intamapInteractive (version 1.2-6)

biasCorr: Bias correction

Description

Identifies and removes biases from measurement networks

Usage

biasCorr(object,regCode = "regCode",...)

Value

Data frame with observations, with the identified biases removed.

Arguments

object

Data frame with observations with same format as observations described in the presentation of the intamap-package)

regCode

the column name of regions in the data polygons, if existing

...

further arguments to the bias correction methods called, see details below

Author

Jon Olav Skoien

Details

Many data sets can consist of data coming from a large number of different measurement networks, using different measurement devices or applying different methods for post-processing the observations. Some of these networks can exist in the same area, e.g. when different authorities are measuring the same, but at different locations (one of them in cities, the other one close to lakes), some networks will only exist as neighbouring networks (networks operated by a municipality or a country). Local networks can also be grouped together as one national data-base, which can again be merged into an international data-base.

One challenge with the merging into data-bases is that there will be inconsistencies between measurements in the different networks, which will again cause difficulties when attempting to map the observations, as done in the intamap-package. The intention of this function is therefore to call other functions that are able to identify and remove such differences, which can be referred to as biases between the networks.

There are at the moment two methods available for bias correction, "UK" and "LM". "UK" is a universal kriging based approach implemented in findBiasUK. This method can only deal with biases between neigbouring networks, but is well capable of taking covariates into account. "LM" is based on local methods for estimating differences between networks, and is implemented in findLocalBias and findRegionalBias. The choice between the methods is given by the parameter biasRemovalMethod in the parameter element of the object, set in getIntamapParams, called from createIntamapObject.

The function will remove biases according to the settings of the parameters removeBias. Below is a list of the functions available for bias corrections. See each individual function for more information about usage.

findBiasUK

The universal kriging based function for finding biases between neighbouring networks

findLocalBias

Find biases for ovelapping networks

removeLocalBias

Removes biases between ovelapping networks

findBoundaryLines

Find points that define adjacent boundaries between regions

findRegionalBias

Find biases for neighbouring networks

removeRegionalBias

Remove biases between neighbouring networks

References

Skoien, J. O., O. P. Baume, E. J. Pebesma, and G. B. M. Heuvelink. 2010. Identifying and removing heterogeneities between monitoring networks. Environmetrics 21(1), 66-84.

See Also

findLocalBias

Examples

# \donttest{
data(meuse)
data(meuse.grid)
observations = data.frame(x = meuse$x,y = meuse$y,value = log(meuse$zinc))
coordinates(observations) = ~x+y
gridded(meuse.grid) = ~x+y
pBoundaries = spsample(observations, 8, "regular",bb = bbox(observations) +  
              matrix(c(-400,-400,400,400),ncol=2),offset=c(0,0))
gridded(pBoundaries) = TRUE
cs = pBoundaries@grid@cellsize[1]/2
dx = cs/5

Srl = list()
nb = dim(coordinates(pBoundaries))[1]
for (i in 1:nb) {
  pt1 = coordinates(pBoundaries)[i,]
  x1 = pt1[1]-cs
  x2 = pt1[1]+cs
  y1 = pt1[2]-cs
  y2 = pt1[2]+cs

  boun = data.frame(x=c(seq(x1,x2,dx),rep(x2,11),seq(x2,x1,-dx),rep(x1,11)),
                    y=c(rep(y1,11),seq(y1,y2,dx),rep(y2,11),seq(y2,y1,-dx)))
  coordinates(boun) = ~x+y
  boun = Polygon(boun)
  Srl[[i]] = Polygons(list(boun),ID = as.character(i))
}
pBoundaries = SpatialPolygonsDataFrame(SpatialPolygons(Srl),
                                      data = data.frame(ID=c(1:nb)))
observations$ID = over(observations, geometry(pBoundaries))
blines = findBoundaryLines(pBoundaries,regCode = "ID")


object = createIntamapObject(observations,meuse.grid,boundaryLines = blines, 
  params = list(removeBias = "regionalBias"))
object = biasCorr(object,regCode= "ID")
object$regionalBias$regionalBias
pBoundaries$bias = NA
pBoundaries$bias[object$regionalBias$regionalBias$ID] = object$regionalBias$regionalBias$ols
spplot(pBoundaries,"bias",sp.layout = list(list("sp.points",observations)))
# }