Seismic Oceanography is a developing research topic where new acoustic methods allow high-resolution teledetection of the thermohaline structure of the ocean. First implementations to study the Ocean Surface Boundary Layer have recently been achieved but remain very challenging due to the weakness and shallowness of such seismic reflectors. In this article, we develop a multifrequency seismic analysis of hydrographic datasets collected in a seasonally stratified midlatitude shelf by ARGO network floats to assess the detectability issue of shallow thermoclines. This analysis, for which sensitivity to the data reduction scheme used by ARGO floats for the transmission of the profiles is discussed, allows characterizing both the depth location and the frequency dependency of the dominant reflective feature of such complex structures. This approach provides the first statistical distribution of the range of variability of the frequency-dependent seismic reflection amplitude of the midlatitude seasonal thermoclines. We introduce a new parameter to quantify the overall capability of a multichannel seismic setup, including the source strength, the fold and the ambient noise level, to detect shallow thermoclines. Seismic source signals are approximated by Ricker wavelets, providing quantitative guidelines to help in the design of seismic experiments targeting such oceanic reflectors. For shallow midlatitude seasonal thermoclines, we show that the detectability is optimal for seismic peak frequencies between 200 and 400 Hz: this means that airgun and Sparker sources are not well suited and that significant improvements of source devices will be necessary before seismic imaging of OSBL structures can be reliably attempted.