def get_rate_estimate(ext_time_array,max_t,index,species_list,plot_posterior=0,pdf=0,n_gen = 100000,burnin = 1000):
sys.stdout.write('\rProcessing species: %i/%i '%(index+1,len(species_list)))
ext_time_array_new = ext_time_array.copy()
ext_time_array_new[ext_time_array_new!=ext_time_array_new] = max_t
ext_time_array_new = ext_time_array_new.astype(float)
w_times = np.sum(ext_time_array_new)
ext_events = len(ext_time_array_new[ext_time_array_new<max_t])
post_samples = []
q = 0.01
likA = np.log(q)*ext_events -q*w_times
for i in range(n_gen):
new_q, hast = update_multiplier(q)
lik = np.log(new_q)*ext_events -new_q*w_times
if lik-likA + hast >= np.log(np.random.random()):
q = new_q
likA = lik
if i > burnin and i % 10==0:
post_samples.append(q)
mean_value = np.mean(post_samples)
lower,upper = cust_func.calcHPD(post_samples,0.95)
if plot_posterior:
plt.figure()
plt.hist(post_samples,100)
plt.xlabel('Extinction rate estimates')
plt.ylabel('Counts')
plt.title(species_list[index])
plt.tight_layout()
pdf.savefig()
plt.close()
#fig.savefig(os.path.join(posterior_plot_dir,'%s.pdf'%species_list[index]),bbox_inches='tight', dpi = 500)
return [mean_value,lower,upper]
def plot_mean_and_interval(div, color, label, fig):
plt.plot(time_axis, np.mean(div, axis=0), color=color, label=label)
min_hpd, max_hpd = np.array(
[cust_func.calcHPD(i, 0.95) for i in div.T]).T
plt.fill_between(time_axis,
min_hpd,
max_hpd,
color=color,
alpha=0.2)
return fig
def main(args):
indir = args.indir
outdir = args.outdir
n_years = int(args.n_years)
n_sim = int(args.n_sim)
allow_status_change = int(args.status_change)
conservation_increase_factor = int(args.conservation_increase_factor)
threat_increase_factor = int(args.threat_increase_factor)
extinction_rates = int(args.extinction_rates)
n_gen = int(args.n_gen)
burnin = int(args.burnin)
plot_diversity_trajectory = int(args.plot_diversity_trajectory)
plot_histograms = int(args.plot_histograms)
plot_posterior = int(args.plot_posterior)
model_unknown_as_lc = int(args.model_unknown_as_lc)
plot_status_trajectories = int(args.plot_status_trajectories)
plot_status_piechart = int(args.plot_status_piechart)
if not os.path.exists(outdir):
os.makedirs(outdir)
species_list_status_file = os.path.join(
indir, 'iucn_data/current_status_all_species.txt')
transition_rates_file = os.path.join(indir,
'sampled_status_change_rates.txt')
en_ext_risk_file = os.path.join(indir,
'en_extinction_risks_all_species.txt')
cr_ext_risk_file = os.path.join(indir,
'cr_extinction_risks_all_species.txt')
species_list_status = pd.read_csv(species_list_status_file, sep='\t')
transition_rates = pd.read_csv(transition_rates_file,
sep='\t',
index_col='status_change')
species_list = pd.read_csv(en_ext_risk_file, sep='\t').iloc[:, 0].values
en_ext_data = pd.read_csv(en_ext_risk_file, sep='\t').iloc[:, 1:].values
cr_ext_data = pd.read_csv(cr_ext_risk_file, sep='\t').iloc[:, 1:].values
# target_species=0
# target_species = 'Sarcogyps calvus'
# target_species = 'Cathartes aura'
# if target_species:
# species_list_status,species_list,en_ext_data,cr_ext_data = select_target_species(target_species,species_list_status,species_list,en_ext_data,cr_ext_data)
# calculate all q-matrices (all species and sim replicates)________________
n_rates = transition_rates.shape[1]
print("\nCalculating species-specific q-matrices ...")
