First, we need to understand what a lapse rate is. The atmospheric lapse rate is the change in temperature with height. For example, on average the temperature cools 6.5 C for every kilometer. Of course, this varies from location to location. We can use observations to determine the lapse rate of a particular location. The observations are gathered from a miniature weather-monitoring station (smaller than a shoe box) that is attached to a balloon and released into the air twice a day at various sites throughout the world (a radiosonde).

The value of the lapse rate is strongly dependent on the amount of water vapor in the air. Dry air cools at about 10 C/km (the 'dry adiabatic lapse rate'), while moist air usually cools at less than 6 C/km ('moist adiabatic lapse rate'). The word adiabatic means that no outside heat is involved in the warming or cooling of the air parcels.

Why are the two lapse rates different? Remember that water vapor in a rising parcel of air will condense when the air becomes cold enough. The phase change from gas to liquid takes a little work from the water molecules. As they are working, they release heat. The heat decreases the cooling that occurs in the air parcel. Therefore, a rising parcel of dry air cools faster than a moist parcel of air. And conversely, a sinking parcel of dry air warms faster than a sinking parcel of moist air.

Now apply these concepts to mountains. Moist air that reaches a mountain range is forced to rise. The rising air cools at the moist adiabatic lapse rate and eventually some or all of the water vapor will condense. If the conditions are right, a cloud will form and it will either rain or snow. The air continues to flow over the mountain range and eventually descends on the other side. The sinking air warms. Since the air has lost much of its moisture through precipitation, it warms at the dry adiabatic lapse rate, which is greater than the moist adiabatic lapse rate. This means the air is now drier and warmer than it was before it encountered the mountain range.