calculate agreement of two maps at zonal level in R - r

Question

I am trying to calculate the agreement of two different cropland maps at zonal/ grid level.
I first created the rasters that show cropland agreement between a classified raster and validation data that was created.
agreement<- sample_raster==raster_sampled
crop_agreement <- sample_raster==1 & raster_sampled==1
I then prepared the tiles for zonal analysis by:
tiles_poly$id<- as.double(tiles_poly$id)
tiles_poly$zone <- tiles_poly$id
tiles<-fasterize(st_as_sf(tiles_poly),raster_aoi, field="id")
This works fine.
However, when I'm trying to calculate the agreement at zonal level, I get an error.
zonal_crop_agreement<- zonal(crop_agreement,tiles,fun='sum', na.rm=TRUE)
Error in dimnames(x) <- dn :
length of 'dimnames' [2] not equal to array extent
As far as I understand, the error arises because the number of columns is not equal between the two rasters. Yet, when I print the length of tiles & crop_agreement, the have the same length.
I think therefore, that the error might be due to the "fun = 'sum'". I also do not exactly understand what that function does.
Does someone have an idea what the problem could be?
The complete code is here:
## INPUT DATA GOES HERE ##
test_name<- "220624_cropland" # use this to name the test. The exported files will have this name
tiles_poly<-readOGR("C:/data_cropland_burkina/Grid_50km.shp") #the tile grid
raster_aoi <- raster("C:/data_cropland_burkina/empty_raster.tif") #sets the resolution and area of interest
k_rast<-raster("C:data_cropland_burkina/Cropland_Burkina_Copernicus.tif") # this is the classified raster I am analyzing
#k_rast <- rast_class
## This is to convert points into raster ##
sample_points<-readOGR("C:/Users/louis/OneDrive/Desktop/data_cropland_burkina/validation_points_poly0.shp", stringsAsFactors = FALSE) #the validation points
sample_points$class_2 <- as.numeric(sample_points$item) # make sure the class column is numeric
sample_raster<- fasterize(st_as_sf(sample_points), raster_aoi, field="item")
## crop the raster (makes the script run faster) ##
#aoi<- crop(raster_aoi,k_rast)
plot(k_rast)
#plot(sample_points, add=TRUE) ##maybe remove this by commenting out
crop_aoi <- function(x) {crop(x,raster_aoi)}
k_rast <- crop_aoi(k_rast)
sample_raster<-crop_aoi(sample_raster)
## ONLY USE THIS IF THE RASTER DOES NOT HAVE THE SAME RESOLUTION AS THE RASTER_AOI ##
#resample all rasters
resize_aoi<- function(x){raster::resample(x,raster_aoi,method="ngb")}
k_rast<-resize_aoi(k_rast)
##Reclassify if the raster is binary. otherwise comment out.
m <- c(0, 2, 1, 1)
rclmat <- matrix(m, ncol=2, byrow=TRUE)
k_rast <- reclassify(k_rast, rclmat)
## get the unique values from each raster ##
k_rast_vals<-sort(unique(na.omit(getValues(k_rast))))
sample_vals<-unique(na.omit(getValues(sample_raster)))
## extract the crop and non-crop pixels from the sample ##
sample_crop<- sample_raster==1
sample_non_crop<- sample_raster==2
sample_raster_binary<-sample_raster>0
## reduce the classified raster to only the pixels covered by the sample ##
raster_sampled<- k_rast*sample_raster_binary
## Calculate the confusion matrix ##
crop_comp <- function(x) cellStats(raster_sampled==x * sample_crop, stat='sum')
non_crop_comp <- function(x)cellStats(raster_sampled==x * sample_non_crop, stat='sum')
results_crop <- rbind(sapply(k_rast_vals,crop_comp))
colnames(results_crop)<- k_rast_vals
results_non_crop<- rbind(sapply(k_rast_vals,non_crop_comp))
colnames(results_crop)<- k_rast_vals
results<- rbind(results_crop, results_non_crop)
rownames(results) <- c("crop_ref", "non_crop_ref")
results<-cbind(results, total = rowSums(results))
results<-rbind(results, total = colSums(results))
results
## ------SECTION 2- Zonal Analysis - show agreement at the grid level------- ##
## Create rasters that show cropland agreement between the classified raster and validation data ##
agreement<- sample_raster==raster_sampled
crop_agreement <- sample_raster==1 & raster_sampled==1
## prepare the tiles for zonal analysis ##
tiles_poly$id<- as.double(tiles_poly$id)
tiles_poly$zone <- tiles_poly$id
tiles<-fasterize(st_as_sf(tiles_poly),raster_aoi, field="id")
## calculate agreement at the zonal/grid level (thus, the statistics on cell alues of the raster within the ones defined by the other dataset - validation data & input crop maps)##
## next line calculates the zonal crop agreement, taking the crop agreement raster created as well as the prepared tiles as input
## na.rm removed missing values from the data if they are indicated as NA.
## the function fun = 'sum' is
zonal_crop_agreement<- zonal(crop_agreement,tiles,FUN='sum', na.rm=TRUE)
zonal_crop_ref <- zonal(sample_crop,tiles,fun= 'sum', na.rm=TRUE)
zonal_crop_agreement<- data.frame(cbind(zonal_crop_agreement, zonal_crop_ref[,2]))
colnames(zonal_crop_agreement)[3]<- "crop_ref"
zonal_crop_agreement<- data.frame(cbind(zonal_crop_agreement,zonal(raster_sampled==1,tiles,fun="sum",na.rm=TRUE)[,2]))
colnames(zonal_crop_agreement)[4]<- "crop_class"
zonal_crop_agreement<- data.frame(cbind(zonal_crop_agreement,zonal(sample_raster,tiles,fun="count",na.rm=TRUE)[,2]))
colnames(zonal_crop_agreement)[5]<- "samp_pxls"