qmatrix_dict_list = []
for i in np.arange(n_rates):
rates = transition_rates.iloc[:, i]
sys.stdout.write('\rProgress: %i %%' % int(((i + 1) / n_rates) * 100))
en_risks_rep = en_ext_data.T[i]
cr_risks_rep = cr_ext_data.T[i]
q_matrix_dict = {}
for j, species in enumerate(species_list):
en_risk = en_risks_rep[j]
cr_risk = cr_risks_rep[j]
status_specific_p_e = np.array(
[
0.000000155728, 0.000041551152, 0.001053050310, en_risk,
cr_risk
]
) # These values are the category specific probabilities of extinction per year calculated from IUCN definition of each category
q_matrix = cust_func.qmatrix(rates, status_specific_p_e)
if conservation_increase_factor != 1:
indeces_lower_triangle = np.tril_indices(q_matrix.shape[0], -1)
q_matrix[indeces_lower_triangle] = q_matrix[
indeces_lower_triangle] * conservation_increase_factor
np.fill_diagonal(q_matrix, 0)
np.fill_diagonal(q_matrix, -np.sum(q_matrix, axis=1))
if threat_increase_factor != 1:
indeces_upper_triangle = np.triu_indices(q_matrix.shape[0], 1)
q_matrix[indeces_upper_triangle] = q_matrix[
indeces_upper_triangle] * threat_increase_factor
np.fill_diagonal(q_matrix, 0)
np.fill_diagonal(q_matrix, -np.sum(q_matrix, axis=1))
q_matrix_dict[species] = q_matrix
qmatrix_dict_list.append(q_matrix_dict)
# get transition rates for DD______________________________________________
dd_changes = []
dd_rates = []
for row_id, change_type in enumerate(transition_rates.index.values):
states = change_type.split('->')
if states[0] == 'DD':
dd_changes.append('-'.join(states))
rates = transition_rates[transition_rates.index ==
change_type].values
dd_rates.append(rates[0])
dd_probs = dd_rates / sum(np.array(dd_rates))
if n_sim == 0:
n_rep = n_rates
else:
n_rep = n_sim
# simulations______________________________________________________________
# add dd frequencies for additional simulation replicates
if n_rep - n_rates >= 0:
resampling_rates_indexes = np.random.choice(np.arange(n_rates),
(n_rep - n_rates))
append_this_dd = np.array(
[i[resampling_rates_indexes] for i in dd_probs])
final_dd_probs = np.concatenate([dd_probs, append_this_dd], axis=1)
# redraw n samples of transition rates to fill all simulation replicates
append_this = np.array(qmatrix_dict_list)[resampling_rates_indexes]
final_qmatrix_dict_list = list(qmatrix_dict_list) + list(append_this)
else:
resampling_rates_indexes = np.random.choice(np.arange(n_rates), n_rep)
final_dd_probs = np.array(
[i[resampling_rates_indexes] for i in dd_probs])
final_qmatrix_dict_list = np.array(
qmatrix_dict_list)[resampling_rates_indexes]
#current_year = datetime.datetime.now().year
#final_year = current_year+int(n_years)
delta_t = n_years
if model_unknown_as_lc:
print('\nSetting all DD and NE species to LC.')
all_lc = True
else:
all_lc = False
if allow_status_change:
status_change = True
else:
print('\nNot simulating future status changes!')
status_change = False
dynamic_qmatrix = True
print('\nStarting simulations ...')
diversity_through_time, te_array, status_through_time = cust_func.run_multi_sim(
n_rep,
delta_t,
species_list_status,
final_dd_probs,
final_qmatrix_dict_list,
outdir,
all_lc=all_lc,
status_change=status_change,
dynamic_qmatrix=dynamic_qmatrix)
# summarize simulation results
sim_species_list = te_array[:, 0].copy()
ext_date_data = te_array[:, 1:].copy()
extinction_occs = np.array(
[len(row[~np.isnan(list(row))]) for row in ext_date_data])
extinction_prob = extinction_occs / ext_date_data.shape[1]
# produce output file for status distribution through time
mean_status_through_time = np.mean(status_through_time, axis=2)
year = np.arange(delta_t + 1).astype(int)
status_df_data = np.round(np.vstack([year, mean_status_through_time
])).astype(int)
status_df = pd.DataFrame(
data=status_df_data.T,
columns=['year', 'LC', 'NT', 'VU', 'EN', 'CR', 'EX'])
status_df.to_csv(os.path.join(outdir,
'status_distribution_through_time.txt'),
sep='\t',
index=False)
np.savetxt(os.path.join(outdir, 'simulated_extinctions_array.txt'),
