NIKLAS - an automatic data control tool
The software NIKLAS is a module for real-time and non-real-time quality check of meteorological data. It was first developed for the Mosel river forecast and warning system for the Landesamt für Umwelt, Wasserwirtschaft und Gewerbeaufsicht Rheinland-Pfalz, Germany. While for small amounts of input data quality control can be conducted manually by experienced observers, this is not possible for larger amounts of data and real-time applications. Automatic algorithms can reduce the manual effort by identifying the suspicious cases that have to be examined by human observers. Data for real-time applications also can be controlled assuring a certain level of quality. NIKLAS validates time series for the following parameters:
- global radiation
- sunshine duration
- air temperature
- dew point temperature
- relative humidity
- wind speed
- air pressure
The quality check procedure starts from simple cases before tackling the more difficult ones, following the steps:
- Detection of gaps in the data
- Detection of physically impossible values
- Detection of constant values
- Detection of values above set thresholds
- Detection of improbable zero values
- Detection of unusually low values
- Detection of unusually high values (Values of 6. and 7. may be correct, though).
Precipitation measurements from 4 stations: the red curve is an example for unrealistically constant values.
Additional check options of NIKLAS are variability and inner consistency, using known relationships between parameters as e.g. temperature and dew point. The upper figure shows an example of a station record of constant precipitation values. The thick red line marks the station recording the unrealistic precipitation data.
As result of each test the examined data records are classified into three quality levels: valid, low/limited quality or bad quality. NIKLAS as a pure quality check tool excludes records with bad quality from the time series and does not replace them with plausible estimations. The NIKLAS software comes as a Windows application with Graphical User Interface (see figure). It has successfully been applied for real-time (Mosel flood warning system, TIMISflood project) as well as for non-real-time (ExUS project) validation tasks (Quirmbach et. al. 2009).