The foundation for time series analysis methods to detect chaos is the notion that phase spaces and dynamics of a nonlinear dynamical system (NDS) can be reconstructed from a single variable, based on Takens embedding theorem (Takens, 1981). Many years ago (Phillips, 1993) I showed that temporal-domain chaos in the presence of anything other than perfect spatial isotropy (and when does that ever happen in the real world?) leads to spatial-domain chaos. This implies an analogous principle in the spatial domain.
Assume an Earth surface system (ESS) characterized by n variables or components xi, i = 1, 2, . . , n, which vary as functions of each other:
ESS = f(x1, x2, , , , xn)
If spatial variation is directional along a gradient y (of e.g., elevation, moisture, insolation) then
dxi/dy = f(x1, x2, , , , xn)
dx2/dy = f(x1, x2, , , , xn)
. . .
. . .
. . .