MMM Core Program

A.  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 dynamics and predictability, the life cycle of mesoscale precipitating weather systems, 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.

1.  Mesoscale Dynamics and Predictability

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.

Dynamics

Understanding the relation between meso- and synoptic-scale flows involves, in part, understanding how atmospheric dynamics change as the Rossby number increases.  At small Rossby number, virtually all dynamical theories rest upon the foundation of quasigeostrophy (QG), which is the leading-order theory in Rossby number.  Muraki (Courant Inst., NYU), Chris Snyder and Richard Rotunno have 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, Snyder, and Muraki have applied QG+1 to idealized baroclinic waves; they have illustrated how familiar synoptic diagnostics such as Q-vectors fit within the QG+1 framework and have shown that the characteristic cyclonic bias in primitive-equation simulations of baroclinic waves arises from a next-order correction in the inversion of potential vorticity.  Muraki, Rotunno and Snyder consider flow over topography with QG+1 and are examining how gravity waves and fore-aft asymmetry enter the topographic flow problem at small Rossby number.  Finally, Greg Hakim (University of Washington), 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.

Gravity waves are another important ingredient of mesoscale flows.  In particular, the mechanisms that initiate and maintain observed high-amplitude gravity-wave events remain poorly understood.  Fuqing Zhang and Chris Davis, together with Steven Koch and Michael Kaplan (both of North Carolina State University), have performed detailed diagnostic analysis of a mesoscale numerical simulation of a well-observed gravity wave event that occurred on 4 January 1994 along the East Coast of the United States.  Both wavelet techniques (to extract the evolving structure of the wave) and potential vorticity inversion (to isolate the flow “imbalance”) were used.  This analysis indicated that a train of gravity waves was generated by geostrophic adjustment in the exit region of a strong upper-level jet streak as it approached the inflection point between the upper-level ridge and trough.  Subsequent to this initiation, the wavetrain underwent a complex interaction with both a mid-tropospheric split front and, later, elevated moist convection.  The convection appeared to coincide with a transfer of wave energy to levels beneath the frontal surface.  There, the wave was at least partially ducted, although a continued interaction with moist convection may also have helped to maintain the wave.

Clarifying processes of wake formation remains a key challenge to understanding the influence of mountains on the atmosphere at a range of scales.  On the mesoscale, wake and vortex flows in the vicinity of cities tend to recirculate pollutants and thus have consequences for regional air quality.  In the bigger picture, the improvement of orographic drag parameterizations to account for wake formation has important implications for weather and climate prediction models.  Previous approaches to wake formation have relied heavily on inferences drawn from either small-amplitude stratified theory or nonlinear shallow-water dynamics.  Craig Epifanio (ASP), Dale Durran (University of Washington), and Rotunno have worked to extend our understanding of mountain wakes by focusing on fully nonlinear aspects of wake formation in a stratified, laminar viscous model.  New aspects of this work include (a) the use of semi-analytic approaches to compute weakly nonlinear (second-order in disturbance amplitude) flow fields; and (b) the development of new methods for understanding and diagnosing vorticity production in fully nonlinear numerical models.  Their analysis suggests that in laminar viscous flows, vertical vorticity is produced over the lee slope of the mountain much as described by weakly nonlinear theory.  However, in the process of wake formation this vorticity is amplified several-fold through nonlinear stretching in a hydraulic jump-like feature, thus resulting in the pronounced vertical vorticity anomalies of the wake.  This result provides a new conceptual model to explain vertical vorticity production in mountain wakes. 

Predictability Tom Hamill (ASP), Snyder, and Rebecca Morss (ASP) have also used a quasi-geostrophic model, along with a three-dimensional variational assimilation (3DVAR) scheme, to explore the statistics of forecast and analysis errors.  They find 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 have shown (see Data Assimilation section) that an ensemble Kalman filter using a small sample can outperform 3DVAR. Zhang, Snyder and Rotunno have completed an evaluation of various uncertainties in the numerical prediction of the “surprise” snowstorm that paralyzed Washington, D.C. on 25 January 2000.  While the high-resolution MM5 forecast of this storm is notably successful at 36 h (see figure 1 at right), they find that reducing the horizontal resolution to 30 km, or using another, equally plausible initial analysis, can significantly degrade the precipitation forecast.  In both cases, the degradation of the forecast is intimately tied to moist processes, which result in the growth of forecast differences at horizontal scales of a few hundred to a few tens of kilometers.  They have also examined, and find incorrect, the notion that a poor fit of the operational analysis to one or two key soundings might have been the source of operational forecast errors. The above experiments have led Zhang et al. to consider more explicitly how initial errors of small scale and small amplitude can alter the subsequent forecast.  Much as was foreseen by Lorenz in 1969, they find that such initial errors grow rapidly (with doubling times as short as an hour), evolve nonlinearly and spread upscale.  The contamination is especially pronounced for the short-term (12 h) mesoscale precipitation forecast.  It is also clear, as noted above, that moist processes are central to this behavior, as small errors grow slowly and evolve almost linearly if latent heating is turned off.  In order to obtain the most faithful representation of the error growth, further simulations are underway, with higher resolution and without convective parameterization on the finest grid.  Idealized simulations are also planned with the goals of generalizing these results and understanding the mechanisms by which errors at the convective-scale influence large scales. Mel Shapiro (MMM/IPA visitor, NOAA/ Environmental Technology Laboratory) conducted research on the effects of the El Niño Southern Oscillation (ENSO) on extratropical baroclinic life cycles and results suggest that different time-mean flows between the El Niño and La Niña regimes give rise to significantly different extratropical cyclone life cycles (see figure 2 at right). Results also suggest that differing inter-annual and intra-seasonal time-mean flows impact upon the accuracy of numerical forecasts.  These results have been presented at the European Geophysical Society Annual Meeting (Nice, Fance; April 2000), the Extratropical Cyclone Workshop (Monterey, CA; Sept. 2000), and in seminars at NCAR, NOAA, and at US and international universities.

Figure 1 (click on graphic to view larger version): Left: MM5 forecasted mean sea-level pressure (every 4 hPa), lowest-sigma level winds and reflectivity valid at 00 and 12Z 25 January 2000. The two-way-nested 3-domain MM5 simulation was initialized at 00Z 24 January 2000 with the highest grid resolution of 3.3 km. Middle and right: Observed mean sea-level pressure, surface winds and composite radar reflectivity valid at 00 and 12Z 25 January 2000. Figure 2 (click on graphic to view larger version): Potential vorticity at three isentropic levels for (top) 1200 UTC 6 February 1998 (El Nino); (bottom) 1200 UTC 5 February 1999 (La Nina). 300-K PV (green lines, 2 and 3 PVUs); 320-K PV (color shading, PVU, as in color bar); 340-K PV (black lines, 2 and 3 PVUs) (ECMWF analyses)

2.  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.

a.  Convective Initiation

Andrew Crook and Donna Tucker (University of Kansas) have continued their study of the generation of Mesoscale Convective Systems (MCSs) in the lee of the Rocky Mountains.  Three flow regimes have been identified under which MCSs form in the lee of the Rockies.  These are southwesterly flow, northwesterly flow and curving southerly flow.  Numerical simulations with the Clark-Hall anelastic model have been performed for all of these flow regimes.  These simulations show convection moving out over the Plains late in the day, however, MCS-sized convection has so far not been generated.  The sensitivity to wind speed, CAPE, and various microphysical parameters is currently being explored to determine the optimal conditions for lee-side MCS generation.

b.  Long-time-scale dynamics of mesoscale convective systems

Crook and Morris Weisman have continued their study on the effect of the convective boundary layer on the organization and strength of supercell thunderstorms.  Early results showed that for the same mean environment, storm strength is reduced in a convective boundary layer compared with that in a horizontally-homogeneous environment and it was argued that turbulence in the boundary layer reduces the efficiency of the overall storm flow.  However, as the surface heat flux and eddy convective energy increases, the supercell strength begins to increase again.  This increase at high surface heat fluxes is due to a preconditioning of the boundary layer; convective updrafts modulate the boundary layer so that little or no further lifting is required by the storm's updraft to bring the air to saturation.

Additional work on supercell thunderstorms involves Weisman, Jeff Trapp (NOAA National Severe Storms Laboratory) and Nolan Atkins (Lyndon State College).  Their work focuses on the observation and simulation of tornadic circulations within quasi-linear convective systems such as squall lines and bow echoes.  A set of idealized simulations have been completed that reproduce many of the characteristics of such systems, including the tendency for surface mesocyclones to develop north of the apex of the bow for environments of moderate to strong low-level environmental vertical wind shear.  Preliminary results suggest that the initial source of these low-level circulations is the downward tilting of the horizontal vorticity associated with the cold pool-updraft interface, forced by the downdrafts of strong, short-lived convective cells near the leading edge of the systems.  The action of Coriolis forcing then promotes the strengthening and upscale growth of the cyclonic member of the tilted vortex couplet, producing a significant mesocyclone at the surface that could promote the development of a tornado.  This process appears quite distinct from that associated with supercell tornadoes, whereby a deep, quasi-steady, dynamically forced rotating updraft usually precedes the development of the tornado.  Attempts are also underway to simulate observed storms for comparison with the idealized cases.

Some inferences of predictability associated with warm season rainfall over the U.S. have been discovered in a study by Rit Carbone, John Tuttle, Dave Ahijevych and Stan Trier. The preliminary findings, derived from a radar-based climatology of warm season precipitation, reveal the existence of rainfall “episodes.”  Episodes are defined as time/space clusters of heavy precipitation that often result from sequences of organized convection such as squall lines, mesoscale convective systems and mesoscale convective complexes.  The episodes exhibit coherent rainfall patterns, characteristic of zonally-propagating structures, under a broad range of atmospheric conditions.  The rainfall patterns are especially prominent under “weakly forced” conditions in mid-summer and are often concurrent with a monsoon condition over the Rocky Mountain cordillera.  Diurnal forcing is dominant near the Rockies and Appalachians whereas semi-diurnal forcing is dominant between the mountain ranges.  Many events travel eastward across vast stretches of North America (up to 2600 km) as shown in the figure summarizing frequency of radar echo in July 1997 (see figure 3 at right).  A most common longitude of origin is at or near the east slope of the continental divide (105 W).  The longevity of episodes (up to 60 h) suggests an intrinsic predictability of warm season rainfall that significantly exceeds the lifetime of individual convective systems and greatly exceeds current predictive skill.  Episodes of order 1000 km dimension and one day duration are commonplace, since the mean recurrence frequency from May through August is almost daily.  This observation serves both to supplement and diminish the significance of transient forcings and phase relationships at the synoptic scale.  It recognizes a continental condition of widespread thermal and hydrodynamic instability and the existence of other processes that routinely excite, maintain and regenerate organized convection.  Since episode propagation speeds are often in excess of 50 and 70 kPa winds, it is suspected that wave-like mechanisms, in the free troposphere and/or the planetary boundary layer, may be responsible for propagation.  Such mechanisms may underlie the causes for the observed rainfall coherence and, once these are understood, offer the opportunity for improved representations and predictions from forecast models.

Figure 3 (click on graphic to view larger figure): Radar echo frequency as a function of longitude and time, July 1997.

Numerous research efforts continued in the study of mesoscale convective vortices (MCVs), mid-tropospheric cyclonic circulations that are produced by convective systems and often persist beyond the decay of nocturnal convection.  Trier, Davis and Ahijevych continued efforts in compiling a three-year (1998-2000) climatology of MCV occurrence, environmental conditions leading to their persistence, and their association with subsequent convection.  Results from the 1999 and 2000 warm seasons are consistent with those from the previously analyzed 1998 data and indicate that MCVs are more common than previously suspected, with a total of 15 to 30 MCVs that persisted beyond the life cycle of the convective system from which they were spawned occurring in a given year.  In about half of these cases, new convection occurred in the vicinity of MCV.  Several cases per year were observed to persist beyond a single diurnal cycle, each probably the beneficiary of reintensification by the new convection.

Davis, Ahijevych and Trier performed a statistical and dynamical analysis of mid-tropospheric mesoscale vortices captured by analyses from the Rapid Update Cycle, version 2 (RUC-2) 1999 warm season.  A total of 203 vortices meeting conditions of an automated algorithm were found.  Of these, 86 were observed to form within organized convection, though only 43 formed within well-defined MCSs.  Dry vortices and those arising from non-MCS convection clustered in the lee of the Rocky Mountains.  The MCS-induced vortices were broadly distributed over the High Plains and Midwest.  A robust relationship was found between intensity and longevity such that there appears to be a maximum vortex lifetime that can be predicted from its maximum intensity.  Prediction of mesoscale vortices by the RUC was examined for a subsample of all cases.  In general, the RUC was able to predict the evolution of vortices once analyzed but had virtually no skill at predicting (12 h in advance) the formation of the vortices.

Trier and Davis used observations and numerical simulations with the MM5 model, nested to 1.5 km horizontal resolution, to examine a multi-day event characterized by a large MCV and a sequence of nocturnal MCSs, one of which produced flash flooding.  Flooding resulted from the overall slow movement of the MCV and its associated convection in weak ambient flow and the reorientation of the leading convective line from the eastern to southwestern flank of the MCV overnight.  This reorientation occurred in conjunction with an elevated (1 km AGL) jet that transported air of high equivalent potential temperature towards the southwestern flank of the MCV and resulted in the propagation of the region of most intense convection in a direction opposite to that of individual convective elements.  A decomposition of the flow into its balanced and unbalanced components revealed a strong balanced component to the 1-km southwesterly flow in the vicinity of the MCV, which suggests that in addition to providing a favorable environment for organized convection through balanced ascent, the MCV may also contribute to the transport of conditionally unstable air that fuels convection overnight and the regeneration of the MCV.  Although convection early in the event appeared driven by cold-pool shear dynamics, the MM5 simulation showed that the convective band responsible for the flooding did not possess a significant subcloud cold pool.