status_through_time[-1],
fmt='%i')
# if target_species:
# posterior = get_rate_estimate_posterior(ext_date_data[0],n_years,0,sim_species_list,n_gen=n_gen,burnin=burnin)
# np.savetxt('/Users/tobias/GitHub/iucn_predictions/doc/figures/Figure_2/figure_data/posterior_samples/%s_gl_no_status_change.txt'%target_species.replace(' ','_'),posterior,fmt='%.8f')
# print('\nPrinted posterior')
if plot_diversity_trajectory:
# plot diversity trajectory of species list________________________________
#colors = ["#9a002e","#df4a3d","#fecd5f","#5cd368","#916200"]
# define time axis
time_axis = np.array(range(len(diversity_through_time[0])))
fig = plt.figure()
y_values = np.mean(diversity_through_time, axis=0)
plt.plot(time_axis,
y_values,
color="#b80033",
label='accounting for GL')
# get upper and lower confidence interval boundaries
min_hpd, max_hpd = np.array(
[cust_func.calcHPD(i, 0.95) for i in diversity_through_time.T]).T
mean_min_max = np.vstack([y_values, min_hpd, max_hpd])
np.savetxt(os.path.join(outdir, 'future_diversity_trajectory.txt'),
mean_min_max,
fmt='%.2f')
plt.fill_between(time_axis,
min_hpd,
max_hpd,
color="#b80033",
alpha=0.2)
#plt.legend()
plt.ylabel('Total diversity')
plt.xlabel('Years from present')
ax = plt.gca()
ax1 = ax.twinx()
# Set the limits of the new axis from the original axis limits
ax1.set_ylim(ax.get_ylim())
current_diversity = diversity_through_time[0, 0]
plt.yticks(
[np.mean(diversity_through_time[:, -1])],
[int(current_diversity - np.mean(diversity_through_time[:, -1]))])
#plt.xticks(modified_q_matrix.year[::10],modified_q_matrix.year[::10])
plt.ylabel('Lost species')
plt.tight_layout()
fig.savefig(os.path.join(outdir, 'future_diversity_trajectory.pdf'),
bbox_inches='tight',
dpi=500)
if plot_status_trajectories:
# color palette
colors = [
"#227a00", "#a5c279", "#f3d248", "#6956cb", "#79262a", "#e34349"
]
# define time axis
time_axis = np.array(range(len(diversity_through_time[0])))
# plot results
def plot_mean_and_interval(div, color, label, fig):
plt.plot(time_axis, np.mean(div, axis=0), color=color, label=label)
min_hpd, max_hpd = np.array(
[cust_func.calcHPD(i, 0.95) for i in div.T]).T
plt.fill_between(time_axis,
min_hpd,
max_hpd,
color=color,
alpha=0.2)
return fig
fig = plt.figure(figsize=(10, 10))
plot_mean_and_interval(status_through_time[0, :, :].T, colors[0], 'LC',
fig)
plot_mean_and_interval(status_through_time[1, :, :].T, colors[1], 'NT',
fig)
plot_mean_and_interval(status_through_time[2, :, :].T, colors[2], 'VU',
fig)
plot_mean_and_interval(status_through_time[3, :, :].T, colors[3], 'EN',
fig)
plot_mean_and_interval(status_through_time[4, :, :].T, colors[4], 'CR',
fig)
plot_mean_and_interval(status_through_time[5, :, :].T, colors[5], 'EX',
fig)
# add title, legend and axis-labels
plt.legend(loc='best', fancybox=True)
plt.title('Diversity trajectory IUCN categories - status change'
) #10x higher conservation
plt.ylabel('Number species in category')
plt.xlabel('Years from present')
ax = plt.gca()
ax1 = ax.twinx()
# Set the limits of the new axis from the original axis limits
ax1.set_ylim(ax.get_ylim())
# annotate final counts with labels
right_ticks = [
int(np.round(np.mean(status_through_time[i, -1, :])))
for i in range(status_through_time.shape[0])
]
plt.yticks(right_ticks, right_ticks)
#plt.xticks(modified_q_matrix.year[::10],modified_q_matrix.year[::10])
plt.tight_layout()
fig.savefig(os.path.join(outdir, 'future_status_trajectory.pdf'),
bbox_inches='tight',
dpi=500)
if plot_status_piechart:
statuses, counts = np.unique(species_list_status.current_status.values,
return_counts=True)
init_status_dict = dict(zip(statuses, counts))
init_status_dict['EX'] = 0
iucn_status_code = {
0: 'LC',
1: 'NT',
2: 'VU',
3: 'EN',
4: 'CR',
5: 'EX',
6: 'DD'
}
status_count_list = []
for status_id in np.arange(status_through_time.shape[0] + 1):
status = iucn_status_code[status_id]
if status in init_status_dict.keys():
