reddPrec (version 3.0.2)

learner_rf: Function of the random forest model (rf)

Description

The function of the random forest model (rf) is used to predict wet/dry (classification) days and the amount of rain on wet days (regression). This function employs the hybrid model of Huerta et al. (2024).

Usage

learner_rf(ref, can, covars)

Arguments

ref

matrix or data.frame containing the covariable data and precipitation value for each point location. This data is used to train (build) the model

can

matrix or data.frame containing the variable data and precipitation value for one single point location. This data is used to execute the build model and get the predictions

covars

vector of character elements containing the covariables names used in the model.

Details

The function should not be used directly, as its main purpose is to be passed as an argument in the core functions (qcPrec, gapFilling and gridPcp) of the package. The core functions can handle whatever condition that avoids the execution of this function. Therefore, there is no need to test different situations on the data (empty values, among others). However, users who want to build their functions should test first before being passed as an argument in the core functions (you can check the other models (learner) functions available in the package). This function encapsulates two types of models: classification and regression, and its output is a single numeric vector of three elements: probability of wet day, amount of wet day, and uncertainty of the amount of wet day. To use this function, the randomForest package should have been installed previously

References

Huerta, A., Serrano-Notivoli, R., & Brönnimann, S. (2024). SC-PREC4SA: A serially complete daily precipitation dataset for South America. https://doi.org/10.31223/X57D8R

Examples

if (FALSE) {
set.seed(123)
# creating random data (three predictors)
lon = rnorm(50,0,1)
lat = rnorm(50,40,1)
dcoast = rnorm(50,200,50)
prec = rnorm(1*50, mean = 1.2, sd = 6)
prec[prec<0] <- 0

# precipitation column should be call as "val"
data_full <- data.frame(lon = lon, lat = lat, dcoast = dcoast, val = prec)

# parameters for learner_rf
ref = data_full[-1, ]
can = data_full[1, ]
covars = c("lon", "lat", "dcoast")

learner_rf(ref = ref, can = can, covars = covars)

# case when "prec" is full wet
prec = rnorm(1*50, mean = 1.2, sd = 6)
prec[prec<0] <- 1

data_full <- data.frame(lon = lon, lat = lat, dcoast = dcoast, val = prec)

# parameters for learner_rf
ref = data_full[-1, ]
can = data_full[1, ]
covars = c("lon", "lat", "dcoast")

learner_rf(ref = ref, can = can, covars = covars)
}