Click on each stand for a better view of the stand and the variograms calculated.
| Sowaqua Creek Study Site |
|---|
|
| The Sowaqua Creek data set was used to test geostatistical proceedures for estimating
stand structure. There were two objectives for this stage of the study: 1. to determine if
stands are distinguishable based on their range or sill values, 2. test different data
collection techniques.
Click on each stand for a better view of the stand and the variograms calculated. |
| Sowaqua Creek Study Site: Range and Sill Differences |
|---|
 |
| The value of the range, the distance at which the value of the semivariance levels off, has often been used to describe a wide range of forest parameters. There is a difference in the vales of the range when the two different methods of extracting the data are compared. The values for the range are more variable (6 to 22 pixels) when calculated using the linear data. The range values calculated for the area data have a lower range of values (8-15). This smaller amount of variation is an indication that the spatial variability of the different stands is relatively consitant; all stands reach a sill value around 10 pixels. |
 |
| There is little difference in the values of the sills calculated using the linear or the
area method. The sill represents the value of semivariance where the semivariance levels off.
As the data used were selected from the same area the variation in the data should be similar.
The sill values which measure the amount of variation indicate that there is a posibility to class the semivariance measures into groups. The graph shows that there are three distinct classes of semivariance values: low semivariance for alder stands, high variance for Douglas-Fir stands, and mid-level variances for hemlock, mixed, and younger aged Douglas-Fir stands. |
| Sowaqua Creek Study Site: Goodness of Fit |  |
|---|
| The goodness of fit statistic calculated gives a measure of how well the
cross sections through the 2D model adjust the experimental variogram. The number is standardized
and unitless which allows the comparison of models between areas. In general, the closer the
measure is to zero the better the fit.
The goodness of fit statistic calculated yeilded some interesting results. Previous research has focused on relationship of the range and sill values to other parameters. Quantitative assesments of the quality of the fit have not usually appeared in the literature. The graph shows the differences in the value of the goodness of fit statistic. It is evident that the fit acheived using the area method is superior to that acheived using the linear measure. |
