predict: Predict allocation suitability

In simonmoulds/lulcc2: Land Use Change Modelling in R

Description

Estimate allocation suitability with predictive models.

Usage

## S3 method for class 'PredictiveModelList'
predict(object, newdata,
  data.frame = FALSE, ...)

Arguments

object	a PredictiveModelList object
newdata	data.frame containing new data
data.frame	logical indicating whether the function should return a matrix (default) or data.frame
...	additional arguments to predict methods

Details

This function is usually called from allocate to calculate land use suitability at each timestep. However, it may also be used to produce suitability maps (see examples).

Value

A matrix.

See Also

predict, allocate

Examples

## Not run: 

## Plum Island Ecosystems

lu <- DiscreteLulcRasterStack(x=stack(pie[1:3]),
                              categories=c(1,2,3),
                              labels=c("Forest","Built","Other"),
                              t=c(0,6,14))

idx <- data.frame(var=c("ef_001","ef_002","ef_003"),
                  yr=c(0,0,0),
                  dynamic=c(FALSE,FALSE,FALSE))
ef <- ExpVarRasterStack(x=stack(pie[4:6]), index=idx)

part <- partition(x=lu, size=0.1, spatial=TRUE, t=0)
train.data <- getPredictiveModelInputData(lu=lu,
                                          ef=ef,
                                          cells=part[["train"]],
                                          t=0)

forest.form <- as.formula("Forest ~ ef_001 + ef_002")
built.form <- as.formula("Built ~ ef_001 + ef_002 + ef_003")
other.form <- as.formula("Other ~ ef_001 + ef_002")

forest.glm <- glm(forest.form, family=binomial, data=train.data)
built.glm <- glm(built.form, family=binomial, data=train.data)
other.glm <- glm(other.form, family=binomial, data=train.data)
glm.mods <- PredictiveModelList(list(forest.glm, built.glm, other.glm),
                                categories=lu@categories,
                                labels=lu@labels)

all.data <- as.data.frame(x=ef, cells=part[["all"]]) 
probmaps <- predict(object=glm.mods, 
                    newdata=all.data, 
                    data.frame=TRUE) 

points <- rasterToPoints(lu[[1]], spatial=TRUE) 
probmaps <- SpatialPointsDataFrame(points, probmaps) 
probmaps <- rasterize(x=probmaps, y=lu[[1]], 
                      field=names(probmaps))
plot(probmaps)


## End(Not run)

simonmoulds/lulcc2 documentation built on Dec. 23, 2021, 2:24 a.m.