pre_dd_modeling_count = init_status_dict[status]
else:
pre_dd_modeling_count = 0
if not status == 'DD':
present_status_count = int(
np.round(np.mean(status_through_time[status_id][0])))
final_status_count = int(
np.round(np.mean(status_through_time[status_id][-1])))
else:
present_status_count = 0
final_status_count = 0
status_count_list.append([
pre_dd_modeling_count, present_status_count, final_status_count
])
status_count_list = np.array(status_count_list).T
colors = np.array([
"#227a00", "#a5c279", "#f3d248", "#6956cb", "#79262a", "#b80033",
'black'
])
labels = np.array(['LC', 'NT', 'VU', 'EN', 'CR', 'EX', 'DD'])
def func(pct, allvals):
absolute = int(np.round((pct / 100. * np.sum(allvals))))
return "{:d}".format(absolute)
fig, axs = plt.subplots(1, 3, figsize=(12, 10))
# status distribution beginning
wedges, texts, autotexts = axs[1].pie(
status_count_list[1][status_count_list[1] > 0],
colors=colors[status_count_list[1] > 0],
autopct=lambda pct: func(
pct, status_count_list[1][status_count_list[1] > 0]),
shadow=False,
textprops=dict(color="w"))
# status distribution end
wedges, texts, autotexts = axs[2].pie(
status_count_list[2][status_count_list[2] > 0],
colors=colors[status_count_list[2] > 0],
autopct=lambda pct: func(
pct, status_count_list[2][status_count_list[2] > 0]),
shadow=False,
textprops=dict(color="w"))
ext = wedges[-1]
# status distribution pre-dd
wedges, texts, autotexts = axs[0].pie(
status_count_list[0][status_count_list[0] > 0],
colors=colors[status_count_list[0] > 0],
autopct=lambda pct: func(
pct, status_count_list[0][status_count_list[0] > 0]),
shadow=False,
textprops=dict(color="w"))
axs[0].set_title('Current (including DD)')
axs[1].set_title('Current (DD corrected)')
axs[2].set_title('Final (%i years)' % delta_t)
final_labels = list(labels[status_count_list[0] > 0]) + ['EX']
plt.legend(wedges + [ext],
final_labels,
title="IUCN status\n(N=%i sp.)" %
status_count_list[2].sum(),
loc="center left",
bbox_to_anchor=(1, 0, 0.5, 1))
fig.savefig(os.path.join(outdir, 'status_pie_chart.pdf'),
bbox_inches='tight',
dpi=500)
if extinction_rates:
# calculate some extinction stats__________________________________________
# estimate extinction rates scaled by year
print('\nEstimating extinction rates from simulation output...')
#ext_date_data = ext_date_data[:10,:]
if plot_posterior:
with PdfPages(
os.path.join(outdir,
'posterior_ext_rate_histograms.pdf')) as pdf:
sampled_rates = np.array([
get_rate_estimate(species_values,
n_years,
i,
sim_species_list,
plot_posterior=plot_posterior,
pdf=pdf,
n_gen=n_gen,
burnin=burnin)
for i, species_values in enumerate(ext_date_data)
])
else:
sampled_rates = np.array([
get_rate_estimate(species_values,
n_years,
i,
sim_species_list,
plot_posterior=plot_posterior,
pdf=0,
n_gen=n_gen,
burnin=burnin)
for i, species_values in enumerate(ext_date_data)
])
# export extinction stats to file
column_names = [
'species', 'rate_e_mean', 'rate_e_lower', 'rate_e_upper',
'simulated_p_e_in_%i_years' % delta_t
]
extinction_prob_df = pd.DataFrame(np.array([
sim_species_list, sampled_rates[:, 0], sampled_rates[:, 1],
sampled_rates[:, 2], extinction_prob
]).T,
columns=column_names)
else:
column_names = ['species', 'simulated_p_e_in_%i_years' % delta_t]
extinction_prob_df = pd.DataFrame(np.array(
[sim_species_list, extinction_prob]).T,
columns=column_names)
extinction_prob_df[column_names[1:]] = extinction_prob_df[
column_names[1:]].astype(float)
extinction_prob_df.to_csv(os.path.join(outdir,
'extinction_prob_all_species.txt'),
sep='\t',
index=False,
float_format='%.5f')
print('\n')
if plot_histograms:
# plot histograms of extinction times
with PdfPages(os.path.join(outdir,
'extinction_time_histograms.pdf')) as pdf:
for i, species in enumerate(te_array[:, 0]):
sys.stdout.write(
'\rPlotting extinction histogram for species %i/%i' %
(i + 1, len(te_array[:, 0])))
plt.figure()
species_te_array = te_array[:, 1:][i]
not_na_values = species_te_array[~np.
