The discovery of great oil reserves located under the northern part of the Caspian Sea has lead to an increased interest in the wave climate in this area. Statistical analysis of wave measurements is insufficient to determine reliable design wave conditions on the northern part of the Caspian Sea, because the amount of available wave measurements is insufficient and the available wave measurements do not contain events with design water levels. Wind and water level statistics, however, are available. Wave statistics can now be obtained by means of a wave generation model. At Delft University of Technology a wave generation model has been developed, named SWAN (Simulating WAves Nearshore). The model calculates the evolution of wind wave spectra. Test cases are needed in order to obtain an insight in the performance of this model. The northern part of the Caspian Sea could be an interesting test case. Water depths are small (3 to 4 meters at the Kashagan oil fields) and therefore many physical processes influence the wave spectra in this area. The commercial interest in the area has lead to the initiation of measurement campaigns, leading to the availability of wind and spectral wave measurements. This project therefore serves two main objectives. Firstly, Witteveen+Bos gains insight in the hydraulic boundary conditions on the northern part of the Caspian Sea. Secondly, SWAN in its current state of development is tested thoroughly on an interesting case. This can ultimately contribute to improvements of SWAN itself. Note that many hydraulic engineers follow the developments around SWAN with great interest, as it can be a valuable tool during various projects. This study is a succession of the MSc project by Van Thiel de Vries (2003). He calibrated SWAN on integral wave parameters, such as Hmo and Tp. The underlying burst pressure measurements of these wave parameters have now become available. In this project the model is re-calibrated with the use of spectral wave measurements. Furthermore its performance is studied quantitatively. The following approach was used: an inventory was made of the requirements and the availability of measurements; the burst pressure measurements were converted to wave spectra; the wave spectra were analysed qualitatively; typical events were selected for calibration; the SWAN model was setup and re-calibrated; the SWAN model's ability to determine the design wave spectrum was tested and quantified. An analysis of various methods for the conversion of burst pressure measurements to wave spectra showed some interesting results. Much controversy exists around the cut-off frequency of the so-called gain correction factor. This is the multiplying factor, applied to transform pressure variation spectra to sea level elevation spectra. This factor has its origin in linear wave theory. The factor must be cut-off at a certain value in order to prevent the introduction of instrument noise in the wave spectrum. The method, chosen at the end of the study, resulted in an increase in significant wave heights and a decrease in integral wave periods compared to those used by Van Thiel de Vries (2003) (although he used the same raw measurements). Contrary to his conclusions that a default configuration of SWAN overestimates wave heights and underestimates wave periods, SWAN had actually performed rather well. The mismatch proved to be due to an inaccuracy in the conversion of the raw data. The SWAN model created by Van Thiel de Vries (2003) was used as a basis. For this research the bathymetric map of this model was raised by 70 cm, after a comparison of the map with recent surveys and pressure measurements. The discrepancy was probably caused by an inconsistency in the definition of the reference levels in the old Russian bathymetric maps, on which it was based. The change in the map led to an improved auto-calibration of the Delft3D model, used for the calculation of winddriven surges.