def data(n_output, rtols, schemes, setups_num):
resultant_data = {}
for scheme in schemes:
resultant_data[scheme] = {}
if scheme == "BDF":
for rtol in rtols:
resultant_data[scheme][rtol] = []
for settings_idx in range(setups_num):
settings = setups[settings_idx]
settings.n_output = n_output
simulation = Simulation(settings)
bdf.patch_particulator(simulation.particulator)
results = simulation.run()
for rtol in rtols:
resultant_data[scheme][rtol].append(results)
else:
for rtol in rtols:
resultant_data[scheme][rtol] = []
for settings_idx in range(setups_num):
settings = setups[settings_idx]
settings.rtol_x = rtol
settings.rtol_thd = rtol
settings.n_output = n_output
simulation = Simulation(
settings, backend=CPU if scheme == "CPU" else GPU)
results = simulation.run()
resultant_data[scheme][rtol].append(results)
return resultant_data
def test_water_mass_conservation(settings_idx, mass_of_dry_air, scheme, coord):
# Arrange
assert scheme in ('BDF', 'CPU', 'GPU')
settings = Settings(w_avg=setups[settings_idx].w_avg,
N_STP=setups[settings_idx].N_STP,
r_dry=setups[settings_idx].r_dry,
mass_of_dry_air=mass_of_dry_air,
coord=coord)
settings.n_output = 50
settings.coord = coord
simulation = Simulation(settings, GPU if scheme == 'GPU' else CPU)
qt0 = settings.q0 + liquid_water_mixing_ratio(simulation)
if scheme == 'BDF':
bdf.patch_particulator(simulation.particulator)
# Act
simulation.particulator.products['S_max'].get()
output = simulation.run()
# Assert
ql = liquid_water_mixing_ratio(simulation)
qt = simulation.particulator.environment["qv"].to_ndarray() + ql
significant = 6 if scheme == 'GPU' else 14 # TODO #540
np.testing.assert_approx_equal(qt, qt0, significant)
if scheme != 'BDF':
assert simulation.particulator.products['S_max'].get(
) >= output['S'][-1]
def __init__(
self, settings, products=None, scipy_solver=False, rtol_thd=1e-10, rtol_x=1e-10
):
env = Parcel(
dt=settings.timestep,
p0=settings.initial_pressure,
q0=settings.initial_vapour_mixing_ratio,
T0=settings.initial_temperature,
w=settings.vertical_velocity,
mass_of_dry_air=44 * si.kg,
)
n_sd = sum(settings.n_sd_per_mode)
builder = Builder(n_sd=n_sd, backend=CPU(formulae=settings.formulae))
builder.set_environment(env)
builder.add_dynamic(AmbientThermodynamics())
builder.add_dynamic(Condensation(rtol_thd=rtol_thd, rtol_x=rtol_x))
volume = env.mass_of_dry_air / settings.initial_air_density
attributes = {
k: np.empty(0) for k in ("dry volume", "kappa times dry volume", "n")
}
for i, (kappa, spectrum) in enumerate(settings.aerosol_modes_by_kappa.items()):
sampling = ConstantMultiplicity(spectrum)
r_dry, n_per_volume = sampling.sample(settings.n_sd_per_mode[i])
v_dry = settings.formulae.trivia.volume(radius=r_dry)
attributes["n"] = np.append(attributes["n"], n_per_volume * volume)
attributes["dry volume"] = np.append(attributes["dry volume"], v_dry)
attributes["kappa times dry volume"] = np.append(
attributes["kappa times dry volume"], v_dry * kappa
)
r_wet = equilibrate_wet_radii(
r_dry=settings.formulae.trivia.radius(volume=attributes["dry volume"]),
environment=env,
kappa_times_dry_volume=attributes["kappa times dry volume"],
)
attributes["volume"] = settings.formulae.trivia.volume(radius=r_wet)
super().__init__(
particulator=builder.build(attributes=attributes, products=products)
)
if scipy_solver:
bdf.patch_particulator(self.particulator)
self.output_attributes = {
"volume": tuple([] for _ in range(self.particulator.n_sd))
}
self.settings = settings
self.__sanity_checks(attributes, volume)
def test_just_do_it(scheme, adaptive, backend_class=CPU):
# Arrange
if scheme == 'BDF' and (not adaptive or backend_class is GPU):
return
