The lithosphere of the Earth is structured as a hierarchical system of volu
mes of different sizes, from about 10 tectonic plates to about 10(25) grain
s of rock. Their relative movement against the forces of friction and cohes
ion is realized to a large extent through earthquakes. The movement is cont
rolled by a wide variety of independent processes, concentrated in the thin
boundary zones between the volumes. The boundary zone has a similar hierar
chical structure, consisting of volumes, separated by boundary zones, etc.
Altogether, this hierarchy of volumes and multitude of processes compose th
e lithosphere into a large non-linear complex system. Upon coarse graining
the integral mesoscale empirical regularities emerge, indicating a wide ran
ge of similarity, collective behavior, and the possibility of earthquake pr
ediction. This approach led to new paradigms in the dynamics of the lithosp
here and, on the practical side, created a capacity to predict from 70 to 9
0% of large earthquakes, with alarms occupying 10-20% of the time-space con
sidered. Such predictions may be used to undertake earthquake preparedness
measures, which would prevent a considerable part of the damage (although f
ar from the total damage). The methodology linking prediction with prepared
ness was developed; it may help a disaster management authority to choose t
he preparedness measures, allowing for the currently realistic accuracy of
predictions. A large-scale experiment in advance prediction of large earthq
uakes worldwide has been launched to test the prediction algorithms. The te
st is unprecedented in rigor and coverage. The forecasts are communicated,
with due discretion, to several dozen leading scientists and administrators
in many countries. Among already predicted earthquakes are all the last ei
ght great ones with magnitude 8 and more. The major drawback is the rate of
false alarms. The possibility is outlined to develop a new generation of p
rediction methods, with fivefold increase in accuracy and the transition to
short-term prediction. The links with prediction of geotechnical and engin
eering disasters are established: scenarios of transitions to a large earth
quake happen to share some features with a broader class of catastrophes. T
his experience now opens as yet untapped possibilities for reduction of tec
hnological disasters. (C) 2001 Elsevier Science B.V. All rights reserved.