Trip to the Netherlands GCSS Conference: August - September 1995
By: William R. Cotton

To view Dr. Cotton's other Jounrals, Click here
This is a summary of my travels in The Netherlands, notes taken at the GCSS Meeting, and summary and conclusions provided by Steve Krueger and Chris Bretherton.

Depart Thursday morning from Fort Collins/Loveland. Flight to Dulles uneventful. Overnight flight to Amsterdam was hard to sleep on because many people talked continuously and loudly. Also a 300 pounds plus person behind me kept bumping my seat and flopping a newspaper he was reading on my head. I don't know how he fit in the seat!

Arrived Amsterdam Sunday 0700 and bike arrived on over-sized luggage pickup area intact. As instructed I sought out the train station which is the most conveniently located train station in an airport I have ever seen.

Took the train to transfer point and as I was told it would arrive in 15 minutes, decided to put the bike together. Train arrived before I had a chance to get everything packed, but got everything on the train. I got the bike adjusted and all finalized by the time I arrived in Utrecht.

Then I spent about an hour trying to locate travel information (VVV) but the Utrecht area map they gave me was worth it.

Following the VVV's directions I pedaled to DeBilt. It was only 3 or 4 miles and after asking directions I finally found the hotel and arrived by 10:00 a.m. After a shower I took a nap to 12:30 p.m.

Then I pedaled to KNMI and had lunch at their cafeteria with Aad Van Ulden. Finding everything ready for next week's meeting, I decided to take a ride to help keep awake. Aad had mentioned there was a glider port only 10 km west of DeBilt. I decided to check it out. I found a nice forested area near the airport, but a high barbed wire fence looking much like a military area surrounding it. If it looks military, smells military, it must be military. After riding around 3/4 of the perimeter on some scenic winding bike paths, I finally found an information headquarters. As I suspected it was an airport with glider operations but only for military.

So continuing around the perimeter I headed back only to get into a full downpour. I got out my rain poncho which kept me dry except for my feet. After returning to my room to get dry shoes I pedaled toward Utrecht to find supper.

The prices in my hotel ran about $30 U.S. per meal. I found an Indonesian restaurant whose owner found me a room next door to park my bike. The food was interesting with many tastes I had not experienced. It ranged from mild to very spicy and cost about $30! Pedaled back to hotel, watched a U.S. TV program with Dutch subtitles and after reading went to sleep by 10:30 p.m. I slept through to 9:30 a.m.-- Vollie I slept in!

After breakfast at the hotel I pedaled to DeBilt to pick up some bread, cheese and tomatoes. My destination today is a supposed glider port near Nieuw-Loosdrecht. Bicycling in The Netherlands is great, just like Denmark. Well-defined bikeways with separate stop lights and signs. Most people pedal bikes about like what I bought in 1954 as the newest thing--three-speed, English style with hugh wheels, an iron frame, kick stand, and chain-guard. They must weigh 50 lbs. Families were out biking, some with camping gear. There were bikers everywhere. Riding clubs all in the same color shirts rode high-performance road bikes. I even saw one guy riding a TREK 7000 series mountain bike.

The only problem I had was the bike trails don't always follow the major roadways. Also, since there are no mountains, I frequently got my directions fouled up. At one point I found myself at Soest about 10 km east of my destination.

After getting help from numerous people (all of whom were very helpful) I finally found the Hilversum Airport. There I met a glider club who were operating a two-winch system plus aerotows. The winching operation was very impressive with light signals and a portable control tower.

I got a ride with an instructor in a ASK 13 for the price of the instructor ($10). After a great launch to approx. 200 m we found a thermal and soared in it to approx. 1200 m. Two other planes were in it, flying closer than I have ever been in a thermal. The instructor continued to ask me to flatten out my bank in turns in the thermals contrary to what I have been taught here. Rolf experienced the same thing flying in Switzerland. I had trouble making coordinated turns at first as it was quite different from the sailplanes I have flown so far. The rudder pedals, in particular, were a pain. As you placed your entire foot into metal frames and had to move your whole leg, instead of just wiggle your toes with your heal fixed in place. It was great ride and I got some great views of the area.

