Previous published work after the Kobe and A degrees zmit earthquakes (1995 and 1999, respectively) demonstrated some reported meteorological and animal behaviour precursors were valid. Predictions were freshly tested for the Christchurch earthquake (M = 7.1, 4 September 2010). An internet survey with nearly 400 valid replies showed relative numbers of reports in precursor categories the day before the quake, were statistically significantly different from those in the preceding three days (excess meteorological events and animal behaviour). The day before the quake, there was also altered relative precursor class occurrence within 56 km compared with further away. Both these confirmed the earlier published work. Owners were woken up by unique pet behaviour 12 times as often in the hour before the quake compared with other hours immediately before (statistically highly significant). Lost and Found pet reports were double normal the week before, and 4.5 times normal both the day before the quake, and 9 days before. (Results were again statistically significant). Unique animal behaviour before the quake was often repeated before the numerous aftershocks. These pet owners claimed an approximate 80% prediction reliability. However, a preliminary telephone survey suggested that animals showing any precursor response are a minority. Some precursors seem real, but usefulness seemed mostly restricted to 7 cases where owners were in, or near, a place of safety through disruptive pet behaviour, and one in which owners were diverted by a pet from being struck by falling fixtures. For a later 22 February 2011 M = 6.3 quake no reports of escape through warning by pets were recorded, which raises serious questions whether such prediction is practically useful, because lives claimed saved are extremely low compared with fatalities. It is shown the lost-pet statistics dates, correspond to ionospheric anomalies recorded using the GPS satellite system and geomagnetic disturbance data, and claimed as precursory. The latter more objective measurements may be the way of the future, but improved statistical treatment should include observations over longer periods of time without earthquakes.