Can't run ENMevaluate with categorical variables

[KanaKammin]

Hello!

ENMeval version 2.0.5.2 produces an error and halts when categorical variables are specified. Otherwise runs great.

My simple code:

library(terra)
library(ENMeval)

occs <- read.csv("occ.csv")
bg<-read.csv("bgPoints.csv")


r <- list.files(pattern="\\.asc$", full.names=T)
envs <- terra::rast(r)

catVars <- c("corine", "ParMat")


tune.args <- list(fc=c("L","LQ","LQH","H","LQHP","LQHPT"), rm=1:5)

e.mx.l <- ENMevaluate(occs = occs, envs = envs, bg = bg, 
                      algorithm = 'maxnet', partitions = 'block', 
                      tune.args = tune.args, categoricals = catVars)

Error message:

*** Running initial checks... ***

* Removed 233 occurrence localities that shared the same grid cell.
* Removed 18 occurrence points with NA predictor variable values.
* Removed 425 background points with NA predictor variable values.
Error in terra::levels(envs[[categoricals[i]]])[[1]][, 2] : 
  incorrect number of dimensions

Any help is appreciated

[jamiemkass]

This is strange because I run a check on multiple categorical variables and it works fine. Is it possible that you are running this with only one environmental variable?

[KanaKammin]

I stacked 7 environmental variables:

> envs
class       : SpatRaster 
size        : 1389, 2054, 7  (nrow, ncol, nlyr)
resolution  : 0.005, 0.005  (x, y)
extent      : 19.37404, 29.64404, 34.80382, 41.74882  (xmin, xmax, ymin, ymax)
coord. ref. : lon/lat WGS 84 (CRS84) (OGC:CRS84) 
sources     : bio10.asc  
              bio17.asc  
              bio2.asc  
              ... and 4 more sources
names       : bio10, bio17, bio2, bio6, corine, ParMat, ... 
> names(envs)
[1] "bio10"  "bio17"  "bio2"   "bio6"   "corine" "ParMat" "WDist" 

To add more info, ENMevaluate works fine if I don't specify any categoricals.

Also, the previous version of ENMeval 2.0.0 worked great with the same vars, bricked with the raster pkg:
env <- brick(stack(asc_files))

Thanks for looking into it!

[jamiemkass]

I think I got it. The original version converted categorical variables to factor rasters in your stack if they weren't already, but the newer versions play it safer and have you make them factor rasters yourself. Turn these into factor types and specify the levels how you want them, then run and let me know if it works.

[KanaKammin]

Thanks to your insights I finally managed to make it work, although I must admit I am lost to the basics of r.
I couldn't figure out how to convert specific layers of a stacked spatraster to factor rasters, so my duct tape implementation was to make a separate folder with the categorical variables, load them in a separate spatRaster with "as.factor", and add it to the continuous rasters. If you know of a simpler implementation, please let me know.

Thanks a lot for your help!

library(terra)
library(ENMeval)

occs <- read.csv("occ.csv")
bg<-read.csv("bgPoints.csv")

r <- list.files(pattern="\\.asc$", full.names=T)
cont <- terra::rast(r)
x <- list.files(path = "cats", pattern="\\.asc$", full.names=T)
categ <-as.factor(x=terra::rast(x))

envs <- c(cont, categ)

catVars <- c("corine", "ParMat")

tune.args <- list(fc=c("L","LQ","LQH","H","LQHP","LQHPT"), rm=1:5)

e.mx.l <- ENMevaluate(occs = occs, envs = envs, bg = bg, 
                      algorithm = 'maxent.jar', partitions = 'block', 
                      tune.args = tune.args, categoricals = catVars)

> envs
class       : SpatRaster 
size        : 1389, 2054, 7  (nrow, ncol, nlyr)
resolution  : 0.005, 0.005  (x, y)
extent      : 19.37404, 29.64404, 34.80382, 41.74882  (xmin, xmax, ymin, ymax)
coord. ref. : lon/lat WGS 84 (CRS84) (OGC:CRS84) 
sources     : bio10.asc  
              bio17.asc  
              bio2.asc  
              ... and 4 more sources
names       : bio10, bio17, bio2, bio6, WDist, corine, ... 
min values  :    ? ,    ? ,   ? ,   ? ,    ? ,     NA, ... 
max values  :    ? ,    ? ,   ? ,   ? ,    ? ,     NA, ... 

*** Running initial checks... ***

* Removed 233 occurrence localities that shared the same grid cell.
* Removed 18 occurrence points with NA predictor variable values.
* Removed 425 background points with NA predictor variable values.
* Assigning variable corine to categorical and changing to integer for maxent.jar...
* Assigning variable ParMat to categorical and changing to integer for maxent.jar...
* Clamping predictor variable rasters...
* Model evaluations with spatial block (4-fold) cross validation and lat_lon orientation...

