Title: | Spatial and Spatio-Temporal Geostatistical Modelling, Prediction and Simulation |
---|---|
Description: | Variogram modelling; simple, ordinary and universal point or block (co)kriging; spatio-temporal kriging; sequential Gaussian or indicator (co)simulation; variogram and variogram map plotting utility functions; supports sf and stars. |
Authors: | Edzer Pebesma [aut, cre] , Benedikt Graeler [aut] |
Maintainer: | Edzer Pebesma <[email protected]> |
License: | GPL (>= 2.0) |
Version: | 2.1-2 |
Built: | 2024-12-16 13:34:41 UTC |
Source: | https://github.com/r-spatial/gstat |
Data obtained from Gomez and Hazen (1970, Tables 19 and 20) on coal ash for the Robena Mine Property in Greene County Pennsylvania.
data(coalash)
data(coalash)
This data frame contains the following columns:
a numeric vector; x-coordinate; reference unknown
a numeric vector; x-coordinate; reference unknown
the target variable
data are also present in package fields, as coalash.
unknown; R version prepared by Edzer Pebesma; data obtained from http://homepage.divms.uiowa.edu/~dzimmer/spatialstats/, Dale Zimmerman's course page
N.A.C. Cressie, 1993, Statistics for Spatial Data, Wiley.
Gomez, M. and Hazen, K. (1970). Evaluating sulfur and ash distribution in coal seems by statistical response surface regression analysis. U.S. Bureau of Mines Report RI 7377.
see also fields manual: https://www.image.ucar.edu/GSP/Software/Fields/fields.manual.coalashEX.Krig.shtml
data(coalash) summary(coalash)
data(coalash) summary(coalash)
Spatio-temporal data set with rural background PM10 concentrations in Germany 2005 (airbase v6).
data("DE_RB_2005")
data("DE_RB_2005")
The format is: Formal class 'STSDF' [package "spacetime"] with 5 slots ..@ data :'data.frame': 23230 obs. of 2 variables: .. ..$ PM10 : num [1:23230] 16.7 31.7 5 22.4 26.8 ... .. ..$ logPM10: num [1:23230] 2.82 3.46 1.61 3.11 3.29 ... ..@ index : int [1:23230, 1:2] 1 2 3 4 5 6 7 8 9 10 ... ..@ sp :Formal class 'SpatialPointsDataFrame' [package "sp"] with 5 slots .. .. ..@ data :'data.frame': 69 obs. of 9 variables: .. .. .. ..$ station_altitude : int [1:69] 8 3 700 15 35 50 343 339 45 45 ... .. .. .. ..$ station_european_code: Factor w/ 7965 levels "AD0942A","AD0944A",..: 1991 1648 1367 2350 1113 1098 1437 2043 1741 1998 ... .. .. .. ..$ country_iso_code : Factor w/ 39 levels "AD","AL","AT",..: 10 10 10 10 10 10 10 10 10 10 ... .. .. .. ..$ station_start_date : Factor w/ 2409 levels "1900-01-01","1951-04-01",..: 152 1184 1577 1132 744 328 1202 1555 1148 407 ... .. .. .. ..$ station_end_date : Factor w/ 864 levels "","1975-02-06",..: 1 1 1 579 1 1 1 1 1 1 ... .. .. .. ..$ type_of_station : Factor w/ 5 levels "","Background",..: 2 2 2 2 2 2 2 2 2 2 ... .. .. .. ..$ station_type_of_area : Factor w/ 4 levels "rural","suburban",..: 1 1 1 1 1 1 1 1 1 1 ... .. .. .. ..$ street_type : Factor w/ 5 levels "","Canyon street: L/H < 1.5",..: 4 1 1 1 1 1 1 1 1 1 ... .. .. .. ..$ annual_mean_PM10 : num [1:69] 20.9 21.8 16.5 20.3 23.3 ... .. .. ..@ coords.nrs : num(0) .. .. ..@ coords : num [1:69, 1:2] 538709 545414 665711 551796 815738 ... .. .. .. ..- attr(*, "dimnames")=List of 2 .. .. .. .. ..$ : chr [1:69] "DESH001" "DENI063" "DEBY109" "DEUB038" ... .. .. .. .. ..$ : chr [1:2] "coords.x1" "coords.x2" .. .. ..@ bbox : num [1:2, 1:2] 307809 5295752 907375 6086661 .. .. .. ..- attr(*, "dimnames")=List of 2 .. .. .. .. ..$ : chr [1:2] "coords.x1" "coords.x2" .. .. .. .. ..$ : chr [1:2] "min" "max" .. .. ..@ proj4string:Formal class 'CRS' [package "sp"] with 1 slot .. .. .. .. ..@ projargs: chr "+init=epsg:32632 +proj=utm +zone=32 +datum=WGS84 +units=m +no_defs +ellps=WGS84 +towgs84=0,0,0" ..@ time :An ?xts? object on 2005-01-01/2005-12-31 containing: Data: int [1:365, 1] 5115 5116 5117 5118 5119 5120 5121 5122 5123 5124 ... - attr(*, "dimnames")=List of 2 ..$ : NULL ..$ : chr "..1" Indexed by objects of class: [POSIXct,POSIXt] TZ: GMT xts Attributes: NULL ..@ endTime: POSIXct[1:365], format: "2005-01-02" "2005-01-03" "2005-01-04" "2005-01-05" ...
EEA, airbase v6
data(DE_RB_2005) str(DE_RB_2005)
data(DE_RB_2005) str(DE_RB_2005)
Estimation of the spatio-temporal anisotropy without an underlying spatio-temporal model. Different methods are implemented using a linear model to predict the temporal gamma values or the ratio of the ranges of a spatial and temporal variogram model or a spatial variogram model to predict the temporal gamma values or the spatio-temporal anisotropy value as used in a metric spatio-temporal variogram.
estiStAni(empVgm, interval, method = "linear", spatialVgm, temporalVgm, s.range=NA, t.range=NA)
estiStAni(empVgm, interval, method = "linear", spatialVgm, temporalVgm, s.range=NA, t.range=NA)
empVgm |
An empirical spatio-temporal variogram. |
interval |
A search interval for the optimisation of the spatio-temporal anisotropy parameter |
method |
A character string determining the method to be used (one of |
spatialVgm |
A spatial variogram definition from the call to |
temporalVgm |
A temporal variogram definition from the call to |
s.range |
A spatial cutoff value applied to the empirical variogram |
t.range |
A temporal cutoff value applied to the empirical variogram |
A linear model is fitted to the pure spatial gamma values based on the spatial distances. An optimal scaling is searched to stretch the temporal distances such that the linear model explains best the pure temporal gamma values. This assumes (on average) a linear relationship between distance and gamma, hence it is advisable to use only those pairs of pure spatial (pure temporal) distance and gamma value that show a considerable increase (i.e. drop all values beyond the range by setting values for s.range
and t.range
).
A spatial and temporal variogram model is fitted to the pure spatial and temporal gamma values respectively. The spatio-temporal anisotropy estimate is the ratio of the spatial range over the temporal range.
A spatial variogram model is fitted to the pure spatial gamma values. An optimal scaling is used to stretch the temporal distances such that the spatial variogram model explains best the pure temporal gamma values.
A metric spatio-temporal variogram model is fitted with joint
component according to the defined spatial variogram spatialVgm
. The starting value of stAni
is the mean of the interval
parameter (see vgmST
for the metric variogram definition). The spatio-temporal anisotropy as estimated in the spatio-temporal variogram is returned. Note that the parameter interval
is only used to set the starting value. Hence, the estimate might exceed the given interval.
A scalar representing the spatio-temporal anisotropy estimate.
Different methods might lead to very different estimates. All but the linear
approach are sensitive to the variogram model selection.
Benedikt Graeler
data(vv) estiStAni(vv, c(10, 150)) estiStAni(vv, c(10, 150), "vgm", vgm(80, "Sph", 120, 20))
data(vv) estiStAni(vv, c(10, 150)) estiStAni(vv, c(10, 150), "vgm", vgm(80, "Sph", 120, 20))
All spatio-temporal variogram models have a different set of parameters. These functions extract the parameters and their names from the spatio-temporal variogram model. Note, this function is as well used to pass the parameters to the optim function. The arguments lower and upper passed to optim should follow the same structure.
extractPar(model) extractParNames(model)
extractPar(model) extractParNames(model)
model |
a spatio-temporal variogram model from |
A named numeric vector of parameters or a vector of characters holding the parameters' names.
Benedikt Graeler
fit.StVariogram
and vgmST
sumMetricModel <- vgmST("sumMetric", space=vgm(30, "Sph", 200, 6), time =vgm(30, "Sph", 15, 7), joint=vgm(60, "Exp", 84, 22), stAni=100) extractPar(sumMetricModel) extractParNames(sumMetricModel)
sumMetricModel <- vgmST("sumMetric", space=vgm(30, "Sph", 200, 6), time =vgm(30, "Sph", 15, 7), joint=vgm(60, "Exp", 84, 22), stAni=100) extractPar(sumMetricModel) extractParNames(sumMetricModel)
Fit a Linear Model of Coregionalization to a Multivariable Sample Variogram; in case of a single variogram model (i.e., no nugget) this is equivalent to Intrinsic Correlation
fit.lmc(v, g, model, fit.ranges = FALSE, fit.lmc = !fit.ranges, correct.diagonal = 1.0, ...)
fit.lmc(v, g, model, fit.ranges = FALSE, fit.lmc = !fit.ranges, correct.diagonal = 1.0, ...)
v |
multivariable sample variogram, output of variogram |
g |
gstat object, output of gstat |
model |
variogram model, output of vgm; if supplied this value is used as initial value for each fit |
fit.ranges |
logical; determines whether the range coefficients (excluding that of the nugget component) should be fitted; or logical vector: determines for each range parameter of the variogram model whether it should be fitted or fixed. |
fit.lmc |
logical; if TRUE, each coefficient matrices of partial sills is guaranteed to be positive definite |
correct.diagonal |
multiplicative correction factor to be applied to partial sills of direct variograms only; the default value, 1.0, does not correct. If you encounter problems with singular covariance matrices during cokriging or cosimulation, you may want to try to increase this to e.g. 1.01 |
... |
parameters that get passed to fit.variogram |
returns an object of class gstat
, with fitted variograms;
This function does not use the iterative procedure proposed by M. Goulard and M. Voltz (Math. Geol., 24(3): 269-286; reproduced in Goovaerts' 1997 book) but uses simply two steps: first, each variogram model is fitted to a direct or cross variogram; next each of the partial sill coefficient matrices is approached by its in least squares sense closest positive definite matrices (by setting any negative eigenvalues to zero).
The argument correct.diagonal
was introduced by experience: by
zeroing the negative eigenvalues for fitting positive definite partial
sill matrices, apparently still perfect correlation may result, leading
to singular cokriging/cosimulation matrices. If someone knows of a more
elegant way to get around this, please let me know.
Edzer Pebesma
variogram, vgm, fit.variogram,
demo(cokriging)
Fits a spatio-temporal variogram of a given type to spatio-temporal sample variogram.
fit.StVariogram(object, model, ..., method = "L-BFGS-B", lower, upper, fit.method = 6, stAni=NA, wles)
fit.StVariogram(object, model, ..., method = "L-BFGS-B", lower, upper, fit.method = 6, stAni=NA, wles)
object |
The spatio-temporal sample variogram. Typically output from |
model |
The desired spatio-temporal model defined through |
... |
further arguments passed to |
lower |
Lower limits used by optim. If missing, the smallest well defined values are used (mostly near 0). |
upper |
Upper limits used by optim. If missing, the largest well defined values are used (mostly |
method |
fit method, pass to |
fit.method |
an integer between 0 and 13 determine the fitting routine (i.e. weighting of the squared residuals in the LSE). Values 0 to 6 correspond with the pure spatial version (see |
stAni |
The spatio-temporal anisotropy that is used in the weighting. Might be missing if the desired spatio-temporal variogram model does contain a spatio-temporal anisotropy parameter (this might cause bad convergence behaviour). The default is |
wles |
Should be missing; only for backwards compatibility, |
The following list summarizes the meaning of the fit.method
argument which is essential a weighting of the squared residuals in the least-squares estimation. Please note, that weights based on the models gamma value might fail to converge properly due to the dependence of weights on the variogram estimate:
fit.method = 0
no fitting, however the MSE between the provided variogram model and sample variogram surface is calculated.
fit.method = 1
Number of pairs in the spatio-temporal bin:
fit.method = 2
Number of pairs in the spatio-temporal bin divided by the square of the current variogram model's value:
fit.method = 3
Same as fit.method = 1
for compatibility with fit.variogram
but as well evaluated in R.
fit.method = 4
Same as fit.method = 2
for compatibility with fit.variogram
but as well evaluated in R.
fit.method = 5
Reserved for REML for compatibility with fit.variogram
, not yet implemented.
fit.method = 6
No weights.
fit.method = 7
Number of pairs in the spatio-temporal bin divided by the square of the bin's metric distance. If stAni
is not specified, the model's parameter is used to calculate the metric distance across space and time:
fit.method = 8
Number of pairs in the spatio-temporal bin divided by the square of the bin's spatial distance. . Note that the 0 distances are replaced by the smallest non-zero distances to avoid division by zero.
fit.method = 9
Number of pairs in the spatio-temporal bin divided by the square of the bin's temporal distance. . Note that the 0 distances are replaced by the smallest non-zero distances to avoid division by zero.
fit.method = 10
Reciprocal of the square of the current variogram model's value:
fit.method = 11
Reciprocal of the square of the bin's metric distance. If stAni
is not specified, the model's parameter is used to calculate the metric distance across space and time:
fit.method = 12
Reciprocal of the square of the bin's spatial distance. . Note that the 0 distances are replaced by the smallest non-zero distances to avoid division by zero.
fit.method = 13
Reciprocal of the square of the bin's temporal distance. . Note that the 0 distances are replaced by the smallest non-zero distances to avoid division by zero.
See also Table 4.2 in the gstat manual for the original spatial version.
Returns a spatio-temporal variogram model, as S3 class StVariogramModel. It carries the temporal and spatial unit as attributes "temporal unit"
and "spatial unit"
in order to allow krigeST
to adjust for different units. The units are obtained from the provided empirical variogram. Further attributes are the optim output "optim.output"
and the always not weighted mean squared error "MSE"
.
Benedikt Graeler
fit.variogram
for the pure spatial case. extractParNames
helps to understand the parameter structure of spatio-temporal variogram models.
# separable model: spatial and temporal sill will be ignored # and kept constant at 1-nugget respectively. A joint sill is used. ## Not run: separableModel <- vgmST("separable", method = "Nelder-Mead", # no lower & upper needed space=vgm(0.9,"Exp", 123, 0.1), time =vgm(0.9,"Exp", 2.9, 0.1), sill=100) data(vv) separableModel <- fit.StVariogram(vv, separableModel, method="L-BFGS-B", lower=c(10,0,0.01,0,1), upper=c(500,1,20,1,200)) plot(vv, separableModel) ## End(Not run) # dontrun
# separable model: spatial and temporal sill will be ignored # and kept constant at 1-nugget respectively. A joint sill is used. ## Not run: separableModel <- vgmST("separable", method = "Nelder-Mead", # no lower & upper needed space=vgm(0.9,"Exp", 123, 0.1), time =vgm(0.9,"Exp", 2.9, 0.1), sill=100) data(vv) separableModel <- fit.StVariogram(vv, separableModel, method="L-BFGS-B", lower=c(10,0,0.01,0,1), upper=c(500,1,20,1,200)) plot(vv, separableModel) ## End(Not run) # dontrun
Fit ranges and/or sills from a simple or nested variogram model to a sample variogram
fit.variogram(object, model, fit.sills = TRUE, fit.ranges = TRUE, fit.method = 7, debug.level = 1, warn.if.neg = FALSE, fit.kappa = FALSE)
fit.variogram(object, model, fit.sills = TRUE, fit.ranges = TRUE, fit.method = 7, debug.level = 1, warn.if.neg = FALSE, fit.kappa = FALSE)
object |
sample variogram, output of variogram |
model |
variogram model, output of vgm; see Details below
for details on how |
fit.sills |
logical; determines whether the partial sill coefficients (including nugget variance) should be fitted; or logical vector: determines for each partial sill parameter whether it should be fitted or fixed. |
fit.ranges |
logical; determines whether the range coefficients (excluding that of the nugget component) should be fitted; or logical vector: determines for each range parameter whether it should be fitted or fixed. |
fit.method |
fitting method, used by gstat. The default method uses
weights $N_h/h^2$ with $N_h$ the number of point pairs and $h$ the
distance. This criterion is not supported by theory, but by practice.
For other values of |
debug.level |
integer; set gstat internal debug level |
warn.if.neg |
logical; if TRUE a warning is issued whenever a sill value of a direct variogram becomes negative |
fit.kappa |
logical; if |
If any of the initial parameters of model
are NA
,
they are given default values as follows. The range parameter
is given one third of the maximum value of object$dist
.
The nugget value is given the mean value of the first three values
of object$gamma
. The partial sill is given the mean of the
last five values of object$gamma
.
Values for fit.method
are 1: weights equal to
$N_j$; 2: weights equal to $N_j/((gamma(h_j))^2)$; 5 (ignore, use
fit.variogram.reml); 6: unweighted (OLS); 7: $N_j/(h_j^2)$.
(from: http://www.gstat.org/gstat.pdf, table 4.2).
returns a fitted variogram model (of class variogramModel
).
This is a data.frame
with two attributes: (i) singular
a logical attribute that indicates whether the non-linear fit
converged (FALSE), or ended in a singularity (TRUE), and (ii)
SSErr
a numerical attribute with the (weighted) sum of
squared errors of the fitted model. See Notes below.
If fitting the range(s) is part of the job of this function,
the results may well depend on the starting values, given in
argument model
, which is generally the case for non-linear
regression problems. This function uses internal C code, which
uses Levenberg-Marquardt.
If for a direct (i.e. not a cross) variogram a sill parameter (partial sill or nugget) becomes negative, fit.variogram is called again with this parameter set to zero, and with a FALSE flag to further fit this sill. This implies that the search does not move away from search space boundaries.
On singular model fits: If your variogram turns out to be a flat, horizontal or sloping line, then fitting a three parameter model such as the exponential or spherical with nugget is a bit heavy: there's an infinite number of possible combinations of sill and range (both very large) to fit to a sloping line. In this case, the returned, singular model may still be useful: just try and plot it. Gstat converges when the parameter values stabilize, and this may not be the case. Another case of singular model fits happens when a model that reaches the sill (such as the spherical) is fit with a nugget, and the range parameter starts, or converges to a value smaller than the distance of the second sample variogram estimate. In this case, again, an infinite number of possibilities occur essentially for fitting a line through a single (first sample variogram) point. In both cases, fixing one or more of the variogram model parameters may help you out.
The function will accept anisotropic sample variograms as input. It will fit a model for a given direction interval if the sample variogram only includes this direction. It is not possible to fit a multiple direction model to each direction of the sample variogram, in this case the model will be fitted to an average of all directions.
