Computer simulation of rockfalls has been widely used in rockfall analysis in recent years, and the coefficient of restitution is an important parameter input that is difficult to determine. Aimed at finding an easy solution to this problem, three stages of laboratory and field tests have been carried out. Rockfall trajectory analysis at a specific site has been done as an application to verify the method developed. In the first stage of laboratory testing, quasi-spherical rock "'balls" made from different rock samples were dropped from 1 m onto polished rock slabs that were clamped on a concrete deck, which can be set to different slope angles. A high-speed video camera was used to record the impact processes, and normal and tangential coefficients of restitution were calculated from the video records. The results show a linear relationship between the normal restitution coefficient and the Schmidt hammer numbers of both the rock slabs and the falling rock "balls", and the slope angle. An empirical equation was then established to calculate the normal coefficient of restitution from those three parameters. However, the correlations between the tangential coefficient of restitution and the above parameters are poor, indicating that the tangential coefficient of restitution is not adequately determined by such rock properties. The second stage of laboratory testing was under more practical conditions. Three different rough rock blocks were used as rock slopes. Angular rock boulders were dropped from different heights onto the rock blocks, and a rock "ball" was also used to make a comparison. The results show that the normal coefficients of restitution from impacts of angular rocks are much smaller than those of rock "'balls", and have a linear correlation with those calculated from the empirical equation obtained by the earlier test. Tests on beds of gravel, soil, rock fragments and sand have also been carried out to obtain the coefficients of restitution of those materials. Finally, field tests have been carried out at a quarry site in Lyttelton Basalt rock boulders of about 0.3 m in diameter were dropped from about 4 m onto rock and debris slopes using an excavator. The values of restitution coefficients obtained are similar to those from laboratory tests but larger than those calculated from the empirical equations due to the effect of weathering and surface roughness of rocks in the field on the Schmidt hammer measurement. Forty basalt boulders were then rolled down a bench slope of about 16 m, three cameras were used to record the rockfall processes. Two different rockfall simulation programs (CRSP and RocFall) were used to simulate the rockfall processes. The simulated bounce heights and velocities from CRSP are close the field trial, while those from RocFall are smaller than the field results. Comprehensive rockfall analysis has been carried out for the Marine apartments, Sumner, where a steep cliff of 35-45 m represents potential rockfall hazards to a car park and proposed buildings at the base. Site investigation, rockfall simulation and risk assessment have been carried out for the site. The results show that without any protection measures, a majority of rocks from the cliff face reach the edge of the car park. The probability of an accident at the car park is moderate (1 in 195 years), while the probability of fatality at the car park is low (8.69* 10-6) and acceptable under the proposed risk criteria for "Major Civil Engineering Projects".