An explicit three-dimensional nonhydrostatic numerical simulation of a tropical cyclone

Summary A nonhydrostatic numerical simulation of a tropical cyclone is performed with explicit representation of cumulus on a meso- scale grid and for a brief period on a meso- scale grid. Individual cumulus plumes are represented by a combination of explicit resolution and a 1.5 level closure predicting turbulent kinetic energy (TKE).The results demonstrate a number of expected and unexpected important scale interaction processes. Within the central core of the developing cyclone, meso- convective regions grow and breakdown into propagating inertiagravity waves throughout the lifecycle of the cyclone. In the early stages, the amplitude of pressure fluctuations associated with the meso- scale convection exceed the central pressure of the cyclone and strongly modulate its intensity. With each meso- scale pulsation, the cyclone core increases in strength, measured by the central pressure deficit. The increasingly strong inertial frequency of the storm core acts to increasingly trap the convection induced heating within the core by balancing the tangential wind against the low central pressure, before the meso- scale convection breaks down and sends the warmth away as a propagating wave. Eventually, the slow manifold's amplitude exceeds the amplitude of the meso- scale oscillations and a stable eye region is formed. As inertial instability increases, increasingly high thermal warmth can be protected in the core, allowing persistent subsidence to form and to clear out the cyclone eye.On the outside of the eye wall, strong inertial stability gradients in the troposphere cause convective warming to split the inflow to the eye wal! and spawn outwardly propagating inertia gravity waves. These waves carry away all of the heating forced by convection that is not inertially trapped by the eye wall and act as a moderating influence on storm intensity.Inertia gravity waves are also spawned in the stratosphere at the top of the eye wall by the revolution of asymmetric cumulus structures. In all instances, the tropospheric waves are coupled to the propagating stratospheric waves which both move at 35 ms^–1, although there are many instances where the stratospheric waves seem to have no tropospheric counterpart. Hence the anvil top forcing and low level breakdown are linked.The outwardly propagating inertia gravity waves act to initiate outer bands of convection. This initiation is with the assistance of low level boundary layer variations of density related to previous convection and to virga falling from the anvil which moistens and destabilizes the mid levels of _e minimum. The convection initiated by these waves does not move substantially outward with the wave, although may appear to develop outward discontinuously.With 12 Figures     

Numerical Simulation of Strongly Stratified Flow Over a Three-Dimensional Hill

A numerical study of stably stratified flow over a three-dimensional hill is presented. Large-eddy simulation is used here to examine in detail the laboratory experimental flows described in the landmark work of Hunt and Snyder about stratified flow over a hill. The flow is linearly stratified and U/Nh is varied from 0.2 to 1.0. Here N and U are the buoyancy frequency and freestream velocity respectively, and h is the height of the hill. The Reynolds number based on the hill height is varied from 365 to 2968. The characteristic flow patterns at various values of U/Nh have been obtained and they are in good agreement with earlier theoretical and experimental results. It is shown that the flow field cannot be predicted by Drazin's theory when recirculation exists at the leeside of the hill even at UNh 1. The wake structure agrees well with a two-dimensional wake assumption when U/Nh 1 but lee waves start to influence the wake structure as U/Nh increases. The dividing-streamline heights obtained in the simulation are in accordance with experimental results and Sheppard's formula. The energy loss along the dividing streamline due to friction/turbulence approximately offsets the energy gained from pressure field. When lee waves are present, linear theory always underestimates the amplitude and overestimates the wavelength of three-dimensional lee waves. The simulated variations of drag coefficients with the parameterK (=ND/ U) are qualitatively consistent with experimental data and linear theory. Here D is the depth of the tank.     

A study of the sensitivity of bare soil evaporation schemes to soil surface wetness,using the coupled soil moisture and surface temperature prediction model,BARESOIL

Summary The performance of evaporation schemes with and approach and their combination within resistance representation of evaporation from bare soil surface is discussed. For this purpose nine schemes, based on different functions of or , on the ratio of the volumetric soil moisture content and its saturated value are used.The quality of the chosen schemes has been evaluated using the results of time integration by the coupled soil moisture and surface temperature prediction model, BARESOIL, using in situ data. A sensitivity analysis was made using two sets of data derived from the volumetric soil moisture content of the top soil layer. One with values below the wilting point (0.17 m³m^–3) and the second with values above 0.20m³m^–3. Data sets were obtained at the experimental site Rimski anevi, Yugoslavia, from the bare surface of a chernozem soil.With 4 Figures     

