def test_returns_correct_data(self):
"""
See if it gets the right type of data each time
:return: Nothing
"""
print("Testing that get_data returns the expected amount of data")
wmo_boxes = np.array([3505, 1, 1, 1])
data_ctd = get_data(wmo_boxes, 1, self.config, self.float_name)
self.assertTrue(data_ctd['long'].shape[1] == 10,
"should return some data when fetching ctd")
data_bot = get_data(wmo_boxes, 2, self.config, self.float_name)
self.assertTrue(data_bot['long'].shape[1] == 33,
"should return some data when fetching argo")
data_argo = get_data(wmo_boxes, 3, self.config, self.float_name)
self.assertTrue(data_argo['long'][0].__len__() == 787,
"should return some data when fetching argo")
def test_returns_no_data(self):
"""
If given some data to find, it returns the correct data
:return: Nothing
"""
print("Testing that get_data will return no data if box is all 0's")
wmo_boxes = np.array([3505, 0, 0, 0])
data = get_data(wmo_boxes, 1, self.config, self.float_name)
self.assertTrue(data.__len__() == 0, "should return no data")
def test_removes_argo_float(self):
"""
See if it removes the argo float currently being analysed
:return: Nothing
"""
print("Testing that get_data will remove the argo being analysed")
wmo_boxes = np.array([3505, 0, 0, 1])
float_removed = "1900479"
data_normal = get_data(wmo_boxes, 3, self.config, self.float_name)
data_removed = get_data(wmo_boxes, 3, self.config, float_removed)
# 7 pieces of historical data are from float 1900479 in the selected box, whereas
# 0 pieces of historical data are from float 3901960 in the selected box, so there
# should be 7 less data in data_removed
self.assertTrue(
data_normal['long'][0].__len__() -
data_removed['long'][0].__len__() == 7,
"Should have removed data associated with the float being processed"
)
def get_region_hist_locations(pa_wmo_numbers, pa_float_name, config):
"""
Uses the WMO boxes and to return all of the historical data in the given area,
excluding the float that is currently being analysed.
:param pa_wmo_numbers: 2D array containing the name of the WMO boxes that cover the area
of interest, and flags for whether we want to use argo, bottle, and/or CTD data
:param pa_float_name: string of the name of the float currently being processed
:param config: Dictionary containing configuration settings. Used to find locations of folders
and file containing data
:return: the latitude, longitude and age of each data point we want
"""
# set up matrices to hold data
grid_lat = []
grid_long = []
grid_dates = []
# go through each of the WMO boxes
for wmo_box in pa_wmo_numbers:
# go through each of the columns denoting whether we should use CTD, bottle, and/or argo
for data_type in range(1, 4):
# get the data
try:
data = get_data(wmo_box, data_type, config, pa_float_name)
# if we have data, combine it with the other data then reset it
if data:
grid_lat = np.concatenate([grid_lat, data['lat'][0]])
grid_long = np.concatenate([grid_long, data['long'][0]])
grid_dates = np.concatenate([grid_dates, data['dates'][0]])
data = []
except:
pass
if grid_lat.__len__() == 0:
raise ValueError(
"get_region_hist_locations found no data for your specification. "
"Are your wmo_boxes files set up correctly?")
grid_long = wrap_longitude(grid_long)
# decimalise dates
grid_dates = change_dates(grid_dates)
return grid_lat, grid_long, grid_dates
def test_returns_correct_shape(self):
"""
See if the returned data has the expected shape
:return: Nothing
"""
print("Testing that get_data returns data with the expected shape")
wmo_boxes = np.array([3505, 1, 0, 0])
data = get_data(wmo_boxes, 1, self.config, self.float_name)
self.assertTrue(
data['sal'].shape == data['ptmp'].shape == data['temp'].shape,
"Ocean characteristic data should be the same shape")
self.assertTrue(
data['long'].shape == data['lat'].shape == data['dates'].shape,
"Spatial/temporal data should be the same shape")
self.assertTrue(data['sal'].shape[1] == data['long'].shape[1],
"Should be a profile for every location")
def get_region_data(pa_wmo_numbers, pa_float_name, config, index,
pa_float_pres):
"""
Get the historical pressure, salinity, and temperature of selected casts
:param pa_wmo_numbers: 2D array containing the name of the WMO boxes that cover the area
of interest, and flags for whether we want to use argo, bottle, and/or CTD data
:param pa_float_name: string of the name of the float currently being processed
:param config: Dictionary containing configuration settings. Used to find locations of folders
and file containing data
:param index: array of indices of selected historical casts
:param pa_float_pres: array of pressures for the float being processed
:return: The salinity, potential temperature, pressure, latitude, longitude, and age of each
historical cast selected to use
"""
# maximum depth to retrieve data from (deepest float measurement + MAP_P_DELTA
max_pres = np.nanmax(pa_float_pres) + config["MAP_P_DELTA"]
# set up empty arrays to hold the data to return
grid_sal = []
grid_ptmp = []
grid_pres = []
grid_lat = []
grid_long = []
grid_dates = []
data = []
