Compare particle and boundary fuel methods for vegetation

[rmcdermo]

It looks like there is already one comparison for these methods:

WUI/ground_vegetation_load.fds

But Ruddy was worried that the grid resolution can have a big impact on the consistency. We need to make sure we span the range of vegetation thickness compared to grid resolution. When the bed thickness is small compared to DZ then the surface behaves like a typical SURF. When the bed thickness is large compared with DZ then the boundary fuel model should be have like each node in the solid phase model is a Lagrangian particle interacting with the boundary layer at a certain height from the ground level.

[ruddymell]

Also, the temperature equation in the model for vegetation represented by particles does not have a conductivity term, while the boundary fuel model does have this term. For consistency with the particle model the conductivity term in the boundary fuel model should be removed.

[mcgratta]

I ran one of the Catchpole cases (EXMC91) at three grid resolutions for both the particle fuel model and the boundary fuel model. The results are summarized here: EXMC91_Sensitivity.pdf Has this ever been done before?

[usfs-fds]

These tests are very important. We did similar tests years ago with a significantly different version of the code. For the Catchpole experiments a better case to compare the the boundary and particle models (and grid sensitivity) would be one with a larger bulk density. The EXMC91 case considered here has a bulk density of 1.99 kg/m^3 and an optical depth of 10.5 cm. So, the grid resolutions of these tests (2.5, 5, 10 cm) is much closer to resolving the decay of radiation intensity in the vegetation than a case with a larger bulk density of 32 kg/m^3 (PPMC2H) which has an optical depth of 1.1 cm. It's clearly better (for physical fidelity [it's 3D] and from your results) to use the particle model if we can. But the concern is that at higher bulk densities it forces the gas-phase to have too fine a grid (i.e., it's costly).

[mcgratta]

I ran the Catchpole PPM2H case at the same three resolutions (2.5 cm, 5 cm, 10 cm) with both the particle method (LPM) and the boundary fuel method (BFM). The results are shown here: PPMC2H_Sensitivity.pdf Clearly there is something amiss with the boundary fuel method, but I have not pinpointed it yet. Have you ever run this case with the old WFDS?

[ruddymell]

The problem is almost certainly due to not having a bound on the mass loss rate. If I run the old WFDS BFM model (but I've put it in the current FDS for consistent comparison) without this bound, I get essentially the same behavior as FDS of a much too large spread rate. With a pyrolysis mass loss rate bound of 0.35 kg/s/m^2 (which is based loosely on the Douglas fir tree burns) the old BFM WFDS (10 cm grid resolution) gives a spread rate result similar to FDS LPM.

[mcgratta]

OK, I forgot about that parameter. I'll take another look.

[mcgratta]

I assume you mean 0.35 kg/s/m^3 ?

[ruddymell]

oops, yes, 0.35 kg/s/m^3

[mcgratta]

I ran the cases with a MINIMUM_BURN_TIME of 91 s. This comes from dividing the bulk density of 32 kg/m^3 by the max burning rate of 0.35 kg/s/m^3. However, the fire front still spreads quickly while the fire width extends the entire length of the fuel, 8 m. By what mechanism in the Boundary Fuel Model does the fuel out in front of the fire not heat up too quickly? When using the Lagrangian particles, the dense foliage literally blocks the heat gases from pre-heating fuel. For the BFM, I don't see that happening.

[mcgratta]

The problem of over-predicting fire rate of spread through vegetation has to do with the char model, and the over-heating of particles or small bits of vegetation. Until a better char model can be developed, FDS employs an upper bound on the reaction rate for char of 1 1/s. This prevents char temperatures from hitting the FDS max temperature of 2500 C, which often leads to excessive spread rates.

New issues will arise, of course, but for now the V&V cases for WUI are behaving reasonably well.

