Glossary#

AGL#: Above ground level. Receptor altitudes are AGL unless you set altitude_ref="msl".
ARL#: The file format for meteorology used by HYSPLIT and STILT, named after NOAA’s Air Resources Laboratory. NOAA publishes HRRR, NAM, GDAS, and other models in this format.
averaging kernel#: How sensitive a column retrieval is to each height in the atmosphere. Applying it weights particles released at each height accordingly.
backend#: Where the simulations run. The choices are local (your computer), slurm (an HPC cluster), and kubernetes (cloud, experimental). Set it in the execution section of config.yaml. See Where To Run.
defaults#: The settings written at the top level of config.yaml. Every variant starts from them and overrides what it changes. A variant with no overrides runs them as they are.
derived variant#: A variant declared with from: <other>. It makes another footprint (a different grid, smoothing, or particle weighting) from that variant’s stored particles without running HYSPLIT again.
empty footprint#: A simulation that ran fine but whose particles never touched the footprint grid, for example because the grid is too small or is not upwind. PYSTILT records it with a small .empty file instead of a NetCDF, and counts the simulation as finished.
footprint#: A map of how much each surface grid cell influenced a measurement. Each value is the concentration change at the receptor per unit of surface emission in that cell, in ppm per (µmol m⁻² s⁻¹). Multiply by an emissions map and sum to get a modeled concentration increase. Stored as a Footprint (a NetCDF file on disk).
hyper-near field#: The area right around the receptor, where the particle cloud hasn’t yet spread out enough to represent mixing well. The hnf_plume setting (on by default) corrects the footprint there with a Gaussian plume model.
HYSPLIT#: NOAA’s particle transport program, written in Fortran. PYSTILT runs it to move particles and includes a copy for Linux and macOS (Intel).
met name#: The name you give a meteorology source in your settings, such as hrrr. With no variants declared, each met source is a variant of the same name, so the same receptor can be run with several meteorology products side by side.
meteorology#: Gridded weather-model fields (winds, temperature, turbulence, boundary layer height) that STILT uses to move particles. Often shortened to met. Must be in ARL format. See Meteorology.
MSL#: Above mean sea level.
n_hours#: How many hours to follow the particles. Negative values run backward in time, which is the usual case for measurements. -24 means one day back.
numpar#: The number of particles released per simulation. More particles give smoother footprints and take longer. 200 to 1000 is typical.
particle#: An imaginary parcel of air released at the receptor and moved backward in time by the winds, plus random turbulence. STILT releases many (numpar) and treats them together as a sample of where the air could have come from.
pressure weighting#: Weighting column particles by the share of the column’s air mass they represent. HYSPLIT releases particles evenly in height, but air thins with height, so a particle high in the column stands for less air.
project#: A folder holding your settings (config.yaml), your receptors (receptors.csv), and all outputs (simulations/). See Project Folders And Reruns.
realization#: One member of a variant declared with realizations: N. The same settings run N times, named <variant>-0 to <variant>-(N-1), and each run gets its own random draws. Realizations are used for transport-error ensembles (Transport Error).
receptor#: Where and when a measurement was made, which is where STILT releases particles. A PointReceptor is one point (a tower inlet), a ColumnReceptor is a vertical column, and a MultiPointReceptor is a set of points, such as a slanted line of sight. See Receptors and Slant Columns.
simulation#: One receptor run under one variant: one set of trajectories and at most one footprint. A project with 100 receptors and 3 variants has 300 simulations.
simulation ID#: The name of a simulation and of its output folder, such as 202307151800_-111.848_40.766_10/hrrr. It is the receptor id, a slash, and the variant name. The receptor id is the receptor time as YYYYMMDDHHMM, then the longitude, latitude, and altitude. Column receptors end in _X instead of an altitude. Multipoint receptors use multi_ and a short hash instead of coordinates.
trajectory#: The path of one particle through time. PYSTILT stores all of a simulation’s particle paths together as a Trajectories table (a Parquet file on disk).
transform#: A step that changes how much each particle counts before the footprint is calculated. Used for column measurements (averaging kernel and pressure weighting) and for gases that decay in the atmosphere. See Particle Weighting (Transforms).
variant#: A named set of settings that every receptor is run under. It is a met source plus any settings that differ from the defaults in config.yaml, such as a different ziscale, a wind-error run, or another footprint grid. See Configuration.