I am trying to model temperature across seasons using (package mgcv) GAMs. I am interested in using terra::predict() to plot the predicted temperatures as a raster.
Ideally, the output would be 4 rasters of the study area to represent the four seasons (winter, spring, summer, fall). However, I do not understand how to predict/if this can be done when using non-spatial (non-raster) with factors/groups (season).
An example of my code looks like:
library(terra)
library(mgcv)
v <- vect(system.file("ex/lux.shp", package="terra"))
r <- rast(system.file("ex/elev.tif", package="terra"))
set.seed(50)
pnts <- spatSample(v, size = 50)
pnts_r <- terra::extract(r, pnts, method = "simple")
pnts_r$temp <- sample(0:25, 50, replace = TRUE)
# Define the sequence of seasons
seasons <- c('winter', 'spring', 'summer', 'fall')
# random add seasons seq
pnts_r$season <- seasons[seq_len(nrow(pnts_r)) %% length(seasons) + 1]
mod_gam <- mgcv::gam(temp ~ elevation + factor(season), data = pnts_r,
family = gaussian(), method = "REML")
p <- predict(mod_gam, type="response")
x <- terra::predict(r, mod_gam, type="response")
#Error in eval(predvars, data, env) : object 'season' not found
#In addition: Warning message:
#In predict.gam(model, d, ...) :
# not all required variables have been supplied in newdata!`
I understand that seasons are not included in this raster, but I do not know if I can stack empty rasters as place holders or temperature modelling per season should be performed individually. I'd also like to bring in more data to look at inter-year variation, but want to start simply with inter-season of one year.
Any help is appreciated
