Main

Data Quality Control of in-situ observations.

Errors in meteorological data can be divided in (almost copied from Gandin, 1988):

• Random errors: caused by many independent factors; distributed more or less symmetrically around zero; do not depend on the measured value. In addition, meteorological data may present random errors correlated in space. Random errors are due to the joint action of:
  • Observational errors: inherent in all data.
  • Representativeness errors: deviations caused by small-scale perturbations (also called micrometeorological error or representativity error or representation error). Their variance is significantly higher than that of observational errors.
• Systematic errors: caused by scale shift of the instrument and an influence of some more or less persistent factor which is not accounted for (or accounted for imprecisely); distributed asymmetrically with respect to zero; often persistent in time. Correlated random errors often behave like systematic errors, however they are usually not persistent in time.
• Rough errors (or large errors): caused by malfunctioning of measuring device and by mistakes during data processing, transmission and reception. Very large rough errors are usually referred as gross errors.

TITAN makes available to the user several tests aiming at detecting those observations that are most likely affected by: gross errors or rough errors; large systematic errors; large representativeness errors.

Good to know:

• the tests are applied sequentially and each test gets its own code, which is assigned to the suspicious observations.
• different test parameters (e.g., thresholds) can be specified for different data sources (or observation providers).

Command line

Typically, a command line looks like:

$>./titan.R --input.files fileIN_1.txt fileIN_2.txt ... fileIN_M.txt --output.file fileOUT --config.files test_1.ini test_2.ini ... test_J.ini --fg.files fgfile_1.ini fgfile_2.txt ... fgfile_B.txt

• --input.files The text files with input data for the in-situ observations to check. One for each observation provider. More info here.
• --output.file The text files with the titan results. More info here.
• --config.files The configurations files for the test. Rather flexible, could be one file or several files (e.g. one for each test). Note that all the input arguments in the configuration files can also be submitted as command line arguments.
• --fg.files The configuration files for reading the background (or first-guess) data on regular grids. More info here.

References

Gandin, L.S., 1988: Complex Quality Control of Meteorological Observations. Mon. Wea. Rev., 116, 1137–1156, https://doi.org/10.1175/1520-0493(1988)116<1137:CQCOMO>2.0.CO;2