def test_equally_spaced_nodes():
x = u.equally_spaced_nodes(5, eps=1e-4)
assert len(x) == 7
d = np.diff(x)
np.testing.assert_almost_equal(d[0], d[1] / 2, 3)
x = u.equally_spaced_nodes(1)
np.testing.assert_almost_equal(x[0], 0.5)
def train_quantiledeltamapping(
reference, historical, variable, kind, quantiles_n=100, window_n=31
):
"""Train quantile delta mapping
Parameters
----------
reference : xr.Dataset
Dataset to use as model reference. Target variable must have a units attribute.
historical : xr.Dataset
Dataset to use as historical simulation. Target variable must have a units attribute.
variable : str
Name of target variable to extract from `historical` and `reference`.
kind : {"+", "*"}
Kind of variable. Used for QDM scaling.
quantiles_n : int, optional
Number of quantiles for QDM.
window_n : int, optional
Centered window size for day-of-year grouping.
Returns
-------
xclim.sdba.adjustment.QuantileDeltaMapping
"""
qdm = sdba.adjustment.QuantileDeltaMapping.train(
ref=reference[variable],
hist=historical[variable],
kind=str(kind),
group=sdba.Grouper("time.dayofyear", window=int(window_n)),
nquantiles=equally_spaced_nodes(int(quantiles_n), eps=None),
)
return qdm
def cannon_2015_rvs(n, random=True):
# Frozen distributions
fd = cannon_2015_dist()
if random:
r = [d.rvs(n) for d in fd]
else:
u = equally_spaced_nodes(n, None)
r = [d.ppf(u) for d in fd]
return map(lambda x: series(x, "pr"), r)
def _make_qm(a, *, group="time.month"):
a = np.atleast_2d(a)
n, m = a.shape
mo = range(1, m + 1)
if group.prop:
q = equally_spaced_nodes(n, None)
dims = ("quantiles", group.prop)
coords = {"quantiles": q, "month": mo}
else:
q = equally_spaced_nodes(m, None)
dims = ("quantiles", )
coords = {"quantiles": q}
a = a[0]
return xr.DataArray(
a,
dims=dims,
coords=coords,
attrs={
"group": group,
"window": 1
},
)
def test_interp_on_quantiles_monthly():
t = xr.cftime_range("2000-01-01",
"2030-12-31",
freq="D",
calendar="noleap")
ref = xr.DataArray(
(-20 * np.cos(2 * np.pi * t.dayofyear / 365) +
2 * np.random.random_sample((t.size, )) + 273.15 + 0.1 *
(t - t[0]).days / 365), # "warming" of 1K per decade,
dims=("time", ),
coords={"time": t},
attrs={"units": "K"},
)
sim = xr.DataArray(
(-18 * np.cos(2 * np.pi * t.dayofyear / 365) +
2 * np.random.random_sample((t.size, )) + 273.15 + 0.11 *
(t - t[0]).days / 365), # "warming" of 1.1K per decade
dims=("time", ),
coords={"time": t},
attrs={"units": "K"},
)
ref = ref.sel(time=slice(None, "2015-01-01"))
hist = sim.sel(time=slice(None, "2015-01-01"))
group = Grouper("time.month")
quantiles = u.equally_spaced_nodes(15, eps=1e-6)
ref_q = group.apply(nbu.quantile, ref, main_only=True, q=quantiles)
hist_q = group.apply(nbu.quantile, hist, main_only=True, q=quantiles)
af = u.get_correction(hist_q, ref_q, "+")
for interp in ["nearest", "linear", "cubic"]:
afi = u.interp_on_quantiles(sim,
hist_q,
af,
group="time.month",
method=interp,
extrapolation="constant")
assert afi.isnull().sum("time") == 0, interp