def testCrpss(self):
crpss_dict = self.crpssDict()
crpss = Crpss(**crpss_dict)
crpss.prepareInput()
crpss.analyze()
#test file names
res_files = [f for r,_,files in os.walk(self.tmp_dir+'/output') for f in files]
test_files = [f for r,_,files in os.walk(self.test_path+'/output') for f in files]
self.assertEqual(sorted(res_files),sorted(test_files))
res_files = [f for r,_,files in os.walk(self.tmp_dir+'/plots') for f in files]
test_files = [f for r,_,files in os.walk(self.test_path+'/plots') for f in files]
self.assertEqual(sorted(res_files),sorted(test_files))
#test values of all .nc files
res_files = sorted([os.path.join(r,f) for r,_,files in os.walk(self.tmp_dir+'/output') for f in files])
test_files = sorted([os.path.join(r,f) for r,_,files in os.walk(self.test_path+'/output') for f in files])
for i,f in enumerate(res_files):
t1 = FileHandler.openNetCDFFile(res_files[i])
t2 = FileHandler.openNetCDFFile(test_files[i])
np.testing.assert_array_almost_equal(t1['variable'], t2['variable'], 3)
#elf.assertTrue((t1['variable']==t2['variable']).all())
self.assertTrue((t1['lon']==t2['lon']).all())
self.assertTrue((t1['lat']==t2['lat']).all())
def testMsss(self):
msssDict = self.msssDict()
msss = Msss(**msssDict)
msss.prepareInput()
msss.analyze()
#test file names
res_files = [f for r,_,files in os.walk(self.tmp_dir+'/output') for f in files]
test_files = [f for r,_,files in os.walk(self.test_path+'/output') for f in files]
self.assertEqual(sorted(res_files),sorted(test_files))
res_files = [f for r,_,files in os.walk(self.tmp_dir+'/plots') for f in files]
test_files = [f for r,_,files in os.walk(self.test_path+'/plots') for f in files]
self.assertEqual(sorted(res_files),sorted(test_files))
#test values of all .nc files
res_files = sorted([os.path.join(r,f) for r,_,files in os.walk(self.tmp_dir+'/output') for f in files])
test_files = sorted([os.path.join(r,f) for r,_,files in os.walk(self.test_path+'/output') for f in files])
#test_files = sorted([os.path.join(r,f) for r,_,files in os.walk(tets_p) for f in files])
# print len(res_files)
# print len(test_files)
for i,f in enumerate(res_files):
tmp_test = cdo.copy(input=test_files[i],options='-f nc')
t1 = FileHandler.openNetCDFFile(res_files[i])
t2 = FileHandler.openNetCDFFile(tmp_test)#test_files[i])
# print '_______________________________________-'
# print res_files[i]
# print test_files[i]
np.testing.assert_array_almost_equal(t1['variable'], t2['variable'], 0)
self.assertTrue((t1['lon']==t2['lon']).all())
self.assertTrue((t1['lat']==t2['lat']).all())
def testMsss(self):
msssDict = self.msssDict()
msss = Msss(**msssDict)
msss.prepareInput()
msss.analyze()
#test file names
res_files = [
f for r, _, files in os.walk(self.tmp_dir + '/output')
for f in files
]
test_files = [
f for r, _, files in os.walk(self.test_path + '/output')
for f in files
]
self.assertEqual(sorted(res_files), sorted(test_files))
res_files = [
f for r, _, files in os.walk(self.tmp_dir + '/plots')
for f in files
]
test_files = [
f for r, _, files in os.walk(self.test_path + '/plots')
for f in files
]
self.assertEqual(sorted(res_files), sorted(test_files))
#test values of all .nc files
res_files = sorted([
os.path.join(r, f)
for r, _, files in os.walk(self.tmp_dir + '/output') for f in files
])
test_files = sorted([
os.path.join(r, f)
for r, _, files in os.walk(self.test_path + '/output')
for f in files
])
#test_files = sorted([os.path.join(r,f) for r,_,files in os.walk(tets_p) for f in files])
# print len(res_files)
# print len(test_files)
for i, f in enumerate(res_files):
tmp_test = cdo.copy(input=test_files[i], options='-f nc')
t1 = FileHandler.openNetCDFFile(res_files[i])
t2 = FileHandler.openNetCDFFile(tmp_test) #test_files[i])
# print '_______________________________________-'
# print res_files[i]
# print test_files[i]
np.testing.assert_array_almost_equal(t1['variable'],
t2['variable'], 0)
self.assertTrue((t1['lon'] == t2['lon']).all())
self.assertTrue((t1['lat'] == t2['lat']).all())
def testTestfiles(self):
from create_testfiles import createMsssInput
std1_goal=1.