Because there is a distinct absence of observations within MCSs that develop significant rotational and divergent signatures in the mid-troposphere, Weisman, Davis and Trier, in collaboration with Dave Jorgenson (NOAA/National Severe Storms Laboratory), Roger Wakimoto (University of California, Los Angeles), Mike Bickerstaff (Texas A&M), and others, have begun planning a new Bow Echo and MCV Experiment (BAMEX).  The goals of this experiment are (1) to obtain kinematic and thermodynamic documentation of the development of system-scale circulation features behind the leading convective line in maturing and decaying MCSs; (2) to understand mechanisms of convective regeneration near MCVs and the dynamics of MCV intensification that appear critical for multi-day events; (3) to understand the cause of damaging surface winds in bow echoes and (4) to assess predictability of long-lived MCSs and their effects on weather.  The experiment is planned for 2002, based in St. Louis, Missouri. BAMEX will feature only mobile facilities and emphasize a system-following strategy in order to sample convective systems over a large portion of the Central U.S.  Two airborne Doppler radar aircraft (P-3 and Electra), a dropsonde jet and a mobile, ground-based observing facility (two radars, boundary-layer profiler, acoustic sounder, soundings and mesonet vehicles) will constitute the main observing facilities. BAMEX builds on the existing National Weather Service observational infrastructure with WSR-88D radars, wind profilers, rawinsondes (with 3-hourly launches) and GOES satellite information.  Collaborating institutions include the National Severe Storms Laboratory, Texas A&M University, University of Alabama, Colorado State University, Penn State University, University of California Los Angeles,  Storm Prediction Center and the NWSFO in St. Louis in addition to NCAR.  The BAMEX Science Overview Document is viewable at http://www.mmm.ucar.edu/bamex/science.html. A detailed study of a cool season mesoscale vortex was performed by Neil Laird (University of Illinois and Illinois State Water Survey), David Kristovich (Illinois State Water Survey), and Miller.  They used the Synthetic Dual-Doppler (SDD) technique to examine the kinematic structure and evolution of the 5 December 1997 winter mesoscale vortex in the vicinity of Lake Michigan (see figure 4 at right).  When such near-steady mesoscale precipitation events advect a distance large enough that the viewing angle from a single Doppler radar (in this case the Grand Rapids Michigan WSR-88D, KGRR) changes by about 30 degrees, then the SDD method can reveal reliable details about the circulation.  During the nearly 2.5 hours that the vortex was in view of the KGRR radar, the reflectivity field exhibited a pronounced asymmetric convective structure with at least three well-defined, inward-spiraling radial snowbands, and a distinct weak-reflectivity region or "eye" near the center of cyclonic circulation.  The SDD results (see figure 5 at right) showed the vortex circulation was comprised of a combination of rotation on the meso-beta scale and convergence on the meso-gamma scale associated with the embedded radial snowbands.  Vertical profiles of derived meso-beta scale, area-mean convergence and vorticity suggest that this winter vortex was likely a warm-core system, similar to both tropical cyclones and polar lows.  However, unlike many Great Lakes vortices, this case featured relatively small surface sensible and latent heat fluxes, and only moderate lake-air temperature differences of 6-9 degrees C were observed over Lake Michigan.  This suggests that dynamic conditions such as horizontal shear across the trough axis and associated convergence zone were more important than topographic or thermal forcing for the development of the vortex. Convective systems in Amazonia Liu and Moncrieff conducted hierarchical three-dimensional numerical modeling of convective systems with two TRMM-LBA soundings and an idealized sounding, with the following preliminary conclusions:  (i) Both the shear-perpendicular squall-type convection observed on 26 January 1999 and the shear-parallel quasi-stationary convective band observed on 23 February 1999 were realized. (ii) The parameterized model (i.e., coarse-grid simulations applying the Kain-Fritsch parameterization) with a grid mesh larger than 10-km failed to adequately produce the evolution of the mesoscale systems realized in the cloud-resolving simulations. (iii) The parameterized model converged toward the cloud-resolving simulation. (iv) Through comparison with coarse-grid explicit simulations, the 5-km - 10-km resolution models hardly benefit from a convective parameterization scheme.

Figure 4(click on graphic to view larger figure): Maximum values of reflectivity in the column from the gridded volumetric scan by the Grand Rapids Michigan WSR-88D (KGRR) on December 5, 1997. Reflectivities (dBZ)follow the color scale on the right. Geographic and political features in the vicinity of the Great Lakes are also shown along with several National Weather Service WSR-88D radars marked by stars and their NWS designations. Figure 5 (click on graphic to view larger version): Reflectivity and winds at 2-km from a synthetic dual-Doppler analysis of gridded WSR-88D level II data for December 5, 1997. Reflectivities (dBZ) follow the color scale on the right. Horizontal vector winds (scaled as the 10 m/s vector near the bottom of the color bar) were derived from separate volume scans at 1023 and 1103 UTC from the Grand Rapids Michigan WSR-88D (KGRR). These gridded radial velocity fields were advected to the 1053 common time and the winds synthesized. The thin line across the western part of the image shows the eastern Lake Michigan shoreline.

c.  Tropical cyclones

Davis and Lance Bosart (The University at Albany, SUNY) continued their study of the transformation of a weak baroclinic disturbance into Hurricane Diana by diagnosing numerical simulations from the NCAR MM5 model  (MM5).  The model is run with three domains, the innermost having a 9-km grid spacing.  Three distinct phases of the evolution were evident.  First, baroclinic and barotropic development, strongly modified by the effects of latent heating occurs.  During the latter part of this phase, the low-level circulation is strengthened through the merging of remote PV anomalies that are generated by condensational heating and then advected toward the incipient storm.  The transformation from cold-core to warm-core vortex occurs in this development stage.  In the second phase, lasting 10-12 h, little deepening occurs.  Spiral bands of convection begin to form and the core of the storm moistens.  The third stage then ensues, driven mainly by the positive feedback between fluxes of latent heat and the increase of the tangential wind.  In this stage, the storm readily develops a clear eye.  Numerous sensitivity simulations revealed that details in the upper-level trough structure preceding the event are important for determining the storm track.  Both track and intensity varied with different choices of cumulus parameterization.  The most realistic intensity is obtained with a fourth domain of 3 km resolution without a cumulus scheme.  At 9 km, the partitioning of latent heating between grid resolved and parameterized has a large bearing on the upper-tropospheric ridge development poleward of the storm and hence a pronounced effect on the storm track.

Jordan Powers and Davis performed an additional numerical simulation of the genesis of Diana, using cloud-resolving 1.2 km grid spacing over a domain of size 1200x1320 km.  The simulation was made possible by the dedication of SCD's IBM supercomputer “blackforest,” with the MM5 model being run in distributed-memory parallel mode over 554 processors – to date, the largest number of processors over which the MM5 has been run.  The simulation was initialized with a weak, synoptic-scale baroclinic disturbance and within 48 h, a tropical storm developed, entirely from grid-resolved convection.  Both individual convective cells and organized convective precipitation bands were evident in the simulation. Further examination of these results will be performed in the next year.

d.  Orographics effects Numerical studies of the 1994 Piedmont flood in the Alps have shown that the orographically modified flow was a critical element for the production of extraordinary rainfall.  To uncover the precise mechanism of orographic rainfall occurring in our full-physics MM5 simulations of the 1994 Piedmont flood, Rotunno and Rossella Ferretti (Univ. L'Aquila, Italy) carried out simulations with the same real-data initial and boundary conditions, but with the real orography replaced by an idealized one.  With excellent agreement between real and idealized orography on the rainfall rate vs. time in the Piedmont area, their analysis of the idealized-orography simulation provides a clear picture of the model's mechanism of orographically induced rainfall.  As noted in previous studies of the 1994 Piedmont case, a moist saturated airflow has a reduced effective static stability and tends to flow over the mountains, while an unsaturated air stream is stable and tries to flow around (toward the left in the northern hemisphere) (see figure 6 at right).  In the 1994 Piedmont case, there was a strong horizontal gradient of moisture; thus the western moist part of the air stream flows over, while the eastern dryer part is deflected westward around the obstacle, and so a convergence is produced between the air streams.  This effect has been explored using a simple version of MM5 wherein the flow, moisture distribution, and idealized orography are varied within the observed range.  Quantitative as well as qualitative rainfall rates and flow features of the full-physics MM5 simulations are captured with the simple model.

Figure 6

Janice Coen and Roelof Bruintjes (MMM/RAP) have continued to investigate the factors that control the timing, quantity, scales, and distribution of precipitation over complex terrain.  Using high-resolution Clark model simulations (as fine as 100 m horizontal grid spacing), they examined summer convective precipitation driven by solar heating in monsoon-like flow against Mexico's Sierra Madres.  Simulations reproduced clusters of small clouds with strong convective updrafts and downdrafts (up to 10 m/s) in bands along the peaks, which deepened and gathered as the solar heating increased through the day.  Within the bands were enhanced convective clusters (see figure 7 at right) in locations clearly tied to specific mountain features ranging from lone mountains to dead-end canyons, with precipitation shadows in small valleys in between, all of which were confirmed by radar and satellite.  These results reinforce earlier conclusions from a variety of precipitating systems (including warm frontal systems in Korea and winter gravity wave-upslope clouds in Arizona) that existing microphysical parameterizations are quite successful at simulating precipitation, provided that dynamics resulting from topographic features and mesoscale and synoptic features introduced through large-scale initialization are sufficiently resolved, vertically as well as horizontally.  As a potential framework for cloud seeding studies, such simulations also caution against current efforts in the literature to generalize conclusions from flat-ground bubble-type model simulations of seeding of individual clouds to actual situations, because cloud microstructure is intimately tied to dynamics of neighboring clouds and topographic influences.

Figure 7 (click on graphic to view larger version): Horizontal cross section of modeled cloud water field (red contours, every 0.25 g/kg) superimposed on radar estimated precipitation field in thesame location, along the edge of the Sierra Madre.

Flows past complex terrain: adequacy of NWP models

Piotr Smolarkiewicz and Michael Cullen (ECMWF) performed an extensive numerical study of stratified flows past the Scandinavian Peninsula to test the adequacy of a numerical weather prediction (NWP) model in handling complex terrain. Key questions are: a) do NWP models (and the ECMWF model in particular) adequately capture the smooth aspects of natural flows past complex terrain? and b) what numerical difficulties do the present NWP models have when they switch to nonhydrostatic formulations in near future?  In order to address these issues, they have performed, overall, several hundred simulations using MMM's nonhydrostatic model EULAG.  These simulations spawned a range of numerical experiments from 2D flows past highly idealized mountains to 3D rotating flows past high-resolution natural orography of the peninsula.  Each experiment was conducted at a range of Froude numbers.  Four numerical methods (explicit/implicit, semi-Lagrangian/Eulerian) were used, and the experiments were conducted using successively lower horizontal resolutions.  The results of the low-resolution experiments were compared with the results of high resolution runs averaged to the appropriately low resolution—a unique convergence study.  The results revealed: a) present NWP models can capture smooth (low resolution aspects of natural flows) surprisingly well but the effective model resolution is 2 to 4 times lower than that of the grid, depending upon the methods employed and the flow fields; b) at high resolution, where the natural terrain becomes truly complex, the semi-Lagrangian approach becomes unreliable because it cannot accurately represent at the lower boundary conditions; c) at high resolution, finite volume Eulerian schemes may require a higher-than-anticipated degree of implicitness in order to represent the lower boundary condition accurately.

e.  Cloud microphysics and precipitation

Charles Knight, working with Jothiram Vivekanandan (NCAR/RAP) and Sonia Lasher-Trapp (NCAR/ASP and Texas A&M University) completed analysis of a detailed data set obtained with the S-Pol radar in PRECIP 98 in central Florida, on precipitation initiation in warm cumulus.  First radar echoes from precipitation were recorded along with Z_DR, a measure of raindrop oblateness and hence raindrop size.  It was already known that early radar echoes could be composed of very low concentrations of raindrops one to several millimeters in diameter; this study confirms that, but also finds that drops of this size are often present low in the clouds well before the distinct increase of the radar echo aloft that is usually attributed to the first precipitation growth.  The studies also reveal a way of detecting and characterizing the first significant precipitation formation by the warm rain process.  (The very early, large drops are so sparse that they constitute virtually zero rainfall.)  This occurs high in the clouds and is characterized by rapid increase in Z_e with low values of Z_DR, signifying more and smaller raindrops, with conversion of significant amounts of cloud water to rain.

To investigate microphysics in mature convective storms, the Severe Thunderstorm Electrification and Precipitation Study (STEPS) field program was conducted in Eastern Colorado and Western Kansas from 22 May to 16 July 2000.  The experiment was organized by Weisman, Knight, and Miller in collaboration with scientists at the Colorado State University (CSU), New Mexico Institute of Mining and Technology (NMIM&T), South Dakota School of Mines and Technology (SDSM&T), the National Severe Storms Laboratory (NSSL), and the National Weather Service (NWS) office in Goodland, Kansas.  The broad goal of STEPS was to achieve a better understanding of the interactions between kinematics, precipitation production, and electrification in severe thunderstorms on the High Plains.  The specific scientific objectives included: 1) understanding the dynamical and microphysical differences between LP, CL and HP (low, classic and high precipitation) supercells; 2) understanding lightning formation and behavior in storms, and how it differs among storm types; and 3) verifying and improving microphysical interpretations from multi-parameter radar.  A particular interest of the program was to document and better understand the mechanisms by which storms produce predominantly positive CG (cloud-to-ground) lightning.  The primary instrumentation included the NCAR S-Pol, CSU CHILL and the NWS Goodland WSR-88D radars for multiparameter and Doppler wind observations, the SDSM&T T28 armored aircraft for microphysical observations within storm updrafts, a three dimensional lightning mapping system from NMIM&T, balloon-borne electrification measurements and mobile mesonets from NSSL, and two mobile sounding units from NCAR.

Events sampled by STEPS ranged from a high-based downburst producing storms to a tornadic supercell.  One LP storm was also sampled at the 30-50 km range from the S-Pol radar.  One of the significant surprises of the field campaign was that positive CG storms were not limited to strongly-sheared supercell hailstorms, as originally thought, but occurred over the entire spectrum of storm types observed.  Preliminary analyses suggest that storms in the STEPS region exhibited a reverse polarity, with positive charge at mid-storm depths and negative charge higher up in the anvil regions.  This finding may have significant implications as to our understanding of both the electrification and microphysical make-up of storms in this part of the world.

Satellite studies of relationships between lightning frequency and thundercloud parameters

John Latham, Hugh Christian and Kevin Driscoll (MSFC/NASA); Alan Blyth (New Mexico Institute of Mining and Technology) and Alan Gadian (University of Manchester Institute of Science and Technology) have extended and refined their lightning frequency model in an effort to provide a more realistic framework from which to examine computationally the relationships that might exist between lightning frequency (which is now being routinely measured from satellites, using NASA/MSFC devices) and a variety of cloud physical parameters, including precipitation rate, updraft speed and non-precipitating ice content.

Model results indicate the existence of a simple relationship between lightning frequency and the upward flux of ice crystals into the thunderstorm anvil.  It follows that, for a particular situation, one can assign a specific mass of non-precipitating ice to an individual lightning stroke.  Therefore it may prove possible-using satellite measurements of global lightning - to estimate the atmospheric loading of ice crystals in thunderstorm anvils, a parameter of climatological importance.  Early results from this work, employing lightning frequency and brightness temperature measurements made by instruments flown on the TRMM (Tropical Rainfall Measurement Mission) satellite, provide encouraging indications that quantitative relations - of global validity - may be found between lightning frequency and both precipitating and non-precipitating ice thunderclouds. 

Glaciation processes in deep convective clouds

Latham, Phil Brown (UK Meteorological Office), Alan Blyth (New Mexico Institute of Mining and Technology), and Tom Choularton and Vaughan Phillips (University of Manchester Institute of Science and Technology) have taken a first step to establish, on a quantitative basis, the processes responsible for the development of glaciation in deep convective clouds.  They have employed the UK Meteorological Office Cloud Resolving Model (CRM) and the University of Manchester Institute of Science and Technology Explicit Microphysics Model (EMM) in analysis of data from airborne studies of a summertime multi-thermal cumulus cloud which developed over New Mexico.  The EMM was utilized in a series of tests designed to assess the sensitivity of cloud glaciation via the Hallett-Mossop process to obtain cloud parameters such as CCN concentration, cloud-base temperature, entrainment, and the freezing and splintering of supercooled raindrops.  Good agreement was found between the predictions of the CRM and the available dynamical and microphysical field observations.  Analysis indicated that the cloud glaciation is explicable in terms of the Hallett-Mossop process, with ice production being dominated by the freezing of supercooled raindrops in the Hallett-Mossop band, and the immediate and continuous production of ice splinters as supercooled droplets freeze onto them.  This work is being extended to encompass more case studies, particularly of tropical convection.

