(1) Why must radiosonde data be checked for accuracy?
Stations in the NWS network take approximately 75,000 observations each year. Each observation can contain thousands of measurements of air pressure, temperature, relative humidity, wind direction, and wind speed. Radiosonde defects, ground equipment problems, and/or atmospheric conditions aloft result in some soundings containing erroneous data. Some errors are easy to detect, such as excessive cooling of the temperature with height (i.e. super-adiabatic lapse rate). Others, such as identifying temperature errors caused by solar radiation, can be difficult.
(2) How does the NOAA/National Centers for Environmental Prediction (NCEP) quality control upper-air data?
Before ingesting rawinsonde data into numerical weather prediction models, NCEP corrects temperature data (for some radiosonde types such as Sippican) for solar and infrared radiation and checks the quality of the data using objective and subjective techniques.
Solar and infrared radiation significantly affects the accuracy of radiosonde temperature data at heights over 15 km. The temperature errors caused by radiation can exceed 1ºC. These errors are difficult for upper-air observers to detect and they are not expected to correct the data. Full correction of the errors is difficult, but NCEP applies an adjustment to the temperature data to reduce the error.
For checking data quality, NCEP utilizes computerized comparisons of the actual upper-air data received from an upper-air station with that generated from a 6-hour numerical weather prediction model. These automated data quality control programs are then combined with knowledge of how the real atmosphere operates. NCEP meteorologists examine the validity of the upper-air data and check the continuity of model comparisons from cycle to cycle. This provides the human element necessary to balance the computer's prognosis. If the temperature, geopotential height, and/or wind data compare poorly, the upper-air data are either deleted or corrected. Typically, most data deletions are for temperature and height data above 10 km. Often NCEP meteorologists find bad data not flagged by the computer or determine that data flagged as bad by the computer are actually good and should be kept. It should be noted that at the present time, RH data are not as closely checked for accuracy as are the other radiosonde data because data analysis tools have yet to be developed.
Data quality control methods at NCEP are not perfect and errors do pass through undetected. Furthermore, NCEP does not quality control soundings before plotting them on upper-air charts. If bad data is disseminated from a station, it will appear on the charts. Thus, it is very important that upper-air observers make every effort to check the validity of the sounding data before they are disseminated.
(3) Does the National Centers for Environmental Information (NCEI) quality control the upper-air data archive?
Yes. Similar to NCEP, NCEI relies on a mix of objective and subjective procedures to quality control rawinsonde data before it is archived. Briefly, each sounding undergoes a series of checks to determine the plausibility of the data. If data is determined to be suspect or clearly erroneous, it is flagged as such, but is not deleted from the archive. Unlike NCEP, NCEI routinely checks RH data quality (primarily for rapid, unrealistic fluctuations in RH). As with NCEP data quality checks, NCEI data quality control is not perfect and erroneous data will not always be flagged.
(4) What can the upper-air station do to improve data quality?
Observers must avoid disseminating poor quality observations. Upper-air station staff should not assume that NCEP or NCEI will detect and correct all the errors in the soundings. Data quality control at NCEP and NCEI will not capture all data errors. Furthermore, NCEP does not quality control soundings before plotting them on upper-air charts. If bad data is disseminated from a station, it will appear on the charts. Thus, it is very important that upper-air observers make every effort to check the validity of the sounding data before they are disseminated.
Station staff must follow all standard NWS procedures for detecting and editing as much erroneous data as possible. However, not even the best observers can detect all data errors. Effective data quality control involves the upper-air observer, NCEP, NCEI, and NWS administrative office support on the regional and national level. Station staff should also have an adequate equipment maintenance program to keep the radiosonde ground tracking system in good working order. Assistance in troubleshooting problems with the radiosondes, ground tracking system or data communications should be directed to the appropriate NWS Regional Office.
(5) NCEP is frequently rejecting upper-air data from a particular station. Why is this happening and what can station staff do to correct the problem?
Here are some possible causes and corrective actions:
(a) Defective radiosondes or ground equipment.
The NWS upper-air network uses about 75,000 radiosondes each year. Some of them will be defective. Likewise, the ground tracking equipment is old and prone to failure or misalignment. Station staff should check to see if there is a correlation between the date when the data rejections started and when a new shipment of radiosondes was first used. If there is clearly a match, the NWS Regional Upper-air Program Manager should be contacted as soon as possible. If frequent winds errors are occurring, the automated radio theodolite (ART) should be checked for correct alignment.
(b) Solar/IR radiation on the radiosonde temperature sensor.
