The use of active acoustic to monitor abiotic structures and processes in the ocean have been gaining ground in oceanography. In some systems, acoustics allow the robust estimation of the depth of the pycnocline or thermocline either directly or indirectly when the physical structures drive the one of organisms. Here, we examined the feasibility of extracting the thermohaline structure (mixed-layer depth, upper and lower thermocline) from echosounder data collected in the oligotrophic Southwestern tropical Atlantic region at two seasons (spring and fall), more precisely in two areas with different thermohaline conditions, at both day and night. For that, we tested three approaches: (i) the vertical extension of the epipelagic community; (ii) the use of acoustic gradients; and (iii) a cross-wavelet approach. Results show that, even if the thermohaline structure impacts the vertical distribution of acoustic scatters, the resultant structuring did not allow for a robust estimation of the thermohaline limits indicating that other oceanographic or biological processes are acting. This result prevents for a fine-scale representation of the upper-layer turbulence from acoustic data. However, studying the proportion of acoustic biomass within each layer provides interesting insights on ecosystem structure in different thermohaline, seasonal and diel scenarios.