Precipitation development in small cumulus clouds

Latham, Alan Blyth (New Mexico Institute of Mining and Technology) Al Cooper, Charles Knight  and Sonia Lasher-Trapp have compared radar and aircraft observations of warm cumulus congestus clouds with results of collision and coalescence calculations in a closed parcel model in order to examine whether or not the initial development of warm rain can be explained solely by taking account of the full range of cloud condensation nuclei (CCN), i.e., including giant and ultra-giant aerosols  (UGA).  The clouds were observed using an instrumented aircraft and high-powered dual-wavelength radar during the Small Cumulus Microphysics Study (SCMS) conducted in Florida in the summer of 1995.  The calculated altitudes of the -5 to 0 dBZ echoes are consistent with the observations in the three cases studied, only when UGA are included.  No correlation was found between the observed altitude of the first 0 dBZ reflectivity echo and the total concentration of droplets in the cloud.  The calculations indicate that this result is consistent with UGA being ingested into the clouds.

Modeling studies of freezing drizzle

Roy Rasmussen and Istvan Geresdi (visitor from the University of Pecs, Hungary) conducted studies on the formation of freezing drizzle in stably stratified clouds with a detailed microphysical model implemented into a 2D version of MM5.  The detailed microphysical model included four categories of ice (pristine ice, rimed ice, aggregated ice, and graupel) and used the moment conserving numerical technique.  These studies were meant to simulate the formation of freezing drizzle in warm fronts, artic and cold fronts, and near low-pressure centers.  Drizzle forms in these synoptic situations primarily through a collision-coalescence process (melting snow is usually not a factor).  Motivation for this work is improved forecasts of freezing drizzle for aircraft icing and surface transportation.

Kevin Manning contributed to this effort by using a bell-shaped mountain 1200 km in length, 1 km high, and with a half width of 100 km to simulate the 5-10 cm/s updraft typically observed in freezing drizzle cases.  The results showed a strong sensitivity to the initial CCN distribution.  For a maritime CCN distribution, drizzle formed 1-2 hours after initialization.  A continental CCN distribution, on the other hand, took 4-5 hours to form drizzle.  The cloud also needed to be deeper and longer in the continental case.

The formation of drizzle was also found to be very sensitive to the formation of ice.  The commonly used Meyers ice formation scheme was found to over-deplete supercooled liquid water, suppressing drizzle formation.  If ice crystals were initiated using the observations of Cooper, then reasonable agreement with observations was achieved.

The key result of this study was the discovery that ice nuclei needs to be depleted for freezing drizzle to form.  If it is not, too much ice forms for both the Meyers and Cooper methods of ice initiation.  This result is significant because most cloud and mesoscale models do not take this effect into account.  Further studies will be required to show the generality of this result.

3.  Mesoscale Data Assimilation

The primary goal of mesoscale data assimilation research is to develop and support state-of-the-art data assimilation systems for application in high-resolution mesoscale models.  These data assimilation systems can be used for a variety of purposes including the assimilation of data from new observing systems, the optimal use of observations, and understanding the observational requirements for accurate precipitation forecasts and optimal strategies for obtaining targeted observations.

a.  Advanced data assimilation systems for community use Development of MM5 3DVAR system Dale Barker, Yong-Run Guo, Wei Huang, and Francois Vandenberghe continued the development of a three-dimensional variational data assimilation (3DVAR) system for use with the MM5 model.  Beginning in October 1999 with a prototype code developed by Vandenberghe, the system has since been extended to provide fully three-dimensional background error covariances and now permits a linearized geostrophic/cyclostrophic balance between mass and wind analyses.  Further tuning of the background error statistics used in MM5 3DVAR has been performed in order to provide a system ready for detailed intercomparison with existing assimilation schemes.  A particular goal of MM5 3DVAR development has been to ensure flexibility of the algorithms for potential use in other assimilation schemes, in particular the WRF 3DVAR system. The capability to assimilate observations from a number of  additional observation sources has been added during the year.  These include geostationary satellite's cloud-track-winds, SSM/I total precipitable water (TPW) and oceanic wind speed and TPW retrieved from ground-based GPS data. An example of the impact of a single TPW observation is shown in figure 8 at the right.

Figure 8 (click on graphic to view larger figure): This figure illustrates the response of the MM5 3DVAR system to a hypothetical single total precipitable water (TPW) observation situated over Taiwan. Particular isosurfaces of specific humidity (blue), temperature (yellow) and pressure (purple) are shown, as is the coupled wind response (cyclonic circulation aloft, anticyclonic below) to the TPW observation. The mass/wind coupling is specified via linearised geostrophic/cyclostrophic balance within 3DVAR.

b.  Optimal use of existing observations and potential benefits of new observing systems

Surface meteorological stations generally have a much higher density compared with the radiosonde observations.  The surface reports, if properly assimilated into the model, can be used to improve the quality of model initial conditions.  In collaboration with Dong-Hyun Shin (Korean Meteorological Administration), Vandenberghe developed an algorithm for the assimilation of surface observations, based on boundary-layer physics.  This algorithm was included in the MM5 3DVAR observation pre-processor, and is shown to assimilate surface observations effectively into the model.

Assimilation of mesoscale observations

With the recent advance in Global Positioning System (GPS) Meteorology, ground-based GPS receivers have become an important instrument that can potentially provide high resolution water vapor observations at low cost.  Since 1996, near real-time GPS sensing of atmospheric water vapor from the NOAA network has been available for assimilation in numerical weather prediction.  Recent developments in GPS tracking and processing techniques have allowed accurate measurement of atmospheric wet delay along a slant path between a GPS satellite and a ground receiver.  This allows high temporal resolution measurements of azimuthal variations in atmospheric water vapor, a technique known as “slant water vapor” (SWV) measurements.  The SWV measurements provide an opportunity to capture the three-dimensional structure of water vapor through tomography, or advanced data assimilation techniques. Vandenberghe and Guo, in collaboration with Yoaz Bar-Sever (Jet Propulsion Laboratory), have developed an observation operator for the variational assimilation of GPS line of sight (LOS) observations.  Comparison between simulated GPS LOS data from very high-resolution (3km) model forecasts and real observations taken at the Lamont ARM site have shown a relatively good agreement.  Research continues to improve the forward operator and to develop its adjoint.

Another important upcoming observing system for meteorology is the atmospheric limb sounding technique that makes use of the radio signals broadcasted by the 24 GPS satellites.  By placing a GPS receiver onboard a low-Earth orbiting satellite, the limb sounding technique can provide a vertical profile of bending angles at the ray perigee point.  Using a local spherical symmetry assumption, the vertical profile of refractivity can be derived from the bending angle data.  Because of the dependence of refractivity on both temperature and water vapor, independent information on one variable (temperature or water vapor) is required to derive the other (water vapor or temperature) from the refractivity observations.  The traditional retrieval technique is to assume the temperature from a global analysis (i.e., ECMWF analysis) is “perfect,” and to derive the water vapor profile from refractivity soundings. Bill Kuo, in collaboration with Tae-Kwon Wee (Seoul National University), has developed a one-dimensional variational assimilation scheme (1DVAR) that makes use of realistic physical constraints and the error statistics of global analysis.  They show that the 1DVAR produces considerably improved temperature and water vapor profile retrieval than the traditional retrieval method.

Juanzhen Sun, in collaboration with Prof. Ching-Long Lin (University of Iowa), applied the four-dimensional variational data assimilation (4DVAR) technique to the retrieval of turbulent eddies in a convective boundary layer (CBL).  Simulated data that emulate Doppler-lidar observations are used in combination with a boundary layer numerical model.  They demonstrated that the microscale turbulent eddy structures in a simulated CBL can be recovered with good accuracy.  Addition of a surface flux model improves the retrieval quality, whereas allowing height-dependent eddy viscosity and diffusivity does not lead to better retrieval quality.  The implementation of the temporal and spatial smoothness penalty function with dynamic tuned coefficients significantly improves the retrieval quality in the presence of different types of errors.  All of the retrieval experiments show higher correlation coefficients between retrieved data and exact data for velocity than for temperature.  The retrieval of temperature is most sensitive to data errors that are correlated in three orthogonal directions.

c.  Data assimilation and ensemble forecasting

Ensemble Kalman filter data assimilation and adaptive observations

The extremely simple form assumed for the background covariances is an important limitation of present atmospheric assimilation schemes.  One route to overcoming this limitation is to use an ensemble of short-range forecasts to estimate the background covariances.  The resulting assimilation schemes are typically referred to as ensemble Kalman filters. Hamill and Snyder (2000) have implemented an ensemble Kalman filter in a quasigeostrophic model and find that both analyses and forecasts are significantly improved relative to a three-dimensional variational scheme (3DVAR).  Benefits increase as the density of available observations decreases and as the number of ensemble members increases; for an observing network yielding analyses of comparable quality to present global analyses, an ensemble Kalman filter using 25 members decreases analysis error by a factor of 40% relative to 3DVAR. Presently, Hamill and Snyder are examining the use of distance-dependent filtering of the estimated covariances.

d.   Research in adaptive observations

Given the location and uncertainty of an observation, the ensemble Kalman filter can provide a quantitative estimate of the impact of that observation on the analysis uncertainty.  Hamill and Snyder have begun a study within the quasigeostrophic model in which this capability is used in adaptive observations.  They have developed a simple and efficient algorithm to choose the locations of observations: using the ensemble Kalman filter, they find the location at which the impact of an observation is estimated to be largest; they then find the next best location given the first observation and continue the process up to the desired number of observations.  They find that as few as half as many adaptive observations are required to produce analyses with the same uncertainty as those obtained for a given fixed observation network.

4.  Real-Time Experimental Numerical Weather Prediction

Jordan Powers, James Bresch, Kevin Manning, Dale Barker, Francois Vandenberghe, Yong-Run Guo and visitor John Michalakes (Argonne National Laboratories) continued efforts to design, implement, and tune a MM5-based forecasting system for Taiwan's Civil Aeronautics Administration (CAA).  The system was expanded to generate four forecasts per day, with the products from these being used by meteorologists and flight forecasters at the CAA in a number of display systems developed in MMM and RAP.  Over the past year, MPG has implemented this system on a Fujitsu VPP5000 supercomputer at Taiwan's Central Weather Bureau.  In evaluating the model for the East Asia forecast region, research and development was undertaken in cumulus parameterization behavior, tropical cyclone bogussing, and first-guess field reanalysis.  The system ingests a broad range of input data, such as that of global models and diverse observation types (e.g., MDCRS, satellite-derived winds and temperatures, typhoon reports, and GPS measurements).  Creation of the system's 3DVAR capability continued, with research addressing control variables, background error correlation length scales, and the assimilation of ground-based GPS precipitable water data.

Manning and Fei Chen (RAP) implemented the OSU land surface model in the real-time systems built for the Army Test and Evaluation Command (ATEC). The land surface model had already been put into the MM5 system, but it had received only limited testing and use (for example, the case study of the Buffalo Creek, Colorado, flash flood event of 1996 by F. Chen, T. Warner, and K. Manning).  Attempts to use it in more general circumstances turned up a number of problems related mostly to initialization and nesting.  The implementation within the real-time ATEC systems involved adapting the system to new sources of input, and adapting the MM5 code to more adequately handle nesting.  These improvements to the ATEC system have been fed back to the MM5 community code.  Additional changes to the preprocessing of certain datasets have also been fed back to the community version of the MM5 model.

In other work for ATEC, Simon Low-Nam, in collaboration with Jennifer Cram and Yubao Liu of RAP, developed quality control (QC) algorithms and formats for different data types used in a real-time Four Dimensional Data Assimilation (rt-FDDA) system.  This system is a frequent update system, similar to the Rapid Update Cycle run by the National Centers for Environmental Prediction (NCEP), but runs with an innermost domain of 3.3 km resolution (as opposed to 20 km for the RUC).  Observation nudging is used to assimilate the observations.  Low-Nam developed QC algorithms for surface mesonet, profiler and RASS data which are ingested on an hourly basis.  Low-Nam also developed a data format strategy which merged these data with WMO observations into a more general format than previously existed.

5.  Model Development

a.  Global MM5

Jimy Dudhia developed an experimental global version of MM5 for use in medium-range forecasting.  This development started in Summer 1999, and was completed in October 1999, when Jim Bresch implemented real-time 5- and 10-day global forecasts with this version.  Real-time forecasts have been produced regularly for a year, and are displayed on the MMM Division Real-time MM5 Web Pages.  The forecasts are initialized with MRF data from NCEP plus a reanalysis using all available surface and sounding data.  The method used to make MM5 global was an innovative technique of interfacing two hemispheric domains at the equator. By allowing each domain to cover a square region that extends slightly beyond the equator the code modification to MM5 is minimized.  Tracing features across the equator shows the interface to be working well.  The model has comparable resolution, and so results can be compared to operational medium-range forecasts, and they appear within the range of variability among such models, often providing good quality forecasts of long-wave patterns to 5 days, and surface patterns to 3 days.  The MM5 global model was parallelized by John Michalakes and Al Bourgeois in Summer 2000 to run on the Division's larger Compaq machines for the Antarctic Mesoscale Forecasting System.  Unlike the regional MM5, the global model requires no boundary file, and can therefore be run indefinitely in principle.  In practice, however, sea-surface temperature variation is needed for realistic long-range simulations.  Potential future applications include medium-range predictability and global 3DVAR studies.

b.  High-resolution Weather Research and Forecast (WRF) Model

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 WRF project is a collaborative effort among a number of organizations and university scientists.  At present, NCAR/MMM, NOAA/NCEP, NOAA/FSL, OU/CAPS, and the Air Force Weather Agency are the principal partners committed to the project.  During the past year, the model development efforts have progressed substantially, with the result that a first version of the WRF system is about to be released to the community for evaluation and testing.

WRF model prototypes for integrating the dynamical equations

Recognizing the research focus within the WRF effort, alternative numerical techniques are being explored and adapted to the WRF framework to facilitate the comparative evaluation of their relative accuracy and efficiency in a controlled computational environment.  Work has been progressing on three candidate prototype solvers; two of these prototypes are split-explicit Eulerian models, while the third is a semi-implicit semi-Lagrangian formulation.  Based on the performance of these prototypes, options will be considered for supporting multiple solvers that are selectable within the framework.