Edzer Pebesma
Pebesma, E.J., 2004. Multivariable geostatistics in S: the gstat package. Computers and Geosciences, 30: 683-691.
library(sp) data(meuse) coordinates(meuse) = ~x+y vgm1 <- variogram(log(zinc)~1, meuse) fit.variogram(vgm1, vgm(1, "Sph", 300, 1)) fit.variogram(vgm1, vgm("Sph")) # optimize the value of kappa in a Matern model, using ugly <<- side effect: f = function(x) attr(m.fit <<- fit.variogram(vgm1, vgm(,"Mat",nugget=NA,kappa=x)),"SSErr") optimize(f, c(0.1, 5)) plot(vgm1, m.fit) # best fit from the (0.3, 0.4, 0.5. ... , 5) sequence: (m <- fit.variogram(vgm1, vgm("Mat"), fit.kappa = TRUE)) attr(m, "SSErr")
library(sp) data(meuse) coordinates(meuse) = ~x+y vgm1 <- variogram(log(zinc)~1, meuse) fit.variogram(vgm1, vgm(1, "Sph", 300, 1)) fit.variogram(vgm1, vgm("Sph")) # optimize the value of kappa in a Matern model, using ugly <<- side effect: f = function(x) attr(m.fit <<- fit.variogram(vgm1, vgm(,"Mat",nugget=NA,kappa=x)),"SSErr") optimize(f, c(0.1, 5)) plot(vgm1, m.fit) # best fit from the (0.3, 0.4, 0.5. ... , 5) sequence: (m <- fit.variogram(vgm1, vgm("Mat"), fit.kappa = TRUE)) attr(m, "SSErr")
Fits variogram parameters (nugget, sill, range) to variogram cloud, using GLS (generalized least squares) fitting. Only for direct variograms.
fit.variogram.gls(formula, data, model, maxiter = 30, eps = .01, trace = TRUE, ignoreInitial = TRUE, cutoff = Inf, plot = FALSE)
fit.variogram.gls(formula, data, model, maxiter = 30, eps = .01, trace = TRUE, ignoreInitial = TRUE, cutoff = Inf, plot = FALSE)
formula |
formula defining the response vector and (possible)
regressors; in case of absence of regressors, use e.g. |
data |
object of class Spatial |
model |
variogram model to be fitted, output of |
maxiter |
maximum number of iterations |
eps |
convergence criterium |
trace |
logical; if TRUE, prints parameter trace |
ignoreInitial |
logical;
if FALSE, initial parameter are taken from model;
if TRUE, initial values of model are
ignored and taken from variogram cloud:
nugget: |
cutoff |
maximum distance up to which point pairs are taken into consideration |
plot |
logical; if TRUE, a plot is returned with variogram cloud and fitted model; else, the fitted model is returned. |
an object of class "variogramModel"; see fit.variogram; if
plot
is TRUE, a plot is returned instead.
Inspired by the code of Mihael Drinovac, which was again inspired by code from Ernst Glatzer, author of package vardiag.
Edzer Pebesma
Mueller, W.G., 1999: Least-squares fitting from the variogram cloud. Statistics and Probability Letters, 43, 93-98.
Mueller, W.G., 2007: Collecting Spatial Data. Springer, Heidelberg.
library(sp) data(meuse) coordinates(meuse) = ~x+y ## Not run: fit.variogram.gls(log(zinc)~1, meuse[1:40,], vgm(1, "Sph", 900,1)) ## End(Not run)
library(sp) data(meuse) coordinates(meuse) = ~x+y ## Not run: fit.variogram.gls(log(zinc)~1, meuse[1:40,], vgm(1, "Sph", 900,1)) ## End(Not run)
Fit Variogram Sills to Data, using REML (only for direct variograms; not for cross variograms)
fit.variogram.reml(formula, locations, data, model, debug.level = 1, set, degree = 0)
fit.variogram.reml(formula, locations, data, model, debug.level = 1, set, degree = 0)
formula |
formula defining the response vector and (possible)
regressors; in case of absence of regressors, use e.g. |
locations |
spatial data locations; a formula with the coordinate variables in the right hand (dependent variable) side. |
data |
data frame where the names in formula and locations are to be found |
model |
variogram model to be fitted, output of |
debug.level |
debug level; set to 65 to see the iteration trace and log likelihood |
set |
additional options that can be set; use |
degree |
order of trend surface in the location, between 0 and 3 |
an object of class "variogramModel"; see fit.variogram
This implementation only uses REML fitting of sill parameters. For each
iteration, an matrix is inverted, with $n$ the number of
observations, so for large data sets this method becomes
demanding. I guess there is much more to likelihood variogram fitting in
package
geoR
, and probably also in nlme
.
Edzer Pebesma
Christensen, R. Linear models for multivariate, Time Series, and Spatial Data, Springer, NY, 1991.
Kitanidis, P., Minimum-Variance Quadratic Estimation of Covariances of Regionalized Variables, Mathematical Geology 17 (2), 195–208, 1985
library(sp) data(meuse) fit.variogram.reml(log(zinc)~1, ~x+y, meuse, model = vgm(1, "Sph", 900,1))
library(sp) data(meuse) fit.variogram.reml(log(zinc)~1, ~x+y, meuse, model = vgm(1, "Sph", 900,1))
Airborne counts of Fulmaris glacialis during the Aug/Sept 1998 and 1999 flights on the Dutch (Netherlands) part of the North Sea (NCP, Nederlands Continentaal Plat).
data(fulmar)
data(fulmar)
This data frame contains the following columns:
year of measurement: 1998 or 1999
x-coordinate in UTM zone 31
y-coordinate in UTM zone 31
sea water depth, in m
distance to coast of the Netherlands, in km.
observed density (number of birds per square km)
Dutch National Institute for Coastal and Marine Management (RIKZ)
E.J. Pebesma, R.N.M. Duin, P.A. Burrough, 2005. Mapping Sea Bird Densities over the North Sea: Spatially Aggregated Estimates and Temporal Changes. Environmetrics 16, (6), p 573-587.
data(fulmar) summary(fulmar) ## Not run: demo(fulmar) ## End(Not run)
data(fulmar) summary(fulmar) ## Not run: demo(fulmar) ## End(Not run)
Given multivariable predictions and prediction (co)variances, calculate contrasts and their (co)variance
get.contr(data, gstat.object, X, ids = names(gstat.object$data))
get.contr(data, gstat.object, X, ids = names(gstat.object$data))
data |
data frame, output of predict |
gstat.object |
object of class |
X |
contrast vector or matrix; the number of variables in
|
ids |
character vector with (selection of) id names, present in data |
From data, we can extract the vector with multivariable
predictions, say $y$, and its
covariance matrix $V$. Given
a contrast matrix in $X$, this function computes the contrast vector
$C=X'y$ and its variance $Var(C)=X'V X$.
a data frame containing for each row in data
the generalized
least squares estimates (named beta.1, beta.2, ...), their
variances (named var.beta.1, var.beta.2, ...) and covariances
(named cov.beta.1.2, cov.beta.1.3, ...)
Edzer Pebesma
Function that creates gstat objects; objects that hold all the information necessary for univariate or multivariate geostatistical prediction (simple, ordinary or universal (co)kriging), or its conditional or unconditional Gaussian or indicator simulation equivalents. Multivariate gstat object can be subsetted.
gstat(g, id, formula, locations, data, model = NULL, beta, nmax = Inf, nmin = 0, omax = 0, maxdist = Inf, force = FALSE, dummy = FALSE, set, fill.all = FALSE, fill.cross = TRUE, variance = "identity", weights = NULL, merge, degree = 0, vdist = FALSE, lambda = 1.0) ## S3 method for class 'gstat' print(x, ...)
gstat(g, id, formula, locations, data, model = NULL, beta, nmax = Inf, nmin = 0, omax = 0, maxdist = Inf, force = FALSE, dummy = FALSE, set, fill.all = FALSE, fill.cross = TRUE, variance = "identity", weights = NULL, merge, degree = 0, vdist = FALSE, lambda = 1.0) ## S3 method for class 'gstat' print(x, ...)
g |
gstat object to append to; if missing, a new gstat object is created |
id |
identifier of new variable; if missing, |
formula |
formula that defines the dependent variable as a linear
model of independent variables; suppose the dependent variable has name
|
locations |
formula with only independent variables that define the
spatial data locations (coordinates), e.g. |
data |
data frame; contains the dependent variable, independent variables, and locations. |
model |
variogram model for this |
beta |
for simple kriging (and simulation based on simple kriging): vector with the trend coefficients (including intercept); if no independent variables are defined the model only contains an intercept and this should be the expected value; for cross variogram computations: mean parameters to be used instead of the OLS estimates |
nmax |
for local kriging: the number of nearest observations that should be used for a kriging prediction or simulation, where nearest is defined in terms of the space of the spatial locations |
nmin |
for local kriging: if the number of nearest observations
within distance |
omax |
maximum number of observations to select per octant (3D) or
quadrant (2D); only relevant if |
maxdist |
for local kriging: only observations within a distance
of |
force |
for local kriging, force neighbourhood selection: in case
|
dummy |
logical; if TRUE, consider this data as a dummy variable (only necessary for unconditional simulation) |
set |
named list with optional parameters to be passed to
gstat (only |
x |
gstat object to print |
fill.all |
logical; if TRUE, fill all of the direct variogram and,
depending on the value of |
fill.cross |
logical; if TRUE, fill all of the cross variograms, if
FALSE fill only all direct variogram model slots in |
variance |
character; variance function to transform to non-stationary covariances; "identity" does not transform, other options are "mu" (Poisson) and "mu(1-mu)" (binomial) |
weights |
numeric vector; if present, covariates are present, and variograms are missing weights are passed to OLS prediction routines resulting in WLS; if variograms are given, weights should be 1/variance, where variance specifies location-specific measurement error; see references section below |
merge |
either character vector of length 2, indicating two ids
that share a common mean; the more general gstat merging of any two
coefficients across variables is obtained when a list is passed, with
each element a character vector of length 4, in the form
|
degree |
order of trend surface in the location, between 0 and 3 |
vdist |
logical; if TRUE, instead of Euclidian distance variogram distance is used for selecting the nmax nearest neighbours, after observations within distance maxdist (Euclidian/geographic) have been pre-selected |
lambda |
test feature; doesn't do anything (yet) |
... |
arguments that are passed to the printing of variogram models only |
to print the full contents of the object g
returned,
use as.list(g)
or print.default(g)
an object of class gstat
, which inherits from list
.
Its components are:
data |
list; each element is a list with the |
model |
list; each element contains a variogram model; names are
those of the elements of |
)
set |
list; named list, corresponding to set |
The function currently copies the data objects into the gstat object, so this may become a large object. I would like to copy only the name of the data frame, but could not get this to work. Any help is appreciated.
Subsetting (see examples) is done using the id
's of the variables,
or using numeric subsets. Subsetted gstat objects only contain cross
variograms if (i) the original gstat object contained them and (ii) the
order of the subset indexes increases, numerically, or given the order
they have in the gstat object.
The merge item may seem obscure. Still, for colocated cokriging, it is
needed. See texts by Goovaerts, Wackernagel, Chiles and Delfiner, or
look for standardised ordinary kriging in the 1992 Deutsch and Journel
or Isaaks and Srivastava. In these cases, two variables share a common
mean parameter. Gstat generalises this case: any two variables may share
any of the regression coefficients; allowing for instance analysis of
covariance models, when variograms were left out (see e.g. R. Christensen's
“Plane answers” book on linear models). The tests directory of the
package contains examples in file merge.R. There is also demo(pcb)
which merges slopes across years, but with year-dependent intercept.
Edzer Pebesma
http://www.gstat.org/ Pebesma, E.J., 2004. Multivariable geostatistics in S: the gstat package. Computers and Geosciences, 30: 683-691.
for kriging with known, varying measurement errors (weights
), see e.g.
Delhomme, J.P. Kriging in the hydrosciences. Advances in Water
Resources, 1(5):251-266, 1978; see also the section Kriging with known
measurement errors in the gstat user's manual, http://www.gstat.org/
library(sp) data(meuse) coordinates(meuse) = ~x+y # let's do some manual fitting of two direct variograms and a cross variogram g <- gstat(id = "ln.zinc", formula = log(zinc)~1, data = meuse) g <- gstat(g, id = "ln.lead", formula = log(lead)~1, data = meuse) # examine variograms and cross variogram: plot(variogram(g)) # enter direct variograms: g <- gstat(g, id = "ln.zinc", model = vgm(.55, "Sph", 900, .05)) g <- gstat(g, id = "ln.lead", model = vgm(.55, "Sph", 900, .05)) # enter cross variogram: g <- gstat(g, id = c("ln.zinc", "ln.lead"), model = vgm(.47, "Sph", 900, .03)) # examine fit: plot(variogram(g), model = g$model, main = "models fitted by eye") # see also demo(cokriging) for a more efficient approach g["ln.zinc"] g["ln.lead"] g[c("ln.zinc", "ln.lead")] g[1] g[2] # Inverse distance interpolation with inverse distance power set to .5: # (kriging variants need a variogram model to be specified) data(meuse.grid) gridded(meuse.grid) = ~x+y meuse.gstat <- gstat(id = "zinc", formula = zinc ~ 1, data = meuse, nmax = 7, set = list(idp = .5)) meuse.gstat z <- predict(meuse.gstat, meuse.grid) spplot(z["zinc.pred"]) # see demo(cokriging) and demo(examples) for further examples, # and the manuals for predict and image # local universal kriging gmeuse <- gstat(id = "log_zinc", formula = log(zinc)~sqrt(dist), data = meuse) # variogram of residuals vmeuse.res <- fit.variogram(variogram(gmeuse), vgm(1, "Exp", 300, 1)) # prediction from local neighbourhoods within radius of 170 m or at least 10 points gmeuse <- gstat(id = "log_zinc", formula = log(zinc)~sqrt(dist), data = meuse, maxdist=170, nmin=10, force=TRUE, model=vmeuse.res) predmeuse <- predict(gmeuse, meuse.grid) spplot(predmeuse)
library(sp) data(meuse) coordinates(meuse) = ~x+y # let's do some manual fitting of two direct variograms and a cross variogram g <- gstat(id = "ln.zinc", formula = log(zinc)~1, data = meuse) g <- gstat(g, id = "ln.lead", formula = log(lead)~1, data = meuse) # examine variograms and cross variogram: plot(variogram(g)) # enter direct variograms: g <- gstat(g, id = "ln.zinc", model = vgm(.55, "Sph", 900, .05)) g <- gstat(g, id = "ln.lead", model = vgm(.55, "Sph", 900, .05)) # enter cross variogram: g <- gstat(g, id = c("ln.zinc", "ln.lead"), model = vgm(.47, "Sph", 900, .03)) # examine fit: plot(variogram(g), model = g$model, main = "models fitted by eye") # see also demo(cokriging) for a more efficient approach g["ln.zinc"] g["ln.lead"] g[c("ln.zinc", "ln.lead")] g[1] g[2] # Inverse distance interpolation with inverse distance power set to .5: # (kriging variants need a variogram model to be specified) data(meuse.grid) gridded(meuse.grid) = ~x+y meuse.gstat <- gstat(id = "zinc", formula = zinc ~ 1, data = meuse, nmax = 7, set = list(idp = .5)) meuse.gstat z <- predict(meuse.gstat, meuse.grid) spplot(z["zinc.pred"]) # see demo(cokriging) and demo(examples) for further examples, # and the manuals for predict and image # local universal kriging gmeuse <- gstat(id = "log_zinc", formula = log(zinc)~sqrt(dist), data = meuse) # variogram of residuals vmeuse.res <- fit.variogram(variogram(gmeuse), vgm(1, "Exp", 300, 1)) # prediction from local neighbourhoods within radius of 170 m or at least 10 points gmeuse <- gstat(id = "log_zinc", formula = log(zinc)~sqrt(dist), data = meuse, maxdist=170, nmin=10, force=TRUE, model=vmeuse.res) predmeuse <- predict(gmeuse, meuse.grid) spplot(predmeuse)
Produces h-scatterplots, where point pairs having specific separation distances are plotted. This function is a wrapper around xyplot.
hscat(formula, data, breaks, pch = 3, cex = .6, mirror = FALSE, variogram.alpha = 0, as.table = TRUE,...)
hscat(formula, data, breaks, pch = 3, cex = .6, mirror = FALSE, variogram.alpha = 0, as.table = TRUE,...)
formula |
specifies the dependent variable |
data |
data where the variable in formula is resolved |
breaks |
distance class boundaries |
pch |
plotting symbol |
cex |
plotting symbol size |
mirror |
logical; duplicate all points mirrored along x=y? (note that correlations are those of the points plotted) |
variogram.alpha |
parameter to be passed as alpha parameter to variogram; if alpha is specified it will only affect xyplot by being passed through ... |
as.table |
logical; if |
... |
parameters, passed to variogram and xyplot |
an object of class trellis; normally the h scatter plot
Data pairs are plotted once, so the h-scatterplot are not symmetric.
Edzer Pebesma
Pebesma, E.J., 2004. Multivariable geostatistics in S: the gstat package. Computers and Geosciences, 30: 683-691.
library(sp) data(meuse) coordinates(meuse) = ~x+y hscat(log(zinc)~1, meuse, c(0, 80, 120, 250, 500, 1000))
library(sp) data(meuse) coordinates(meuse) = ~x+y hscat(log(zinc)~1, meuse, c(0, 80, 120, 250, 500, 1000))
Image gridded data, held in a data frame, keeping the right aspect ratio for axes, and the right cell shape
## S3 method for class 'data.frame' image(x, zcol = 3, xcol = 1, ycol = 2, asp = 1, ...) xyz2img(xyz, zcol = 3, xcol = 1, ycol = 2, tolerance = 10 * .Machine$double.eps)
## S3 method for class 'data.frame' image(x, zcol = 3, xcol = 1, ycol = 2, asp = 1, ...) xyz2img(xyz, zcol = 3, xcol = 1, ycol = 2, tolerance = 10 * .Machine$double.eps)
x |
data frame (or matrix) with x-coordinate, y-coordinate, and z-coordinate in its columns |
zcol |
column number or name of z-variable |
xcol |
column number or name of x-coordinate |
ycol |
column number or name of y-coordinate |
asp |
aspect ratio for the x and y axes |
... |
arguments, passed to image.default |
xyz |
data frame (same as |
tolerance |
maximum allowed deviation for coordinats from being exactly on a regularly spaced grid |
image.data.frame plots an image from gridded data, organized
in arbritrary order, in a data frame. It uses xyz2img and
image.default for this. In the S-Plus version, xyz2img
tries to make an image object with a size such that it will plot with
an equal aspect ratio; for the R version, image.data.frame uses the
asp=1
argument to guarantee this.
xyz2img returns a list with components: z
, a matrix
containing the z-values; x
, the increasing coordinates of the
rows of z
; y
, the increasing coordinates of the columns
of z
. This list is suitable input to image.default.
I wrote this function before I found out about levelplot
,
a Lattice/Trellis function that lets you control the aspect ratio by
the aspect
argument, and that automatically draws a legend, and
therefore I now prefer levelplot over image
. Plotting points
on a levelplots is probably done with providing a panel function and
using lpoints
.
(for S-Plus only – ) it is hard (if not impossible) to get exactly right
cell shapes (e.g., square for a square grid) without altering the size of
the plotting region, but this function tries hard to do so by extending
the image to plot in either x- or y-direction. The larger the grid, the
better the approximation. Geographically correct images can be obtained
by modifiying par("pin")
. Read the examples, image a 2 x 2 grid,
and play with par("pin")
if you want to learn more about this.
Edzer Pebesma
library(sp) data(meuse) data(meuse.grid) g <- gstat(formula=log(zinc)~1,locations=~x+y,data=meuse,model=vgm(1,"Exp",300)) x <- predict(g, meuse.grid) image(x, 4, main="kriging variance and data points") points(meuse$x, meuse$y, pch = "+")
library(sp) data(meuse) data(meuse.grid) g <- gstat(formula=log(zinc)~1,locations=~x+y,data=meuse,model=vgm(1,"Exp",300)) x <- predict(g, meuse.grid) image(x, 4, main="kriging variance and data points") points(meuse$x, meuse$y, pch = "+")
The jura data set from Pierre Goovaerts' book (see references
below). It contains four data.frame
s: prediction.dat, validation.dat
and transect.dat and juragrid.dat, and three data.frame
s with
consistently coded land use and rock type factors, as well as geographic
coordinates. The examples below show how to transform these into
spatial (sp) objects in a local coordinate system and in geographic
coordinates, and how to transform to metric coordinate reference
systems.
data(jura)
data(jura)
The data.frames
prediction.dat and validation.dat contain the following fields:
X coordinate, local grid km
Y coordinate, local grid km
see book and below
see book and below
mg cadmium topsoil
mg cobalt topsoil
mg chromium topsoil
mg copper topsoil
mg nickel topsoil
mg lead topsoil
mg zinc topsoil
The data.frame
juragrid.dat only has the first four fields.