isnan(list(species_te_array))]
heights, bins = np.histogram(not_na_values,
np.arange(0, delta_t + 10, 10))
percent = heights / n_rep
plt.bar(bins[:-1], percent, width=10, align="edge")
plt.ylim(0, 0.5)
#survival_prob = 1-sum(percent)
#if survival_prob >= 0.5:
# text_color = 'green'
#else:
# text_color = 'red'
ax = plt.gca()
#plt.text(0.05, 0.7, 'survival probability: %.2f'%survival_prob,color=text_color, horizontalalignment='left',verticalalignment='baseline', transform=ax.transAxes)
# annotate last bar
if ax.patches[-1].get_height() > 0:
ax.text(ax.patches[-1].get_x() + 3,
np.round(ax.patches[-1].get_height() + 0.001, 4),
'**',
fontsize=12,
color='black')
plt.title('%s - Extinct in %i years: %i/%i' %
(species, delta_t, sum(heights), n_rep))
plt.xlabel('Years from present')
plt.ylabel('Fraction of simulations')
plt.tight_layout()
pdf.savefig() # saves the current figure into a pdf page
plt.close()
print('\n')
def main(args):
# get user input___________________________________________________________
seed = args.seed
try:
random_seed = int(seed)
print('Simulating with user-set starting seed %i.'%random_seed)
except:
random_seed = np.random.randint(999999999)
print('Simulating with randomely generated starting seed %i.'%random_seed)
np.random.seed(random_seed)
infile = args.input_data
outdir = args.outdir
n_years = int(args.n_years)
n_sim = int(args.n_sim)
n_extinct_taxa = int(args.n_extinct_taxa)
allow_status_change = int(args.status_change)
conservation_increase_factor = int(args.conservation_increase_factor)
threat_increase_factor = int(args.threat_increase_factor)
extinction_rates = int(args.extinction_rates)
n_gen = int(args.n_gen)
burnin = int(args.burnin)
plot_diversity_trajectory = int(args.plot_diversity_trajectory)
plot_histograms = int(args.plot_histograms)
plot_posterior = int(args.plot_posterior)
model_unknown_as_lc = int(args.model_unknown_as_lc)
plot_status_trajectories = int(args.plot_status_trajectories)
plot_status_piechart = int(args.plot_status_piechart)
if not os.path.exists(outdir):
os.makedirs(outdir)
np.savetxt(os.path.join(outdir,'starting_seed.txt'),np.array([random_seed]),fmt='%i')
input_data = cust_func.load_obj(infile)
species_input_data, dd_probs = input_data
species_list = np.array([i[0] for i in species_input_data])
current_status_list = np.array([i[1] for i in species_input_data])
q_matrix_list = [i[2] for i in species_input_data]
n_rates = dd_probs.shape[1]
#__________________________________________________________________________
# modify q-matrices, if set by user________________________________________
final_qmatrix_list = []
q_matrix_list_copy = np.array(q_matrix_list).copy()
for q_matrix_list_i in q_matrix_list_copy:
q_matrix_list_temp = []
for q_matrix in q_matrix_list_i:
if conservation_increase_factor != 1:
indeces_lower_triangle = np.tril_indices(q_matrix.shape[0],-1)
q_matrix[indeces_lower_triangle] = q_matrix[indeces_lower_triangle] * conservation_increase_factor
np.fill_diagonal(q_matrix,0)
np.fill_diagonal(q_matrix, -np.sum(q_matrix,axis=1))
if threat_increase_factor != 1:
indeces_upper_triangle = np.triu_indices(q_matrix.shape[0],1)
q_matrix[indeces_upper_triangle] = q_matrix[indeces_upper_triangle] * threat_increase_factor
np.fill_diagonal(q_matrix,0)
np.fill_diagonal(q_matrix, -np.sum(q_matrix,axis=1))
q_matrix_list_temp.append(q_matrix)
final_qmatrix_list.append(q_matrix_list_temp)
# turn into a dict with the n qmatrices for each species
final_qmatrix_dict = dict(zip(species_list,final_qmatrix_list))
#__________________________________________________________________________
# if more n_rep are set than there are q-matrices, resample________________
if n_sim <= n_rates:
sample_columns = np.random.choice(np.arange(n_rates),size=n_sim,replace=False)
# since there are only as many cr and en p(ex) estimates as there are provided GL values, we may have to resample some (but make sure all are present at least once)
else:
sample_columns1 = np.random.choice(np.arange(n_rates),size=n_rates,replace=False)
sample_columns2 = np.random.choice(np.arange(n_rates),size=(n_sim-n_rates),replace=True)
sample_columns = np.concatenate([sample_columns1,sample_columns2])
#__________________________________________________________________________
# run simulations__________________________________________________________
delta_t = n_years
model_ne_as_dd = False
dynamic_qmatrix = True
if model_ne_as_dd:
current_status_list[current_status_list=='NE'] = 'DD'
if model_unknown_as_lc:
print('\nSetting all DD and NE species to LC.')
all_lc=True
else:
all_lc=False
if allow_status_change:
status_change=True
else:
print('\nNot simulating future status changes!')
status_change=False
print('\nStarting simulations ...')