During the year, Skamarock, Michalakes, and Dudhia implemented the Eulerian, terrain-following height coordinate, split-explicit, flux-form prototype within the WRF model computational framework.  The implementation includes both 2nd and 3rd order Runge Kutta (RK2, RK3) time integration methods that allow the use of high-order upwind advection operators and do not suffer from the large dispersion errors found in leapfrog schemes.  The RK2 scheme, while computationally more efficient, was found to have significant timestep constraints for upwind advection operators higher than 3rd order.  The 3rd order timesplit Runge Kutta scheme, developed by Lou Wicker (NOAA/National Severe Storms Laboratory) and Skamarock, allows the use of both upwind (odd ordered) and centered (even ordered) high-order flux-divergence operators.  It also exhibits low dispersion errors and allows a larger time step when used with either odd or even higher order advection operators.  The schemes appear robust, and, as anticipated, the higher order schemes give superior solutions at marginal resolution. Skamarock and Michalakes have demonstrated that the timesplit methods, as implemented in the WRF modeling framework, are efficient for parallel processing computer architectures.  Higher order spatial discretizations, and the predictor-corrector type approaches embodied in the Runge-Kutta timesplit schemes (see above), can bring with them the need for more inter-node communication in distributed memory computer architectures.  The RK3 scheme allows larger timesteps with increased accuracy, which largely ameliorates the increased computations and communications needed in the parallel applications.  Tradeoffs between communications and computations can be made, and benchmarking is continuing with the aim of examining these issues within the context of overall model performance and efficiency. A variety of test cases covering a broad range of scales have been performed with this first prototype, including simulations of synoptic-scale baroclinic waves in a periodic channel with a 100-km horizontal grid (see figure 9 at right), and supercell evolution in a homogeneous environment with a 1-km grid (see figure 10at right).  These simulations and others previously conducted provide benchmarks for the WRF prototype with published solutions from other models, and demonstrate the robustness and accuracy of the new approaches used in the WRF prototype.

Figure 9 (click on graphic to view larger figure): Moist Baroclinic Wave Simulation. Height coordinate model (dx=100 km, dz=250 m, dt=600 s; Surface temperature, surface winds, cloud and rain water. Figure 10 (click on graphic to view larger figure): Supercell Thunderstorm Simulation. Height coordinate model, (dx = dy = 2 km, dz = 500 m, dt = 12 s, 160 x 160 x 20 km domain ); Surface temperature, surface winds and cloud field at 2 hours

Development of the Eulerian prototype in mass coordinates by Klemp, Skamarock, and Dudhia has also progressed substantially.  Skamarock has extended the initial 2-D test code to three dimensions, including moist processes, and implemented the 3rd order Runge-Kutta time integration scheme (described above) in such a manner that either leapfrog or RK3 time stepping can be selected with a simple switch.  This prototype has been tested through supercell storm simulations, and integration into the WRF computational framework is beginning.

Development of the semi-implicit semi-Lagrangian prototype is being led by Jim Purser (NOAA/National Centers for Environmental Prediction).  To attain a high formal order of accuracy for the spatial operations of differentiation and quadrature, Purser has developed a package of efficient “compact” or “Pade” schemes.  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.  Purser has developed algorithms for the efficient implementation of schemes on either serial or parallel computers.  The solver for this prototype is presently under development and will be evaluated in comparison with the other prototypes.

Also within the context of the WRF model solvers, Skamarock, Stan Benjamin (NOAA/Forecast Systems Laboratory) Riener Bleck and Zuwen He (both University of Miami) have been examining 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.  Zuwen He is developing a hybrid approach based on an explicit solution technique that splits the integration of the acoustic modes from the coordinate-surface movement.  This approach may be generally more robust because 3D Helmholtz solvers are not needed, but problems associated with the detrimental effects associated with splitting need to be solved.

WRF computational framework

The WRF model computational framework has undergone substantial development over the past year by Michalakes, Shu-Hua Chen (Air Force Weather Agency), Dudhia, Dave Gill, and Skamarock.  A solver for an Eulerian, timesplit, flux-form equation set has been implemented within the framework as the first prototype WRF model.  Distributed-memory, shared-memory, and hybrid parallelism is supported within the framework, and significant effort has been made to preserve efficiency across both vector and microprocessor-based architectures.  Tests with this prototype revealed significant differences in model performance based on the storage order of the 3D variables in the model.  In these tests, it was found that scalar architectures preferred the vertical dimension innermost in storage (z,x,y), whereas vector architectures preferred having the horizontal dimensions innermost in storage (x,y,z).  Based on the results from benchmark tests with alternative loop and storage order, WRF prototypes are being written with a single horizontal dimension innermost followed by the vertical dimension (x,z,y), because this option provides the best compromise in achieving efficiency across both scalar and vector architectures.

The WRF software architecture consists of three distinct model layers: a solver layer that is usually written by scientists, a driver layer that is responsible for allocating and deallocating space and controlling the integration sequence and I/O, and a mediation layer that glues these pieces together.  A novel aspect of this modeling system is its use of a data registry.  The data registry, designed and implemented by Michalakes, is the single place where developers list model variables and their characteristics.  A preprocessor uses this list to construct all memory allocations within Fortran 95, calls, variable declarations, I/O, and other time-consuming and potentially error-prone coding, throughout the model hierarchy.  Thus, the addition of variables is a simple task in the WRF modeling system, making applications needing additional variables easy to accommodate.  These advanced features of the WRF framework should make it possible to easily integrate other solvers and model components in the future.

In order to streamline the handling of I/O throughout the many components of the overall system, Michalakes, and Leslie Hart and Jacques Middlecoff (both NOAA/Forecast Systems Laboratory) have designed and implemented an Application Program Interface (API) that provides a standard way of specifying and accessing data within the model that is independent of any particular I/O package. For the initial version of WRF, they are using the API to implement the model I/O based on the NetCDF format.  Other data formats, such as HDF and GRIB, will be coupled to the I/O API as the code matures.  At present, work is continuing to integrate the new I/O interface within the WRF software framework and refine the formats used for the NetCDF output files.

WRF model physics

Dudhia and Chen have collaborated with Klemp, Michalakes, and Skamarock to incorporate a variety of physics options into the WRF prototype to be released in Fall 2000.  In this early stage of model development, only existing physics schemes are being implemented, with most of them being adapted from other existing models or packages.  These packages include microphysical options (Kessler, Lin et.al., and NCEP simple ice), a boundary-layer/surface option (MRF, soil thermal diffusion), cumulus parameterizations (Kain-Fritsch, Betts-Miller-Janjic), sub-grid turbulence (constant K diffusion, Smagorinsky, TKE) and short-wave (simple MM5 scheme, Goddard) and long-wave (RRTM) radiation.  Outside collaborators for these packages have included Song-You Hong (formerly National Centers for Environmental Prediction/Environmental Modeling Center (NCEP/EMC)), Tom Black (NCEP/EMC), Wei-Kuo Tao (NASA Goddard), Jack Kain (National Severe Storms Laboratory), Ming Xue (Oklahoma University /Center for Analysis and Predictions of Storms), and John Brown (NOAA/Forecast Systems Laboratory).  Work continues on incorporating a land-surface model with Fei Chen  (NCAR/RAP).  These physics options constitute sufficient physics sophistication to allow WRF to be used for preliminary real-data testing.  A few cases have been tested and the model has been able to reproduce well-known solutions.

The development of WRF physics is an ongoing research effort that must rely significantly on community participation.  Therefore, a standard physics interface has been designed in order to streamline participation in developing and adapting physics to WRF.  In this interface, the model solver calls a generic driver for each class of physics, which in turn calls the specific desired package.  Thus, user developed packages plug into the physics driver through the standard interface, and remain isolated from the model solver.

WRF real data initialization

Gill has worked with Dudhia and collaborated with Brent Shaw, Paul Schultz, and John Smart (all NOAA/Forecast Systems Laboratory), to bring real-data analyses into the WRF model.  This initialization uses programs developed at FSL that process GRIB format files to interpolate fields from operational and archived gridded meteorological data sets onto the WRF grid.  It also incorporates static information, such as terrain elevation maps.  Gill and Dudhia have ingested the FSL data into the WRF model, and have also written an interpolator to initialize WRF from MM5 model fields for comparison purposes.  Preliminary real-data tests have been run with the WRF prototype and compared to MM5, and indicate the implementation is working properly.

WRF model data assimilation

Dale Barker has conducted numerous discussions and meetings with other WRF data assimilation developers, primarily at NCEP and FSL, to define the exact requirements of the WRF 3DVAR system, to assign a priority to each component and to agree on an approach for building a basic, usable WRF 3DVAR system by the summer of 2001.  Although each centre's data assimilation efforts have been primarily focused on their individual systems, an agreement has been reached to use a prototype WRF 3DVAR top-level code developed for MM5 3DVAR as a framework for inclusion of particular algorithms at a later date.  Barker has been incorporating all of the basic WRF 3DVAR functionality into this prototype that is compatible with requirements for the MM5 system.  Two remaining capabilities not yet addressed are conversion to an unstaggered grid and to BUFR input format for observational data.

First WRF model release to community

Klemp, Skamarock, Michalakes, Dudhia, Chen, Gill, and Wang have been working in cooperation with WRF developers at NOAA/FSL to assemble a first release of the WRF model to the community in the first quarter of FY01.  This first version is regarded as a beta release for friendly users and will offer an early opportunity for broader involvement in the development of WRF.  It will allow interested users to participate in evaluation and testing of the basic model system, to begin porting physic to WRF and developing new physics, and to provide early feedback that will shape future directions and capabilities for WRF. 

This first release integrates the fully compressible nonhydrostatic equations in scalar-conserving flux form using a time-split small step for acoustic modes.  Large time steps utilize the Runge-Kutta techniques discussed above and 2nd to 6th order advection operators can be specified.  The vertical coordinate is a terrain following height coordinate that allows variable resolution with height.  It is initially a single domain version and contains map-scale factors for conformal projections.  The model code is written in standard Fortran 90 and is self-contained.  It will run in parallel on both shared-memory and distributed memory platforms.  The initial format for model output file will be in NET CDF.

The model can be configured to run either idealized or real data simulations.  For idealized simulations, periodic, symmetric, or open radiative lateral boundary conditions are available.  For real-data cases, initial fields are interpolated from GRIB or MM5 files and model physics can be selected from the above mentioned options.  Lateral boundary conditions are specified and merged to the interior with a relaxation zone.

The WRF system will be supported for community use in a manner similar to that provided for the MM5 system.  As other prototypes for the WRF dynamic core are implemented within the WRF framework, they will also be made available for evaluation and testing.

WRF project management

Klemp prepared a WRF management plan that was approved by the Interagency Working Group (IWG).  The management structure consists of a WRF Oversight Board (WOB) to monitor the progress of the project, a WRF Science Board to provide guidance on user needs and technical issues, and five development teams (containing some 13 working groups) to design and implement the model system.  The WRF Oversight Board held its first meeting in January, and formally adopted the management plan and appointed Klemp as overall WRF Coordinator.  In March, the WRF development teams held a planning workshop in Boulder to review the status of development efforts, and refine the plans and schedules for future work.  The First WRF Users Workshop was held at NCAR in June to inform prospective WRF users of progress and plans in developing the model and to discuss the interests and desires of this user group.  Over 110 people from all over the country attended. Following the Users Workshop, the WRF Science Board held its first meeting and discussed plans presented in the Workshop.

Further information on the WRF project and the first model release is available on the WRF web site at: http://wrf-model.org

c.  Adaptive-Grid Nonhydrostatic Global Model

Joseph Prusa (Dept. of Mechanical Engineering, Iowa State University) and Piotr Smolarkiewicz developed an adaptive grid-refinement approach for the nonoscillatory forward-in-time (NFT) nonhydrostatic global model.  The major focus was the design of a generalized mathematical framework for the implementation of deformable coordinates and its efficient numerical coding in a generic Eulerian/semi-Lagrangian NFT format.  A key prerequisite of the adaptive grid is a time-dependent coordinate transformation.  It results in numerous Christoffel symbols, or grid metric terms in the equations that arise due to the motion, non-orthogonality and curvature of the deformable coordinate system.  By design, the zonal and meridional coordinates of the grid points do not depend on altitude.  This greatly reduces the number of metric terms.  Tests have established the efficiency and accuracy of the deformable-grid code.  A particular benefit of the adaptive mesh is an optional meridional stretching that clusters grid points near the equator while moving them further from the poles.  This can reduce the condition number of the elliptic operator significantly leading to a substantially faster code.

Spurious vortical structures in under-resolved flow simulations

Dimitris Drikakis (Queen Mary, University of London) and Smolarkiewicz have investigated the formation of spurious vortical structures in incompressible flow simulations employing Godunov-type methods.  Their work is motivated by the earlier studies of Brown and Minion (Journal of Computational Physics, 1995, 1997) who demonstrated for a variety of numerical schemes (and for the upwind-biased methods in particular) that spurious vortices can occur in under-resolved flow simulations.  The aim of the present effort is threefold: (i) to identify deficiencies in various Godunov-type methods leading to spurious flow structures, (ii) to examine the numerical mechanisms responsible for these artifacts, and (iii) to propose modifications of Godunov-type methods in order to recover the correct solutions even under insufficient grid resolution.  Their results reveal that the occurrence of spurious solutions depends strongly on the Godunov-type method employed.  They show that in addition to the dissipation other factors can also contribute to numerical artifacts.  These include a vortical instability arising from the numerical discretization of the advective terms in the primmitive variable formulation of the Navier-Stokes equations, the balance of dissipation among the different discretized terms in a Godunov-type method, as well as order of accuracy of the interpolation used to discretize the wave-speed dependent term of the Godunov flux.

3-D Instability of counter-rotating vortices

Andreas Dornbrack (DLR) and Smolarkiewicz investigated the three-dimensional instabilities of counter-rotating vortices in the atmospheric boundary layer.  The study employs high-resolution simulations, using MMM's nonhydrostatic model EULAG, for a realistic set of parameters of vortex core radius, distance and initial circulation.  The results quantify atmospheric states where the spacing rules could be changed in order to reach higher frequencies of aircraft landing and take-off without jeopardy to aviation.  This research involves cooperation with Lufthansa and the German Air Control at Frankfurt.

Large-eddy simulations of solar convection

Smolarkiewicz and Julian Elliott (University of Reading) investigated solar convection and differential rotation using LES modeling without explicit viscosity.  Previous studies have employed artificial viscosity, which, by necessity, has been many orders of magnitude greater than the true molecular value, and have only been able to reach Reynolds number of order 50.  Their new simulations are able to reach much higher Reynolds numbers (based on the grid spacing), and reveal a richness of turbulent structure: rotationally-aligned convection cells are still seen near the equator (which, although significantly distorted, are analogous to the banana cells seen in laminar solutions), and a prograde equatorial zonal jet is seen, in keeping with deductions from helioseismology.  Additionally, there is evidence for a breaking of the Taylor-Proudman state of rotation constant on cylinders, which was difficult to achieve in previous simulations.  These studies were carried out using the same MMM nonhydrostatic global code, EULAS, as has been successfully employed in modeling the Earth's atmosphere, demonstrating the viability of a unified modeling approach to fluid flow in spherical geometry.

Generation and propagation of coastal solutions

Patrick C. Gallacher, Steve Piacsek, and Alex Varn Warnas (all NRL) and Smolarkiewicz studied the generation and propagation of solutions in idealized coastal flows (e.g., Straits of Messina, Yellow Sea, and mid-Atlantic Bight).  They used the Boussinesq option of the nonhydrostatic model EULAG, developed in MMM as a small-to-mesoscale atmospheric model.  One technical challenge posed to the model is to capture accurately flow response to a tidal forcing past a steep (in atmospheric terms) orography of the sea bottom typical of coastal regions with a characteristic abrupt transition between the continental shelf and deep ocean.  A series of idealized high-resolution (20 m and 1 m in the horizontal and vertical, respectively) 2D simulations of diuurnal flow evolution in Yellow Sea appears both revealing and promising.  In the absence of meaningful explicit subgrid-scale turbulence model, an implicit subgrid-scale model within the NFT approach - the algorithmic basis of EULAG - appears beneficial for high-Reynolds-number (LES type) simulations of breaking solutions.  Furthermore, a rigorously designed elliptic solver for pressure, free of splitting at vilinear boundaries, benefits simulations of coastal flows.  This indicates the viability of a unified modeling approach for atmospheric/oceanic flows.

