def inertial_period(lat):
"""Calculate the inertial period as:
.. math::
Ti = \\frac{2\\pi}{f} = \\frac{T_{sd}}{2\\sin\\phi}
Parameters
----------
lat : array_like
latitude in decimal degrees north [-90..+90]
Returns
-------
Ti : array_like
period in seconds
Examples
--------
>>> from oceans import sw_extras as swe
>>> lat = 30.
>>> swe.inertial_period(lat)/3600
23.934849862785651
"""
lat = np.asanyarray(lat)
return 2 * np.pi / sw.f(lat)
def inertial_period(lat):
"""
Calculate the inertial period as:
.. math::
Ti = \\frac{2\\pi}{f} = \\frac{T_{sd}}{2\\sin\\phi}
Parameters
----------
lat : array_like
latitude in decimal degrees north [-90..+90]
Returns
-------
Ti : array_like
period in seconds
Examples
--------
>>> from oceans import sw_extras as swe
>>> lat = 30.
>>> swe.inertial_period(lat)/3600
23.934849862785651
"""
lat = np.asanyarray(lat)
return 2 * np.pi / sw.f(lat)
def qg_stability(N2, ubar, z, k=1., lat=50., structure=True):
f0 = sw.f(lat)
beta = 2 * 7.29e-5 * np.cos(lat * pi / 180.) / 6371.e3
dz = np.abs(z[1] - z[0])
S = (f0**2) / N2
L = stretching_matrix(z.size, S, dz)
k2 = k**2
L2 = L - np.eye(ubar.size) * k2
U = np.eye(ubar.size) * ubar
Qy = np.eye(
ubar.size) * (beta - np.array(np.matrix(L) * np.matrix(ubar).T))
L3 = L2.copy()
for i in range(ubar.size):
L3[i, :] = L2[i, :] * ubar[i]
A = L3 + Qy
B = L2.copy()
evals, evecs = sp.linalg.eig(A, B)
imax = evals.imag.argmax()
eval_max = evals[imax]
evec_max = evecs[:, imax]
if structure:
PSI, X, Z = wave_structure(evec_max, z, k)
return eval_max, evec_max, PSI, X, Z
else:
return eval_max, evec_max
def qg_stability(N2,ubar,z,k=1.,lat=50.,structure=True):
f0 = sw.f(lat)
beta = 2*7.29e-5*np.cos(lat*pi/180.)/6371.e3
dz = np.abs(z[1]-z[0])
S = (f0**2)/N2
L= stretching_matrix(z.size,S,dz)
k2 = k**2
L2 = L - np.eye(ubar.size)*k2
U = np.eye(ubar.size)*ubar
Qy = np.eye(ubar.size)*(beta - np.array( np.matrix(L)*np.matrix(ubar).T ) )
L3 = L2.copy()
for i in range(ubar.size):
L3[i,:] = L2[i,:]*ubar[i]
A = L3 + Qy
B = L2.copy()
evals,evecs = sp.linalg.eig(A,B)
imax = evals.imag.argmax()
eval_max = evals[imax]
evec_max = evecs[:,imax]
if structure:
PSI, X, Z = wave_structure(evec_max,z,k)
return eval_max,evec_max,PSI,X,Z
else:
return eval_max,evec_max
def set_params(lat, dt=3., dz=1., max_depth=100., mld_thresh=1e-4, dt_save=1., rb=0.65, rg=0.25, rkz=0., beta1=0.6, beta2=20.0, heat_ON=True, winds_ON=True, emp_ON=True, drag_ON=True):
"""
This function sets the main paramaters/constants used in the model.
These values are packaged into a dictionary, which is returned as output.
Definitions are listed below.
CONTROLS (default values are in [ ]):
lat: latitude of profile
dt: time-step increment. Input value in units of hours, but this is immediately converted to seconds.[3 hours]
dz: depth increment (meters). [1m]
max_depth: Max depth of vertical coordinate (meters). [100]
mld_thresh: Density criterion for MLD (kg/m3). [1e-4]
dt_save: time-step increment for saving to file (multiples of dt). [1]
rb: critical bulk richardson number. [0.65]
rg: critical gradient richardson number. [0.25]
rkz: background vertical diffusion (m**2/s). [0.]
beta1: longwave extinction coefficient (meters). [0.6]
beta2: shortwave extinction coefficient (meters). [20]
winds_ON: True/False flag to turn ON/OFF wind forcing. [True]
emp_ON: True/False flag to turn ON/OFF freshwater forcing. [True]
heat_ON: True/False flag to turn ON/OFF surface heat flux forcing. [True]
drag_ON: True/False flag to turn ON/OFF current drag due to internal-inertial wave breaking. [True]
OUTPUT is dict with fields containing the above variables plus the following:
dt_d: time increment (dt) in units of days
g: acceleration due to gravity [9.8 m/s^2]
cpw: specific heat of water [4183.3 J/kgC]
f: coriolis term (rad/s). [sw.f(lat)]
ucon: coefficient of inertial-internal wave dissipation (s^-1) [0.1*np.abs(f)]
"""
params = {}
params['dt'] = 3600.0*dt
params['dt_d'] = params['dt']/86400.
params['dz'] = dz
params['dt_save'] = dt_save
params['lat'] = lat
params['rb'] = rb
params['rg'] = rg
params['rkz'] = rkz
params['beta1'] = beta1
params['beta2'] = beta2
params['max_depth'] = max_depth
params['g'] = 9.81
params['f'] = sw.f(lat)
params['cpw'] = 4183.3
params['ucon'] = (0.1*np.abs(params['f']))
params['mld_thresh'] = mld_thresh
params['winds_ON'] = winds_ON
params['emp_ON'] = emp_ON
params['heat_ON'] = heat_ON
params['drag_ON'] = drag_ON
return params
def set_params(dt=3., dz=1., max_depth=100., mld_thresh=1e-4, lat=74., dt_save=1., rb=0.65, rg=0.25, rkz=0., beta1=0.6, beta2=0.2):
"""
This function sets the main paramaters/constants used in the model.
These values are packaged into a dictionary, which is returned as output.
Definitions are listed below.