I biked back to DeBilt through some forests and farmland, a pleasant ride.

Then I biked to Altrecht to a pizza place and got a pizza and beer for about $10.

The Dutch seem about as conservative about soaring as they are about biking. All the sailplanes I saw were canvas covered--no glass ships here! Aside for lower maintenance costs the old ships fly slow (approx. 30 kts) and have low sink rates so they can stay in the weak thermals and low cloud bases pretty well.

On biking, besides the standard old style bikes being the norm, no one wears helmets, and they look peculiarly at mine, especially with the mirror. Few wear shorts, but instead standard long pants.

Today I decided to bike to Terlet to the east (approx. 70 km) where I was told that a very impressive commercial soaring operation is located. I began about 10:00 a.m. after sleeping in to 8:30 (Vollie take note!). I went through Zeist and beyond passed over a hilly, forested area and on to Wordenberg. As usual I lost the bikeway in the towns where the highway bi-passed but not the bikes. It seems the biking signs all pointed to the same town which much later (on my return) I found meant through traffic! Oh well! In one town I believe Schurensee, I asked directions of a man dressed in a suit with four children in hand also dressed in suits. The man said he didn't speak English but proceeded to tell me the Lord made six days for work and one Sunday for rest. Well I knew I was lost, but not that bad! So, I proceeded on, lost as I was, and eventually found the bike route which again paralleled the main road.

Along the way I passed many small farms; dairy, I believe, judging from the manure smells. Many of the farm homes as well as the homes in town had very nice flower gardens. I passed through another forest area so thick there was no underbrush and cars on the highway were instructed to turn on lights.

After 3.5 hours I found my way to Terlet. The area is a highland area with strong winds, more open land scattered with heather about and smaller trees and bushes.

The gliderport is on top of a ridge. Here was an impressive commercial glider operation comparable to Turf near Phoenix. The difference was Turf operates a fleet of 6 or more tow planes. At Terlet they had a single tow plane, an old 2-place motorized glider, and a fleet of high-performance 2-place and single place glass sailplanes. They were all lined up in six side-by-side rows, with several behind the lead planes like the start to the Indy car race.

The trick was a superwinch that operated six winch lines. A truck hauled the six lines up from the winch like the start to a sulky horse race. The lines were dropped parachutes and weak links attached, then hooked on the planes when they were ready to launch. The planes were individually launched but in rapid succession.

The weak link by the way in a metal contraption holding a piece of metal rated to break upon stressed for a two-place or single-place plane (different metal pieces).

I asked the cost for a flight and since it was 2.5 times what it cost at Hilversum. I decided to pass. I liked the club atmosphere better at Hilversum anyway.

I had my lunch and at 3:00 p.m. started back. The ride back was tiring as the winds picked up to 12 to 15 kts dead on the nose.

Oh, I forgot, on the way out and again on my return I passed a windy meadow area near Ede. There there was a kite convention. There must have been 100 or more kites flying in the area, some huge, some big and many small ones. Some contest maybe? Loads of cars were parked along the roadside. It was the only area I had trouble biking because the bikeways were cluttered with family groups on bikes. Mom, pop, the kids, grandparents, etc. all moving at snail's paces and me without a bell yelling "bike passing on left" again and again. I guess they got the idea.

Finally I broke out of the pack and headed along. I was getting pretty tired toward the end, especially my (sore) END. I decided I didn't want to wait and pay a big price for food besides, so I took a 3km (out and return) side-track and stopped at McDonald's--yes me at McDs! I had 2 fish sandwiches, fries, and a vanilla shake for about $6.50 US. Lots of fat and energy to make up for seven hours of continuous biking! When I stepped outside it was raining, so I had to suit-up. Got back to the hotel and ordered a cheesecake to take to my room.

The I relished in a hot shower and then ate my cheesecake with hot drinks prepared from the hot water faucet. Now I'm pooped and going to bed. Hopefully I won't wake up after an hour of sleep like last night. I read from 11:00 p.m to 11:30 p.m. Hence the reason I feel like sleeping in--All part of the adjustment process.

