The current practice in fatigue crack growth assessment for metallic components usually starts with a sufficiently long initial crack, which is an strong assumption. However, especially for high cycle fatigue, the crack propagation at early stage towards the initial crack length can account for about 90% of the structural fatigue life. And considerable uncertainties in the fatigue life prediction come from the short crack regime especially the microstructurally short crack (MSC) regime where heterogeneous polycrystalline microstructure induce oscillating crack growth rate. Therefore the MSC growth and uncertainty from microstructural variability needs a better quantification for a certain material under different loading scenario. The present paper proposes a framework to quantify the time distribution to intial long crack (TDILC) by Sobol’ decomposition based reduced order modeling (SDROM) built upon microstructurally short fatigue crack growth model based on crystal plasticity finite element method (CPFEM) and fatigue indicator parameter (FIPs), in a faster and more tractable manner. The feasibility test is carried on a high strength low alloy steel aggregate of regular hexagonal grains with only orientation as random variable, under constant amplitude load. The effectiveness of the SDROM will also be discussed by comparing to brute force Monte Carlo Simulations.