CONTROLS (default values are in [ ]):
dt: time-step increment. Input value in units of hours, but this is immediately converted to seconds.[3 hours]
dz: depth increment (meters). [1m]
max_depth: Max depth of vertical coordinate (meters). [100]
mld_thresh: Density criterion for MLD (kg/m3). [1e-4]
dt_save: time-step increment for saving to file (multiples of dt). [1]
lat: latitude of profile (degrees). [74.0]
rb: critical bulk richardson number. [0.65]
rg: critical gradient richardson number. [0.25]
rkz: background vertical diffusion (m**2/s). [0.]
beta1: longwave extinction coefficient (meters). [0.6]
beta2: shortwave extinction coefficient (meters). [20]
OUTPUT is dict with fields containing the above variables plus the following:
dt_d: time increment (dt) in units of days
g: acceleration due to gravity [9.8 m/s^2]
cpw: specific heat of water [4183.3 J/kgC]
f: coriolis term (rad/s). [sw.f(lat)]
ucon: coefficient of inertial-internal wave dissipation (s^-1) [0.1*np.abs(f)]
"""
params = {}
params['dt'] = 3600.0*dt
params['dt_d'] = params['dt']/86400.
params['dz'] = dz
params['dt_save'] = dt_save
params['lat'] = lat
params['rb'] = rb
params['rg'] = rg
params['rkz'] = rkz
params['beta1'] = beta1
params['beta2'] = beta2
params['max_depth'] = max_depth
params['g'] = 9.81
params['f'] = sw.f(lat)
params['cpw'] = 4183.3
params['ucon'] = (0.1*np.abs(params['f']))
params['mld_thresh'] = mld_thresh
return params
def __init__(self): # Beta = meridional rate of change of the Coriolis parameters (m^-1 s^-1) r = 6.371e+6 # Earth's radius (m) omega = 7.292115e-5 # Earth's rotation rate (s^-1) beta = 2.*omega*np.cos(np.pi*soest.LAT16_100/180.)/r # Coriolis parameter (s^-1) f = sw.f(soest.LAT16_100) # Rho = typical density of the sea water above the pycnocline (kg m^-3) # Since we are considering a incompressible ocean (i. e., Boussinesq approximation) it will be # constant rho = 1025.0 self.beta = beta self.rho = rho self.f = f
def __init__(self):
# Beta = meridional rate of change of the Coriolis parameters (m^-1 s^-1)
r = 6.371e+6 # Earth's radius (m)
omega = 7.292115e-5 # Earth's rotation rate (s^-1)
beta = 2. * omega * np.cos(np.pi * scow.latitude / 180.) / r
# Coriolis parameter (s^-1)
f = sw.f(scow.latitude)
# Rho = typical density of the sea water above the pycnocline (kg m^-3)
# Since we are considering a incompressible ocean (i. e., Boussinesq approximation) it will be
# constant
rho = 1025.0
self.beta = beta
self.rho = rho
self.f = f
def GM(lat, N, N0, b=1000, oned=False):
try:
import GM81.gm as gm
except ImportError:
raise ImportError("Please install the GM81 package.")
# Coriolis frequency
f = sw.f(lat=12)
# frequency
omg = np.logspace(np.log10(1.01 * f), np.log10(N), 401)
# horizontal wavenumber
k = 2 * np.pi * np.logspace(-6, -2, 401)
# mode number
j = np.arange(1, 100)
# reshape to allow multiplication into 2D array
Omg = np.reshape(omg, (omg.size, 1))
K = np.reshape(k, (k.size, 1))
J = np.reshape(j, (1, j.size))
# frequency spectra (KE and PE)
K_omg_j = gm.K_omg_j(Omg, J, f, N, N0, b)
P_omg_j = gm.P_omg_j(Omg, J, f, N, N0, b)
# wavenumber spectra (KE and PE)
K_k_j = gm.K_k_j(K, J, f, N, N0, b)
P_k_j = gm.P_k_j(K, J, f, N, N0, b)
# sum over modes
K_omg = np.sum(K_omg_j, axis=1)
P_omg = np.sum(P_omg_j, axis=1)
K_k = np.sum(K_k_j, axis=1)
P_k = np.sum(P_k_j, axis=1)
# compute 1D spectra from 2D spectra
K_k_1d = gm.calc_1d(k, K_k)
P_k_1d = gm.calc_1d(k, P_k)
return (omg, K_omg, P_omg, k, K_k_1d, P_k_1d)
def qg_stability_2d(N2, ubar, vbar, z, k, l, lat, structure=False):
f0 = sw.f(lat)
beta = 2 * 7.29e-5 * np.cos(lat * pi / 180.) / 6371.e3
dz = np.abs(z[1] - z[0])
S = (f0**2) / N2
L = stretching_matrix(z.size, S, dz)
k2 = k**2 + l**2
L2 = L - np.eye(ubar.size) * k2
Qy = np.eye(
ubar.size) * (beta - np.array(np.matrix(L) * np.matrix(ubar).T))
Qx = np.eye(ubar.size) * np.array(np.matrix(L) * np.matrix(vbar).T)
Q = k * Qy - l * Qx
L3 = np.empty_like(L2)
for i in range(ubar.size):
L3[i, :] = L2[i, :] * (ubar[i] * k + vbar[i] * l)
A = (L3 + Q)
B = L2.copy()
try:
evals, evecs = sp.linalg.eig(A, B)
imax = evals.imag.argmax()
eval_max = evals[imax]
evec_max = evecs[:, imax]
except:
eval_max = np.nan + 1.j * np.nan
evec_max = np.nan * z
if structure:
PSI, X, Z = wave_structure(evec_max, z, k)
return eval_max, evec_max, PSI, X, Z
else:
return eval_max, evec_max
def qg_stability_2d(N2,ubar,vbar,z,k,l,lat,structure=False):
f0 = sw.f(lat)
beta = 2*7.29e-5*np.cos(lat*pi/180.)/6371.e3
dz = np.abs(z[1]-z[0])
S = (f0**2)/N2
L= stretching_matrix(z.size,S,dz)
k2 = k**2 + l**2
L2 = L - np.eye(ubar.size)*k2
Qy = np.eye(ubar.size)*(beta - np.array( np.matrix(L)*np.matrix(ubar).T ))
Qx = np.eye(ubar.size)*np.array( np.matrix(L)*np.matrix(vbar).T )
Q = k*Qy - l*Qx
L3 = np.empty_like(L2)
for i in range(ubar.size):
L3[i,:] = L2[i,:]*(ubar[i]*k + vbar[i]*l)
A = (L3 + Q)
B = L2.copy()
try:
evals,evecs = sp.linalg.eig(A,B)
imax = evals.imag.argmax()
eval_max = evals[imax]
evec_max = evecs[:,imax]
except:
eval_max = np.nan + 1.j*np.nan
evec_max = np.nan*z
if structure:
PSI, X, Z = wave_structure(evec_max,z,k)
return eval_max,evec_max,PSI,X,Z
else:
return eval_max,evec_max
def geostrophic_current(ix, lat):
g = sw.g(lat.mean())
f = sw.f(lat.mean())
v = ix * g / f
return v
def eqmodes(N2, z, nm, pmodes=False):
'''
This function computes the equatorial velocity modes
========================================================
Input:
N2 - Brunt-Vaisala frequency data array
z - Depth data array (equaly spaced)
lat - Latitude scalar
nm - Number of modes to be computed
pmodes - If the return of pressure modes is required.