std2_goal=2.
corr_goal=0.5
std_ratio = std2_goal/std1_goal
cond_bias = corr_goal - std1_goal/std2_goal
msss_goal = corr_goal**2 + cond_bias**2
length = 50
(hindcasts,observations) = createMsssInput(1960, std1_goal, std2_goal, corr_goal, length)
msssDict = self.msssDict()
msssDict['decadals'] = hindcasts.keys()
msssDict['observation'] = 'hadcrut3v'
msssDict['leadtimes'] = '1'
msss = Msss(**msssDict)
msss.input1Remapped = hindcasts
msss.input2Remapped = hindcasts
msss.observationRemapped = observations
msss.analyze()
goal_dict = {'msss_goal': '1-1/baseline1_output_mpi-esm-lr_decs4e_input1/accuracy/1_1_tas_baseline1_output_mpi-esm-lr_decs4e_1960-2009_msss.nc',
'corr_goal': '1-1/baseline1_output_mpi-esm-lr_decs4e_input1/accuracy/1_1_tas_baseline1_output_mpi-esm-lr_decs4e_1960-2009_correlation.nc',
'cond_bias': '1-1/baseline1_output_mpi-esm-lr_decs4e_input1/accuracy/1_1_tas_baseline1_output_mpi-esm-lr_decs4e_1960-2009_conditional_bias.nc',
'std_ratio': '1-1/baseline1_output_mpi-esm-lr_decs4e_input1/accuracy/1_1_tas_baseline1_output_mpi-esm-lr_decs4e_1960-2009_std_ratio.nc'}
for key,val in goal_dict.iteritems():
t1 = FileHandler.openNetCDFFile(msss.outputDir+val,mode='var')
self.assertAlmostEqual(np.round(t1[12,12],2), locals()[key], 1)
def testTestfiles(self):
from create_testfiles import createMsssInput
std1_goal = 1.
std2_goal = 2.
corr_goal = 0.5
std_ratio = std2_goal / std1_goal
cond_bias = corr_goal - std1_goal / std2_goal
msss_goal = corr_goal**2 + cond_bias**2
length = 50
(hindcasts, observations) = createMsssInput(1960, std1_goal, std2_goal,
corr_goal, length)
msssDict = self.msssDict()
msssDict['decadals'] = hindcasts.keys()
msssDict['observation'] = 'hadcrut3v'
msssDict['leadtimes'] = '1'
msss = Msss(**msssDict)
msss.input1Remapped = hindcasts
msss.input2Remapped = hindcasts
msss.observationRemapped = observations
msss.analyze()
goal_dict = {
'msss_goal':
'1-1/baseline1_output_mpi-esm-lr_decs4e/msss/1_1_tas_baseline1_output_mpi-esm-lr_decs4e_1960-2009_msss.nc',
'corr_goal':
'1-1/baseline1_output_mpi-esm-lr_decs4e/msss/1_1_tas_baseline1_output_mpi-esm-lr_decs4e_1960-2009_correlation.nc',
'cond_bias':
'1-1/baseline1_output_mpi-esm-lr_decs4e/msss/1_1_tas_baseline1_output_mpi-esm-lr_decs4e_1960-2009_conditional_bias.nc',
'std_ratio':
'1-1/baseline1_output_mpi-esm-lr_decs4e/msss/1_1_tas_baseline1_output_mpi-esm-lr_decs4e_1960-2009_std_ratio.nc'
}
for key, val in goal_dict.iteritems():
t1 = FileHandler.openNetCDFFile(msss.outputDir + val, mode='var')
print key
print locals()[key]
print t1[12, 12]
self.assertAlmostEqual(np.round(t1[12, 12], 2), locals()[key], 2)
def testTestfiles(self):
def crps(var,H,O):
x = (H-O)/var
crps = -var * (1./np.sqrt(np.pi) - 2. * stats.norm.pdf(x) - x * (2. * stats.norm.cdf(x) - 1.))
