Based on the strong-motion data set from the 1999 Chi-Chi, Taiwan, earthquake and a shaking damage statistics database, we investigated the correlations between strong ground motions and earthquake damage (fatalities and building collapses) through a regression analysis. As a result, the current earthquake intensity scale I-t is placed on a more reliable instrumental basis. This is necessary for the real-time seismic monitoring operation in Taiwan where programs for earthquake rapid reporting (RRS) and earthquake: early warning (EWS) are actively pursued. It is found that the earthquake damage statistics give a much closer correlation with the peak ground velocity (PGV) than with the peak ground acceleration (PGA). The empirical relationship between PGV and the intensity I-t determined in this study can be expressed as I-t = 2..14 x log(10)(PGV) + 1.89. This PGV-based intensity is particularly useful in real-time applications for damage prediction and assessment, as the damage impact of high PGV is much more important for mid-rise and high-rise buildings that are characteristic of a modern society. For smaller earthquakes (M <5), the PGV-intensity correlation also out-performs the PGA-intensity correlation, as large sharp PGA spikes are often observed for rather small nondamaging events at close-in distances. However, as the lower level intensity is conventionally defined through human feelings, for even smaller events (M <3) humans are more sensitive to PGA than to PGV. Since the RRS and EWS operations are mainly dealing with large and damaging earthquakes, the above PGV-based empirical relationship should prove to be more appropriate in these real-time operations.