Default is False
========================================================
Output:
Si - Equatorial modes matrix with MxN dimension being:
M = z array size
N = nm
Rdi - Deformation Radii array
Fi - Pressure modes matrix with MxN dimension being:
M = z array size
N = nm
Returned only if input pmodes=True
made by Hélio Almeida, Iury Sousa and Wandrey Watanabe
Laboratório de Dinâmica Oceânica - Universidade de São Paulo
2016
'''
#needed to the problem
lat = 0
#nm will be the number of baroclinic modes
nm -= 1 #Barotropic mode will be added
#defines the orthonormalization function
onorm = lambda f: f/np.sqrt(dz*((np.array(f[1:])**2+\
np.array(f[:-1])**2)/(2*H)).sum())
# defines function to keep consistency multiply by plus/minus
# to make the leading term positive
plus_minus = lambda f: -f if (np.sign(f[1]) < 0) else f
dz = np.abs(z[1] - z[0])
# assembling matrices
# N2 matrix
N2 = np.diag(N2[1:-1], 0)
# 2nd order difference matrix
A = np.diag(np.ones(z.size-2)*-2.,0)+\
np.diag(np.ones(z.size-3)*1.,-1)+\
np.diag(np.ones(z.size-3)*1.,1)
A = A / (dz**2)
A = np.matrix(A)
N2 = np.matrix(N2)
#C = A*N2
N02 = -1 / N2
N02[np.isinf(N02)] = 0
C = N02 * A
# solve the eigenvalue problem
egval, egvec = np.linalg.eig(C)
i = np.argsort(egval)
egval = egval[i]
egvec = egvec[:, i]
# truncate the eigenvalues and eigenvectors for the number of modes needed
ei = egval[:nm]
Si = egvec[:, :nm]
# Applying Dirichlet boundary condition at top and bottom
# adding a row of zeros at bottom and top
Si = np.append(np.matrix(np.zeros(nm)), Si, axis=0)
Si = np.append(Si, np.matrix(np.zeros(nm)), axis=0)
Si = np.array(Si)
H = np.abs(z).max()
# normalizing to get orthonormal modes
Si = np.array(list(map(onorm, Si.T))).T
# to keep consistency multiply by plus/minus
# to make the leading term positive
Si = np.array(list(map(plus_minus, Si.T))).T
# compute the deformation radii [km]
beta = (7.2921150e-5 * 2 * np.cos(np.deg2rad(lat * 1.))) / 6371000
c = np.sqrt(1 / ei)
radii = np.sqrt(c / beta)
#append external deformation radius
radii = np.hstack([(np.sqrt(9.81 * H) / np.abs(sw.f(lat))), radii])
#converting to km
radii *= 1e-3
#BAROTROPIC MODE
no = 1
fb = np.ones((Si.shape[0], 1)) * no
sb = np.expand_dims(np.linspace(0, no + 1, Si.shape[0]), axis=1)
# trying to compute the pmodes based on the polarization
# relation between velocity/pressure modes
#
# F_j(z)=-g.he_j d/dzS_j
#
####
if pmodes == True:
Fi = np.zeros(Si.shape)
for i in np.arange(nm):
Fi[1:-1,i] =\
(-1/ei[i])*((Si[1:-1,i]-Si[0:-2,i])/dz)
#Aplying Neuman boundary condition d/dzFi=o @0,-H
Fi[0], Fi[-1] = Fi[1], Fi[-2]
# normalizing to get orthonormal modes
Fi = np.array(list(map(onorm, Fi.T))).T
# to keep consistency multiply by plus/minus
# to make the leading term positive
Fi = np.array(list(map(plus_minus, Fi.T))).T
Si = np.hstack([sb, Si])
Fi = np.hstack([fb, Fi])
return Si, radii, Fi
else:
Si = np.hstack([sb, Si])
return Si, radii
def model_timestep(T,
S,
U,
V,
z,
I,
L,
E,
P,
tau_x,
tau_y,
dt,
nabla_b=None,
Ekman_Q_flux=None,
use_static_stability=True,
use_mixed_layer_stability=True,
use_shear_stability=True,
use_Ekman_flux=False,
use_MLI=False,
tracer=None,
vert_diffusivity=None,
verbose=False,
I1=0.62,
I2=None,
lambda1=.6,
lambda2=20,
T0=0,
S0=34,
rho0=None,
alpha=None,
beta=None,
f=sw.f(40),
return_MLD=False,
advection=False,
l_poly=1e4,
phi=5e-2,
T_cdw=1,
S_cdw=34.5,
shelf_thick=350,
debug=False,
T_out=-1,
S_out=33.8,
return_dTdS=False,
return_TSrelax=False,
return_vel=False):
# define initial variables
c = 4218 # heat capacity (J kg^-1 K^-1)
if I2 is None:
I2 = 1 - I1
if I1 + I2 != 1:
raise Exception('Shortwave insolation amplitudes need to sum to unity')
if rho0 is None:
rho0 = sw.dens(S0, T0, 0)
if alpha is None:
alpha = -sw.alpha(
S0, T0, 0
) * rho0 # multiply by rho to fit into nice equation of state (see get_rho function)
if beta is None:
beta = sw.beta(
S0, T0, 0
) * rho0 # # multiply by rho to fit into nice equation of state (see get_rho function)
dz = z[1] - z[0]
if use_Ekman_flux and Ekman_Q_flux is None:
raise Exception(
'Using Ekman-induced buoyacy flux but no buoyancy gradients were given.'