Solar/IR radiation can affect accuracy of the temperature data (and height data) at altitudes typically above 200 hPa. Corrections are applied to the data, but sometimes they do not work adequately. If temperature/height rejections are mostly above 200 hPa and are cyclical in nature (high in the summer, low in the winter or vice versa) then solar/IR radiation is probably causing the errors. There is little station staff can do to correct this problem. NWS HQ is working with the radiosonde manufacturers to improve stratospheric temperature measurements.
(c) Incorrect station data.
All station data in the upper-air computer (release point latitude/longitude, release point height, release point pressure correction, etc.) must be correct and kept up to date. Station staff should pay special attention to the pressure correction between the station barometer height and the height of the balloon release point and make sure it is correctly entered into the upper-air computer station data file. If this value is wrong, upper-air height errors will result. NWS HQ works with NCEP to make sure the station data used for the numerical weather prediction models is correct.
(d) Observer Error.
Improper handling of the radiosonde, balloon, or flight train prior to release can introduce data errors. Also, incomplete or incorrect data quality control by the observer can cause problems as well. This includes:
- Incorrect measurement of the station pressure. This value must be as accurate as possible to ensure that the radiosonde is properly base lined.
- Incomplete deletion of bad radiosonde data. All observers are required to check the validity of their soundings prior to dissemination. The upper-air users guide provides instructions on how to identify and edit bad data.
- Incorrect deletion of radiosonde data. Some observers will try to identify and keep one or two "good" points in a layer of clearly erroneous data so that the data profile is preserved by interpolation of data through the missing data layer. Trying to identify which points are good or bad is very difficult and not necessary. All data in the erroneous layer(s) must be deleted.
- Making the radiosonde flight train length too short. The sun can warm the balloon skin to temperatures well above that of the ambient atmosphere. The radiosonde flight train length must be of sufficient length to ensure that heat radiating from the balloon will not contaminate the temperature measurements, especially in the stratosphere.
- Incorrect inflation of the balloon. Under-inflation will cause slow ascension rates, which can lead to inadequate sensor ventilation and loss of wind data (from limiting angles). In most cases, balloons should be inflated so that they have an average ascension rate no less than 275 meters/minute.
- Manipulation of the radiosonde sensor boom or other radiosonde components in a manner not in accordance with the radiosonde preparation instructions. With good intent, some observers manipulate the sensor boom in the belief that it will provide better data or do no harm. In most cases, these changes do not work and cause erroneous data. Following all the radiosonde preparation instructions will help ensure a problem free sounding.
Lastly, there have been cases where upper-air data errors are associated with one or two observers at the station who are making some honest mistakes. This can be verified by comparing the observer log with the data error reports issued by NCEP over a several week period. If there is a match, corrective action (i.e., retraining) needs to be taken by the station manager.
(6) What is the NWS upper-air station performance score?
The NWS Observing Systems Branch developed the station performance score program to help monitor and improve NWS upper-air data availability, quantity, and quality. One of the goals of the program was to keep the workload required to score the stations to a minimum and automate the procedure as much as possible. Realizing that many stations across the network use different ground equipment, balloon inflation gas, and radiosondes, the variables and weighting factors that make up the equation were adjusted accordingly to make the scoring as fair as possible to all upper-air stations.
(7) If a NWS upper-air station receives a low performance score are upper-air station staff to blame?
Not always. A low score may be caused by a number of factors beyond the control of station staff. Sometimes undetectable defects in radiosonde components and balloons, weather conditions, or problems with the data communications lines can cause a low score. Therefore, the station score reflects the performance of the entire upper-air program, which includes upper-air observers and program managers, as well as staff involved in equipment maintenance, logistics, and data communications. Upper-air station managers should use the scores as only one of many tools for assessing how well their station is performing.
(8) Why should a station's performance score be reduced when poor or missing observations were not caused by station staff?
First, NWS management realizes that missing, late, or poor quality observations are not always the observers fault. ALL upper-air stations, at one time or another, will get an "unfair" hit on their monthly score. It is random and this is helping keep the network scores as fair as possible. The final score will be an average over a full year and this will help smooth out low scores that may have occurred during the year.
One of the goals of the station performance score program is keep workloads required to generate the scores to a minimum. Going through each sounding to determine who or what caused a missing/poor quality sounding and then adjusting the scores accordingly would be an overwhelming task (the NWS network disseminates over 5,500 soundings/month). Therefore, generation of the scores is automated as much as possible.
Back to the Upper-air Observations Main Page
