What is the Proper Set Up for Wind in outdoor Conditions

[Victor]

Hi 

I am trying to model a simple vent release in outdoor condition (open boundaries) with wind blowing to a particular direction with specified velocity. 

I noticed that there are at least two simple ways to define the wind:

#1. Under MISC

&MISC TMPA=34., LAPSE_RATE=-0.01, U0 = 2. /

#2. Create SURF_ID

&SURF ID='WIND',VEL=-2.0,TMP_FRONT=34., PROFILE='ATMOSPHERIC',Z0=10., PLE=0.12 / 

I have tried running a simplified case (specify nothing else but wind only) with #1 only, #2 only and both #1&#2. The results differ. 

Using #2 only, seems like the velocity that I specified does not reach far from boundary of the SURF_ID. In this case, velocity 2m/s will slowly drop and loose its impact at a certain distance within my mesh, just a few meters from SURF_ID boundary. My vent is located at the middle of the mesh, and as a result the release does not see wind impact in this case.

Using #1 only, I got almost uniform velocity throughout my mesh along the wind direction, with just slight fluctuation

Using both #1 and #2, I noticed a weird impact that a velocity greater than 2m/s (around 5m/s) flowing towards wind direction, followed by a so-called 'wind pocket'  with velocity of much less than 2m/s (around 0.7m/s). 

With these observations, for simulating outdoor condition, I am beginning to feel that using only #1 is more appropriate. Can anyone give view on this? 

And if you agree with using #1, my next problem is that now I could not specify the PLE, as I could only do so under SURF_ID (#2). Will this still be modelled as atmospheric profile? 

Or is there any better way to represent a simple wind profile set up for outdoor?

Appreciate your kind feedback. 

Victor

[Kevin]

Actually, if you use U0, you must also use VEL on the SURF line. The purpose of U0 is to establish a uni-directional wind field at the very start of the simulation. If you only use VEL, then you must wait a certain amount of time for the wind field to be established. U0 is just a way to speed up the establishment of a wind field.

I would dedicate a certain amount of time in your simulation just to establish the wind field. If you notice anything strange or unusual, set up a simple case and submit it to the Issue Tracker.

[Randy McDermott]

See also the discussions on the use of OPEN vents on p. 55, MEAN_FORCING on p. 39, and outdoor flows p. 107. in the latest user guide:

https://drive.google.com/folderview?id=0B_wB1pJL2bFQaDJaOFNnUDR4LXM#list

In particular, it is not recommended to use OPEN vents for flows that are predominantly parallel to the boundary.  The reasons are discussed in the guide.

--
You received this message because you are subscribed to the Google Groups "FDS and Smokeview Discussions" group.
To unsubscribe from this group and stop receiving emails from it, send an email to fds-smv+u...@googlegroups.com.
To post to this group, send email to fds...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/fds-smv/e8c3968e-8aab-42d6-8f89-c02f69590913%40googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

[Victor]

Hi Kevin 

Thanks for your reply. I have run it for quite long time (around 5 minutes) and the wind of 2m/s set under SURFID does not reach my vent. If it needs even longer, I am afraid the dispersion from vent already ceased by the time the wind reach there, so no point.

We initially used both U0 and VEL on SURF line but we noticed a weird wind profile. I will submit it in issue tracker, complete with results figure and FDS file, hope you can take a look

So you think the best way is to set up both U0 and VEL on SURF line? Would it not 'double counting' the velocity that I intend to set?

We predominantly use FDS for outdoor simulation, so I just need to get this simple thing right before moving forward.  

Appreciate your help

Cheers, Victor

[Victor]

Hi Randy

Thanks for your help. I replied your message but it seems that I might have clicked the wrong button and replied directly to your email. Let me know if you have my message.

Looking forward to getting your feedback

Cheers, Victor

[Kevin]

I need to see a simple example. Use a fairly coarse grid, just to help identify the problem quickly.

The use of U0 is not double counting. All that U0 does is to initialize the x-component of velocity to this value. Nothing more. If you just set U0, and nothing else, your wind will eventually disappear.

[Randy McDermott]

First, it must be stated that outdoor flows with FDS are research areas.  This work should not be considered validated.

Attached is a case that I developed for another user who wanted to maintain an atmospheric profile.  Imposing a profile at one boundary makes no sense.  If you want that profile to persist, the forcing and the other boundary conditions must work together to maintain the profile.  The FDS OPEN bc is ill-conditioned for this purpose.  It should not be used for the top or sides of the domain.  The reason for this is discussed in the user guide.

