Weather Research and Forecasting –
Stochastic Time-Inverted Lagrangian
(WRF-STILT) model
Derek Mallia
LAIR tutorial
PART I: STILT basics
Backward trajectory simulation using the
WRF-STILT model for Hurricane Sandy
Basics of the STILT model
• The STILT model is a 4D (x,y,z,t) Lagrangian particle dispersion
model (LDPM) for atmospheric transport
• The model was primarily developed to derive upstream influence
regions for some receptor that we are trying to model for
• Can also be used qualitatively to determine the origin of your air mass
• Driven by meteorological models/analyses such as the ECMWF,
• However, WRF output cannot be directly used with the STILT model
without converting the netcdf file format to “ARL” format which is
used by the HYSPLIT model core in STILT
• For optimal results, 3-hourly averaged winds are required (better mass
For non-WRF analyses
• Go to:
• I generally use these to get a “quick” first look
• Also good for using as an outer domain
• I.E if your particles leaves your outer WRF domain, you can still
track them beyond that domain if you nest it with something like
GDAS which is “global”
• Downside of these products?
• Generally, lower spatial and
temporal resolution
• Worse mass conservation (see
Nehrkorn et at. 2010)
• Not necessarily tailored to your
GDAS domain
WRF domain
Uinta Basin example:
In the inversion…
Outside of the inversion…
Relevant papers for details
Lin, J.C., C. Gerbig, S.C. Wofsy, et al., A near-field tool for simulating the
upstream influence of atmospheric observations: The Stochastic Time-Inverted
Lagrangian Transport (STILT) model, J. Geophy. Res., 108(D16), 4493,
doi:10.1029/2002JD003161, 2003.
Nehrkorn, T., J. Eluszkiewicz, S.C. Wofsy, Lin, J.C., C. Gerbig, M. Longo, and
S. Freitas. Coupled Weather Research and Forecasting--Stochastic TimeInverted Lagrangian Transport (WRF-STILT) Model, Meteorology and
Atmospheric Physics, 107, 51-64, 2010.
Cool model bro, where do
I get this model?
• First, need to download the STILT model from:
• Also need to register
• Better directions can be found here with FORTRAN hotfix:
Fixes random model crash…
How to run guide:
• Model uses name list file to run simular to that of WRF, however,
can be ran using a “R” using the scripts provided in the stiltR
• Makes the model less clunky to use, however it requires becoming slightly
familiar with R (very easy to use)
• Need help? Feel free to bug me (WBB 723)
Running STILT at the FORTRAN level
• For now lets just copy the hymodelc executable (has all of the
necessary bug fixes) along with my modified stiltR code and
meteorological input files which can be found here:
• /uufs/
• Now lets get familiar with running the STILT at the
FORTRAN level…
Two “namelist” files for STILT…
See STILThowtoV8.pdf for variable
Only these lines need to be
See STILThowtoV8.pdf formatting of this
• Type hymodelc into you command line once we have finished
editing our CONTROL file…
• If the CONTROL file was properly edited you code should be
running! (will take a few moments)
Sample output file…
Running STILT using the stiltR interface
• Lets cd into the stiltR
directory and get familiar
with some of the most
important subroutines
• Once your have your
bearings lets open up
with your favorite text editor
• Lets enter R now (type R into
the command line within the
stiltR directory and run our
tutorial code
d02 & 3
Sample output
Units: PPM / μmoles m-2 s-1
PART II: STILT advanced settings
STILT model installed now what?
• Need to get WRF data in ARL format and need to output
specific variables that are required by the STILT model
• Involves installing a converter program and editing the WRF
registry + recompiling the WRF
• Also had to fix some bugs as well
• No worries, I supplied the ARL converter code, additional
FORTRAN fixes, and easy to read instructions for WRF registry
• Directions for ARL converter code installation and running can be
found here:
– Most of the starting steps can be skipped as I did this already!
– Don’t bother with having more than 100 WRF vertical levels… lots of ISSUES
– ZSG code can be found here (for STEP #8)
Additional steps…
• Additional all of the files found here:
• This subroutines will need to be copy pasted into your
merged_hysplit directory that was used to compile the original
STILT model code
• HELPFUL TIP, copy the merged_hysplit directory and make a new
one which will be used to recompile STILT with the update code above
• This is needed to fix some minor mass conservation issues and random
particles dropping out for no reason (has to do with particle being
dropped from memory as they leave the sub domains)
• If you use stiltR, copy and paste Trajec.r to your stiltR directory which
adds additional mgmin variable to prevent above issue
• If not, additional line needs to be added to the SETUP.CFG file with
the following MGMIN = 2000,
Sample output…
Footprints x Emissions = Footprint of emission source (mole fraction)
Units: PPM / μmoles m-2 s-1
Units: micromoles / m2 s
PPM /μmoles m-2 s-1 * μmoles / m2 s = PPM of stuff

similar documents