STRUCTURAL ANALYSIS BASED ON SEMIVARIOGRAMS

The chemical data were firstly analysed using standard statistical methods at different cut-off grades. These methods were applied m order to define the statistical characteristics, such as distribution laws, means and variances, and correlation coefficients. The investigated variables were G (grade), T (thickness) and GT (grade x thickness or accumulation value). Two-parameter, log-normal distribution models have been obtained for these variables.

The distribution characteristics of the mineralization, based on the variables GT and T, have been investigated by means of sermvariograms which were calculated with relatively long or short lags, depending on whether they utilised data from the first or second-phase drilling as described above.

Firstly, large-scale semivanograms were calculated over distances of a few kilometres using lags of approximately 300 m and were based on data obtained from the initial drilling phase at 300 x 400 m spacings.

Secondly, small-scale semivanograms were calculated over a few hundred metres with lags of 50 m and were based on the data of the second drilling phase. Several cut-off grades were applied; however, only the results related to two of them are presented here, i.e. a cut-off of zero (or no cut-off) and a cut-off of 200 ppm U3O8. Semivanograms were calculated for four geographic directions, i.e N-S, E-W, NE-SW and NW-SE.

4.1 Large-Scale Semivanograms

The large-scale experimental semivanograms are shown in Figures 2 and 3, from which the following information was deduced:

(a) The GT and T semivanograms (with no cut-off) show very similar characteristics and an anisotropy which conforms to the main geologic directions, NE-SW and NW-SE, which are parallel and perpendicular respectively, to the drainage axis

(b) A large nugget effect (Co), (Figure 3) is suggested by the behaviour of the experimental semivanogram, and could be caused by tfie occurrence of smaller-scale structures undetected by the 300 x 400 m drill grid and/or inadequate sampling and analytical procedures (i e. contamination during drilling or sampling and/or unsuitable sampling intervals). From the results of the drilling at 20 m centres and from exposures in trenches it was noted that the major portion of the nugget effect was due to small-scale structures of high variability.

(c) Spherical-type models can easily be fitted to the individual experimental sermvariograms

(d) A hole effect is present, particularly m the NW-SE direction, i e. perpendicular to the drainage axis This effect could possibly indicate a periodicity m the distribution of sample values [8, 9]

c.

03

0 2 0 1

-0 2 NE-SW

larga «cal« »ami-v»nogram»

——— ——— NE - SW

— — — — - NW - SE

sami-vanograin modal

Figure 3

Large-scale relative expérimental vanograms of the variable GT (grade x thickness) and fitted model; cut-off 2OO ppm U3Og.

os

0 4

0 3

0 2 0 1

ill scale semi vanograms - ——— NE-SW - — — — NW-SE

. large-scale

O N ^ 30

* N< 30 N Number of pairs

venogram models (3> first points of

/, the large-scale sami vanogram

Figure 4

Small-scale relative expérimental variograms of thé variable GT (grade x thickness) compared with the starting part of the fitted models of the large-scale variograms; no cut-off applied.

— 0

small-acala

sami-var iogram —————•—

• N> 30

• MOO N: Numbar of pain __ __ smatl-acala spharical

variogram modal

^^^•M. larga-seala variogram modal Q first points of tha

larga-acala variograma

Figure 5

Small-scale relative (overall} experimental variogram of the variable GT(grade x thickness} compared with the fitted model of the large-scale variograms from Figure 3; cut-off 200 ppm U³Og.

(e) A zonal anisotropy is evident from the behaviour of the semivariograms along the NW-SE and NE-SW directions up to a distance of 1 500 m (Figure 2). This anisotropy is typical of deposits displaying sequences of alternately richer and poorer mineralized zones [10]. The anisotropy weakens in the semivariograms by applying a cut-off of 200 ppm U308, possibly because a considerable amount of data is eliminated.

Therefore, a simpler model with isotropic structures (Figure 3) was fitted for ore resource assessment purposes.

4.2 Small-Scale Semivariograms

The small-scale experimental semivariograms are shown in Figures 4 and 5 and have the following characteristics:

(a) Nugget effect The study of the experimental semivariograms and their models is inconclusive in this regard for lack of sampling values at very close distance from each other. Figure 4 suggests the presence of a nugget effect with 0.6 sill, while the values of the first three points in Figure 5 could be equally fitted by the spherical model alone or by a combination of spherical and nugget effect models. However, detail mapping and sampling of trenches proved the presence of high-grade discontinuities in the mineralisation at small scale.

Thus the nugget effect was used in modelling the experimental semivariograms. In this deposit the nugget effect results primarily from the occurrence of small-scale structures but incomplete recovery and errors in the assaying procedures cannot be excluded.

(b) An additional small structure with a range of less than 100 m for GT and slightly over 100 m for T is present.