I slept the night through--hooray! Today I thought I'd take a train to Amsterdam and then bike to Volendam. I got going by 9:00 a.m. and hardly left DeBilt when it started raining. I put on my poncho over may rain jacket to keep my pack dry. Took the train to Amsterdam and ferry to the north side. I followed a river for a while when it really started pouring. After 4 or 5 miles along a pretty riverside trail I decided to turn back. Returning to Amsterdam I biked around town and even took a glass-topped tourist boat. Besides the views from the boat, I enjoyed the "greenhouse" effect. Remember today is Monday, my fast day, so I am kind of lazy. Also, my butt is still sore from yesterday.

I particularly enjoyed the old boats along the canals, especially the old working sailboats with huge lee boards that people have converted to homes.

I returned to DeBilt and got charged $6 for my bike on the train. I'll have to fold it and put it in my bag to fool them in future.

I found a message from Bjorn at the hotel, so I know he is about but haven't seen him yet.

After a nap I contacted Bjorn and walked into DeBilt with him to keep him company while he ate (remember it is Monday, my fast day). We found a Ching-Indo restaurant, what ever that means where he had a dish that sounded Indian, but tasted more Chinese.

After a full nights sleep and breakfast, I put Bjorn on the cross bar of my bike with a towel for cushioning and headed for Bilthoven which is purported to have bike rental at the train station. It's about a 3 km ride and I worked up a sweat in my rainsuit pedaling Bjorn and me. I also got a taste of Bjorn's ponytail! Bjorn started getting a charley horse in his leg, holding it out from the bike.

He was able to rent a standard Dutch ironsides single-speed with a coaster bake for about $19 per week.

We then headed southeast on a cool and drizzly morning. Bjorn wasn't so sure he wanted to do this since he neglected to bring rain gear--to The Netherlands? I told him to wear shorts and loaned him my parka which he never used as it never got beyond light drizzle.

We took to narrow, paved roads south and east of Dricebergen-Rijsenmberg. The road had a canal (like our irrigation canals) running along at least one side, with boats (row boats) in front of most houses. Every so often we encountered a castle, some looking pretty old, others probably 19th century. It was a pretty pastoral landscape with plenty of cousheisa smells.

We finally reached the river a top a dyke that was put in place for flood prevention. Eventually we walked our bikes across the gates of a lock, then along dam gates across the river. We then came to a closed, barred gate.

Now what do we do? We saw no signs that we interpreted as "keep out" getting there. Of course, that doesn't mean there weren't any. I found a call box and pushed the button asking if we could get through? On the other side a recreational park was noted on the map. After some discussion, we in English and the guy on the speaker in Dutch, I gathered we were supposed to put 5 Kl (approx. $3) in some box before he opened it. Couldn't find the box and just as we started to turn back across the dam the gate opened about 3 ft and I walked through. They promptly closed it with me out and Bjorn still inside. Now what? After Bjorn and the voice in the box had an English/Dutch discussion, the gate opened again and he quickly slid through.

Once in the recreation area we began exploring it. I found a sailboat rental place and was tempted to rent one. But they wanted 65 Kl for the afternoon or approx. $40 and it was cool, windy and it looked like it would rain. So we pedaled on to a marina that was closed up and moved some chairs under an overhanging roof and had lunch while it rained.

Our lunch consisted of cheese and bread which I had purchased earlier at DeBilt, some apples from our breakfast, and some pears which I picked up beneath a tree along the road. They were great!

We headed back by way of Wijk bij Duurstede, taking a ferry across the river. From there we took some single-lane paved side roads through Cothen (drove through a windmill), Doorn, and again on side roads to Drieberge-Rijsenburg. We headed north through wooded countryside and bike lanes to Austerlitz, the rough woods to Zeist. Zeist is quite a wealthy looking city. Actually I haven't seen anything run down or close to slums here. But Zeist really stands out. I got a pastry and Bjorn cashed some money and, bucking the wind, we headed back to DeBilt. Ah! a nice hot shower again.