Use of precipitation data for diabatic initialization to improve the tropical analysis of divergence and moisture

Summary First, we review the present status of diabatic initialization used for numerical weather prediction and conclude that the deficiency of diabatic initialization mostly stems from the shortcoming in evaluating diabatic heating rates accurately, particularly the release of latent heat by cumulus convection. This indicates the need to adjust the initial conditions for physical processes, and Krishnamurti and his colleagues introduced in 1984 the concept of physical initialization. Since cumulus convection is most sensitive to input data among many physical processes, the adjustment of atmospheric input data to a prediction model to produce desired initial precipitation rates is referred to as cumulus initialization.In this article we describe a general approach to diabatic initialization with a special emphasis on cumulus initialization. We present the results of forecasting experiments with a version of the NCAR Community Climate Model (CCM) to demonstrate the efficacy of a cumulus initialization procedure to ameliorate the spinup problern of precipitation. Finally, we discuss application of the present methodology of cumulus initialization for a stability-dependent mass-flux cumulus parameterization of CCM2 to pave the way to complete the diabatic normal mode initialization package for CCM2. Note that the present cumulus initialization scheme can be used to assimilate into the atmospheric analysis of the tropics the precipitation rates estimated by satellite radiometric imagery data.With 8 FiguresThe National Center for Atmospheric Research is sponsored by the National Science Foundation     

Propagation of the sea-surface temperature anomalies in the Tropical Atlantic

Summary The evolving modes of the sea-surface temperature (SST) in the Tropical Atlantic on the short interannual (IA) timescale were obtained by performing the extended empirical orthogonal function (EEOF) analyses on this variable separately for the 106-year (1871–1976) and 20-year (1881–1900; 1901–1920; 1921–1940; 1941–1960) periods. The equatorial and inter-hemispheric patterns manifest in the first EEOF mode of each analysis as part of the short IA evolution of the SST anomalies in the Tropical Atlantic. Another outstanding feature of the first EEOF mode of each analysis concerns the propagations of the SST anomalies in the meridional direction within the 20°N–20°S band and in the zonal direction in the sector 40°W–20°W. For all analyses, the SST anomalies propagate northward from the equator to 15°N and southward from 20°N to 15°N, with the same sign anomalies merging approximately at 15°N. On the other hand, the SST anomalies propagate westward in the sector 40°W–20°W with a propagation rate close to that of the phase speed of the fastest baroclinic Rossby wave in the ocean. So, the observed propagations of the SST anomalies in the 20°N–20°S band might result from the combined effect of the surface oceanic currents in this band and the baroclinic Rossby waves in the ocean.     

Northern hemisphere atmospheric response to changes of atlantic ocean SST on decadal time scales: a GCM experiment

Analyses indicate that the Atlantic Ocean seasurface temperature (SST) was considerably colder at the beginning than in the middle of the century. In parallel, a systematic change in the North Atlantic sea-level pressure (SLP) pattern was observed. To find out whether the SST and SLP changes analyzed are consistent, which would indicate that the SST change was real and not an instrumental artifact, a response experiment with a low-resolution (T21) atmospheric GCM was performed. Two perpetual January simulations were conducted, which differ solely in the Atlantic Ocean (40° S-60° N) SST: the cold simulation utilizes the SSTs for the period 1904–1913; the warm simulation uses the SSTs for the period 1951–1960. Also, a control run with the model's standard SST somewhat between the cold and warm SST was made. For the response analysis, a rigorous statistical approach was taken. First, the null hypothesis of identical horizontal distributions was subjected to a multivariate significance test. Second, the level of recurrence was estimated. The multivariate statistical approaches are based on hierarchies of test models. We examined three different hierarchies: a scale-dependent hierarchy based on spherical harmonics (S), and two physically motivated ones, one based on the barotropic normal modes of the mean 300 hPa flow (B) and one based on the eigenmodes of the advection diffusion operator at 1000 hPa (A). The intercomparison of the cold and warm experiments indicates a signal in the geostrophic stream function that in the S-hierarchy is significantly nonzero and highly recurrent. In the A-hierarchy, the low level temperature field is identified as being significantly and recurrently affected by the altered SST distribution. The SLP signal is reasonably similar to the SLP change observed. Unexpectedly, the upper level stream-function signal does not appear to be significantly nonzero in the B-hierarchy. If, however, the pairs of experiments warm versus control and cold versus control are examined in the B-hierarchy, a highly significant and recurrent signal emerges. We conclude that the cold versus warm response is not a small disturbance that would allow the signal to be described by eigenmodes of the linear system. An analysis of the three-dimensional structure of the signal leads to the hypothesis that two different mechanisms are acting to modify the model's mean state. At low levels, local heating and advection are dominant, but at upper levels the extratropical signal is a remote responce to modifications of the tropical convection.This paper was presented at the International Conference on Modelling of Global Climate Change and Variability, held in Hamburg 11–15 September 1989 under the auspices of the Meteorological Institute of the University of Hamburg and the Max Planck Institute for Meteorology. Guest Editor for these papers is Dr. L. Dilmenil.AWI Publication no. 254     

