def test_below_threshold_without_fuzzy_factor(self):
"""Test if the fixed threshold is above the value in the data."""
plugin = Threshold(0.6)
result = plugin.process(self.cube)
expected_result_array = np.zeros_like(self.cube.data).reshape(
1, 1, 5, 5)
self.assertArrayAlmostEqual(result.data, expected_result_array)
def iterate_over_threshold(self, cubelist, threshold):
"""
Iterate over the application of thresholding to multiple cubes.
Args:
cubelist (iris.cube.CubeList):
Cubelist containing cubes to be thresholded.
threshold (float):
The threshold that will be applied.
Returns:
iris.cube.CubeList:
Cubelist after thresholding each cube.
"""
cubes = iris.cube.CubeList([])
for cube in cubelist:
threshold_cube = BasicThreshold(
threshold,
fuzzy_factor=self.fuzzy_factor,
comparison_operator=self.comparison_operator,
)(cube.copy())
# Will only ever contain one slice on threshold
for cube_slice in threshold_cube.slices_over(
find_threshold_coordinate(threshold_cube)):
threshold_cube = cube_slice
cubes.append(threshold_cube)
return cubes
def test_threshold_fuzzy_miss_high_threshold(self):
"""Test when a point is not within the fuzzy high threshold area."""
plugin = Threshold(3.0, fuzzy_factor=self.fuzzy_factor)
result = plugin.process(self.cube)
expected_result_array = np.zeros_like(self.cube.data).reshape(
1, 1, 5, 5)
self.assertArrayAlmostEqual(result.data, expected_result_array)
class Test__add_threshold_coord(IrisTest):
"""Test the _add_threshold_coord method"""
def setUp(self):
"""Set up a cube and plugin for testing."""
self.cube = set_up_variable_cube(np.ones((3, 3), dtype=np.float32))
self.plugin = Threshold([1])
self.plugin.threshold_coord_name = self.cube.name()
def test_basic(self):
"""Test a scalar threshold coordinate is created"""
result = self.plugin._add_threshold_coord(self.cube, 1)
self.assertEqual(result.ndim, 3)
self.assertIn("air_temperature", [coord.standard_name for coord in
result.coords(dim_coords=True)])
threshold_coord = result.coord("air_temperature")
self.assertEqual(threshold_coord.var_name, "threshold")
self.assertEqual(threshold_coord.attributes,
{"spp__relative_to_threshold": "above"})
self.assertAlmostEqual(threshold_coord.points[0], 1)
self.assertEqual(threshold_coord.units, self.cube.units)
def test_long_name(self):
"""Test coordinate is created with non-standard diagnostic name"""
self.cube.rename("sky_temperature")
self.plugin.threshold_coord_name = self.cube.name()
result = self.plugin._add_threshold_coord(self.cube, 1)
self.assertIn("sky_temperature", [coord.long_name for coord in
result.coords(dim_coords=True)])
def test_value_error(self):
"""Test method catches ValueErrors unrelated to name, by passing it a
list of values where a scalar is required"""
with self.assertRaises(ValueError):
self.plugin._add_threshold_coord(self.cube, [1, 1])
def test_basic(self):
"""Test that the plugin returns an iris.cube.Cube."""
fuzzy_factor = 0.95
threshold = 0.1
plugin = Threshold(threshold, fuzzy_factor=fuzzy_factor)
result = plugin.process(self.cube)
self.assertIsInstance(result, Cube)
def test_threshold_dimension_added(self):
"""Test that a threshold dimension coordinate is added."""
plugin = Threshold(0.1)
result = plugin.process(self.cube)
expected_coord = DimCoord([0.1], long_name='threshold',
units=self.cube.units)
self.assertEqual(result.coord('threshold'), expected_coord)
def test_threshold_below_fuzzy_miss(self):
"""Test not meeting the threshold in fuzzy below-threshold-mode."""
plugin = Threshold(
2.0, fuzzy_factor=self.fuzzy_factor, below_thresh_ok=True)
result = plugin.process(self.cube)
expected_result_array = np.ones_like(self.cube.data).reshape(
1, 1, 5, 5)
self.assertArrayAlmostEqual(result.data, expected_result_array)
def test_threshold_fuzzy(self):
"""Test when a point is in the fuzzy threshold area."""
plugin = Threshold(0.6, fuzzy_factor=self.fuzzy_factor)
result = plugin.process(self.cube)
expected_result_array = np.zeros_like(self.cube.data).reshape(
1, 1, 5, 5)
expected_result_array[0][0][2][2] = 1.0/3.0
self.assertArrayAlmostEqual(result.data, expected_result_array)
def test_threshold_asymmetric_bounds_below(self):
"""Test when a point is below asymmetric fuzzy threshold area."""
bounds = (0.51, 0.9)
plugin = Threshold(0.6, fuzzy_bounds=bounds)
result = plugin.process(self.cube)
expected_result_array = np.zeros_like(self.cube.data).reshape(
1, 1, 5, 5)
self.assertArrayAlmostEqual(result.data, expected_result_array)
def correct_where_input_true(self, selector_val):
"""
Replace points in the output_cube where output_land matches the
selector_val and the input_land does not match, but has matching
points in the vicinity, with the nearest matching point in the
vicinity in the original nearest_cube.
