RR_24 is the precipitation collected over the last 24 hours. The Qc1 check precipcollected_flag identifies if an RR_24 value is valid (fd=1), erroneous (fd=3) or is the accumulation (fd=2) of a number (N) of consecutive 24 hour periods, where N>1. For the case fd=2, the accumulated value shall be distributed over the N 24 hours periods.
For each day the calculated
RR_24= model_value*observed_accumulation/sum_of_model_values
where the model values for the N missing values of RR_24 are calculated by an interpolation from the other stations making measurements at the same time and for which fd=1. At present, the interpolation uses an inverse 2-D squared distance weighting. Other weighting and interpolation techniques might be added later.
The calculated RR_24 updates the corrected value in the kvalobs database.
A warning is given if fmis=0 and original!=-1.
The standard configuration options detailed in the user guide also apply to Redistribution.
Further configuration options are:
| Option | Default | Description |
|---|---|---|
| endpoint_cflags | fmis=[04]&fd=2&fhqc=0 | specify the flags for accumulation endpoints. Endpoints not matching these flags will not be detected |
| endpoint_uflags | no constraints | |
| missingpoint_cflags | fmis=3&fd=2 | specify the flags for the missing points before the accumulation |
| missingpoint_uflags | no constraints | |
| neighbor_cflags | fd=1 | specify requirements for the neighbor stations used for the interpolations. Neighbor stations that do not match these flags will be ignored. |
| neighbor_uflags | U2=0 | |
| warn_and_stop_cflags | fhqc=)0( | define for which flags for the accumulation point or the missing points before the accumulation the only action should be to print a warning and stop |
| warn_and_stop_uflags | no constraints | |
| update_flagchange | fd=7;fmis=3->fmis=1 | flags to change for the accumulation endpoint and the missing points |
| ParamId | — | selects the parameter ids to use. The only value tested here is 110 (RR_24). |
| tids | — | accepted typeids |
| measurement_hour | 6 | specifies the hour at which the data are expected. Series containing data with other times will be rejected with a warning. |
| min_neighbors | 1 | specifies the minimum number of neighbor stations with good data |
| max_neighbors | 5 | specifies the maximum number of neighbor stations to use |
| InterpolationDistance | 100 | radius in km inside which to search neighbor stations |
| warning_distance_closest_neighbor | 50 | minimum neighbor distance to print an INFO message about not so good neighbors |
| days_before_no_neighbor_warning | 14 | specifies the number of days after which a WARN is written to the logfile if there are not enough neighbor data |
| days_before_redistributing_zeroes_warning | 14 | specifies the number of days after which a WARN is written to the logfile if a non-dry accumulation has only dry neighbors |
This algorithm has only been tested for paramid 110.
There might be many different WARN messages in the log file /var/log/kvalobs/kvqc2d.log from this algorithm. These messages are (as of version 1.3.14pre3):
original = missing/rejected for endpoint – an endpoint (i.e. a row matching endpoint_?flags) has a missing original value; this might also indicate a row which somehow obtained the same flags as an endpoint.fmis=0 and original!=-1 for endpoint – fmis should only be 0 for dry accumulations flagg-specfd=2 and original!=corrected for endpoint – redistribution has apparently been performed without setting fd=7endpoint matches warn_and_stop_flags or missing point … matches warn_and_stop_flags for accumulation – a row matches warn_and_stop_?flags, which is never by defaultmissing point … has different fd flag than endpoint – a redistribution seems to have been done without setting fd flags consistenlymissing point … has fmis!=3 while fd=2 for endpoint – normally, all missing rows (if they exist) should have fmis=3 before redistributionmissing point … has fmis!=1 while fd=7 for endpoint – normally, all missing rows should have fmis=1 after redistributionfound … but expected … rows with corrected values for accumulation … – normally, only one row should have a corrected value if the endpoint has fd=2, and all rows if fd=6/7redistributed sum … starting … differs from original in endpoint … – the sum of corrected values in the accumulation period does not match the original value in the endpoint row; unless fhqc!=0, an attempt will be made to fix this automaticallyfhqc mess for accumulation from .. to endpoint … – some rows in the accumulation period have fhqc=0, while others have fhqc!=0suspicious row … before endpoint ..; giving up – a row with missing original value has been found before the assumed accumulation period; to avoid a wrong redistribution in case the correct start date for the accumulation has not been detected, the algorithm stopsno really good neighbors at obstime … for accumulation ending … – the nearest neighbor is quite far away (more than warning_distance_closest_neighbor)accumulation … > 0 would be redistributed to zeros for endpoint … – there is non-zero accumulated precipitation, but all accessible neighbor stations are dry so that the redistribution cannot be performedcould not avoid difference of … between distributed sum and accumulated value at endpoint … – some precipitation has been lost due to numerical (rounding) problems