In addition the data.frame
s jura.pred, jura.val and jura.grid also
have inserted third and fourth fields giving geographic coordinates:
Longitude, WGS84 datum
Latitude, WGS84 datum
The points data sets were obtained from http://home.comcast.net/~pgoovaerts/book.html, which seems to be no longer available; the grid data were kindly provided by Pierre Goovaerts.
The following codes were used to convert prediction.dat
and validation.dat
to jura.pred
and jura.val
(see examples below):
Rock Types: 1: Argovian, 2: Kimmeridgian, 3: Sequanian, 4: Portlandian, 5: Quaternary.
Land uses: 1: Forest, 2: Pasture (Weide(land), Wiese, Grasland), 3: Meadow (Wiese, Flur, Matte, Anger), 4: Tillage (Ackerland, bestelltes Land)
Points 22 and 100 in the validation set
(validation.dat[c(22,100),]
) seem not to lie exactly on the
grid originally intended, but are kept as such to be consistent with
the book.
Georeferencing was based on two control points in the Swiss grid system shown as Figure 1 of Atteia et al. (see above) and further points digitized on the tentatively georeferenced scanned map. RMSE 2.4 m. Location of points in the field was less precise.
Data preparation by David Rossiter ([email protected]) and Edzer Pebesma; georeferencing by David Rossiter
Goovaerts, P. 1997. Geostatistics for Natural Resources Evaluation. Oxford Univ. Press, New-York, 483 p. Appendix C describes (and gives) the Jura data set.
Atteia, O., Dubois, J.-P., Webster, R., 1994, Geostatistical analysis of soil contamination in the Swiss Jura: Environmental Pollution 86, 315-327
Webster, R., Atteia, O., Dubois, J.-P., 1994, Coregionalization of trace metals in the soil in the Swiss Jura: European Journal of Soil Science 45, 205-218
data(jura) summary(prediction.dat) summary(validation.dat) summary(transect.dat) summary(juragrid.dat) # the following commands were used to create objects with factors instead # of the integer codes for Landuse and Rock: ## Not run: jura.pred = prediction.dat jura.val = validation.dat jura.grid = juragrid.dat jura.pred$Landuse = factor(prediction.dat$Landuse, labels=levels(juragrid.dat$Landuse)) jura.pred$Rock = factor(prediction.dat$Rock, labels=levels(juragrid.dat$Rock)) jura.val$Landuse = factor(validation.dat$Landuse, labels=levels(juragrid.dat$Landuse)) jura.val$Rock = factor(validation.dat$Rock, labels=levels(juragrid.dat$Rock)) ## End(Not run) # the following commands convert data.frame objects into spatial (sp) objects # in the local grid: require(sp) coordinates(jura.pred) = ~Xloc+Yloc coordinates(jura.val) = ~Xloc+Yloc coordinates(jura.grid) = ~Xloc+Yloc gridded(jura.grid) = TRUE # the following commands convert the data.frame objects into spatial (sp) objects # in WGS84 geographic coordinates # example is given only for jura.pred, do the same for jura.val and jura.grid # EPSG codes can be found by searching make_EPSG() jura.pred <- as.data.frame(jura.pred) coordinates(jura.pred) = ~ long + lat proj4string(jura.pred) = CRS("+init=epsg:4326")
data(jura) summary(prediction.dat) summary(validation.dat) summary(transect.dat) summary(juragrid.dat) # the following commands were used to create objects with factors instead # of the integer codes for Landuse and Rock: ## Not run: jura.pred = prediction.dat jura.val = validation.dat jura.grid = juragrid.dat jura.pred$Landuse = factor(prediction.dat$Landuse, labels=levels(juragrid.dat$Landuse)) jura.pred$Rock = factor(prediction.dat$Rock, labels=levels(juragrid.dat$Rock)) jura.val$Landuse = factor(validation.dat$Landuse, labels=levels(juragrid.dat$Landuse)) jura.val$Rock = factor(validation.dat$Rock, labels=levels(juragrid.dat$Rock)) ## End(Not run) # the following commands convert data.frame objects into spatial (sp) objects # in the local grid: require(sp) coordinates(jura.pred) = ~Xloc+Yloc coordinates(jura.val) = ~Xloc+Yloc coordinates(jura.grid) = ~Xloc+Yloc gridded(jura.grid) = TRUE # the following commands convert the data.frame objects into spatial (sp) objects # in WGS84 geographic coordinates # example is given only for jura.pred, do the same for jura.val and jura.grid # EPSG codes can be found by searching make_EPSG() jura.pred <- as.data.frame(jura.pred) coordinates(jura.pred) = ~ long + lat proj4string(jura.pred) = CRS("+init=epsg:4326")
Function for simple, ordinary or universal kriging (sometimes called external drift kriging), kriging in a local neighbourhood, point kriging or kriging of block mean values (rectangular or irregular blocks), and conditional (Gaussian or indicator) simulation equivalents for all kriging varieties, and function for inverse distance weighted interpolation. For multivariable prediction, see gstat and predict
krige(formula, locations, ...) krige.locations(formula, locations, data, newdata, model, ..., beta, nmax = Inf, nmin = 0, omax = 0, maxdist = Inf, block, nsim = 0, indicators = FALSE, na.action = na.pass, debug.level = 1) krige.spatial(formula, locations, newdata, model, ..., beta, nmax = Inf, nmin = 0, omax = 0, maxdist = Inf, block, nsim = 0, indicators = FALSE, na.action = na.pass, debug.level = 1) krige0(formula, data, newdata, model, beta, y, ..., computeVar = FALSE, fullCovariance = FALSE) idw(formula, locations, ...) idw.locations(formula, locations, data, newdata, nmax = Inf, nmin = 0, omax = 0, maxdist = Inf, block, na.action = na.pass, idp = 2.0, debug.level = 1) idw.spatial(formula, locations, newdata, nmax = Inf, nmin = 0, omax = 0, maxdist = Inf, block = numeric(0), na.action = na.pass, idp = 2.0, debug.level = 1) idw0(formula, data, newdata, y, idp = 2.0)
krige(formula, locations, ...) krige.locations(formula, locations, data, newdata, model, ..., beta, nmax = Inf, nmin = 0, omax = 0, maxdist = Inf, block, nsim = 0, indicators = FALSE, na.action = na.pass, debug.level = 1) krige.spatial(formula, locations, newdata, model, ..., beta, nmax = Inf, nmin = 0, omax = 0, maxdist = Inf, block, nsim = 0, indicators = FALSE, na.action = na.pass, debug.level = 1) krige0(formula, data, newdata, model, beta, y, ..., computeVar = FALSE, fullCovariance = FALSE) idw(formula, locations, ...) idw.locations(formula, locations, data, newdata, nmax = Inf, nmin = 0, omax = 0, maxdist = Inf, block, na.action = na.pass, idp = 2.0, debug.level = 1) idw.spatial(formula, locations, newdata, nmax = Inf, nmin = 0, omax = 0, maxdist = Inf, block = numeric(0), na.action = na.pass, idp = 2.0, debug.level = 1) idw0(formula, data, newdata, y, idp = 2.0)
formula |
formula that defines the dependent variable as a linear
model of independent variables; suppose the dependent variable has name
|
locations |
object of class |
data |
data frame: should contain the dependent variable, independent variables, and coordinates, should be missing if locations contains data. |
newdata |
object of class |
model |
variogram model of dependent variable (or its residuals),
defined by a call to vgm or fit.variogram; for |
beta |
for simple kriging (and simulation based on simple kriging): vector with the trend coefficients (including intercept); if no independent variables are defined the model only contains an intercept and beta should be the simple kriging mean |
nmax |
for local kriging: the number of nearest observations that should be used for a kriging prediction or simulation, where nearest is defined in terms of the space of the spatial locations. By default, all observations are used |
nmin |
for local kriging: if the number of nearest observations
within distance |
omax |
see gstat |
maxdist |
for local kriging: only observations within a distance
of |
block |
block size; a vector with 1, 2 or 3 values containing the size of a rectangular in x-, y- and z-dimension respectively (0 if not set), or a data frame with 1, 2 or 3 columns, containing the points that discretize the block in the x-, y- and z-dimension to define irregular blocks relative to (0,0) or (0,0,0)—see also the details section of predict. By default, predictions or simulations refer to the support of the data values. |
nsim |
integer; if set to a non-zero value, conditional simulation
is used instead of kriging interpolation. For this, sequential Gaussian
or indicator simulation is used (depending on the value of
|
indicators |
logical, only relevant if |
na.action |
function determining what should be done with missing values in 'newdata'. The default is to predict 'NA'. Missing values in coordinates and predictors are both dealt with. |
debug.level |
debug level, passed to predict; use -1 to see progress in percentage, and 0 to suppress all printed information |
... |
for krige: arguments that will be passed to gstat;
for |
idp |
numeric; specify the inverse distance weighting power |
y |
matrix; to krige multiple fields in a single step, pass data
as columns of matrix |
computeVar |
logical; if TRUE, prediction variances will be returned |
fullCovariance |
logical; if FALSE a vector with prediction variances will be returned, if TRUE the full covariance matrix of all predictions will be returned |
Function krige
is a simple wrapper method around gstat
and predict for univariate kriging prediction and conditional
simulation methods available in gstat. For multivariate prediction or
simulation, or for other interpolation methods provided by gstat (such as
inverse distance weighted interpolation or trend surface interpolation)
use the functions gstat and predict directly.
Function idw
performs just as krige
without a model being
passed, but allows direct specification of the inverse distance weighting
power. Don't use with predictors in the formula.
For further details, see predict.
if locations
is not a formula, object of the same class as
newdata
(deriving from Spatial
); else a data frame
containing the coordinates of newdata
. Attributes columns
contain prediction and prediction variance (in case of kriging) or the
abs(nsim)
columns of the conditional Gaussian or indicator
simulations
krige0
and idw0
are alternative functions with reduced
functionality and larger memory requirements; they return numeric vectors
(or matrices, in case of multiple dependent) with predicted values only;
in case computeVar
is TRUE, a list with elements pred
and
var
is returned, containing predictions, and (co)variances (depending
on argument fullCovariance
).
locations specifies which coordinates in data
refer to spatial coordinates
Object locations knows about its own spatial locations
used in case of unconditional simulations; newdata needs to be of class Spatial
Daniel G. Krige is a South African scientist who was a mining engineer
when he first used generalised least squares prediction with spatial
covariances in the 50's. George Matheron coined the term kriging
in the 60's for the action of doing this, although very similar approaches
had been taken in the field of meteorology. Beside being Krige's name,
I consider "krige" to be to "kriging" what "predict" is to "prediction".
Edzer Pebesma
N.A.C. Cressie, 1993, Statistics for Spatial Data, Wiley.
Pebesma, E.J., 2004. Multivariable geostatistics in S: the gstat package. Computers and Geosciences, 30: 683-691.
library(sp) data(meuse) coordinates(meuse) = ~x+y data(meuse.grid) gridded(meuse.grid) = ~x+y m <- vgm(.59, "Sph", 874, .04) # ordinary kriging: x <- krige(log(zinc)~1, meuse, meuse.grid, model = m) spplot(x["var1.pred"], main = "ordinary kriging predictions") spplot(x["var1.var"], main = "ordinary kriging variance") # simple kriging: x <- krige(log(zinc)~1, meuse, meuse.grid, model = m, beta = 5.9) # residual variogram: m <- vgm(.4, "Sph", 954, .06) # universal block kriging: x <- krige(log(zinc)~x+y, meuse, meuse.grid, model = m, block = c(40,40)) spplot(x["var1.pred"], main = "universal kriging predictions") # krige0, using user-defined covariance function and multiple responses in y: # exponential variogram with range 500, defined as covariance function: v = function(x, y = x) { exp(-spDists(coordinates(x),coordinates(y))/500) } # krige two variables in a single pass (using 1 covariance model): y = cbind(meuse$zinc,meuse$copper,meuse$lead,meuse$cadmium) x <- krige0(zinc~1, meuse, meuse.grid, v, y = y) meuse.grid$zinc = x[,1] spplot(meuse.grid["zinc"], main = "zinc") meuse.grid$copper = x[,2] spplot(meuse.grid["copper"], main = "copper") # the following has NOTHING to do with kriging, but -- # return the median of the nearest 11 observations: x = krige(zinc~1, meuse, meuse.grid, set = list(method = "med"), nmax = 11) # get 25%- and 75%-percentiles of nearest 11 obs, as prediction and variance: x = krige(zinc~1, meuse, meuse.grid, nmax = 11, set = list(method = "med", quantile = 0.25)) # get diversity (# of different values) and mode from 11 nearest observations: x = krige(zinc~1, meuse, meuse.grid, nmax = 11, set = list(method = "div"))
library(sp) data(meuse) coordinates(meuse) = ~x+y data(meuse.grid) gridded(meuse.grid) = ~x+y m <- vgm(.59, "Sph", 874, .04) # ordinary kriging: x <- krige(log(zinc)~1, meuse, meuse.grid, model = m) spplot(x["var1.pred"], main = "ordinary kriging predictions") spplot(x["var1.var"], main = "ordinary kriging variance") # simple kriging: x <- krige(log(zinc)~1, meuse, meuse.grid, model = m, beta = 5.9) # residual variogram: m <- vgm(.4, "Sph", 954, .06) # universal block kriging: x <- krige(log(zinc)~x+y, meuse, meuse.grid, model = m, block = c(40,40)) spplot(x["var1.pred"], main = "universal kriging predictions") # krige0, using user-defined covariance function and multiple responses in y: # exponential variogram with range 500, defined as covariance function: v = function(x, y = x) { exp(-spDists(coordinates(x),coordinates(y))/500) } # krige two variables in a single pass (using 1 covariance model): y = cbind(meuse$zinc,meuse$copper,meuse$lead,meuse$cadmium) x <- krige0(zinc~1, meuse, meuse.grid, v, y = y) meuse.grid$zinc = x[,1] spplot(meuse.grid["zinc"], main = "zinc") meuse.grid$copper = x[,2] spplot(meuse.grid["copper"], main = "copper") # the following has NOTHING to do with kriging, but -- # return the median of the nearest 11 observations: x = krige(zinc~1, meuse, meuse.grid, set = list(method = "med"), nmax = 11) # get 25%- and 75%-percentiles of nearest 11 obs, as prediction and variance: x = krige(zinc~1, meuse, meuse.grid, nmax = 11, set = list(method = "med", quantile = 0.25)) # get diversity (# of different values) and mode from 11 nearest observations: x = krige(zinc~1, meuse, meuse.grid, nmax = 11, set = list(method = "div"))
Cross validation functions for simple, ordinary or universal point (co)kriging, kriging in a local neighbourhood.
gstat.cv(object, nfold, remove.all = FALSE, verbose = interactive(), all.residuals = FALSE, ...) krige.cv(formula, locations, ...) krige.cv.locations(formula, locations, data, model = NULL, ..., beta = NULL, nmax = Inf, nmin = 0, maxdist = Inf, nfold = nrow(data), verbose = interactive(), debug.level = 0) krige.cv.spatial(formula, locations, model = NULL, ..., beta = NULL, nmax = Inf, nmin = 0, maxdist = Inf, nfold = nrow(locations), verbose = interactive(), debug.level = 0)
gstat.cv(object, nfold, remove.all = FALSE, verbose = interactive(), all.residuals = FALSE, ...) krige.cv(formula, locations, ...) krige.cv.locations(formula, locations, data, model = NULL, ..., beta = NULL, nmax = Inf, nmin = 0, maxdist = Inf, nfold = nrow(data), verbose = interactive(), debug.level = 0) krige.cv.spatial(formula, locations, model = NULL, ..., beta = NULL, nmax = Inf, nmin = 0, maxdist = Inf, nfold = nrow(locations), verbose = interactive(), debug.level = 0)
object |
object of class gstat; see function gstat |
nfold |
integer; if larger than 1, then apply n-fold cross validation;
if |
remove.all |
logical; if TRUE, remove observations at cross validation locations not only for the first, but for all subsequent variables as well |
verbose |
logical; if FALSE, progress bar is suppressed |
all.residuals |
logical; if TRUE, residuals for all variables are returned instead of for the first variable only |
... |
other arguments that will be passed to predict
in case of |
formula |
formula that defines the dependent variable as a linear
model of independent variables; suppose the dependent variable has name
|
locations |
data object deriving from class |
data |
data frame (deprecated); should contain the dependent variable, independent
variables, and coordinates; only to be provided if |
model |
variogram model of dependent variable (or its residuals), defined by a call to vgm or fit.variogram |
beta |
only for simple kriging (and simulation based on simple kriging); vector with the trend coefficients (including intercept); if no independent variables are defined the model only contains an intercept and this should be the simple kriging mean |
nmax |
for local kriging: the number of nearest observations that should be used for a kriging prediction or simulation, where nearest is defined in terms of the space of the spatial locations. By default, all observations are used |
nmin |
for local kriging: if the number of nearest observations
within distance |
maxdist |
for local kriging: only observations within a distance
of |
debug.level |
print debugging information; 0 suppresses debug information |
Leave-one-out cross validation (LOOCV) visits a data point, and predicts the value at that location by leaving out the observed value, and proceeds with the next data point. (The observed value is left out because kriging would otherwise predict the value itself.) N-fold cross validation makes a partitions the data set in N parts. For all observation in a part, predictions are made based on the remaining N-1 parts; this is repeated for each of the N parts. N-fold cross validation may be faster than LOOCV.
data frame containing the coordinates of data
or those
of the first variable in object
, and columns of prediction and
prediction variance of cross validated data points, observed values,
residuals, zscore (residual divided by kriging standard error), and fold.
If all.residuals
is true, a data frame with residuals for all
variables is returned, without coordinates.
locations specifies which coordinates in data
refer to spatial coordinates
Object locations knows about its own spatial locations
Leave-one-out cross validation seems to be much faster in plain (stand-alone) gstat, apparently quite a bit of the effort is spent moving data around from R to gstat.