)
if use_MLI and MLI_flux is None:
raise Exception('Using MLI but no horizontal buoyancy gradient given.')
T = T.copy()
S = S.copy()
U = U.copy()
V = V.copy()
# so don't overwrite data
if tracer is not None:
tracer = tracer.copy()
# make initial heat profile
I_profile = -I / dz * (
I1 * np.exp(-z / lambda1) *
(np.exp(-dz / 2 / lambda1) - np.exp(dz / 2 / lambda1)) +
I2 * np.exp(-z / lambda2) *
(np.exp(-dz / 2 / lambda2) - np.exp(dz / 2 / lambda2)))
L_profile = np.zeros(len(z))
L_profile[0] = L / dz
Q_profile = I_profile + L_profile
if use_Ekman_flux:
A = 0.1 # eddy viscosity m^2 s^-1
z_Ek = np.sqrt(A / np.abs(f))
if verbose:
print('Using Ekman depth of %d m' % z_Ek)
z_Ek_ind = np.where(z > z_Ek)[0][0]
Q_Ek_profile = np.zeros(len(z))
Q_Ek_profile[0:z_Ek_ind] = Ekman_Q_flux / z_Ek * dz
Q_profile += Q_Ek_profile
if advection == True:
Tf = sw.fp(S, z)
#freezing temperature for the column using salinity and pressure
Tf_mean = np.mean(T[51::] - Tf[51::])
#find temperature of no motion
v_profile = ((T - Tf) - Tf_mean) * phi
#create the velocity profile
v_profile[0:50] = 0
#there is no mean advection in the top 25 m
h_interface = (np.abs(v_profile[51::] - 0)).argmin() + 50
#find the depth of the point of no motion
inv_time_length = np.zeros(shape=len(z))
inv_time_length = np.absolute(v_profile) / l_poly
#create 1/tau
#Create the relaxation profiles for S and T
T_relax = np.zeros(shape=len(z))
T_relax = np.tanh((z - z[h_interface]) /
100) * (T_cdw - T_out) / 2 + (T_cdw + T_out) / 2
T_relax[h_interface:len(z)] = T_relax[h_interface:len(z)] + 0.6 * (
(z[h_interface:len(z)] - z[h_interface]) / 1000)
S_relax = np.zeros(shape=len(z))
S_relax = np.tanh((z - z[h_interface]) /
100) * (S_cdw - S_out) / 2 + (S_cdw + S_out) / 2
S_relax[h_interface:len(z)] = S_relax[h_interface:len(z)] + 0.25 * (
(z[h_interface:len(z)] - z[h_interface]) / 1000)
# update temperature
if advection == False:
dTdt = Q_profile / (c * rho0)
else:
dTdt = Q_profile / (c * rho0) + inv_time_length * (T_relax - T)
if debug == True:
print('inv_time_length*(T_relax-T): ',
inv_time_length[0:100] * (T_relax[0:100] - T[0:100]))
if use_MLI:
mld_ind = get_mld_ind(T, S, U, V, z)
mld = z[mld_ind]
C_e = 0.06
g = 9.81
# gravitational acceleration (m s^-2)
c = 4218
# heat capacity (J kg^-1 degC^-1
MLI_dTdt = -C_e * nabla_b**2 * mld**2 * rho0 / (np.abs(f) * alpha * g)
vert_profile = 4 / mld * (1 - 2 * z / mld) * (
16 + 10 *
(1 - 2 * z / mld)**2) / 21 # this is vertical derivative of mu(z)
vert_profile[mld_ind::] = 0
vert_profile[0:mld_ind] -= np.mean(
vert_profile[0:mld_ind]
) # just to ensure that no heat added to system
dTdt += MLI_dTdt * vert_profile
T += dTdt * dt
# update salinity
dSdt_0 = S[0] * (E - P) / dz / 1000
S[0] += dSdt_0 * dt
if advection == True:
dSdt = inv_time_length * (S_relax - S)
S += dSdt * dt
if use_MLI:
rho = get_rho(T, S, T0, S0, rho0, alpha, beta)
half_mld_ind = int(mld_ind / 2)
if np.any(np.diff(rho[half_mld_ind::]) < 0):
if verbose:
print(
'Need to homogenize discontinuity at base of previous mixed layer'
)
# get rid of discontinuity at base of previous mixed layer
# homogenize the region of water from mld/2 to z*
# z* is the shallowest value (> mld/2) such that the homogenized rho <= rho(z*)
zstar_ind = mld_ind.copy()
while np.mean(rho[half_mld_ind:zstar_ind]) >= rho[zstar_ind]:
if verbose:
print('Deepening z*...')