B.  Cloud and Surface Processes Parameterizations

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.  The emphasis is on understanding how the moist atmosphere, land and ocean surface, and hydrological processes interact and how these processes can be quantified.  Nonhydrostatic fine-scale modeling using large-eddy simulation and cloud-resolving models is an important component of our approach to these problems since they allow high-resolution definition of the mesoscale and microscale systems involved, and are therefore a good means for testing methods to quantify the effects of these processes on larger scales.  Critical to the success of our program is the evaluation of these models against detailed observational studies of the underlying physical processes.  The goals of this program contribute to the objectives of the Global Change Research Program (GCRP) Global Energy and Water-cycle Experiment (GEWEX) and the Cloud System Study (GCSS) component of GEWEX.

1. Deep Convective Cloud Systems

a.  Cloud systems on long time scales

Cloud-radiation interaction over the tropical western Pacific

Cloud-resolving model (CRM) simulations of Tropical Ocean and Global Atmosphere Program Coupled Ocean-Atmosphere Response Experiment (TOGA COARE) cloud systems were used by Xiaoqing Wu and Mitchell Moncrieff to quantify uncertainties involving convection, cloud and radiation parameterizations in global circulation models and numerical weather prediction models.  Comparing the results from the CRM, a single-column model of the NCAR Community Climate Model version 3 (CCM3) (SCM) and TOGA COARE observational datasets, they found that the cloud scheme greatly under-predicts both cloud condensate and cloud fraction, required by the radiation scheme to match the observed top-of-atmosphere (TOA) radiative flux.  Using an off-line calculation of radiation with CRM-generated temperature, moisture and cloud fields they found that the radiation scheme under-predicts the TOA and surface radiative fluxes.  Also, the representation of cloud geometry association and inhomogeneity in the radiation scheme need to be improved in order to produce accurate radiative heating profiles and radiative fluxes.

Explicit ocean-atmosphere interaction on the mesoscale

Wu and Moncrieff used a CRM to quantify the effects of sea surface temperature (SST) variability, large-scale dynamics and surface fluxes on radiative-convective equilibrium.  They examined SST feedback using a CRM coupled to a one-dimensional ocean model.  The coupled system was run to a state of radiative-convective-dynamical equilibrium.  The effects of cloud systems on the equilibrium state were investigated by analyzing two 40-day simulations with different ice microphysics parameterizations.  Changes in sedimentation flux of ice particles and ice water content had large effects on the atmospheric and surface equilibrium states.  The equilibrium associated with smaller ice water content featured a warmer and moister atmosphere, more deep clouds and smaller SST than that associated with larger ice water content.  This work is now being used to improve the cloud-radiation parameterizations in CCM3, with particular regard to the cloud-overlap assumptions.

Explicit and parameterized realizations of oceanic convective cloud systems

Changhai Liu, Moncrieff and Wojciech Grabowski examined convection and cloud processes in a hierarchy of two-dimensional numerical realizations of cloud systems observed during the 19-26 December 1992 period of TOGA COARE.  The hierarchy consists of cloud-resolving simulations with a 2-km grid, and two sets of 15-km resolution simulations with parameterized convection.  By comparing with the cloud-resolving model, shortcomings in the Kain-Fritsch convection parameterization scheme were quantified:  (a) The entraining plumes in the parameterization excessively overshoot the tropopause producing a cold bias mostly through adiabatic cooling.  The attendant moisture detrainment overproduces cirrus. (b) Parameterized downdrafts cause a surface cold bias. (c) The scheme fails to represent the trimodal convection (cumulonimbus reaching the tropopause, cumulus congestus around the melting level, and shallow convection regimes) realized by the cloud-resolving simulation and also seen in observations.  The lack of shallow convection and cumulus congestus leads to an overprediction of the low-level moisture. (d) The simulations are sensitive to the magnitude of moisture feedback from the convective parameterization to the grid scale but less sensitive to whether the moisture is in vapor or condensate form.  These deficiencies are mostly a consequence of the single-plume entrainment model for updrafts and downdrafts used in the parameterization scheme.

Large-scale tropical circulations driven by convection and SST gradients

Jun-Ichi Yano (University of Hamburg, Germany), Moncrieff and Grabowski produced a two-dimensional theory of the interaction of moist convection waves with Walker-type mean circulations.  The drying and cooling of the boundary layer by convective and mesoscale downdrafts balances the wind-induced perturbations of surface fluxes (a boundary-layer entropy equilibrium).  The moist thermodynamic state affects the stability of the large-scale circulation by controlling the impact of wind perturbations on surface fluxes.  The mean circulation is unstable through positive-feedback among enhanced low-level convergence, latent heating and surface flux.  Linear instability selects the size of the ascending region in the mean circulation and suggests that a Walker-type quasi-steady state depends on the spontaneous generation of wave modes.  When the domain size is of the order of a few thousand kilometers, the first baroclinic-type circulation is too unstable to be maintained even in a quasi-steady sense.  A double-cell structure in the cloud-resolving simulations is generated.  The instability mechanism is distinct from other tropical large-scale instabilities (e.g., wave-CISK and WISHE) wherein the higher wave number modes are unstable.  When nonlinear advection is included, growing oscillatory modes are realized in qualitative agreement with observed spontaneous generation of convective wave modes.

Multi-scale organization of convection and intraseasonal tropical variability

Grabowski and Moncrieff investigated the large-scale organization of tropical deep convection in idealized two-dimensional 40-day cloud-resolving simulations.  The initial state had a constant 10 m/s easterly wind and a uniform sea surface temperature.  A prescribed temperature tendency mimics the mean radiative cooling of the tropical troposphere.  A 20,000-km computational domain allows interactions among moist convection, mesoscale organization and surface exchange on a wide range of scales.  The large-scale organization of convection and its eastward propagation resembles the supercluster/Madden-Julian Oscillation organization seen in satellite observations.  The large-scale organization is basically explained in terms of convectively coupled Kelvin wave dynamics.  Mesoscale convective systems organized on scales of several hundred kilometers move east-to-west within the envelope of convection.  Convective momentum transport and the effects of the convective systems on temperature and moisture near the surface are key processes.  The convective momentum transport was approximated by a nonlinear analytic theory of organized convection.

Coupling cloud processes with global dynamics using the cloud resolving convection parameterization (CRCP)

Grabowski applied the Cloud Resolving Convection Parameterization (CRCP) to couple cloud-scale processes to the large-scale dynamics using the anelastic nonhydrostatic global model EULAS developed by Piotr Smolarkiewicz and collaborators.  The CRCP technique applies a 2D cloud model to represent subgrid scale processes (e.g., cloud dynamics and microphysics) in each column of the global model.  The modeling setup is a radiative-convective equilibrium on a rotating constant SST aquaplanet.  In addition, a pilot simulation considering an aquaplanet with a realistic meridional SST distribution was performed.  Radiation transfer model was also added into the CRCP 2D model framework.  The global CRCP simulations feature pronounced large-scale organization of convection within the equatorial waveguide.  Prominent equatorial “super cloud clusters” develop in these idealized simulations which bear a strong resemblance to the Madden-Julian Oscillation observed in the terrestrial tropics.  Sensitivity experiments with 2D cloud models oriented either in the E-W or the N-S direction (and thus considering coupling between either zonal or meridional cloud-scale and large-scale wind components) demonstrate the importance of zonal momentum transport by mesoscale convective systems for the large-scale convection organization.  Simulations using larger horizontal resolution of the global model were performed to demonstrate robustness of the simulated MJO-like structures.

b.  Convectively generated tropical ice clouds

The properties of deep tropical stratiform clouds have been recently studied by Andrew Heymsfield, Aaron Bansemer, William Hall, James Dye and Jeffrey Stith of NCAR and Tony Grainger (University of North Dakota, UND).  They collected data from the UND Citation aircraft during the tropical rain measuring mission (TRMM) field programs in Florida, Brazil, and Kwajalein, Marshall Islands.  The microphysical measurements are probably the most complete set of in-situ data in subtropical and tropical clouds to date, as it includes particle size distributions and habit information over a range of sizes from 10 microns to 5 cm using recently developed instruments.

Flight patterns, involving slow spiral descents through cloud layers while drifting with the wind, were developed to provide information on how ice particles grow and then melt in layer clouds.  One example of this process is illustrated in figure 11 at the right.  The figure shows a vertical profile of particle size distributions through a cloud layer at Kwajalein.  Each horizontal line comprises a size distribution, which was generated by taking a two-dimensional plot of concentration versus maximum dimension and assigning colors to represent the concentrations at each size.  The figure shows how the size distribution broadens from cloud top downwards to approximately 4.3 km, as a result of aggregation.  These aggregates are formed and grow by collecting particles in the 100 to 400 micron size range; as a result, particles in this size range are depleted.  At the 4.3 km level, the particles encounter the 0°C level and begin to melt.  By the base of the melting layer, drop size distributions conform to the results from previous studies. Curves were fitted to the measured size distributions to develop a means of retrieving particle size distributions from radar data, either from the TRMM radar in space or from ground-based radars.  They observed a systematic variation in the fit parameters with height below cloud top.  Within the melting layer, these parameters also varied in a systematic way, and by the base of the melting layer approached the well-known Marshall-Palmer type drop size distribution.  As part of the research effort, they developed new software to process particle size distribution data.  This software is an important advance in the field, and is freely available to the community.

Figure 11 (click on graphic to view larger figure): Moist Baroclinic Wave Simulation. Height coordinate model (dx=100 km, dz=250 m, dt=600 s; Surface temperature, surface winds, cloud and rain water.

c.  Parameterizations for ice crystal cross-sectional area Accurate parameterizations for the cross-sectional area of ice crystals are important for the treatment of ice clouds in numerical models, determining of ice cloud properties from satellite and other remote-sensor measurements, and understanding fundamental processes of ice crystal growth and aggregation.  The cross-sectional area of an ice crystal influences its optical properties, and affects its mass/area ratio from which crystal fall speed and ice-mass flux are calculated.  Diameter and area are both measured by most particle imaging and particle collection instruments.  Larry Miloshevich and Heymsfield have analyzed observational data to develop parameterizations for ice crystal cross-sectional area as a function of crystal diameter.  Figure 12 to the right shows the parameterization in terms of the “area ratio,” which is the ratio of a crystal's area to the area of a circumscribed circle, an indicator of ice crystal shape and the degree of elongation and openness in its structure.  The parameterization shows that crystals are compact when small, but become elongated and more open as they grow.  The analysis was also performed in height intervals within the clouds, and the slope of the curve is found to decrease from cloud top downward, for reasons concerning fundamental processes of ice crystal growth and aggregation.  These results will allow substantial improvement in the treatment of cloud optical properties, ice mass, and vertical transport of ice in numerical models and in remote sensing techniques.

Figure 12 (click on graphic to view larger figure): Light curves show the ice crystal area ratio as a function of crystal size for different datasets, and the bold curve is an average of the asets. Small crystals are and compact, becoming more elongated and open as they grow, affecting crystal opti properties, fall speeds, and calculations of crystal mass.

d.  Possible technique for control of global warming

 In a highly preliminary independent study, John Latham is exploring in a more comprehensive and quantitative manner than hitherto the idea that it may be possible - in a controlled manner - to modify (enhance) the albedo of maritime cumulus clouds for incoming solar radiation, thereby, in principle, compensating for warming effects.  The technique involves production of fine sprays of seawater at the ocean surface.  Some fraction of the droplets, which are excellent CCN, would be carried up to cloud-base levels by convection.  If sufficiently numerous they would become activated preferentially, so that clouds (which in any event would have formed) would apportion their liquid water over more droplets, thereby increasing the cloud albedo.  More quantification and technological examination is required before it will be clear whether a field experiment is justified.

2.  Boundary Layer Flows

a.  Marine stratocumulus regime

Collaborating with Ruby Leung and Steven Ghan (Pacific Northwest National Laboratory; PNNL), Chin-Hoh Moeng continued to analyze large-eddy simulations (LES) of the marine stratocumulus PBL, searching for (1) a proper way to incorporate the countergradient effect into the Level 2.5 turbulence model used in the PNNL GCM; and (2) a better method to estimate the variances and covariances of the temperature and moisture fields, which are used to parameterize the cloud fraction and liquid water amount in many GCMs.

Entrainment of air from the overlying free atmosphere across the top of marine stratocumulus cloud decks is central to their subsequent evolution.  This has been extensively studied by LES, but confirming observational data has been difficult to obtain because of the difficulty in measuring the entrainment rate.  Recent developments in aircraft instrumentation and observational techniques have led to renewed interest in carrying out a definitive observational study focused on measuring entrainment rate, as well as the variables that control it.  Bjorn Stevens (University of California, Los Angeles), Donald Lenschow, Gabor Vali (University of Wyoming), Chris Bretherton (University of Washington), Alan Bandy (Drexel University) and Hermann Gerber (Gerber Scientific) are planning to carry out such a study, the Dynamics and Chemistry of Marine Stratocumulus (DYCOMS-II), using the NCAR C-130 aircraft in July 2001 off the coast of southern California.

b.  Fair weather cumulus

 Simulation of shallow cumulus over land

The PBL cloud case selected last year for the GEWEX (Global Energy and Water Cycle Experiment) Cloud System Study (GCSS) Working Group 1 intercomparison study was shallow cumulus over land.  The case was based on an idealization of observations from 21 June 1997 at the Southern Great Plains ARM site, in which the cloud field was strongly forced by the surface fluxes.  The simulated cloud field compared well among the eight LES groups that participated in this study.  It also agreed well with the observations.  Comparison between the stratocumulus-topped PBL and the cumulus-topped PBL (including both subcloud mixed-layer and cumulus layer) shows similar features in their turbulence flux profiles.  In both PBLs, the total heat and water fluxes are linear with height, from the surface up to near cloud top; and their buoyancy flux (if surface buoyancy flux is positive) usually decreases linearly with height in the subcloud layer, increases sharply at the cloud base, and then remains quite uniform in the cloud layer before it drops to zero just above cloud top.

c.  Studying the subgrid-scale motion of LES

The increasing use of large-eddy simulation (LES) in addressing PBL problems in which small-scale turbulence eddies become important (e.g., entrainment across the PBL top, second-order chemical reactions, and interactions with the Earth's surface) compels us to reevaluate how accurately subgrid-scale (SGS) eddies are parameterized in LES.  To examine SGS motions in the surface layer of the PBL, Moeng, Peter Sullivan and Lenschow worked with Tom Horst and Steve Oncley (ATD), Jeff Weil (CIRES, MMM visitor), Charles Meneveau, Marc Parlange, and Jan Kleissl (Johns Hopkins University), John Wyngaard (Penn State University), and Bjorn Stevens and Jianjun Duan (University of California, Los Angeles), to carry out a field study (SGS-2000) over a flat, uniform surface in the San Joaquin Valley of California.  Data were collected during September 2000 using an array of fourteen 3-dimensional sonic anemometers at two levels oriented approximately normal to the prevailing wind.  The velocity and temperature data will be used to decompose the turbulence velocities into a 2-dimensional filtered  (i.e., resolved-scale) field and its SGS fluctuations by applying a weighted average over the data measured along the sonic array for the cross-wind filter and a low-pass filter over time for the stream wise filter (via Taylor's hypothesis).  They intend to use the filtered and SGS fields as surrogates for the LES resolved and SGS fields to improve our understanding of the relationship between resolved and SGS eddies in LES.