Trace gas exchange at the air/water interface: Measurements of mass accommodation coefficients

A liquid jet of 90 m diameter and variable length has been utilized to determine absorption rates and, hence, mass accommodation coefficients , of atmospheric trace gases. The compounds investigated are HCl (0.01), HNO₃ (0.01), N₂O₅ (0.005), peroxyacetyl nitrate (>0.001), and HONO (0.005). It is concluded that the absorption of these trace gases by liquid atmospheric water is not significantly retarded by interfacial mass transport. The strengths and limitations of the liquid jet technique for measuring mass accommodation coefficients are explored.     

Models of the wave boundary layer

A general approach to model the structure of the wave boundary layer, based on the nonlinear Reynolds equations in a curvilinear system of coordinates, is described. Both spectral and numerical grid models are used. The energetic interactions between wind and wave in terms of Miles' parameter are studied. For waves outrunning or running against the wind, the range of the inverse flux of energy is found. For waves running slower than the wind, quadratic growth of is established. Vertical profiles of the wave momentum flux for different fetches are given. Following P. Janssen, a one-dimensional analytical model of the wave boundary layer is suggested. The effect of waves on the drag coefficient is analyzed.     

Application of a “Big-Tree” model to regional climate modeling: a sensitivity study

Summary ¶In order to better understand land-atmosphere interactions and increase the predictability of climate models, it is important to investigate the role of forest representation in climate modeling. Corresponding to the big-leaf model commonly employed in land surface schemes to represent the effects of a forest, a so called big-tree model, which uses multi-layer vegetation to represent the vertical canopy heterogeneity, was introduced and incorporated into the National Center for Atmospheric Research (NCAR) regional climate model RegCM2, to make the vegetation model more physically based. Using this augmented RegCM2 and station data for China during 1991 Meiyu season, we performed 10 experiments to investigate the effects of the application of the big-tree model on the summer monsoon climate.With the big-tree model incorporated into the regional climate model, some climate characteristics, e.g. the 3-month-mean surface temperature, circulation, and precipitation, are significantly and systematically changed over the model domain, and the change of the characteristics differs depending on the area. Due to the better representation of the shading effect in the big-tree model, the temperature of the lower layer atmosphere above the plant canopy is increased, which further influences the 850hPa temperature. In addition, there are significant decreases in the mean latent heat fluxes (within 20–30W/m²) in the three areas of the model domain.The application of the big-tree model influences not only the simulated climate of the forested area, but also that of the whole model domain, and its impact is greater on the lower atmosphere than on the upper atmosphere. The simulated rainfall and surface temperature deviate from the originally simulated result and are (or seem to be) closer to the observations, which implies that an appropriate representation of the big-tree model may improve the simulation of the summer monsoon climate.We also find that the simulated climate is sensitive to some big-tree parameter values and schemes, such as the shape, height, zero-plane displacement height and mixing-length scheme. The simulated local/grid differences may be very large although the simulated areal-average differences may be much lower. The area-average differences in the monthly-mean surface temperature and heat fluxes can amount to 0.5°C and 4W/m², respectively, which correspond to maximum local/grid differences of 3.0°C and 40W/m² respectively. It seems that the simulated climate is most sensitive to the parameter of the zero-plane displacement among the parameters studied.     

The validation of various schemes for parameterizing evaporation from bare soil for use in meteorological models: A numerical study using in situ data

Parameterization of evaporation from a non-plant-covered surface is very important in the hierarchy strategy of modelling land surface processes. One of the representations frequently used in its computation is the resistance formulation. The performance of the evaporation schemes using the , , and their combination resistance approaches to parameterize evaporation from bare soil surfaces is discussed. For that purpose, the nine schemes, based on a different dependence of and on volumetric soil moisture content and its saturated value, are used.The tests of performances of the considered schemes are based on time integrations by the land surface module (BARESOIL) using observed data. The 23 data sets at a bare surface experimental site in Rimski anevi, Yugoslavia on chernozem soil, were used for the resistance algorithm evaluation. The quality of the schemes was compared with the observed values of the latent heat flux using several statistical parameters.     

