From all possible path models we chose the following scheme, here arranged as a lower-triangular matrix. Each of the variables along the left of the matrix are assumed to be influenced by those along the top, and the cell entries suggest what we expect to find given very simple theories of environmental interaction. The matrix is therefore exploratory and is not intended to represent the structure of a more sophisticated process model. For example, we expect that elevation should depress temperature, which itself decreases the amount of vegetation, so that the link between elevation and vegetation would be positive (although the direct link between these two is negative). Any such causal argument will be modified by regional conditions, effects of scale, and temporal influences occurring at hourly to millenial scales.

Note that there are n = 9 variables in the data, and they can be interrelated causally in 9! = 362,880 ways (and almost a million ways if subsets of variables are selected).