PE EIE[R-Rg RESEARCH ON - HJ Andrews Experimental Forest

LINEAR
Y~(T)
2nd ORDER
Y2 (T)
3rd ORDER
Y 3 (T)
INPUT
X (T)
OUTPUT
Y (T )
}
nth ORDER
Yi (T )
Figure 10 . Illustration of nonlinear system representation by Volterra Series.
the memory, an exact match could be made
to each of the input values . However, n o
economy in the description would have bee n
affected . Instead, fewer polynomials can b e
used. This introduces an error, but due to th e
nature of the operation, it will be the leas t
error possible .
Methods using the above procedure hav e
been introduced by Jacobi (1966), Harder
and Zand (1969), and Brandstetter an d
Amorocho (1970) . It is this method which is
being considered for use in this study . There
are certain distinct advantages associated with
the process . The method is quite general and
can handle a variety of problems such as runoff,
chemical quality of runoff waters, and
suspended sediment predictions, when inpu t
values are properly weighted by functions
describing the physical processes involved in
each case. Once the response functions or
kernels of the system are evaluated, they may
be used for predictions given any sequence o f
inputs. It requires a minimum amount of
data, possibly 1 to 2 years of good records .
Systems can be used in cascade, lik e
predict predict
precip.-runoff quality or sediment
It can be used for quantitative evaluations o f
the changes created by any type of watershe d
management procedure by evaluating th e
kernels of the original and the manage d
hydrologic system. Emphasis will be given to
establishing weighted functions for the bes t
description of the physical process, and o n
deriving tools for the greatest economizatio n
of the procedure .
67