I figure I have ridden over 400 km since Friday. My butt is a bit sore, otherwise OK. Bjorn did well on the single speed and kept with me all the way.

0830 Begin meeting. Welcome B. Cotton. Chris Bretherton plan for meeting.

Bjorn overview on entrainment.

Taylor's entrainment hypothesis origin of this model.

Separation boundary "laminar super-layer" - Townsend

Basic lab experiments - Deardorff's Tank experiments - 2 layer fluid - sharing grids. Summarizes selected results from lab exps.

Issue of similarity - lab

Stratocumulus -- how does it differ from lab exps?

Debate of quantitative value of field stratocu obs vs. lab exps.

Recent attempts at synthesis. Breidenthal's work.


  • No consensus
  • A = 0.2 law not well established, but doesn't matter
  • Much less known about stratocu
  • Previous lab and theoretical work often contradictory
  • Need for more observations

Peter Bechtold

  • Shows Stull's fig. of passive, forced, and active convection
  • Approaches to simulating cloud population - Arakawa Schubert etc. scale dependence?
  • What is the horizontal mesoscale convection (10-50 km) based on Kain and Fritsch (1993)
  • Hyrbrid mesoscale
  • How to separate cloudiness by convective scheme and stratiform cloudiness?
  • Shows a fractional cloudiness parameterization scheme derived from LES a function of not defined--may be related to variance? I guess its Mellor's definition of

    Joao Teixeira Describes ECMWF single column model. Shows prog. eqs. for liquid water and cloud fraction and parameterizations of cloud coverage entrainment rates, dissipation of cloud fraction, etc. Describes Monin-Obukov surface layer implementation. Too many eqs. certainly can't absorb all!

    Model is available by writing letter to director request it (ECMWF).

    Branko Kosovic

    Describes his "nonlinear" subgrid scale scheme. Overcomes deficiencies of linear models.

    • No backscatter
    • Can't do normal stress effects of secondary flows
    • Inconsistent with respect to invariant transformation

    Nonlinear model has 2 more terms added to standard Smagorinsky closure due to strain rate tensor and rotation rate tensor. Simplest nonlinear scheme that has all rotational properties.

    Nonlinear model extracts energy better from the mean flow and also does backscatter.

    TKE is a little different between linear and nonlinear SCS schemes but temperature fluxes are identical. Smoke fluxes are slightly different. In TKE budget, largest difference in transport term near the inversion.

    Overall nonlinear model has small effect for smoke case but it may be more important in a sheared boundary layer case.

    It is not known if coefficients are stability dependent--they may be.

    Chin-Hoh Moeng

    Effect of cloud top cooling on entrainment rate. She varies downwelling radiative flux 60, 40, 20. Potential temperature flux varies linearly with FR. TKE doesn't vary linearly. Cloud top height or varies with FR.

    is approx. linearly proportional to average TKE.

    Steve Krueger

    Stratocu to trade cu transitions. Showed slides of different cloud regimes over Porto Santo. Discusses his 2D cloud-resolving model and application to transition from solid stratus to cu.

    Chris Bretherton

    More stratocumulus to cumulus transitions. PM mesoscale cellular convection. Agee (1987) fig. showing regions of MCC.

    Explanations of MCC's

    • constant flux BC's coupled with low (Fiedler, 1994; Qu and Randall, 1995)
    • Narrow up and broad downdrafts in cu (Bjerknes, 1939; Van Delden, 1985
    • Clouds preferentially???
    • Cooperative instability with deep gravity waves. Clark, Hauf, Kuettner, 19??
    • Drizzle

    Hsu and Sun (1991) 3D with turbulence model doing transports

    Chris shows 3D simulation for ASTEX 4 June -- nocturnal based on Porto Santo sounding 500 m . Cloud spacing approx. 50 km controlled by up moist/down dry.


    • Broad aspect ratios driven by latent heating (up moist/down dry)
    • Commented that so far cloud fraction is same whether or not MCC occurred.
    • Suggests that open vs. closed cells are the same dynamically except that you can see cu while the other is obscured by shallow anvil-like clouds.

