RO Parameters#
Class |
Parameter (equation) |
Parameter (code) |
Unit |
Physical meaning |
---|---|---|---|---|
Deterministic Parameters |
Linear parameters |
|||
\(R\) |
R |
month\(^{-1}\) |
Bjerknes feedback |
|
\(F_1\) |
F1 |
K m\(^{-1}\) month\(^{-1}\) |
Delayed oceanic feedback efficiency |
|
\(F_2\) |
F2 |
m K\(^{-1}\) month\(^{-1}\) |
Recharge/discharge efficiency |
|
\(\varepsilon\) |
epsilon |
month\(^{-1}\) |
Basin adjustment |
|
Nonlinear parameters |
||||
\(b_T\) |
b_T |
K\(^{-1}\) month\(^{-1}\) |
Deterministic quadratic nonlinearities |
|
\(c_T\) |
c_T |
K\(^{-2}\) month\(^{-1}\) |
Deterministic cubic nonlinearities |
|
\(d_T\) |
d_T |
m\(^{-1}\) month\(^{-1}\) |
Deterministic quadratic nonlinearities |
|
\(b_h\) |
b_h |
K\(^{-2}\) m month\(^{-1}\) |
Deterministic quadratic nonlinearities |
|
Noise parameters |
\(\sigma_T\) |
sigma_T |
K month\(^{-\frac{1}{2}}\) (n_T = 1)K month\(^{-1}\) (n_T = 0) |
Stochastic forcing amplitude |
\(\sigma_h\) |
sigma_h |
m month\(^{-\frac{1}{2}}\) (n_h = 1)m month⁻¹ (n_h = 0) |
Stochastic forcing amplitude |
|
\(B\) |
B |
K⁻¹ |
State-dependence noise |
|
\(m_T\) |
m_T |
month⁻¹ |
red noise decorrelation rate |
|
\(m_h\) |
m_h |
month⁻¹ |
red noise decorrelation rate |
|
Noise option parameters |
\(n_T\) |
n_T |
no unit |
noise option |
\(n_h\) |
n_h |
no unit |
noise option |
|
\(n_g\) |
n_g |
no unit |
noise option |
|
Noise forcing |
\(\xi_T\) |
xi_T |
no unit |
red noise |
\(\xi_h\) |
xi_h |
no unit |
red noise |
|
\(w_T\) |
w_T |
month\(^{-\frac{1}{2}}\) |
white noise |
|
\(w_h\) |
w_h |
month\(^{-\frac{1}{2}}\) |
white noise |
|
External forcing |
\(E_T\) |
E_T |
K month⁻¹ |
extental forcing |
\(E_h\) |
E_h |
m month⁻¹ |
extental forcing |
|
Variable |
\(T\) |
T |
K |
ENSO SST anomaly |
\(h\) |
h |
m |
ENSO therocline anomaly |
Linear and Nonlinear Parameters:
\(R\) (R
), \(F_1\) (F1
), \(F_2\) (F2
), \(\varepsilon\) (epsilon
), \(b_T\) (b_T
), \(c_T\) (c_T
), \(d_T\) (d_T
), \(b_h\) (b_h
)
'parameter': []
→ interpreted as zero (e.g.,'R': []
is the same as'R': [0.0]
)'parameter': [value]
→ only the annual mean is used (e.g.,'R': [-0.05]
means the annual mean value of –0.05 is used for R)'parameter': [mean, amplitude, phase (in radians)]
→ annual mean and annual seasonality are used (e.g.,'R': [-0.05, 0.05, np.pi]
means R has an annual mean of –0.05 and a seasonal variation with amplitude 0.05 and phase shift \(\pi\))'parameter': [mean, amp₁, phase₁ (in radians), amp₂, phase₂ (in radians)]
→ annual mean, annual seasonality, and semi-annual seasonality are used (e.g.,'R': [-0.05, 0.05, np.pi, 0.01, np.pi/2]
means R has an annual mean of –0.05, annual amplitude 0.05 with phase \(\pi\), and semi-annual amplitude 0.01 with phase \(\pi/2\))Any other form → invalid (e.g.,
'parameter': [mean, amp]
will raise an error)
Noise Parameters:
\(\sigma_T\) (sigma_T
), \(\sigma_h\) (sigma_h
), \(B\) (B
), \(m_T\) (m_T
), \(m_h\) (m_h
)
These follow the same format as the linear and nonlinear parameters, meaning that seasonality can be included for RO simulations.
However, RO parameter fitting does not support seasonality in these parameters. Therefore, it is strongly recommended to use annual means only for consistency during RO fitting.
Noise Option Parameters:
\(n_T\) (n_T
), \(n_h\) (n_h
), \(n_g\) (n_g
)
n_T
andn_h
are flags that specify the noise type for the \(T\) and \(h\) equations, respectively:0
: red noise1
: white noise
n_g
is a flag for the multiplicative noise type in the \(T\) equation:0
: linear multiplicative noise \((1 + B \cdot T)\)1
: Heaviside-based multiplicative noise \((1 + B \cdot H(T) \cdot T)\)2
: no multiplicative noise (i.e., additive only)
'parameter': [integer]
or'parameter': integer
is a valid option; any other format will raise an error.Mixed noise types (e.g.,
n_T = 0
,n_h = 1
) are allowed in the solver but not allowed during parameter fitting. Therefore, it is strongly recommended to use a consistent noise type (either white or red) for both the \(T\) and \(h\) equations.