In shallow soil layers of urban landscape, soil suction is greatly affected by vegetation and atmosphere. In this study, a field monitoring was conducted for recording variations of soil suction, air relative humidity and air temperature at different distances from a tree and at a constant depth under natural environmental conditions. A computational model is built using the genetic programming (GP) method for describing the functional relationship between soil suction and the four input variables. Based on the performance analysis, the efficiency and reliability of the developed computational model are validated. It can be concluded that the proposed computational model based on the artificial intelligence can sufficiently describe the relationship between the soil suction as a crucial parameter in geotechnical engineering and four input variables with acceptable errors.