Table 2. Line 2 Variogram of Line 1 and Line 2 FIDLER Date (1 Foot Aboveground) (Continued) Number of Pairs Lag Number 1 2 3 360 34 22 10 161 Distance (Inches) 5 10 15 1800 Variogram (x 10 3 58.37 56.78 55.91 124.1 It does not seem possible though, to use these results from Line 1 to model the short distance behavior of the variogram over all the area covered by the 100-foot grid, since the counting error depends on the level of activity. At very low levels along Line 2, we get pure noise (Table 2), signifying no correlation structure. A question at this point is: analyses consistent?" “Are the l-foot and the 100-foot scale The answer is yes. Seen from a 100-foot grid, the details of the l-foot scale structures merely appear as noise. The value Cy = 0.0127, which may be called the “apparent nugget effect,” is the variance of that noise. Since we are working at a 100-foot scale, we do not need a precise modeling of microstructures and can instead use a simplified macroscopic model with a discontinuity of magnitude Co The T(x) Field We do not have direct access to T(x), the true correction term, and what we analyze are the estimated residuals from the regression: T(x,) = P(x,) - aF(x,) ~ B= TOx,) + fe(x,) -anG@I. AD Figure 7 shows the raw variograms of these residuals computed separately for FIDLER > 5,000 and FIDLER < 5,000. The variograms of the differences {log Am(x) - a,F(x) - B,} on Lines 1 and 2 data are also plotted for comparison (a, and §, regression coefficients were calculated using pooled Lines 4 and 2 data). Since the Pu to Am ratio is approximately constant in the log scale, these differences should be comparable to values of T(x). Each of these variograms stabilizes at a certain level, called the "sill," indicating a stationary behavior of T(x). This was to be expected since the drift effect is already accounted for by the guess field aF(x) + 8. The value at the sill is equal to the overall variance of the 382