diversity_through_time,te_array,status_through_time,time_until_n_extinctions_list = cust_func.run_multi_sim(n_sim,delta_t,species_list,current_status_list,dd_probs,final_qmatrix_dict,sample_columns,outdir,all_lc=all_lc,status_change=status_change,dynamic_qmatrix=dynamic_qmatrix,n_extinct_taxa=n_extinct_taxa)
# summarize simulation results
sim_species_list = te_array[:,0].copy()
ext_date_data = te_array[:,1:].copy()
extinction_occs = np.array([len(row[~np.isnan(list(row))]) for row in ext_date_data])
extinction_prob = extinction_occs/ext_date_data.shape[1]
# produce output file for status distribution through time
mean_status_through_time = np.mean(status_through_time,axis=2)
year = np.arange(delta_t+1).astype(int)
status_df_data = np.round(np.vstack([year,mean_status_through_time])).astype(int)
status_df = pd.DataFrame(data = status_df_data.T,columns=['year','LC','NT','VU','EN','CR','EX'])
status_df.to_csv(os.path.join(outdir,'status_distribution_through_time.txt'),sep='\t',index=False)
np.savetxt(os.path.join(outdir,'simulated_extinctions_array.txt'),status_through_time[-1],fmt='%i')
pd.DataFrame(data=te_array).to_csv(os.path.join(outdir,'te_all_species.txt'),sep='\t',header=False,index=False)
#__________________________________________________________________________
#__________________________________________________________________________
# if target_species:
# posterior = get_rate_estimate_posterior(ext_date_data[0],n_years,0,sim_species_list,n_gen=n_gen,burnin=burnin)
# np.savetxt('/Users/tobias/GitHub/iucn_predictions/doc/figures/Figure_2/figure_data/posterior_samples/%s_gl_no_status_change.txt'%target_species.replace(' ','_'),posterior,fmt='%.8f')
# print('\nPrinted posterior')
#__________________________________________________________________________
if n_extinct_taxa:
np.savetxt(os.path.join(outdir,'time_until_%i_extinctions.txt'%n_extinct_taxa),time_until_n_extinctions_list,fmt='%.2f')
fig = plt.figure()
plt.hist(time_until_n_extinctions_list)
plt.xlabel('Time in years')
plt.ylabel('N')
plt.title('Simulated years until %i extinctions (%i simulations)'%(n_extinct_taxa,n_sim))
fig.savefig(os.path.join(outdir,'time_until_%i_extinctions.pdf'%n_extinct_taxa),bbox_inches='tight', dpi = 500)
#__________________________________________________________________________
if plot_diversity_trajectory:
# plot diversity trajectory of species list________________________________
#colors = ["#9a002e","#df4a3d","#fecd5f","#5cd368","#916200"]
# define time axis
time_axis = np.array(range(len(diversity_through_time[0])))
fig = plt.figure()
y_values = np.mean(diversity_through_time, axis =0)
plt.plot(time_axis,y_values,color="#b80033", label='accounting for GL')
# get upper and lower confidence interval boundaries
min_hpd, max_hpd = np.array([cust_func.calcHPD(i,0.95) for i in diversity_through_time.T]).T
mean_min_max = np.vstack([y_values,min_hpd,max_hpd])
np.savetxt(os.path.join(outdir,'future_diversity_trajectory.txt'),mean_min_max,fmt='%.2f')
plt.fill_between(time_axis, min_hpd, max_hpd,
color="#b80033", alpha=0.2)
#plt.legend()
plt.ylabel('Total diversity')
plt.xlabel('Years from present')
ax = plt.gca()
ax1 = ax.twinx()
# Set the limits of the new axis from the original axis limits
ax1.set_ylim(ax.get_ylim())
current_diversity = diversity_through_time[0,0]
plt.yticks([np.mean(diversity_through_time[:,-1])],[int(current_diversity-np.mean(diversity_through_time[:,-1]))])
#plt.xticks(modified_q_matrix.year[::10],modified_q_matrix.year[::10])
plt.ylabel('Lost species')
plt.tight_layout()
fig.savefig(os.path.join(outdir,'future_diversity_trajectory.pdf'),bbox_inches='tight', dpi = 500)
#__________________________________________________________________________
#__________________________________________________________________________
if plot_status_trajectories:
# color palette
colors = ["#227a00","#a5c279","#f3d248","#6956cb","#79262a","#e34349"]
# define time axis
time_axis = np.array(range(len(diversity_through_time[0])))
# plot results
def plot_mean_and_interval(div,color,label,fig):
plt.plot(time_axis,np.mean(div,axis=0),color=color,label=label);
min_hpd, max_hpd = np.array([cust_func.calcHPD(i,0.95) for i in div.T]).T
plt.fill_between(time_axis, min_hpd, max_hpd, color=color, alpha=0.2);
return fig
fig = plt.figure(figsize=(10,10))
plot_mean_and_interval(status_through_time[0,:,:].T,colors[0],'LC',fig)
plot_mean_and_interval(status_through_time[1,:,:].T,colors[1],'NT',fig)
plot_mean_and_interval(status_through_time[2,:,:].T,colors[2],'VU',fig)
plot_mean_and_interval(status_through_time[3,:,:].T,colors[3],'EN',fig)
plot_mean_and_interval(status_through_time[4,:,:].T,colors[4],'CR',fig)
plot_mean_and_interval(status_through_time[5,:,:].T,colors[5],'EX',fig)
# add title, legend and axis-labels
plt.legend(loc='best',fancybox=True)
plt.title('Diversity trajectory IUCN categories - status change') #10x higher conservation
plt.ylabel('Number species in category')
plt.xlabel('Years from present')
ax = plt.gca()
ax1 = ax.twinx()
# Set the limits of the new axis from the original axis limits
ax1.set_ylim(ax.get_ylim())
# annotate final counts with labels
right_ticks = [int(np.round(np.mean(status_through_time[i,-1,:]))) for i in range(status_through_time.shape[0])]
plt.yticks(right_ticks,right_ticks)
#plt.xticks(modified_q_matrix.year[::10],modified_q_matrix.year[::10])
plt.tight_layout()
fig.savefig(os.path.join(outdir,'future_status_trajectory.pdf'),bbox_inches='tight', dpi = 500)
#__________________________________________________________________________
#__________________________________________________________________________
if plot_status_piechart:
statuses, counts = np.unique(current_status_list,return_counts=True)
init_status_dict = dict(zip(statuses, counts))
init_status_dict['EX'] = 0
iucn_status_code = {0:'LC', 1:'NT', 2:'VU', 3:'EN', 4:'CR', 5:'EX', 6:'DD'}
status_count_list = []
for status_id in np.arange(status_through_time.shape[0]+1):
status = iucn_status_code[status_id]
if status in init_status_dict.keys():
pre_dd_modeling_count = init_status_dict[status]
else:
pre_dd_modeling_count = 0
if not status == 'DD':
present_status_count = int(np.round(np.mean(status_through_time[status_id][0])))
final_status_count = int(np.round(np.mean(status_through_time[status_id][-1])))
else:
present_status_count = 0
final_status_count = 0
status_count_list.append([pre_dd_modeling_count,present_status_count,final_status_count])
status_count_list = np.array(status_count_list).T
colors = np.array(["#227a00","#a5c279","#f3d248","#6956cb","#79262a","#b80033",'black'])
labels = np.array(['LC', 'NT', 'VU', 'EN', 'CR', 'EX', 'DD'])
def func(pct, allvals):
absolute = int(np.round((pct/100.*np.sum(allvals))))
return "{:d}".format(absolute)
fig, axs = plt.subplots(1, 3,figsize=(12,10))
# status distribution beginning
wedges, texts, autotexts =axs[1].pie(status_count_list[1][status_count_list[1] >0], colors= colors[status_count_list[1] >0], autopct=lambda pct: func(pct, status_count_list[1][status_count_list[1] >0]), shadow=False,textprops=dict(color="w"))
# status distribution end
wedges, texts, autotexts =axs[2].pie(status_count_list[2][status_count_list[2] >0], colors= colors[status_count_list[2] >0], autopct=lambda pct: func(pct, status_count_list[2][status_count_list[2] >0]), shadow=False,textprops=dict(color="w"))
ext = wedges[-1]
# status distribution pre-dd
wedges, texts, autotexts =axs[0].pie(status_count_list[0][status_count_list[0] >0], colors= colors[status_count_list[0] >0], autopct=lambda pct: func(pct, status_count_list[0][status_count_list[0] >0]), shadow=False,textprops=dict(color="w"))
axs[0].set_title('Current (including DD)')
axs[1].set_title('Current (DD corrected)')
axs[2].set_title('Final (%i years)'%delta_t)
final_labels = list(labels[status_count_list[0] >0]) + ['EX']
plt.legend(wedges+[ext], final_labels,title="IUCN status\n(N=%i sp.)"%status_count_list[2].sum(),loc="center left",bbox_to_anchor=(1, 0, 0.5, 1))
fig.savefig(os.path.join(outdir,'status_pie_chart.pdf'),bbox_inches='tight', dpi = 500)
#__________________________________________________________________________
#__________________________________________________________________________
if extinction_rates:
