generateRF {CoSMoS}

Simulation of random field with given marginals and spatiotemporal properties

Description

Generates random field with given marginals and spatiotemporal properties, just provide (1) the output of fitVAR function, and (2) the number of time steps to simulate.

Usage

generateRF(n, STmodel)

Arguments

n

number of fields (time steps) to simulate

STmodel

list of arguments resulting from fitVAR function

Details

Referring to the documentation of fitVAR for details on computational complexity of the fitting algorithm, here we report indicative simulation CPU times for some settings, assuming that the model parameters are already evaluated. CPU times refer to a Windows 10 Pro x64 laptop with Intel(R) Core(TM) i7-6700HQ CPU @ 2.60GHz, 4-core, 8 logical processors, and 32GB RAM.
CPU time:
m = 30, p = 1, n = 1000: ~17s
m = 30, p = 1, n = 10000: ~75s
m = 30, p = 5, n = 100: ~280s
m = 30, p = 5, n = 1000: ~302s
m = 50, p = 1, n = 1000 : ~160s
m = 50, p = 1, n = 10000 : ~570s where m denotes the side length of a square field (mxm)

Examples

## The example below refers to the simulation of few random fields of
## size 10x10 with AR(1) temporal correlation for the sake of illustration.
## For a more effective visualization and reliable performance assessment,
## we suggest to generate a larger number of fields (e.g. 100 or more)
## of size about 30X30.
## See section 'Details' for additional information on running times
## with different settings.

fit <- fitVAR(
  spacepoints = 10,
  p = 1,
  margdist ='burrXII',
  margarg = list(scale = 3, shape1 = .9, shape2 = .2),
  p0 = 0.8,
  stcsid = "clayton",
  stcsarg = list(scfid = "weibull", tcfid = "weibull",
                 copulaarg = 2,
                 scfarg = list(scale = 20, shape = 0.7),
                tcfarg = list(scale = 1.1, shape = 0.8))
)

sim <- generateRF(n = 12,
                    STmodel = fit)
checkRF(sim,
          lags = 10,
          nfields = 12)