Updates self.output_cube.data
Args:
selector_val (int):
Value of mask to replace if needed.
Intended to be 1 for filling land points near the coast
and 0 for filling sea points near the coast.
"""
# Find all points on output grid matching selector_val
use_points = np.where(self.input_land.data == selector_val)
# If there are no matching points on the input grid, no alteration can
# be made. This tests the size of the y-coordinate of use_points.
if use_points[0].size is 0:
return
# Get shape of output grid
ynum, xnum = self.output_land.shape
# Using only these points, extrapolate to fill domain using nearest
# neighbour. This will generate a grid where the non-selector_val
# points are filled with the nearest value in the same mask
# classification.
(y_points, x_points) = np.mgrid[0:ynum, 0:xnum]
selector_data = griddata(
use_points,
self.nearest_cube.data[use_points],
(y_points, x_points),
method="nearest",
)
# Identify nearby points on regridded input_land that match the
# selector_value
if selector_val > 0.5:
thresholder = BasicThreshold(0.5)
else:
thresholder = BasicThreshold(0.5, comparison_operator="<=")
in_vicinity = self.vicinity(thresholder(self.input_land))
# Identify those points sourced from the opposite mask that are
# close to a source point of the correct mask
mismatch_points = np.logical_and(
np.logical_and(
self.output_land.data == selector_val,
self.input_land.data != selector_val,
),
in_vicinity.data > 0.5,
)
# Replace these points with the filled-domain data
self.output_cube.data[mismatch_points] = selector_data[mismatch_points]
def test_threshold_asymmetric_bounds_above(self):
"""Test when a point is above asymmetric fuzzy threshold area."""
bounds = (0.0, 0.45)
plugin = Threshold(0.4, fuzzy_bounds=bounds)
result = plugin.process(self.cube)
expected_result_array = np.zeros_like(self.cube.data).reshape(
1, 1, 5, 5)
expected_result_array[0][0][2][2] = 1.
self.assertArrayAlmostEqual(result.data, expected_result_array)
def test_threshold_boundingbelowzero(self):
"""Test fuzzy threshold of below-zero."""
bounds = (-1.0, 1.0)
plugin = Threshold(0.0, fuzzy_bounds=bounds, below_thresh_ok=True)
result = plugin.process(self.cube)
expected_result_array = np.full_like(
self.cube.data, fill_value=0.5).reshape(1, 1, 5, 5)
expected_result_array[0][0][2][2] = 0.25
self.assertArrayAlmostEqual(result.data, expected_result_array)
def test_threshold_boundingzero_above(self):
"""Test fuzzy threshold of zero where data are above upper-bound."""
bounds = (-0.1, 0.1)
plugin = Threshold(0.0, fuzzy_bounds=bounds)
result = plugin.process(self.cube)
expected_result_array = np.full_like(
self.cube.data, fill_value=0.5).reshape(1, 1, 5, 5)
expected_result_array[0][0][2][2] = 1.
self.assertArrayAlmostEqual(result.data, expected_result_array)
def test_threshold_negative(self):
"""Test a point when the threshold is negative."""
plugin = Threshold(-1.0,
fuzzy_factor=self.fuzzy_factor,
below_thresh_ok=True)
result = plugin.process(self.cube)
expected_result_array = np.ones_like(self.cube.data).reshape(
1, 1, 5, 5)
self.assertArrayAlmostEqual(result.data, expected_result_array)
def test_threshold_below_fuzzy(self):
"""Test a point in fuzzy threshold in below-threshold-mode."""
plugin = Threshold(
0.6, fuzzy_factor=self.fuzzy_factor, below_thresh_ok=True)
result = plugin.process(self.cube)
expected_result_array = np.ones_like(self.cube.data).reshape(
1, 1, 5, 5)
expected_result_array[0][0][2][2] = 2.0/3.0
self.assertArrayAlmostEqual(result.data, expected_result_array)
def test_above_threshold_without_fuzzy_factor(self):
"""Test if the fixed threshold is below the value in the data."""
