Models abound for predicting earthquakes from past seismicity and detailed tectonic maps, analogous to those suggested for prediction of volcanic hazards. The Izmit earthquake of 17 August 1999 in Turkey was among the most savage in recent years and killed thousands. It was as powerful (magnitude 7.8) as the celebrated 1906 San Francisco earthquake, and like it stemmed from movement on a continental-scale strike-slip fault. The North Anatolian Fault is almost as well studied as the San Andreas line, and seismicity was known to be heading westwards well before the Izmit catastrophe. Indeed, the Izmit area was predicted to be next on the list, yet no preparation had been made, even by Turkish tectonicians who had been involved in seismic analysis. Chinese geoscientists take a different approach to seismic prediction than those in the west – over the last few centuries, hundreds of thousand Chinese people have perished in earthquakes. They are trying to organise local people to monitor possible precursors to earthquakes, such as rises in water levels in wells and strange behaviour of animals. They have had some notable successes, including preparation for one earthquake in recent years that saved an estimated 80 thousand people in one particularly hazard prone city. The Geological Survey of Israel has been testing a well known correlation between the times of anomalous radon emissions from the ground and earthquakes along the Aqaba Fault that controls the Dead Sea. Over a 7-year period, hourly scintillation-counter readings of radon emissions from springs, wells and especially gravels near known active faults allowed a rigorous test of a possible prediction system, because in that time there were almost 800 minor earthquakes (Steinitz, G. et al. 2003. Statistically significant relation between radon flux and weak earthquakes in the Dead Sea rift valley. Geology, v. 31, p, 505-508). For events beneath the Dead Sea rift, there is a good correlation between the start of radon emission increases and earthquakes, which suggests that about 3-days warning could be given, if the monitoring was widely deployed. The same cannot be said for small tremors with a source outside of the active fault zones. The success may possibly be because sufficient radon to be easily detected is generated by radioactive decay of uranium in a phosphorite bed that underlies the study area. Radon escape to the surface is possibly eased when microfractures begin to open as strains build before an earthquake.