Discovering new sources of high energy gamma rays of galactic (e.g. in the galactic center or the Fermi Bubbles) and extragalactic origin (such as quasars or blazars) is of great intereest to the astroparticle physics community. Currently the HAWC and LHAASO observatories, located in the Northern hemisphere, use detectors based on the Cherenkov effect in water to continuously detect secondary particles from atmospheric showers initiated by primary gamma rays in the atmosphere. Since there is no equivalent experiment in the Southern hemisphere, the future gamma-ray observatory SWGO will complete the sky coverage for high duty cycle and wide field of view observatories. Its proposed site above 4400 m will be located in the Southern Andes, with Chile, Argentina and Peru as candidate countries. The SWGO reference configuration consists of an array of water tanks with two circular cores: the inner core, reaching 160 m radius, with 88% sensitive area and the outer ring, reaching 300 m radius, with a 5% fill-factor. To identify gamma ray sources, primary particles need to be reconstructed from the air showers reaching the detector array, obtaining their energy, direction and type. A gamma/hadron separator describes the characteristics of the air showers to distinguish between gamma rays, considered as signal, and hadrons (i.e. cosmic rays) that are considered noise. This thesis proposes an alternative gamma/hadron separator variable to distinguish be- tween types of atmospheric showers by using the arrival time distribution of secondary particles reaching SWGO. To define the best new time-based variable we use the CORSIKA software to simulate the development of air showers in the atmosphere up to the arrival of secondary particles at the array of water Cherenkov tanks. The analysis was done using the geomagnetic conditions of the Imata candidate site in Arequipa, Perú, located at 4500 meters above sea level. We considered as primaries photons and protons with a vertical trajectory in the center of the array in the energy range from 1 to 100 TeV. The optimal separation parameter found is the time for the 15% percentile of arriving particles inside a ring of 100 to 150 m. Following the calculation and evaluation of the simulation sample, the recognized signal is ≳ 88% on average, and the background rejection is ≳ 90%. Both performances are comparable to using the standard muon count variable.