Prior to the field experiment, Moeng and Sullivan compared the above filters against the standard spatial filters (both sharp wavelength cutoff and Gaussian) used in LES.  A time series of turbulent motions from a high-resolution LES (where the horizontal and vertical grid spacings are about 6 m and 4.5 m, respectively) was used to mimic the field measurement setup.  They found: 1) the surrogate filter mimics a Gaussian filter much better than a sharp wave-cutoff filter; 2) the sharp wave-cutoff filter, which is currently used in the NCAR LES code, can produce negative SGS variances and energy and hence violates reliability conditions; and 3) The instantaneous Leonard term is a large term in the SGS stress for both the sharp wave-cutoff and Gaussian filters.  The current NCAR LES code incorrectly assumes the Leonard term is zero for a sharp wave cutoff filter.  The Leonard and cross terms are comparable in magnitude but tend to have opposite signs, and hence tend to cancel.  Consequently, the SGS-SGS interaction term of the stress, albeit smaller in magnitude compared to the Leonard and cross terms, becomes important in determining the net SGS stress.

Sullivan, collaborating with MMM visitors Berengere Dubrulle (Chargee de Recherche at CNRS), Jean-Phillipe Laval (University of California, Los Angeles), and Joseph Werne (Colorado Research Associates) began evaluation of a new subgrid scale model using high resolution LES and direct numerical simulation (DNS) datasets.  The proposed model differs from conventional SGS approaches in that the scheme attempts to estimate the subgrid scale velocity instead of the subgrid scale flux.  An important feature of the model is that no closure constants are involved; the scheme follows directly from the governing equations.  The model assumes that the non-linear interaction terms in the subgrid scale velocity equation are negligible compared to the resolved-resolved and resolved-subgrid interactions and thus is a generalization of rapid-distortion theory (this is refered to as non-locality, i.e., the small scales dynamics are primarily determined by the larger scales).  The assumptions of the model are substantiated with apriori tests using LES of free-convection and shear-buoyancy driven PBLs.  The model will be tested using data from SGS-2000 and compared with direct numerical simulations of stably stratified shear flow to determine the feasibility of incorporating it into LES codes.

d.  Clear-air boundary layers

The stable boundary layer

A stable boundary layer experiment Cooperative Atmosphere Surface Exchange Study (CASES-99), was successfully carried out in Kansas in October 1999.  Jielun Sun and Sean Burns, with help from Xuhui Lee (Yale University), deployed thermocouples at 34 levels on the main tower.  Thermocouple temperatures for the entire field experiment, processed by Burns, are now available.  Sun and Burns have started to analyze the CASES-99 data in collaboration with other CASES-99 investigators.  Preliminary analysis by Sun, Burns, Lenschow, and Gerrit Oosterhuis (Wageningen University, Netherlands) demonstrate the complexity of the stable boundary layer.  They found that interactions between the thermal and shear instability could lead to turbulence that is both spatially and temporally intermittent.  Sun and Burns are also working with NCAR/ATD and others in the CASES-99 community on estimating radiative flux divergence, which they found to be at a maximum in the early evening, at around 10 Wm-2.  Strong turbulent mixing can strongly reduce the radiative flux divergence, as expected.

Last year, Margaret LeMone reported finding a strong linear relationship between half-hour averaged temperature at 2 m and elevation during the early morning hours.  CASES-97 data indicated that the elevation dependence corresponded to about half the vertical gradient as measured by radiosondes released during the same half-hour period.  If the Froude number was sufficiently high, the potential temperature was constant.  This year, Kyoko Ikeda and LeMone looked at data through the night, and found that it took several hours for the linear relationship to develop.  They hypothesized that the minimum time for the relationship to develop can be defined in terms of advection, i.e., a time scale defined by L/U, where L is the length-scale of the CASES-97 array, and U is the wind speed.

Evening transition

Under LeMone’s supervision, Andrew Church, a SOARS student, studied the evolution of the wind in the late afternoon on three (Intensive Operations Period) IOP days, to see what characteristics distinguished a day with early-onset wind veering (identified by Julie Lundquist of the University of Colorado) from two others for which the boundary-layer winds remained steady and well-mixed until around sunset as is more commonly observed.  Typically, veering is associated with decoupling of the boundary layer from the surface as the effects of friction are decreased.  Church and LeMone looked at both vertical velocity variance and turbulence kinetic energy (TKE) as indicators of boundary layer mixing.  They found that mixing (as defined by TKE) on one of the ‘normal’ veering days remained vigorous until sunset, but could not distinguish between the two other days.

Lagrangian modeling of dispersion in the PBL

In collaboration with Moeng and Sullivan, Jeffrey Weil (University of Colorado) is modeling scalar dispersion in the convective boundary layer (CBL) using velocity fields from LES.  The LES fields drive a Lagrangian “particle” model of a passive scalar, for which the mean concentration field is obtained from the probability distribution of the particles.  Most recently, they simulated dispersion from a surface and near-surface point source, with stability ranging from strong to weak convection.  The results showed a decrease in the dispersion rate with decreasing stability as a result of the smaller turbulence time and length scales.  For the mean concentration field, the main result was a slight increase in the peak surface concentration and an extension of the high concentration region to greater distances downstream of the source.  Although limited in scope and number, field observations have corroborated the results, and in particular, the variation of the surface concentration distribution with stability and distance.  The Lagrangian modeling fills an important gap in understanding dispersion due to the very limited number of and scatter in plume observations for surface and near-surface releases.

3. Surface-Atmosphere Interactions

a.  Land-atmosphere interactions Sullivan, Edward Patton (Pennsylvania State University), and Moeng coupled their large-eddy simulation code for the clear PBL to the NOAA (National Center for Environmental Prediction / Oregon State University / Air Force / Office of Hydrology) land-surface model (LSM).  The coupled LES-LSM code was developed around the Message Passing Interface (MPI) so as to take advantage of the capabilities of massively parallel computer architectures (simulations utilizing more than 100 processors have been carried out).  Our first simulations focused on the interaction between the free-convection (no mean wind) boundary layer and land surfaces with heterogeneous soil moisture content as a function of incoming solar radiation in 3D domains of size (30 x 5 x 2) km with resolution (50 x 50 x 20) m.  The preliminary results show that the presence or absence of soil moisture strongly modifies the components of the surface energy balance (sensible and latent atmospheric heat fluxes, and soil heat flux) which in turn alters the surface forcing of the PBL (see figure 12 at right). For the free convection cases, simulated complex boundary layer circulations are generated near the transition from high to low soil moisture content.  The horizontal extent of these boundary layer circulations is asymmetric with respect to the soil moisture transition and typically increases with decreasing solar radiation.

Figure 12 (click on graphic to view larger figure): Large-eddy simulation of a coupled land-atmosphere system. Response of the atmospheric boundary layer to heterogeneous soil moisture. The dramatic changes in boundary layer structure result from the non-linear dependence of soil properties on soil moisture.

Turbulent fluxes are most commonly measured using aircraft over limited areas and time periods, and towers at fixed locations.  However, area-averaged turbulent fluxes are required to improve model physics in numerical models.  Jielun Sun combined tower and aircraft flux data with satellite and meteorological mesonet data collected during BOREAS to develop regression formulae for sensible and latent heat fluxes for various land classifications.  The improved formulae are based on observed aircraft or tower fluxes, and external variables such as the downward solar and longwave radiation, wind, and satellite-derived land classifications.  By using the empirical formulae for the land classification map over the BOREAS southern and northern study areas, sensible and latent heat flux maps can be constructed at the spatial resolution of the satellite, then area-averaged for comparison with model predictions.  Collaborators included Larry Mahrt (Oregon State University), Forrest Hall (NASA Goddard Space Flight Center), Jing Chen (Canada Centre for Remote Sensing), Valentijn Pauwels (Princeton University), Harry McCaughey (Queen's University, Canada), Alan Betts (Atmospheric Research), and Alan Barr (Atmospheric Environmental Service, Canada).

The Southern Great Plains (SGP) experiment provided atmospheric and hydrological data simultaneously over a variety of surface types.  By studying the atmospheric response to the spatial variation of soil moisture, Sun, in collaboration with Mahrt, Ian MacPherson (National Research Council, Canada), Ron Dobosy (NOAA), Jay Famiglietti  (University of Texas at Austin), and Tom Jackson and Bill Kustas (USDA ARS Hydrology Lab), found that atmospheric moisture flux responds to the spatial variation of surface type during the early morning if the wind is not strong.  This spatial response stops when the turbulence decays in the late afternoon.  Large turbulence eddies around noon and early afternoon can also reduce the influence of the spatial variation of surface type on the atmospheric moisture flux.  The spatial variation of the atmospheric moisture flux was found to be more sensitive to the spatial variation of vegetation than to that of soil moisture.  The strength of the atmospheric moisture flux may be inversely related to the soil moisture if the atmospheric moisture flux is dominated by the turbulence intensity or buoyancy flux.  The relationship between the responses of the atmospheric boundary layer to the spatial variation of soil moisture will be further examined over additional flight tracks.

Observational and modeling studies to characterize the role of surface heterogeneity on boundary layer structure and its diurnal evolution have been conducted based on the CASES-97 dataset.  The CASES-97 array, located on the lower Walnut River Watershed east of Wichita, Kansas, consisted of a 60-km triangle with 915-MHz wind profile/minisodar data at the vertices.  Within the triangle measurements included: surface energy fluxes, soil moisture, and temperatures from 10 stations located according to land use and elevation; meteorological data from the flux stations and three additional stations; precipitation data from the surface array (an additional array set up by Oregon State University); the Wichita WSR-88D radar; and the NCAR S-Pol radar.  During intensive observing periods, the Wyoming King Air and NOAA Twin Otter flew patterns to determine flux and mean profiles, boundary layer height, and horizontal gradients; and radiosondes were released each 90 minutes from the three-profiler sites.  In addition to ongoing work with the CASES-97 and -99 data sets, LeMone, William Blumen and Robert Grossman (University of Colorado), have been holding meetings with colleagues in the Boulder area to discuss the next CASES field programs.

LeMone and Ikeda, along with Robert Grossman (University of Colorado), Wayne Angevine and Stuart McKeen (NOAA Aeronomy Lab), Bob McMillen (NOAA/Oak Ridge), and  Kuo-Nan Liou and Steven Ou (Univesity of California, Los Angeles) are in the final stages of completing temperature and humidity budgets for the late-morning boundary layer using surface, aircraft, radiosonde, and synoptic data for three CASES-97 Intensive Observing Periods with contrasting surface characteristics.  They have:

(a) Performed error analyses on the major budget terms  (time-tendency, vertical flux divergence, and horizontal advection).  Because of uncertainty in the advection term, they are also estimating gradients from NCEP Eta and RUC model products.

(b) Computed the radiative heating by applying the Fu-Liou algorithm to the 1530, 1700, and 1830 UTC soundings from the vertices of the experimental domain assuming reasonable concentrations of trace gases and no aerosols.  They added a correction for aerosols (due to field burning) by inserting a range of reasonable visibilities in an algorithm developed by McKeen (NOAA). Estimated radiative heating is of the order of 0.1 K/hr, with slightly larger heating rates for the two earlier days, for which aircraft videos indicated poorer visibility.

(c) Used wind and temperature data from the stacks of flux legs to show that mesoscale eddies were present on 10 May, when air trajectories were along contours of land use.  Potential temperatures in the lower PBL reached a maximum over the eastern edge of the watershed, and flight legs at all levels in the lower-mid PBL revealed along-track convergence consistent with a vertical velocity of up to 5 cm/s at the mixed-layer top, in the presence of a subsidence above the PBL of about 2 cm/s, indicated by descent of features in the soundings.  This circulation probably contributed to the relatively large error in the budgets on this day.

Fei Chen and David Yates (RAP), LeMone, Ikeda, Grossman (University of Colorado), and Haruyasu Nagai (Japanese Atomic Energy Commission) compared aircraft-derived fluxes of sensible  and latent heat to those predicted from three land surface models (LSMs): the Oregon State (OSU) LSM, the LSM developed by Nagai (SOLVEG), and the NCAR LSM.  Ikeda and LeMone processed the aircraft data, by first smoothing the flux time series for all low-level (~30 m) flight legs, and then averaging the fluxes at common longitudes to obtain fluxes over the average leg position in the north-south direction, with a spacing of about 1 km. Chen, Yates, and Nagai combined surface, remotely sensed atmospheric, precipitation, and surface data to force each LSM to generate a month-long 1-km gridded surface-flux dataset over the 71 x 74 km² CASES-97 area.  These data were used to generate surface fluxes beneath the morning flight tracks using positions and times from Ikeda and LeMone, and then averaged for the morning hours.  The model and aircraft fluxes showed a roughly similar pattern, with higher humidity fluxes over the cropped areas than over the pasture areas, particularly early in the experiment when the grass was dormant.  However, the correlation between aircraft and surface fluxes was quite low, as compared to the correlation between model fluxes and surface observations.

LeMone examined the individual and averaged smoothed flux legs to show that the poor correlation between the observed aircraft fluxes and the modeled surface fluxes is partially an observational sampling problem, but also appears to reflect atmospheric processes systematically concentrating energy flux.  It has long been known that large eddies in the unstable PBL collect rising plumes in their upwelling regions, leading to greater turbulence levels and flux than elsewhere.  Under such circumstances, sensible and latent heat fluxes are positively correlated, and their sum can exceed what would be expected from the local energy budget.  On the other hand, surface fluxes (and the fluxes predicted by the LSMs) are constrained by the surface energy budget, leading to a negative correlation between latent and sensible heat fluxes.  The smoothed fluxes for the lowest level flight legs reflect both processes:  on the day with the most variation in surface properties, sensible and latent heat flux were negatively correlated, while on the day with relatively uniform surface processes, the fluxes were positively correlated.  Averaging over all the morning flight legs for each day made the correlation less positive on all three days but did not eliminate positive correlation entirely on some days, because some maxima and minima were persistent.  Grossman and Yates are examining aircraft videos and surface data to understand the reasons for these persistent patterns, which were present even without obvious mesoscale circulations.  They have found a broad, flat area on the east side of the watershed that is systematically associated with larger-than-average sensible and latent heat fluxes.

Nicole Molders (University of Leipzig), Jimy Dudhia, Chen, and LeMone used CASES-97 data to compare the boundary layer response in the coupled MM5 modeling system to two other models: the OSU LSM, and the German HTSVS LSM.  Preliminary results show that the CASES-97 data are valuable to help to understand the simulated land-atmosphere interactions and to further improve the land-surface models.

b.  Ocean-atmosphere interaction

Sullivan, James McWilliams (University of California, Los Angeles), and Moeng continued to investigate air-sea interaction using direct numerical simulations.  Building on their previous results for neutrally stratified flow, they began examining the interaction of stratified turbulence and surface gravity waves.  In this study, both unstable (hot waves) and stable (cold waves) stratification are considered.  They found, as expected, higher turbulence levels and surface form drag for unstable stratification compared to the neutral case, but with a clear dependence on the wave speed.  An important question is whether the friction velocity is a better scaling parameter than the commonly used mean wind speed at 10 m for wave growth.  For the ideal cases considered in the simulations, friction velocity was a better scaling parameter over the entire range of atmospheric stabilities.

