predictRainfall: Predicts the rainfall rate based on a trained model and MSG...

In environmentalinformatics-marburg/Rainfall: Functions for training, predictiion and validation of an MSG based optical rainfall retrieval

Description

Functions calculates predictors which are required by the model and uses them for prediction

Usage

predictRainfall(model, inpath = NULL, sceneraster = NULL, rainmask = NULL,
  sunzenith = NULL, date = NULL, useOptimal = TRUE,
  scaleparam = model$scaleParam, min_x = NULL, max_x = NULL,
  type = "tif")

Arguments

model	The final model from either caret's train or rfe. Use model$fit$finalModel or model$finalModel to get it
inpath	Path to the MSG data
sceneraster	If no inpath is specified: The Meteosat data from which rainfall should be predicted. Load them with getChannels
rainmask	A raster indicates areas which are not raining with NA values
sunzenith	If no inpath is specified: sunzenith optional. Only needed if included in the predictor variables
date	If no inpath is specified: date optional. only needed if jday is included in the predictor variables
useOptimal	if model is a rfe object: Logical. Use the optimal variables from rfe or those less variables which lead to a model performance within one sd of the optimal model?
scaleparam	A data.frame created with calcScalingStats
type	If inpath!=NULL then the data type of the MSG raster

Value

A Raster Layer containing predicted rainfall

Author(s)

Hanna Meyer

Examples

# stack the msg scenes:
msg_example <-getChannels(inpath=system.file("extdata/msg",package="Rainfall"))

reference <- raster(system.file("extdata/radar",
"201007121650_radolan_SGrid.rst",package="Rainfall"))
values(reference)[values(reference<0.06)]=NA

data(rfemodel)

#predict on the new scene (don't expect good results from the small model!)
pred<-predictRainfall(model=rfeModel, sceneraster=msg_example, rainmask=reference)
validate(obs=reference,pred=pred)

environmentalinformatics-marburg/Rainfall documentation built on May 16, 2019, 7:49 a.m.