For fluid flow one has a well-accepted mathematical model: the Navier-
Stokes equations. Why, then, is the problem of turbulence so intractab
le? One major difficulty is that the equations appear insoluble in any
reasonable sense. (A direct numerical simulation certainly yields a '
'solution'', but it provides little understanding of the process per s
e.) However, three developments are beginning to bear fruit: (1) The d
iscovery, by experimental fluid mechanicians, of coherent structures i
n certain fully developed turbulent flows; (2) the suggestion, by Ruel
le, Takens and others, that strange attractors and other ideas from dy
namical systems theory might play a role in the analysis of the govern
ing equations, and (3) the introduction of the statistical technique o
f Karhunen-Loeve or proper orthogonal decomposition, by Lumley in the
case of turbulence. Drawing on work on modeling the dynamics of cohere
nt structures in turbulent flows done over the past ten years, and con
centrating on the near-wall region of the fully developed boundary lay
er, we describe how these three threads can be drawn together to weave
low-dimensional models which yield new qualitative understanding. We
focus on low wave number phenomena of turbulence generation, appealing
to simple, conventional modeling of inertial range transport and ener
gy dissipation.