Edzer Pebesma
library(sp) data(meuse) coordinates(meuse) <- ~x+y m <- vgm(.59, "Sph", 874, .04) # five-fold cross validation: x <- krige.cv(log(zinc)~1, meuse, m, nmax = 40, nfold=5) bubble(x, "residual", main = "log(zinc): 5-fold CV residuals") # multivariable; thanks to M. Rufino: meuse.g <- gstat(id = "zn", formula = log(zinc) ~ 1, data = meuse) meuse.g <- gstat(meuse.g, "cu", log(copper) ~ 1, meuse) meuse.g <- gstat(meuse.g, model = vgm(1, "Sph", 900, 1), fill.all = TRUE) x <- variogram(meuse.g, cutoff = 1000) meuse.fit = fit.lmc(x, meuse.g) out = gstat.cv(meuse.fit, nmax = 40, nfold = 5) summary(out) out = gstat.cv(meuse.fit, nmax = 40, nfold = c(rep(1,100), rep(2,55))) summary(out) # mean error, ideally 0: mean(out$residual) # MSPE, ideally small mean(out$residual^2) # Mean square normalized error, ideally close to 1 mean(out$zscore^2) # correlation observed and predicted, ideally 1 cor(out$observed, out$observed - out$residual) # correlation predicted and residual, ideally 0 cor(out$observed - out$residual, out$residual)
library(sp) data(meuse) coordinates(meuse) <- ~x+y m <- vgm(.59, "Sph", 874, .04) # five-fold cross validation: x <- krige.cv(log(zinc)~1, meuse, m, nmax = 40, nfold=5) bubble(x, "residual", main = "log(zinc): 5-fold CV residuals") # multivariable; thanks to M. Rufino: meuse.g <- gstat(id = "zn", formula = log(zinc) ~ 1, data = meuse) meuse.g <- gstat(meuse.g, "cu", log(copper) ~ 1, meuse) meuse.g <- gstat(meuse.g, model = vgm(1, "Sph", 900, 1), fill.all = TRUE) x <- variogram(meuse.g, cutoff = 1000) meuse.fit = fit.lmc(x, meuse.g) out = gstat.cv(meuse.fit, nmax = 40, nfold = 5) summary(out) out = gstat.cv(meuse.fit, nmax = 40, nfold = c(rep(1,100), rep(2,55))) summary(out) # mean error, ideally 0: mean(out$residual) # MSPE, ideally small mean(out$residual^2) # Mean square normalized error, ideally close to 1 mean(out$zscore^2) # correlation observed and predicted, ideally 1 cor(out$observed, out$observed - out$residual) # correlation predicted and residual, ideally 0 cor(out$observed - out$residual, out$residual)
Simulating a conditional/unconditional Gaussian random field via kriging and circulant embedding
krigeSimCE(formula, data, newdata, model, n = 1, ext = 2)
krigeSimCE(formula, data, newdata, model, n = 1, ext = 2)
formula |
the formula of the kriging predictor |
data |
spatial data frame that conditions the simulation |
newdata |
locations in space where the Gaussian random field shall be simulated |
model |
a vgm model that defines the spatial covariance structure |
n |
number of simulations |
ext |
extension factor of the circulant embedding, defaults to 2 |
A spatial data frame as defined in newdata
with n
simulations.
Benedikt Graeler
Davies, Tilman M., and David Bryant: "On circulant embedding for Gaussian random fields in R." Journal of Statistical Software 55.9 (2013): 1-21. See i.e. the supplementary files at (retrieved 2018-05-25): https://www.jstatsoft.org/index.php/jss/article/downloadSuppFile/v055i09/v55i09.R
# see demo('circEmbeddingMeuse')
# see demo('circEmbeddingMeuse')
Function for ordinary global and local and trans Gaussian spatio-temporal kriging on point support
krigeST(formula, data, newdata, modelList, beta, y, ..., nmax = Inf, stAni = NULL, computeVar = FALSE, fullCovariance = FALSE, bufferNmax=2, progress=TRUE) krigeSTTg(formula, data, newdata, modelList, y, nmax=Inf, stAni=NULL, bufferNmax=2, progress=TRUE, lambda = 0)
krigeST(formula, data, newdata, modelList, beta, y, ..., nmax = Inf, stAni = NULL, computeVar = FALSE, fullCovariance = FALSE, bufferNmax=2, progress=TRUE) krigeSTTg(formula, data, newdata, modelList, y, nmax=Inf, stAni=NULL, bufferNmax=2, progress=TRUE, lambda = 0)
formula |
formula that defines the dependent variable as a linear
model of independent variables; suppose the dependent variable has name
|
data |
ST object: should contain the dependent variable and independent variables. |
newdata |
ST object with prediction/simulation locations in space and time; should contain attribute columns with the independent variables (if present). |
modelList |
object of class |
y |
matrix; to krige multiple fields in a single step, pass data
as columns of matrix |
beta |
The (known) mean for simple kriging. |
nmax |
The maximum number of neighbouring locations for a spatio-temporal local neighbourhood |
stAni |
a spatio-temporal anisotropy scaling assuming a metric spatio-temporal space. Used only for the selection of the closest neighbours. This scaling needs only to be provided in case the model does not have a stAni parameter, or if a different one should be used for the neighbourhood selection. Mind the correct spatial unit. Currently, no coordinate conversion is made for the neighbourhood selection (i.e. Lat and Lon require a spatio-temporal anisotropy scaling in degrees per second). |
... |
further arguments used for instance to pass the model into vgmAreaST for area-to-point kriging |
computeVar |
logical; if TRUE, prediction variances will be returned |
fullCovariance |
logical; if FALSE a vector with prediction variances will be returned, if TRUE the full covariance matrix of all predictions will be returned |
bufferNmax |
factor with which nmax is multiplied for an extended search radius (default=2). Set to 1 for no extension of the search radius. |
progress |
whether a progress bar shall be printed for local spatio-temporal kriging; default=TRUE |
lambda |
The value of lambda used in the box-cox transformation. |
Function krigeST
is a R implementation of the kriging function from
gstat using spatio-temporal covariance models following the
implementation of krige0
. Function krigeST
offers some
particular methods for ordinary spatio-temporal (ST) kriging. In particular,
it does not support block kriging or kriging in a distance-based
neighbourhood, and does not provide simulation.
If data
is of class sftime
, then newdata
MUST be
of class stars
or sftime
, i.e. mixing form old-style
classes (package spacetime) and new-style classes (sf, stars, sftime)
is not supported.
An object of the same class as newdata
(deriving from
ST
). Attributes columns contain prediction and prediction
variance.
Edzer Pebesma, Benedikt Graeler
Benedikt Graeler, Edzer Pebesma, Gerard Heuvelink. Spatio-Temporal Geostatistics using gstat. The R Journal 8(1), 204–218. https://journal.r-project.org/archive/2016/RJ-2016-014/index.html
N.A.C. Cressie, 1993, Statistics for Spatial Data, Wiley.
Pebesma, E.J., 2004. Multivariable geostatistics in S: the gstat package. Computers and Geosciences, 30: 683-691.
krige0
, gstat
, predict
, krigeTg
library(sp) library(spacetime) sumMetricVgm <- vgmST("sumMetric", space = vgm( 4.4, "Lin", 196.6, 3), time = vgm( 2.2, "Lin", 1.1, 2), joint = vgm(34.6, "Exp", 136.6, 12), stAni = 51.7) data(air) suppressWarnings(proj4string(stations) <- CRS(proj4string(stations))) rural = STFDF(stations, dates, data.frame(PM10 = as.vector(air))) rr <- rural[,"2005-06-01/2005-06-03"] rr <- as(rr,"STSDF") x1 <- seq(from=6,to=15,by=1) x2 <- seq(from=48,to=55,by=1) DE_gridded <- SpatialPoints(cbind(rep(x1,length(x2)), rep(x2,each=length(x1))), proj4string=CRS(proj4string(rr@sp))) gridded(DE_gridded) <- TRUE DE_pred <- STF(sp=as(DE_gridded,"SpatialPoints"), time=rr@time) DE_kriged <- krigeST(PM10~1, data=rr, newdata=DE_pred, modelList=sumMetricVgm) gridded(DE_kriged@sp) <- TRUE stplot(DE_kriged)
library(sp) library(spacetime) sumMetricVgm <- vgmST("sumMetric", space = vgm( 4.4, "Lin", 196.6, 3), time = vgm( 2.2, "Lin", 1.1, 2), joint = vgm(34.6, "Exp", 136.6, 12), stAni = 51.7) data(air) suppressWarnings(proj4string(stations) <- CRS(proj4string(stations))) rural = STFDF(stations, dates, data.frame(PM10 = as.vector(air))) rr <- rural[,"2005-06-01/2005-06-03"] rr <- as(rr,"STSDF") x1 <- seq(from=6,to=15,by=1) x2 <- seq(from=48,to=55,by=1) DE_gridded <- SpatialPoints(cbind(rep(x1,length(x2)), rep(x2,each=length(x1))), proj4string=CRS(proj4string(rr@sp))) gridded(DE_gridded) <- TRUE DE_pred <- STF(sp=as(DE_gridded,"SpatialPoints"), time=rr@time) DE_kriged <- krigeST(PM10~1, data=rr, newdata=DE_pred, modelList=sumMetricVgm) gridded(DE_kriged@sp) <- TRUE stplot(DE_kriged)
conditional/unconditional spatio-temporal simulation based on turning bands
krigeSTSimTB(formula, data, newdata, modelList, nsim, progress = TRUE, nLyrs = 500, tGrid = NULL, sGrid = NULL, ceExt = 2, nmax = Inf)
krigeSTSimTB(formula, data, newdata, modelList, nsim, progress = TRUE, nLyrs = 500, tGrid = NULL, sGrid = NULL, ceExt = 2, nmax = Inf)
formula |
the formula of the kriging predictor |
data |
conditioning data |
newdata |
locations in space and time where the simulation is carried out |
modelList |
the spatio-temporal variogram (from |
nsim |
number of simulations |
progress |
boolean; whether the progress should be shown in progress bar |
nLyrs |
number of layers used in the turning bands approach (default = 500) |
tGrid |
optional explicit temporal griding that shall be used |
sGrid |
optional explicit spatial griding that shall be used |
ceExt |
expansion in the circulant embedding, defaults to 2 |
nmax |
number of nearest neighbours that shall e used, defaults to 'Inf' meaning all available points are used |
a spatio-temporal data frame with nSim
simulations
Benedikt Graeler
Turning bands
Lantuejoul, C. (2002) Geostatistical Simulation: Models and Algorithms. Springer.
Matheron, G. (1973). The intrinsic random functions and their applications. Adv. Appl. Probab., 5, 439-468.
Strokorb, K., Ballani, F., and Schlather, M. (2014) Tail correlation functions of max-stable processes: Construction principles, recovery and diversity of some mixing max-stable processes with identical TCF. Extremes, Submitted.
Turning layers
Schlather, M. (2011) Construction of covariance functions and unconditional simulation of random fields. In Porcu, E., Montero, J.M. and Schlather, M., Space-Time Processes and Challenges Related to Environmental Problems. New York: Springer.
# see demo('circEmbeddingMeuse')
# see demo('circEmbeddingMeuse')
TransGaussian (ordinary) kriging function using Box-Cox transforms
krigeTg(formula, locations, newdata, model = NULL, ..., nmax = Inf, nmin = 0, maxdist = Inf, block = numeric(0), nsim = 0, na.action = na.pass, debug.level = 1, lambda = 1.0)
krigeTg(formula, locations, newdata, model = NULL, ..., nmax = Inf, nmin = 0, maxdist = Inf, block = numeric(0), nsim = 0, na.action = na.pass, debug.level = 1, lambda = 1.0)
formula |
formula that defines the dependent variable as a linear
model of independent variables; suppose the dependent variable has name
|
locations |
object of class |
newdata |
Spatial object with prediction/simulation locations;
the coordinates should have names as defined in |
model |
variogram model of the TRANSFORMED dependent variable, see vgm, or fit.variogram |
nmax |
for local kriging: the number of nearest observations that should be used for a kriging prediction or simulation, where nearest is defined in terms of the space of the spatial locations. By default, all observations are used |
nmin |
for local kriging: if the number of nearest observations
within distance |
maxdist |
for local kriging: only observations within a distance
of |
block |
does not function correctly, afaik |
nsim |
does not function correctly, afaik |
na.action |
function determining what should be done with missing values in 'newdata'. The default is to predict 'NA'. Missing values in coordinates and predictors are both dealt with. |
lambda |
value for the Box-Cox transform |
debug.level |
debug level, passed to predict; use -1 to see progress in percentage, and 0 to suppress all printed information |
... |
other arguments that will be passed to gstat |
Function krigeTg
uses transGaussian kriging as explained in
https://www.math.umd.edu/~bnk/bak/Splus/kriging.html.
As it uses the R/gstat krige function to derive everything, it needs in
addition to ordinary kriging on the transformed scale a simple kriging
step to find m from the difference between the OK and SK prediction
variance, and a kriging/BLUE estimation step to obtain the estimate
of .
For further details, see krige and predict.
an SpatialPointsDataFrame object containing the fields:
m
for the m (Lagrange) parameter for each location;
var1SK.pred
the correction obtained by
muhat
for the mean estimate at each location;
var1SK.var
the simple kriging variance;
var1.pred
the OK prediction on the transformed scale;
var1.var
the OK kriging variance on the transformed scale;
var1TG.pred
the transGaussian kriging predictor;
var1TG.var
the transGaussian kriging variance, obtained by
Edzer Pebesma
N.A.C. Cressie, 1993, Statistics for Spatial Data, Wiley.
library(sp) data(meuse) coordinates(meuse) = ~x+y data(meuse.grid) gridded(meuse.grid) = ~x+y v = vgm(1, "Exp", 300) x1 = krigeTg(zinc~1,meuse,meuse.grid,v, lambda=1) # no transform x2 = krige(zinc~1,meuse,meuse.grid,v) summary(x2$var1.var-x1$var1TG.var) summary(x2$var1.pred-x1$var1TG.pred) lambda = -0.25 m = fit.variogram(variogram((zinc^lambda-1)/lambda ~ 1,meuse), vgm(1, "Exp", 300)) x = krigeTg(zinc~1,meuse,meuse.grid,m,lambda=-.25) spplot(x["var1TG.pred"], col.regions=bpy.colors()) summary(meuse$zinc) summary(x$var1TG.pred)
library(sp) data(meuse) coordinates(meuse) = ~x+y data(meuse.grid) gridded(meuse.grid) = ~x+y v = vgm(1, "Exp", 300) x1 = krigeTg(zinc~1,meuse,meuse.grid,v, lambda=1) # no transform x2 = krige(zinc~1,meuse,meuse.grid,v) summary(x2$var1.var-x1$var1TG.var) summary(x2$var1.pred-x1$var1TG.pred) lambda = -0.25 m = fit.variogram(variogram((zinc^lambda-1)/lambda ~ 1,meuse), vgm(1, "Exp", 300)) x = krigeTg(zinc~1,meuse,meuse.grid,m,lambda=-.25) spplot(x["var1TG.pred"], col.regions=bpy.colors()) summary(meuse$zinc) summary(x$var1TG.pred)
rearrange data frame for plotting with levelplot
map.to.lev(data, xcol = 1, ycol = 2, zcol = c(3, 4), ns = names(data)[zcol])
map.to.lev(data, xcol = 1, ycol = 2, zcol = c(3, 4), ns = names(data)[zcol])
data |
|
xcol |
x-coordinate column number |
ycol |
y-coordinate column number |
zcol |
z-coordinate column number range |
ns |
names of the set of z-columns to be viewed |
data frame with the following elements:
x |
x-coordinate for each row |
y |
y-coordinate for each row |
z |
column vector with each of the elements in columns |
name |
factor; name of each of the stacked |
image.data.frame, krige;
for examples see predict;
levelplot
in package lattice.
This data set gives locations and top soil heavy metal concentrations (ppm), along with a number of soil and landscape variables, collected in a flood plain of the river Meuse, near the village Stein. Heavy metal concentrations are bulk sampled from an area of approximately 15 m x 15 m.
data(meuse.all)
data(meuse.all)
This data frame contains the following columns:
sample number
a numeric vector; x-coordinate (m) in RDM (Dutch topographical map coordinates)
a numeric vector; y-coordinate (m) in RDM (Dutch topographical map coordinates)
topsoil cadmium concentration, ppm.; note that zero cadmium values in the original data set have been shifted to 0.2 (half the lowest non-zero value)
topsoil copper concentration, ppm.
topsoil lead concentration, ppm.
topsoil zinc concentration, ppm.
relative elevation
organic matter, as percentage
flooding frequency class
soil type
lime class
landuse class
distance to river Meuse (metres), as obtained during the field survey
logical; indicates whether this is a sample taken in a pit
logical; indicates whether the sample is part of
the meuse
(i.e., filtered) data set; in addition to the samples
in a pit, an sample (139) with outlying zinc content was removed
logical; indicates whether the sample is used as part of the subset of 98 points in the various interpolation examples of Burrough and McDonnell
sample
refers to original sample number. Eight samples
were left out because they were not indicative for the metal content of
the soil. They were taken in an old pit. One sample contains an outlying
zinc value, which was also discarded for the meuse (155) data set.
The actual field data were collected by Ruud van Rijn and Mathieu Rikken; data compiled for R by Edzer Pebesma
P.A. Burrough, R.A. McDonnell, 1998. Principles of Geographical Information Systems. Oxford University Press.
data(meuse.all) summary(meuse.all)
data(meuse.all) summary(meuse.all)
This data set gives a point set with altitudes, digitized from the 1:10,000 topographical map of the Netherlands.
data(meuse.alt)
data(meuse.alt)
This data frame contains the following columns:
a numeric vector; x-coordinate (m) in RDM (Dutch topographical map coordinates)
a numeric vector; y-coordinate (m) in RDM (Dutch topographical map coordinates)
altitude in m. above NAP (Dutch zero for sea level)
data(meuse.alt) library(lattice) xyplot(y~x, meuse.alt, aspect = "iso")
data(meuse.alt) library(lattice) xyplot(y~x, meuse.alt, aspect = "iso")
Gridded data for the NCP (Nederlands Continentaal Plat, the Dutch part of the North Sea), for a 5 km x 5 km grid; stored as data.frame.
data(ncp.grid)
data(ncp.grid)
This data frame contains the following columns:
x-coordinate, UTM zone 31
y-coordinate, UTM zone 31
sea water depth, m.
distance to the coast of the Netherlands, in km.
identifier for administrative sub-areas
Dutch National Institute for Coastal and Marine Management (RIKZ); data compiled for R by Edzer Pebesma
data(ncp.grid) summary(ncp.grid)
data(ncp.grid) summary(ncp.grid)
Calculate, for a given variogram model, ordinary block kriging standard errors as a function of sampling spaces and block sizes
ossfim(spacings = 1:5, block.sizes = 1:5, model, nmax = 25, debug = 0)
ossfim(spacings = 1:5, block.sizes = 1:5, model, nmax = 25, debug = 0)
spacings |
range of grid (data) spacings to be used |
block.sizes |
range of block sizes to be used |
model |
variogram model, output of |
nmax |
set the kriging neighbourhood size |
debug |
debug level; set to 32 to see a lot of output |
data frame with columns spacing
(the grid spacing),
block.size
(the block size), and kriging.se
(block kriging
standard error)
The idea is old, simple, but still of value. If you want to map a variable with a given accuracy, you will have to sample it. Suppose the variogram of the variable is known. Given a regular sampling scheme, the kriging standard error decreases when either (i) the data spacing is smaller, or (ii) predictions are made for larger blocks. This function helps quantifying this relationship. Ossfim probably refers to “optimal sampling scheme for isarithmic mapping”.
Edzer Pebesma
Burrough, P.A., R.A. McDonnell (1999) Principles of Geographical Information Systems. Oxford University Press (e.g., figure 10.11 on page 261)
Burgess, T.M., R. Webster, A.B. McBratney (1981) Optimal interpolation and isarithmic mapping of soil properties. IV Sampling strategy. The journal of soil science 32(4), 643-660.
McBratney, A.B., R. Webster (1981) The design of optimal sampling schemes for local estimation and mapping of regionalized variables: 2 program and examples. Computers and Geosciences 7: 335-365.
## Not run: x <- ossfim(1:15,1:15, model = vgm(1,"Exp",15)) library(lattice) levelplot(kriging.se~spacing+block.size, x, main = "Ossfim results, variogram 1 Exp(15)") ## End(Not run) # if you wonder about the decrease in the upper left corner of the graph, # try the above with nmax set to 100, or perhaps 200.
## Not run: x <- ossfim(1:15,1:15, model = vgm(1,"Exp",15)) library(lattice) levelplot(kriging.se~spacing+block.size, x, main = "Ossfim results, variogram 1 Exp(15)") ## End(Not run) # if you wonder about the decrease in the upper left corner of the graph, # try the above with nmax set to 100, or perhaps 200.