zstar_ind += 1
T[half_mld_ind:zstar_ind] = np.mean(T[half_mld_ind:zstar_ind])
S[half_mld_ind:zstar_ind] = np.mean(S[half_mld_ind:zstar_ind])
if tracer is not None:
tracer[:, half_mld_ind:zstar_ind] = np.atleast_2d(
np.mean(tracer[:, half_mld_ind:zstar_ind], axis=1)).T
elif verbose:
print('No need to homogenize base of previous mixed layer')
# update momentum
# first rotate momentum halfway
angle = -f * dt / 2 # currently assuming this is in rad
U, V = rotate(angle, U, V)
# then add wind stress
mld_ind = get_mld_ind(T, S, U, V, z)
mld = z[mld_ind]
U[0:mld_ind] += tau_x / mld / rho0 * dz * dt
V[0:mld_ind] += tau_y / mld / rho0 * dz * dt
# then rotate second half
U, V = rotate(angle, U, V)
if use_static_stability:
T, S, U, V = static_stability(T, S, U, V, z, T0, S0, rho0, alpha, beta)
if get_mld_ind(T, S, U, V, z) == (T.size - 1):
use_mixed_layer_stability = False
use_shear_stability = False
if use_mixed_layer_stability:
T, S, U, V = mixed_layer_stability(T,
S,
U,
V,
z,
T0,
S0,
rho0,
alpha,
beta,
verbose=verbose)
if use_shear_stability:
T, S, U, V = shear_stability(T,
S,
U,
V,
z,
T0,
S0,
rho0,
alpha,
beta,
verbose=verbose)
if vert_diffusivity is not None:
dTdt_vd = np.zeros(len(T))
dTdt_vd[1:-1] = np.diff(np.diff(T)) / dz**2
T += vert_diffusivity * dTdt_vd * dt
dSdt_vd = np.zeros(len(S))
dSdt_vd[1:-1] = np.diff(np.diff(S)) / dz**2
S += vert_diffusivity * dSdt_vd * dt
dUdt = np.zeros(len(U))
dUdt[1:-1] = np.diff(np.diff(U)) / dz**2
U += vert_diffusivity * dUdt * dt
dVdt = np.zeros(len(V))
dVdt[1:-1] = np.diff(np.diff(V)) / dz**2
V += vert_diffusivity * dVdt * dt
if tracer is not None:
dtdt = np.zeros(shape=tracer.shape)
dtdt[:, 1:-1] = np.diff(np.diff(tracer, axis=1), axis=1) / dz**2
tracer += vert_diffusivity * dtdt * dt
return_variables = (T, S, U, V)
if tracer is not None:
return_variables += (tracer, )
if return_MLD:
return_variables += (get_mld(T, S, U, V, z), )
if return_dTdS:
return_variables += (
dTdt,
dSdt,
)
if return_TSrelax:
return_variables += (
T_relax,
S_relax,
)
if return_vel:
return_variables += (v_profile, )
return return_variables
from datetime import datetime
# set up initial parameters (these are also set as defaults in the PWP code, so don't need to put them in)
T0 = 0
# reference temperature (degC)
S0 = 34
# reference salinity (parts per thousand; ppt)
rho0 = 1025
# reference density (kg m^-3)
alpha = -sw.alpha(
S0, T0, 0) * rho0 # thermal expansion coefficient (kg m^-3 degC^-1)
beta = sw.beta(S0, T0,
0) * rho0 # haline contraction coefficient (kg m^-3 ppt^-1 )
latitude = -75
# degrees north
f = sw.f(latitude) # planetary vorticity
fwflux1 = np.loadtxt('/home/erobo/ThompsonResearch/Data/fwflux1_oneyear.csv',
delimiter=',')
qnet1 = np.loadtxt('/home/erobo/ThompsonResearch/Data/qnet1_oneyear.csv',
delimiter=',')
taux1 = np.loadtxt('/home/erobo/ThompsonResearch/Data/taux1_oneyear.csv',
delimiter=',')
tauy1 = np.loadtxt('/home/erobo/ThompsonResearch/Data/tauy1_oneyear.csv',
delimiter=',')
temp_profile = np.loadtxt(
'/home/erobo/ThompsonResearch/Data/temp_profile_snapshot.txt')
salt_profile = np.loadtxt(
'/home/erobo/ThompsonResearch/Data/salt_profile_snapshot.txt')
depth_profile = np.loadtxt(
'/home/erobo/ThompsonResearch/Data/Schodlok_depths.txt')
# isotropic
Eiso = np.empty((292, omg.size))
for i in range(omg.size):
kiso, Eiso[:, i] = calc_ispec(k, l, E[:, :, i])
# linear dispersion relationship
kr = 2 * pi * kiso * 1.e-3
kr2 = kr**2
m = np.logspace(-3, 0., 500)
#omgr = 2*pi*omg/8600.
N2 = (1.7e-5)
f2 = sw.f(59.247177)**2
b = 1.e3
jmax = 100
#for j in range(jmax):
for j in range(m.size):
#m = (pi*j)/b
m2 = m[i]**2
omgr = np.sqrt((f2 * m2 + N2 * kr2) / (kr2 + m2))
krp = 1.e3 * kr / (2 * pi)
if j == 0:
omgrp = 3600 * omgr / (2 * pi)
else:
omgrp = np.vstack([omgrp, 3600 * omgr / (2 * pi)])
def model_timestep(T,
S,
U,
V,
z,
I,
L,
E,
P,
tau_x,
tau_y,
dt,
nabla_b=None,
Ekman_Q_flux=None,
use_static_stability=True,
use_mixed_layer_stability=True,
use_shear_stability=True,
use_Ekman_flux=False,
use_MLI=False,
tracer=None,
vert_diffusivity=None,
verbose=False,
I1=0.62,
I2=None,
lambda1=.6,
lambda2=20,
T0=17,
S0=36,
rho0=None,
alpha=None,
beta=None,
f=sw.f(40),
return_MLD=False):
# define initial variables
c = 4218 # heat capacity (J kg^-1 K^-1)
if I2 is None:
I2 = 1 - I1
if I1 + I2 != 1:
raise Exception('Shortwave insolation amplitudes need to sum to unity')
if rho0 is None:
rho0 = sw.dens(S0, T0, 0)
if alpha is None:
alpha = -sw.alpha(
S0, T0, 0
) * rho0 # multiply by rho to fit into nice equation of state (see get_rho function)
if beta is None:
beta = sw.beta(
S0, T0, 0
) * rho0 # # multiply by rho to fit into nice equation of state (see get_rho function)
dz = z[1] - z[0]
if use_Ekman_flux and Ekman_Q_flux is None:
raise Exception(
'Using Ekman-induced buoyacy flux but no buoyancy gradients were given.'
)
if use_MLI and nabla_b is None:
raise Exception('Using MLI but no horizontal buoyancy gradient given.')