return crps
def condBias(H,O):
H_ensmean = np.mean(H, axis=1)
r = np.corrcoef(H_ensmean, O)[0,1]
std_H = np.std(H_ensmean)
std_O = np.std(O)
cond_bias = r * std_O/std_H
return cond_bias
from create_testfiles import createCrpssInput, getCrpssTimeseries
mse_goal = 4
ensspread_goal = 3
ensemble_size = 10
length = 20
#create timeseries
(t,hind,obs) = getCrpssTimeseries(mse_goal, ensspread_goal, ensemble_size, length)
(hindcasts,observations) = createCrpssInput(1960, hind, obs, ensemble_size, length)
crpssDict = self.crpssDict()
crpssDict['decadals'] = hindcasts.keys()
crpssDict['observation'] = 'HadCrut'
crpssDict['leadtimes'] = '1'
crpss = Crpss(**crpssDict)
crpss.inputRemapped = hindcasts
crpss.observationRemapped = observations
crpss.analyze()
factor = length/(length-2.)
factor2 = ensemble_size/(ensemble_size-1.)
ensspread = np.mean((factor2*np.var(hind,axis=1)))**0.5
hind_mean = np.mean(hind,axis=1)
#calculte crossvalidated mean
hind_mean_anom = hind_mean.copy()
crossval = np.zeros((length))
for i in range(length):
tmp = np.delete(hind_mean,i)
crossval[i] = np.mean(tmp)
hind_mean_anom[i] = hind_mean[i] - crossval[i]
hind_mean = hind_mean_anom.copy()
#remove conditional bias
bias = condBias(hind, obs)
for i in range(length):
hind[i,:] = hind[i,:] - crossval[i]
hind[i,:] = hind[i,:] + (bias-1)*hind_mean[i]
hind_mean_corrected = np.mean(hind,axis=1)
RMSE = (factor*np.mean((np.mean(hind,axis=1)-obs)**2))**0.5
#calculate crpss_goal
crps1 = np.zeros((length))
crps2 = np.zeros((length))
for i in range(length):
crps1[i] = crps(ensspread, hind_mean_corrected[i], obs[i])
crps2[i] = crps(RMSE,hind_mean_corrected[i],obs[i])
crpss_goal = 1 - np.mean(crps1)/np.mean(crps2)
#test spreadscore
tf = crpss.outputDir+'1-1/baseline1_output_mpi-esm-lr_decs4e_input1/ensemble_spread/1_1_tas_baseline1_output_mpi-esm-lr_decs4e_1960-1979_ensspread_vs_referror.nc'
t1 = FileHandler.openNetCDFFile(tf,mode='var')
self.assertAlmostEqual(t1[12,12],ensspread/RMSE,2)
#test crpss ens vs ref
tf = crpss.outputDir+'1-1/baseline1_output_mpi-esm-lr_decs4e_input1/ensemble_spread/1_1_tas_baseline1_output_mpi-esm-lr_decs4e_1960-1979_ens-vs-ref_crpss.nc'
t1 = FileHandler.openNetCDFFile(tf,mode='var')
self.assertAlmostEqual(t1[12,12],crpss_goal,2)