def test_plot_allmodelruns(self):
from spotpy.examples.spot_setup_hymod_python import spot_setup as sp
sp = sp()
sampler = spotpy.algorithms.dream(sp,
parallel="seq",
dbname='test_plot_allmodelruns',
dbformat="ram",
sim_timeout=5)
sampler.sample(50)
modelruns = []
for run in sampler.getdata():
on_run = []
for i in run:
on_run.append(i)
on_run = np.array(on_run)[:-9]
print(on_run)
modelruns.append(on_run.tolist())
test_plot_allmodelruns = spotpy.analyser.plot_allmodelruns(
modelruns,
sp.evaluation(),
dates=range(1,
len(sp.evaluation()) + 1))
fig_name = "bestmodel.png"
# approximately 8855 KB is the size of an empty matplotlib.pyplot.plot, so
# we expecting a plot with some content without testing the structure of the pot just
# the size
self.assertGreaterEqual(os.path.getsize(fig_name), 8855)
os.remove(fig_name)
fig = plt.figure(figsize=(16, 9))
ax = plt.subplot(1, 1, 1)
q5, q25, q75, q95 = [], [], [], []
for field in fields:
q5.append(np.percentile(results[field][-100:-1], 2.5))
q95.append(np.percentile(results[field][-100:-1], 97.5))
ax.plot(q5, color='dimgrey', linestyle='solid')
ax.plot(q95, color='dimgrey', linestyle='solid')
ax.fill_between(np.arange(0, len(q5), 1),
list(q5),
list(q95),
facecolor='dimgrey',
zorder=0,
linewidth=0,
label='parameter uncertainty')
ax.plot(spot_setup.evaluation(), 'r.', label='data')
ax.set_ylim(-50, 450)
ax.set_xlim(0, 729)
ax.legend()
fig.savefig('python_hymod.png', dpi=300)
#########################################################
# Example plot to show the convergence #################
fig = plt.figure(figsize=(12, 16))
plt.subplot(2, 1, 1)
for i in range(int(max(results['chain'])) + 1):
index = np.where(results['chain'] == i)
plt.plot(results['like1'][index], label='Chain ' + str(i + 1))
plt.ylabel('Likelihood value')
plt.legend()
# Get fields with simulation data
fields=[word for word in results.dtype.names if word.startswith('sim')]
# Example plot to show remaining parameter uncertainty #
fig= plt.figure(figsize=(9,6))
ax = plt.subplot(1,1,1)
q5,q25,q75,q95=[],[],[],[]
for field in fields:
q5.append(np.percentile(results[field][-100:-1],2.5))# ALl 100 runs after convergence
q95.append(np.percentile(results[field][-100:-1],97.5))# ALl 100 runs after convergence
ax.plot(q5,color='dimgrey',linestyle='solid')
ax.plot(q95,color='dimgrey',linestyle='solid')
ax.fill_between(np.arange(0,len(q5),1),list(q5),list(q95),facecolor='dimgrey',zorder=0,
linewidth=0,label='simulation uncertainty')
ax.plot(spot_setup.evaluation(),color='red', markersize=2,label='data')
ax.set_ylim(-50,450)
ax.set_xlim(0,729)
ax.set_ylabel('Discharge [l s-1]')
ax.set_xlabel('Days')
ax.legend()
fig.savefig('DREAM_simulation_uncertainty_Hymod.png',dpi=150)
#########################################################
# Example plot to show the convergence #################
spotpy.analyser.plot_gelman_rubin(results, r_hat, fig_name='DREAM_r_hat.png')
########################################################
plt.xlabel('Iteration')
fig.savefig('SCEUA_objectivefunctiontrace.png', dpi=150)
# Plot the best model run
#Find the run_id with the minimal objective function value
bestindex, bestobjf = spotpy.analyser.get_minlikeindex(results)
# Select best model run
best_model_run = results[bestindex]
#Filter results for simulation results
fields = [
word for word in best_model_run.dtype.names if word.startswith('sim')
]
best_simulation = list(best_model_run[fields])
fig = plt.figure(figsize=(9, 6))
ax = plt.subplot(1, 1, 1)
ax.plot(best_simulation,
color='black',
linestyle='solid',
label='Best objf.=' + str(bestobjf))
ax.plot(spot_setup.evaluation(),
'r.',
markersize=3,
label='Observation data')
plt.xlabel('Number of Observation Points')
plt.ylabel('Discharge [l s-1]')
plt.legend(loc='upper right')
fig.savefig('SCEUA_best_modelrun.png', dpi=150)
