Abstract: Salinity is a key variable used to explain the distribution of species within estuaries and the functioning of estuarine ecosystems. Despite the relevance of average conditions and extreme events, the existing schemes of zonation are only based on average values; the extreme values obtained through high-resolution long-term data have not been still integrated in an appropriate methodology capable of recognizing salinity types. Therefore, the background variability of salinity has been frequently underestimated in ecological studies. The primary goal of this research is the identification of ecologically significant salinity zones capable of encompassing the entire estuarine regime. A two-step methodological approach was developed: (1) the reconstruction of long-term salinity series using the Deflt3D hydrodynamic model, the analog method and a subset of short-term representative states of river flow and tidal level and (2) the identification of different zones using eight descriptors of the salinity regime (i.e., the median, the range and the intensity, duration and frequency of extreme events) and the application of a combination of two clustering techniques of unsupervised learning (Self-Organizing Maps (SOM) and K-Means). Thus, five ecologically significant salinity types that are representative of estuarine variability were identified based on a salinity series reconstructed using a validated method. Differences in the mean values of salinity among typologies allow explaining patterns in the general descriptors of benthic macroinvertebrates assemblages (i.e., richness and diversity). If extreme salinity conditions are also considered, typologies increase their ecological significance and they are able to recognize differences in species composition.