kalmanFilter.prior

Set or return the prior for a Kalman filter

Syntax


obj = obj.prior(X)
obj = obj.prior(ens)
obj = obj.prior(  X, whichPrior)
obj = obj.prior(ens, whichPrior)
[  X, whichPrior] = obj.prior
[ens, whichPrior] = obj.prior
obj = obj.prior(‘delete’)

Description


obj = obj.prior(X)

Provide the prior for a Kalman filter direclty as a numeric array. Overwrites any previously specified prior. Each row is a state vector element, and each column is an ensemble member. Each element along the third dimension holds a particular prior in an evolving set. If using a static prior, X should have a single element along the third dimension. Otherwise, X should have one element along the third dimension per assimilation time step. In this case, the Kalman filter will use the indicated prior for each assimilated time step.


obj = obj.prior(ens)

Provide the prior using ensemble objects. To implement a static prior, the input should be a scalar ensemble object that implements a static prior.

For evolving priors, the input may either be: 1. A scalar ensemble object that implements an evolving ensemble, or 2. A vector of ensemble objects that all implement static ensembles. In the first case, the number of priors for the assimilation will match the number of ensembles in the evolving set. In the second case, the number of priors for the assimilation will match the number of elements in the vector of ensemble objects. The number of evolving priors must match the number of assimilation time steps (although see the syntax below for relaxing that requirement). If using a vector of ensemble objects, all the objects must implement ensembles with the same number of state vector elements (nRows) and ensemble members (nMembers).


obj = obj.prior(  X, whichPrior)
obj = obj.prior(ens, whichPrior)

Indicate which prior to use in each assimilation time step. This syntax allows the number of priors to differ from the number of time steps.


[  X, whichPrior] = obj.prior
[ens, whichPrior] = obj.prior

Returns the current prior for the Kalman filter object, and indicates which prior is used in each assimilation time step.


obj = obj.prior(‘delete’)

Deletes the current prior(s) from the Kalman filter object.

Input Arguments

X

X

numeric matrix [nState x nMembers x 1|*nTime*|*nPrior*]

The prior(s) for the Kalman filter. A numeric array with one row per state vector element, and one column per ensemble member. Each element along the third dimension holds a unique prior. If using a static prior, X should have one element along the third dimension. If the number of priors matches the number of time steps, uses the appropriate prior for each time step. If the number of priors is neither 1 nor the number of time steps, use the whichPrior input to indicate which prior to use in each assimilation time step. If the prior for a state vector element includes NaN values in a particular time step, then the updated state vector element will be NaN in that time step.

ens

ens

scalar ensemble object <static | evolving [nPrior]> | vector, <static> ensemble objects [nPrior]

The prior for the Kalman filter, provided via ensemble objects. For a static prior, a scalar ensemble object that implements a static ensemble. For an evolving ensemble, either a scalar ensemble object that implements an evolving ensemble, OR a vector of ensemble objects that implement static ensembles. If providing a vector of ensemble objects, each object must implement an ensemble with the same number of state vector elements (rows) and ensemble members (columns). If the number of priors matches the number of time steps, uses the appropriate prior for each time step. If the number of priors is neither 1 nor the number of time steps, use the whichPrior input to indicate which prior to use in each assimilation time step. If the prior for a state vector element includes NaN values in a particular time step, then the updated state vector element will be NaN in that time step.

whichPrior

whichPrior

vector, positive integers [nTime]

Indicates which prior to use in each assimilation time step. Must have one element per assimilation time step.

Output Arguments

obj

obj

scalar kalmanFilter object

The kalmanFilter object with the udpated prior.

X

X