Geoscience Reference
In-Depth Information
Fig. 7.2 Concept of trend illustrated by means of two-dimensional graphs (After Davis 1973 ).
Value (vertical scale) is plotted against distance (horizontal scale). (a) Collection of original data
points and smooth curve on which they lie; (b) Straight-line trend fitted to the observations; (c)
Quadratic trend; (d) Cubic trend. Shadings represent positive and negative residuals from the
trends (Source: Agterberg 1984 , Fig. 2)
Some problems of trend surface analysis can be illustrated by the fitting of curves
to data collected along a line. An example is shown in Fig. 7.2 (after Davis 1973 ). The
observations fall on a smooth curve which, however, is not a low-degree polynomial.
The magnitudes of the residuals decrease when the degree of the polynomial is
increased. This artificial example illustrates two important points:
1. A good fit is not necessarily the object of trend analysis. Instead of this, the aim
usually is to divide the data into a regional trend component and local residuals
which can be linked to separate spatial processes of regional and local signifi-
cance, respectively. Examples of interpretation of trends and residuals will be
given later in this chapter.
2. Each curve-fitting of Fig. 7.2 yields residuals that form a continuous curve.
Adjoining residuals along a line therefore are not statistically independent.
Instead of this, the residuals are spatially correlated.
Search WWH ::




Custom Search