The -ftime option prints the forecast time stamp
such as the forecast hour, accumulation period, etc. This option is used
by many other inventory options.
The -set_ftime option does the inverse of the
The -set_ftime option takes the forecast time stamp
in text format, and alters the grib metadata.
Types of time stamps
The simplest forecast time stamps are for an analysis and a forecast at a point in time.
Note that wgrib2 considers a "0 hour forecast" to be an "analysis". In theory, code table
4.3 could be used to determine whether the field is a "0 hour forecast" or "analysis".
In practice, models like the GFS will use an analysis for the initial conditions, and
the "0 hour forecast" will be the same as the analysis. So code table 4.3 is ignored.
Single analysis or single forecast
The next simplest time stamp involves an statistical operator and single time period
from a forecast. Processing is done continously or at a high frequency.
- anl analysis
- 12 hr fcst 12 hour forecast
Single forecast, examples of common statiistical operators (Code Table 4.10)
Grib allows for processed sets of data. For example, you have
have analyses from May 1, .., May 31. You can create a grib message which is
the May mean. You can then combine the May means from 1991..2020, and
produce a 1991-2020 climatology. Another example is the take the forecasts
from a single forecast run. You can take days 7..14, and create the 2nd week forecast.
- 174-180 hour acc fcst: accumulated forecast between 174-180 hours
- 174-180 hour ave fcst: average forecast between 174-180 hours
- 174-180 hour max fcst: maximum forecast between 174-180 hours
- 174-180 hour min fcst: minmum forecast between 174-180 hours
- 174-180 hour OPR fcst: operator of forecast between 174-180 hours, OPR from Code Table 4.10
Multiple analyses(forecasts), constant time between analyses (forecast start times)
The next type of forecast time stamp is used for processing forecast from a single forecast run.
The "(xxx)++" notation means that xxx is incremented to the next forecast time.
- 240@3 hour ave(anl),missing=0 average of 240 analyses with 3 hour interval between analyses, no missing analyses
- 248@3 hour ave(6 hour fcst),missing=0 average of 240 6-hour fcsts with 3 hour interval of forecast start times
- 248@3 hour ave(0-3 hour ave fcst),missing=0, average of 240 "0-3 hour ave forecasts", with 3 hour interval
between forecast start times
Multiple forecasts from a single forecast run
LAF (Lagged Average Forecasts) which are forecasts with different forecast start times and
the same verification time. This is the easiest to produce ensemble forecast.
- 9@1 hour max(0-1 hour max fcst)++,missing=0: max(0-1 hour max fcst, 1-2 hour max fcst, .., 8-9 hour max fcst)
Note: this can be encoded as "0-9 hour max fcst"
Lagged Average Forecasts, old to new forecasts
- LAF[..]++ reference time++, fcst_time-- (verification time does not change)
Lagged Average Forecasts, new to old forecasts
Grib allows fields that are produced by multiple statistical operators.
This is how you can generate climatologies.
- LAF[..]-- reference time--, fcst_time++ (verification time does not change)
- 30@1 year ave(124@6 hour ave(anl)),missing=0: reference time is 1981050100
"124@6 hour ave(anl)" is the average of (00Z May 1, 06Z May2, .., 18Z May 31), month average
"30@1 year ave(X)" is the 30 year average of the X
So a 1981-2010 May climatology has been precisely defined
- 30@1 year ave(31@24 hour ave(anl)),missing=0: reference time is 1981050100
This is the 1981-2020 climatology of the May 00Z analyses
It was noted that the old ftime had problems with the more complicated time stamps and
ftime2 was developed as the replacement. Currently (5/2021), you can use the
old ftime by -ftime1 and the new ftime by -ftime2 .
The default -ftime will use -ftime2.
The -config option will tell which version is the default.
Note: -ftime is called by many other options like -s and -S.
$ wgrib2 grib2.polar -ftime
1.1:0:24 hour fcst
1.2:0:24 hour fcst