The environment of near space atmosphere is very complex. Its spatial and temporal changes are hard to be characterized and modeled. With 11 years of TIMED/SABER atmospheric density data, the climate means and standard deviations at 38°N in 20-100km were obtained by the method of global gridding and mathematical statistics. Quantitative results were used to represent and analyze the characteristics of static slow climate changes and dynamic transient atmospheric disturbances. The results show that the atmospheric density at 38°N varies remarkably with altitude, seasons and longitude. A modeling method was set up, where the atmospheric density in near space can be represented as the sum of the climate means and atmospheric disturbances. A self-regression model was established for the atmospheric random disturbances. Model simulations were taken and compared with the lidar-observed density data, showing that the modeling method is feasible.