seem that the agreement is very poor since all raw variograms lie above One should remember, however, that the presence of a the fitted model. It is only at drift results in an upward bias in the raw variograms. short distances, perhaps, that the raw variograms may coincide with the the raw varioThis coincidence indeed happens here: underlying ones. gram in the east-west direction becomes tangent to the fitted model at Therefore, the agreement is actually very distances less than 200 feet. good, especially since the model resulted from an automatic identification procedure based on quite a different approach than that used to calculate the raw variograms. There is no theorem that says there must be some direction for which the drift effect is not felt, at least at short distances, but it often occurs in practice. It is of interest to comment on the nugget effect of "0.0127 found here. Typically, it can originate from 1. measurement errors 2. microstructures at a scale much smaller than that at which the observations are made (as occurs with golden nuggets in gold mines, where the ore grade varies discontinuously from inside to outside nuggets). But the interpretation of the nugget effect is more complex for the FIDLER data. Basically, the value C_ = 0.0127 results from extrapola- tions to the origin of a variogram computed at a 100-foot scale. This is clearly shown in Figure 6. Yet, if the phenomenon is analyzed at a much finer scale, it emerges that the discontinuity of the variogram at the origin is only apparent. The "true" variogram probably resembles the dashed curve in Figure 6, decreasing steeply to zero, with a small nugget effect left to account for measurement errors. Such conjecture is based on the variogram computed from Line 1 FIDLER data (Table 2). This table clearly indicates that the F_ (x) field is continuous, as can be expected since two close FIDLER readings integrate radiation from a large common area. A nugget effect still shows, but of the order of 1073 (rather than 1072 as found above), corresponding to a counting error ofcount ~ 7 percent. Table 2. Line 1 Variogram of Line 1 and Line 2 FIDLER Data (1 Foot Aboveground) Lag Number 1 2 3 48 Number Distance Variogram 45 30 15 224 5 10 15 240 1.116 1.455 2.818 20.07 of Pairs 381 (Inches) (x 1073)