def test_density(self):
"""Test density"""
outfile = 'test_xarray.nc'
analysis_file = 'test_xarray_analysis.nc'
o = OceanDrift(loglevel=20)
o.set_config('environment:fallback:land_binary_mask', 0)
t1 = datetime.now()
t2 = t1 + timedelta(hours=6)
o.seed_elements(time=t1, lon=4, lat=60, number=100, origin_marker=0)
o.seed_elements(time=[t1, t2],
lon=4.2,
lat=60.2,
number=100,
origin_marker=1)
o.seed_elements(time=[t1, t2],
lon=4.1,
lat=60.1,
number=100,
origin_marker=2)
reader_x = reader_oscillating.Reader('x_sea_water_velocity',
amplitude=1,
zero_time=t1)
reader_y = reader_oscillating.Reader('y_sea_water_velocity',
amplitude=1,
zero_time=t2)
o.add_reader([reader_x, reader_y])
o.set_config('drift:horizontal_diffusivity', 10)
o.run(duration=timedelta(hours=12), time_step=1800, outfile=outfile)
#o.plot(fast=True)
density_pixelsize_m = 5000
H, Hsub, Hsurf, lon_array, lat_array = o.get_density_array(
pixelsize_m=density_pixelsize_m)
ox = opendrift.open_xarray(outfile, analysis_file=analysis_file)
Hx, lon_arrayx, lat_arrayx = ox.get_density_xarray(
pixelsize_m=density_pixelsize_m)
Hx = Hx[0] # Presently only for origin_marker = 0
self.assertAlmostEqual(lon_array[0], 3.94, 1)
self.assertAlmostEqual(lon_array[-1], 4.76, 1)
self.assertAlmostEqual(lon_arrayx[0], 3.90, 1)
self.assertAlmostEqual(lon_arrayx[-1], 4.67, 1)
self.assertEqual(Hx.shape, H.shape)
Hsum = H.sum(axis=1).sum(axis=1)
Hxsum = Hx.sum(axis=1).sum(axis=1)
self.assertEqual(Hsum[0], 118)
self.assertEqual(Hxsum[0], 118)
self.assertEqual(Hsum[-1], 300)
self.assertEqual(Hxsum[-1], 300)
os.remove(outfile)
os.remove(analysis_file)
reader_y = reader_oscillating.Reader('y_sea_water_velocity',
amplitude=1,
zero_time=t2)
o.add_reader([reader_x, reader_y])
o.set_config('drift:horizontal_diffusivity', 10)
o.run(duration=timedelta(hours=24),
time_step=900,
time_step_output=1800,
outfile=outfile)
#%%
# Opening the output file lazily with Xarray.
# This will work even if the file is too large to fit in memory, as it
# will read and process data chuck-by-chunk directly from file using Dask.
# (See also `example_river_runoff.py <https://opendrift.github.io/gallery/example_river_runoff.html>`_)
oa = opendrift.open_xarray(outfile)
#%%
# Calculating histogram
# The histogram may be stored/cached to a netCDF file for later re-use,
# as the calculation may be time consuming for huge output files.
h = oa.get_histogram(pixelsize_m=500)
#%%
# Plot the cumulative coverage of first seeding (origin_marker=0)
b = h.isel(origin_marker=0).sum(dim='time')
oa.plot(background=b.where(b > 0),
fast=True,
show_elements=False,
vmin=0,
vmax=1000,
reader_y = reader_oscillating.Reader('y_sea_water_velocity',
amplitude=1,
zero_time=t2)
o.add_reader([reader_x, reader_y])
o.set_config('drift:horizontal_diffusivity', 10)
o.run(duration=timedelta(hours=24),
time_step=900,
time_step_output=1800,
outfile=outfile)
#%%
# Opening the output file lazily with Xarray.
# This will work even if the file is too large to fit in memory, as it
# will read and process data chuck-by-chunk directly from file using Dask.
# Note that the analysis file will be re-used if existing. Thus this file should be deleted after making any changes to the simulation above.
o = opendrift.open_xarray(outfile, analysis_file='simulation_density.nc')
#%%
# Making two animations, for each of the two seedings / origin_markere.
# The calculated density fields will be stored/cached in the analysis file
# for later re-use, as their calculation may be time consuming
# for huge output files.
# Note that other analysis/plotting methods are not yet adapted
# to datasets opened lazily with open_xarray
for om in [0, 1]:
o.animation(density=True,
density_pixelsize_m=500,
fast=False,
corners=[4.0, 6, 59.5, 61],
origin_marker=om,
show_elements=False,
amplitude=.5,
zero_time=t2)
o.add_reader([reader_x, reader_y])
o.set_config('drift:horizontal_diffusivity', 300)
o.set_config('general:coastline_action', 'previous')
o.run(duration=timedelta(hours=48),
time_step=1800,
time_step_output=3600,
outfile=outfile)
#%%
# Opening the output file lazily with Xarray.
# This will work even if the file is too large to fit in memory, as it
# will read and process data chuck-by-chunk directly from file using Dask.
# Note that the analysis file will be re-used if existing. Thus this file should be deleted after making any changes to the simulation above.
o = opendrift.open_xarray(outfile, analysis_file=analysis_file)
#%%
# Animation of the spatial density of river runoff water.
# Although there are the same number of elements from each river, the density plots are
# weighted with the actual runoff at time of seeding. This weighting can be done/changed
# afterwards without needing to redo the simulation.
# The calculated density fields will be stored/cached in the analysis file
# for later re-use, as their calculation may be time consuming
# for huge output files.
# Note that other analysis/plotting methods are not yet adapted
# to datasets opened lazily with open_xarray
runoff = np.abs(np.cos(np.arange(number * 2) * 2 * np.pi /
(number))) # Impose a temporal variation of runoff
runoff[0:number] *= .1 # Let River 1 have 10% of the runoff of River 2
#%%
# Seed elements at 5 different locations/longitudes
lons = [4, 4.2, 4.3, 4.32, 4.6]
t = datetime.now()
o = OceanDrift(loglevel=20)
for i, lon in enumerate(lons):
o.seed_elements(lon=lon, lat=60, radius=3000, number=2000, time=t, origin_marker_name='Lon %f' % lon)
o.set_config('environment:constant:y_sea_water_velocity', .1)
o.run(steps=15, outfile=of)
#%%
# Calculate spatial density of elements at 1500m grid spacing
oa = opendrift.open_xarray(of)
oa.ds = oa.ds.where(oa.ds.status==0)
d = oa.get_histogram(pixelsize_m=1500, weights=None)
dom = d.argmax(dim='origin_marker', skipna=True)
dom = dom.where(d.sum(dim='origin_marker')>0)
dom.name = 'Dominating source'
#%%
# Show which of the 5 sources are dominating within each grid cell
oa.animation(background=dom, show_elements=False, bgalpha=1,
legend=oa.origin_marker, colorbar=False, vmin=0, vmax=4)
#%%
# .. image:: /gallery/animations/example_dominating_0.gif