Crowdsourcing incentive mechanism design has raised numerous interests from research communities in recent years. While most research focuses on contribution-based payment allocation, a solid crowdsourcing incentive mechanism should encourage users to both devote efforts to complete the task and refer other users to join into participation. In this paper, we adopt a data structure called incentive tree which has a unique advantage in incentivizing participants for solicitation. Furthermore, we consider the crowdsourcing scenario where the contribution model is submodular and time-sensitive, which is more realistic compared to the linear summation model adopted by previous works. Under this model, we design a reward mechanism based on the incentive tree, and prove that this mechanism satisfies several economic properties such as continuing contribution incentive, continuing solicitation incentive, θ-reward proportional to contribution, early contribution incentive, and sybil-proofness. We implemented our incentive mechanism and conducted extensive performance evaluations. The evaluation results confirm our theoretical analysis.