By using the outcrop shale of Longmaxi-3, a series of experiments including in-situ AFM Force Map, SEM-EDS before and after water vapor saturation and uniaxial/triaxial compression tests are conducted to discover the changing pattern of shale mechanical properties from micro and macro level. The current research first measures the micro elastic modulus distribution changes of shale under water condition, also verifying the water-softening effects. Expectation Maximization algorithm programmed by C language is utilized to define Gaussian Mixture Model of micro elastic modulus under different moisture. Self-Consistent Scheme by C program and Mori-Tanaka Scheme is adopted to upscale the micro modulus based on the results of EM. In the comparison with compression test results, M-T scheme behaves better on estimating modulus of originally dry shale core than SC scheme, however, the homogenization technique cannot achieve the accurate estimation of core sample modulus based on micro modulus. The combination of AFM morphology and SEM-EDS verifies the component CaSO4 of newborn particles when moisture is low. Due to strong heterogeneity of shale, the relationship between elastic modulus and water saturation is negative and non-linear, homogenization technique is not proper to upscale micro modulus of sample processed by water or other factors.