def gen_sst_maps(base_fp = EV['HOME'] + '/Desktop/lcrb_sst/'):
from simpleNIPA import NIPAphase
from numpy import ravel
from smapFuncts import sstMap
lcrb_prcp = get_lcrb_prcp()
sst, slp, mei, phaseind = get_climate_data()
models = {}
for bootconf in [1.0]:
for corrconf in [.99]:
fig, axes = plt.subplots(nrows = 3, ncols = 2, figsize = (12, 12))
for phase, ax in zip(phaseind, ravel(axes)):
print 'Starting %s, %.2f, %.2f' % (phase, corrconf, bootconf)
model = NIPAphase(lcrb_prcp, sst, mei, phaseind[phase])
model.bootcorr(corrconf = corrconf, bootconf = bootconf, quick = True)
model.gridCheck(lim = 6)
model.crossvalpcr()
if not model.flags['noSST']:
fig, ax, m = sstMap(model, fig = fig, ax = ax)
ax.set_title('%s, %.2f, %i' % (phase, model.correlation, model.n_post_grid))
else:
print '%s has no significant SST'
models[phase] = model
fp = base_fp + '%i_%i_shift' % (int(corrconf * 100), int(bootconf*100))
fig.savefig(fp); plt.close(fig)
return models
def gen_model(phase = 'lanina', cc = 0.95, bc = 0.80, quick = True):
from simpleNIPA import NIPAphase
from numpy import ravel
from smapFuncts import sstMap
lcrb_prcp = get_lcrb_prcp()
sst, slp, mei, phaseind = get_climate_data()
model = NIPAphase(lcrb_prcp, sst, mei, phaseind[phase])
model.bootcorr(corrconf = cc, bootconf = bc, quick = quick)
model.gridCheck(lim = 5, ntim = 3, debug = True)
model.crossvalpcr()
return model
def gen_models(quick = False):
from simpleNIPA import NIPAphase
from numpy import ravel
from smapFuncts import sstMap
models = {}
lcrb_prcp = get_lcrb_prcp()
sst, slp, mei, phaseind = get_climate_data()
for phase in phaseind:
model = NIPAphase(lcrb_prcp, sst, mei, phaseind[phase])
model.bootcorr(ntim = 100, corrconf = 0.99, bootconf = 0.80, quick = quick)
model.gridCheck()
model.crossvalpcr()
if not model.flags['noSST']: model.genEnsemble()
models[phase] = model
return models
def gen_models(n = 100, cc = 0.90, bc = 0.80, quick = False):
from simpleNIPA import NIPAphase
from numpy import ravel
from smapFuncts import sstMap
models = {}
lcrb_prcp = get_lcrb_prcp()
sst, slp, mei, phaseind = get_climate_data()
for phase in ['lanina', 'neutneg', 'neutpos', 'elnino', 'allyears']:
model = NIPAphase(lcrb_prcp, sst, mei, phaseind[phase])
model.bootcorr(ntim = n, corrconf = cc, bootconf = bc, quick = quick)
model.gridCheck(lim = 5, ntim = 3)
model.crossvalpcr()
if not model.flags['noSST']: model.genEnsemble()
models[phase] = model
return models
def make_monte_file(n = 10000, lcrb = False, corrconf = 0.90, bootconf = 0.80, quick = False):
from simpleNIPA import NIPAphase
from numpy import ravel, isnan, arange, nan
from random import sample
from smapFuncts import sstMap
from matplotlib import pyplot as plt
if quick:
filename = '%ixx' % (int(corrconf*100))
else:
filename = '%i%i' % (int(corrconf*100), int(bootconf*100))
if lcrb:
filename = filename + '_lcrb.csv'
n = 1000
else:
filename = filename + '_%i.csv' % (n)
#_Initialize dictionary to hold NIPA models for different phases
models = {}
#_Initialize files with input filename
all_file, sig_file = init_monte_files(filename)
#_Get the clim_data
clim_data = get_lcrb_prcp()
#_Get atmospheric data
sst, slp, mei, phaseind = get_climate_data()
#_Only want to analyze the four phases. Can do All years separately.
phases = ['lanina', 'neutneg', 'neutpos', 'elnino']
#_Begin monte carlo for n rounds.
for mc in range(1, n + 1):
print 'Starting round %i' % mc
#_Create random index
idx = arange(90)
if not lcrb:
idx = sample(xrange(90),90)
#_Initialize array for testing if all phases are significant
sig = []
#_Loop through phases and generate models
for phase in phases:
model = NIPAphase(clim_data[idx], sst, mei, phaseind[phase])
model.bootcorr(ntim = 100, corrconf = corrconf, bootconf = bootconf, quick = quick)
model.gridCheck(debug = True)
if not model.flags['noSST']:
model.crossvalpcr()
else:
model.correlation = nan
#_Store model
models[phase] = model
#_Write line
line = '%i,%s,%.2f,%i\n' % (mc, phase, model.correlation,model.n_post_grid)
all_file.write(line)
#_If the model correlation is not nan, append True to sig
if model.flags['noSST']:
sig.append(False)
else:
sig.append(True)
if all(sig):
print 'All significant, combining models'
sig_file.write('%i\n' % mc)
#_Close files
all_file.close; sig_file.close()
return
def make_monte_file(n = 10000, filename = '9980.csv', lcrb = False,
corrconf = 0.90, bootconf = 0.8, quick = True):
from simpleNIPA import NIPAphase
from numpy import ravel, isnan
from random import sample
from smapFuncts import sstMap
from matplotlib import pyplot as plt
if lcrb:
filename = filename[:4] + '_lcrb.csv'
n = 1
#_Initialize dictionary to hold NIPA models for different phases
models = {}
#_Initialize files with input filename
all_file, sig_file = init_monte_files(filename)
#_Get the clim_data
clim_data = get_lcrb_prcp()
#_Get atmospheric data
sst, slp, mei, phaseind = get_climate_data()
#_Only want to analyze the four phases. Can do All years separately.
phases = ['lanina', 'neutneg', 'neutpos', 'elnino']
#_Begin monte carlo for n rounds.
for mc in range(1, n + 1):
print 'Starting round %i' % mc
#_Create random index
idx = xrange(90)
if not lcrb:
idx = sample(xrange(90),90)
#_Initialize array for testing if all phases are significant
sig = []
#_Loop through phases and generate models
for phase in phases:
model = NIPAphase(clim_data[idx], sst, mei, phaseind[phase])
model.bootcorr(ntim = 100, corrconf = corrconf, bootconf = bootconf, quick = quick)
model.gridCheck(); model.crossvalpcr()
#_Store model
models[phase] = model
#_Write line
line = '%i,%s,%.2f,%i\n' % (mc, phase, model.correlation,model.n_post_grid)
all_file.write(line)
#_If the model correlation is not nan, append True to sig
sig.append(~isnan(model.correlation))
# if all(sig):
# print 'All significant, combining models'
# [models[phase].genEnsemble() for phase in phases]
# clim_data, hindcast, ensemble = combine_models(models)
#
# #_Create and write the long line for the significant data
# line = write_sig_data(mc, clim_data, hindcast, ensemble)
# sig_file.write(line)
#_Close files
all_file.close; sig_file.close()
return