An atmospheric sounding is a vertical profile of temperature, moisture, pressure, and wind from the surface up through the stratosphere, taken from a single column of air at a single time. The profile is usually plotted on a Skew-T Log-P diagram with wind barbs alongside so that the temperature, dewpoint, and wind structure can be read together.
Where the data come from
The most common source is a radiosonde, a small instrument package flown on a weather balloon that transmits pressure, temperature, humidity, and position back to a ground station as it rises. The National Weather Service releases radiosondes at roughly 90 sites across the country twice daily, at 00 and 12 UTC, and additional flights are launched around expected severe weather episodes.
Not all soundings come from balloons. Aircraft ascent and descent profiles, ground-based profilers, satellite retrievals, and VAD wind profiles from WSR-88D radars all produce partial soundings at higher frequency than a RAOB. Numerical model output can also be sampled at a point to produce a forecast sounding for a location and time when no observation exists.
What a sounding actually shows
The temperature and dewpoint curves together define the moisture and stability of every layer. Where the two curves are close, the air is nearly saturated. Where they diverge, the layer is dry, and if a dry layer sits above a moist boundary layer, the profile hints at a cap or at potential for evaporatively cooled downdrafts.
The wind profile shows how flow evolves with height and is the source of every shear and helicity diagnostic. Lifted parcel calculations use the temperature and dewpoint profiles to trace an imaginary parcel upward, and the area between that path and the environmental temperature curve becomes CAPE, CIN, LCL, LFC, and equilibrium level.
How forecasters use it
A convective forecast starts by finding representative soundings for the region of interest and reading them for the ingredients: sufficient moisture in the low levels, sufficient instability aloft, sufficient shear, and either a weak cap or a way to break it. Comparing a morning sounding with an expected afternoon modification, using surface heating, moisture advection, and mid-level cooling, is a core piece of severe weather forecasting.
Model soundings extend that workflow forward in time and to locations without an observation. Meteorologists routinely pull a forecast sounding for the exact hour and grid point of interest, then compare it with the nearest real observation to check for model bias in the boundary layer or capping structure.
What a sounding does not do
A sounding is a single column. It cannot resolve mesoscale features between two RAOB sites, and a boundary just downstream can render the observed profile unrepresentative within an hour. Balloon drift also means a nominally local sounding can end up sampling air fifty or more kilometers downwind by the time it reaches the upper troposphere.
The instrument itself has known error characteristics. Sensors have small lags, solar heating can bias daytime temperatures aloft, and the humidity sensor loses skill at very cold temperatures. These matter less for gross ingredients and more for the fine detail of very dry or very cold layers. When something looks wrong in a sounding, checking neighboring sites and satellite is usually a better step than trusting a single feature outright.