Issue with long runs in ENMeval

[Oscar Toro]

Hi everybody,

I was wondering if anubody has faced issues in ENMeval, when running long or demanding analyses in R. Particularly, I have had continuous problems trying to use "jackknife" occurrence permutation method, for which R crashes showing an issue related to rJava. I tested a small dataset (11 occurrences) on 5 raster layers and after a few seconds, R sends the following error:

java.io.FileNotFoundException: C:\Users\BOTANI~1\AppData\Local\Temp\RtmpOWRZRY\file11cc2f1a4b7c\plots\species_roc.png (Acceso denegado)
    at java.io.RandomAccessFile.open0(Native Method)
    at java.io.RandomAccessFile.open(Unknown Source)
    at java.io.RandomAccessFile.<init>(Unknown Source)
    at javax.imageio.stream.FileImageOutputStream.<init>(Unknown Source)
    at com.sun.imageio.spi.FileImageOutputStreamSpi.createOutputStreamInstance(Unknown Source)
    at javax.imageio.ImageIO.createImageOutputStream(Unknown Source)
    at javax.imageio.ImageIO.write(Unknown Source)
    at density.Runner.writePlots(Runner.java:2449)
    at density.Runner.writeCumulativeIndex(Runner.java:1947)
    at density.Runner.start(Runner.java:493)
    at mebridge.fit(mebridge.java:27)

Error in rJava::.jcall(mxe, "S", "fit", c("autorun", "-e", afn, "-o",  :  java.lang.NullPointerException

The command line for this rutine was the following:

c = ENMevaluate(bis, cli, method="jackknife", bg, RMvalues = c(1,2,3), fc = c("L", "LQ"), bin.output=TRUE,clamp=TRUE, parallel=FALSE)

I suspect that there is some issue in my computer to save files in the Temp folder, since the error is found when jackknife is applied. I tested other methods of partitioning with no issue at sight. I am using a computer with Windows 7 (64x).

Has anybody get this error before? Is there any way to get around this problem??

Thanks a lot!!

Oscar

[Jamie M. Kass]

So this problem only occurs for jackknife partitions, but not any others? Have you tried doing a random partition with the same number of bins (11?)? If this also crashes, I'd say it could be a Java memory problem, but might be a temp file thing too (I'm not a Java expert). Maybe try this to up the RAM Java uses and see if anything changes:

options(java.parameters = "-Xmx8000m")

Jamie Kass
PhD Candidate
City College of NY

[Oscar Toro]

Hi Jaimie,

Thank you very much for the input. I did try doing random partitions with the same number of bins (11) and, unfortunately, the system still crashes. There seems to be an issue with runs that require intense writing of temporary files, as required with processes that partition occurrence data. Increasing memory to Java system also did not change the situation. I wonder if there are restrictions from my OS, operating Java...

Oscar

[Jamie M. Kass]

Yes, this sounds like a Java issue. Please try changing the “8000” to an even larger number — if it’s a memory issue, a larger value may work. Otherwise, not sure I can help. When we finally release the new version of ENMeval with maxnet capability, you could use that.

[bkara...@gmail.com]

I had no problems with running ENMeval for 11 records, jackknife validation, 1-10 beta values, all combinations of features and parallel processing on 6gb ram and 4 cores. However, I suggest you to do garbage collection with gc() before running memory demanding operations.

24 Ekim 2017 Salı 11:35:35 UTC+3 tarihinde Oscar Toro yazdı: