def compute(dm, do):
""" Computes Mean Absolute Error"""
if dm is None and do is None: # just want the doc
return {
"Name": "Mean Absolute Error",
"Abstract": "Compute Full Average of " +
"Absolute Difference Between Model And Observation",
"Contact": "Peter Gleckler <*****@*****.**>",
}
mae = MV.average(MV.absolute(MV.subtract(dm, do)))
return float(mae)
def compute(dm, do):
""" Computes Mean Absolute Error"""
if dm is None and do is None: # just want the doc
return {
"Name": "Mean Absolute Error",
"Abstract": "Compute Full Average of " +
"Absolute Difference Between Model And Observation",
"Contact": "*****@*****.**",
}
absdif = MV2.absolute(MV2.subtract(dm, do))
mae = cdutil.averager(absdif, axis='xyt', weights='weighted')
# mae = MV.average(MV.absolute(MV.subtract(dm, do))) - depricated ... did
# not include area weights
return float(mae)
def compute(dm, do):
"""Computes Mean Absolute Error"""
if dm is None and do is None: # just want the doc
return {
"Name": "Mean Absolute Error",
"Abstract": "Compute Full Average of " +
"Absolute Difference Between Model And Observation",
"Contact": "*****@*****.**",
}
absdif = MV2.absolute(MV2.subtract(dm, do))
mae = cdutil.averager(absdif, axis="xy", weights="weighted")
# mae = MV.average(MV.absolute(MV.subtract(dm, do))) - depricated ... did
# not include area weights
return float(mae)
def fingerprint_agreement_percentages(D,SM=None):
pdsi_eof=da.get_orientation(D.ALL.solver)*D.ALL.solver.eofs()[0]
mask = D.ALL.obs[0].mask
if SM is None:
SM = b.SoilMoisture(mask)
SM30_eof=da.get_orientation(SM.solvers["30cm"] )*SM.solvers["30cm"].eofs()[0]
SM2m_eof=da.get_orientation(SM.solvers["2m"] )*SM.solvers["2m"].eofs()[0]
samesign=cmip5.cdms_clone(np.sign(pdsi_eof)*np.sign(SM30_eof),pdsi_eof)
aw=cdutil.area_weights(pdsi_eof)
test_area=np.ma.sum(MV.absolute(samesign)*aw)
samesign_area=np.ma.sum(MV.masked_where(samesign<1,samesign)*aw)
print "PDSI and 30cm have same sign in "+str(samesign_area/test_area*100)+"% of area"
samesign=cmip5.cdms_clone(np.sign(pdsi_eof)*np.sign(SM2m_eof),pdsi_eof)
samesign_area=np.ma.sum(MV.masked_where(samesign<1,samesign)*aw)
print "PDSI and 2m have same sign in "+str(samesign_area/test_area*100)+"% of area"
samesign=cmip5.cdms_clone(np.sign(SM30_eof)*np.sign(SM2m_eof),pdsi_eof)
samesign_area=np.ma.sum(MV.masked_where(samesign<1,samesign)*aw)
print "30cm and 2m have same sign in "+str(samesign_area/test_area*100)+"% of area"
ulat = ud.getLatitude()
if not isinstance(udat, TV): markError('Slice does not return TV')
f = cdms2.open(os.path.join(pth, 'u_2000.nc'))
uf = f['u']
vel = MV2.sqrt(ud * ud + vd * vd)
vel.id = 'velocity'
vel2 = MV2.sqrt(udat * udat + vdat * vdat)
vel2.id = 'velocity'
vel2.units = ud.units
if not MV2.allequal(vel, vel2): markError('Slice operators do not compare')
x1 = uf + 1.0
x2 = 1.0 - ud
x11 = -uf
x12 = MV2.absolute(ud)
x3 = uf + x2
x4 = 1.0 + ud
x5 = uf - 1
x6 = ud * uf
x7 = ud / x2
x8 = 1 / uf
x9 = 3 * ud
x10 = uf**3
x13 = MV2.add.reduce(uf)
x14 = MV2.add.reduce(ud)
x15 = x9.astype(numpy.float32)
if not x15.dtype.char == numpy.sctype2char(numpy.float32):
markError('astype error')
## arrayrange(start, stop=None, step=1, typecode=None, axis=None, attributes=None, id=None)
def compute(dm,do):
""" Computes Mean Absolute Error"""
mae = MV.average(MV.absolute(MV.subtract(dm,do)))
return float(mae)
def calc_PDSI(Z, BTthresh):
#Allocate space for effective wetness/dryness
Ud = Z * 0.