T = T.copy()
S = S.copy()
U = U.copy()
V = V.copy()
# so don't overwrite data
if tracer is not None:
tracer = tracer.copy()
# make initial heat profile
I_profile = -I / dz * (
I1 * np.exp(-z / lambda1) *
(np.exp(-dz / 2 / lambda1) - np.exp(dz / 2 / lambda1)) +
I2 * np.exp(-z / lambda2) *
(np.exp(-dz / 2 / lambda2) - np.exp(dz / 2 / lambda2)))
L_profile = np.zeros(len(z))
L_profile[0] = L / dz
Q_profile = I_profile + L_profile
if use_Ekman_flux:
A = 0.1 # eddy viscosity m^2 s^-1
z_Ek = np.sqrt(A / np.abs(f))
if verbose:
print('Using Ekman depth of %d m' % z_Ek)
z_Ek_ind = np.where(z > z_Ek)[0][0]
Q_Ek_profile = np.zeros(len(z))
Q_Ek_profile[0:z_Ek_ind] = Ekman_Q_flux / z_Ek * dz
Q_profile += Q_Ek_profile
# update temperature
dTdt = Q_profile / (c * rho0)
if use_MLI:
mld_ind = get_mld_ind(T, S, U, V, z)
mld = z[mld_ind]
C_e = 0.06
g = 9.81
# gravitational acceleration (m s^-2)
c = 4218
# heat capacity (J kg^-1 degC^-1
MLI_dTdt = -C_e * nabla_b**2 * mld**2 * rho0 / (np.abs(f) * alpha * g)
vert_profile = 4 / mld * (1 - 2 * z / mld) * (
16 + 10 *
(1 - 2 * z / mld)**2) / 21 # this is vertical derivative of mu(z)
vert_profile[mld_ind::] = 0
vert_profile[0:mld_ind] -= np.mean(
vert_profile[0:mld_ind]
) # just to ensure that no heat added to system
dTdt += MLI_dTdt * vert_profile
T += dTdt * dt
# update salinity
dSdt = S[0] * (E - P) / dz / 1000
S[0] += dSdt * dt
if use_MLI:
rho = get_rho(T, S, T0, S0, rho0, alpha, beta)
half_mld_ind = int(mld_ind / 2)
if np.any(np.diff(rho[half_mld_ind::]) < 0):
if verbose:
print(
'Need to homogenize discontinuity at base of previous mixed layer'
)
# get rid of discontinuity at base of previous mixed layer
# homogenize the region of water from mld/2 to z*
# z* is the shallowest value (> mld/2) such that the homogenized rho <= rho(z*)
zstar_ind = mld_ind.copy()
while np.mean(rho[half_mld_ind:zstar_ind]) >= rho[zstar_ind]:
if verbose:
print('Deepening z*...')
zstar_ind += 1
T[half_mld_ind:zstar_ind] = np.mean(T[half_mld_ind:zstar_ind])
S[half_mld_ind:zstar_ind] = np.mean(S[half_mld_ind:zstar_ind])
if tracer is not None:
tracer[:, half_mld_ind:zstar_ind] = np.atleast_2d(
np.mean(tracer[:, half_mld_ind:zstar_ind], axis=1)).T
elif verbose:
print('No need to homogenize base of previous mixed layer')
# update momentum
# first rotate momentum halfway
angle = -f * dt / 2 # currently assuming this is in rad
U, V = rotate(angle, U, V)
# then add wind stress
mld_ind = get_mld_ind(T, S, U, V, z)
mld = z[mld_ind]
U[0:mld_ind] += tau_x / mld / rho0 * dz * dt
V[0:mld_ind] += tau_y / mld / rho0 * dz * dt
# then rotate second half
U, V = rotate(angle, U, V)
if use_static_stability:
T, S, U, V, tracer = static_stability(T,
S,
U,
V,
z,
T0,
S0,
rho0,
alpha,
beta,
tracer=tracer,
verbose=verbose)
if use_mixed_layer_stability:
T, S, U, V, tracer = mixed_layer_stability(T,
S,
U,
V,
z,
T0,
S0,
rho0,
alpha,
beta,
tracer=tracer,
verbose=verbose)
if use_shear_stability:
T, S, U, V, tracer = shear_stability(T,
S,
U,
V,
z,
T0,
S0,
rho0,
alpha,
beta,
tracer=tracer,
verbose=verbose)
if vert_diffusivity is not None:
dTdt = np.zeros(len(T))
dTdt[1:-1] = np.diff(np.diff(T)) / dz**2
T += vert_diffusivity * dTdt * dt
dSdt = np.zeros(len(S))
dSdt[1:-1] = np.diff(np.diff(S)) / dz**2
S += vert_diffusivity * dSdt * dt
dUdt = np.zeros(len(U))
dUdt[1:-1] = np.diff(np.diff(U)) / dz**2
U += vert_diffusivity * dUdt * dt
dVdt = np.zeros(len(V))
dVdt[1:-1] = np.diff(np.diff(V)) / dz**2
V += vert_diffusivity * dSdt * dt
if tracer is not None:
dtdt = np.zeros(shape=tracer.shape)
dtdt[:, 1:-1] = np.diff(np.diff(tracer, axis=1), axis=1) / dz**2
tracer += vert_diffusivity * dtdt * dt
return_variables = (
T,
S,
U,
V,
)
if tracer is not None:
return_variables += (tracer, )
if return_MLD:
return_variables += (get_mld(T, S, U, V, z), )
return return_variables
def vmodes(N2, z, nm, lat, ubdy='N', lbdy='N'):
f0 = sw.f(lat)
dz = np.abs(z[1] - z[0])
# f2n2 array
f2n2 = (f0**2) / N2
# assembling matrices
# N2F2 matrix
N2F2 = np.eye(z.size - 1)
for i in range(0, z.size - 1):
N2F2[i, i] = (f2n2[i] + f2n2[i + 1]) / 2