I recommend you play with U0 and VEL_T(1), and possibly the turbulence intensity on the top of the domain, to match your profile in the mean.  If you have second-order statistics (they matter for dispersion!) then try to match those as well.

In replying to this thread, I just now realize that I have not written up the TURBULENT_DISPERSION model that may be applied to particle tracers.  If you are using the model for outdoor pollutant dispersion, I recommend the Lagrangian approach.  The Eulerian approach gives way too much numerical dissipation under typical outdoor flow grid resolutions.  See, for example, the random_walk_1 and _2 cases in the WUI verification folder:

https://code.google.com/p/fds-smv/source/browse/trunk/FDS/trunk/Verification/WUI/random_walk_1.fds

To view this discussion on the web visit https://groups.google.com/d/msgid/fds-smv/ccd768ee-d8b8-46bd-bcf9-c9389e7309bf%40googlegroups.com.

[JRTatitsa]

Dear All,

I did quite a few experiments in setting up wind fields a couple of years ago using FDS5, with grids up to 200 m. high and 2 km. in length, using meshes down to 1 m^3. I think you have covered the findings well, but some observations may be interesting.

Firstly, using U0 together with an atmospheric profile for the upwind vent, worked well, but with the upper surface OPEN, momentum gradually diffused into the upper atmosphere. The problem was solved by making the upper surface CLOSED, but with a tangential velocity which matched the atmospheric profile. Turbulence from the upper atmosphere was supressed in this way, which is unfortunate, but with a large depth for the boundary layer I do not think this makes too much difference. This covers the case for neutral atmospheric stability, as implicit in the lapse rate and wind profile settings. Varying lapse rate did not seem to have much effect however (This was FDS5, I will repeat for FDS 6)

Creating unstable atmospheric conditions proved fairly easy - I made a heated surface at ground level. Velocity profile and turbulence corresponding to current analytic models were generated in a few minutes of simulated time.

Creating stable atmospheric profiles proved difficult - it appeared that you would need many kilometres of upwind simulation to achieve a realistic turbulence profile. This is still an area for investigation for me. I need to revisit this issue with FDS6, which has more features.

One important observation - using LES for this kind of outdoor gas and smoke dispersion is important - there are many phenomena which arise in terms of interaction with terrain and with gas and smoke cloud dynamics which are not apparent in standard analytic models such as Gaussian dispersion models. The problem is that standard models work in terms of long term averages. I repeated some of the phenomena from the FDS5 simulations in small scale outdoor experiments. As an example of the difference between steady state modelling and real life - in mildly unstable conditions, evaporation from a pool would build up a well formed plume. Then a turbulent eddy (gust) would come along and wipe the surface of the pool clean. Virtually all of the dispersion mass transfer was due to these turbulent eddies. The implication - that toxic vapour dispersion from a pool may match current models on average over a longer period, but this average may be the results of periods of relatively high concentration interspersed with short periods of very low concentration, which is about the worst kind of discrepancy between model and reality.

Wit best regards, Robert Taylor

[Brian L]

HI JRTatisa,

Any updates with your work? Did you find anything with FDS6 regarding this issue of outdoor flows?

[Randy McDermott]

Since 2014 Kuldeep Prasad has been doing work to couple FDS with WRF for the goal of modeling outdoor pollutant dispersion.  In my view, this represents the current best guidance on outdoor flows with FDS.  Here is an example input file:

https://github.com/firemodels/fds-smv/blob/development/Verification/WRF/wrf_time_prof_ramp.fds

On Tue, Apr 5, 2016 at 12:05 PM, Brian L <liqui...@gmail.com> wrote:

HI JRTatisa,

Any updates with your work? Did you find anything with FDS6 regarding this issue of outdoor flows?

--

You received this message because you are subscribed to the Google Groups "FDS and Smokeview Discussions" group.
To unsubscribe from this group and stop receiving emails from it, send an email to fds-smv+u...@googlegroups.com.
To post to this group, send email to fds...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/fds-smv/383ee755-4554-41d8-96d1-94d33d665634%40googlegroups.com.

[Chris Kohler]

We made for a Project a Wind-Profil 0 m, 0 m/s - 2.5 m, 2.0 m/s - 5.0 m, 4.0 m/s - 7.5 m, 6 m/s - 10 m, 8 m/s which is actually in the range of a Beaufort 4-Wind (Beaufort 4 = 5.5.-7.9 m/s) which is actually quite high. 