settings = Settings(dt_output=10 * si.second)
settings.adaptive = adaptive
if scheme == 'BDF':
settings.dt_max = settings.dt_output # TODO #334 'BDF')
elif not adaptive:
settings.dt_max = 1 * si.second
simulation = Simulation(settings, backend_class)
if scheme == 'BDF':
bdf.patch_particulator(simulation.particulator)
# Act
output = simulation.run()
r = np.array(output['r']).T * si.metres
n = settings.n / (settings.mass_of_dry_air * si.kilogram)
# Assert
condition = (r > 1 * si.micrometre)
NTOT = n_tot(n, condition)
N1 = NTOT[:int(1 / 3 * len(NTOT))]
N2 = NTOT[int(1 / 3 * len(NTOT)):int(2 / 3 * len(NTOT))]
N3 = NTOT[int(2 / 3 * len(NTOT)):]
n_unit = 1 / si.microgram
assert min(N1) == 0.0 * n_unit
assert .6 * n_unit < max(N1) < .8 * n_unit
assert .17 * n_unit < min(N2) < .18 * n_unit
assert .35 * n_unit < max(N2) < .41 * n_unit
assert .1 * n_unit < min(N3) < .11 * n_unit
assert .27 * n_unit < max(N3) < .4 * n_unit
# TODO #527
if backend_class is not GPU:
assert max(output['ripening rate']) > 0
def __init__(self, settings, products=None):
env = Parcel(
dt=settings.dt,
mass_of_dry_air=settings.mass_of_dry_air,
p0=settings.p0,
q0=settings.q0,
T0=settings.T0,
w=settings.w,
)
n_sd = settings.n_sd_per_mode * len(settings.aerosol.modes)
builder = Builder(n_sd=n_sd, backend=CPU(formulae=settings.formulae))
builder.set_environment(env)
attributes = {
"dry volume": np.empty(0),
"dry volume organic": np.empty(0),
"kappa times dry volume": np.empty(0),
"n": np.ndarray(0),
}
for mode in settings.aerosol.modes:
r_dry, n_in_dv = settings.spectral_sampling(
spectrum=mode["spectrum"]).sample(settings.n_sd_per_mode)
V = settings.mass_of_dry_air / settings.rho0
N = n_in_dv * V
v_dry = settings.formulae.trivia.volume(radius=r_dry)
attributes["n"] = np.append(attributes["n"], N)
attributes["dry volume"] = np.append(attributes["dry volume"],
v_dry)
attributes["dry volume organic"] = np.append(
attributes["dry volume organic"], mode["f_org"] * v_dry)
attributes["kappa times dry volume"] = np.append(
attributes["kappa times dry volume"],
v_dry * mode["kappa"][settings.model],
)
for attribute in attributes.values():
assert attribute.shape[0] == n_sd
np.testing.assert_approx_equal(
np.sum(attributes["n"]) / V,
Sum(
tuple(
settings.aerosol.modes[i]["spectrum"]
for i in range(len(settings.aerosol.modes)))).norm_factor,
significant=5,
)
r_wet = equilibrate_wet_radii(
r_dry=settings.formulae.trivia.radius(
volume=attributes["dry volume"]),
environment=env,
kappa_times_dry_volume=attributes["kappa times dry volume"],
f_org=attributes["dry volume organic"] / attributes["dry volume"],
)
attributes["volume"] = settings.formulae.trivia.volume(radius=r_wet)
if settings.model == "Constant":
del attributes["dry volume organic"]
builder.add_dynamic(AmbientThermodynamics())
builder.add_dynamic(Condensation())
products = products or (
PySDM_products.ParcelDisplacement(name="z"),
PySDM_products.Time(name="t"),
PySDM_products.PeakSupersaturation(unit="%", name="S_max"),
PySDM_products.AmbientRelativeHumidity(unit="%", name="RH"),
PySDM_products.ParticleConcentration(
name="n_c_cm3",
unit="cm^-3",
radius_range=settings.cloud_radius_range),
PySDM_products.ParticleSizeSpectrumPerVolume(
radius_bins_edges=settings.wet_radius_bins_edges),
PySDM_products.ActivableFraction(),
)
particulator = builder.build(attributes=attributes, products=products)
if settings.BDF:
bdf.patch_particulator(particulator)
self.settings = settings
super().__init__(particulator=particulator)
def run_parcel(
w,
sol2,
N2,
rad2,
n_sd_per_mode,
RH0=1.0,
T0=294 * si.K,
p0=1e5 * si.Pa,
n_steps=50,
mass_of_dry_air=1e3 * si.kg,
dt=2 * si.s,
):
products = (
PySDM_products.WaterMixingRatio(unit="g/kg", name="ql"),
