Prompt estimation of phytoplankton biomass is critical in determining the ecological quality of freshwaters. Remote Sensing (RS) may provide new opportunities to integrate with situ traditional monitoring techniques. Nonetheless, wide regional and temporal variability in freshwater optical constituents makes it difficult to design universally applicable RS protocols. Here, we assessed the potential of two neural networks-based models, namely the Case 2 Regional CoastColour (C2RCC) processor and the Mixture Density Network (MDN), applied to MSI Sentinel-2 data for monitoring Chlorophyll (Chl) content in three monomictic volcanic lakes while accounting for the effect of their specific water circulation pattern on the remotely-sensed and in situ data relation. Linear mixed models were used to test the relationship between the remote sensing indices calculated through C2RCC (INN ) and MDN (IMDN ), and in situ Chl concentration. Both indices proved to explain a large portion of the variability in the field data and exhibited a positive and significant relationship between Chl concentration and satellite data, but only during the mixing phase. The significant effect of the water circulation period can be explained by the low responsiveness of the RS approaches applied here to the low phytoplankton biomass, typical of the stratification phase. Sentinel-2 data proved their valuable potential for the remote sensing of phytoplankton in small inland water bodies, otherwise challenging with previous sensors. However, caution should be taken, since the applicability of such an approach on certain water bodies may depend on hydrological and ecological parameters (e.g., thermal stratification and seasonal nutrient availability) potentially altering RS chlorophyll detection by neural networks-based models, despite their alleged global validity.