Skip to contents

This function provides semi-analytical updates for the disturbance probability.

Usage

disturbance_DA(
  X,
  Y,
  R,
  H,
  aqq,
  bqq,
  disturbance,
  mu0 = rep(0.8, 0.8),
  V0 = diag(rep(0.25, 2)),
  pD = c(0.99, 0.01)
)

Arguments

X

numeric matrix (dim = M*N, where M is the ensemble number while N is the number of variables): model forecasts for each variable of each ensemble member.

Y

numeric vector (dim = L, the number of observations): vector stores the available observations for the corresponding variables.

R

numeric matrix (dim = L*L): the covariance matrix of the observations.

H

integral vector (dim = L): mapping operator (Y -> X. e.g., c(1,3,4) represents the fact that we have three observations covered the 1st, 2nd, and 4th state variables).

aqq

numeric vector (dim = N): the vector of the `shape` parameter for the gamma distribution for each state variable.

bqq

numeric vector (dim = N): the vector of the `rate` parameter for the gamma distribution for each state variable.

disturbance

logical (dim = 1): if the current site experienced the disturbance (1) or not (0).

mu0

numeric vector (dim = B where B is the number of biomass variables): percent decreases of biomass to be used for the disturbance model.

V0

numeric matrix (dim B*B): covariance to be used for the disturbance model

pD

prior of disturbed Vs. undisturbed probability, default is c(0.99, 0.01).

Value

returns a vector of mua (dim = N, vector of means for each updated state variable) and pa (dim = N*N, covariance matrix for the updated state variables).