Erosion of cohesive sediments is a ubiquitous phenomenon in estuarine and intertidal environments. Several methods have been proposed to determine the surface erosion threshold (τc0), which are still debatable because of the numerous and uncertain definitions. Based on erosion microcosm experiments, we have compared three different methods using (1) eroded mass (EM), (2) erosion rate (ER), and (3) suspended sediment concentration (SSC), and suggested a suitable method for revealing the variation of erodibility in intertidal sediments. Erosion experiments using a microcosm system were carried out in the Muuido tidal flat, west coast of South Korea. The mean values of τc0 for three methods were: 0.20 ± 0.08 Pa (EM); 0.18 ± 0.07 Pa (ER); and (3) 0.17 ± 0.09 Pa (SSC). The SSC method yielded the lowest τc0, due to the outflow of suspended sediment from the erosion chamber of the microcosm. This was because SSC gradually decreased with time after depleting the erodible sediment at a given bed shear stress (τb). Therefore, the regression between SSC and applied τb might skew an x-intercept, resulting in the underestimation (or “not-determined”) of τc0. The EM method yielded robust and accurate (within the range of τb step at which erosion begins) results. The EM method represents how the erodible depth thickens as τb increases and therefore seems better suited than the SSC and ER methods for representing depth-limited erosion of cohesive sediments. Furthermore, this study identified the spatiotemporal variations of τc0 by EM method in an intertidal flat. The τc0 in mud flat was about two times higher than that in mixed flat. Compared to the end of tidal emersion, the sediment was 10–40% more erodible at the beginning stage.