad range of stability and source height.
Wave-vortex interaction in large scale stratified rotating fluids: shallow water (RSW) and Boussinesq; Long term effects and turbulence; geostrophic balance, long (weather) cycles, nonlinear response
RSW viewed as a coupled system of wave and vortical modes exhibits chaotic patterns of nonlinear oscillations, with multiple time scales, ranging from hours (gravity waves), to days (eddy turnover) to annual/biannual; and longer weather cycles. As a first step, David Gurarie (visitor, Case Western Reserve University) developed novel averaging and renormalization methods to produce the effective mean field theory of RSW on slow time scales. Gurarie's analysis revealed the principal effect of wave-vortex interaction to be modulated quasigeostrophy, which slowly drives the basic parameters of oscillation. Gurarie's theory predicts new phases, patterns and cycles of nonlinear oscillations, that were confirmed by numerical simulations and analysis of the truncated (Lorenz-type) RSW-systems.
Three plots demonstrate long time series (2500 days) of three vortical modes
and three pairs of associated wave-amplitudes for the RSW triad with small Rossby/Froude
number 
=.02.
The first plot shows the complete systems with all scales resolved up to smallest
(hour) scale of gravity waves [See Figure];
the second plot shows the first renormalized system on time scale 
,
with small scale oscillations averaged out [See Figure];
the third plot shows the result of second renormalization, where solutions are
constructed by analytic formulae, based on slow adiabatic invariants of the
system evolving on time scale 
[See Figure].
Work in progress with Sergey Danilov (Russian Academy of Sciences) aims to extend these results to large (multi-mode) systems, to include forcing-dissipation, to explore the effect of modulation on QGS turbulence, to study modulation in multilayer and continuously stratified fluids, to explore nonlinear response to forcing cycles, and to explore possible mechanisms of quasi-cyclicity (QBO).
Turbulent transport and diffusion in stratified flows
Gurarie and Valery Klyatskin (Institute of Atmospheric Physics, Moscow) recently developed new ideas in turbulent transport, as an alternative to the standard "mean field" theories. This approach focuses on the dynamics and "statistical geometry/topography" of iso-levels, their pdf’s, and Fokker-Planck evolution. Several models were studied, in particular passive tracers advected by surface waves, and compressible fluids. They showed new non-diffusive features of turbulent transport, like clustering of tracer concentration, and found its quantitative description. The current goal is to extend these methods and results to stratified fluids, with Jackson Herring and Yoshi Kimura (University of Nagoya, Japan).
d. Chemistry in Clouds Research
The focus of the Chemistry in Clouds Program is to examine gaseous and aqueous chemistry, chemical species transport and dynamics affecting transport and chemistry at the cloud scale. Ongoing projects within the program include observational and numerical analyses of the Stratosphere-Troposphere Experiments: Radiation, Aerosols and Ozone (STERAO)-Deep Convection experiment datasets, Large Eddy Simulation (LES) modeling studies of the PBL, and a model-based aerosol transport study.
The STERAO-Deep Convection experiment, which was conducted during the summer of 1996, had the major goals of investigating NOx production by lightning and transport of chemical constituents by thunderstorms. For more information on STERAO-Deep Convection field project and earlier results refer to the 1997 and 1998 Annual Scientific Report. Work in the Chemistry in Clouds Program associated with the STERAO experiment are focused on analyses of the observations and on numerical modeling efforts to augment the observational analyses.
Dye and collaborators from MMM, NCAR/ACD, NOAA NSSL, NOAA Aeronomy Lab, and ONERA (France) continued their analysis and interpretation of the multi-faceted airborne in situ chemical, airborne and surface based Doppler radar, and total lightning (intra-cloud as well as cloud-to-ground) measurements. In the past year, the analyses concentrated on the 10 July 1996 storm. The major results from these observational analyses are:
(1) Comparison of NO, CO and ozone measurements made in the inflow region of the storm with measurements made in the anvil suggest that 50 to 90 percent of the enhanced NO observed in the anvil of the 10 July storm was produced by lightning.
(2) The observed enhancements in NO from lightning had to have been almost exclusively from intra-cloud (IC) lightning. There were only a few cloud-to-ground (CG) flashes during the period of aircraft observations and the IC/CG flash ratio was in excess of 100:1 over the period for which there are observations of NO.
(3) The large variations in total lightning flash rates, flash durations, and flash lengths during the 5-hour lifetime of the storm.
(4) Abrupt increases in IC lightning flash rates, but decreases in CG flash rates, occurred during periods of rapid storm intensification for both the 10 and 12 July storms.
(5) Analysis of lightning measurements from the ONERA lightning interferometer shows that the cumulative length of lightning channels per minute in the 10 July storm vary from 500 to over 2000 km. This includes leader, recoil streamer, and return stroke channels.
William Skamarock, Mary Barth (joint appointment with ACD), and Dye simultaneously worked with simulations of the 10 July storm. An NOx (NO+NO2) budget of the storm is being produced by augmenting the observations with the simulations. For the 10 July storm, the North Dakota Citation aircraft traversed to anvil outflow at many levels. Cross sections of NO, CO, ozone and other species, along with their fluxes, have been objectively analyzed [See Figure].
