StatAnalysis not reading the stats for the specified period

[wuxx66]

I am using StatAnalysis to read the stats files from INIT_BEG to INIT_END and then aggregate the stats, but it reads all stats files.

My stat input directory is something like /path/EnsembleStat/{init?fmt=%Y%m%d%H}. In the directory, it has the file name "ensemble_stat_{valid?fmt=%Y%m%d}_%H{0000}V.stat". For example, "/path/EnsembleStat/2026021600/ensemble_stat_20260216_120000V.stat"

My config file is as below

PROCESS_LIST = StatAnalysis
LOOP_BY = INIT
INIT_TIME_FMT = %Y%m%d%H
INIT_BEG = 2026020100
INIT_END = 2026020600
INIT_INCREMENT=24

GRID_STAT_OUTPUT_DIR = /path/EnsembleStat

MODEL1 = AGlobal4E
#MODEL1_OBTYPE = GPM
MODEL1_OBTYPE = ANALYSIS

MODEL1_STAT_ANALYSIS_LOOKIN_DIR = {GRID_STAT_OUTPUT_DIR}

MODEL1_STAT_ANALYSIS_DUMP_ROW_TEMPLATE = 00Z/{MODEL1}{valid?fmt=%Y%m%d%H}.stat
MODEL1_STAT_ANALYSIS_OUT_STAT_TEMPLATE = 00Z/{MODEL1}{fcst_lead?fmt=%H}ECNT{valid?fmt=%Y%m%d%H}.stat

STAT_ANALYSIS_OUTPUT_DIR = {OUTPUT_BASE}
STAT_ANALYSIS_CONFIG_FILE = {PARM_BASE}/met_config/STATAnalysisConfig_wrapped

MODEL_LIST = {MODEL1}
DESC_LIST =
FCST_LEAD_LIST = 12
OBS_LEAD_LIST =
FCST_VALID_HOUR_LIST = 12
FCST_INIT_HOUR_LIST = 0
OBS_VALID_HOUR_LIST =
OBS_INIT_HOUR_LIST =
FCST_VAR_LIST =
OBS_VAR_LIST =
FCST_UNITS_LIST =
OBS_UNITS_LIST =
FCST_LEVEL_LIST =
OBS_LEVEL_LIST =
VX_MASK_LIST =
INTERP_MTHD_LIST =
INTERP_PNTS_LIST =
FCST_THRESH_LIST =
OBS_THRESH_LIST =
COV_THRESH_LIST =
ALPHA_LIST =
LINE_TYPE_LIST =

GROUP_LIST_ITEMS = FCST_VALID_HOUR_LIST
LOOP_LIST_ITEMS = FCST_LEAD_LIST

If I do as below
GRID_STAT_OUTPUT_DIR = /path/EnsembleStat/{init?fmt=%Y%m%d%H}

MODEL1_STAT_ANALYSIS_LOOKIN_DIR = {GRID_STAT_OUTPUT_DIR}

I will get an error

WARNING:
WARNING: get_filenames() -> can't stat "/path/EnsembleStat/ALL"
WARNING:
ERROR :
ERROR : process_search_dirs() -> no STAT files found in the directories specified!
ERROR :
ERROR :
ERROR : main() -> encountered an error value of 1. Calling clean_up() and usage() before exiting.
ERROR :

*** Model Evaluation Tools (METV12.0.2) ***

It seems "{init?fmt=%Y%m%d%H}" is not passed to the stat directory "GRID_STAT_OUTPUT_DIR"

Could you help me have a look what is wrong in my config?

Thank you

Xiaoxi

[j-opatz]

Hello Xiaoxi,

And thank you for your question. It looks like you are currently encountering an issue with StatAnalysis where it is not selecting the subset of dates that you would like, and substitutes "ALL" instead of the initialization times you expect.

After discussions with our lead METplus wrappers engineer, the issue is coming from the MODEL1_STAT_ANALYSIS_LOOKIN_DIR setting, as you discovered. By default, this setting looks for ALL available statistical files in the directory, and ignores any timing information that would otherwise allow a user to select a subset of the data in the directory.

There are two suggested options to get around this, but both are admittedly not straight-forward.

You could add a UserScript instance to the PROCESS_LIST to loop through the init times (USER_SCRIPT_RUNTIME_FREQ = RUN_ONCE_PER_INIT_OR_VALID), and create a symbolic link for each init directory into another directory, then pass that newly created directory as the MODEL1_STAT_ANALYSIS_LOOKIN_DIR.

The scripting would do something like this:

mkdir /path/EnsembleStat_stage;
ln -s /path/EnsembleStat/{init?fmt=%Y%m%d%H} /path/EnsembleStat_stage/

then set

MODEL1_STAT_ANALYSIS_LOOKIN_DIR = /path/EnsembleStat_stage

The second option would be to list the init directories in MODEL1_STAT_ANALYSIS_LOOKIN_DIR by hand. This is messy and not easy to adjust between runs, but it should work. For example,

MODEL1_STAT_ANALYSIS_LOOKIN_DIR =
  /path/EnsembleStat/2026020100,
  /path/EnsembleStat/2026020200,
  /path/EnsembleStat/2026020300,
  /path/EnsembleStat/2026020400,
  /path/EnsembleStat/2026020500,
  /path/EnsembleStat/2026020600

I understand that these are not straightforward solutions, and we agree that the behavior you are looking for should be available in StatAnalysis. As a result of your finding, I've gone ahead and created a new issue documenting the enhancement to process directories with timing information. Hopefully when this work is completed, your original set up will work as intended! Please note that our ability to work on these issues is limited by time and funding, so we cannot provide a time when this functionality will be available.

[wuxx66]

Thank you @j-opatz for creating an issue for this. Hopefully this can be solved in future. In the meanwhile, I will try your work around options.

[j-opatz]

Hi Xiaoxi,

I wanted to see if either of the two workaround methods I suggested helped you solve this issue.

If so, I'd like to go ahead and mark this Discussion as solved and close it out.

[wuxx66]

Hi @j-opatz , I think I have tried. One of the methods works, but for operational purpose, it is not really flexible to list all the files per month in the StatAnalysis config file. We have tried a work around way. Copy all the files in the required month to a temporary directory and StatAnalysis looks into this directory. After the job is done, then remove the directory. We are still testing it, but feel free to close this discussion. Thank you!

[j-opatz]

Thank you Xiaoxi,

I'm sorry the method that does work is not very compatible with your operational needs.

Hopefully the work on #3234 will help alleviate the need to use this work around.