Uw = Z * 0.
X1 = Z * 0.
X2 = Z * 0.
X3 = Z * 0.
Pe = Z * 0.
BT = Z * 0.
montho = MV.zeros(
Z.shape) #log how many months wet/dry spells have been going on
X = MV.zeros(Z.shape)
#Start in month 1, inheriting no wetness/dryness from previous month
i = 0
XX = Z[i] / 3. #PDSI value if there is no wet or dry spell underway
X1i = MV.where(XX > 0, XX, 0)
X1i = MV.where(XX > 1, 0, X1i)
X2i = MV.where(XX < 0, XX, 0)
X2i = MV.where(XX < -1, 0, X2i)
X3i = MV.where(MV.absolute(XX) > 1, XX, 0)
#how many months has a wet/dry spell lasted?
nmonths = MV.where(MV.absolute(XX) > 1, 1, 0)
montho[i] = nmonths
X[i] = XX
X1[i] = X1i
X2[i] = X2i
X3[i] = X3i
for i in range(Z.shape[0])[1:10]: #Loop over time
XX = Z[i] / 3.
nmonths = montho[i - 1] #months so far in dry spell
#These value will be used when calculating the probability of ending a wet/dry spell.
#An established spell will be maintained at |PDSI|>=0.5 with a value of |Z| >=0.15
Ud[i] = Z[i] - 0.15
Uw[i] = Z[i] + 0.15
this_month_wet = XX > 0
this_month_dry = XX < 0
this_month_neutral = XX == 0
#PDSI value if initiating a dry spell
X2i = MV.where((X2[i - 1] * (1 - 0.103) + XX) < XX,
X2[i - 1] * (1 - 0.103) + XX, XX)
X1i = MV.where((X1[i - 1] * (1 - 0.103) + XX) > XX,
X1[i - 1] * (1 - 0.103) + XX, XX)
transitional_value = MV.where(this_month_wet, X1[i], 0)
transitional_value = MV.where(this_month_dry, X2[i], 0)
extreme_transitional_value = MV.absolute(transitional_value) > 1
last_month_neutral = X3[i - 1] == 0
last_month_dry = X3[i - 1] < 0
last_month_wet = X3[i - 1] > 0
#Moisture needed to end drought or wet spell:
Ze_drought = -2.691 * X3[i - 1] - 1.5
Ze_wet_spell = -2.691 * X3[i - 1] + 1.5
Ze = MV.where(last_month_dry, Ze_drought, 0)
Ze = MV.where(last_month_wet, Ze_wet_spell, Ze)
#Probability that last month's wet or dry spell will end
P_end_wet = end_spell_P(Ud, Ze_wet_spell, nmonths, i, typ="wet")
P_end_dry = end_spell_P(Uw, Ze_drought, nmonths, i, typ="dry")
P_end_spell = MV.where(last_month_dry, P_end_dry, 0)
P_end_spell = MV.where(last_month_wet, P_end_wet, P_end_spell)
Pe[i] = P_end_spell
# If this was a wet month
condition = this_month_wet
#If last month was in a dry spell
condition1 = np.ma.logical_and(condition, last_month_dry)
#If the dry period is definitely ending
dry_definitely_ending = P_end_dry >= 100
condition2 = np.ma.logical_and(condition1, dry_definitely_ending)
#PDSI = wet transitional value
Xi = MV.where(condition2, X1i, 0)
# If wet transitional value indicates very wet conditions
very_wet = X1i > 1
# set X3 to wet transitional value
condition3 = np.logical_and(condition2, very_wet)
X3i = MV.where(condition3, X1i, 0)
X1i = MV.where(condition3, 0, X1i)
#Log beginning of wet spell
nmonths = MV.where(condition3, 1, 0)
#If the dry period is definitely NOT ending
dry_definitely_NOT_ending = P_end_dry <= 0
condition2 = np.ma.logical_and(condition1, dry_definitely_NOT_ending)
X2i = MV.where(condition2, 0, X2i)
#If last month was in a wet spell
condition1 = np.ma.logical_and(condition, last_month_wet)
#If the wet period is definitely ending
wet_definitely_ending = P_end_wet >= 100
condition2 = np.ma.logical_and(condition1, wet_definitely_ending)
#PDSI = dry transitional value
Xi = MV.where(condition2, X2i, Xi)
# If dry transitional value indicates very dry conditions
very_dry = X2i <= -1
# set X3 to dry transitional value
condition3 = np.logical_and(condition2, very_dry)
X3i = MV.where(condition3, X2i, X3i)
X2i = MV.where(condition3, 0, X1i)
#Log beginning of dry spell
nmonths = MV.where(condition3, 1, nmonths)
#If the wet spell is definitely NOT ending
wet_definitely_NOT_ending = P_end_wet <= 0
condition2 = np.ma.logical_and(condition1, wet_definitely_NOT_ending)
X1i = MV.where(condition2, 0, X1i)
#if last month was neutral but this month has a positive Z:
condition1 = np.ma.logical_and(condition, last_month_neutral)
wet_spell_establishing = X1i > 0.5
condition2 = np.logical_and(condition1, wet_spell_establishing)
#PSDI = wet transitional value
Xi = MV.where(condition2, X1i, Xi)
#Begin wet spell
nmonths = MV.where(condition2, 1, nmonths)
#if wet spell is atarting off strong
wet_strong = X1i >= 1
condition3 = np.logical_and(condition2, wet_strong)
#Set wet persistence value to wet transitional value
X3i = MV.where(condition3, X1i, X3i)
X1i = MV.where(condition3, 0, X1i)
#If this was a dry month
condition = this_month_dry
#If last month was in a wet spell
condition1 = np.ma.logical_and(condition, last_month_wet)
#If the wet period is definitely ending
wet_definitely_ending = P_end_wet >= 100
condition2 = np.ma.logical_and(condition1, wet_definitely_ending)
#PDSI = dry transitional value
Xi = MV.where(condition2, X2i, Xi)
# If dry transitional value indicates very dry conditions
very_dry = X2i <= -1
# set X3 to dry transitional value
condition3 = np.logical_and(condition2, very_dry)
X3i = MV.where(condition3, X2i, X3i)
X2i = MV.where(condition3, 0, X1i)
#Log beginning of dry spell
nmonths = MV.where(condition3, 1, nmonths)
#If the wet spell is definitely NOT ending
wet_definitely_NOT_ending = P_end_wet <= 0
condition2 = np.ma.logical_and(condition1, wet_definitely_NOT_ending)
X1i = MV.where(condition2, 0, X1i)
#If last month was in a dry spell
condition1 = np.ma.logical_and(condition, last_month_dry)
#If the dry period is definitely ending
dry_definitely_ending = P_end_dry >= 100
condition2 = np.ma.logical_and(condition1, dry_definitely_ending)
#PDSI = wet transitional value
Xi = MV.where(condition2, X1i, 0)
# If wet transitional value indicates very wet conditions
very_wet = X1i > 1
# set X3 to wet transitional value
condition3 = np.logical_and(condition2, very_wet)
X3i = MV.where(condition3, X1i, 0)
X1i = MV.where(condition3, 0, X1i)
#Log beginning of wet spell
nmonths = MV.where(condition3, 1, 0)
#If the dry period is definitely NOT ending
dry_definitely_NOT_ending = P_end_dry <= 0
condition2 = np.ma.logical_and(condition1, dry_definitely_NOT_ending)
