Water erosion affects, in Tunisia, nearly 3 million hectares of agricultural land, and constitutes a threat to the sustainability of small lakes in the hilly regions. Twenty six lakes have been identified in the central part of Tunisia covering the Ridge area and Cap Bon. To predict the siltation in these lakes, we proposed a simple and practical model classification assessing the sediment fluxes at the outlet of the small watersheds and the sediment load of the receiving lakes. We further explored the links between the sediment prediction model parameters and various terrain attributes of the contributing catchments. Three statistical methods are used for the data analysis: the Principal Component Analysis method (PCA), the hierarchical classification method, and finally the step wise linear regression analysis. We identified three different classes of lakes. The first class is less vulnerable to the silting risk and is located on the southern and eastern borders, west of the ridge and the coastal plains. The lakes of this group are characterized by a low rate of silting, a large drainage area, a low relief, a relatively hierarchical hydrographic network and an effect of precipitation and little intense runoff. A second class which is the most degraded environments covers almost the entire semi-arid zone of Central Tunisia. This class has a very abrasive potential watershed, explained by high flow coefficients related, mainly to higher erosive rainfall intensities associated with a moderate or accentuate topography, structure of soil over marl and poor drainage and unimproved surface. The third class exists in the north and south of the Ridge. The catchments of this class are characterized by a moderate to high sedimentation rate which is governed by a more or less marl soil structure and an intense hydrodynamic compounded by the steep slopes of these basins. The represented lake will server to quantify and predict the sedimentation.