Abstract
There are many kinds of quantitative models such as landuse models which analyze values by spatial unit. But there is no effective measures which show the goodness-of-fit of an estimated distribution to an observed distribution of the value. In this study, the authors propose measures to represent spatial fit between spatial distributions which they call Spatial Fit Indicators (SFIs). SFIs are defined as transportation costs of errors in order to reproduce the observed distribution from the estimated one. They can be regarded as indicators to show the degree of spatial discrepancy between the two distributions. In the present study, four kinds of SFIs are proposed to be corresponded with the purposes for evaluating distributions. Being corresponding with each SFI, a measure named Spatial Fit Grade GSM is derived from the population distribution of the SFI to absolutely evaluate each estimated distribution. The values of SFGs vary between -1 and 1, and the way of interpretation with them is very similar to that with the correlation coefficient.
