Geostatistical simulation: SGS with block data, ensembles, p-field, LMC and collocated co-simulation, SIS, SNESIM, annealing

SgsBlockSimulate: sequential Gaussian simulation (simple kriging with known mean from the k nearest informed nodes plus all block-average data with discretised point-to-block covariances). ConditionalEnsemble: E-type mean, conditional variance and quantiles over realisations with seeds seed + r. PfieldSimulate: m + s Y with an unconditional correlated Gaussian field Y. LmcConditionalSimulate: exact conditional joint simulation of p coregionalised fields. CollocatedCosimulate: sequential Gaussian co-simulation by collocated cokriging (Markov model 1). SisMarkovBayes: sequential indicator simulation with optional collocated Markov-Bayes soft data. SnesimSimulate: SNESIM by scanning the training image for the data event. AnnealingSimulate: value swaps to a target variogram. Covariance models are lists (model "Exp", "Sph" or "Gau"; sill, range, nugget); Philox streams, paths and results match the Python arm.

Usage

SgsBlockSimulate(
  data_coords,
  data_values,
  targets,
  model,
  mean = 0,
  k = 12,
  blocks = NULL,
  seed = 0
)

ConditionalEnsemble(
  data_coords,
  data_values,
  targets,
  model,
  mean = 0,
  k = 12,
  n_real = 20,
  probs = c(0.1, 0.5, 0.9),
  seed = 0
)

PfieldSimulate(krige_mean, krige_sd, coords, model, seed = 0)

LmcConditionalSimulate(
  data_coords,
  data_values,
  targets,
  components,
  means = NULL,
  seed = 0
)

CollocatedCosimulate(
  data_coords,
  data_values,
  targets,
  secondary,
  model,
  rho,
  k = 12,
  seed = 0
)

SisMarkovBayes(
  data_coords,
  data_categories,
  targets,
  proportions,
  models,
  k = 12,
  soft = NULL,
  B = NULL,
  seed = 0
)

SnesimSimulate(training, nx, ny, template, conditioning = NULL, seed = 0)

Arguments

data_coords: Data coordinates (matrix rows or list).
data_values: Data values (for LmcConditionalSimulate a matrix, NA missing).
targets: Target coordinates.
model: Covariance model list.
mean: Known mean.
k: Neighbourhood size.
blocks: List of list(points, value) block data, or NULL.
seed: Philox seed.
n_real: Number of realisations.
probs: Quantile levels.
krige_mean, krige_sd: Local kriging means and standard deviations.
coords: Node coordinates.
components: List of list(B, model) LMC components.
means: Variable means, or NULL.
secondary: Secondary values at the targets.
rho: Primary-secondary correlation.
data_categories: Data categories (0-based).
proportions: Category proportions.
models: Indicator covariance models per category.
soft: Soft probabilities at the targets (rows), or NULL.
B: Markov-Bayes calibration coefficients per category.
training: Training image (matrix rows are y, columns x).
nx, ny: Grid size.
template: List of c(di, dj) offsets.
conditioning: Data frame or matrix with columns i, j, value (0-based), or NULL.

Value

A list.

References

Deutsch, C. V. and Journel, A. G. (1998). GSLIB, 2nd ed. Liu, Y. and Journel, A. G. (2009). Computers and Geosciences 35, 527-547. Goovaerts, P. (1997). Geostatistics for Natural Resources Evaluation. Froidevaux, R. (1993). In Geostatistics Troia 92, 73-84. Almeida, A. S. and Journel, A. G. (1994). Math. Geology 26, 565-588. Zhu, H. and Journel, A. G. (1993). In Geostatistics Troia 92. Strebelle, S. (2002). Math. Geology 34, 1-21. Deutsch, C. V. and Cowan, P. W. (1996). In GSLIB, section V.6.

Examples

m <- list(model = "Exp", sill = 1, range = 2)
SgsBlockSimulate(rbind(c(0, 0)), 1.5, rbind(c(0, 0), c(5, 0)), m, seed = 1)$simulated
#> [1]  1.5000000 -0.5186549