This study compares predicted seismic thresholds to actual thresholds for a high-resolution, three-component digital network in the southern part of the Great Basin of Nevada, centered on Yucca Mountain, the designated site of a high-level nuclear-waste geologic repository. In order to do this comparison, it was necessary to quantify carefully the statistical properties of noise, including mean, variance, and correlation, and the statistical properties of signals, including mean, variance, and site effects. It was also necessary to relate the noise levels to the long-term average (lta) and the signal levels to the short-term average (sta) used in conventional signal-detector algorithms. Combining these attributes with a random-earthquake simulator enables one to predict the seismic threshold of the network for a given source area.
Initial predicted thresholds for the network were found to be approximately 0.3 to 0.4 magnitude unit less than the actual thresholds achieved by the network. There are several reasons for this, but the major reason seems to be the inability to time small signals that are automatically detected. A predicted threshold map for the network was constructed by accounting for the difference initially seen between predicted and actual thresholds. This map indicates that the network detects all events within 10 km of Yucca Mountain with three or more stations down to a threshold of roughly ML −0.5, whereas events near the fringes of the network are similarly detected down to a threshold of roughly ML +0.5. The simulations show that the network-estimated magnitude is biased upward as true magnitude approaches the threshold; this bias may be as large as roughly +0.3 for events near the threshold.