Satellites and weather balloons measure temperature in the upper air, far above the surface, taking measurements through the troposphere and stratosphere.
The atmosphere has a number of layers. The troposphere is closest to the surface and above it lies the stratosphere. The troposphere is thickest in the tropics - around 20km deep - and thinnest near the poles - where it is around 7km deep. In the troposphere, temperatures decrease with height. In the stratosphere, temperatures increase with height as the air absorbs ultraviolet radiation from the sun. Temperatures in these layers are important for understanding and predicting the weather and climate.
Global mean temperatures in the troposphere have increased overall since the start of the record. As with surface temperatures, there are short-lived peaks associated with El Niño events and cooling periods associated with La Niña and occasional volcanic eruptions.
Global stratospheric temperatures have decreased overall, although there has been relatively little change in the observed temperaturessince the late 1990s. Overlain on these long-term changes, there are sharp warm peaks associated with the major volcanic eruptions of El Chichón in 1982 and Mt. Pinatubo in 1991.
Atmospheric temperature can be measured directly by weather balloons and indirectly by satellites. The two series shown here are based on satellite measurements of microwave radiation emitted by oxygen molecules in the atmosphere. The intensity and frequency of the microwave radiation detected by the satellite are related to the temperature and the altitude of the oxygen molecules. By measuring the intensity at different frequencies the microwave measurements can be used to work out how temperature changed at different altitudes. Unforunately, because of limitations in the methods, temperatures can't be estimated at precise altitudes. Different altitudes get smudged together into broad layers for which we can work out the temperatures. The series shown above are for broad layers of the lower troposphere and the lower stratosphere.
There are numerous difficulites that arise in trying to accurately assess long-term change in atmospheric temperatures. Because of the smearing between altitudes, the lower tropospher measurements are influenced by surface temperatures. The surface temperature varies during the day, cooling at night and warming once the sun comes up. As satellites age, their orbits drift meaning that they measure at different times of day and, as a result, the contribution of the surface warms or cools leading to potentially spurious changes in the calculated tropospheric temperatures. The groups who work on the satellite data sets apply corrections during processing to account for these effects. In a similar way, measurements of the mid and upper troposphere are affected by radiation from the cooling stratosphere.
There are other difficulites with assessing long-term change from satellite instruments. For example, when one satellite is replaced by another a careful assessment needs to be made of any potential "jumps" in the measurements. These can be assessed by comparing measurements from the two satellites during an overlap period. Unfortunately, such overlaps are not always long enough to get a perfectly reliable transfer.
Temperatures in the upper air are affected by a number of long term drivers. Each one affects the stratosphere and troposphere in distinctive ways. Greenhouse gases are expected to warm the troposphere and cool the stratosphere. In contrast, increases in solar radiation would warm both troposphere and stratosphere. Loss of ozone in the stratosphere (caused by ChloroFluoroCarbons, CFCs), is also expected to cool the stratosphere. Aerosols in the stratosphere from volcanic eruptions warm the stratosphere and cool the troposphere.
In summarising the inferred causes of changes in the troposphere, the IPCC AR5 said "It is likely that anthropogenic forcings, dominated by GHGs, have contributed to the warming of the troposphere since 1961 and very likely that anthropogenic forcings, dominated by the depletion of the ozone layer due to ozone-depleting substances, have contributed to the cooling of the lower stratosphere since 1979." but also note that "Uncertainties in radiosonde and satellite records make assessment of causes of observed trends in the upper troposphere less confident than an assessment of the overall atmospheric temperature changes."
More detailed information is available from Remote Sensing Systems.