Data: 126 soil augerings on a 100 x 100m square grid, with 6 columns and 21 rows. Grid is oriented with long axis North-north-west to South-south-east Origin of grid is South-south-east point, 100m outside grid.
Original data are part of a soil survey carried out by P.A. Burrough in 1967. The survey area is located on the chalk downlands on the Berkshire Downs in Oxfordshire, UK. Three soil profile units were recognised on the shallow Rendzina soils; these are Ia - very shallow, grey calcareous soils less than 40cm deep over chalk; Ct - shallow to moderately deep, grey-brown calcareous soils on calcareous colluvium, and Cr: deep, moderately acid, red-brown clayey soils. These soil profile classes were registered at every augering.
In addition, an independent landscape soil map was made by interpolating soil boundaries between these soil types, using information from the changes in landform. Because the soil varies over short distances, this field mapping caused some soil borings to receive a different classification from the classification based on the point data.
Also registered at each auger point were the site elevation (m), the depth to solid chalk rock (in cm) and the depth to lime in cm. Also, the percent clay content, the Munsell colour components of VALUE and CHROMA , and the lime content of the soil (as tested using HCl) were recorded for the top two soil layers (0-20cm and 20-40cm).
Samples of topsoil taken as a bulk sample within a circle of radius 2.5m around each sample point were used for the laboratory determination of Mg (ppm), OM1 %, CEC as mequ/100g air dry soil, pH, P as ppm and K (ppm).
data(oxford)
data(oxford)
This data frame contains the following columns:
profile number
x-coordinate, field, non-projected
y-coordinate, field, non-projected
elevation, m.
soil class, obtained by classifying the soil profile at the sample site
soil class, obtained by looking up the site location in the soil map
Munsell colour component VALUE, 0-20 cm
Munsell colour component CHROMA, 20-40 cm
Lime content (tested using HCl), 0-20 cm
Munsell colour component VALUE, 0-20 cm
Munsell colour component CHROMA, 20-40 cm
Lime content (tested using HCl), 20-40 cm
soil depth, cm
depth to lime, cm
percentage clay, 0-20 cm
percentage clay, 20-40 cm
Magnesium content (ppm), 0-20 cm
organic matter (%), 0-20 cm
CES as mequ/100g air dry soil, 0-20 cm
pH, 0-20 cm
Phosphorous, 0-20 cm, ppm
K (potassium), 0-20 cm, ppm
oxford.jpg
, in the gstat package external directory (see
example below), shows an image of the soil map for the region
P.A. Burrough; compiled for R by Edzer Pebesma
P.A. Burrough, R.A. McDonnell, 1998. Principles of Geographical Information Systems. Oxford University Press.
data(oxford) summary(oxford) # open the following file with a jpg viewer: system.file("external/oxford.jpg", package="gstat")
data(oxford) summary(oxford) # open the following file with a jpg viewer: system.file("external/oxford.jpg", package="gstat")
PCB138 measurements in sediment at the NCP, which is the Dutch part of the North Sea
data(pcb)
data(pcb)
This data frame contains the following columns:
measurement year
x-coordinate; UTM zone 31
y-coordinate; UTM zone 31
distance to coast of the Netherlands, in km.
sea water depth, m.
PCB-138, measured on the sediment fraction smaller than
63 , in
dry matter; BUT SEE NOTE BELOW
year; as factor
A note of caution: The PCB-138 data are provided only to be able to re-run the analysis done in Pebesma and Duin (2004; see references below). If you want to use these data for comparison with PCB measurements elsewhere, or if you want to compare them to regulation standards, or want to use these data for any other purpose, you should first contact mailto:[email protected]. The reason for this is that several normalisations were carried out that are not reported here, nor in the paper below.
Pebesma, E. J., and Duin, R. N. M. (2005). Spatial patterns of temporal change in North Sea sediment quality on different spatial scales. In P. Renard, H. Demougeot-Renard and R. Froidevaux (Eds.), Geostatistics for Environmental Applications: Proceedings of the Fifth European Conference on Geostatistics for Environmental Applications (pp. 367-378): Springer.
data(pcb) library(lattice) xyplot(y~x|as.factor(yf), pcb, aspect = "iso") # demo(pcb)
data(pcb) library(lattice) xyplot(y~x|as.factor(yf), pcb, aspect = "iso") # demo(pcb)
Creates a variogram plot
## S3 method for class 'gstatVariogram' plot(x, model = NULL, ylim, xlim, xlab = "distance", ylab = attr(x, "what"), panel = vgm.panel.xyplot, multipanel = TRUE, plot.numbers = FALSE, scales, ids = x$id, group.id = TRUE, skip, layout, ...) ## S3 method for class 'variogramMap' plot(x, np = FALSE, skip, threshold, ...) ## S3 method for class 'StVariogram' plot(x, model = NULL, ..., col = bpy.colors(), xlab, ylab, map = TRUE, convertMonths = FALSE, as.table = TRUE, wireframe = FALSE, diff = FALSE, all = FALSE)
## S3 method for class 'gstatVariogram' plot(x, model = NULL, ylim, xlim, xlab = "distance", ylab = attr(x, "what"), panel = vgm.panel.xyplot, multipanel = TRUE, plot.numbers = FALSE, scales, ids = x$id, group.id = TRUE, skip, layout, ...) ## S3 method for class 'variogramMap' plot(x, np = FALSE, skip, threshold, ...) ## S3 method for class 'StVariogram' plot(x, model = NULL, ..., col = bpy.colors(), xlab, ylab, map = TRUE, convertMonths = FALSE, as.table = TRUE, wireframe = FALSE, diff = FALSE, all = FALSE)
x |
object obtained from the method variogram, possibly containing directional or cross variograms, space-time variograms and variogram model information |
model |
in case of a single variogram: a variogram model, as obtained from vgm or fit.variogram, to be drawn as a line in the variogram plot; in case of a set of variograms and cross variograms: a list with variogram models; in the spatio-temporal case, a single or a list of spatio-temporal models that will be plotted next to each other for visual comparison. |
ylim |
numeric; vector of length 2, limits of the y-axis |
xlim |
numeric; vector of length 2, limits of the x-axis |
xlab |
character; x-axis label |
ylab |
character; y-axis label |
panel |
panel function |
multipanel |
logical; if TRUE, directional variograms are plotted in different panels, if FALSE, directional variograms are plotted in the same graph, using color, colored lines and symbols to distinguish them |
plot.numbers |
logical or numeric; if TRUE, plot number of point pairs next to each plotted semivariance symbol, if FALSE these are omitted. If numeric, TRUE is assumed and the value is passed as the relative distance to be used between symbols and numeric text values (default 0.03). |
scales |
optional argument that will be passed to |
ids |
ids of the data variables and variable pairs |
group.id |
logical; control for directional multivariate variograms: if TRUE, panels divide direction and colors indicate variables (ids), if FALSE panels divide variables/variable pairs and colors indicate direction |
skip |
logical; can be used to arrange panels, see |
layout |
integer vector; can be used to set panel layout: c(ncol,nrow) |
np |
logical (only for plotting variogram maps); if TRUE, plot number of point pairs, if FALSE plot semivariances |
threshold |
semivariogram map values based on fewer point pairs than threshold will not be plotted |
... |
any arguments that will be passed to the panel plotting functions
(such as |
col |
colors to use |
map |
logical; if TRUE, plot space-time variogram map |
convertMonths |
logical; if TRUE, |
as.table |
controls the plotting order for multiple panels, see |
wireframe |
logical; if TRUE, produce a wireframe plot |
diff |
logical; if TRUE, plot difference between model and sample variogram; ignores |
all |
logical; if TRUE, plot sample and model variogram(s) in single wireframes. |
Please note that in the spatio-temporal case the levelplot and wireframe plots use the spatial distances averaged for each time lag avgDist
. For strongly varying spatial locations over time, please check the distance columns dist
and avgDist
of the spatio-temporal sample variogram. The lattice::cloud
function is one option to plot irregular 3D data.
returns (or plots) the variogram plot
currently, plotting models and/or point pair numbers is not supported when a variogram is both directional and multivariable; also, three-dimensional directional variograms will probably not be displayed correctly.
Edzer Pebesma
variogram, fit.variogram, vgm variogramLine,
library(sp) data(meuse) coordinates(meuse) = ~x+y vgm1 <- variogram(log(zinc)~1, meuse) plot(vgm1) model.1 <- fit.variogram(vgm1,vgm(1,"Sph",300,1)) plot(vgm1, model=model.1) plot(vgm1, plot.numbers = TRUE, pch = "+") vgm2 <- variogram(log(zinc)~1, meuse, alpha=c(0,45,90,135)) plot(vgm2) # the following demonstrates plotting of directional models: model.2 <- vgm(.59,"Sph",926,.06,anis=c(0,0.3)) plot(vgm2, model=model.2) g = gstat(NULL, "zinc < 200", I(zinc<200)~1, meuse) g = gstat(g, "zinc < 400", I(zinc<400)~1, meuse) g = gstat(g, "zinc < 800", I(zinc<800)~1, meuse) # calculate multivariable, directional variogram: v = variogram(g, alpha=c(0,45,90,135)) plot(v, group.id = FALSE, auto.key = TRUE) # id and id pairs panels plot(v, group.id = TRUE, auto.key = TRUE) # direction panels # variogram maps: plot(variogram(g, cutoff=1000, width=100, map=TRUE), main = "(cross) semivariance maps") plot(variogram(g, cutoff=1000, width=100, map=TRUE), np=TRUE, main = "number of point pairs")
library(sp) data(meuse) coordinates(meuse) = ~x+y vgm1 <- variogram(log(zinc)~1, meuse) plot(vgm1) model.1 <- fit.variogram(vgm1,vgm(1,"Sph",300,1)) plot(vgm1, model=model.1) plot(vgm1, plot.numbers = TRUE, pch = "+") vgm2 <- variogram(log(zinc)~1, meuse, alpha=c(0,45,90,135)) plot(vgm2) # the following demonstrates plotting of directional models: model.2 <- vgm(.59,"Sph",926,.06,anis=c(0,0.3)) plot(vgm2, model=model.2) g = gstat(NULL, "zinc < 200", I(zinc<200)~1, meuse) g = gstat(g, "zinc < 400", I(zinc<400)~1, meuse) g = gstat(g, "zinc < 800", I(zinc<800)~1, meuse) # calculate multivariable, directional variogram: v = variogram(g, alpha=c(0,45,90,135)) plot(v, group.id = FALSE, auto.key = TRUE) # id and id pairs panels plot(v, group.id = TRUE, auto.key = TRUE) # direction panels # variogram maps: plot(variogram(g, cutoff=1000, width=100, map=TRUE), main = "(cross) semivariance maps") plot(variogram(g, cutoff=1000, width=100, map=TRUE), np=TRUE, main = "number of point pairs")
Plot a point pairs, identified from a variogram cloud
## S3 method for class 'pointPairs' plot(x, data, xcol = data$x, ycol = data$y, xlab = "x coordinate", ylab = "y coordinate", col.line = 2, line.pch = 0, main = "selected point pairs", ...)
## S3 method for class 'pointPairs' plot(x, data, xcol = data$x, ycol = data$y, xlab = "x coordinate", ylab = "y coordinate", col.line = 2, line.pch = 0, main = "selected point pairs", ...)
x |
object of class "pointPairs", obtained from the function plot.variogramCloud, containing point pair indices |
data |
data frame to which the indices refer (from which the variogram cloud was calculated) |
xcol |
numeric vector with x-coordinates of data |
ycol |
numeric vector with y-coordinates of data |
xlab |
x-axis label |
ylab |
y-axis label |
col.line |
color for lines connecting points |
line.pch |
if non-zero, symbols are also plotted at the middle of
line segments, to mark lines too short to be visible on the plot;
the color used is |
main |
title of plot |
... |
arguments, further passed to |
plots the data locations, with lines connecting the point pairs identified (and refered to by indices in) x
Edzer Pebesma
### The following requires interaction, and is therefore outcommented #data(meuse) #coordinates(meuse) = ~x+y #vgm1 <- variogram(log(zinc)~1, meuse, cloud = TRUE) #pp <- plot(vgm1, id = TRUE) ### Identify the point pairs #plot(pp, data = meuse) # meuse has x and y as coordinates
### The following requires interaction, and is therefore outcommented #data(meuse) #coordinates(meuse) = ~x+y #vgm1 <- variogram(log(zinc)~1, meuse, cloud = TRUE) #pp <- plot(vgm1, id = TRUE) ### Identify the point pairs #plot(pp, data = meuse) # meuse has x and y as coordinates
Plot a sample variogram cloud, possibly with identification of individual point pairs
## S3 method for class 'variogramCloud' plot(x, identify = FALSE, digitize = FALSE, xlim, ylim, xlab, ylab, keep = FALSE, ...)
## S3 method for class 'variogramCloud' plot(x, identify = FALSE, digitize = FALSE, xlim, ylim, xlab, ylab, keep = FALSE, ...)
x |
object of class |
identify |
logical; if TRUE, the plot allows identification of a series of individual point pairs that correspond to individual variogram cloud points (use left mouse button to select; right mouse button ends) |
digitize |
logical; if TRUE, select point pairs by digitizing a region with the mouse (left mouse button adds a point, right mouse button ends) |
xlim |
limits of x-axis |
ylim |
limits of y-axis |
xlab |
x axis label |
ylab |
y axis label |
keep |
logical; if TRUE and |
... |
parameters that are passed through to plot.gstatVariogram (in case of identify = FALSE) or to plot (in case of identify = TRUE) |
If identify
or digitize
is TRUE, a data frame of class
pointPairs
with in its rows the point pairs identified (pairs of
row numbers in the original data set); if identify is F, a plot of the
variogram cloud, which uses plot.gstatVariogram
If in addition to identify
, keep
is also TRUE, an object
of class variogramCloud
is returned, having attached to it attributes
"sel" and "text", which will be used in subsequent calls to plot.variogramCloud
with identify
set to FALSE, to plot the text previously identified.
If in addition to digitize
, keep
is also TRUE, an object of
class variogramCloud
is returned, having attached to it attribute
"poly", which will be used in subsequent calls to plot.variogramCloud
with digitize
set to FALSE, to plot the digitized line.
In both of the keep = TRUE
cases, the attribute ppairs
of
class pointPairs
is present, containing the point pairs identified.
Edzer Pebesma
variogram, plot.gstatVariogram, plot.pointPairs, identify, locator
library(sp) data(meuse) coordinates(meuse) = ~x+y plot(variogram(log(zinc)~1, meuse, cloud=TRUE)) ## commands that require interaction: # x <- variogram(log(zinc)~1, loc=~x+y, data=meuse, cloud=TRUE) # plot(plot(x, identify = TRUE), meuse) # plot(plot(x, digitize = TRUE), meuse)
library(sp) data(meuse) coordinates(meuse) = ~x+y plot(variogram(log(zinc)~1, meuse, cloud=TRUE)) ## commands that require interaction: # x <- variogram(log(zinc)~1, loc=~x+y, data=meuse, cloud=TRUE) # plot(plot(x, identify = TRUE), meuse) # plot(plot(x, digitize = TRUE), meuse)
The function provides the following prediction methods: simple, ordinary, and universal kriging, simple, ordinary, and universal cokriging, point- or block-kriging, and conditional simulation equivalents for each of the kriging methods.
## S3 method for class 'gstat' predict(object, newdata, block = numeric(0), nsim = 0, indicators = FALSE, BLUE = FALSE, debug.level = 1, mask, na.action = na.pass, sps.args = list(n = 500, type = "regular", offset = c(.5, .5)), ...)
## S3 method for class 'gstat' predict(object, newdata, block = numeric(0), nsim = 0, indicators = FALSE, BLUE = FALSE, debug.level = 1, mask, na.action = na.pass, sps.args = list(n = 500, type = "regular", offset = c(.5, .5)), ...)
object |
|
newdata |
data frame with prediction/simulation locations; should
contain columns with the independent variables (if present) and the
coordinates with names as defined in |
block |
block size; a vector with 1, 2 or 3 values containing the size of a rectangular in x-, y- and z-dimension respectively (0 if not set), or a data frame with 1, 2 or 3 columns, containing the points that discretize the block in the x-, y- and z-dimension to define irregular blocks relative to (0,0) or (0,0,0)—see also the details section below. By default, predictions or simulations refer to the support of the data values. |
nsim |
integer; if set to a non-zero value, conditional simulation
is used instead of kriging interpolation. For this, sequential Gaussian
or indicator simulation is used (depending on the value of
|
indicators |
logical; only relevant if |
BLUE |
logical; if TRUE return the BLUE trend estimates only, if FALSE return the BLUP predictions (kriging) |
debug.level |
integer; set gstat internal debug level, see below for useful values. If set to -1 (or any negative value), a progress counter is printed |
mask |
not supported anymore – use na.action; logical or numerical vector; pattern with valid values in newdata (marked as TRUE, non-zero, or non-NA); if mask is specified, the returned data frame will have the same number and order of rows in newdata, and masked rows will be filled with NA's. |
na.action |
function determining what should be done with missing values in 'newdata'. The default is to predict 'NA'. Missing values in coordinates and predictors are both dealt with. |
sps.args |
when newdata is of class |
... |
ignored (but necessary for the S3 generic/method consistency) |
When a non-stationary (i.e., non-constant) mean is used, both for simulation and prediction purposes the variogram model defined should be that of the residual process, not that of the raw observations.
For irregular block kriging, coordinates should discretize the area relative to (0), (0,0) or (0,0,0); the coordinates in newdata should give the centroids around which the block should be located. So, suppose the block is discretized by points (3,3) (3,5) (5,5) and (5,3), we should pass point (4,4) in newdata and pass points (-1,-1) (-1,1) (1,1) (1,-1) to the block argument. Although passing the uncentered block and (0,0) as newdata may work for global neighbourhoods, neighbourhood selection is always done relative to the centroid values in newdata.
If newdata is of class SpatialPolygons or
SpatialPolygonsDataFrame, then the block
average for each of the polygons or polygon sets is
calculated, using spsample to discretize the
polygon(s). Argument sps.args
controls the parameters
used for spsample. The "location" with respect to
which neighbourhood selection is done is for each polygon the
SpatialPolygons polygon label point; if you use local neighbourhoods
you should check out where these points are—it may be well
outside the polygon itself.
The algorithm used by gstat for simulation random fields is the
sequential simulation algorithm. This algorithm scales well to large or
very large fields (e.g., more than $10^6$ nodes). Its power lies in using
only data and simulated values in a local neighbourhood to approximate the
conditional distribution at that location, see nmax
in krige
and gstat. The larger nmax
, the better the approximation,
the smaller nmax
, the faster the simulation process. For selecting
the nearest nmax
data or previously simulated points, gstat uses
a bucket PR quadtree neighbourhood search algorithm; see the reference
below.
For sequential Gaussian or indicator simulations, a random path through
the simulation locations is taken, which is usually done for sequential
simulations. The reason for this is that the local approximation of the
conditional distribution, using only the nmax
neareast observed
(or simulated) values may cause spurious correlations when a regular
path would be followed. Following a single path through the locations,
gstat reuses the expensive results (neighbourhood selection and solution
to the kriging equations) for each of the subsequent simulations when
multiple realisations are requested. You may expect a considerable speed
gain in simulating 1000 fields in a single call to predict,
compared to 1000 calls, each for simulating a single field.
The random number generator used for generating simulations is the native
random number generator of the environment (R, S); fixing randomness by
setting the random number seed with set.seed()
works.
