def mean_regstat(timeres, variable):
uptake = fb_id.invf_uptake[timeres]
df = FeedbackAnalysis.INVF(
fb_id.co2['year'],
fb_id.temp['year'],
uptake,
variable
)
regstat = [df.regstats()[model] for model in uptake]
index = regstat[0].index
columns = regstat[0].columns
array = np.zeros([len(index), len(columns)])
for i, j in product(range(len(index)), range(len(columns))):
values = []
for model_regstat in regstat:
val = model_regstat.iloc[i,j]
if val != np.nan:
values.append(val)
array[i][j] = np.nanmean(values)
mean_regstat = pd.DataFrame(array)
mean_regstat.index = index
mean_regstat.columns = columns
return mean_regstat
def window_af(self, emission_rate=0.02):
"""
"""
land_df = FeedbackAnalysis.INVF(self.co2, self.temp, self.uptake,
'Earth_Land')
ocean_df = FeedbackAnalysis.INVF(self.co2, self.temp, self.uptake,
'Earth_Ocean')
land = land_df.params()
ocean = ocean_df.params()
beta = (land['beta'] + ocean['beta'])
u_gamma = (land['u_gamma'] + ocean['u_gamma'])
alpha = (beta + u_gamma)
alpha_mean = alpha.mean(axis=1)
alpha_std = alpha.std(axis=1)
b = 1 / np.log(1 + emission_rate)
u = 1 - b * alpha_mean
af = 1 / u
return {'af': af, 'alpha_mean': alpha_mean, 'alpha_std': alpha_std}
def all_regstat(timeres, variable):
uptake = fb_id.invf_uptake[timeres]
df = FeedbackAnalysis.INVF(
fb_id.co2['year'],
fb_id.temp['year'],
uptake,
variable
)
all_regstats = {model : pd.DataFrame(df.regstats()[model])
for model in uptake
}
return all_regstats
def fb_seasonal_regional_inv2(save=False):
timeres = ['winter', 'summer']
uptake = {time : fb_id.invf_uptake[time] for time in timeres}
all_subplots = ['42'+str(i) for i in range(1,9)]
vars = product(['Earth', 'South', 'Tropical', 'North'], timeres)
# Note u_gamma is used to compare with beta.
parameters = ['beta', 'u_gamma']
colors = ['green', 'red']
bar_width = 1.5
bar_position = np.arange(-1, 1) * bar_width
fig = plt.figure(figsize=(14,10))
ax = {}
axl = fig.add_subplot(111, frame_on=False)
axl.tick_params(labelcolor="none", bottom=False, left=False)
fb_reg_inv = {}
ymin_land, ymax_land = [], []
ymin_ocean, ymax_ocean = [], []
for subplot, var in zip(all_subplots, vars):
variable = var[0] + '_Land'
time = var[1]
df = FeedbackAnalysis.INVF(
fb_id.co2['year'],
fb_id.temp['year'],
uptake[time],
variable
)
params = df.params()
for parameter, color, bar_pos in zip(parameters, colors, bar_position):
param = params[parameter]
param_mean = param.mean(axis=1)
param_std = param.std(axis=1)
fb_reg_inv[variable, parameter] = pd.DataFrame({'Mean': param_mean,
'90_conf_interval': 1.645*param_std})
ax[subplot] = fig.add_subplot(subplot)
ax[subplot].bar(param.index + 5 + bar_pos,
param_mean,
yerr=1.645*param_std,
width=bar_width,
color=color)
if time == 'winter' and subplot != "421":
ymin_land.append((param_mean - 1.645*param_std).min())
ymax_land.append((param_mean + 1.645*param_std).max())
if time == 'summer' and subplot != "422":
ymin_ocean.append((param_mean - 1.645*param_std).min())
ymax_ocean.append((param_mean + 1.645*param_std).max())
delta = 0.05
for subplot in ["423", "425", "427"]:
ax[subplot].set_ylim([min(ymin_land) - delta * abs(min(ymin_land)),
max(ymax_land) + delta * abs(max(ymax_land))])
for subplot in ["424", "426", "428"]:
ax[subplot].set_ylim([min(ymin_ocean) - delta * abs(min(ymin_ocean)),
max(ymax_ocean) + delta * abs(max(ymax_ocean))])
ax["421"].set_xlabel("Winter", fontsize=14, labelpad=5)
ax["421"].xaxis.set_label_position('top')
ax["422"].set_xlabel("Summer", fontsize=14, labelpad=5)
ax["422"].xaxis.set_label_position('top')
for subplot, region in zip(["421", "423", "425", "427"],
['Earth', 'South', 'Tropical', 'North']):
ax[subplot].set_ylabel(region, fontsize=14, labelpad=5)
axl.set_xlabel("Middle year of 10-year window", fontsize=16, labelpad=10)
axl.set_ylabel(r'Feedback parameter (yr$^{-1}$)',
fontsize=16,
labelpad=40
)
if save:
plt.savefig(fb_id.FIGURE_DIRECTORY + f"fb_inv_regional_seasonal2.png")
return fb_reg_inv
def fb_seasonal_regional_inv(save=False):
timeres = ['winter', 'summer']
uptake = {time : fb_id.invf_uptake[time] for time in timeres}
all_subplots = ['22'+str(i) for i in range(1,5)]