Momentum transport across the air-sea interface in coastal zones was studied by Sun and Burns as part of the Shoaling Wave Experiment (SHOWEX) off the coast of Duck, NC using the NOAA LongEZ aircraft in collaboration with Tim Crawford, Jerry Crescenti, and Jeff French (NOAA Air Resources Laboratory), Ed Dumas and Chris Vogel (NOAA ATDD), Douglas Vandemark (NASA Goddard Space Flight Center), Pierre Mourad (University of Washington), and Larry Mahrt (Oregon State University).  Sun was involved in designing aircraft flights and coordinating joint flights with the SURPAS TwinOtter aircraft.  Both Sun and Burns have been analyzing the aircraft data, particularly momentum transport in the coastal zone and the wave spectra derived from three laser altimeters aboard the LongEZ.

This is the first dataset where both atmospheric and oceanic measurements were obtained as functions of distance from shore.  The results show that the spatial variation of the friction velocity with offshore distance is much larger with offshore flow than with on-shore flow.  With on-shore flow, the friction velocity is strongly correlated with surface waves.  In addition, the variation of the neutral drag coefficient is well correlated with the atmospheric bulk Richardson number.  With offshore flow, the observed momentum flux significantly decreases with offshore distance.  Within the first few km offshore, the relationship between the friction velocity and the mean square slope of the short waves, and the relationship between the neutral drag coefficient and the atmospheric bulk Richardson number are obscured by the direct influence of the upstream land surface on the measured turbulence.  These relationships for offshore flow agree well with those for on-shore flow if the fetch is beyond the immediate influence of the land surface.  The results in this study suggest that the effects of the strong turbulence advected over the ocean from over the nearby land may lead to ambiguous physical interpretation of the correlation between the momentum flux and wave state.

By using three laser altimeters in a triangle aboard the aircraft, the 2-dimensional wave spectra over a region can be derived from the laser distance measurements.  The Doppler effect, which can be removed from the wave spectra by flying two legs on opposite headings, allows measurement of the phase speed of the waves.  Phase speed cannot be obtained from traditional wave measurements from buoys.  Sun and Burns found that this technique captures not only relatively long swells, but also short wind-generated surface waves (~2m).  The aircraft-derived wavelengths of the swell compared well with buoy observations in the same area.  In addition, the aircraft-derived phase speed agreed with the phase speed calculated from the dispersion relationship.

c.  Chemical transports and transformations

Large eddy simulations of chemical species in the convective boundary layer, have been performed by Sullivan, Mary Barth, and Edward Patton and Kenneth Davis (Pennsylvania State University) to examine how boundary layer dynamics, plant canopies, and heterogeneously spaced emissions influence the distribution of chemical constituents, in particular, isoprene.  Results from these numerical experiments indicate the following:

1) The plant canopy does not affect top-down mixing in the convective boundary layer.  However, bottom-up mixing results in smaller gradient and variance functions (i.e., enhanced mixing) at locations up to three times the canopy height.  The enhanced mixing results from the formation of canopy-scale structures thought to be generated by the elevated velocity shear generated near the top of the plants. 2) Enhanced mixing in the plant canopy, in which species such as isoprene are emitted, decreases the amount of isoprene transported to regions above the canopy because the efficient mixing allows for reaction with the hydroxyl radical to occur uninhibited by turbulent segregation. 3) When more complex chemistry is considered for the PBL without a plant canopy, very little segregation occurs between isoprene and hydroxyl radical, a reaction that plays an important role in ozone production.  A preliminary analysis of the covariance budget showed that the reaction of NO and HO2 to produce OH coupled with positive covariance between NO and isoprene reduces the segregation between isoprene and OH. 4) Imposed large-scale spatial segregation by horizontally heterogeneous source distributions with a plant canopy are found to decrease dramatically average reaction rates (see figure 13 at right).

Figure 13 (click on graphic to view larger figure): (A): A plan view of the four emission schemes for the comparison of homogeneous versus heterogeneous emissions. (B): Vertical profiles of the intensity of segregation derived from large-eddy simulations of the PBL transporting simple second-order chemical species. Compare cases of horizontally homogeneous emission schemes with and without a plant canopy and cases that are influenced by the canopy but with horizontally heterogeneous emissions according to (A).

There are now available many datasets of aircraft measurements of trace gas species that have been collected throughout the troposphere over all parts of the Earth.  These measurements have been analyzed statistically, and the results show that the variances of concentrations for non-conserved species with a surface source or sink are related to the lifetimes of the species.  The relationship that has been found observationally is that the standard deviation of species concentration is typically approximately inversely proportional to the square root of the species lifetime.  Lenschow has developed a simple analytical global model of diffusion that predicts this power law, and also predicts the scaling coefficient in this relationship, in reasonable agreement with the observations.  This offers a tool for predicting the lifetimes of trace species whose concentrations can be measured, but whose sources and sinks are poorly known.

d.  Wildfire Research With Larry Radke and Craig Walther (ATD), Terry Clark and Janice Coen completed analysis of airborne data of wildland fires obtained during WiFE (WildFire Experiment).  An upcoming paper in the Canadian Journal of Remote Sensing summarizes this first suite of airborne instruments gathered to observe the dynamics of wildland fires, which have generally been hidden by thick smoke until these infrared observations.  The images obtained with the Thermacam, NCAR's infrared imager, provided some striking examples of explosive wildfire behavior, and suggested that fire spreads in more dynamic ways than previously believed.  Six Thermacam images (see figure 14 at right) taken from the NSF/NCAR C130 while flying near Glacier National Park, spaced 0.75 sec. apart, show how a flame-filled finger bursts forward from the fireline about 100 m along the ground at approximately 100 miles per hour before collapsing, leaving hints of ignited materials in its wake. Clark and Coen have been analyzing infrared video imagery collected from two perspectives during FROSTFIRE, a prescribed burn in the boreal forest outside Fairbanks, Alaska.  Clark was onboard the U.S. Forest Service Piper Navajo with colleagues Larry Radke (ATD) and Bob Higgins (NASA Ames), while Coen was operating a similar infrared imager on the ground below, with University of Colorado colleagues Shankar Mahalingam, John Daily, and Yottana Khunatorn.  After developing and applying more advanced techniques to remove aircraft motions, Clark has been examining the airborne data collected during periods of dramatic runs in the fires below, once when the fire leapt across a 100 m wide clear strip expected to retain it.  Coen's perspective of these fires from a mountaintop site across the valley has revealed other examples of forward bursts from the fire as it raced up the slopes.  These sequences are giving unique insights into the mechanisms by which fires propagate, and demonstrate that the image registration (i.e. identifying the same spatial position in each image frame) has proven successful enough to next use the IR images to extract fire winds on moving platforms using image flow analysis (see more information at NCAR/MMM's Infrared Data Collection and Analysis web site).

Figure 14 (click on graphic to view larger figure): A time sequence of 6 images (spaced 0.75 sec apart) of infrared imagery from the McDonald Creek fire, observed during WiFE. Brightest color (yellow) is the hottest.

Working with Don Latham (U.S.  Forest Service Fire Sciences Laboratory), Clark and Coen continued advances in NCAR's coupled atmosphere-fire model and applied them in simulations that showed that atmospheric and fire-induced motions are inseparable.  While atmospheric wind and temperature structure can have a big impact on fire behavior, conversely, fire-induced winds are also a crucial factor in fire development.  Some conditions can lead to blowups, with rapid intensification of spread rate and heat output, so much that the fire winds overwhelm those in its surroundings.  Three-dimensional animated visualizations of these modeling animations produced with Don Middleton (SCD) greatly improved their insight into these simulations, revealing organized motions that would otherwise have been difficult to see (see more information and simulations at the NCAR/MMM Coupled Fire-Atmosphere Modeling web site).

Working with Michael Reeder and Morwenna Griffiths at Monash University in Melbourne, Australia, Clark continued the numerical modeling of Victoria fires.  Clark and his Australian colleagues made significant progress on their simulation of a 1997 fire in the Dandenong Mountains near Melbourne.  Their work uses the Bureau of Meteorology data for initial and boundary conditions.  This case study was started during Clark's visit to Monash in summer 2000. This is the first real-fire application of the new fire code, which after a few minor fixes worked exceedingly well.

4.  Chemistry, Aerosols, and Dynamics Interactions Research

The foci of studies on chemistry, aerosols, and dynamics interactions are to examine the effect of physical and dynamical processes on chemical species and to study the effect of chemistry on aerosols and cloud condensation nuclei.  Ongoing projects within the program include observational and numerical analyses of the Stratosphere-Troposphere Experiments: Radiation, Aerosols and Ozone (STERAO)-Deep Convection experiment datasets, observational and numerical analysis of the Indian Ocean Experiment (INDOEX), and aerosol and cloud chemistry process studies.

a.  STERAO project

Lightning

The STERAO-Deep Convection experiment, which was conducted during the summer of 1996, had the major goals of investigating NO_x production by lightning and transport of chemical constituents by thunderstorms.  Work in the Chemistry, Aerosol, and Dynamics Interactions Program associated with the STERAO experiment includes analysis of the electricity and storm dynamics, synthesis of the numerical model results and analysis work to determine the production of NO_x from lightning, numerical simulations of chemical constituents, and calculation of photolysis frequencies in and near the storm.

Eric Defer and Dye continued their analysis of the lightning data, and airborne and ground based Doppler radar measurements recorded during STERAO-A.  From a flash-by-flash analysis for the 10 July 1996 storm, they observed that short-duration flashes (duration < 1 m/s) were located in cells where high reflectivity was recorded at high altitude and in regions of >10 m/s vertical velocity.  This same analysis is currently being performed by Defer and Jessica Hagan for other STERAO-A storms.  Defer and Dye developed a new technique to distinguish the negative cloud-to-ground flashes from the other flashes based on the typical signature of VHF radiation observed for the downward negative stepped leader-return stroke process.  This technique is being applied to several STERAO storms as well as storms observed in other field campaigns.  Defer also studied the simultaneous measurements of the cloud-to-ground flashes recorded by the National Lightning Detection Network (NLDN) and the ONERA interferometric mapper.  For the 10 July 1996 STERAO-A storm, Defer concluded that both instruments reported time and spatial locations consistently for the cloud-to-ground flash population.  Defer noted for the 10 July 19996 storm that the NLDN was characterized by 100% detection efficiency for negative cloud-to-ground flashes.

Almost all previous studies of NO_x production by lightning have estimated NO_x production on a per flash basis, yet it is known that there is large variability from flash to flash in terms of energy, peak currents, and time duration.  For example, in the 10 and 12 July storms, time durations range from <100 us to >1 s.  In order to better account for these flash to flash variations and provide a more physically based estimate of NO_x production, Defer has used interferometer measurements and knowledge of characteristics of different lightning processes to derive estimates of total flash length.  This new approach gives for the first time estimates of total discharge lengths within a flash. For the 10 July storm Defer estimates total path lengths per flash ranging from 0.5 to 1500 km or 15 to 300 km per minute when the variation in flash rate and flash types are considered.

NO_x fluxes

William Skamarock, Barth, and Dye calculated an NO_x (NO+NO₂) budget of the 10 July storm by augmenting the observations with numerical simulations.  For the 10 July storm, the North Dakota Citation aircraft traversed the anvil outflow at many levels. Cross sections of NO, CO, ozone and other species, along with their fluxes, have been objectively analyzed.  By using the simulation to determine the convective flux of non-lightning-produced NO and Defer's estimates of flash lengths, estimates of NO_x produced by lightning were constructed from the objective analyses (6.8x10²⁰ molecules per meter flash channel length).  These agree well with earlier estimates by Stith and Dye (between 2 x 10²⁰ and 10²² molecules/m) using a completely different approach.

Chemical species transport and photolysis Barth, Amy Stuart (Stanford University), and Skamarock used numerical simulations to calculate the amount of mass transported to the anvil region and the scavenging efficiencies of soluble tracers for the 10 July 1996 storm.  They found that the mass change in the anvil region for a highly soluble tracer is -0.5 x 105 to 0 kg for a 3-hour period, while for a passive scalar that has the same initial concentration as the soluble tracers the mass change was 2.3 x 105 kg.  Highly soluble scalars (Henry's Law coefficient > 105 Molar/atm) had scavenging efficiencies of greater than 55%. Numerical simulations of the 10 July 1996 STERAO deep convective storm coupled with gas and aqueous phase chemistry by Barth show that the concentration of low-soluble species, such as O3, NO, and NO2, were controlled by transport and gas-phase chemistry (see figure 15 at right).   The aqueous chemistry was found to be important only for SO2 and HCHO depletion.  The concentration of moderately to highly soluble species, e.g., HCHO, H2O2, and HNO3, was substantially reduced in the anvil region when ice, snow, and hail captured these soluble species from the liquid drops.  An analysis of the chemical and physical processes affecting the chemical species will be performed next. Skamarock and Barth collaborated with Richard Ramaroson, A.-L. Brasseur, and A. Delannoy (all from ONERA) to determine the effect of cloud particles on the actinic flux and photolysis frequencies.  They used 10 July 1996 STERAO model output of cloud water, rain, ice, snow, and hail mixing ratios for spectral radiation calculations.  Results show an enhancement of actinic flux of 2 to 5 compared to clear air values near the top edge of the storm.  The enhanced actinic flux increases the photolysis frequencies of O3, NO2, H2O2, and HCHO by 2 to 2.5 times in this same region.  The shadow effect of the cloud reduced photolysis frequencies by 50-80%.

Figure 15 (click on graphic to view larger figure)

b.  INDOEX

Greg McFarquhar and Heymsfield examined how anthropogenic aerosols from the Indian subcontinent affected cloud microphysical properties using data from the Indian Ocean Experiment (INDOEX).  They showed that regimes with many condensation nuclei (CN), i.e., polluted regions, contained almost three times as many small droplets as compared to the low CN (clean) regimes.  Conversely, in the clean regimes mean droplet diameters were 33% larger in the polluted regimes, and more frequent drizzle was measured.  However, in both regions bulk cloud properties such as liquid water content, vertical velocity, and cloud horizontal dimensions were similar.  Studies that parameterized the cloud microphysical properties in terms of bulk variables were used to estimate potential indirect radiative effects associated with the emission of anthropogenic pollutants from the Indian subcontinent due to changing sizes of cloud particles.  Future studies will further examine effects of aerosols and entrainment on drizzle suppression and cloud radiative properties, possibly through participation in ACE-Asia (Asian-Pacific Regional Aerosol Characterization Experiment).

c.  Aerosol and Cloud Chemistry Process Studies

Patrick Chuang (ASP postdoctoral fellow) participated in a field experiment to measure the time scale for condensational growth of aerosol particles.  It has been hypothesized for at least 15 years that there exist particles on which water uptake is inhibited because the particles are coated with an organic film.  However, little effort to directly study this hypothesis for atmospheric aerosols has been made, although circumstantial evidence suggests that the hypothesis is plausible.  If such particles do exist, they would need to be accounted for in our understanding of cloud microphysics (and therefore affect the way anthropogenic aerosols impact climate change), as well as the respiratory effects of fine particles.   Chuang, Darrel Baumgardner (NCAR and Universidad Nacional Autonoma de Mexico, UNAM) and Al Cooper (NCAR) who both provided equipment, Mike Hannigan  (University of Denver) and Graciela Raga (UNAM) who both made aerosol filter measurements, and Cristina Facchini and Sandro Fuzzi (both at Institute of Atmospheric  and Oceanic Sciences, Consiglio Nazionale delle Ricerche, Italy) who made aerosol surface tension measurements, conducted the field experiment in September 2000.  Preliminary findings indicate that during three days (25-27 September) such particles were found to be very rare.