When mean coefficient are not supplied, they are generated as well from their conditional distribution (assuming multivariate normal, using the generalized least squares BLUE estimate and its estimation covariance); for a reference to the algorithm used see Abrahamsen and Benth, Math. Geol. 33(6), page 742 and leave out all constraints.
Memory requirements for sequential simulation: let n be the product of
the number of variables, the number of simulation locations, and the
number of simulations required in a single call. the gstat C function
gstat_predict
requires a table of size n * 12 bytes to pass the
simulations back to R, before it can free n * 4 bytes. Hopefully, R does
not have to duplicate the remaining n * 8 bytes when the coordinates
are added as columns, and when the resulting matrix is coerced to
a data.frame
.
Useful values for debug.level
: 0: suppres any output except
warning and error messages; 1: normal output (default): short data report,
program action and mode, program progress in %, total execution time;
2: print the value of all global variables, all files read and written,
and include source file name and line number in error messages; 4: print
OLS and WLS fit diagnostics; 8: print all data after reading them; 16:
print the neighbourhood selection for each prediction location; 32:
print (generalised) covariance matrices, design matrices, solutions,
kriging weights, etc.; 64: print variogram fit diagnostics (number
of iterations and variogram model in each iteration step) and order
relation violations (indicator kriging values before and after order
relation correction); 512: print block (or area) discretization data
for each prediction location. To combine settings, sum their respective
values. Negative values for debug.level
are equal to positive,
but cause the progress counter to work.
For data with longitude/latitude coordinates (checked by
is.projected
), gstat uses great circle distances in km to compute
spatial distances. The user should make sure that the semivariogram
model used is positive definite on a sphere.
a data frame containing the coordinates of newdata
, and columns
of prediction and prediction variance (in case of kriging) or the
columns of the conditional Gaussian or indicator simulations
Edzer Pebesma
N.A.C. Cressie, 1993, Statistics for Spatial Data, Wiley.
Pebesma, E.J., 2004. Multivariable geostatistics in S: the gstat package. Computers and Geosciences, 30: 683-691.
# generate 5 conditional simulations library(sp) data(meuse) coordinates(meuse) = ~x+y v <- variogram(log(zinc)~1, meuse) m <- fit.variogram(v, vgm(1, "Sph", 300, 1)) plot(v, model = m) set.seed(131) data(meuse.grid) gridded(meuse.grid) = ~x+y sim <- krige(formula = log(zinc)~1, meuse, meuse.grid, model = m, nmax = 10, beta = 5.9, nsim = 5) # for speed -- 10 is too small!! # show all 5 simulation spplot(sim) # calculate generalised least squares residuals w.r.t. constant trend: g <- gstat(NULL, "log.zinc", log(zinc)~1, meuse, model = m) blue0 <- predict(g, newdata = meuse, BLUE = TRUE) blue0$blue.res <- log(meuse$zinc) - blue0$log.zinc.pred bubble(blue0, zcol = "blue.res", main = "GLS residuals w.r.t. constant") # calculate generalised least squares residuals w.r.t. linear trend: m <- fit.variogram(variogram(log(zinc)~sqrt(dist.m), meuse), vgm(1, "Sph", 300, 1)) g <- gstat(NULL, "log.zinc", log(zinc)~sqrt(dist.m), meuse, model = m) blue1 <- predict(g, meuse, BLUE = TRUE) blue1$blue.res <- log(meuse$zinc) - blue1$log.zinc.pred bubble(blue1, zcol = "blue.res", main = "GLS residuals w.r.t. linear trend") # unconditional simulation on a 100 x 100 grid xy <- expand.grid(1:100, 1:100) names(xy) <- c("x","y") gridded(xy) = ~x+y g.dummy <- gstat(formula = z~1, dummy = TRUE, beta = 0, model = vgm(1,"Exp",15), nmax = 10) # for speed -- 10 is too small!! yy <- predict(g.dummy, xy, nsim = 4) # show one realisation: spplot(yy[1]) # show all four: spplot(yy)
# generate 5 conditional simulations library(sp) data(meuse) coordinates(meuse) = ~x+y v <- variogram(log(zinc)~1, meuse) m <- fit.variogram(v, vgm(1, "Sph", 300, 1)) plot(v, model = m) set.seed(131) data(meuse.grid) gridded(meuse.grid) = ~x+y sim <- krige(formula = log(zinc)~1, meuse, meuse.grid, model = m, nmax = 10, beta = 5.9, nsim = 5) # for speed -- 10 is too small!! # show all 5 simulation spplot(sim) # calculate generalised least squares residuals w.r.t. constant trend: g <- gstat(NULL, "log.zinc", log(zinc)~1, meuse, model = m) blue0 <- predict(g, newdata = meuse, BLUE = TRUE) blue0$blue.res <- log(meuse$zinc) - blue0$log.zinc.pred bubble(blue0, zcol = "blue.res", main = "GLS residuals w.r.t. constant") # calculate generalised least squares residuals w.r.t. linear trend: m <- fit.variogram(variogram(log(zinc)~sqrt(dist.m), meuse), vgm(1, "Sph", 300, 1)) g <- gstat(NULL, "log.zinc", log(zinc)~sqrt(dist.m), meuse, model = m) blue1 <- predict(g, meuse, BLUE = TRUE) blue1$blue.res <- log(meuse$zinc) - blue1$log.zinc.pred bubble(blue1, zcol = "blue.res", main = "GLS residuals w.r.t. linear trend") # unconditional simulation on a 100 x 100 grid xy <- expand.grid(1:100, 1:100) names(xy) <- c("x","y") gridded(xy) = ~x+y g.dummy <- gstat(formula = z~1, dummy = TRUE, beta = 0, model = vgm(1,"Exp",15), nmax = 10) # for speed -- 10 is too small!! yy <- predict(g.dummy, xy, nsim = 4) # show one realisation: spplot(yy[1]) # show all four: spplot(yy)
Get or set progress indicator
get_gstat_progress() set_gstat_progress(value)
get_gstat_progress() set_gstat_progress(value)
value |
logical |
return the logical value indicating whether progress bars should be given
Edzer Pebesma
set_gstat_progress(FALSE) get_gstat_progress()
set_gstat_progress(FALSE) get_gstat_progress()
Creates a trellis plot for a range of variogram models, possibly with nugget; and optionally a set of Matern models with varying smoothness.
show.vgms(min = 1e-12 * max, max = 3, n = 50, sill = 1, range = 1, models = as.character(vgm()$short[c(1:17)]), nugget = 0, kappa.range = 0.5, plot = TRUE, ..., as.groups = FALSE)
show.vgms(min = 1e-12 * max, max = 3, n = 50, sill = 1, range = 1, models = as.character(vgm()$short[c(1:17)]), nugget = 0, kappa.range = 0.5, plot = TRUE, ..., as.groups = FALSE)
min |
numeric; start distance value for semivariance calculation beyond the first point at exactly zero |
max |
numeric; maximum distance for semivariance calculation and plotting |
n |
integer; number of points to calculate distance values |
sill |
numeric; (partial) sill(s) of the variogram model |
range |
numeric; range(s) of the variogram model |
models |
character; variogram model(s) to be plotted |
nugget |
numeric; nugget component(s) for variogram models |
kappa.range |
numeric; if this is a vector with more than one element, only a range of Matern models is plotted with these kappa values |
plot |
logical; if TRUE, a plot is returned with the models specified; if FALSE, the data prepared for this plot is returned |
... |
passed on to the call to xyplot |
as.groups |
logical; if TRUE, different models are plotted with different lines in a single panel, else, in one panel per model |
returns a (Trellis) plot of the variogram models requested; see examples. I do currently have strong doubts about the “correctness” of the “Hol” model. The “Spl” model does seem to need a very large range value (larger than the study area?) to be of some value.
If plot is FALSE, a data frame with the data prepared to plot is being returned.
the min
argument is supplied because the variogram
function may be discontinuous at distance zero, surely when a positive
nugget is present.
Edzer Pebesma
show.vgms() show.vgms(models = c("Exp", "Mat", "Gau"), nugget = 0.1) # show a set of Matern models with different smoothness: show.vgms(kappa.range = c(.1, .2, .5, 1, 2, 5, 10), max = 10) # show a set of Exponential class models with different shape parameter: show.vgms(kappa.range = c(.05, .1, .2, .5, 1, 1.5, 1.8, 1.9, 2), models = "Exc", max = 10) # show a set of models with different shape parameter of M. Stein's representation of the Matern: show.vgms(kappa.range = c(.01, .02, .05, .1, .2, .5, 1, 2, 5, 1000), models = "Ste", max = 2)
show.vgms() show.vgms(models = c("Exp", "Mat", "Gau"), nugget = 0.1) # show a set of Matern models with different smoothness: show.vgms(kappa.range = c(.1, .2, .5, 1, 2, 5, 10), max = 10) # show a set of Exponential class models with different shape parameter: show.vgms(kappa.range = c(.05, .1, .2, .5, 1, 1.5, 1.8, 1.9, 2), models = "Exc", max = 10) # show a set of models with different shape parameter of M. Stein's representation of the Matern: show.vgms(kappa.range = c(.01, .02, .05, .1, .2, .5, 1, 2, 5, 1000), models = "Ste", max = 2)
The text below was copied from the original sic2004 event, which is no longer online available.
The variable used in the SIC 2004 exercise is natural ambient radioactivity measured in Germany. The data, provided kindly by the German Federal Office for Radiation Protection (BfS), are gamma dose rates reported by means of the national automatic monitoring network (IMIS).
In the frame of SIC2004, a rectangular area was used to select 1008 monitoring stations (from a total of around 2000 stations). For these 1008 stations, 11 days of measurements have been randomly selected during the last 12 months and the average daily dose rates calculated for each day. Hence, we ended up having 11 data sets.
Prior information (sic.train): 10 data sets of 200 points that are identical for what concerns the locations of the monitoring stations have been prepared. These locations have been randomly selected (see Figure 1). These data sets differ only by their Z values since each set corresponds to 1 day of measurement made during the last 14 months. No information will be provided on the date of measurement. These 10 data sets (10 days of measurements) can be used as prior information to tune the parameters of the mapping algorithms. No other information will be provided about these sets. Participants are free of course to gather more information about the variable in the literature and so on.
The 200 monitoring stations above were randomly taken from a larger set of 1008 stations. The remaining 808 monitoring stations have a topology given in sic.pred. Participants to SIC2004 will have to estimate the values of the variable taken at these 808 locations.
The SIC2004 data (sic.val, variable dayx): The exercise consists in using 200 measurements made on a 11th day (THE data of the exercise) to estimate the values observed at the remaining 808 locations (hence the question marks as symbols in the maps shown in Figure 3). These measurements will be provided only during two weeks (15th of September until 1st of October 2004) on a web page restricted to the participants. The true values observed at these 808 locations will be released only at the end of the exercise to allow participants to write their manuscripts (sic.test, variables dayx and joker).
In addition, a joker data set was released (sic.val, variable joker), which contains an anomaly. The anomaly was generated by a simulation model, and does not represent measured levels.
data(sic2004) #
data(sic2004) #
The data frames contain the following columns:
this integer value is the number (unique value) of the monitoring station chosen by us.
X-coordinate of the monitoring station indicated in meters
Y-coordinate of the monitoring station indicated in meters
mean gamma dose rate measured during 24 hours, at day01. Units are nanoSieverts/hour
same, for day 02
...
...
...
...
...
...
...
...
the data observed at the 11-th day
the joker data set, containing an anomaly not present in the training data
the data set sic.grid provides a set of points on a regular grid (almost
10000 points) covering the area; this is convenient for interpolation;
see the function makegrid
in package sp.
The coordinates have been projected around a point located in the South West of Germany. Hence, a few coordinates have negative values as can be guessed from the Figures below.
Data: the German Federal Office for Radiation Protection (BfS), https://www.bfs.de/EN/home/home_node.html, data provided by Gregoire Dubois, R compilation by Edzer Pebesma.
https://wiki.52north.org/bin/view/AI_GEOSTATS/WebHome
data(sic2004) # FIGURE 1. Locations of the 200 monitoring stations for the 11 data sets. # The values taken by the variable are known. plot(y~x,sic.train,pch=1,col="red", asp=1) # FIGURE 2. Locations of the 808 remaining monitoring stations at which # the values of the variable must be estimated. plot(y~x,sic.pred,pch="?", asp=1, cex=.8) # Figure 2 # FIGURE 3. Locations of the 1008 monitoring stations (exhaustive data sets). # Red circles are used to estimate values located at the questions marks plot(y~x,sic.train,pch=1,col="red", asp=1) points(y~x, sic.pred, pch="?", cex=.8)
data(sic2004) # FIGURE 1. Locations of the 200 monitoring stations for the 11 data sets. # The values taken by the variable are known. plot(y~x,sic.train,pch=1,col="red", asp=1) # FIGURE 2. Locations of the 808 remaining monitoring stations at which # the values of the variable must be estimated. plot(y~x,sic.pred,pch="?", asp=1, cex=.8) # Figure 2 # FIGURE 3. Locations of the 1008 monitoring stations (exhaustive data sets). # Red circles are used to estimate values located at the questions marks plot(y~x,sic.train,pch=1,col="red", asp=1) points(y~x, sic.pred, pch="?", cex=.8)
The text below is copied from the data item at ai-geostats, https://wiki.52north.org/bin/view/AI_GEOSTATS/WebHome
data(sic97) #
data(sic97) #
The data frames contain the following columns:
this integer value is the number (unique value) of the monitoring station
rainfall amount, in 10th of mm
See the pdf that accompanies the original file for a description of the data. The .dxf file with the Swiss border is not included here.
Gregoire Dubois and others.
https://wiki.52north.org/bin/view/AI_GEOSTATS/WebHome
data(sic97) image(demstd) points(sic_full, pch=1) points(sic_obs, pch=3)
data(sic97) image(demstd) points(sic_full, pch=1) points(sic_obs, pch=3)
Plot map matrix of prediction error variances and covariances
spplot.vcov(x, ...)
spplot.vcov(x, ...)
x |
Object of class SpatialPixelsDataFrame or SpatialGridDataFrame, resulting from a krige call with multiple variables (cokriging |
... |
remaining arguments passed to spplot |
The plotted object, of class trellis; see spplot
in
package sp.
Edzer Pebesma
The Südliche Tullnerfeld is a part of the Danube river basin in central Lower Austria and due to its homogeneous aquifer well suited for a model-oriented geostatistical analysis. It contains 36 official water quality measurement stations, which are irregularly spread over the region.
data(tull)
data(tull)
The data frames contain the following columns:
X location in meter
Y location in meter
Station name
Station name
Station name
Station name
Station name
Station name
Station name
Station name
Station name
Station name
Station name
Station name
Station name
Station name
Station name
Station name
Station name
Station name
Station name
Station name
Station name
Station name
Station name
Station name
Station name
Station name
Station name
Station name
Station name
Station name
Station name
Station name
Station name
Station name
Station name
Station name
This data set was obtained on May 6, 2008 from http://www.ifas.jku.at/e5361/index_ger.html. The author of the book that uses it is found at: http://www.ifas.jku.at/e2571/e2604/index_ger.html
Werner G. Müller, Collecting Spatial Data, 3rd edition. Springer Verlag, Heidelberg, 2007
data(tull) # TULLNREG = read.csv("TULLNREG.csv") # I modified tulln36des.csv, such that the first line only contained: x,y # resulting in row.names that reflect the station ID, as in # tull36 = read.csv("tulln36des.csv") # Chlorid92 was read & converted by: #Chlorid92=read.csv("Chlorid92.csv") #Chlorid92$Datum = as.POSIXct(strptime(Chlorid92$Datum, "%d.%m.%y")) summary(tull36) summary(TULLNREG) summary(Chlorid92) # stack & join data to x,y,Date,Chloride form: cl.st = stack(Chlorid92[-1]) names(cl.st) = c("Chloride", "Station") cl.st$Date = rep(Chlorid92$Datum, length(names(Chlorid92))-1) cl.st$x = tull36[match(cl.st[,"Station"], row.names(tull36)), "x"] cl.st$y = tull36[match(cl.st[,"Station"], row.names(tull36)), "y"] # library(lattice) # xyplot(Chloride~Date|Station, cl.st) # xyplot(y~x|Date, cl.st, asp="iso", layout=c(16,11)) summary(cl.st) plot(TULLNREG, pch=3, asp=1) points(y~x, cl.st, pch=16)
data(tull) # TULLNREG = read.csv("TULLNREG.csv") # I modified tulln36des.csv, such that the first line only contained: x,y # resulting in row.names that reflect the station ID, as in # tull36 = read.csv("tulln36des.csv") # Chlorid92 was read & converted by: #Chlorid92=read.csv("Chlorid92.csv") #Chlorid92$Datum = as.POSIXct(strptime(Chlorid92$Datum, "%d.%m.%y")) summary(tull36) summary(TULLNREG) summary(Chlorid92) # stack & join data to x,y,Date,Chloride form: cl.st = stack(Chlorid92[-1]) names(cl.st) = c("Chloride", "Station") cl.st$Date = rep(Chlorid92$Datum, length(names(Chlorid92))-1) cl.st$x = tull36[match(cl.st[,"Station"], row.names(tull36)), "x"] cl.st$y = tull36[match(cl.st[,"Station"], row.names(tull36)), "y"] # library(lattice) # xyplot(Chloride~Date|Station, cl.st) # xyplot(y~x|Date, cl.st, asp="iso", layout=c(16,11)) summary(cl.st) plot(TULLNREG, pch=3, asp=1) points(y~x, cl.st, pch=16)
Calculates the sample variogram from data, or in case of a linear model is given, for the residuals, with options for directional, robust, and pooled variogram, and for irregular distance intervals.
In case spatio-temporal data is provided, the function variogramST
is called with a different set of parameters.
## S3 method for class 'gstat' variogram(object, ...) ## S3 method for class 'formula' variogram(object, locations = coordinates(data), data, ...) ## Default S3 method: variogram(object, locations, X, cutoff, width = cutoff/15, alpha = 0, beta = 0, tol.hor = 90/length(alpha), tol.ver = 90/length(beta), cressie = FALSE, dX = numeric(0), boundaries = numeric(0), cloud = FALSE, trend.beta = NULL, debug.level = 1, cross = TRUE, grid, map = FALSE, g = NULL, ..., projected = TRUE, lambda = 1.0, verbose = FALSE, covariogram = FALSE, PR = FALSE, pseudo = -1) ## S3 method for class 'gstatVariogram' print(x, ...) ## S3 method for class 'variogramCloud' print(x, ...)
