Below are Skew-T and Pseudo-Adiabatic charts of upper-air radiosonde soundings that show examples of erroneous temperature, dew point, or winds data. All sounding plots are provided courtesy of the University of Wyoming (bad link).
Temperature Errors
An example of a temperature sensor or ground system failure is shown in the sounding above. At about 350 hPa the temperature rapidly warms unrealistically and becomes erratic. Above this level, the temperature data seems to recover, but it's difficult to determine at what level the data becomes useful again. A pressure sensor failure may also cause a similar temperature profile.
The sounding plot above shows an example of the wet bulb" effect. Note at about 750 hPa and again at about 680 hPa the temperature profile shows rapid, unrealistic cooling or a super-adiabatic lapse rate. As the radiosonde ascended though a cloud or precipitation, the temperature sensor became wet. When the temperature sensor was exposed to drier air at about 750 hPa and 680 hPa, water on the sensor began to evaporate, causing the temperature to cool rapidly.
Above is another sounding showing the "wet-bulb" effect. In this case, the radiosonde exited a cloud and a super-adiabatic lapse rate occurred between about 680 hPa and 650 hPa.
The sounding shown above on the left may look good, but actually has a temperature profile that is too cold. This was determined by comparing this sounding with other near-by soundings taken at the same time. On the right is one such sounding. Note that its temperature profile is warmer. Comparing soundings with other nearby soundings can be helpful in detecting bad data. The cause for the cold-biased temperatures may be from faulty sensor calibration at the factory or a radiosonde defect.
Above are two soundings taken 12 hours apart from the same NWS upper-air station. Between the sounding times, a cold front passed over the station. Note how much warmer the sounding on the right is from the one on the left (taken 12 hours earlier). The freezing level increased from about 2,700 to 4,500 meters and mandatory height levels increased anywhere from 50 to 200 meters. From this information and examination of upper-air charts, it's clear that most of the 12:00Z sounding data is erroneous. A poorly calibrated radiosonde or sensor defect is likely the cause.
Operational soundings within thunderstorms are avoided as much as possible since they provide data that is not representative of the surrounding environment. A radiosonde that enters or goes near a thunderstorm will likely produce the sounding shown above. Note the erratic temperatures (i.e., super-adiabatic lapse rates) occurring between about 650 hPa and 400 hPa. The wind profile within this layer appears erratic as well. The erratic profiles are likely the result of the "wet-bulb" effect on the temperature sensor and rapidly changing atmospheric conditions typically found inside thunderstorms.
Dew Point Errors
The sounding above shows an example of a humidity sensor failure beginning at the 700 hPa level. At about this level, the radiosonde exited a cloud and the dew point rapidly dropped to below - 40 degrees C. Dew points stayed below - 40 degrees until about the 450 hPa level. This dry layer is likely real, but the radiosonde humidity sensor (a carbon hygristor) reported relative humidity values biased too low. Chamber tests have been done by the NWS that revealed this type of sensor failure.
A water or ice covered relative humidity sensor can cause the resulting dew point temperatures to be too moist. Sometimes the sensor recovers (i.e., dries out) and provides accurate measurements. However, in the example above (showing data plotted to about 15 hPa), the relative humidity sensor (a H-Humicap) became water or ice covered near the beginning of the sounding and never dried out. The result are unrealistic, moist dew point depressions throughout much of the sounding, especially in the stratosphere (above 150 hPa) where there is typically little moisture.
As part of an upper-air sounding, a complete surface weather observation (including surface dew point) is taken near the time of balloon release. This observation can be helpful in identifying erroneous radiosonde data near the surface. In the example above, note the rapid drying just off the surface (i.e., rapidly falling dew point temperature) and the very dry layers at 850 hPa and at about 675 hPa. The rapid drying near the surface is unrealistic because the temperature profile (a dry adiabatic lapse rate) and wind speeds (10 knots) near the surface indicate that the layer is well mixed and such a moisture profile should not occur. In this example, the relative humidity sensor (a H-Humicap) was reporting values biased too dry.
Sometimes the relative humidity sensor or some kind of radiosonde defect will cause the dew point data to become erratic as shown in the example above. The dew point plot above shows rapid, unrealistic changes in moisture just off the surface.
Wind Errors
A radio-theodolite is used to track the position of the ascending radiosonde. If the observer fails to get the radio-theodolite "locked" on the radiosonde after balloon release, a wind profile similar to the one shown above will likely result. Note the unrealistically high wind speeds below 500 hPa (one level at about 525 hPa shows 220 knot winds) and the erratic wind directions above 500 hPa.
If a radiosonde gets near or enters a thunderstorm, a sounding similar to one shown above will likely result. Winds inside or near thunderstorms are typically quite different from the surrounding environment. In this example, note the high wind speeds (about 65 knots) around 700 hPa. Comparisons of this wind profile to surrounding soundings taken nearby showed that these winds were too high. They more likely represented the airflow near or within the thunderstorm, rather than the synoptic-scale environment.
