A study on estimation of a ground motion prediction equation for three versions of Cumulative Absolute Velocity (CAV) is conducted using 320 three-component strong motion records from 49 earthquakes with moment magnitude between 5.0 and 6.1. Earthquakes are occurred in Zagros and hypocentral distance ranges from 10 to 193 km. Three well-known definitions of CAV are calculated for each record. The versions of CAV are: 1- CAV obtained by integration of absolute acceleration over whole duration (CAVTotal), 2- CAV determined from sum of integrations of absolute acceleration over one second intervals with PGA greater than 0.025g (CAVSTD), and 3- CAV calculated by integration of absolute acceleration over portions with PGA greater than 0.005 (CAV5). Calculation of CAVTotal and CAV5 from all of records passing a primarily quality control leads to nonzero values. However, nonzero CAVSTD values are obtained from only 156 records with maximum hypocentral distance of 119 km. For the regression analysis we used two models; one a simple model with only magnitude and geometrical spreading terms, and in the other model we added the magnitude second power term and a term related to anelastic attenuation. The results show that the CAVTotal regression by the simplest model is the best model fit. Standard deviations of the attenuation coefficients as well as analysis of residuals demonstrate that CAVTotal is the most predictable version of CAV. On the other hand predictability of CAV5 is less than others.

Keywords