# Copy the cube as the cube.data is used as the basis for comparison.
cube = self.cube.copy()
plugin = Threshold(0.1)
result = plugin.process(cube)
expected_result_array = self.cube.data.reshape(1, 1, 5, 5)
expected_result_array[0][0][2][2] = 1.0
self.assertArrayAlmostEqual(result.data, expected_result_array)
def test_threshold_point_nan(self):
"""Test behaviour for a single NaN grid cell."""
# Need to copy the cube as we're adjusting the data.
self.cube.data[0][2][2] = np.NAN
msg = "NaN detected in input cube data"
plugin = Threshold(
2.0, fuzzy_factor=self.fuzzy_factor, below_thresh_ok=True)
with self.assertRaisesRegex(ValueError, msg):
plugin.process(self.cube)
def test_threshold_asymmetric_bounds_upper_below(self):
"""Test when a point is in upper asymmetric fuzzy threshold area
and below-threshold is requested."""
bounds = (0.0, 0.6)
plugin = Threshold(0.4, fuzzy_bounds=bounds, below_thresh_ok=True)
result = plugin.process(self.cube)
expected_result_array = np.ones_like(self.cube.data).reshape(
1, 1, 5, 5)
expected_result_array[0][0][2][2] = 0.25
self.assertArrayAlmostEqual(result.data, expected_result_array)
def test_threshold_asymmetric_bounds_middle(self):
"""Test when a point is on the threshold with asymmetric fuzzy
bounds."""
bounds = (0.4, 0.9)
plugin = Threshold(0.5, fuzzy_bounds=bounds)
result = plugin.process(self.cube)
expected_result_array = np.zeros_like(self.cube.data).reshape(
1, 1, 5, 5)
expected_result_array[0][0][2][2] = 0.5
self.assertArrayAlmostEqual(result.data, expected_result_array)
def test_threshold(self):
"""Test the basic threshold functionality."""
# Copy the cube as the cube.data is used as the basis for comparison.
cube = self.cube.copy()
fuzzy_factor = 0.95
plugin = Threshold(0.1, fuzzy_factor=fuzzy_factor)
result = plugin.process(cube)
# The single 0.5-valued point => 1.0, so cheat by * 2.0 vs orig data.
expected_result_array = (self.cube.data * 2.0).reshape(1, 1, 5, 5)
self.assertArrayAlmostEqual(result.data, expected_result_array)
def test_threshold_unit_conversion(self):
"""Test data are correctly thresholded when the threshold is given in
units different from that of the input cube. In this test two
thresholds (of 4 and 6 mm/h) are used on a 5x5 cube where the
central data point value is 1.39e-6 m/s (~ 5 mm/h)."""
expected_result_array = np.zeros((2, 5, 5))
expected_result_array[0][2][2] = 1.
plugin = Threshold([4.0, 6.0], threshold_units='mm h-1')
result = plugin.process(self.rate_cube)
self.assertArrayAlmostEqual(result.data, expected_result_array)
def test_masked_array_fuzzybounds(self):
"""Test masked array are handled correctly when using fuzzy bounds.
Masked values are preserved following thresholding."""
bounds = (0.6 * self.fuzzy_factor, 0.6 * (2. - self.fuzzy_factor))
plugin = Threshold(0.6, fuzzy_bounds=bounds)
result = plugin.process(self.masked_cube)
expected_result_array = self.masked_cube.data.reshape((1, 1, 5, 5))
expected_result_array[0][0][2][2] = 1.0/3.0
self.assertArrayAlmostEqual(result.data.data, expected_result_array)
self.assertArrayEqual(
result.data.mask, self.masked_cube.data.mask.reshape((1, 1, 5, 5)))
def test_data_precision_preservation(self):
"""Test that the plugin returns an iris.cube.Cube of the same float
precision as the input cube."""
threshold = 0.1
plugin = Threshold(threshold, fuzzy_factor=self.fuzzy_factor)
f64cube = self.cube.copy(data=self.cube.data.astype(np.float64))
f32cube = self.cube.copy(data=self.cube.data.astype(np.float32))
f64result = plugin.process(f64cube)
f32result = plugin.process(f32cube)
self.assertEqual(f64cube.dtype, f64result.dtype)
self.assertEqual(f32cube.dtype, f32result.dtype)
def test_threshold_negative(self):
"""Test a point when the threshold is negative."""
