In several studies it is necessary to determine the magnitude of extreme flows in a river. Having an adequate register of observeddischarge it is possible to adjust a probability density function (fdp) that allows estimating events associated with a high returnperiod (i.e. 100 years). In ungauged basins, such as the majority of basins in the world are, other methodologies are used, suchas the Synthetic Unity Hydrograph proposed by the United State Soil Conservation Service (UH-SCS). The UH-SCS model wasevaluated in watersheds of the Norte de Santander department (COL), in its capacity to estimate extreme flows, and to exploreits possible regionalization. The evaluation is done by comparing estimates of Q100, using the Frequency Factors method andthe UH-SCS model. Discharge and precipitation time series were obtained from the IDEAM network, selecting 19 basins basedon their drainage area, climatological stations density and records length. Geomorphology was characterized using ArcMap™on the ASTER-GDEM digital elevation model. Using information available on geology, soils, vegetation cover, and assuminga wet antecedent moisture condition (AMC-III), values of the median of relative Q100 error (e-Q100) of + 507% and + 406%were obtained for the fdp Gumbel and Log-Pearson. Using dry antecedent moisture condition (AMC-I) e-Q100 low to + 36%and + 17%. It was possible to minimize e-Q100 by calibrating the Curve Number (CN) parameter. A satisfactory regionalizationfunction for CN was not found. Applying SCS-HU under AMC-III condition, Q100 is greatly overestimated. It is possible tominimize the error by considering AMC-I and reduce CN, a counter-intuitive situation since extreme flows are associated withwet weather conditions (i.e. Ano Nina). Improvements in the characterization of rainfall and soils in Norte de Santander shouldbe investigated.