Analysis of Astex-Stratocumulus Observational Data Using a Mass-Flux Approach

A mass-flux approach is applied to observational data obtained in a convective boundary layer topped with stratocumulus clouds. The observational data were obtained from aircraft measurements during the Atlantic Stratocumulus Transition Experiment (ASTEX). A conditional sampling method is used to calculate average updraft and downdraft values. The vertical fluxes calculated with the mass-flux approach are found to be proportional to the real (measured) fluxes, with a proportionality factor being about 0.6. This value is predicted by theory for two variables having a joint Gaussian distribution function; proportionality factor = 2^-1 0.637. The horizontal fractional entrainment and detrainment rates calculated from the data ( 1–2 × 10^-2 m^-1) are an order of magnitude higher than the rates obtained by large eddy simulations for cumulus convection ( 2–3 × 10^-3 m^-1) and two orders of magnitude higher than those used in modelling cumulus convection with a mass-flux scheme in an operational weather forecast model ( 3 × 10^-4 m^-1). A numerical mass-flux model for the thermodynamics was developed and showed that results are in good agreement when compared with measured profiles of the liquid water content.     

Bioaccumulation of Cobalt in Parmelia sulcata

Six locations across mainland Portugal were selected for exposing Parmelia sulcata, for a one-year period (8 months for one site), with simultaneous measurement of total (dry + wet) deposition (one-month periods). The exposed lichens and the total (dry + wet) deposition were analysed for cobalt contents by INAA (instrumental neutron activation analysis) and ICP-MS (inductively coupled plasma mass spectroscopy), respectively. The designated wet deposition was evaluated through the collected water volume; the designated dry deposition was assessed after the (dried) residual mass of the wet deposition. An excellent agreement between Co contents in exposed lichens and the cumulative (1) Co contents in the dry deposition, (2) dry deposition, and (3) wet deposition has been found for the locations with alternate drought and precipitation months, high dry deposition, and high Co contents in the latter. Continuous rainfall was found to hinder the Co accumulation in the lichen due to its release from the lichen and/or lower Co contents in the dry deposition. At three locations, P. sulcata Co contents, after subtraction of the background (before exposure), equalled or exceeded the Co contents in the cumulative dry deposition at the end of the exposure time. The optimal exposure period for this species likely depends on the exposure conditions.     

A study of a fair weather boundary layer in TOGA-COARE: Parameterization of surface fluxes in large scale and regional models for light wind conditions

The study focuses on a way to parameterize the effect of subgrid scale convective motions on surface fluxes in large scale and regional models for the case of light surface winds. As previously proposed, these subgrid effects are assumed to scale with the convection intensity through the relationship: where is the mean velocity of the wind, U₀ the velocity of the mean wind, w_* the free convection velocity, and an empirical coefficient to be determined. Both observations and numerical simulation are presently used to determine the free convection coefficient .Large eddy simulation of a fair weather convective boundary layer case observed during TOGA-COARE is performed. Comparisons between observations and the simulation of surface properties and vertical profiles in the planetary boundary layer are presented. The simulated vertical turbulent fluxes of heat, moisture and buoyancy range well within estimates from aircraft measurements.The most important result is that the true free convection coefficient , directly estimated from simulation, leads to a value of 0.65, smaller than the ones estimated from temporal and spatial variances. Using observations and simulation, estimates of from temporal and spatial variances are obtained with similar values 0.8. From both theoretical derivations and numerical computations, it is shown that estimates of the true from variances are possible but only after applying a correction factor equal to 0.8. If this correction is not used, is overestimated by about 25%. The time and space sampling problem is also addressed in using numerical simulations.     