self.cube.data[0][2][2] = -0.75
plugin = Threshold(-1.0,
fuzzy_factor=self.fuzzy_factor,
comparison_operator='<')
result = plugin.process(self.cube)
expected_result_array = np.zeros_like(self.cube.data).reshape(
1, 1, 5, 5)
expected_result_array[0][0][2][2] = 0.25
self.assertArrayAlmostEqual(result.data, expected_result_array)
def test_threshold_unit_conversion_fuzzy_factor(self):
"""Test for sensible fuzzy factor behaviour when units of threshold
are different from input cube. A fuzzy factor of 0.75 is equivalent
to bounds +/- 25% around the threshold in the given units. So for a
threshold of 4 (6) mm/h, the thresholded exceedance probabilities
increase linearly from 0 at 3 (4.5) mm/h to 1 at 5 (7.5) mm/h."""
expected_result_array = np.zeros((2, 5, 5))
expected_result_array[0][2][2] = 1.
expected_result_array[1][2][2] = 0.168
plugin = Threshold([4.0, 6.0], threshold_units='mm h-1',
fuzzy_factor=0.75)
result = plugin.process(self.rate_cube)
self.assertArrayAlmostEqual(result.data, expected_result_array)
def test_multiple_thresholds(self):
"""Test multiple thresholds applied to the cube return a single cube
with multiple arrays corresponding to each threshold."""
thresholds = [0.2, 0.4, 0.6]
plugin = Threshold(thresholds)
result = plugin.process(self.cube)
expected_array12 = np.zeros_like(self.cube.data).reshape(1, 1, 5, 5)
expected_array12[0][0][2][2] = 1.
expected_array3 = expected_array12 * 0.
expected_result_array = np.vstack(
[expected_array12, expected_array12, expected_array3])
self.assertIsInstance(result, Cube)
self.assertArrayAlmostEqual(result.data, expected_result_array)
def test_data_type_change_for_ints(self):
"""Test that the plugin returns an iris.cube.Cube of float32 type
if the input cube is of int type. This allows fuzzy bounds to be used
which return fractional values."""
fuzzy_factor = 5./6.
threshold = 12
self.cube.data = np.arange(25).reshape(1, 5, 5)
plugin = Threshold(threshold, fuzzy_factor=fuzzy_factor)
result = plugin.process(self.cube)
expected = np.round(np.arange(0, 1, 1./25.)).reshape(1, 1, 5, 5)
expected[0, 0, 2, 1:4] = [0.25, 0.5, 0.75]
self.assertEqual(result.dtype, 'float32')
self.assertArrayEqual(result.data, expected)
def test_multiple_thresholds(self):
"""Test multiple thresholds applied to the cube return a single cube
with multiple arrays corresponding to each threshold."""
thresholds = [0.2, 0.4, 0.6]
plugin = Threshold(thresholds)
result = plugin.process(self.cube)
all_zeroes = np.zeros_like(self.cube.data).reshape(1, 1, 5, 5)
one_exceed_point = all_zeroes.copy()
one_exceed_point[0][0][2][2] = 1.
expected_result_array = np.vstack(
[one_exceed_point, one_exceed_point, all_zeroes])
# transpose array to reflect realization coordinate re-ordering
expected_result_array = np.transpose(expected_result_array,
[1, 0, 2, 3])
self.assertIsInstance(result, Cube)
self.assertArrayAlmostEqual(result.data, expected_result_array)
def iterate_over_threshold(self, cubelist, threshold):
"""
Iterate over the application of thresholding to multiple cubes.
Args:
cubelist (Iris.cube.CubeList):
Cubelist containing cubes to be thresholded.
threshold (float):
The threshold that will be applied.
Returns:
cubes (Iris.cube.CubeList):
Cubelist after thresholding each cube.
"""
cubes = iris.cube.CubeList([])
for cube in cubelist:
threshold_cube = (
BasicThreshold(
threshold, fuzzy_factor=self.fuzzy_factor,
below_thresh_ok=self.below_thresh_ok
).process(cube.copy()))
# Will only ever contain one slice on threshold
for cube_slice in threshold_cube.slices_over('threshold'):
threshold_cube = cube_slice
cubes.append(threshold_cube)
return cubes
def test_threshold_le(self):
"""Test a point when we are in le threshold mode."""
plugin = Threshold(0.5, comparison_operator='<=')
name = "probability_of_{}_below_threshold"
expected_name = name.format(self.cube.name())
expected_attribute = "below"
result = plugin.process(self.cube)
expected_result_array = np.ones_like(self.cube.data).reshape(
1, 1, 5, 5)
expected_result_array[0][0][2][2] = 1
self.assertEqual(result.name(), expected_name)
self.assertEqual(
result.coord(
var_name="threshold").attributes['spp__relative_to_threshold'],
expected_attribute)
self.assertArrayAlmostEqual(result.data, expected_result_array)
