Precise location of seismic hypocenters continues to be a significant challenge in seismology. Because the focal depths are usually much smaller than the epicentral distance, we improved the simplex optimum method for seismic location from the commonly-used tetrahedral optimum to triangular optimum. Tests using synthetic data of various noise-levels and real data suggest that the triangular optimum approach is more effective than the tetrahedral optimum approach, as reflected by the higher accuracy and the capability of automatically identifying seismic records with unreasonably large errors in arrival time picks. For instance, for a seismic network with 6 stations evenly distributed on a circle and one at the center of the circle with a radius of 25 km, tests using noise-free synthetic data show that the maximum error of the epicenter location for events inside the network is as small as 0.046 km, and the located depths and origin times are almost perfectly accurate; and for events that are up to 150 km from the center of the network, the maximum errors for depth, origin time and epicenter location are 7 km, 0.6 s and 0.89 km, respectively. When arrival times have uniformly distributed random errors of (-0.1 s, 0.1 s), for events inside the network, the maximum errors for depth, origin time and epicenter location are 2 km, 0.3 s and 1.345 km, respectively; and for the location of earthquakes up to 150 km from the center of the network, the maximum errors for depth, origin time and epicenter location are 8 km, 2.3 s and 10.106 km, respectively. Results from other network sizes and noise- levels and the comparisons with commonly-used event location methods will be presented.