QualiMET 2.0 The new Quality Control System of Deutscher Wetterdienst Reinhard Spengler Deutscher Wetterdienst Department Observing Networks and Data Quality Assurance of Meteorological Data Michendorfer Chaussee 23 D-14473 Potsdam Tel.: +49 (0)69 8062-5200 E-mail: reinhard.spengler@dwd.de 1
Overview 1. Background of planning and creating QualiMET 2.0 2. Difference between QualiMET 1.3 and QualiMET 2.0 3. Decision tree & usage of remote sensing data / nowcasting data 4. Calculation of climatological values 5. Co-operation with other NMSs 2
Background of planning and creating QualiMET 2.0 The previous system has been in operation for about 15 years. We had many human resources for visual observations and for operating interactive QC procedures. DWD decided to run a fully automated observing network as of 2020. Interactive QC is currently distributed to seven regional offices Future: one central QC group Human QC takes to much time and is often biased by subjective judgement. 3
The previous system QualiMET 1.3 Station Data QualiMET 1.3 QualiMET- Client Doubtful values Correct values User database 4
The new system QualiMET 2.0 Station Data QualiMET 2.0 QualiMET- Client Doubtful values Correct values User database 5
Decision tree & usage of remote sensing / nwc data (1/6) To be continued: 4 levels of Quality Control 1. Real-time at station site (fully automated) 2. Real-time at central office (24/7, semi-automated) 3. Non real-time at central office ( max. 3 days) 4. Climatological Quality Control ( 1 year) To be continued: 5 steps of Quality Control 1. Completeness, availability 2. Climatological limits 3. Temporal consistency 4. Internal consistency 5. Spatial consistency 6
Decision tree & usage of remote sensing / nwc data (2/6) Usage of satellite data (cloud cover, cloud type, cloud mask, radiation, sunshine duration, etc.) at levels 2, 3 and 4 to check hourly, daily and monthly values to decide if available data is acceptable to get substitute data to correct inaccurate data to get substitute data to close gaps in time series Usage of radar products at levels 2 and 3 to check hourly and daily values in the same way as satellite data Usage of station site classification (CIMO) and other metadata example on following slide 7
Decision tree & usage of remote sensing / nwc data (3/6) - Example: sudden drop in temperature 5cm above ground Weather conditions - no clouds full sunshine - no precipitation - nearly calm - 8
Decision tree & usage of remote sensing / nwc data (4/6) - Example: sudden drop in temperature 5cm above ground Decision tree in previous system Drop of temperature too large! precipitation (gauge)? yes no data correct full sunshine? flagging (end) yes operator has to decide result?? no data correct flagging (end) 9
Decision tree & usage of remote sensing / nwc data (5/6) - Example: sudden drop in temperature 5cm above ground Decision tree in future system Drop of temperature too large! precipitation (gauge)? yes data correct no precipitation (radar)? flagging (end) yes data correct no full sunshine? flagging (end) flagging (end) yes other influence? (search in meta data) yes automatic correction no decision of operator no data correct flagging (end) 10
Decision tree & usage of remote sensing / nwc data (6/6) - Example: sudden drop in temperature 5cm above ground Result - Drop in temperature was caused bei shadow of a radio tower (see picture) 11
Remote sensing / model data - Check-algorithms for precipitation by using radar products 12
Remote sensing / model data - spatial check of daily precipitation with a new linear model 40 mm 50 mm = 30 mm 30 mm 0 mm Information from neighboring station (considering dataset of the last 10 years) Search / identify of suitable predictors Interpolation of Data into a grid of 1km 40 mm 50 mm 30 mm =?? mm 0 mm 13
Results from Sensor calibration checks - should these data be considered in QA of time series? Maintanance interval of wind-sensor is 12 months The re-calibration check detects an non-acceptable deviation in wind-speed How should we deal with such a result relating to the time series, especially of climatological data??? 14
Calculation of climatological values All climate data are calculated using data sets from the synoptic stations. This means that QC is based on data with the highest possible temporal resolution: 1-minute data, 10-minute data,... Any change in high temporal resolution data will automatically lead to adaptation of the climate data (possible for up to 30 days back) A quality flag is assigned to each measurement value. During the calculation of climate data, this information is passed on to the condensed values. The concept of data condensation and quality flagging will be expanded until 2017 to be applicable to the new procedures by using remote sensing and model data. In future, all users of the data will see whether the values are original data or have been corrected or added. 15
Co-operation with other NMSs The new developments to the method and software require intense coordination, also at international level. For this reason, we have intensified our contacts to MeteoSwiss and ZAMG during 2014. I would be very pleased if we could work together on improving our procedures for the quality control of meteorological data and data management. 16
Thank you very much for your attention Are there any Weitere Fragen? questions? Vielen Dank für Ihre Aufmerksamkeit