A mechanism of Alpine lee cyclogenesis as revealed by a quasigeostrophic variational filter

Summary Mechanisms associated with Alpine lee cyclogenesis during the early phase of their generation are investigated using a variational quasigeostrophic filter technique. It was possible to extract the quasigeostrophic signal from the available analyzed real data set.The results presented here are for the 11–12 March 1982, an example of so-called orographically induced lee cyclogenesis. Non-quasigeostrophic fields, calculated as a difference between observations and the quasigeostrophic fields, show significant magnitudes indicating the possible importance of non-quasigeostrophic processes. A dipole structure in the residual geopotential field was observed, similar to the results of numerical model experiments. Also, a strong upper-level non-quasigeostrophic divergence was found in the Alpine region 24 hours prior to lee cyclogenesis, lasting for 6–12 hours. On the other hand, quasigeostrophic results indicate only a local effect of mountain slopes, suggesting possibly a dominant role of the low-level blocking. A hypothetical scenario of Alpine lee cyclogenesis is proposed, based on results obtained here.With 14 Figures     

Arctic hazes in summer over Greenland and the North American Arctic: II. Nature and concentrations of accumulation-mode and giant particles

Airborne measurements made during August 1985 over Greenland and its environs show that both accumulation-mode (0.1 m D2.0 m) and giant (D2 m) particles were present in relatively high concentrations in arctic haze layers and that the accumulation-mode particles dominated light scattering. Particles with diameters (D) between 1 and 4 m consisted predominately of mixed materials, small and dense inclusions, and probably organic compounds containing sulfur. Many of the particles from 0.1 to 1 m in diameter were also of mixed composition, with sulfuric acid, ammonium sulfate and organics probably the dominant constituents.     

A study of gravity waves generated by convective systems in eta model forecasts

Summary Spectral filtering and visualization reveal a multitude of gravity waves in the fine mesh (29 km) Eta model output, extending into the large scale, masking the vertical motions associated with Rossby-modes. The 6 h forecast of the Eta model from 00 UTC 22 December 1994 initialization contains gravity waves recognizable over a broad spectrum of wavelengths as concentric circles radiating outward from centers of model convection. Contamination of the signal is therefore identifiable in any spectral range, as rings. Postprocessing techniques are presented to eliminate gravity wave contamination and to reveal the meteorologically significant signal. Low-pass filtering is found to be an effective means of extracting the Rossby signal, but with loss in Rossby signal that renders the results obtained with a high resolution model no better than what could have been obtained from a coarser resolution model. Time filtering, is suggested as a more effective method for eliminating the effects of fast modes while maintaining mesoscale details of the slow modes.With 9 Figures     

ALPEX-Simulation

Summary In a project ALPEX-Simulation, sponsored by the Österreichischer Fond zur Förderung der wissenschaftlichen Forschung (FWF), all eight cases of ALPEX-SOP cyclones were numerically simulated with a fine mesh isentropic model of the atmosphere. These numerical simulations in six-hourly intervals allow a deeper insight into the synoptics and dynamics of the cyclogeneses in the Western Mediterranean, especially into the genesis of the two basic types of cyclones: the so-called Überströmungs-type and Vorderseiten-type. In the first phase of cyclogenesis of the Überströmungs-type, the blocking and flow splitting of the cold air due to the Alps and the canalization between the Alps and the Massif Central are important. Cold air flows cyclonically around the western part of the Alps, creating a vorticity maximum at the south western edge of the Alpine, bow and leads also to an enhanced PV. In connection with warm air in the Mediterranean, a strong baroclinic zone is generated. The interaction between the arriving PV maximum in the upper troposphere and the enhanced PV at the bottom leads to cyclogenesis in the Western Mediterranean. In the case of the Vorderseiten-type warm air advection dominates with the exception of a shallow layer of cold air in the inner Po-Valley, which is shielded by the Alpine ridge. A well-pronounced PV maximum builds up and couples with the PV maximum arriving at upper levels, even before the cold air, coming from the north-west, has surrounded the Alps. The cold air only intensifies the development by raising the baroclinity. Therefore, the Vorderseiten-cyclogenesis is an orographically modified cyclogenesis, in the course of which the cyclonic development is triggered by the Alps, whereas the Überströmungs-cyclogenesis is an orographically induced cyclogenesis i.e. a true lee cyclogenesis.With 14 FiguresDied in a tragic traffic accident on June 6, 1993.     

