Event location capabilities of a single array and of two arrays in terms of random and biased observational errors have been simulated on the computer. The approach was to obtain a Gaussian distribution of the parameters azimuth, slowness (DT/DΔ) and arrival-time differences which are used in one- and two-array epicenter determinations. Specifying their mean values and variances, a large number of paired values of these parameters are simulated by using a random number generation routine. These observations determine a distribution of event locations in geographic space, from which the axes of the 95 per cent confidence ellipse were calculated. In case of one array, the semi-axes of the confidence ellipse are tabulated as a function of epicentral distance, and fictive variances of azimuth and slowness. The joint event location capability for two arrays is easily computed but not tabulated because of the regional changes in the shape and orientation of the confidence ellipses. The above analysis was supplemented with epicenter determinations using real NORSAR and LASA data. In the West Indies area, the average location error was found to be 310, 161 and 139 km for NORSAR, LASA, and NORSAR and LASA combined.
A model for individual subarray time corrections for the NORSAR array is used for simulating biased location errors, which would be very large if left uncorrected. The results obtained roughly indicate that the array siting area may account for about 45 per cent of the observed slowness anomalies, but no improvement was obtained in the azimuth anomalies. The slowness bias makes it difficult to utilize the NORSAR DT/DΔ measurements for investigations of heterogeneities in the mantle.