X2i = MV.where(condition2, 0, X2i)
#if last month was neutral but this month has a negative Z:
condition1 = np.ma.logical_and(condition, last_month_neutral)
dry_spell_establishing = X2i <= 0.5
condition2 = np.logical_and(condition1, dry_spell_establishing)
#PSDI = dry transitional value
Xi = MV.where(condition2, X2i, Xi)
#Begin dry spell
nmonths = MV.where(condition2, 1, nmonths)
#if dry spell is atarting off strong
dry_strong = X2i <= -1
condition3 = np.logical_and(condition2, dry_strong)
#Set dry persistence value to dry transitional value
X3i = MV.where(condition3, X2i, X3i)
X2i = MV.where(condition3, 0, X2i)
X1i = MV.where(X1i < 0, 0, X1i)
X2i = MV.where(X2i > 0, 0, X2i)
#if the probability the spell has ended is not 100% and it wasn't a neutral month
condition = P_end_spell < 100
X3i = MV.where(condition, X3[i - 1] * .897 + XX, X3i)
#If the sign of X3 didn't change over the time step
sign_same = np.sign(X3[i - i]) == np.sign(X3i)
condition1 = np.ma.logical_and(condition, sign_same)
#If X3 is still extreme
extreme = MV.absolute(X3i) > 0.5
condition2 = np.ma.logical_and(condition1, extreme)
#The spell continues
nmonths = MV.where(condition2, nmonths + 1, nmonths)
#Otherwise, the spell ends
condition2 = np.ma.logical_and(condition1, ~extreme)
nmonths = MV.where(condition2, 0, nmonths)
#If the sign did change
sign_different = ~sign_same
condition1 = np.ma.logical_and(condition, sign_different)
#IF the X3 anomaly is large
very_extreme = MV.absolute(X3i) >= 1
condition2 = np.ma.logical_and(condition1, very_extreme)
#new dry or wet spell begins
nmonths = MV.where(condition2, 1, nmonths)
#otherwise neutral conditions continue
condition2 = np.ma.logical_and(condition1, ~very_extreme)
nmonths = MV.where(condition2, 0, nmonths)
montho[i] = nmonths
#Decide what PDSI is
#If there is no event established or in progress
no_event = nmonths == 0
tending_toward_wet = X1i > -X2i
wet_condition = np.logical_and(no_event, tending_toward_wet)
Xi = MV.where(wet_condition, X1i, Xi)
tending_toward_dry = X1i <= -X2i
dry_condition = np.logical_and(no_event, tending_toward_dry)
Xi = MV.where(dry_condition, X2i, Xi)
#If a wet or dry spell has begun
event_begun = nmonths == 1
began_last_month = X3i == 0
condition = np.logical_and(event_begun, began_last_month)
condition1 = np.logical_and(condition, tending_toward_wet)
Xi = MV.where(condition1, X1i, Xi)
condition1 = np.logical_and(condition, tending_toward_dry)
Xi = MV.where(condition1, X2i, Xi)
began_this_month = X3i != 0
condition = np.logical_and(event_begun, began_this_month)
pdsi_not_assigned = Xi == 0
Xi = MV.where(np.logical_and(pdsi_not_assigned, condition), X3i, Xi)
#If beyond the first month of a wet or dry spell
Xi = MV.where(nmonths > 1, X3i, Xi)
X[i] = Xi
X1[i] = X1i
X2[i] = X2i
X3[i] = X3i
#BACKTRACK
ending_or_confirmed = np.logical_or(P_end_spell >= 100,
P_end_spell <= 0)
condition = np.logical_and(montho[i] == 1, ending_or_confirmed)
