Tutorial 2: An atmospheric boundary layer
This tutorial adds two features of a real atmosphere to Tutorial 1: A neutral surface layer: the Earth’s rotation (Coriolis
force) and a temperature inversion that caps the boundary layer. It is ERF’s Neutral_ABL_Flat case, solved
directly for the steady state. The run takes about 15 seconds.
You will learn to: drive the flow with a geostrophic wind, prescribe a temperature sounding and check the result with ERF’s own check script.
The physical problem
Above the boundary layer, the wind is in geostrophic balance: the Coriolis force balances the large-scale pressure gradient, and the wind is \((u_g, v_g) = (10, 0)\) m/s. Near the ground, friction slows the wind and turns it to the left (in the northern hemisphere). The result is the Ekman spiral.
The potential temperature is 300 K up to 700 m and rises above, 0.03 K/m. This stable inversion limits the depth of the boundary layer.
The input file
Tests/Canonical_RANS/Neutral_ABL_Flat/inputs_steady_neutral_flat differs from tutorial 1 in these lines:
# A larger horizontal domain, split into four boxes
mesh.n_cell = 8 8 64
mesh.max_grid_size = 4 4
# Temperature sounding: 300 K to 700 m, 309 K at 1000 m, 0.03 K/m through the top
bc.theta_profile_z = 0. 700. 1000.
bc.theta_profile = 300. 300. 309.
bc.zhi_theta_grad = 0.03
# Coriolis forcing with a geostrophic wind of 10 m/s (f = 1e-4 1/s)
forcing.coriolis = 1
forcing.coriolis_3d = 0
forcing.latitude = 90.
forcing.rotation_period = 125663.7061435917
forcing.geostrophic = 1
forcing.ug = 10.
forcing.vg = 0.
# Initial profile: the log law with u* = 0.35 m/s (tutorial 1: 0.25 m/s)
bc.inflow_ustar = 0.35
# Solver: four z-line sweeps, CFL cap 2
steady.sweep = zjacobi
steady.cfl_max = 2.
steady.zline_sweeps = 4
steady.residual_ref_iterations = 100
steady.residual_drop_orders = 2
turb.tke_floor = 1.e-6
The pressure-gradient force of tutorial 1 is replaced by the geostrophic forcing. The lines
turb.max_geom_lscale and turb.face_mean are removed, so ERF’s defaults apply: a length-scale cap of 30 m and
arithmetic face averages.
Why only two orders of residual drop? The boundary layer slowly entrains the warm air of the inversion and warms
(by about 0.01 K per day) without ever stopping. The \(\theta\) and \(w\) residuals therefore level off
about 2.5 orders below their starting value. A drop of 2 orders, together with steady.residual_tol, still gives a
converged wind field.
Run the case
mkdir -p runs/flat && cd runs/flat
mpirun -n 2 ../../build-rel/gust ../../Tests/Canonical_RANS/Neutral_ABL_Flat/inputs_steady_neutral_flat
The run ends with
2840 2.000e+00 zjacobi 1.793e-05 6.202e-06 5.720e-06 1.488e-05 7.754e-06 6.324e-08 6.816e-03
gust: steady after 2840 iterations (2840 over the run)
Unlike tutorial 1, all six residuals are non-zero: the Coriolis force couples \(u\) and \(v\), and the inversion couples \(w\) and \(\theta\). The largest wind speed is 11.1 m/s, slightly above the geostrophic wind (a low-level jet below the inversion), and \(u_* = 0.377\) m/s.
Check the result with ERF’s script
ERF provides a check script for this case. It reads the turbulence plotfile, which uses ERF’s variable names:
python3 ../../extern_erf/check_neutral.py --physics steady_neutral_flat_turb02840 steady_neutral_flat_turb02840/surf_hist.dat
It prints one line per check and ends with 0 check(s) failed. Among the checks:
u_star [m/s] plausible 0.377402 0.35 0.15 abs yes
KE(k=0)/u*^2 vs 1/Cmu0^2 3.23251 3.2325 0.05 rel yes
|U|(k=0) vs log law 4.0235 4.0235 0.1 rel yes
What you have seen
Coriolis forcing and a sounding are set in the
forcingandbcnamespaces.With rotation and stratification the solver needs a smaller CFL cap than the column (2 instead of 50).
ERF’s check scripts work directly on GUST’s turbulence plotfile.
Next: Tutorial 3: Flow over a hill adds terrain.