It shows a zonal anisotropy orientated as in the large-scale structure (Figure 4).

(c) The behaviour of the semivariograms up to a distance of approximately 400 m indicates the presence of a hole-effect structure. The semivariograms show a tendency to increase again at distances over 400 m; a periodically increasing behaviour may exist over larger distances.

(d) An unexpected result is the high amplitude of the small-scale structure, particularly at cut-off 200 ppm U3O8

(Figure 5), which exceeds the sill value of the corresponding large-scale structure (Figures 3 and 5).

5. APPLICATIONS

Several further geostatistical operations can be carried out, based on the semivariogram models obtained from the structural analysis. The first group of applications comprises simply the calculation of the dispersion and of the extension or estimation variances. The second group of applications comprises the estimation of the local (block) and global in-situ reserves by using kriging techniques: the estimation is then optimal in that the

estimation variances are minimized. Calculations of estimation variances, including kriging variance, are very useful for optimizing drilling patterns and spacings in exploration and development drilling.

These techniques have been applied to the Napperby surf icial uranium deposit with the two objectives of defining optimum drill spacings and of obtaining preliminary global reserves for an economic and technical evaluation.

Results are given in Akin [11]. However, only the large-scale semivariograms have been used so far and only linear techniques were applied, although the log-normal distribution characteristics of the variables might require the use of log-normal kriging techniques, e.g. [6]. Furthermore, the small-scale structures, so far detected only in one area of the deposit by detailed drilling, might or might not have validity for the whole deposit and thus require the computation of local variograms (of course based on additional detailed drilling) for more precise results on smaller scales, particularly at higher cut-offs.

Further applications, in the more evolved development stage, again based on the obtained semivariogram models, include advanced geostatistical techniques, such as disjunctive kriging and simulation techniques, for the estimation of recoverable reserves and for mine planning purposes.

6. CONCLUSION

The investigation of the structural characteristics confirms and quantifies the presence of an erratically distributed component and relatively high variability of the mineralization in the Napperby uranium deposit which is in excellent agreement with observations from other deposits of the same type. This type of distribution is characterized by the large nugget effect and by the presence of a relatively small-scale structure with a range of approximately 100 m or less. However, a large-scale structure with a range of over 1 000 m in all directions is also present, which reflects a large-scale continuity of the mineralization. An anisotropy is clearly detected by the semivariograms on both scales. The main anisotropy direction coincides with the drainage axis and is, therefore, clearly controlled geologically.

ACKNOWLEDGEMENT

The authors wish to thank the management of Uranerzbergbau-GmbH for permission to publish this paper and the colleagues of Uranerz Australia (Pty) Ltd., in particular J. Borshoff, who carried out the field and evaluation work and provided the data which formed the base for the present study. G. Möller prepared the drawings.

REFERENCES

[I] HAYCRAFT, J.A., Sampling of the Yeelirrie uranium deposit. Western Australia, In. Aust. IMM Sampling Symp., Melbourne (1976) 51-62.

[2] DICKSON, B.L., Uranium series disequilibrium in the carnotite deposits of Western Australia, this Volume.

[3] FRENCH, R.R., ALLEN, J.H., Lake Way uranium deposit, Wiluna, Western Australia, this Volume.

[4] HAMBLETON-JONES, B.B., HEARD, R., TOENS, P.D., Exploration for surficial uranium deposits, this Volume.

[5] HAMBLETON-JONES, B.B., Surficial uranium deposits in Namibia, this Volume.

[6] JOURNEL, A.G., HUIJBREGTS, C.J., Mining geostatistics. Academic Press, London (1978) 600.

[7] TOENS, P.D., HAMBLETON-JONES, B.B., Definition and classification of surficial uranium deposits, this Volume.

[8] SERRA, J., Échantillonnage et estimation locale des phénomènes de transition miniers. Doctoral Thesis, IRSID and C4, Fontainebleau (1967).

[9] CAMISANI-CALZOLARI, F.A.G.M., Geostatistical appraisal of a tabular uranium deposit in South Africa, In.

2nd NATO Symp. ASI Lake Tahoe, California (1983).

[10] GUARASCIO, M., Optimization of sampling patterns and of block estimation methods, Proc. Adv. Gp. Mtg., IAEA, Vienna (1979) 145-178.

[II] AKIN, H., Beispiel einer geostatistischen Vorratsermittlung für eine sedimentäre Uranlagerstätte (Napperby/Zentralaustralien), in Klassifikation von Lagerstättenvorräten mit Hilfe der Geostatistik, Schriftenreihe der GDMB, Hft. 39, Vers. Chem. (1983) 101-109.

BENEFICIATION OF SURFICIAL URANIUM DEPOSITS