*** Running ENMeval v2.0.5 with maxent.jar v3.4.3 using the predicts package v0.1.19 ***
  |=============================================================================| 100%
Making model prediction rasters...
ENMevaluate completed in 7 minutes 30.6 seconds.

[brackejt]

I think I'm running into the same problem, potentially, but can't seem to correct it. Any suggestions?

occ <- read.csv("Presences.csv")

current <- paste0(getwd(), "/Current")
env_files <- list.files(current, pattern = "\.tif$", full.names = TRUE)
env <- brick(stack(env_files))

categ.vars <- c("LC20_BPS_220_wgs84.tif", "LC20_EVC_220_wgs84.tif", "LC20_EVH_220_wgs84.tif", "LC20_EVT_220_wgs84.tif")

me1 <- maxent(env, occ, factors = categ.vars, path = results_folder,
args = c("writeplotdata=TRUE",
"writebackgroundpredictions=TRUE",
"randomseed=TRUE",
"askoverwrite=FALSE",
"autofeature=TRUE"
)
)

number <- nrow(read.csv(paste0(results_folder, "/species_samplePredictions.csv")))
noquote(paste0("you have ", number, " occurrence points after duplicates occupying the same cell and points missing some environmental data have been removed."))

if(number >= 50){
enmeval_results <- ENMevaluate(occ, env, tune.args = list(fc = c("L","LQ","H", "LQH", "LQHP", "LQHPT"), rm = 1:5), partitions = "randomkfold", partition.settings = list(kfolds = 10), algorithm = "maxent.jar", parallel = FALSE, categoricals = categ.vars)
}else{
enmeval_results <- ENMevaluate(occ, env, tune.args = list(fc = c("L","LQ","H", "LQH", "LQHP", "LQHPT"), rm = 1:5), partitions = "jackknife", algorithm = "maxent.jar", parallel = FALSE, categoricals = categ.vars)
}

[jamiemkass]

Please make sure your categorical rasters are factor type before running ENMeval. Are they factor type?

[brackejt]

Hi Jamie,

Thanks for responding. I tried several times to make the categorical variables factor type (all the variables are categorical in this model) using the as.factor function, and was unsuccessful. Probably because I'm a bad coder and I'm not familiar with the function. If you have any code suggestions to implement, I'd happily take them.

[jamiemkass]

If you send me the code you are running and the errors you are getting, I can try to help.

[brackejt]

Hi Jamie, Sorry for the delay. I've attached my code and I cropped my files and presence points down to almost nothing, so the entire script can finish in a minute when it doesn't error out. And it doesn't error if I run the script as if they are all continous variables (categ.vars <- ()). But it fails when I define categoricals. Any help you can provide would be appreciated! Thanks, and hope you are well! Jason

…

On Mon, Jul 27, 2026 at 10:38 AM Jamie M. Kass ***@***.***> wrote: *jamiemkass* left a comment (jamiemkass/ENMeval#175) <#175 (comment)> If you send me the code you are running and the errors you are getting, I can try to help. — Reply to this email directly, view it on GitHub <#175?email_source=notifications&email_token=AKYC4755XRLUGOYUME7VU2L5G5SO5A5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTKMBZGI3TCMZZHA22M4TFMFZW63VHMNXW23LFNZ2KKZLWMVXHJLDGN5XXIZLSL5RWY2LDNM#issuecomment-5092713985>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AKYC47YMMCA7R3TK7FKYNWD5G5SO5AVCNFSNUABEKJSXA33TNF2G64TZHMZDSOBWGQYDIMZ3JFZXG5LFHMZTMMBYHE4DSMBYG6QXMAQ> . Triage notifications, keep track of coding agent tasks and review pull requests on the go with GitHub Mobile for iOS <https://github.com/notifications/mobile/ios/AKYC476YSJGCCFTRTTBDX7T5G5SO5A5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTKMBZGI3TCMZZHA22M4TFMFZW63VHMNXW23LFNZ2KKZLWMVXHJKTGN5XXIZLSL5UW64Y> and Android <https://github.com/notifications/mobile/android/AKYC476DC6QQ4JKZGXPYXCD5G5SO5A5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTKMBZGI3TCMZZHA22M4TFMFZW63VHMNXW23LFNZ2KKZLWMVXHJLTGN5XXIZLSL5QW4ZDSN5UWI>. Download it today! You are receiving this because you commented.Message ID: ***@***.***>

-- Jason Bracken (he/him/his) PhD Candidate, Research Associate Jezkova Global Change Biology Lab Miami University Oxford, Ohio 45056 484-354-4798 ***@***.***