How to include time variable when making spatiotemporal predictions with sdmTMB over a raster stack in R?

Question

I'm using the sdmTMB package in R to fit a spatiotemporal model, and I'm trying to make predictions over a raster stack of environmental variables. However, because this is a spatiotemporal model, I need to include a time variable in my predictions. I've tried to convert the raster stack to a dataframe, but I’m unsure how to incorporate time across the raster predictions. Apologies for my example code as the points and rasters don't spatially overlap.

library(sdmTMB)
library(terra)

# Data and mesh setup
data("pcod")
mesh <- make_mesh(pcod, xy_cols = c("X", "Y"), cutoff = 10)

# Model fitting
m <- sdmTMB(
  density ~ 1, 
  time_varying = ~ 0 + depth_scaled + depth_scaled2,
  data = pcod, mesh = mesh,
  time = "year",
  family = binomial(link = "logit"),
  spatial = "on",
  spatiotemporal = "ar1"
)

# Example raster stack
elev <- rast(system.file("ex/elev.tif", package = "terra"))
slope <- terrain(elev, "slope")
aspect <- terrain(elev, "aspect")

# Create a raster stack
raster_stack <- c(elev, slope, aspect)

# Converting raster stack to dataframe for predictions
vars_stack_df <- terra::as.data.frame(raster_stack, xy = TRUE) |> 
  rename(X = x, Y = y) 

# Attempt to predict on raster data
tmb_ras_p <- predict(mod_tmb, newdata = vars_stack_df)