The surface area of mangrove ecosystems in Senegal have fluctuated substantially over several decades. Satellite data at 10 to 30-m resolution, which has been available since the 1980s, has allowed the mapping and quantification of these dynamics. However, the plant formations have reorganized internally; this has not been well-documented, possibly because there is no established method for detecting zonation of the Senegalese mangrove. This paper proposes a two-step method for mapping the zonation of the Saloum Delta mangrove. First, mangrove surfaces were detected using machine learning methods from an object-based time series. Finally, a typology was developed through object-based clustering using time-series metrics derived from the harmonic regression modeling of the vegetation fraction. A comparison with field data allowed us to determine the number of classes and discriminating variables. The results showed that the selected method resulted in an overall accuracy of 97.55% (Kappa = 95.42) for the land cover at 4 classes (Water, Mangrove, Salt flats, Other land covers), with an F1-Score of 98.91% for the mangroves. Second, our results suggest that the annual trend of the vegetation fraction at the object scale is effective in differentiating mangrove zonation into three classes based on canopy density and stand height (HM: High mangrove; LDM: Low and dense mangrove; LOM: Low and open mangrove). Finally, the temporal stability of the classes and uncertainty around the magnitudes of the plant fraction values per class were assessed by Bayesian inference. An overall accuracy of 85.5% can be expected to identify the zonation typology on an inter-annual scale. This mapping technique can be used to characterize the rate of change in zonation in response to environmental changes and to guide management strategies.