# 1st order difference matrix
A = np.eye(z.size)[:, :-1]
for i in range(1, z.size):
A[i, i - 1] = -1
A = A / dz
# linear operator C = A*N2F2*(A.transpose())
A = np.matrix(A)
N2F2 = np.matrix(N2F2)
C = A * N2F2 * A.transpose()
if ubdy == 'D':
C = C[1:, 1:]
if lbdy == 'D':
C = C[:-1, :-1]
# solve the eigenvalue problem
lam, fi = np.linalg.eig(C)
i = np.argsort(lam)
ei = lam[i]
fi = fi[:, i]
ei = ei[:nm]
fi = fi[:, :nm]
if ubdy == 'D':
fi = np.append(np.matrix(np.zeros(nm)), fi, axis=0)
if lbdy == 'D':
fi = np.append(fi, np.matrix(np.zeros(nm)), axis=0)
# normalizing to get orthonormal modes
H = np.abs(z).max()
for i in range(nm):
s = np.sqrt(
dz *
((np.array(fi[1:, i])**2 + np.array(fi[:-1, i])**2) / 2 / H).sum())
fi[:, i] = fi[:, i] / s
# to keep consistency multiply by plus/minus to make the leading term positive
for i in range(nm):
if np.sign(fi[1, i]) < 0:
fi[:, i] = -fi[:, i]
# compute the deformation radii [km]; the barotropic radius is computed by sqrt(gH)/f0
radii = np.zeros(nm)
radii[0] = np.sqrt(9.81 * H) / np.abs(f0) / 1000
radii[1:] = 1 / np.sqrt(ei[1:]) / 1000
return fi, radii
def plotsec(opname,
rad,
pastafig,
pathadcp,
corsta='b',
cond=False,
ts=False,
fmed=False,
interp=False,
radcomplicada=False,
mdrcalc=True):
vsec,latv,lonv,lat,lon,section,lonsta,latsta = [],[],[],[],[],[],[],[]
adcp, lat_adcp, lon_adcp = [], [], []
radname = 'r' + os.path.split(rad)[1]
if mdrcalc:
vsec, lat, lon, section = extsect(rad, sup=True)
else:
vsec, lat, lon, section = extsect(rad, sup=False)
# opname = section.items.values[0].split('0')[0]
if ts:
tscomp(section)
plt.savefig(pastafig + opname + '/ts/' + 'ts' + radname + '.png',
dpi=100,
format='png')
plt.close('all')
lat_adcp, lon_adcp, adcp = adcpvel(lat,
lon,
pathadcp,
dya=600,
interp=interp)
if mdrcalc:
vsec = mdr(vsec, adcp)
# lonsta = np.nanmean(section.minor_xs('lon'),axis=0)
# latsta = np.nanmean(section.minor_xs('lat'),axis=0)
latv = lat[:-1] + np.diff(lat) / 2
lonv = lon[:-1] + np.diff(lon) / 2
if fmed:
f = sw.f(latv)
vsec = vsec * (f / np.mean(f))
if cond:
# SALINIDADE
plotprop(z=section.minor_xs('sp'),
lat=latsta,
lon=lonsta,
cor=corsta,
csteps=0.5,
title=u'Salinidade Prática',
propmin=34,
propmax=37)
plt.savefig(pastafig + opname + '/salinidade/' + 'sal' + radname +
'.png',
dpi=100,
format='png')
plt.close('all')
# TEMPERATURA
plotprop(z=section.minor_xs('pt'),
lat=latsta,
lon=lonsta,
cor=corsta,
csteps=2.5,
title=u'Temperatura Potencial ($\\theta_0$)',
propmin=0,
propmax=28)
plt.savefig(pastafig + opname + '/temperatura/' + 'temp' + radname +
'.png',
dpi=100,
format='png')
plt.close('all')
# MASSAS D'ÁGUA
plotprop(z=section.minor_xs('psigma0'),
lat=latsta,
lon=lonsta,
cor=corsta,
wmass=True,
title=u'Massas D\'água ($\\rho_0$)')
plt.savefig(pastafig + opname + '/massas/' + 'mass' + radname + '.png',
dpi=100,
format='png')
plt.close('all')
# DENSIDADE POTENCIAL
plotprop(z=section.minor_xs('psigma0'),
lat=latsta,
lon=lonsta,
cor=corsta,
csteps=0.5,
title=u'Densidade Potencial ($\\rho_0$)',
propmin=23,
propmax=28)
plt.savefig(pastafig + opname + '/densidade/' + 'dens' + radname +
'.png',
dpi=100,
format='png')
plt.close('all')
# reff = adcp.iloc[150,:].values-vsec.iloc[150,:].values
# vsec += reff
if radcomplicada:
latv[3] = -2.0933934592636545
lonv[3] = -38.442823159524877
if mdrcalc:
ttitle = u'Velocidade Geostrófica (MDR)'
else:
ttitle = u'Velocidade Geostrófica (Isopicnal = 32.15)'
# Velocidade Geostrófica
plotprop(z=vsec,
cmap='RdBu_r',
lat=latv,
lon=lonv,
cor=corsta,
csteps=0.1,
title=ttitle,
propmin=-1.5,
propmax=1.5,
lim=-300)
plt.savefig(pastafig + opname + '/geostrofia/' + 'geos' + radname + '.png',
dpi=100,
format='png')
plt.close('all')
plotprop(z=vsec,
lat=latv,
lon=lonv,
cor=corsta,
z2=adcp,
lat2=lat_adcp,
lon2=lon_adcp,
cor2='g',
csteps=0.1,
title=u'Velocidade Geostrófica X Velocidade Observada',
cmap='RdBu_r',
propmin=-1.5,
propmax=1.5,
comp=True,
lim=-300)
plt.savefig(pastafig + opname + '/velocidade/' + 'vel' + radname +
'adcp.png',
dpi=100,
format='png')
plt.close('all')
"""
params = {}
params['dt'] = 3600.0*dt
params['dt_d'] = params['dt']/86400.