## S3 method for class 'gstat' variogram(object, ...) ## S3 method for class 'formula' variogram(object, locations = coordinates(data), data, ...) ## Default S3 method: variogram(object, locations, X, cutoff, width = cutoff/15, alpha = 0, beta = 0, tol.hor = 90/length(alpha), tol.ver = 90/length(beta), cressie = FALSE, dX = numeric(0), boundaries = numeric(0), cloud = FALSE, trend.beta = NULL, debug.level = 1, cross = TRUE, grid, map = FALSE, g = NULL, ..., projected = TRUE, lambda = 1.0, verbose = FALSE, covariogram = FALSE, PR = FALSE, pseudo = -1) ## S3 method for class 'gstatVariogram' print(x, ...) ## S3 method for class 'variogramCloud' print(x, ...)
object |
object of class |
data |
data frame where the names in formula are to be found |
locations |
spatial data locations. For variogram.formula: a
formula with only the coordinate variables in the right hand (explanatory
variable) side e.g. For variogram.default: list with coordinate matrices, each with the number of rows matching that of corresponding vectors in y; the number of columns should match the number of spatial dimensions spanned by the data (1 (x), 2 (x,y) or 3 (x,y,z)). |
... |
any other arguments that will be passed to variogram.default (ignored) |
X |
(optional) list with for each variable the matrix with regressors/covariates; the number of rows should match that of the correspoding element in y, the number of columns equals the number of regressors (including intercept) |
cutoff |
spatial separation distance up to which point pairs are included in semivariance estimates; as a default, the length of the diagonal of the box spanning the data is divided by three. |
width |
the width of subsequent distance intervals into which data point pairs are grouped for semivariance estimates |
alpha |
direction in plane (x,y), in positive degrees clockwise from positive y (North): alpha=0 for direction North (increasing y), alpha=90 for direction East (increasing x); optional a vector of directions in (x,y) |
beta |
direction in z, in positive degrees up from the (x,y) plane; |
optional a vector of directions
tol.hor |
horizontal tolerance angle in degrees |
tol.ver |
vertical tolerance angle in degrees |
cressie |
logical; if TRUE, use Cressie”s robust variogram estimate; if FALSE use the classical method of moments variogram estimate |
dX |
include a pair of data points $y(s_1),y(s_2)$ taken at locations $s_1$ and $s_2$ for sample variogram calculation only when $||x(s_1)-x(s_2)|| < dX$ with and $x(s_i)$ the vector with regressors at location $s_i$, and $||.||$ the 2-norm. This allows pooled estimation of within-strata variograms (use a factor variable as regressor, and dX=0.5), or variograms of (near-)replicates in a linear model (addressing point pairs having similar values for regressors variables) |
boundaries |
numerical vector with distance interval upper boundaries; values should be strictly increasing |
cloud |
logical; if TRUE, calculate the semivariogram cloud |
trend.beta |
vector with trend coefficients, in case they are known. By default, trend coefficients are estimated from the data. |
debug.level |
integer; set gstat internal debug level |
cross |
logical or character; if FALSE, no cross variograms are computed
when object is of class |
formula |
formula, specifying the dependent variable and possible covariates |
x |
object of class |
grid |
grid parameters, if data are gridded (not to be called directly; this is filled automatically) |
map |
logical; if TRUE, and |
g |
NULL or object of class gstat; may be used to pass settable parameters and/or variograms; see example |
projected |
logical; if FALSE, data are assumed to be unprojected,
meaning decimal longitude/latitude. For projected data, Euclidian
distances are computed, for unprojected great circle distances
(km). In |
lambda |
test feature; not working (yet) |
verbose |
logical; print some progress indication |
pseudo |
integer; use pseudo cross variogram for computing time-lagged spatial variograms? -1: find out from coordinates – if they are equal then yes, else no; 0: no; 1: yes. |
covariogram |
logical; compute covariogram instead of variogram? |
PR |
logical; compute pairwise relative variogram (does NOT check whether variable is strictly positive) |
If map is TRUE (or a map is passed), a grid map is returned containing the (cross) variogram map(s). See package sp.
In other cases, an object of class "gstatVariogram" with the following fields:
np |
the number of point pairs for this estimate;
in case of a |
dist |
the average distance of all point pairs considered for this estimate |
gamma |
the actual sample variogram estimate |
dir.hor |
the horizontal direction |
dir.ver |
the vertical direction |
id |
the combined id pair |
If cloud is TRUE: an object of class variogramCloud
, with the field
np
encoding the numbers of the point pair that contributed to a
variogram cloud estimate, as follows. The first point is found by 1 + the
integer division of np by the .BigInt
attribute of the returned
object, the second point by 1 + the remainder of that division.
as.data.frame.variogramCloud returns no np
field,
but does the decoding into:
left |
for variogramCloud: data id (row number) of one of the data pair |
right |
for variogramCloud: data id (row number) of the other data in the pair |
In case of a spatio-temporal variogram is sought see variogramST
for details.
variogram.default
should not be called by users directly,
as it makes many assumptions about the organization of the data, that
are not fully documented (but of course, can be understood from reading
the source code of the other variogram
methods)
Successfully setting gridded() <- TRUE
may trigger a branch that
will fail unless dx and dy are identical, and not merely similar
to within machine epsilon.
variogram.line
is DEPRECATED; it is and was never meant as a variogram
method, but works automatically as such by the R dispatch system. Use
variogramLine instead.
Edzer Pebesma
Cressie, N.A.C., 1993, Statistics for Spatial Data, Wiley.
Cressie, N., C. Wikle, 2011, Statistics for Spatio-temporal Data, Wiley.
Pebesma, E.J., 2004. Multivariable geostatistics in S: the gstat package. Computers and Geosciences, 30: 683-691.
print.gstatVariogram,
plot.gstatVariogram,
plot.variogramCloud;
for variogram models: vgm,
to fit a variogram model to a sample variogram:
fit.variogram
variogramST
for details on the spatio-temporal sample variogram.
library(sp) data(meuse) # no trend: coordinates(meuse) = ~x+y variogram(log(zinc)~1, meuse) # residual variogram w.r.t. a linear trend: variogram(log(zinc)~x+y, meuse) # directional variogram: variogram(log(zinc)~x+y, meuse, alpha=c(0,45,90,135)) variogram(log(zinc)~1, meuse, width=90, cutoff=1300) # GLS residual variogram: v = variogram(log(zinc)~x+y, meuse) v.fit = fit.variogram(v, vgm(1, "Sph", 700, 1)) v.fit set = list(gls=1) v g = gstat(NULL, "log-zinc", log(zinc)~x+y, meuse, model=v.fit, set = set) variogram(g) if (require(sf)) { proj4string(meuse) = CRS("+init=epsg:28992") meuse.ll = sf::st_transform(sf::st_as_sf(meuse), sf::st_crs("+proj=longlat +datum=WGS84")) # variogram of unprojected data, using great-circle distances, returning km as units print(variogram(log(zinc) ~ 1, meuse.ll)) }
library(sp) data(meuse) # no trend: coordinates(meuse) = ~x+y variogram(log(zinc)~1, meuse) # residual variogram w.r.t. a linear trend: variogram(log(zinc)~x+y, meuse) # directional variogram: variogram(log(zinc)~x+y, meuse, alpha=c(0,45,90,135)) variogram(log(zinc)~1, meuse, width=90, cutoff=1300) # GLS residual variogram: v = variogram(log(zinc)~x+y, meuse) v.fit = fit.variogram(v, vgm(1, "Sph", 700, 1)) v.fit set = list(gls=1) v g = gstat(NULL, "log-zinc", log(zinc)~x+y, meuse, model=v.fit, set = set) variogram(g) if (require(sf)) { proj4string(meuse) = CRS("+init=epsg:28992") meuse.ll = sf::st_transform(sf::st_as_sf(meuse), sf::st_crs("+proj=longlat +datum=WGS84")) # variogram of unprojected data, using great-circle distances, returning km as units print(variogram(log(zinc) ~ 1, meuse.ll)) }
Generates a semivariance values given a variogram model
variogramLine(object, maxdist, n = 200, min = 1.0e-6 * maxdist, dir = c(1,0,0), covariance = FALSE, ..., dist_vector, debug.level = 0)
variogramLine(object, maxdist, n = 200, min = 1.0e-6 * maxdist, dir = c(1,0,0), covariance = FALSE, ..., dist_vector, debug.level = 0)
object |
variogram model for which we want semivariance function values |
maxdist |
maximum distance for which we want semivariance values |
n |
number of points |
min |
minimum distance; a value slightly larger than zero is usually used to avoid the discontinuity at distance zero if a nugget component is present |
dir |
direction vector: unit length vector pointing the direction in x (East-West), y (North-South) and z (Up-Down) |
covariance |
logical; if TRUE return covariance values, otherwise return semivariance values |
... |
ignored |
dist_vector |
numeric vector or matrix with distance values |
debug.level |
gstat internal debug level |
a data frame of dimension (n
x 2), with columns distance and gamma
(semivariances or covariances), or in case dist_vector
is a matrix, a
conforming matrix with semivariance/covariance values is returned.
variogramLine is used to generate data for plotting a variogram model.
Edzer Pebesma
variogramLine(vgm(5, "Exp", 10, 5), 10, 10) # anisotropic variogram, plotted in E-W direction: variogramLine(vgm(1, "Sph", 10, anis=c(0,0.5)), 10, 10) # anisotropic variogram, plotted in N-S direction: variogramLine(vgm(1, "Sph", 10, anis=c(0,0.5)), 10, 10, dir=c(0,1,0)) variogramLine(vgm(1, "Sph", 10, anis=c(0,0.5)), dir=c(0,1,0), dist_vector = 0.5) variogramLine(vgm(1, "Sph", 10, anis=c(0,0.5)), dir=c(0,1,0), dist_vector = c(0, 0.5, 0.75))
variogramLine(vgm(5, "Exp", 10, 5), 10, 10) # anisotropic variogram, plotted in E-W direction: variogramLine(vgm(1, "Sph", 10, anis=c(0,0.5)), 10, 10) # anisotropic variogram, plotted in N-S direction: variogramLine(vgm(1, "Sph", 10, anis=c(0,0.5)), 10, 10, dir=c(0,1,0)) variogramLine(vgm(1, "Sph", 10, anis=c(0,0.5)), dir=c(0,1,0), dist_vector = 0.5) variogramLine(vgm(1, "Sph", 10, anis=c(0,0.5)), dir=c(0,1,0), dist_vector = c(0, 0.5, 0.75))
Calculates the sample variogram from spatio-temporal data.
variogramST(formula, locations, data, ..., tlags = 0:15, cutoff, width = cutoff/15, boundaries = seq(0, cutoff, width), progress = interactive(), pseudo = TRUE, assumeRegular = FALSE, na.omit = FALSE, cores = 1)
variogramST(formula, locations, data, ..., tlags = 0:15, cutoff, width = cutoff/15, boundaries = seq(0, cutoff, width), progress = interactive(), pseudo = TRUE, assumeRegular = FALSE, na.omit = FALSE, cores = 1)
formula |
formula, specifying the dependent variable. |
locations |
A STFDF or STSDF containing the variable; kept for
compatibility reasons with variogram, either |
data |
|
... |
any other arguments that will be passed to the underlying
|
tlags |
integer; time lags to consider or in case |
cutoff |
spatial separation distance up to which point pairs are included in semivariance estimates; as a default, the length of the diagonal of the box spanning the data is divided by three. |
width |
the width of subsequent distance intervals into which
data point pairs are grouped for semivariance estimates, by default the
|
boundaries |
numerical vector with distance interval upper boundaries; values should be strictly increasing |
progress |
logical; if TRUE, show text progress bar |
pseudo |
integer; use pseudo cross variogram for computing time-lagged spatial variograms? -1: find out from coordinates – if they are equal then yes, else no; 0: no; 1: yes. |
assumeRegular |
logical; whether the time series should be assumed regular. The first time step is assumed to be representative for the whole series. Note, that temporal lags are considered by index, and no check is made whether pairs actually have the desired separating distance. |
na.omit |
shall all |
cores |
number of cores to use in parallel |
The spatio-temporal sample variogram contains besides the fields
np
, dist
and gamma
the spatio-temporal fields,
timelag
, spacelag
and avgDist
, the first of which indicates the time lag
used, the second and third different spatial lags. spacelag
is the midpoint in the spatial
lag intervals as passed by the parameter boundaries
, whereas avgDist
is the average
distance between the point pairs found in a distance interval over all temporal lags (i.e. the
averages of the values dist
per temporal lag.) To compute variograms for space lag and
time lag
, the pseudo cross variogram
is averaged over all time
lagged observation sets
and
available (weighted by the number of pairs involved).
Edzer Pebesma, Benedikt Graeler
Cressie, N.A.C., 1993, Statistics for Spatial Data, Wiley.
Cressie, N., C. Wikle, 2011, Statistics for Spatio-temporal Data, Wiley.
Pebesma, E.J., 2004. Multivariable geostatistics in S: the gstat package. Computers and Geosciences, 30: 683-691.
plot.StVariogram
,
for variogram models: vgmST
,
to fit a spatio-temporal variogram model to a spatio-temporal sample variogram:
fit.StVariogram
# The following spatio-temporal variogram has been calcualted through # vv = variogram(PM10~1, r5to10, width=20, cutoff = 200, tlags=0:5) # in the vignette "st". data(vv) str(vv) plot(vv)
# The following spatio-temporal variogram has been calcualted through # vv = variogram(PM10~1, r5to10, width=20, cutoff = 200, tlags=0:5) # in the vignette "st". data(vv) str(vv) plot(vv)
Generates a surface of semivariance values given a spatio-temporal variogram model (one of separable, productSum, sumMetric, simpleSumMetric or metric)
variogramSurface(model, dist_grid, covariance = FALSE)
variogramSurface(model, dist_grid, covariance = FALSE)
model |
A spatio-temporal variogram model generated through |
dist_grid |
A data.frame with two columns: |
covariance |
Whether the covariance should be computed instead of the variogram (default: FALSE). |
A data.frame with columns spacelag
, timelag
and gamma
.
Benedikt Graeler
See variogramLine
for the spatial version and fit.StVariogram
for the estimation of spatio-temporal variograms.
separableModel <- vgmST("separable", space=vgm(0.86, "Exp", 476, 0.14), time =vgm( 1, "Exp", 3, 0), sill=113) data(vv) if(require(lattice)) { plot(vv, separableModel, wireframe=TRUE, all=TRUE) } # plotting of sample and model variogram plot(vv, separableModel)
separableModel <- vgmST("separable", space=vgm(0.86, "Exp", 476, 0.14), time =vgm( 1, "Exp", 3, 0), sill=113) data(vv) if(require(lattice)) { plot(vv, separableModel, wireframe=TRUE, all=TRUE) } # plotting of sample and model variogram plot(vv, separableModel)
Generates a variogram model, or adds to an existing model.
print.variogramModel
prints the essence of a variogram model.
vgm(psill = NA, model, range = NA, nugget, add.to, anis, kappa = 0.5, ..., covtable, Err = 0) ## S3 method for class 'variogramModel' print(x, ...) ## S3 method for class 'variogramModel' plot(x, cutoff, ..., type = 'l') as.vgm.variomodel(m)
vgm(psill = NA, model, range = NA, nugget, add.to, anis, kappa = 0.5, ..., covtable, Err = 0) ## S3 method for class 'variogramModel' print(x, ...) ## S3 method for class 'variogramModel' plot(x, cutoff, ..., type = 'l') as.vgm.variomodel(m)
psill |
(partial) sill of the variogram model component, or model: see Details |
model |
model type, e.g. "Exp", "Sph", "Gau", or "Mat". Can be a character vector of model types combined with c(), e.g. c("Exp", "Sph"), in which case the best fitting is returned. Calling vgm() without a model argument returns a data.frame with available models. |
range |
range parameter of the variogram model component; in case of anisotropy: major range |
kappa |
smoothness parameter for the Matern class of variogram models |
nugget |
nugget component of the variogram (this basically adds a nugget compontent to the model); if missing, nugget component is omitted |
add.to |
the variogram model to which we want to add a component (structure) |
anis |
anisotropy parameters: see notes below |
x |
a variogram model to print or plot |
... |
arguments that will be passed to |
covtable |
if model is |
Err |
numeric; if larger than zero, the measurement error variance component that will not be included to the kriging equations, i.e. kriging will now smooth the process Y instead of predict the measured Z, where Z=Y+e, and Err is the variance of e |
m |
object of class |
cutoff |
maximum distance up to which variogram values are computed |
type |
plot type |
If only the first argument (psill
) is given a
character
value/vector indicating one or more models, as in vgm("Sph")
,
then this taken as a shorthand form of vgm(NA,"Sph",NA,NA)
,
i.e. a spherical variogram with nugget and unknown parameter values;
see examples below. Read fit.variogram to find out how
NA
variogram parameters are given initial values for a fitting
a model, based on the sample variogram. Package automap
gives further options for automated variogram modelling.
If a single model is passed, an object of class variogramModel
extending data.frame
.
In case a vector ofmodels is passed, an object of class
variogramModelList
which is a list of variogramModel
objects.
When called without a model argument, a data.frame with available models is returned, having two columns: short (abbreviated names, to be used as model argument: "Exp", "Sph" etc) and long (with some description).
as.vgm.variomodel tries to convert an object of class variomodel (geoR) to vgm.
Geometric anisotropy can be modelled for each individual simple model by giving two or five anisotropy parameters, two for two-dimensional and five for three-dimensional data. In any case, the range defined is the range in the direction of the strongest correlation, or the major range. Anisotropy parameters define which direction this is (the main axis), and how much shorter the range is in (the) direction(s) perpendicular to this main axis.
In two dimensions, two parameters define an anisotropy ellipse, say
anis = c(30, 0.5)
. The first parameter, 30
, refers to
the main axis direction: it is the angle for the principal direction
of continuity (measured in degrees, clockwise from positive Y, i.e. North).
The second parameter, 0.5
, is the anisotropy ratio, the ratio
of the minor range to the major range (a value between 0 and 1). So,
in our example, if the range in the major direction (North-East) is 100,
the range in the minor direction (South-East) is 0.5 x 100 = 50.
In three dimensions, five values should be given in the form anis
= c(p,q,r,s,t)
. Now, $p$ is the angle for the principal direction of
continuity (measured in degrees, clockwise from Y, in direction of X),
$q$ is the dip angle for the principal direction of continuity (measured
in positive degrees up from horizontal), $r$ is the third rotation angle
to rotate the two minor directions around the principal direction defined
by $p$ and $q$. A positive angle acts counter-clockwise while looking
in the principal direction. Anisotropy ratios $s$ and $t$ are the ratios
between the major range and each of the two minor ranges. The anisotropy code
was taken from GSLIB. Note that in http://www.gslib.com/sec_gb.html
it is reported that this code has a bug. Quoting from this
site: “The third angle in all GSLIB programs operates in the opposite
direction than specified in the GSLIB book. Explanation - The books
says (pp27) the angle is measured clockwise when looking toward
the origin (from the postive principal direction), but it should be
counter-clockwise. This is a documentation error. Although rarely used,
the correct specification of the third angle is critical if used.”
(Note that anis = c(p,s)
is equivalent to anis = c(p,0,0,s,1)
.)
The implementation in gstat for 2D and 3D anisotropy was taken from the gslib (probably 1992) code. I have seen a paper where it is argued that the 3D anisotropy code implemented in gslib (and so in gstat) is in error, but I have not corrected anything afterwards.
Edzer Pebesma
Pebesma, E.J., 2004. Multivariable geostatistics in S: the gstat package. Computers and Geosciences, 30: 683-691.
Deutsch, C.V. and Journel, A.G., 1998. GSLIB: Geostatistical software library and user's guide, second edition, Oxford University Press.
