Studies have proven that merging hands-on and online learning can result in an enhanced experience in learning science. In contrast to traditional online learning, multiple in-classroom activities may be involved in an augmented-reality (AR)-embedded e-learning process and thus could reduce the effects of individual differences. Using a three-stage AR-embedded instructional process, we conducted an experiment to investigate the influences of individual differences on learning earth science phenomena of "day, night, and seasons" for junior highs. The mixed-methods sequential explanatory design was employed. In the quantitative phase, factors of learning styles and ICT competences were examined alongside with the overall learning achievement. Independent "t" tests and ANCOVAs were employed to achieve inferential statistics. The results showed that overall learning achievement was significant for the AR-embedded instruction. Nevertheless, neither of the two learner factors exhibited significant effect on learning achievement. In the qualitative phase, we analyzed student interview records, and a wide variation on student's preferred instructional stages were revealed. These findings could provide an alternative rationale for developing ICT-supported instruction, as our three-stage AR-embedded comprehensive e-learning scheme could enhance instruction adaptiveness to disperse the imparities of individual differences between learners.