The determination of flood peak discharges in small torrential catchments relies in many cases only on simple estimation formulas as no flood record data is available and simulation models are too expensive in their application. Many of these estimation formulas have been proposed by different authors, but a systematic testing in small catchments was not carried out up to now in Switzerland. In the present paper the results of the evaluation of 12 estimation formulas are described. The tests have been carried out with runoff data from seven small torrential catchments with areas between 0.5 and 3.5 km² where time series between 25 and 95 years have been available. More than 50 % of the tested formulas had to be classified as unsuitable for estimating flood peak discharges in small torrential catchments. As no single formula is reliable under all conditions an approach combining the five most reliable estimation formulas is described and proposed for future flood peak estimations. One of these estimation methods used in this multi-method approach is the new approach Clark-WSL. It allows the determination of the design peak flow and the corresponding hydrograph for a given return period. The necessary catchment parameters are objectively determined during a field inspection. The Clark-WSL method is based on the conceptual rainfall-runoff model of Clark. This model combines the concepts of linear storage and linear translation. The storage is assumed to be at the outlet of the catchment. Linear translation is considered by an area-time diagram. The return period of the design flood event is matched to the one of the initiating design storm rainfall, which has to be derived from the corresponding maps. The volume runoff coefficient ψ is assumed to depend on the event rainfall and the water storage capacity, which is to be determined in the field according to a decision scheme.