params['dz'] = dz
params['dt_save'] = dt_save
params['lat'] = lat
params['rb'] = rb
params['rg'] = rg
params['rkz'] = rkz
params['beta1'] = beta1
params['beta2'] = beta2
params['max_depth'] = max_depth
params['g'] = 9.81
params['f'] = sw.f(lat)
params['cpw'] = 4183.3
params['ucon'] = (0.1*np.abs(params['f']))
params['mld_thresh'] = mld_thresh
params['winds_ON'] = winds_ON
params['emp_ON'] = emp_ON
params['heat_ON'] = heat_ON
params['drag_ON'] = drag_ON
return params
def prep_data(met_dset, prof_dset, params):
def qg_stability_2d(N2,
ubar,
vbar,
z,
k,
l,
lat,
hx=0.,
hy=0.,
structure=False):
f0 = sw.f(lat)
beta = 2 * 7.29e-5 * np.cos(lat * pi / 180.) / 6371.e3
# calculate surface shear
ubarz, vbarz = ubar[0] - ubar[1], vbar[0] - vbar[1]
dz = np.abs(z[1] - z[0])
S = (f0**2) / N2
L = stretching_matrix(z.size, S, dz)
k2 = k**2 + l**2
L2 = L - np.eye(ubar.size) * k2
Qy = np.eye(
ubar.size) * (beta - np.array(np.matrix(L) * np.matrix(ubar).T))
Qx = np.eye(ubar.size) * np.array(np.matrix(L) * np.matrix(vbar).T)
# bottom bc with topography
Qx[-1, -1] += (f0 / dz) * hx
Qy[-1, -1] += (f0 / dz) * hy
Q = k * Qy - l * Qx
L3 = np.empty_like(L2)
for i in range(ubar.size):
L3[i, :] = L2[i, :] * (ubar[i] * k + vbar[i] * l)
A = (L3 + Q)
B = L2.copy()
# append surface boundary condition
#A = np.vstack([np.zeros_like(z),A])
#B = np.vstack([np.zeros_like(z),B])
#A = np.hstack([np.zeros(z),A])
#B = np.hstack([np.zeros(z),B])
#print(A.shape)
#print(B.shape)
# the upper boundary condition
#A[0,0] = (ubar[0]*k+vbar[0]*l)
#A[0,1] = -(ubar[0]*k+vbar[0]*l)
#A[0,0] += -(k*ubarz + l*vbarz)
#B[0,0], B[0,1] = dz,-dz
#print(A[0,:5])
# bottom boundary condition satisfied
# (may be should include topographic gradients)
try:
evals, evecs = sp.linalg.eig(A, B)
imax = evals.imag.argmax()
eval_max = evals[imax]
evec_max = evecs[0:, imax]
except:
eval_max = np.nan + 1.j * np.nan
evec_max = np.nan * z
#print(eval_max)
if structure:
PSI, X, Z = wave_structure(evec_max, z, k)
return eval_max, evec_max, PSI, X, Z
else:
return eval_max, evec_max
#Le = np.linspace(750.,10.,30)
#k = 2*pi/(Le*1.e3)
#gr = []
#for i in range(k.size):
# c,psi,PSI,X,Z = qg_stability(N2i,Up,-zi,k=k[i],lat=58)
# gr.append(c.imag*k[i])
#Le = np.linspace(1000.,1.,50)
Le = np.logspace(3, 5., 50)
k = 2 * pi / (Le * 1.e3)
k = np.hstack([-np.flipud(k), k])
l = k.copy()
f0 = sw.f(-58)
S = f0**2 / N2i
N = N2i.size
evals, evecs = pmodes(N, S, -zi, nn=10, dz=zi[2] - zi[1])
xi = integrate.trapz(evecs[:, 1]**3, zi) / zi.max()
delta = .25 * ((np.sqrt(xi**2 + 4.) - xi)**2)
# project Ui onto evecs
#A = evecs
#ATA = np.dot(A.T,A)
#ATU = np.dot(A.T,Up)
#alpha = np.dot(np.linalg.inv(ATA),ATU)
au, av = modal_projection(evecs, Up), modal_projection(evecs, Vp)
ds1 = xr.open_dataset('../data/model_stats/S0.01_gridded_stats.nc')
ds1 = ds1.drop('index')
ds1.set_coords(['alpha', 'slope', 'seed', 'period'])
df1 = ds1.to_dataframe()
ds5 = xr.open_dataset('../data/model_stats/S0.005_gridded_stats.nc')
ds5 = ds5.drop('index')
ds5.set_coords(['alpha', 'slope', 'seed', 'period'])
df5 = ds5.to_dataframe()
df = pd.concat([df1, df5])
cg = 9.8*df.t0m1_mean/4/np.pi
cg1 = 9.8*df1.t0m1_mean/4/np.pi
cg5 = 9.8*df5.t0m1_mean/4/np.pi
c_llc = 9.8*dfl.t0m1_mean/4/np.pi
f = sw.f(31.5)
ind=df.alpha<0.9
ind1=df1.alpha<0.9
ind5=df5.alpha<0.9
corr_df = np.corrcoef(df.vorticity[ind].values, cg[ind]*df.hs_grad[ind]/df.hs_mean[ind]/df.slope[ind])[0, 1]
cor_llc = np.corrcoef(dfl.vorticity, c_llc*dfl.hs_grad/dfl.hs_mean/dfl.EKE_psi_slope)[0,1]
a, b, __, __, __ = linregress(df.vorticity[ind]/f, cg[ind]*df.hs_grad[ind]/df.hs_mean[ind]/df.slope[ind]/f)
############################################
# Normalized scatterplot
############################################
plt.figure(figsize=(8,8))
plt.plot(df1.vorticity[ind1]/f, cg1[ind1]*df1.hs_grad[ind1]/df1.hs_mean[ind1]/df1.slope[ind1]/f, 'o',
color='steelblue', alpha=.5, label='KE = 0.01 m$^2$/s$^2$')
Nb = np.nanmean(N2)
P = np.arange(0., N2.size)
P0 = np.ones(N2.size)
plt.figure()
rdi = np.zeros(nmodes)
for n in np.arange(1, nmodes):
P[0] = (-1)**n * np.sqrt(2 * N2[0] / Nb)
P[-1] = np.sqrt(2 * N2[-1] / Nb)
rdi[n] = np.abs((n * np.pi * f / (Nb * N2.size - 1)))
for z in np.arange(1, N2.size):
P[z] = np.sqrt(2 * N2[z] / Nb) * np.cos(
n * np.pi / (Nb * N2.size - 1) * np.sum(N2[z:]))
plt.plot(np.flipud(P), -np.arange(0, N2.size), linewidth=1.6)
plt.plot(np.flipud(P0), -np.arange(0, N2.size), linewidth=1.6)
plt.plot(np.tile(0, N2.size), -np.arange(0, N2.size), '--k', linewidth=1.6)
return rdi
prof_max = 5000.
z = np.arange(0., prof_max + 1) * -1
d = 900.
f = sw.f(23.)
N2 = 10**4 * np.exp(z / d) * f**2
#N2=np.ones(z.size)*5
plt.figure()
plt.plot(N2, z)
rdi = wkb(N2, f, nmodes=3)