[Kuldeep]

 

Atmospheric flows are usually modeled with the Weather Research and Forecasting (WRF) software. WRF mesoscale simulations are based on RANS models and use a parameterized turbulence model. WRF calculations are usually performed  at 1-10 km resolution.  At such resolution it is difficult to study the effect of buildings / communities. Since we are interested in flows over urban domains (cities such as Los Angeles, Washington DC, Baltimore),  high resolutions LES can be helpful for studying plume dynamics and mixing over cities.  We would like to  obtain large mesoscale winds from a WRF calculation or from observations (weather stations on towers, lidars etc) and use the meteorological data to nudge the mean FDS flow field. We have done some preliminary simulations over large cities at a resolution of 10- 40 m, where the fine scale dynamics is predicted by LES, while the mesoscale winds are obtained from WRF. The code that Randy has posted will enable you to do this coupling between FDS and WRF.

 

Modeling atmospheric flows is quite challenging. FDS has all the functionality and physics, but we need to make sure that it is getting the right input data. Radiative heating of the earth's surface, soil moisture content, evaporation and energy balance at the surface are important parameters which can affect boundary layer dynamics.

 

We are starting to look at how well FDS can simulate a convective, well-mixed, day-time, unstable boundary layer. I am also interested in looking at neutral and stable, night-time flows. If this group is aware of well controlled / documented experimental data for atmospheric boundary layers that can be used to validate FDS for atmospheric flows, I would be interested in working with you.

[Espen S Gåserud]

Hi Randy,

The link you provided to Kuldeep Prasad's work does not work anymore. Do you have an updated link or some more info so that I can find it via google?

regards

Espen S Gåserud

[Espen S Gåserud]

Hi,

 

Using U0, MEAN_FORCING(1)=.TRUE. and DT_MEAN_FORCING under MISC, will this produce a typical atmospheric wind profile with zero velocity at ground level and gradually increasing wind speed as function of height above ground (e.g. parabolic)?

 

I have made a simple test case, but the free-flow velocity seems to be constant at all heights above the ground. I have only ran it for a short time (10 s), so it could be that it needs to run longer for the profile to develop.

 

Attached is the fds-file for your information.

Appreciate your kind feedback.

Espen

[Randy McDermott]

Here is the updated link after we reorganized the repos.

https://github.com/firemodels/fds/tree/master/Verification/WRF

--

You received this message because you are subscribed to the Google Groups "FDS and Smokeview Discussions" group.

To unsubscribe from this group and stop receiving emails from it, send an email to fds-smv+unsubscribe@googlegroups.com.

To post to this group, send email to fds...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/fds-smv/6092fc7d-b2cb-432b-ad59-e8069092fb03%40googlegroups.com.

[Randy McDermott]

Regarding the atmospheric profile, yes, you need to run much longer than 10 s to see the boundary layer develop.

However, there has also been work done to specify the z-profile with mean forcing.  You can explore the guides and some of the examples in

https://github.com/firemodels/fds/tree/master/Validation/LNG_Dispersion

Note, however, that we have been working on the forcing capabilities and the next release may have significant improvements.  But these improvements will alter the time scale that is required.  I don't want to go into this further right now.  Just be aware that DT_MEAN_FORCING is a parameter of the problem and depends heavily on the way the forcing was implemented.

[Espen S Gåserud]

Great, thanks! Using RAMP I was able to get the velocity profile I wanted (power law profile with U=10 m/s at Z=10 m, and exponent p=0.3).

 

I noticed a thing that I found kind of strange. When I first specified U0=10.0 under MISC, my velocity profile was multiplied with a factor of ten. But when I changed this the U0=1.0 I go the correct profile. Any idea why? It appears to be connected toe use of RAMP, because when I did not use RAMP I did not get the same increase in velocity.

-Espen-

To unsubscribe from this group and stop receiving emails from it, send an email to fds-smv+u...@googlegroups.com.

To post to this group, send email to fds...@googlegroups.com.

[Randy McDermott]

The ramp is a factor F(z)*U0.

To unsubscribe from this group and stop receiving emails from it, send an email to fds-smv+unsubscribe@googlegroups.com.

To post to this group, send email to fds...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/fds-smv/6412c748-208c-483e-b510-3331e1be6d39%40googlegroups.com.