OBS_DOMAINdxx with more than 99 observations

[cross85]

I'm trying to run obsgrid for a period of six months, with hourly data from a surface station.
I've got all formatted in little_r, and obsgrid runs well, but it only get the first 99 values, after that overwrite a file named "'OBS_DOMAIN1**'"
Does anyone know how to use a longer period to include more data (100 or more) for observation nudging?

[allenea]

1. What is your max_number_of_obs variable set at in the namelist file. 2) What does the output obsgrid.out file (output from running obsgrid) say?

[cross85]

@allenea the obsgrid.out is ok, the metoa_em* and wrfsfdda_d* are generated ok.

This is the first three records of the namelist:

&record1
 start_year                  =  2017
 start_month                 =    01
 start_day                   =    01
 start_hour                  =    00
 end_year                    =  2017
 end_month                   =    07
 end_day                     =    02
 end_hour                    =    00
 interval                    = 21600
/

&record2
 grid_id                     = 4
 obs_filename                = './OBS/OBS'
 remove_data_above_qc_flag   = 32768
 remove_unverified_data      = .TRUE.
 trim_domain                 =.FALSE.

/

&record3
 max_number_of_obs           = 100000
 fatal_if_exceed_max_obs     = .TRUE.
/

The problem is the output of OBS_DOMAINdxx, because after OBS_DOMAIN199 is generated starts creating a file named "'OBS_DOMAIN1**'" and rewrited in every time.

OBS_DOMAIN199 date is 2017-01-13_06:00 (output is every 3 hours)
'OBS_DOMAIN1**' date is 2017-07-01_21:00 (which is rewrited every output after 199)

I also tried to find a fix for this, and I just found this post with the same problem, but also without a solution.

[cross85]

I finally did the OBS_DOMAIN file with a script instead of using OBSGRID.
Maybe using RT_fdda_reformat_obsnud.pl is better instead of using obsgrid for long runs.

[zwang-leeds]

In src/driver.F90, go to L785, L797, L809 and change "i2.2" into another such as "i4.4", and compile. Then the 1** can be extracted and they don't need to share the same 1** file :)