def test_nash_sutcliff():
"""
test nash sutcliff
"""
x = random.randn(10)
ok_(scores.nash_sutcliffe(x, x) == 1)
ok_(scores.nash_sutcliffe(x, 5 * x) > 0)
ok_(scores.nash_sutcliffe(x, -x) < 0)
def plot_total_precip_and_temp_re_1d(ax_pr,
ax_temp,
crcm5_manager,
rot_latlon_projection,
areas2d,
model_data,
mask=None):
"""
plot relative error of total precipitation in time
"""
assert isinstance(crcm5_manager, Crcm5ModelDataManager)
assert isinstance(ax_pr, Axes)
if mask is None:
i_model0, j_model0 = model_data.metadata["ix"], model_data.metadata[
"jy"]
mask = crcm5_manager.get_mask_for_cells_upstream(i_model0, j_model0)
#print model_data.time[0], model_data.time[-1]
#####Precipitation
ts_prec_mod, dt_mod = crcm5_manager.get_monthly_sums_over_points(
mask,
"PR",
areas2d=areas2d,
start_date=model_data.time[0],
end_date=model_data.time[-1])
cruManager = CRUDataManager(
path="data/cru_data/CRUTS3.1/cru_ts_3_10.1901.2009.pre.dat.nc",
var_name="pre")
ts_prec_obs = cruManager.get_monthly_timeseries_using_mask(
mask,
crcm5_manager.lons2D,
crcm5_manager.lats2D,
areas2d,
start_date=model_data.time[0],
end_date=model_data.time[-1])
mod = np.array(
ts_prec_mod.data) * dt_mod.seconds * 1000.0 #converting to mm/month
sta = np.array(ts_prec_obs.data) #CRU data used was initially in mm/month
#print sta
#print mod
#print mod.shape
#print sta.shape
assert len(sta) == len(mod)
ax_pr.annotate("r = {0:.2f}".format(float(np.corrcoef([mod, sta])[0, 1])),
xy=(0.7, 0.8),
xycoords="axes fraction")
ax_pr.annotate("ns = {0:.2f}".format(scores.nash_sutcliffe(mod, sta)),
xy=(0.7, 0.9),
xycoords="axes fraction")
ax_pr.plot(ts_prec_mod.time, (mod - sta) / sta, color="k", linewidth=2)
ax_pr.xaxis.set_major_formatter(DateFormatter("%y/%m"))
ax_pr.xaxis.set_major_locator(MonthLocator(bymonth=list(range(1, 13, 2))))
ax_pr.set_title("CRCM5 versus CRU")
ax_pr.set_ylabel("Total Precip (Pmod - Pobs)/Pobs")
#Temperature
ts_temp_mod = crcm5_manager.get_spatial_integral_over_mask_of_dyn_field(
mask,
areas2d,
var_name="TT",
level=1000,
level_kind=level_kinds.PRESSURE,
path_to_folder=crcm5_manager.samples_folder,
file_prefix="dp")
print("ts_temp_mod time interval: ", ts_temp_mod.time[0],
ts_temp_mod.time[-1])
ts_temp_mod = ts_temp_mod.time_slice(model_data.time[0],
model_data.time[-1])
cruManager = CRUDataManager(
path="data/cru_data/CRUTS3.1/cru_ts_3_10.1901.2009.tmp.dat.nc",
var_name="tmp")
ts_temp_obs = cruManager.get_monthly_timeseries_using_mask(
mask,
crcm5_manager.lons2D,
crcm5_manager.lats2D,
areas2d,
start_date=model_data.time[0],
end_date=model_data.time[-1])
mod = np.array(ts_temp_mod.get_ts_of_monthly_means().data) / sum(
areas2d[mask == 1]) #converting to mm/month
sta = np.array(ts_temp_obs.data) / sum(
areas2d[mask == 1]) #CRU data used was initially in mm/month
assert len(sta) == len(mod)
ax_temp.annotate("r = {0:.2f}".format(float(np.corrcoef([mod, sta])[0,
1])),
xy=(0.7, 0.8),
xycoords="axes fraction")
ax_temp.annotate("ns = {0:.2f}".format(scores.nash_sutcliffe(mod, sta)),
xy=(0.7, 0.9),
xycoords="axes fraction")
ax_temp.plot(ts_prec_mod.time, (mod - sta), color="k", linewidth=2)
ax_temp.xaxis.set_major_formatter(DateFormatter("%y/%m"))
ax_temp.xaxis.set_major_locator(
MonthLocator(bymonth=list(range(1, 13, 2))))
ax_temp.set_title("CRCM5 versus CRU")
ax_temp.set_ylabel("Temperature, deg., Tmod - Tobs")
pass