s_orig = i
#Find the month when the previous wet or dry spell ended
Pe_trunc = Pe[:s_orig]
months_previous = Pe_trunc.shape[0]
Pe_relevant = MV.where(
np.repeat(condition.asma()[np.newaxis], months_previous, axis=0),
Pe_trunc, 0)
#HACKY PYTHON REPLACEMENT FOR MATLAB FIND LAST
I, J, K = np.ma.where(Pe_relevant >= 100)
datapoints = zip(
J, K
) #spatial grid points at which the previous wet or dry spell ended
#need to find the most recent month of end for each grid point
stringdata = [str(x) for x in datapoints]
unique_points = np.unique(stringdata)
for point in unique_points:
lat, lon = ast.literal_eval(point)
index = np.max(np.where(np.array(stringdata) == point)
[0]) #find the last place at which the point occurs
s = I[index]
#Find months prior to the 100% transition where the probability of the termination of the previous spell was above the BTthresh probability
r = s
if (Pe[s - 1, lat, lon] > 0 and Pe[s - 1, lat, lon] < 100):
for c in np.arange(1, s)[::-1]:
if (Pe[c, lat, lon] > BTthresh and Pe[c, lat, lon] < 100):
r = c
else:
break
#extend backtracking period forward from the month when the dry or wet spell had a 100% of terminating to the month
#when the next dry or wet spell officially began
F = np.arange(r, s)
if len(F) == 0:
F = np.array([r])
if s < s_orig:
F = np.arange(F[0], s_orig + 1)
#determine whether backtracking is occuring as the result of the initiation of a wet or dry spell
m = np.min(np.where(montho[s:s_orig + 1, lat, lon] == 1)[0])
if X[s + m, lat, lon] > 0:
pon = 0 #wet spell
else:
pon = 1 #dry spell
#Backtrack and assign new PDSI values
possible_PDSI = [X1[F, lat, lon][0], X2[F, lat, lon][0]]
the_month = len(F) - 1
last_month_pon = 0
ponflip = lambda thing: 1 if thing is 0 else 0
while the_month >= 0:
#if next month is in a dry or wet phase, assign this month the transitional PDSI
# value associated with that phase as long as it is non-zero
if possible_PDSI[pon][the_month] != 0:
X[F[the_month], lat, lon] = possible_PDSI[pon][the_month]
else:
#if transitional value for next month's phase is zero, assign
# this month to the other transitional phase
X[F[the_month], lat,
lon] = possible_PDSI[ponflip(pon)][the_month]
pon = ponflip(pon)
#if both transitional PDSI values were zero
if X[F[the_month], lat, lon] == 0:
the_X3 = X3[F[the_month], lat, lon]
t = np.sign(the_X3)
especially_strong = np.abs(the_X3) >= 1
consistent_with_next_month_phase = (
1 - t) / 2 == last_month_pon
if (especially_strong or consistent_with_next_month_phase):
#Assign X3 as this month's PDSI value
X[F[the_month], lat, lon] = the_X3
if X[F[the_month], lat, lon] > 0:
pon = 0
else:
pon = 1
#if X3 doesn't satisfy the above conditions, leave PDSI =0
else:
if last_month_pon != 0:
pon = last_month_pon #assign this month to the dry or wet phase that eventually develops
# indicate whether backtracking assigned this month to a wet or dry phase
BT[F[the_month]] = pon
last_month_pon = pon
the_month = the_month - 1