# Note u_gamma is used to compare with beta.
parameters = [['beta', '_Land'], ['u_gamma', '_Land']]
vars = ['Earth', 'South', 'Tropical', 'North']
colors = ['gray', 'blue', 'orange', 'green']
bar_width = 1.25
bar_position = np.arange(-2, 2) * bar_width
fig = plt.figure(figsize=(14,10))
ax = {}
axl = fig.add_subplot(111, frame_on=False)
axl.tick_params(labelcolor="none", bottom=False, left=False)
fb_reg_inv = {}
for subplot, pro in zip(all_subplots, product(parameters, timeres)):
parameter, time = pro
for var, color, bar_pos in zip(vars, colors, bar_position):
variable = var + parameter[1]
df = FeedbackAnalysis.INVF(
fb_id.co2['year'],
fb_id.temp['year'],
uptake[time],
variable
)
param = df.params()[parameter[0]]
param_mean = param.mean(axis=1)
param_std = param.std(axis=1)
fb_reg_inv[(variable, parameter[0])] = pd.DataFrame({'Mean': param_mean,
'90_conf_interval': 1.645*param_std})
ax[subplot] = fig.add_subplot(subplot)
ax[subplot].bar(param.index + 5 + bar_pos,
param_mean,
yerr=1.645*param_std,
width=bar_width,
color=color)
ax["221"].set_xlabel("Winter", fontsize=14, labelpad=5)
ax["221"].xaxis.set_label_position('top')
ax["222"].set_xlabel("Summer", fontsize=14, labelpad=5)
ax["222"].xaxis.set_label_position('top')
for subplot, region in zip(["221", "223"],
[r'$\beta$', r'$u_{\gamma}$']):
ax[subplot].set_ylabel(region, fontsize=14, labelpad=5)
axl.set_xlabel("Middle year of 10-year window", fontsize=16, labelpad=10)
axl.set_ylabel(r'Feedback parameter (yr$^{-1}$)',
fontsize=16,
labelpad=40
)
if save:
plt.savefig(fb_id.FIGURE_DIRECTORY + f"fb_inv_regional_year.png")
return fb_reg_inv
def fb_seasonal_inv(save=False):
timeres = ['winter', 'summer']
uptake = {time : fb_id.invf_uptake[time] for time in timeres}
fig = plt.figure(figsize=(14,10))
ax = {}
axl = fig.add_subplot(111, frame_on=False)
axl.tick_params(labelcolor="none", bottom=False, left=False)
fb_inv_df = {}
for subplot, time in zip(["211", "212"], timeres):
df = FeedbackAnalysis.INVF(
fb_id.co2['year'],
fb_id.temp['year'],
uptake[time],
"Earth_Land"
)
whole_df = feedback_regression(time, "Earth_Land")[0].mean(axis=1)
beta = df.params()['beta']
gamma = df.params()['u_gamma'] # Note u_gamma is used to compare with beta.
whole_beta = whole_df['beta']
whole_gamma = whole_df['u_gamma']
beta_mean = beta.mean(axis=1)
beta_std = beta.std(axis=1)
gamma_mean = gamma.mean(axis=1)
gamma_std = gamma.std(axis=1)
fb_inv_df[time] = pd.DataFrame(
{
'whole_beta': whole_beta, 'whole_gamma': whole_gamma,
'beta_mean': beta_mean, 'beta_std': beta_std,
'gamma_mean': gamma_mean, 'gamma_std': gamma_std
}
)
ax[subplot] = fig.add_subplot(subplot)
ax[subplot].axhline(ls='--', color='k', alpha=0.5, lw=0.8)
ax[subplot].axhline(y= whole_beta, ls='--', color='green', alpha=0.8, lw=1)
ax[subplot].axhline(y= whole_gamma, ls='--', color='red', alpha=0.8, lw=1)
bar_width = 3
ax[subplot].bar(beta.index + 5 - bar_width / 2,
beta_mean,
yerr=beta_std * 1.645,
width=bar_width,
color='green',
label=r'$\beta$')
ax[subplot].bar(gamma.index + 5 + bar_width / 2,
gamma_mean,
yerr=gamma_std * 1.645,
width=bar_width,
color='red',
label=r'$u_{\gamma}$')
ax[subplot].legend()
ax[subplot].set_ylabel(time.title(), fontsize=14, labelpad=5)
axl.set_xlabel("First year of 10-year window", fontsize=16, labelpad=10)
axl.set_ylabel('Feedback parameter (yr $^{-1}$)',
fontsize=16,
labelpad=40
)
if save:
plt.savefig(fb_id.FIGURE_DIRECTORY + f"fb_inv_seasonal.png")
return fb_inv_df
temp = {
"year": xr.open_dataset(OUTPUT_DIR + f'TEMP/spatial/output_all/HadCRUT/year.nc'),
"month": xr.open_dataset(OUTPUT_DIR + f'TEMP/spatial/output_all/HadCRUT/month.nc')
}
invf_models = os.listdir(INV_DIRECTORY)
invf_uptake = {'year': {}, 'month': {}}
for timeres in invf_uptake:
for model in invf_models:
model_dir = INV_DIRECTORY + model + '/'
invf_uptake[timeres][model] = xr.open_dataset(model_dir + f'{timeres}.nc')
variables = ['Earth_Land', 'South_Land', 'Tropical_Land', 'North_Land']
fa = {}
for var in variables:
fa[var] = FeedbackAnalysis.INVF(co2['year'], temp['year'], invf_uptake['year'], var).params()['beta']
def check_sum(model, year):
earth = fa['Earth_Land'][model][year]
south = fa['South_Land'][model][year]
tropical = fa['Tropical_Land'][model][year]
north = fa['North_Land'][model][year]
return earth - south - tropical - north
for m, y in product(['CAMS', 'Rayner', 'JENA_s76'], [1980, 1990, 2000]):
print(check_sum(m,y))