15.11.2014 Views

principles and applications of microearthquake networks

principles and applications of microearthquake networks

principles and applications of microearthquake networks

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

5. Generalized Inversion <strong>and</strong><br />

Nonlinear Optimization<br />

Many scientific problems involve estimating parameters from nonlinear<br />

models or solving nonlinear differential equations that describe the physical<br />

processes. In the previous chapter, the problem <strong>of</strong> finding the<br />

minimum-time ray path between the earthquake source <strong>and</strong> a receiving<br />

station involves solving a set <strong>of</strong> first-order differential equations. These<br />

equations are approximated by a set <strong>of</strong> nonlinear algebraic equations,<br />

which are then solved by an iterative procedure involving the solution <strong>of</strong> a<br />

set <strong>of</strong> linear equations. In the next chapter, calculation <strong>of</strong> earthquake<br />

hypocenters will be treated as a nonlinear optimization problem which<br />

also involves solving a set <strong>of</strong> linear equations. Therefore, solving linear<br />

equations is frequently required in <strong>microearthquake</strong> research as it is for<br />

other sciences. In fact, 75% <strong>of</strong> scientific problems deal with solving a set<br />

<strong>of</strong> rn simultaneous linear equations for n unknowns, as estimated by<br />

Dahlquist <strong>and</strong> Bjorck (1974, p. 137).<br />

In solving linear equations, it is convenient to use matrix notation as<br />

commonly developed in linear algebra or matrix calculus. Readers are<br />

assumed to be familiar with vectors <strong>and</strong> matrices <strong>and</strong> their properties.<br />

Among numerous texts on these topics, we recommend the following:<br />

Lanczos (1956, 1961), Noble (1969), G. W. Stewart (1973), <strong>and</strong> Strang<br />

(1976). In general, components <strong>of</strong> a matrix can be either real or complex<br />

numbers. However, in most <strong>microearthquake</strong> <strong>applications</strong>, we deal with<br />

real numbers. Hence, we will be concerned here only with real matrices.<br />

In matrix notation, the problem <strong>of</strong> solving a set <strong>of</strong> linear equations may<br />

be written as<br />

(5.1) AX = b<br />

where A is a real matrix <strong>of</strong> rn rows by n columns, x is an n-dimensional<br />

vector, <strong>and</strong> b is an m-dimensional vector. Our problem is to solve for the<br />

105

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!