For the validity of variogram models on the sphere, see Huang, Chunfeng, Haimeng Zhang, and Scott M. Robeson. On the validity of commonly used covariance and variogram functions on the sphere. Mathematical Geosciences 43.6 (2011): 721-733.
show.vgms to view the available models, fit.variogram, variogramLine, variogram for the sample variogram.
vgm() vgm("Sph") vgm(NA, "Sph", NA, NA) vgm(, "Sph") # "Sph" is second argument: NO nugget in this case vgm(10, "Exp", 300) x <- vgm(10, "Exp", 300) vgm(10, "Nug", 0) vgm(10, "Exp", 300, 4.5) vgm(10, "Mat", 300, 4.5, kappa = 0.7) vgm( 5, "Exp", 300, add.to = vgm(5, "Exp", 60, nugget = 2.5)) vgm(10, "Exp", 300, anis = c(30, 0.5)) vgm(10, "Exp", 300, anis = c(30, 10, 0, 0.5, 0.3)) # Matern variogram model: vgm(1, "Mat", 1, kappa=.3) x <- vgm(0.39527463, "Sph", 953.8942, nugget = 0.06105141) x print(x, digits = 3); # to see all components, do print.data.frame(x) vv=vgm(model = "Tab", covtable = variogramLine(vgm(1, "Sph", 1), 1, n=1e4, min = 0, covariance = TRUE)) vgm(c("Mat", "Sph")) vgm(, c("Mat", "Sph")) # no nugget
vgm() vgm("Sph") vgm(NA, "Sph", NA, NA) vgm(, "Sph") # "Sph" is second argument: NO nugget in this case vgm(10, "Exp", 300) x <- vgm(10, "Exp", 300) vgm(10, "Nug", 0) vgm(10, "Exp", 300, 4.5) vgm(10, "Mat", 300, 4.5, kappa = 0.7) vgm( 5, "Exp", 300, add.to = vgm(5, "Exp", 60, nugget = 2.5)) vgm(10, "Exp", 300, anis = c(30, 0.5)) vgm(10, "Exp", 300, anis = c(30, 10, 0, 0.5, 0.3)) # Matern variogram model: vgm(1, "Mat", 1, kappa=.3) x <- vgm(0.39527463, "Sph", 953.8942, nugget = 0.06105141) x print(x, digits = 3); # to see all components, do print.data.frame(x) vv=vgm(model = "Tab", covtable = variogramLine(vgm(1, "Sph", 1), 1, n=1e4, min = 0, covariance = TRUE)) vgm(c("Mat", "Sph")) vgm(, c("Mat", "Sph")) # no nugget
Variogram plots contain symbols and lines; more control over them can be gained by writing your own panel functions, or extending the ones described here; see examples.
vgm.panel.xyplot(x, y, subscripts, type = "p", pch = plot.symbol$pch, col, col.line = plot.line$col, col.symbol = plot.symbol$col, lty = plot.line$lty, cex = plot.symbol$cex, ids, lwd = plot.line$lwd, model = model, direction = direction, labels, shift = shift, mode = mode, ...) panel.pointPairs(x, y, type = "p", pch = plot.symbol$pch, col, col.line = plot.line$col, col.symbol = plot.symbol$col, lty = plot.line$lty, cex = plot.symbol$cex, lwd = plot.line$lwd, pairs = pairs, line.pch = line.pch, ...)
vgm.panel.xyplot(x, y, subscripts, type = "p", pch = plot.symbol$pch, col, col.line = plot.line$col, col.symbol = plot.symbol$col, lty = plot.line$lty, cex = plot.symbol$cex, ids, lwd = plot.line$lwd, model = model, direction = direction, labels, shift = shift, mode = mode, ...) panel.pointPairs(x, y, type = "p", pch = plot.symbol$pch, col, col.line = plot.line$col, col.symbol = plot.symbol$col, lty = plot.line$lty, cex = plot.symbol$cex, lwd = plot.line$lwd, pairs = pairs, line.pch = line.pch, ...)
x |
x coordinates of points in this panel |
y |
y coordinates of points in this panel |
subscripts |
subscripts of points in this panel |
type |
plot type: "l" for connected lines |
pch |
plotting symbol |
col |
symbol and line color (if set) |
col.line |
line color |
col.symbol |
symbol color |
lty |
line type for variogram model |
cex |
symbol size |
ids |
gstat model ids |
lwd |
line width |
model |
variogram model |
direction |
direction vector |
labels |
labels to plot next to points |
shift |
amount to shift the label right of the symbol |
mode |
to be set by calling function only |
line.pch |
symbol type to be used for point of selected point pairs, e.g. to highlight point pairs with distance close to zero |
pairs |
two-column matrix with pair indexes to be highlighted |
... |
parameters that get passed to lpoints |
ignored; the enclosing function returns a plot of class trellis
Edzer Pebesma
library(sp) data(meuse) coordinates(meuse) <- c("x", "y") library(lattice) mypanel = function(x,y,...) { vgm.panel.xyplot(x,y,...) panel.abline(h=var(log(meuse$zinc)), color = 'red') } plot(variogram(log(zinc)~1,meuse), panel = mypanel)
library(sp) data(meuse) coordinates(meuse) <- c("x", "y") library(lattice) mypanel = function(x,y,...) { vgm.panel.xyplot(x,y,...) panel.abline(h=var(log(meuse$zinc)), color = 'red') } plot(variogram(log(zinc)~1,meuse), panel = mypanel)
Compute point-point, point-area or area-area variogram values from point model
vgmArea(x, y = x, vgm, ndiscr = 16, verbose = FALSE, covariance = TRUE)
vgmArea(x, y = x, vgm, ndiscr = 16, verbose = FALSE, covariance = TRUE)
x |
object of class SpatialPoints or SpatialPolygons |
y |
object of class SpatialPoints or SpatialPolygons |
vgm |
variogram model, see vgm |
ndiscr |
number of points to discretize an area, using spsample |
verbose |
give progress bar |
covariance |
logical; compute covariances, rather than semivariances? |
semivariance or covariance matrix of dimension length(x)
x lenght(y)
Edzer Pebesma
library(sp) demo(meuse, ask = FALSE, echo = FALSE) vgmArea(meuse[1:5,], vgm = vgm(1, "Exp", 1000)) # point-point vgmArea(meuse[1:5,], meuse.area, vgm = vgm(1, "Exp", 1000)) # point-area
library(sp) demo(meuse, ask = FALSE, echo = FALSE) vgmArea(meuse[1:5,], vgm = vgm(1, "Exp", 1000)) # point-point vgmArea(meuse[1:5,], meuse.area, vgm = vgm(1, "Exp", 1000)) # point-area
Function that returns the covariances for areas based on spatio-temporal point variograms for use in the spatio-temporal area-to-point kriging
vgmAreaST(x, y = x, model, ndiscrSpace = 16, verbose = FALSE, covariance = TRUE)
vgmAreaST(x, y = x, model, ndiscrSpace = 16, verbose = FALSE, covariance = TRUE)
x |
spatio-temporal data frame |
y |
spatio-temporal data frame |
model |
spatio-temporal variogram model for point support |
ndiscrSpace |
number of discretisation in space |
verbose |
Boolean: default to FALSE, set to TRUE for debugging |
covariance |
Boolean: whether the covariance shall be evaluated, currently disfunction and set to TRUE |
The covariance between 'x' and 'y'.
Benedikt Graeler
# see demo('a2pinST')
# see demo('a2pinST')
Constructs a spatio-temporal variogram of a given type checking for a minimal set of parameters.
vgmST(stModel, ..., space, time, joint, sill, k, nugget, stAni, temporalUnit)
vgmST(stModel, ..., space, time, joint, sill, k, nugget, stAni, temporalUnit)
stModel |
A string identifying the spatio-temporal variogram model (see details below). Only the string before an optional "_" is used to identify the model. This mechanism can be used to identify different fits of the same model ( |
... |
unused, but ensure an exact match of the following parameters. |
space |
A spatial variogram. |
time |
A temporal variogram. |
joint |
A joint spatio-temporal variogram. |
sill |
A joint spatio-temporal sill. |
k |
The weighting of the product in the product-sum model. |
nugget |
A joint spatio-temporal nugget. |
stAni |
A spatio-temporal anisotropy; the number of space units equivalent to one time unit. |
temporalUnit |
length one character vector, indicating the temporal unit (like secs) |
The different implemented spatio-temporal variogram models have the following required parameters (see as well the example section)
A variogram for space
and time
each and a joint spatio-temporal sill
(variograms may have a separate nugget effect, but their joint sill will be 1) generating the call
vgmST("separable", space, time, sill)
A variogram for space
and time
each, and the weighting of product k
generating the call
vgmST("productSum", space, time, k)
A variogram (potentially including a nugget effect) for space
, time
and joint
each and a spatio-temporal anisotropy ratio stAni
generating the call
vgmST("sumMetric", space, time, joint, stAni)
A variogram (without nugget effect) for space
, time
and joint
each, a joint spatio-temporal nugget
effect and a spatio-temporal anisotropy ratio stAni
generating the call
vgmST("simpleSumMetric", space, time, joint, nugget, stAni)
A spatio-temporal joint
variogram (potentially including a nugget effect) and stAni
generating the call
vgmST("metric", joint, stAni)
Returns an S3 object of class StVariogramModel
.
Benedikt Graeler
fit.StVariogram
for fitting, variogramSurface
to plot the variogram and extractParNames
to better understand the parameter structure of spatio-temporal variogram models.
# separable model: spatial and temporal sill will be ignored # and kept constant at 1-nugget respectively. A joint sill is used. separableModel <- vgmST("separable", space=vgm(0.9,"Exp", 147, 0.1), time =vgm(0.9,"Exp", 3.5, 0.1), sill=40) # product sum model: spatial and temporal nugget will be ignored and kept # constant at 0. Only a joint nugget is used. prodSumModel <- vgmST("productSum", space=vgm(39, "Sph", 343, 0), time= vgm(36, "Exp", 3, 0), k=15) # sum metric model: spatial, temporal and joint nugget will be estimated sumMetricModel <- vgmST("sumMetric", space=vgm( 6.9, "Lin", 200, 3.0), time =vgm(10.3, "Lin", 15, 3.6), joint=vgm(37.2, "Exp", 84,11.7), stAni=77.7) # simplified sumMetric model, only a overall nugget is fitted. The spatial, # temporal and jont nuggets are set to 0. simpleSumMetricModel <- vgmST("simpleSumMetric", space=vgm(20,"Lin", 150, 0), time =vgm(20,"Lin", 10, 0), joint=vgm(20,"Exp", 150, 0), nugget=1, stAni=15) # metric model metricModel <- vgmST("metric", joint=vgm(60, "Exp", 150, 10), stAni=60)
# separable model: spatial and temporal sill will be ignored # and kept constant at 1-nugget respectively. A joint sill is used. separableModel <- vgmST("separable", space=vgm(0.9,"Exp", 147, 0.1), time =vgm(0.9,"Exp", 3.5, 0.1), sill=40) # product sum model: spatial and temporal nugget will be ignored and kept # constant at 0. Only a joint nugget is used. prodSumModel <- vgmST("productSum", space=vgm(39, "Sph", 343, 0), time= vgm(36, "Exp", 3, 0), k=15) # sum metric model: spatial, temporal and joint nugget will be estimated sumMetricModel <- vgmST("sumMetric", space=vgm( 6.9, "Lin", 200, 3.0), time =vgm(10.3, "Lin", 15, 3.6), joint=vgm(37.2, "Exp", 84,11.7), stAni=77.7) # simplified sumMetric model, only a overall nugget is fitted. The spatial, # temporal and jont nuggets are set to 0. simpleSumMetricModel <- vgmST("simpleSumMetric", space=vgm(20,"Lin", 150, 0), time =vgm(20,"Lin", 10, 0), joint=vgm(20,"Exp", 150, 0), nugget=1, stAni=15) # metric model metricModel <- vgmST("metric", joint=vgm(60, "Exp", 150, 10), stAni=60)
Precomputed variogram for PM10 in data set air
data(vv)
data(vv)
data set structure is explained in variogramST.
## Not run: # obtained by: library(spacetime) library(gstat) data(air) suppressWarnings(proj4string(stations) <- CRS(proj4string(stations))) rural = STFDF(stations, dates, data.frame(PM10 = as.vector(air))) rr = rural[,"2005::2010"] unsel = which(apply(as(rr, "xts"), 2, function(x) all(is.na(x)))) r5to10 = rr[-unsel,] vv = variogram(PM10~1, r5to10, width=20, cutoff = 200, tlags=0:5) ## End(Not run)
## Not run: # obtained by: library(spacetime) library(gstat) data(air) suppressWarnings(proj4string(stations) <- CRS(proj4string(stations))) rural = STFDF(stations, dates, data.frame(PM10 = as.vector(air))) rr = rural[,"2005::2010"] unsel = which(apply(as(rr, "xts"), 2, function(x) all(is.na(x)))) r5to10 = rr[-unsel,] vv = variogram(PM10~1, r5to10, width=20, cutoff = 200, tlags=0:5) ## End(Not run)
This is the Walker Lake data sets (sample and exhaustive data set), used in Isaaks and Srivastava's Applied Geostatistics.
data(walker)
data(walker)
This data frame contains the following columns:
Identification Number
Xlocation in meter
Ylocation in meter
V variable, concentration in ppm
U variable, concentration in ppm
T variable, indicator variable
This data sets was obtained from the data sets on ai-geostats, https://wiki.52north.org/bin/view/AI_GEOSTATS/WebHome
Applied Geostatistics by Edward H. Isaaks, R. Mohan Srivastava; Oxford University Press.
library(sp) data(walker) summary(walker) summary(walker.exh)
library(sp) data(walker) summary(walker) summary(walker.exh)
Daily average wind speeds for 1961-1978 at 12 synoptic meteorological stations in the Republic of Ireland (Haslett and raftery 1989). Wind speeds are in knots (1 knot = 0.5418 m/s), at each of the stations in the order given in Fig.4 of Haslett and Raftery (1989, see below)
data(wind)
data(wind)
data.frame wind
contains the following columns:
year, minus 1900
month (number) of the year
day
average wind speed in knots at station RPT
average wind speed in knots at station VAL
average wind speed in knots at station ROS
average wind speed in knots at station KIL
average wind speed in knots at station SHA
average wind speed in knots at station BIR
average wind speed in knots at station DUB
average wind speed in knots at station CLA
average wind speed in knots at station MUL
average wind speed in knots at station CLO
average wind speed in knots at station BEL
average wind speed in knots at station MAL
data.frame wind.loc
contains the following columns:
Station name
Station code
Latitude, in DMS, see examples below
Longitude, in DMS, see examples below
mean wind for each station, metres per second
This data set comes with the following message: “Be aware that the dataset is 532494 bytes long (thats over half a Megabyte). Please be sure you want the data before you request it.”
The data were obtained on Oct 12, 2008, from: http://www.stat.washington.edu/raftery/software.html The data are also available from statlib.
Locations of 11 of the stations (ROS, Rosslare has been thrown out because it fits poorly the spatial correlations of the other stations) were obtained from: http://www.stat.washington.edu/research/reports/2005/tr475.pdf
Roslare lat/lon was obtained from google maps, location Roslare. The mean wind value for Roslare comes from Fig. 1 in the original paper.
Haslett and Raftery proposed to use a sqrt-transform to stabilize the variance.
Adrian Raftery; imported to R by Edzer Pebesma
These data were analyzed in detail in the following article:
Haslett, J. and Raftery, A. E. (1989). Space-time Modelling with Long-memory Dependence: Assessing Ireland's Wind Power Resource (with Discussion). Applied Statistics 38, 1-50.
and in many later papers on space-time analysis, for example:
Tilmann Gneiting, Marc G. Genton, Peter Guttorp: Geostatistical Space-Time Models, Stationarity, Separability and Full symmetry. Ch. 4 in: B. Finkenstaedt, L. Held, V. Isham, Statistical Methods for Spatio-Temporal Systems.
data(wind) summary(wind) wind.loc library(sp) # char2dms wind.loc$y = as.numeric(char2dms(as.character(wind.loc[["Latitude"]]))) wind.loc$x = as.numeric(char2dms(as.character(wind.loc[["Longitude"]]))) coordinates(wind.loc) = ~x+y ## Not run: # fig 1: library(maps) library(mapdata) map("worldHires", xlim = c(-11,-5.4), ylim = c(51,55.5)) points(wind.loc, pch=16) text(coordinates(wind.loc), pos=1, label=wind.loc$Station) ## End(Not run) wind$time = ISOdate(wind$year+1900, wind$month, wind$day) # time series of e.g. Dublin data: plot(DUB~time, wind, type= 'l', ylab = "windspeed (knots)", main = "Dublin") # fig 2: #wind = wind[!(wind$month == 2 & wind$day == 29),] wind$jday = as.numeric(format(wind$time, '%j')) windsqrt = sqrt(0.5148 * as.matrix(wind[4:15])) Jday = 1:366 windsqrt = windsqrt - mean(windsqrt) daymeans = sapply(split(windsqrt, wind$jday), mean) plot(daymeans ~ Jday) lines(lowess(daymeans ~ Jday, f = 0.1)) # subtract the trend: meanwind = lowess(daymeans ~ Jday, f = 0.1)$y[wind$jday] velocity = apply(windsqrt, 2, function(x) { x - meanwind }) # match order of columns in wind to Code in wind.loc: pts = coordinates(wind.loc[match(names(wind[4:15]), wind.loc$Code),]) # fig 3, but not really yet... dists = spDists(pts, longlat=TRUE) corv = cor(velocity) sel = !(as.vector(dists) == 0) plot(as.vector(corv[sel]) ~ as.vector(dists[sel]), xlim = c(0,500), ylim = c(.4, 1), xlab = "distance (km.)", ylab = "correlation") # plots all points twice, ignores zero distance # now really get fig 3: ros = rownames(corv) == "ROS" dists.nr = dists[!ros,!ros] corv.nr = corv[!ros,!ros] sel = !(as.vector(dists.nr) == 0) plot(as.vector(corv.nr[sel]) ~ as.vector(dists.nr[sel]), pch = 3, xlim = c(0,500), ylim = c(.4, 1), xlab = "distance (km.)", ylab = "correlation") # add outlier: points(corv[ros,!ros] ~ dists[ros,!ros], pch=16, cex=.5) xdiscr = 1:500 # add correlation model: lines(xdiscr, .968 * exp(- .00134 * xdiscr))
data(wind) summary(wind) wind.loc library(sp) # char2dms wind.loc$y = as.numeric(char2dms(as.character(wind.loc[["Latitude"]]))) wind.loc$x = as.numeric(char2dms(as.character(wind.loc[["Longitude"]]))) coordinates(wind.loc) = ~x+y ## Not run: # fig 1: library(maps) library(mapdata) map("worldHires", xlim = c(-11,-5.4), ylim = c(51,55.5)) points(wind.loc, pch=16) text(coordinates(wind.loc), pos=1, label=wind.loc$Station) ## End(Not run) wind$time = ISOdate(wind$year+1900, wind$month, wind$day) # time series of e.g. Dublin data: plot(DUB~time, wind, type= 'l', ylab = "windspeed (knots)", main = "Dublin") # fig 2: #wind = wind[!(wind$month == 2 & wind$day == 29),] wind$jday = as.numeric(format(wind$time, '%j')) windsqrt = sqrt(0.5148 * as.matrix(wind[4:15])) Jday = 1:366 windsqrt = windsqrt - mean(windsqrt) daymeans = sapply(split(windsqrt, wind$jday), mean) plot(daymeans ~ Jday) lines(lowess(daymeans ~ Jday, f = 0.1)) # subtract the trend: meanwind = lowess(daymeans ~ Jday, f = 0.1)$y[wind$jday] velocity = apply(windsqrt, 2, function(x) { x - meanwind }) # match order of columns in wind to Code in wind.loc: pts = coordinates(wind.loc[match(names(wind[4:15]), wind.loc$Code),]) # fig 3, but not really yet... dists = spDists(pts, longlat=TRUE) corv = cor(velocity) sel = !(as.vector(dists) == 0) plot(as.vector(corv[sel]) ~ as.vector(dists[sel]), xlim = c(0,500), ylim = c(.4, 1), xlab = "distance (km.)", ylab = "correlation") # plots all points twice, ignores zero distance # now really get fig 3: ros = rownames(corv) == "ROS" dists.nr = dists[!ros,!ros] corv.nr = corv[!ros,!ros] sel = !(as.vector(dists.nr) == 0) plot(as.vector(corv.nr[sel]) ~ as.vector(dists.nr[sel]), pch = 3, xlim = c(0,500), ylim = c(.4, 1), xlab = "distance (km.)", ylab = "correlation") # add outlier: points(corv[ros,!ros] ~ dists[ros,!ros], pch=16, cex=.5) xdiscr = 1:500 # add correlation model: lines(xdiscr, .968 * exp(- .00134 * xdiscr))