# set up current maximum depth and number of columns
max_depth = 0
how_many_cols = 0
# set up variable to save beginning index for each set of data
starting_index = 0
# go through each of the WMO boxes
for wmo_box in pa_wmo_numbers:
# go through each of the columns denoting whether we should use CTD, bottle, and/or argo
for data_type in range(1, 4):
# get the data
try:
data = get_data(wmo_box, data_type, config, pa_float_name)
if data:
# Sometimes the data comes in wrapped in as a 3d array, so convert to 2d
if data['pres'].__len__() == 1:
data['pres'] = data['pres'][0]
data['sal'] = data['sal'][0]
data['ptmp'] = data['ptmp'][0]
data['lat'] = data['lat'][0].reshape(-1, 1)
data['long'] = data['long'][0].reshape(-1, 1)
data['dates'] = data['dates'][0].reshape(-1, 1)
# check the index of each station to see if it should be loaded
data_length = data['lat'][0].__len__()
data_indices = np.arange(0, data_length) + starting_index
# remember location of last entry
starting_index = starting_index + data_length
# load each station
for i in range(0, data_length):
good_indices = np.argwhere(index == data_indices[i])
if good_indices.__len__() > 0:
# only use non-NaN values
not_nan = np.argwhere(
np.isnan(data['pres'][:, i]) == 0)
# get the non-NaN values
pres = data['pres'][not_nan, i]
sal = data['sal'][not_nan, i]
ptmp = data['ptmp'][not_nan, i]
# remove values where pressure exceeds the maximum we want
too_deep = np.argwhere(pres > max_pres)
pres = np.delete(pres, too_deep[:, 0])
sal = np.delete(sal, too_deep[:, 0])
ptmp = np.delete(ptmp, too_deep[:, 0])
new_depth = pres.__len__()
how_many_rows = np.max([new_depth, max_depth])
# if the new data we are adding is longer than our columns, we need to
# fill in NaNs in the other columns
if new_depth > max_depth != 0:
grid_pres = np.append(
grid_pres,
np.ones(
(how_many_cols, new_depth - max_depth))
* np.nan,
axis=1).reshape(
(how_many_cols, how_many_rows))
grid_ptmp = np.append(
grid_ptmp,
np.ones(
(how_many_cols, new_depth - max_depth))
* np.nan,
axis=1).reshape(
(how_many_cols, how_many_rows))
grid_sal = np.append(
grid_sal,
np.ones(
(how_many_cols, new_depth - max_depth))
* np.nan,
axis=1).reshape(
(how_many_cols, how_many_rows))
# if the new data we are adding is shorter than our columns,
# then we need to fill in the rest with NaNs so it's the same length
elif new_depth < max_depth:
pres = np.append(
pres,
np.ones(
(max_depth - new_depth, 1)) * np.nan)
ptmp = np.append(
ptmp,
np.ones(
(max_depth - new_depth, 1)) * np.nan)
sal = np.append(
sal,
np.ones(
(max_depth - new_depth, 1)) * np.nan)
# if we don't have any data saved yet, create the grid matrix with the
# first data set
if grid_pres.__len__() == 0:
grid_pres = pres.reshape((1, pres.__len__()))
grid_ptmp = ptmp.reshape((1, pres.__len__()))
grid_sal = sal.reshape((1, pres.__len__()))
# if we already have data saved, add the new data to the saved data
else:
grid_pres = np.append(grid_pres, pres).reshape(
how_many_cols + 1, how_many_rows)
grid_ptmp = np.append(grid_ptmp, ptmp).reshape(
how_many_cols + 1, how_many_rows)
grid_sal = np.append(grid_sal, sal).reshape(
how_many_cols + 1, how_many_rows)
# save the latitude, longitude, and date of the new data
grid_lat = np.append(grid_lat, data['lat'][0, i])
grid_long = np.append(grid_long, data['long'][0,
i])
grid_dates = np.append(grid_dates,
data['dates'][0, i])
# readjust our values so we know what column to add the new data to,
# and what shape we should expect the data to be
max_depth = grid_pres.shape[1]
how_many_cols = grid_pres.shape[0]
except:
pass
# convert longitude to 0 to 360 degrees
try:
grid_long = wrap_longitude(grid_long)
# make sure salinity, pressure, and potential temperature data have all the same NaNs
sal_nans = np.argwhere(np.isnan(grid_sal))
for nan in sal_nans:
grid_pres[nan[0], nan[1]] = np.nan
grid_ptmp[nan[0], nan[1]] = np.nan
pres_nans = np.argwhere(np.isnan(grid_pres))
for nan in pres_nans:
grid_sal[nan[0], nan[1]] = np.nan
grid_ptmp[nan[0], nan[1]] = np.nan
ptmp_nans = np.argwhere(np.isnan(grid_ptmp))
for nan in ptmp_nans:
grid_sal[nan[0], nan[1]] = np.nan
grid_pres[nan[0], nan[1]] = np.nan
grid_dates = change_dates(grid_dates)
# transpose data
grid_sal = grid_sal.T
grid_pres = grid_pres.T
grid_ptmp = grid_ptmp.T
except:
raise Exception("NO DATA FOUND")
# we have encountered a problem where some data coming in is all NaN
# these columns need to be removed from the data set
nans = 0
for column in range(grid_sal.shape[1]):
if np.all(np.isnan(grid_sal[:, column - nans])) or \
np.all(np.isnan(grid_pres[:, column - nans])):
grid_sal = np.delete(grid_sal, column - nans, 1)
grid_ptmp = np.delete(grid_ptmp, column - nans, 1)
grid_pres = np.delete(grid_pres, column - nans, 1)
grid_lat = np.delete(grid_lat, column - nans)
grid_long = np.delete(grid_long, column - nans)
grid_dates = np.delete(grid_dates, column - nans)
nans += 1
if nans > 0:
print(
"Warning: found ", nans,
" all NaNs in your dataset. These water columns have been removed")
return grid_sal, grid_ptmp, grid_pres, grid_lat, grid_long, grid_dates