def test_bin_spatial_counts(self):
""" this will test both good and bad points within the region.
1) 2 inside the domain
2) outside the bbox but not in domain
3) completely outside the domain
we will check that only 1 event is placed in the grid and ensure its location is correct.
"""
# create some arbitrary grid
nx = 8
ny = 10
dh = 0.1
x_points = numpy.arange(nx) * dh
y_points = numpy.arange(ny) * dh
origins = list(itertools.product(x_points, y_points))
# grid is missing first and last block
origins.pop(0)
origins.pop(-1)
cart_grid = CartesianGrid2D(
[Polygon(bbox) for bbox in compute_vertices(origins, dh)],
dh
)
catalog = MockCatalog()
catalog.region = cart_grid
test_result = catalog.spatial_counts()
# we have tested 2 inside the domain
self.assertEqual(numpy.sum(test_result), 2)
# we know that (0.05, 0.15) corresponds to index 0
self.assertEqual(test_result[0], 1)
# we know that (0.15, 0.05) corresponds too index 9
self.assertEqual(test_result[9], 1)
def test_polygon_mask_all_masked(self):
polygons = [Polygon(bbox) for bbox in compute_vertices(self.origins, self.dh)]
n_poly = len(polygons)
# this should mask every polygon for easy checking
test_grid = CartesianGrid2D(polygons, self.dh, mask=numpy.zeros(n_poly))
self.assertEqual(n_poly, test_grid.num_nodes)
numpy.testing.assert_array_equal(test_grid.bbox_mask, 1)
numpy.testing.assert_array_equal(test_grid.poly_mask, 0)
def setUp(self):
# create some arbitrary grid
self.nx = 8
self.ny = 10
self.dh = 0.1
x_points = numpy.arange(self.nx)*self.dh
y_points = numpy.arange(self.ny)*self.dh
self.origins = list(itertools.product(x_points, y_points))
# grid is missing first and last block
self.origins.pop(0)
self.origins.pop(-1)
self.num_nodes = len(self.origins)
# this is kinda ugly, maybe we want to create this in a different way, class method?
self.cart_grid = CartesianGrid2D([Polygon(bbox) for bbox in compute_vertices(self.origins, self.dh)], self.dh)
def load_ascii(cls, ascii_fname, start_date=None, end_date=None, name=None, swap_latlon=False):
""" Reads Forecast file from CSEP1 ascii format.
The ascii format from CSEP1 testing centers. The ASCII format does not contain headers. The format is listed here:
Lon_0, Lon_1, Lat_0, Lat_1, z_0, z_1, Mag_0, Mag_1, Rate, Flag
For the purposes of defining region objects and magnitude bins use the Lat_0 and Lon_0 values along with Mag_0.
We can assume that the magnitude bins are regularly spaced to allow us to compute Deltas.
The file is row-ordered so that magnitude bins are fastest then followed by space.
Args:
ascii_fname: file name of csep forecast in .dat format
swap_latlon (bool): if true, read forecast spatial cells as lat_0, lat_1, lon_0, lon_1
"""
# Load data
data = numpy.loadtxt(ascii_fname)
# this is very ugly, but since unique returns a sorted list, we want to get the index, sort that and then return
# from the original array. same for magnitudes below.
all_polys = data[:,:4]
all_poly_mask = data[:,-1]
sorted_idx = numpy.sort(numpy.unique(all_polys, return_index=True, axis=0)[1], kind='stable')
unique_poly = all_polys[sorted_idx]
# gives the flag for a spatial cell in the order it was presented in the file
poly_mask = all_poly_mask[sorted_idx]
# create magnitudes bins using Mag_0, ignoring Mag_1 bc they are regular until last bin. we dont want binary search for this
all_mws = data[:,-4]
sorted_idx = numpy.sort(numpy.unique(all_mws, return_index=True)[1], kind='stable')
mws = all_mws[sorted_idx]
# csep1 stores the lat lons as min values and not (x,y) tuples
bboxes = [tuple(itertools.product(bbox[:2], bbox[2:])) for bbox in unique_poly]
if swap_latlon:
bboxes = [tuple(itertools.product(bbox[2:], bbox[:2])) for bbox in unique_poly]
# the spatial cells are arranged fast in latitude, so this only works for the specific csep1 file format
dh = float(unique_poly[0,3] - unique_poly[0,2])
# create CarteisanGrid of points
region = CartesianGrid2D([Polygon(bbox) for bbox in bboxes], dh, mask=poly_mask)
# get dims of 2d np.array
n_mag_bins = len(mws)
n_poly = region.num_nodes
# reshape rates into correct 2d format
rates = data[:,-2].reshape(n_poly, n_mag_bins)
# create / return class
if name is None:
name = os.path.basename(ascii_fname[:-4])
gds = cls(start_date, end_date, magnitudes=mws, name=name, region=region, data=rates)
return gds
def test_bin_probability(self):
# we have tested 2 inside the domain
nx = 8
ny = 10
dh = 0.1
x_points = numpy.arange(nx) * dh
y_points = numpy.arange(ny) * dh
origins = list(itertools.product(x_points, y_points))
# grid is missing first and last block
origins.pop(0)
origins.pop(-1)
cart_grid = CartesianGrid2D(
[Polygon(bbox) for bbox in compute_vertices(origins, dh)], dh)
catalog = MockCatalog()
catalog.region = cart_grid
test_result = catalog.spatial_event_probability()
self.assertEqual(numpy.sum(test_result), 2)
# we know that (0.05, 0.15) corresponds to index 0 from the above test
self.assertEqual(test_result[0], 1)
self.assertEqual(test_result[9], 1)
def setUp(self):
# create some arbitrary grid
self.nx = 8
self.ny = 10
self.dh = 1
x_points = numpy.arange(1.5, self.nx) * self.dh
y_points = numpy.arange(1.5, self.ny) * self.dh
# spatial grid starts at (1.5, 1.5); so the event at (1, 1) should be removed.
self.origins = list(itertools.product(x_points, y_points))
self.num_nodes = len(self.origins)
self.cart_grid = CartesianGrid2D([Polygon(bbox) for bbox in compute_vertices(self.origins, self.dh)], self.dh)
# define dummy cat
date1 = strptime_to_utc_epoch('2009-01-01 00:00:00.0000')
date2 = strptime_to_utc_epoch('2010-01-01 00:00:00.0000')
date3 = strptime_to_utc_epoch('2011-01-01 00:00:00.0000')
catalog = [(b'1', date1, 1.0, 1.0, 1.0, 1.0),
(b'2', date2, 2.0, 2.0, 2.0, 2.0),
(b'3', date3, 3.0, 3.0, 3.0, 3.0)]
self.test_cat1 = CSEPCatalog(data=catalog)
def test_to_from_dict(self):
self.assertEqual(self.cart_grid,
CartesianGrid2D.from_dict(self.cart_grid.to_dict()))