#recalculate running tally of months in a dry or wet spell
nmonths[lat, lon] = montho[F[0] - 1, lat, lon]
if ((the_month < len(F) - 1) and (F[-1] < Z.shape[0])):
F = np.append(F[the_month:], F[-1] + 1)
for newj in range(len(F)):
if np.abs(X[F[newj], lat, lon]) > 1:
if np.sign(X[F[newj], lat,
lon]) == np.sign(X[F[newj] - 1, lat,
lon]):
nmonths[lat, lon] = nmonths[lat, lon] + 1
else:
nmonths[lat, lon] = 1
else:
nmonths[lat, lon] = 0
montho[F[newj], lat, lon] = nmonths[lat, lon]
PDSI = X
PDSI.id = "pdsi"
PDSI.setAxisList(Z.getAxisList())
return PDSI
def testNegAbs(self):
x11 = -self.other_u_file
x12 = MV2.absolute(self.u_file)
self.assertTrue(MV2.allequal(x11 + x12[0], 0))
ulat = ud.getLatitude()
if not isinstance(udat, TV): markError('Slice does not return TV')
f = cdms2.open(os.path.join(get_sample_data_dir(),'u_2000.nc'))
uf = f['u']
vel = MV2.sqrt(ud*ud + vd*vd)
vel.id = 'velocity'
vel2 = MV2.sqrt(udat*udat + vdat*vdat)
vel2.id = 'velocity'
vel2.units = ud.units
if not MV2.allequal(vel,vel2): markError('Slice operators do not compare')
x1 = uf+1.0
x2 = 1.0-ud
x11 = -uf
x12 = MV2.absolute(ud)
x3 = uf+x2
x4 = 1.0+ud
x5 = uf-1
x6 = ud*uf
x7 = ud/x2
x8=1/uf
x9 = 3*ud
x10=uf**3
x13 = MV2.add.reduce(uf)
x14 = MV2.add.reduce(ud)
x15 = x9.astype(numpy.float32)
if not x15.dtype.char==numpy.sctype2char(numpy.float32): markError('astype error')
## arrayrange(start, stop=None, step=1, typecode=None, axis=None, attributes=None, id=None)
## Just like range() except it returns a variable whose type can be specfied
def calculate_Z(PET, P, WCTOP, WCBOT, year1, year2):
"""
Written by Kate Marvel. Adapted from code by Park Williams (LDEO).
Calculates the Palmer z-index
Inputs:
PET: Potential evapotranspiration (mm)
PR: Precipitation (mm). NOTE CMIP5 standard is kg/m s-2 so need to run convert_to_mm on CMIP5 output
WCBOT: Soil moisture holding capacity in bottom layer (mm)
WCTOP: Water holding capacity in top layer (mm)
year1: start date of calibration period
year2: stop date of calibration period
Outputs:
PL: Potential loss
ET: Actual evapotranspiration
TL: Total loss
RO: Runoff
R: Recharge
SSS: Updated surface soil moisture
SSU: Updated moisture of underlying soil layers
"""
#INITIALIZE VARS
WCTOT = WCBOT + WCTOP # Total water holding capacity of the soil layers
SS = WCTOP # Surface soil moisture start at full capacity
SU = WCBOT # Underlying layer soil moisture start at full capacity
SP = SS + SU # Combined surface and underlying soil moisture
#Get arrays for output
pldat = MV.zeros(PET.shape)
spdat = MV.zeros(PET.shape)
prdat = MV.zeros(PET.shape)
rdat = MV.zeros(PET.shape)
tldat = MV.zeros(PET.shape)
etdat = MV.zeros(PET.shape)
rodat = MV.zeros(PET.shape)
sssdat = MV.zeros(PET.shape)
ssudat = MV.zeros(PET.shape)
#get rid of badly calibrated negative PET
PET = MV.where(PET < 0, 0, PET)
#fake 10 year spinup
PET_ex = pad_by_10(PET, year1, year2)
P_ex = pad_by_10(P, year1, year2)
#get rid of badly calibrated negative PET