    Andreas Chlond - Max Planck Institute

    LES model. Added a single-moment bin model of microphysics. Include radiative effects in droplet growth. Have 12 classes - ice particles 1-64 to simulate contrails. For boundary layer clouds they use a bulk microphysics scheme (Lupkes et al., 1989) this scheme is tested for closed-box type calculations.

    Radiation scheme. Use spherical harmonic spatial grid method (low truncations 2?; simplified version of Frank Evans scheme). Solutions obtained using conjugant gradient method? Bounded cascade model to describe spatial distribution of cloud liquid-water. Claims errors of approx. 2% for truncations of 5 to 7.

    Dave Lewellen

    Shows variation in "A" with . Shows strong sensitivity to . Bulk of smoke fluxes come from reasonable scales even for very coarse . Little sensitivity to sub-grid parameterization once m.

    Physical Variations

    Smoke cooling varied from 10 varied linearly with was constant.

    Absorption coefficient (var. thickness of cooling) falls off as layer deepens, but not too much and nearly constant.

    Then he lowered cooling rate (made top of smoke transparent) -- entrainment values dropped sharply, but was flat, and dropped.

    Cloud with fixed cooling

    Complication of radiation coupled to LWC. Set radiative cooling as in smoke, but otherwise cloud could do its thing. With ; 50 200 m all give same fluxes, etc.

    Varied moisture amounts, e.g., Randall-Deardorff criteria and increases to a max. at their point and remains constant after that and cloud fraction goes from 1 to smaller values.

    For fully coupled cloud (radiation and cond/evp) higher resolution was need than fixed radiation case. But not a lot of higher resolution.

  • 8/31/95-Thursday
  • Malcolm MacVean
    • "Smoke Case" showed results of his calcs - with 3.2 km and 6.4 km domain - not much sensitivity so smaller domain was selected.
    • Showed hour-to-hour variability for his runs--some differences in amplitude but shapes some.
    • Somewhat bigger variations in smoke flux.
    • Residual in energy budget near inversion were large, otherwise small.

    Rate of rise of smoke top varies with method of calculating buoyancy in equation. Hour-to-hour variations in total smoke flux were very large!

    Showed plots for all participants--huge scatter!

    "A" values varied greatly from monotone to non-monotone schemes. Variability (hour-to-hour) greater in non-monotone schemes.

    Clear differences in smoke topped height between monotone and non-monotone.

    Sub-grid contributions to smoke fluxes are consistently small. Horizontal velocity variances agree among monotone schemes, less so for non-monotone. Same with .

    Theta flux shows more scatter but sub-grid fluxes are very small compared with total.

    Skewness values are more consistent except above inversion where monotone schemes yield small values (near zero) of .

    Eddy viscosity (diagnosed) varies quite a bit with MacVean and Bjorn showing a max near inversion while others don't. Reasonable consistency in resolved TKE.

    High Vertical Resolution

    (4 people)

    values smaller. Smoke heights closer. Layer average TKE shows more spread but converging? Total smoke fluxes showed most variation. Radiative cooling still appears to extend one grid point above inversion.

    2D Results

    Restricted domain results less statistical significance in 2D and 3D. is consistently larger in 3D than 2D. Peter Bechtold did only 1D model.

    Bjorn's Summary of Many Experiments

    • Changed eddy Prandtle number - No effect.
    • Changed diffusion length scale - No effect.
    • Moved point where K is calculated - No effect.
    • Used non-monotone operator - Same as group.
    • Monotonic one direction - No change except reduced cost.
    • Ran compressible model - Little change
    • Ran anelastic model - Little change.

    Increased resolution increased entrainment if only. Increase gave about same resolution.

    P. Mason - If simulations are correct model suggests what is important is local distribution of . You can't distinguish numerical diffusion from Smagorinsky diffusion by scaling.

    Hans Cuijpers

    Shows differences in predictions of for smoke case for cases using centered differences and using monotonic schemes.

    Shows case where most radiative cooling occurring at smoke top and case where most is within the mixed layer.


