Locating Underground Nuclear Explosions Using Combined Seismic and Satellite Imagery Analysis

With the aid of commercially available LANDSAT satellite images taken over the Pokaran test area in the Rajasthan Desert, India and a recently developed and particularly sensitive statistical change detection method, we have obtained all accurate location for the May 11, 1998, Indian nuclear test site without using seismic data. Its separation from the seismic determination the epicenter of the nuclear tests on May 11, 1998, as provided by the seismological monitoring stations of the prototype International Data Center (pIDC), is 4.1 km. The indication from this pilot study of the 1998 Indian nuclear tests, namely that LANDSAT imagery with a spatial ground resolution of 30 m can be used to locate signals of the typical dimension of a spall zone, was subjected to a validity check using data from nuclear test sites in Eastern Kazakhstan, China and Nevada (USA). The conclusion from this check is that statistical change detection using commercial LANDSAT imagery is well-suited to detect signals associated with spall zones or with test preparation activity. An often-discussed proposal to reduce routine inspection effort while maintaining the timeliness of an inspection regime is to replace scheduled inspections with a smaller number of randomly chosen, unannounced inspections. The unpredictability aspect of such measures is appealing, as they would seem to place the potential violator in a permanent state of uncertainty and thus serve to deter illegal activity. On-site interim inspections at nuclear reactors with unirradiated plutonium inventories (e.g. MOX-fueled reactors) are obvious candidates. In the model presented here, the inspector's equilibrium strategy is indeed mixed: he must choose his interim inspection time randomly. This is, however, a consequence of the modelling assumption that the inspections are unobservable, or, as mentioned in the introduction, that the inspector cannot make use of his information. In the context of routine verification under the Nuclear Weapons Non-Proliferation Treaty, Sanborn contrasts the intuitive attractiveness of unannounced, random inspections both with the substantial practical difficulties of implementing them and with the disadvantages to the inspected party in trying to accommodate them. Significantly, for on-site inspections, randomization is not advantageous. Avenhaus and Canty derive a general solution for the equivalent sequential inspection game which arises when the inspections arc observable. The equilibrium inspection strategies in that game are not mixed: rather they are deterministic. This means that the inspection schedule can be common knowledge without prejudice to the inspector