Using the simulation to determine the convective flux of non-lightning-produced NO, estimates of NOx produced by lightning were constructed from the objective analyses (approximately 1021 molecules per meter flash channel length). These agree well with earlier estimates produce by Stith and Dye (between 1020 and 2 x 1022 molecules/m) using a completely different approach.
Analysis work was performed on the processes affecting the distribution of soluble tracers in COMMAS model simulations. Barth found that accretion of cloud water by snow and hail are important processes in determining the distribution of a soluble tracer. Barth found that a highly soluble tracer, similar to nitric acid, is depleted by 40 percent compared to a passive tracer in the anvil region of the storm. This depletion is much less than is expected from wet deposition parameterizations. Learning the reasons for this discrepancy is ongoing, but it is likely due to the assumption of Henry's Law equilibrium in the parameterizations.
Simulations with chemically reactive species were also performed. From these simulations, it was found that the aqueous chemistry did not substantially alter the total mixing ratio of a species such as hydrogen peroxide, but instead the role of ice, snow and hail in capturing the species reduced the mixing ratios of the species in the anvil region. Further analysis of these simulations is being pursued.
NSSL Contributions to STERAO Analysis
NSSL scientists Thomas Matejka and Diana Bartels collaborated with Dye on the analysis and interpretation of data collected during the field phase of the STERAO. Analysis of Doppler radar data for STERAO cases of 11-12 July 1996 revealed two convective events markedly different in airflow and cellular propagation. The former case exhibited almost no downdrafts in the analysis (consistent with all in situ and visual measurements); the latter case contained downdrafts of up to 4 m s**-1. These differences illustrate how differently specific cases may redistribute ambient and lightning-generated chemical species.
To better understand the evolution of the cases, the steady and evolving airflow components are being separated. To do this properly, the most stationary frame of reference of each mesoscale convective system must be known. Since the Doppler radar data for each analysis time are themselves collected over quite long times, assumptions about the most stationary frame must be made prior to the wind analysis. A new method was developed to deduce the most stationary frame of reference from Doppler radar data (airborne or surface-based) from one or more radars at three consecutive times. These objectively determined estimates of the most stationary frame of reference are significant improvements on those obtained from previous methods and should enable the airflows to be decomposed in an instructive way.
More information can be obtained at the NSSL/MRB accomplishments WWW site (see http://mrd3.nssl.ucar.edu/accomplishments/FY99/STERAO.html).
LES studies of the Planetary Boundary Layer
To understand the effect of boundary layer processes on the rate of chemical reactions, chemistry was coupled with LES. Edward Patton (visitor, University of Minnesota), Kenneth Davis (University of Minnesota), Sullivan, and Barth coupled a simple chemistry mechanism with a canopy-scale LES in which the plant canopy acts as a spatially distributed source of the chemical species, isoprene. The variances and fluxes of all the scalars peak near the canopy top. It was found that a species with a first-order decay rate of less than 15 minutes has a much smaller variance and vertical flux than a passive tracer. However, a simplified isoprene chemistry, with the loss rate dependent on the local hydroxyl radical mixing ratio, generates a similar variance and vertical flux as a passive tracer. This work is currently being documented with a detailed analysis of the processes affecting the variance and flux of each scalar.
Numerical experiments are also underway by Patton, Sullivan, Kenneth Davis, and Barth using Sullivan's nested-grid LES coupled with both the plant canopy and simple chemistry to investigate the influence of horizontally inhomogeneous chemical emissions and canopy storage on full PBL scale statistics of chemically active hydrocarbons. An example is isoprene that is emitted by a variety of plant species and has important effects on the ozone, carbon monoxide and hydroxyl radical concentrations in the VR2. Preliminary results show that segregation of a simple isoprene species from the hydroxyl radical reduces the isoprene's rate of reaction. Segregation increases the variance and vertical flux of isoprene species. The heterogeneous source distribution acts to increase a species' variance and to increase the segregation of reactive species.
Barth, Patton, and Kenneth Davis simulated the chemistry in the convective boundary layer using realistic hydrocarbon chemistry appropriate for rural conditions. They found that there is very little segregation (<5%) between reactive chemical species. This result is true for conditions for shear-driven flows (geostrophic winds of 5 m/s and 10 m/s) and flow with no geostrophic wind. Segregation between reactive species increased when the available NOx (NO + NO2) was decreased by an order of magnitude. Decreasing the NOx decreased the competition between NOx and the hydrocarbons for the OH radical, thereby linking the highly reactive OH radical and hydrocarbon concentrations more closely. Further analysis of the processes that affect the isoprene variance is underway.
Tropospheric aerosol studies
Barth and Andrew Church (visitor, SOARS, University of New Mexico) examined the transport and distribution of sulfate aerosols and industrially produced black carbon from two regions, southeast China and Mexico City. Using the global sulfur model they found that aerosols from southeast China were transported long distances, while those from Mexico City were found regionally. The distribution of black carbon was found to be sensitive to the prescribed transfer rate of hydrophobic black carbon to hydrophilic black carbon, which is susceptible to rainout. Future work with the global sulfur model includes describing ammonia and its interactions with the sulfate aerosol.