[ruddymell]

I think any V&V case for vegetation and char oxidation needs to have the input file modified so that, for example in the case of USFS_Corsica/FDS_Input_Files/Test_2_0.fds, the following entries on &MATL ID='CHAR' SPEC_ID = 'PRODUCTS', 'AIR' NU_SPEC = 8.09,-7.2 are changed to SPEC_ID = 'OXYGEN', 'CARBON DIOXIDE' NU_SPEC = -1.65, 2.63 where oxygen and carbon dioxide are declared as species and NU_MATL=0.02 in &MATL ID='CHAR'. This for the char degradation kinetics scheme that's in the user guide.

Do you agree?

[mcgratta]

This would mean that we cannot use the SIMPLE_CHEMISTRY combustion model, for which there are only three TRACKED_SPECIES---AIR, FUEL, PRODUCTS. To do what you suggest, we would have to write out the reaction and track explicitly the species NITROGEN, OXYGEN, CARBON DIOXIDE, CARBON MONOXIDE, WATER VAPOR, FUEL, SOOT. This adds a considerable expense to an already costly simulation. Question is--- will this make a dramatic impact on the results?

[rmcdermo]

Have we figured out how to conserve mass and energy with variable specific heats, etc., without tracking O2, N2 separately? I thought this was still an issue.

On Fri, Jul 24, 2020 at 3:02 PM Kevin McGrattan notifications@github.com wrote:

This would mean that we cannot use the SIMPLE_CHEMISTRY combustion model, for which there are only three TRACKED_SPECIES---AIR, FUEL, PRODUCTS. To do what you suggest, we would have to write out the reaction and track explicitly the species NITROGEN, OXYGEN, CARBON DIOXIDE, CARBON MONOXIDE, WATER VAPOR, FUEL, SOOT. This adds a considerable expense to an already costly simulation. Question is--- will this make a dramatic impact on the results?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/firemodels/fds/issues/7921#issuecomment-663687031, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABBWXQOJVU2VFREKPIEFLKDR5HLE3ANCNFSM4KJVXNIQ .

-- Sent from my iPhone

[drjfloyd]

Could we just track AIR, FUEL1, FUEL2, PRODUCTS1, and PRODUCTS2 and have two REAC. Five speices vs. three and two fast reactions vs. 1. This would add more cost, but not a prohibitive amount. FIRE is generally a low single digit percentage of the run time and DIVG and MASS wind up around 20 to 30 % and not all of that is in loops over all species. What total cost would this really add? 15 to 20 %?

[mcgratta]

Sure, we can do anything we want, but the question is whether it's worth the cost.

[rmcdermo]

My view is that this is research right now and that we have to be careful about moving forward with slight errors that we end up tuning around. When we are that concerned about cost, we will likely be forced into using level set. Ultimately, we all agree (I think) that we need validation cases to be the arbiter of these approximations. Personally, I would err on the side of being tight with mass and energy conservation to start. I think the inputs can be streamlined vis-a-vis an N_REACTION=2 type of parameter.

[mcgratta]

I'm running the USFS/Corsica cases using both lumped and primitive species approaches. I first want to establish that we get the same results.

[ruddymell]

You may need to add the option in the input file to remove vegetation that has fully oxidized (only ash remains). For marginal fire spread conditions (in this case due to no imposed wind) this can make the difference between the fire surviving or not. If the consumed veg is not removed, it can sufficiently block the inflow to limit the O2 supply, etc. This will also more accurately represent the collapse of the fuel bed after consumption. I needed to do this in FDS for the crown fire initiation cases that I'm running now. They are working fairly well without a bound on the char mass loss rate. It will be good to see how the lumped and primitive species compare.

[mcgratta]

I re-ran the USFS/Corsica cases using 6 species (N2, O2, CO2, Soot, H2O, Fuel) compared to 3 (AIR, FUEL, PRODUCTS). The extra CPU time was only 5% greater, and the results are comparable. It's not clear whether the char oxidation done using the proper species O2/CO2 vs AIR/PRODUCTS had a significant effect, but the cost increase is not bad. I'll commit the cases and results to the Validation Guide and look at the other cases.

[mcgratta]

I'm going to close this issue. It has been used to document progress in WUI modeling. Most of the results are now in the V&V documents. We can address specific issues as they arise.