# calculate some extinction stats
# estimate extinction rates scaled by year
print('\nRunning %i MCMCs to estimate species-specific extinction rates from simulation output...'%len(species_list))
#ext_date_data = ext_date_data[:10,:]
if plot_posterior:
with PdfPages(os.path.join(outdir,'posterior_ext_rate_histograms.pdf')) as pdf:
sampled_rates = np.array([get_rate_estimate(species_values,n_years,i,sim_species_list,plot_posterior=plot_posterior,pdf=pdf,n_gen=n_gen,burnin=burnin) for i,species_values in enumerate(ext_date_data)])
else:
sampled_rates = np.array([get_rate_estimate(species_values,n_years,i,sim_species_list,plot_posterior=plot_posterior,pdf=0,n_gen=n_gen,burnin=burnin) for i,species_values in enumerate(ext_date_data)])
# export extinction stats to file
column_names = ['species','rate_e_mean','rate_e_lower','rate_e_upper','simulated_p_e_in_%i_years'%delta_t]
extinction_prob_df = pd.DataFrame(np.array([sim_species_list,sampled_rates[:,0],sampled_rates[:,1],sampled_rates[:,2],extinction_prob]).T,columns=column_names)
else:
column_names = ['species','simulated_p_e_in_%i_years'%delta_t]
extinction_prob_df = pd.DataFrame(np.array([sim_species_list,extinction_prob]).T,columns=column_names)
extinction_prob_df[column_names[1:]] = extinction_prob_df[column_names[1:]].astype(float)
extinction_prob_df.to_csv(os.path.join(outdir,'extinction_prob_all_species.txt'),sep='\t',index=False,float_format='%.8f')
print('\n')
#__________________________________________________________________________
#__________________________________________________________________________
if plot_histograms:
# plot histograms of extinction times
with PdfPages(os.path.join(outdir,'extinction_time_histograms.pdf')) as pdf:
for i,species in enumerate(te_array[:,0]):
sys.stdout.write('\rPlotting extinction histogram for species %i/%i'%(i+1,len(te_array[:,0])))
plt.figure()
species_te_array = te_array[:,1:][i]
not_na_values = species_te_array[~np.isnan(list(species_te_array))]
heights, bins = np.histogram(not_na_values,np.arange(0,delta_t+10,10))
percent = heights/n_sim
plt.bar(bins[:-1],percent,width=10, align="edge")
plt.ylim(0,0.5)
#survival_prob = 1-sum(percent)
#if survival_prob >= 0.5:
# text_color = 'green'
#else:
# text_color = 'red'
ax = plt.gca()
#plt.text(0.05, 0.7, 'survival probability: %.2f'%survival_prob,color=text_color, horizontalalignment='left',verticalalignment='baseline', transform=ax.transAxes)
# annotate last bar
if ax.patches[-1].get_height() > 0:
ax.text(ax.patches[-1].get_x()+3, np.round(ax.patches[-1].get_height()+0.001,4), '**', fontsize=12, color='black')
plt.title('%s - Extinct in %i years: %i/%i'%(species,delta_t,sum(heights),n_sim))
plt.xlabel('Years from present')
plt.ylabel('Fraction of simulations')
plt.tight_layout()
pdf.savefig() # saves the current figure into a pdf page
plt.close()
print('\n')
#__________________________________________________________________________
# 10961-7245.00
# 10961-8018.00
# test rate and see how many exitncitons it produces
ext_species = []
for i in np.arange(10000):
true_rate = 0.000217
n_sim = 10961
ext_time_array = (np.random.exponential(1./true_rate, n_sim)).astype(int)
max_t=500
ext_time_array[ext_time_array>max_t] = 0
values,counts = np.unique(ext_time_array,return_counts=True)
n_ext = 10961-counts[0]
ext_species.append(n_ext)
np.mean(ext_species)
cust_func.calcHPD(ext_species,0.95)
# compare extinction rates:
highlight_dd_ne = True
plot_species = 'Strigops habroptila'
ex_mode_1_file = '/Users/tobias/GitHub/iucn_extinction_simulator/data/iucn_sim_output/aves/future_sim_0/extinction_prob_all_species.txt'
ex_mode_2_file = '/Users/tobias/GitHub/iucn_extinction_simulator/data/iucn_sim_output/aves/future_sim_1_pex/extinction_prob_all_species.txt'
ex_mode_1 = pd.read_csv(ex_mode_1_file,sep='\t').rate_e_mean.values
ex_mode_2 = pd.read_csv(ex_mode_2_file,sep='\t').rate_e_mean.values
species_names = pd.read_csv(ex_mode_2_file,sep='\t').species.values
status_info_file = '/Users/tobias/GitHub/iucn_extinction_simulator/data/iucn_sim_output/aves/iucn_data/species_data.txt'
status_array = pd.read_csv(status_info_file,sep='\t',header=None).iloc[:,1].values
species_list_status_array = pd.read_csv(status_info_file,sep='\t',header=None).iloc[:,0].values