PySDM_products.PeakSupersaturation(name="S max"),
PySDM_products.AmbientRelativeHumidity(name="RH"),
PySDM_products.ParcelDisplacement(name="z"),
)
formulae = Formulae()
const = formulae.constants
pv0 = RH0 * formulae.saturation_vapour_pressure.pvs_Celsius(T0 - const.T0)
q0 = const.eps * pv0 / (p0 - pv0)
env = Parcel(dt=dt,
mass_of_dry_air=mass_of_dry_air,
p0=p0,
q0=q0,
w=w,
T0=T0)
aerosol = AerosolARG(M2_sol=sol2, M2_N=N2, M2_rad=rad2)
n_sd = n_sd_per_mode * len(aerosol.modes)
builder = Builder(backend=CPU(), n_sd=n_sd)
builder.set_environment(env)
builder.add_dynamic(AmbientThermodynamics())
builder.add_dynamic(Condensation())
builder.add_dynamic(Magick())
attributes = {
k: np.empty(0)
for k in ("dry volume", "kappa times dry volume", "n")
}
for i, mode in enumerate(aerosol.modes):
kappa, spectrum = mode["kappa"]["CompressedFilmOvadnevaite"], mode[
"spectrum"]
r_dry, concentration = ConstantMultiplicity(spectrum).sample(
n_sd_per_mode)
v_dry = builder.formulae.trivia.volume(radius=r_dry)
specific_concentration = concentration / builder.formulae.constants.rho_STP
attributes["n"] = np.append(
attributes["n"], specific_concentration * env.mass_of_dry_air)
attributes["dry volume"] = np.append(attributes["dry volume"], v_dry)
attributes["kappa times dry volume"] = np.append(
attributes["kappa times dry volume"], v_dry * kappa)
r_wet = equilibrate_wet_radii(
r_dry=builder.formulae.trivia.radius(volume=attributes["dry volume"]),
environment=env,
kappa_times_dry_volume=attributes["kappa times dry volume"],
)
attributes["volume"] = builder.formulae.trivia.volume(radius=r_wet)
particulator = builder.build(attributes, products=products)
bdf.patch_particulator(particulator)
output = {product.name: [] for product in particulator.products.values()}
output_attributes = {
"n": tuple([] for _ in range(particulator.n_sd)),
"volume": tuple([] for _ in range(particulator.n_sd)),
"critical volume": tuple([] for _ in range(particulator.n_sd)),
"critical supersaturation":
tuple([] for _ in range(particulator.n_sd)),
}
for _ in range(n_steps):
particulator.run(steps=1)
for product in particulator.products.values():
value = product.get()
output[product.name].append(value[0])
for key, attr in output_attributes.items():
attr_data = particulator.attributes[key].to_ndarray()
for drop_id in range(particulator.n_sd):
attr[drop_id].append(attr_data[drop_id])
error = np.zeros(len(aerosol.modes))
activated_fraction_S = np.zeros(len(aerosol.modes))
activated_fraction_V = np.zeros(len(aerosol.modes))
for j, mode in enumerate(aerosol.modes):
activated_drops_j_S = 0
activated_drops_j_V = 0
RHmax = np.nanmax(np.asarray(output["RH"]))
for i, volume in enumerate(output_attributes["volume"]):
if j * n_sd_per_mode <= i < (j + 1) * n_sd_per_mode:
if output_attributes["critical supersaturation"][i][-1] < RHmax:
activated_drops_j_S += output_attributes["n"][i][-1]
if output_attributes["critical volume"][i][-1] < volume[-1]:
activated_drops_j_V += output_attributes["n"][i][-1]
Nj = np.asarray(output_attributes["n"])[j * n_sd_per_mode:(j + 1) *
n_sd_per_mode, -1]
max_multiplicity_j = np.max(Nj)
sum_multiplicity_j = np.sum(Nj)
error[j] = max_multiplicity_j / sum_multiplicity_j
activated_fraction_S[j] = activated_drops_j_S / sum_multiplicity_j
activated_fraction_V[j] = activated_drops_j_V / sum_multiplicity_j
Output = namedtuple(
"Output",
[
"profile",
"attributes",
"aerosol",
"activated_fraction_S",
"activated_fraction_V",
"error",
],
)
return Output(
profile=output,
attributes=output_attributes,
aerosol=aerosol,
activated_fraction_S=activated_fraction_S,
activated_fraction_V=activated_fraction_V,
error=error,
)