Intermittent smkinven "duplicate and SIGSEGV error" in pt_wildfire sector

SMOKE
1

Hi all,

I am finding intermittent smkinven SIGSEGV failure while running ptfire onetime step in our forecasting pipeline.

I am using daily HMS input from:https://satepsanone.nesdis.noaa.gov/pub/FIRE/web/HMS/Fire_Points/Text/2026/06/hms_fire20260605.txt. Creating .orl files for smoke (5.2.1, EMP 2022 v1) using this script: https://drive.google.com/file/d/1OHm7CiaeVjJVANlSQwQVsVSCNhsK0PBp/view?usp=drive_link

With the first 4 days input ptinv and ptday file, the fire sector (inline) processed smoothly; however, it showed an issue after that. the smkinven log is here:

smkinven_log.txt (12.1 KB). The successful and failed ptinv, ptday files are here: ORLS - Google Drive

Is it an HMS raw data issue, or any smoke setup issue? Previously, I had to compile smkinven with rdinvdata to read orls in SMOKE 5.2.1, and the read file issue was resolved.

Any suggestion or advice is greatly appreciated.

Thanks,

Priom

2

As a first suggestion: compile and run the debug/trap-everything/traceback version of smkinv to find out where the failure is happening: this version will give you a “traceback” giving the line numbers of the source files for which the error is occurring.

(You may need to look at The Easy Guide® for Building SMOKE to see how to build the debug-version.)

3

Your SMOKE logs are showing duplicate record warnings based on multiple ptinv inputs. Do these inventories contain the same data? For the ORL ptinv you should only have one record per release point and process regardless of the number of days that you assume the fire occurs; The ptday file(s) will handle the daily temporal aspect. If moving to a single ptinv does not fix your issue I suggest compiling SMOKE with debugging, as Carlie suggests, and sharing the traceback.

Additionally, SMOKE is not regularly tested with the ORL inventory file format as the EMPs have gone exclusively to FF10s. I suggest moving to the point FF10 format ( 6.2.8. PTINV: Point source annual or average day emissions ) following the ptinv/ptday ptfire files in the 2022v2 platform (https://gaftp.epa.gov/Air/emismod/2022/v2/2022emissions/2022he_fires_inventory_25sep2025.zip).

4

Before recompiling, could you please guide me regarding this error or issues? My system works for monthly test runs, benchmarks and the 3 forecast cycles smoothly with the same system, therefore, I want to make sure completely.

Thank you so much.

image
image2072×1105 115 KB

5

Lines 166 and 167 of procinven:

        J = INDEXA( I )
        S = INRECA( J )

I suspect that J is exceeding the number of input records (source X pollutants) in INRECA, which should not happen. Please try reducing your ptinv into one file without source duplicates.

6

Thank you so much for the guidance. I am grateful. I have used an assembly script, that concatenate orl files (ptinv dedup also). Feeding SMOKE, one concatenated ptinv, ptday files for 3 days cycle; It worked. Log is here:

tail-log.txt (46.4 KB)