    • Location of radiative cooling important
    • High resolution runs--gradients better preserved placing more radiative cooling in mixed layer
    • Real cloud exhibits different budgets - v not conserved
    • Evolving concept - differences in coarse resolution runs attributable to placement of cooling and not to differences in "dynamical resolution" of entrainment.

    Joan Cuxart

    Sensitivity of dry entrainment to turbulent Prandtl number and mixing length in 1D and 3D simulations.

    New model code, non-hydrostatic built from grid up. 1D simulation can be tweaked to compared well with LES using in 1D gives better results.

    Chris B.

    Shows dependence of A.

    Macolm again

    Shows values of for cases of surface heating and cloud top cooling things fall on line if

    ASTEX Lagrangian I and 2 cases

    Chris and Peter Duynkerke describe the Lagrangian experiments. Peter - Lagrangian 1 -

    Drizzle rate: a lot of scatter about 1 mm/day. Estimates of . Stephan de Roode--Turbulent structure of 1st Lagrangian SST increased 4.5 C in 4 days subsidence 0.5 cm s.

    estimated ?

    KNMI sponsored a fancy full course dinner in Utrecht. It took 4 hrs! Got back to my room at 11:30 p.m.

  • 9/1/95-Friday
  • Chris Bretherton

    2D simulations of ASTER Lagrangian with HUSCI.

    Lagrangian 1 -- Forced with observed subsidence, geostrophic winds SST's. Exhibited a deepening of BL with time and transition to cumulus with a decoupled BL.

    At end of simulation cloud fraction remained approx. 100% near top of BL which is an overestimate of cloud cover.

    Lagrangian 2 -- Subsidence assumed constant throughout period. Simulation gives a drop in BL height followed by a growth while it was observed to rise steadily. Was it the model or data driving it?

    Steve Krueger

    2D simulation using the UUCEM.

    Lagrangian 1 -- Results comparable to Chris except that cloud fraction was much less than 100%. Shows substantial difference in simulation due to radiation.

    Lagrangian 2 -- Steve's results more similar to Chris' than either are to observations.

    1D models

    Joao Texceira

    Lagrangian 1 -- BL depth varied reasonably. Cloud cover not bad. LWP ok earlier on but poor later on. What maintained the cloud was not the stratocu para. but large scale condensation from radiation cooling.

    Sensitivity -- Not sensitive to radiation scheme frequency of activation. Small time steps smooth oscillations but cloud disappears.

    Lagrangian 2 -- Produces convection immediately and produces stratus cloud similar to Chris and Steve.

    Erik van Meijgaard

    Single column model for global-wide ECHAM4. Nonlocal diffusion (Troen and Mahrt, 1986; Holtslag and Moeng, 1991. Cloud scheme generally based on Sundqvist's scheme.

    Lagrangian 2 -- LWC's predicted much less than observed partly because mass flux scheme very active. Cloud cover looked very good.

    Lagrangian 1 -- Cloud base very low. Cloud cover solid stratus early on changed to cu and lower cover later as observed. LWC looks good (at 10 hrs later on it creates too high a cloud. Sensitivity experiments with mass flux scheme. Modified entrainment rate which results in mass flux decreasing with height. Increasing entrainment rates doesn't improve results.


    LWC underestimated by standard scheme. Increase of lateral entrainment puts increased RH at low levels, but this is in wrong direction.

    Pier Siebsma

    Mostly discussion of representativeness of sounding. Diagnosed fractional entrainment;detrainment rates much larger than generally used in GCMs and detrainment is larger than entrainment.

    Steve Krueger

    Presents Don Lenschow's proposed measurement strategy for entrainment.


    Steve K.

    Lagrangian 1 -- Drizzle rates reasonably consistent. Sensible heat fluxes are consistent latent heat fluxes are more variable.

    Lagrangian 2 -- Lots of disagreement on cloud fraction. Drizzle rates vary a lot with obs being close to zero. Some models 2-3 mm/day.

    ASTEX Lagrangians

    First Lagrangian: Cases 1,3 -- General agreement with obs BL structure.

    Second Lagrangian: Cases 2,4 -- General agreement with cloud but not cloud evolution. Reasons not clear. Candidates: Initial soundings and spin-up time.

    Next steps:

    1. Clarify; standardize simulation procedures (relaxation; winds)
    2. Increase participation; submit results.
    3. Sensitivity issues; Case 2 initialization. Role of drizzle? (Address)
    4. Assemble more extensive aircraft etc. obs for verification. (Time scale?)
    5. Group paper; Summarize results (deadline date for inclusion?)

    The Smoke Cloud Case Study of the GCSS Boundary Layer Cloud WG
    Summary of Results and Ongoing Work from Aug 95 Workshop

    Chris Bretherton


    1. Even if the subgrid fluxes are at all levels smaller than the resolved fluxes, LES results are still sensitive to the numerical algorithm and subgridscale turbulence scheme.
      Centered difference advection schemes produced negative smoke, spurious temperature rises, and substantial spurious heat fluxes in the layer above the inversion. On the positive side, they also tended to produce relatively small entrainment rates, with typical A values varied from 0.5-1.
      In the control run, monotone schemes tended to have 'A' of 0.8-1.6, 50-100 they produce more numerical diffusion. The numerical diffusion appears to be comparable to the explicit turbulent diffusion. In one model (D. Stevens), no explicit diffusion was used but the results were comparable to other monotone schemes.
      A simple model of advection of an oscillating interface coupled to a mixed layer (B. Stevens) below produces entrainment rates comparable to those found in the models. This purely numerical entrainment mechanism as opposed to a more physical explanation such as wave-breaking, eddy rebound, etc. may be responsible for much of the entrainment seen in the LES runs.
    2. It was generally agreed that the vertical grid spacing specified in the control run (25 m) was too large to fully resolve the important motions at the entrainment interface. Enhanced vertical resolution decreases 'A' and the entrainment rate. D. Lewellen's results, corroborated by B. Stevens, suggest 'A' decreases to 0.4-0.5 if a 10 m or less vertical grid spacing is used and is insensitve to further increases in vertical resolution. The entrainment process was not found to be sensitive to changes in the horizontal resolution between 50 and 320 m. However, 2D high resolution simulations by B. Stevens and D. Stevens suggest that if vertical and horizontal grid spacing are both increased, the decrease in 'A' may be much smaller than simply increasing vertical resolution alone, confusing the issue a bit. Two considerations may contribute to the need for higher vertical resolution:
      Radiative cooling in the inversion and top of the smoke layer is too vertically concentrated to be resolved.
      Scaling arguments provide a height scale zu = wstar**2/delta-b for undulations of the inversion. For the smoke cloud case, zu is approximately 3 m. The LES cannot resolve these undulations unless dz is comparable or less than zu, possibly leading to a poor representation of the entrainment-forcing motions at the entrainment interface.