Barth and Sonia Kreidenweis (Colorado State University) led the Cloud Chemistry Case for the Cloud Modeling Workshop that was held in August 2000. Barth and Rynda Hudman (SOARS student, San Jose State) participated in the intercomparison of chemistry solvers for gas and aqueous chemistry with two different box model solution techniques.  Barth also helped Susan Durlak (NCAR/ASP) with the development of cloud chemistry in the NCAR aerosol box model (MAPS), which participated in the aerosol parcel model comparison of the cloud chemistry case.  The summary of this case is currently being documented by Barth and Kreidenweis.

FOCUSED SCIENCE PROGRAMS

The MMM Division contributes to and receives directed funding for several of the high-priority national programs identified in the ATM-UCAR long-range plans.  These programs are the (A) Environment, (B) U. S. Global Change Research Program (USGCRP), and (C) High Performance Computing and Communication (HPCC).  Descriptions of MMM's contributions to these programs are outlined below.

A. Environment

MMM plays a key role in the U. S. Weather Research Program (USWRP) by hosting the office of the USWRP Lead Scientist and the chairman for the World Weather Research Program (WWRP), and through active participation in the NCAR USWRP Program.

1. U. S. Weather Research Program (USWRP)

During the past year, the Lead Scientist position was held by Robert Gall.  The role of the Lead Scientist is to assure the quality and relevance of the USWRP science through the development of short- and long-range scientific goals and objectives, to plan meetings and workshops needed to accomplish these goals, and to evaluate results and coordinate with various external bodies such as the World Weather Research Program (WWRP).

Activities during 2000 were divided into four broad categories:

Administering and evaluating research, forecast demonstration, and training projects.
Organizing workshops and symposia.
Establishing cooperative and collaborative projects, as appropriate, with programs of the WCRP and the WGNE.
Advising WMO Members about weather prediction research.

Activities conducted in FY 00 through the office of the Lead Scientist included

Maintain the USWRP Website at a .25 FTE level (currently Sherrie Fredrick).  This year the web site was completely revamped.  A web designer was consulted to provide a new look and new organization to the site.  The possibility of changing the URL of the site to something very simple has been explored, the proposal is USWRP.org.
Publish and post reports.  The PDT 9 and 10 reports were completed this year and have appeared in the Bulletin of the AMS.  In addition, a “Vision Document” that outlines the full USWRP Plan and its structure is in preparation and nearing completion.  It also will be bound and available as needed.  All the documents mentioned here are available on the web site, http://uswrp.ucar.edu/uswrp.html.
Develop plans to facilitate the transfer of research results into operations.  A workshop on “The Research Needs of the Private Sector” was held in Palm Springs, 29 November - 1 December 2000.
Complete implementation plans.  The two implementation plans have now been completed.  The focus of the two Associate Lead Scientist positions will now be on assessing the progress made by the program toward achieving the plans laid out in the implementation plans.
Conduct meetings of the USWRP/SSC.  There was a Retreat of the IWG held in Columbia, Maryland, 9-10 May.  The Fall USWRP/SSC meeting was held in Washington, DC, 1-4 October.  The primary theme of that SSC meeting was to examine the possibility of expanding the foci of the Program beyond the three current themes; Hurricane Landfall, Quantitative Precipitation Forecasting and the Optimal Mix of Observations.  One result of this discussion was a recommendation that the OLS conduct a Prospectus Development Team meeting on Air quality.
Conduct and administratively support the 2^nd USWRP Science Symposium.  The 2^nd USWRP Science Symposium was held in Boulder, 27-29 March 2000.  It was decided that this symposium would focus on the DA/QPF/OM theme of the Program.  This coming year the main focus would be Hurricane Landfall and that in the future the main theme of the symposium would alternate between these two themes.
Facilitate the Conduit Working Group.  This group held its first meeting at NASA Goddard on 10 December, 1999. 
Continue to support the Impacts and Use Assessment Committee, IUAC.  The committee met during the SSC meeting in Washington, DC, 1-4, October, 2000.  The IUAC will be charged this year with developing an Implementation Plan for the societal impacts component of the USWRP. This plan will be in parallel to the Implementation Plans for Hurricane Landfall and DA/QPF/OM.
Begin development of plans for field programs outlined in the Hurricane Landfall and DA/QPF/Optimal Mix Implementation Plans.  This included:

• THORpex.  The lead scientist participated this year in development of international plans for THORpex as co-chair of the World Weather Research Program/International Science Working Group (WWRP/ISWG) for THORpex.  The ISWG met in Nice, France, 26-27 April 2000, to discuss and suggest modifications to a draft preliminary proposal for THORpex.  The preliminary proposal was presented to the WWRP Science Steering Committee (SSC) and the WWRP Working Group in Numerical Experimentation (WGNE) for their approval.  Both groups endorsed the preliminary proposal and have requested a full science plan by October 2001.  
• IHOP and PACJET.  Planning for both programs is proceeding.  During the IWG retreat held in Columbia Maryland, 9-10 May 2000, it was decided to postpone IHOP until 2002 so that it would not conflict with the joint NASA/NOAA hurricane field program (noted below) in 2001.  PACJET is proceeding in early 2001, mostly as planned.  Descriptions of both programs are available at http://uswrp.ucar.edu/uswrp.html  under field programs.
• Field activities related to Hurricane Landfall.  Planning for a joint field program focused on hurricanes in the summer of 2001 involving NOAA and NASA occurred during the AMS Tropical Conference in Ft. Lauderdale, FL on 2 June, 2000.  The NASA component of the program is referred to as CAMEX 4.

Other activities to stimulate and promote the international community (WWRP) to cooperate with the USWRP.  Richard Carbone is serving a four-year term as Chair of the WMO/WWRP Science Steering Committee.  The Committee has one large meeting per year and conducts the remainder of its business via email year ‘round.  In this capacity, Carbone also serves on the WMO Commission on Atmospheric Sciences Advisory Working Group, under President Anton Eliassen. 

2. World Weather Research Program (WWRP)

In 1999, by resolution of the WMO Congress XIII, the World Weather Research Program (WWRP) became an official program.  Coordination of international research activities related to weather prediction is now consolidated under the WWRP and the Working Group on Numerical Experimentation (WGNE).  Significant developments include the approval of four WWRP projects, each of which has a U. S. contribution and some of which are USWRP related.  Richard Carbone has continued to fulfill the role of WWRP Chairman.

Below are highlights of approved and developing activities under the WWRP and WWRP research and development projects.

Projects

In 2000, Carbone’s principal activities have been associated with the Mesoscale Alpine Programme, Sydney 2000 Forecast Demonstration Project, THORpex, and the Mediterranean Experiment (MEDEX).  These activities have ranged from review of field activities(MAP), field phase conduct (S2K), review and coordination with WGNE (THORpex), review and science advisory functions (MEDEX).  There have also been activities associated with sand and dust storms (Arab League).

Mesoscale Alpine Programme (MAP):  Studying airflow and precipitation in steep terrain in the extra-tropical cyclone context in .  Field phase completed by mid-November 1999 with an emphasis on heavy rainfall and flooding as well as wind hazards.
Sydney 2000 (S2K):  Conducting very short-range prediction and nowcasting of high impact weather in large metropolitan area during and after the Sydney 2000 Olympics.  Field phase to be held 2 September through 18 November 2000.
In-Flight Icing (IFI):  Understanding icing conditions in water and mixed-phase clouds.
Tropical Cyclone Landfall (TCL):  Improving predictions of wind, rainfall, and use/impact of forecast information associated with landfall of tropical cyclones.
Mediterranean Cyclones (MEDEX):  Studying high impact weather events in Mediterranean regions with emphasis on high winds and heavy precipitation and flooding.
Continuation of FASTEX-Type Research:  Conducting research related to predictability and prediction of cyclones of oceanic (or remote-continent) origin given the present sparsity of in situ observations and the under utilization of remote sensing data from space.
Sand and Dust Storms (ASAPRO):  Improving observational infrastructure for soil uptake, atmospheric surface and boundary layer structures, upper air observations, and limited area model experimentation.
Urban Environment and Flooding:  Studying air pollution, heavy rain and urban flooding.
China Rainfall and Flooding:  Experimenting with limited area model forecasts.

Workshops and Symposia

Three meetings were organized or conducted in fiscal year 2000:

Sydney 2000 Forecast Demonstration Workshop was held in conjunction with the Sydney 2000 Forecast Demonstration Project from 30 October through 9 November 2000 in Sydney, Australia.  Forecasters from around the world participated in an experiential course of study that includes assessing the nowcasting potential of advanced forecast systems; the forecast process relationship to current scientific understanding; political/decision-making aspects; and societal impacts of improved prediction.
1^st WWRP Workshop on QPF Verification will be held in Prague, Czech Republic, during 13-18 May 2001.  The workshop emphasis will be on mesoscale and convective scale verification of heavy rainfall amounts but also on winter precipitation.
1^st WWRP Symposium on QPF will be held at Reading University, Reading, UK, 2-6 September 2002.  The workshop emphasis will be on high impact weather including heavy rain events.

WGNE, WCRP/USGCRP Collaboration

Extensive interaction with WGNE concerning development, oversight, and participation in THORpex, including joint research participation, review evaluations and recommendations.  Members of the WGNE represent R&D at NCEP, NRL, ECMWF, UKMO, DWD, JMA, METEOFrance, etc.

Extensive interaction with both US and International GEWEX including presentations at the GEWEX/SSG related to cooperative activities in THORpex, and Warm Season Rainfall and Floods Projects.  In the US, with NOAA/OGP/GAPP, NASA/GEWEC, and Interagency Water Cycle initiative.  Promising avenues of collaboration are developing with respect the North American Monsoon Experiment (NAME), which is joint GAPP and CLIVAR/VAMOS) and ensemble prediction of rainfall for hydrological applications (probably under the water cycle initiative).  Carbone serves as a SWG Member of NAME, representing high-frequency weather research aspects.

Pacific CLIVAR has general interest in observations and process studies over the north Pacific and specific interest in WW/US-WRP observing technologies.  Carbone is a member of the 1^st Pacific CLIVAR Workshop Organizing Committee and will present to this international group, plans for THORpex and related US/WWRP project activities for the purpose of defining improved global observations and joint process studies.

NCAR USWRP Program

In March of 2000 NCAR began the second year of its second two-year USWRP science program and embarked.  MMM received funding in both of the two-year cycles to support activities in the areas of numerical precipitation forecasting, mesoscale data assimilation, research in adaptive observations, large domain studies of warm season precipitation systems, and baroclinic wave studies.  Specific research highlights from FY 00 appear earlier in this Annual Scientific Report under the heading of Prediction and Precipitating Weather Systems (PPWS) Program.

B. U. S. Global Change Research Program

Within the U.S. Global Change Research Program (GCRP), MMM contributes to three of the eight programs.  The eight programs are the Global Tropospheric Chemistry Program (GTCP), Climate Modeling, Analysis, and Prediction (CMAP), CLImate VARiability and Predictability (CLIVAR), Role of Clouds, Energy, and Water (ROCEW), Geosystems Databases, Earth System History (ESH), Geospace Environment Modeling (GEM), and Coupling, Energetics, and Dynamics of Atmospheric Regions (CEDAR).  Descriptions of MMM's contributions to GTCP, CLIVAR, and ROCEW are outlined below.

1. Global Tropospheric Chemistry Program (GTCP)

To better understand the links between clouds, aerosols, and chemistry, MMM has instituted a Chemistry in Clouds Research program which is concerned with gaseous and aqueous chemistry, chemical species transport and dynamics affecting transport, and chemistry at cloud scale.  Other areas of GTCP research include:

The development of techniques to estimate fluxes of trace reactive species in the planetary boundary layer from measurements of mean vertical gradients or turbulent fluctuations of species concentrations, along with correlative meteorological measurements;
The understanding of the chemical and physical processes that contribute to the redistribution of chemical species by deep convection using a coupled convective cloud model to chemistry;
The study of aerosol composition and the study of the locations of formation of new sulfate aerosols;
The utilization of analyzed observations of storm structure, airflow, and lightning from STERAO-A coupled with 3D cloud scale simulations of selected storms to examine the vertical distribution of NO_x produced by lightning; and
The development and testing of airborne instruments that detect particles active in clouds, with increased attention to the characterization of soluble particles.

Specific research highlights from FY 00 appear earlier in this Annual Scientific Report under the headings of Cloud and Surface Processes Parameterizations (CaSPP) Program.

2. CLImate VARiability and Predictability (CLIVAR)

One of the goals of the CLIVAR program is to study the seasonal-to-interannual climate variability and predictability of the global ocean-atmosphere land system.  MMM's focus in this area is in the oceanic-atmospheric interactions.  Areas of research include:

The performance of cloud-resolving numerical simulations of tropical cloud systems on time scales of up to the intraseasonal, the comparison of the results to observations made in TOGA COARE, and the development of dynamics models of organized convective cloud systems;
The study of explicit effects of cloud systems on the surface energy budget;
The study of the role of cloud systems and attendant physical processes on upper-troposphere moisture distribution and its evolution over the tropical oceans; and
The improvement of the parameterization of precipitation convective cloud systems in the CCM3.

3. Role of Clouds, Energy, and Water (ROCEW)

The third USGCRP activity that MMM contributes to is ROCEW.  The primary purpose of ROCEW is to investigate the roles that precipitating convection, and cirrus and stratocumulus clouds, play in the Earth's energy balance, and the interactions of the radiative fields with these clouds.  The NCAR Clouds in Climate Program (CCP), a collaborative program within NCAR between MMM and the Climate and Global Dynamics (CGD) Division, helps facilitate the studies in the following areas:

The use of large-eddy simulations (LES) to explicitly calculate turbulence and boundary layer cloud motions;
The evaluation and development of convective parameterizations schemes, through the use of cloud-resolving models; and
The application of cloud-resolving models of precipitation cloud systems to understand and represent unresolved processes in climate and weather models related to cloud systems.

Specific research highlights from FY 00 appear earlier in this Annual Scientific Report under the heading of Cloud and Surface Processes Parameterizations (CaSPP) Program (click on the key words).

C. High Performance Computing and Communications (HPCC)

There are three basic components of the HPCC:  (1) High Performance Computing Systems; (2) Advanced Software Technology and Algorithms; and (3) Human Resource Development through HPCC.  MMM's focus is primarily on advancing software technology through:

Development of models on shared-memory massively parallel machines (Wildfire Research);
Development of ultra-high resolution models for turbulence simulations (Geophysical Turbulence Research); and
Development and testing of a multi-layer parallel implementation for the Weather Research and Forecast (WRF) Model. 

Specific research highlights from FY 00 appear earlier in this Annual Scientific Report under the headings that appear above.