La forme rotationnelle des équations de la dynamique atmosphérique et les invariants du mouvement de l'air

Résumé Après avoir rappelé quelques notions de calcul tensoriel et les équations de la dynamique atmosphérique en coordonnées généralisées, l'auteur établit l'équation fondamentale de la rotationnelle absolue de l'air. Diverses formes de cette équation sont mentionnées. La première de ces formes est celle d'un bilan qui permet de définir le transport et le taux de production de la rotationnelle absolue de l'air. De la forme particulière de ce transport, il résulte que ce bilan se réduit à un bilan dans un espace à deux dimensions. Deux exemples illustrent les formules générales, le premier, en coordonnées sphériques (, ,r), le second, en coordonnées (, , ) où est un scalaire quelconque (pression atmosphérique, température potentielle, ...). Une autre forme de l'équation fondamentale est celle qui donne le taux d'accroissement individuel d'une composante de la rotationnelle absolue. Cette forme conduit à l'équation d'Ertel. En terminant, l'auteur généralise l'invariant d'Ertel-Rossby.

---

Summary Having recalled some results of the tensor analysis and the equations of atmospheric motion in general coordinates, the writer establishes the fundamental equation of the absolute vorticity. Different forms of this equation are mentioned. One of these is the equation of balance from which it is possible to deduce the definition of transport and production of vorticity. It is shown that the equation of balance of the absolute vorticity may be reduced to an equation of balance in two-dimensional space. Two examples illustrate the general equations and formulas: the first example, in spherical coordinates (, ,r), the second one in the coordinates (, , ) where is an arbitrary single-valued scalar quantity (pressure, potentiel temperature, ...). Another form of the absolute vorticity equation expresses the individual change of an arbitrary component of the absolute vorticity. This form leads toErtel's equation. Finally, the writer generalizes theErtel-Rossby invariant.

---

Zusammenfassung Nach einer Rekapitulation einiger Resultate der Tensorrechnung und der atmosphärischen Bewegungsgleichungen in generalisierten Koordinaten wird die absolute Wirbelgleichung aufgestellt und es werden verschiedene Formen dieser Gleichung erwähnt. Eine derselben ist die Bilanzformel, aus der die Definition des Wirbeltransports und der Anteil der Wirbelbildung abgeleitet werden können. Es wird gezeigt, daß die Bilanzgleichung des absoluten Wirbels auf eine Bilanzgleichung im zweidimensionalen Raum reduziert werden kann. Die allgemeinen Formeln werden durch zwei Beispiele erläutert: das erste in Kugelkoordinaten (, ,r), das zweite in den Koordinaten (, , ), wo eine beliebige skalare Größe (luftdruck, potentielle Temperatur usw.) darstellt. Eine andere Form der absoluten Wirbelgleichung gibt die individuelle Zunahme einer beliebigen Komponente der absoluten Wirbelstärke wieder; diese Form führt zurErtelschen Gleichung. Zum Schlusse wird dieErtel-Rossbysche Invariante verallgemeinert.

     

A note on ‘separation’ over short wind waves

Whether or not separation occurs in the airflow over wind-generated water waves is partly a question of semantics but also has an important bearing on the wave generation process. In the present paper we use a rather formal and perhaps narrow definition of separation and show that it does not occur where the shear stress is zero but only in conjunction with wave breaking. This is unlikely to happen except in the presence of quite strong surface drift velocities in the water. A similar connection between separation, surface drift and wave breaking has recently been established by Banner and Melville (1976).The effects of increasing wave amplitude or steepness are investigated with a numerical model of the airflow over water waves. Variations in the depth of the closed streamline region are predicted. The model is also used to investigate the possible importance of surface drift velocities.On leave 1976, Atmospheric Environment Service, Downsview, Ont., Canada.     

Parameterizing turbulent diffusion through the joint probability density

The convective mass flux parameterization often used in meteorological modeling expresses the vertical flux of a transported scalar as proportional to the product of the difference in mean values of the scalar in updrafts and downdrafts and their characteristic velocity. The proportionality factor is a constant to be specified. We show that this proportionality factor also appears in the relaxed eddy accumulation technique of Businger and Oncley. That associates the surface-layer flux of a scalar with the product of the standard deviation of vertical velocity and the mean concentration difference between updrafts and downdrafts.We show that this constant (b) is determined uniquely by the joint probability density (jpd) of vertical velocity and the scalar. Using large-eddy simulation, we generate this jpd for a conservative scalar diffusing through a convective boundary layer. It has quite different forms in top-down and bottom-up diffusion geometries. The bottom-up jpd is fairly well represented by a jointly Gaussian form and implies b ~ 0.6, in good agreement with the surface-layer value reported by Businger and Oncley. The top-down jpd is strikingly non-Gaussian and gives b ~ 0.47. Updrafts carry the bulk of the scalar flux - 70% in the bottom-up case, 60% in the top-down case.The National Center for Atmospheric Research is sponsored by the National Science Foundation.