    3. From 1 + 2 and cross sections, it was clear that implicit or explicit diffusion, rather than resolved overturning, was the entrainment mechanism in all of the LES runs. Both horizontal and vertical resolution would have to be increased 5-10 fold to see reolved scale overturning, which is not computationally feasible in the immediate future. There was debate about how important a drawback this was.
    4. M. MacVean showed that most models had large residuals in the kinetic energy budget at the inversion. He traced this to an energetic inconsistency between the usual finite-difference representation of the buoyancy term in the vertical momentum equation and the vertical advection term in the potential temperature equation. The implied conversion between potential and kinetic energy due to buoyancy flux are different in the two equations, leading to a nonconservation of total energy that can be quite severe in a poorly resolved inversion such as occurs in the control run. He showed how the buoyancy could be rewritten to remove this inconsistency in a scheme using centered time differencing, and found that this significantly decreased A in the control simulation and almost entirely removed the residual in the KE budget. He could not find an analogous way to improve energy conservation at the inversion for forward-in-time schemes,
    5. Sensitivity studies of the control results to changes in forcing strength and distribution and inversion strength were performed by several groups (Lewellen, MacVean, B. Stevens, Siebesma and Cuijpers, Moeng, Rand and Bretherton). The main purpose of these studies was to test whether the parameterization we/wstar = A/Ri did correctly describe the effects of changing the forcing. Contradictory results ranging from scaling laws that we/wstar = c*Ri**(-0.6) to we/wstar = c*Ri**(-2.5) were obtained. Differences may partially be due to temporal variability in entrainment rate and turbulent transports, especially for high Ri. High vertical resolution studies of Lewellen supported an Ri**(-1) scaling. The entrainment rate also decreased markedly if more of the cooling took place below the entrainment region.
    6. Despite the above sensitivities, many resolved features of the convective turbulence, such as skewness profiles, were at least qualitatively similar between all LES models. Not surprisingly, most differences between models appear to be associated with the dynamics and subgridscale processes at the entrainment interface.
    7. Several groups also performed 2D simulations, and all found much higher values of A between 3-5 and of entrainment rate, compared to the 3D runs.
    8. Two groups also tried one-dimensional models (P. Bechtold and J. Cuxart). Bechtold required enhanced vertical diffusion of smoke to maintain a well-mixed smoke profile. Cuxart found A = 2.3, lowering to 1.5 with 5 m vertical resolution, with a stability dependent Prandtl number. More general conclusions could be drawn if more 1D modelling groups tried to run this case.
    9. Some groups (Cuijpers, Lewellen) also considered the impact of adding moist thermodynamics back into the control run without mkaking the cloud radiatively active and found this tended to increase A.

    Planned Papers and Future Work

    1. A group paper, with M. MacVean as lead author, will be prepared during this winter, including the above conclusions and sample results. This paper will focus exclusively on results presented at the workshop. (Malcolm intends to write a separate note about point 4 above, as well).
    2. One or more further papers in which we try to resolve some of the issue raised in this workshop, in particular points 2, 3, and 5. These papers will involve further simulations and analysis, with limited intercomparison of results between groups, and further work will be coordinated using email, WWW and phone. Groups indicating some interest in further participation included Lewellen, MacVean, B. Stevens, Siebesma and Cuijpers, Rand and Bretherton and D. Stevens

    Principal topics to investigate further:

    Nail down effect of not resolving radiative cooling and placing too much cooling in the inversion above the actual turbulent mixing. It was agreed to remove cooling in the inversion from future simulations by only allowing points with S 0.5 to contribute to the smoke path and radiative cooling. This is analogous to a stratocumulus cloud, where intense cooling requires the presence of condensate in a mixture, which typically requires that the mixture be at least 90 air.
    Continue investigating the effects of vertical and horizontal resolution, particularly to nail down whether low horizontal resolution but very high vertical resolution is sufficient and how the required vertical resolution can be predicted a priori. One tool that may be useful here is simulations with a nested region of both horizotnal and vertical refinement in the entrainment zone (could be investigated bu D. Stevens and B. Stevens)
    Try to get agreement between groups on the dependence of we/wstar on the forcing, perhaps using higher vertical resolution in the inversion.
    By increasing the radiative forcing dramatically, do a simulation with much lower Ri (on the order of 10) in which the entrainment process should be fully resolvable and test whether A agrees with lab experiments for this case. For such a case, a direct numerical simulation with constant viscosity and diffusivity should also be meaningful.

    Group agreed to met at Clermont-Ferrand next August.

    After the meeting Steve Krueger (on Bjorn's single speed rental) and I took a bike ride to the north. Steve had a hard time with the brisk head winds.

    We followed winding back country roads with the usual canals along side. In Westbroek we took pictures of a large operating windmill. Very picturesque. From there we found a recreational lake with many sailboarders on it. Steve wants to get back there to sail Sunday or Monday.

    We then followed a bike path around the lake and then back to the hotel.

    Bjorn, Steve, the student from Penn State working with John Wyngaard and I went to dinner in DeBilt. Had a good meal and a few real great tasting beers.

  • 8/2/95-Saturday
  • Start of the long day back to the Fort. Bicycled to Utrecht, broke down my bike and headed by train to the airport. It is convenient to have the train go right into the station. Too bad DIA doesn't do it too.