Atmospheric monitoring is an integral part of the design of the Cherenkov Telescope Array (CTA), as atmospheric conditions affect the observations by Imaging Atmospheric Cherenkov Telescopes (IACT) in multiple ways. The variable optical properties of the atmosphere are a major contribution to the systematic uncertainty in the determination of the energy and flux of the gamma photons. Both the development of the air-shower and the production of Cherenkov light depend on the molecular profile of the atmosphere. Additionally, the rapidly changing aerosol profile, affecting the transmission of the Cherenkov light, needs to be monitored on short time scales. Establishing a procedure to select targets based on current atmospheric conditions can increase the efficiency of the use of the observation time. The knowledge of atmospheric properties of the future CTA locations and their annual and short-term variations in advance is essential so that the atmospheric calibration can be readily applied to first scientific data. To this end, some devices are already installed at one or both of the selected sites. These include a Sun/Moon photometer, all-sky cameras for cloud detection and stellar photometry, detectors of night sky background and weather stations; a small optical telescope to measure atmospheric extinction using stellar photometry (FRAM) is ready to be deployed soon. Aerosol climatology at both sites will be soon assessed in a dedicated campaign using the ARCADE LIDAR. The atmospheric calibration strategy for the CTA contains dedicated instruments and methods for each task, namely: the combination of the FRAM and Raman Lidars to characterize the aerosol extinction profile along the line-of-sight of the observed target, the combination of an all-sky-camera and a ceilometer to characterize the atmosphere in the direction of possible future observations. These will be complemented by data from satellites and global data assimilation models (validated by a radiosonde campaign). Moreover, information about the atmosphere can be extracted directly from the data using the Cherenkov Transparency Coefficient (CTC) method.