The general problem of an appropriate erosion modelling can be seen in a lack of available data and in the validation and verification of the methodologies applied. The article includes two significant and challenging topics, i.e. the evaluation of the sediment amounts in a catchment and the validation of the methodologies used. The importance of the sediment estimation can be found not only in the modelling and prediction fields but also in the terms of engineering practice. The significance of erosion model validation lies in the confidence in the model itself and in the detection of its applicability and relevance. In this study, results obtained from the physically-based EROSION-3D model was compared with the sediment yield of a small agricultural catchment in central Poland. The estimation of sediment yield from the agricultural catchment has been conducted using the empirical method (USLE) coupled with a sediment delivery ratio (USLE-SDR) and verified by reservoir measurement. Firstly, the application of a physically-based EROSION-3D model has been done based on a continuous rainfall series for the selected period and afterwards the results have been compared with the sediment yield obtained by the empirical methods in order to test a model´s performance. The results of the paper point to the comparison of the results obtained by two different approaches, i.e. the physically-based and empirical methods together with the validation of the methods through the acoustic depth measurements.