PET = MV.where(PET < 0, 0, PET)
#run 2-layer bucket model
for i in range(PET_ex.shape[0]):
PR = WCTOT - SP
PL, ET, TL, RO, R, SSS, SSU = bucket2d(PET_ex[i], P_ex[i], WCBOT,
WCTOP, SS, SU)
ET = MV.where(ET < 0, 0, ET)
SS = SSS
SU = SSU
SP = SS + SU
if i >= 120:
pldat[i - 120] = PL
spdat[i - 120] = SS + SU
prdat[i - 120] = PR
rdat[i - 120] = R
tldat[i - 120] = TL
etdat[i - 120] = ET
rodat[i - 120] = RO
sssdat[i - 120] = SSS
ssudat[i - 120] = SSU
rdat = MV.where(rdat >= prdat, prdat, rdat)
tldat = MV.where(tldat >= pldat, pldat, tldat)
for X in [spdat, pldat, prdat, rdat, tldat, etdat, rodat]:
X.setAxisList(PET.getAxisList())
#calculate means over calibration period
SPSUM = cdutil.ANNUALCYCLE.climatology(spdat(time=(year1, year2)))
PLSUM = cdutil.ANNUALCYCLE.climatology(pldat(time=(year1, year2)))
PRSUM = cdutil.ANNUALCYCLE.climatology(prdat(time=(year1, year2)))
RSUM = cdutil.ANNUALCYCLE.climatology(rdat(time=(year1, year2)))
TLSUM = cdutil.ANNUALCYCLE.climatology(tldat(time=(year1, year2)))
ETSUM = cdutil.ANNUALCYCLE.climatology(etdat(time=(year1, year2)))
PESUM = cdutil.ANNUALCYCLE.climatology(PET(time=(year1, year2)))
ROSUM = cdutil.ANNUALCYCLE.climatology(rodat(time=(year1, year2)))
PSUM = cdutil.ANNUALCYCLE.climatology(P(time=(year1, year2)))
# CAFEC: Climatology Appropriate for Existing Conditions (Palmer 1965, p 12)
# Calculate the CAFEC coefficients
#Coefficient of evaporation: fraction of mean ET to mean potential ET
alpha = pdsi_coeff(ETSUM, PESUM)
#Coef of Recharge: ratio of mean recharge to mean potential recharge
beta = pdsi_coeff(RSUM, PRSUM)
#Coefficient of Runoff: Ratio of mean runoff to mean potential runoff
gamma = pdsi_coeff(ROSUM, SPSUM)
#Coefficient of Loss: Ratio of mean moisture loss to mean potential moisture loss
delta = pdsi_coeff(TLSUM, PLSUM)
TRAT = (PESUM + RSUM + ROSUM) / (PSUM + TLSUM)
#CAFEC precipitation (needed to maintain "normal" moisture)
nyears = PET.shape[0] / 12
repeat_it = lambda coef: np.ma.repeat(coef, nyears, axis=0)
Phat = PET * repeat_it(alpha) + prdat * repeat_it(
beta) + spdat * repeat_it(gamma) - pldat * repeat_it(delta)
# Moisture departure (difference between precip and precip needed for normal conditions)
DD = P - Phat
DD.setAxisList(P.getAxisList())
SABSD = MV.absolute(DD)
DBAR = cdutil.ANNUALCYCLE.climatology(SABSD(time=(year1, year2)))
# Weird empirical scaling thing to standardize moisture availability departures. THIS IS DUMB.
AKHAT = 1.5 * MV.log10((TRAT + 2.8 * 25.4) / DBAR) + 0.5
AKHAT = MV.where(AKHAT < 0, 0, AKHAT)
annsum = MV.sum(AKHAT * DBAR, axis=0)
AK = 17.67 * 25.4 * AKHAT / annsum
Z = DD / 25.4 * repeat_it(AK)
#Force Z between -16 and 16
Z = MV.where(Z > 16, 16, Z)
Z = MV.where(Z < -16, -16, Z)
Z.setAxisList(P.getAxisList())
Z.id = "z"
Z.units = "mm"
Z.info = "Created by Kate Marvel using calc_Z.py"
Z.creation_date = datetime.datetime.now().isoformat()
Z.long_name = "Palmer Z index"
Z.standard_name = "z_index"
return Z