Statistical Evaluation of Exploration of Gold Pla-
cers in Adola Area (Southern Ethiopia)
Milan Hamrla
with 11 textfigures and 14 tables
Contents
Introduction.........................127
Geology..........................129
Basic geological information.................129
Placers.........................130
Mining ..........................131
Methods applied......................131
Workable reserves.....................131
Estimate of cost and payable tenor Umit............132
Exploration.........................133
General.........................133
Statistical characteristics of placers..............133
Comparison of exploration data and mining results........138
Analysis of grid density..................139
Sampling ...........................145
Costs and p>erformance...................150
Conclusions.........................152
Statistična ocena raziskav zlatonosnih naplavin na območju Adole (Južna
Etiopija).........................152
References.........................153
Introduction
The state owned Adola gold mine is situated in Sidamo province in
Southern Ethiopia, some 500 km from Addis Ababa (Fig. 1). It is accessible
by a good all-weather road. The gold-bearing placers occupy the tributary
valleys of the Dawa Parma river. The known placers are confined to a
slightly arched elongated zone of about 150 X 20 km, running N—S. The
undulated moderately hiUy country is covered by dense forest in its nor-
thern part, and the southern one is a semiarid shrub and savanna land.
The area is sparsely populated by semi-nomadic cattle-raising Guji tribe.
Alluvial and deluvial gold was apparently discovered in 1936 during
Italian occupation. Since then exploration and exploitation activities were
carried out with varying intensity up to the present day. Many experts
127
came endeavouring to improve the mining and to bring it on an up-to-date
standard. Mechanised mining was introduced in 1956, but the most impor-
tant share in production falls to handwork. The discontinuous exploration
in the recent years and mining of the richest placers by handworkers
decrease rapidly the reserves of gold.
After a break of several years some exploration was done again in the
second half of 1969 and in the first quarter of 1970. Several larger placers
were tested by pitting and drilling. An attempt was made the same time
to assess the characteristics of the placers and other parameters by means
of statistical techniques as there was a mass of data available. The possi-
bility of comparing exploitation data with those obtained in exploration
was useful in evaluation.
This paper is published with the permission of the Vice Minister of
the Ministry of mines, Addis Ababa.
Geology
Basic geological information
No detailed geological and geomorphological study of the goldbearing
area was ever made. Some information was collected by Jelene (1966
a, b).
The area is built up by pre-Cambrian rocks, the older series of which
is composed of gneisses, seridte-chlorite-amphibolitic schists and quartzi-
tes with emplacements of granites, massive amphibolites and even ser-
pentinites. It is unconformably overlain by a younger series of meta-con-
glomerates and quartzites. There are remnants of young flood basalts in
the northern sector of the area (Fig. 1), capping tìie highest ridges and
belonging to the young Rift valley vdLcanism or even to the Trap series
according to M o h r (1962).
The alluvial gold-bearing placers are confined to depressions in the
present relief, representing mainly dry old beds of rivers and streams.
The absence of basaltic rocks in gravel indicates the pre-volcanic age of
the fluvial cycle of erosion. According to Möhr (1962) there is no firm
datum line for the effusions of these lavas, which might even be of Qua-
ternary age. In any case, however, Adola volcanics must be very young.
Two phases are of importance in the development of Adola placers.
The landscape at the end of an erosion cycle, probably a peneplained one,
was abruptly uphfted, and the erosion created valleys in which the gold-
bearing placers accumulated. It miost have been one rapid and short event
as there is only one relatively thin gravel bed directly overlying the ba-
sement. After a calm phase in which clayey overburden accumulated over
gravel, a second, uplift lowered again the local base-level of erosion. It
manifests itself by the young and recent erosion. The new drainage pattern
has developed on the old one.
Fig. 1. Drainage pattern and larger placers in part of the Adola area
SI. 1. Dolinska mreža in večja prodišča na območju Adele
9 — Geologija 14	129
Alluvial and deluvial placers originated probably during Quaternary
pluvials. The local movement of the floor has been caused by faulting of
the Rift system, which has continued up to the present times.
The origin of the primary gold is unknown. Angular gravel and local-
ly very coarse nuggets prove that the primary deposits are very near, and
so does the gold-bearing deluvium. The placers extend along the general
trend of the basement. The primary gold may have originated and have
been derived from the following possible sources:
(a)	Hydrothermal quartz veins with sulphides, connected to igneous
activity,
(b)	Lens-Ике quartz inclusions in massive amphibohtes, interpreted
tentatively as products of lateral secretion (Jelene, 1966 a, b).
(c)	Mineralization in quartzites,
(d)	Meta-conglomerates as ancient fluviatile deposits.
Gold was found intimately dispersed in quartz pebbles, and fine gold
was found in quartz detritus on slopes as well. Locally abundant black
tourmaline appears in vein quartz. Osmiridium minerals have been found
together with gold in some placers. These facts would rather point to the
primary auriferous quartz veins of hydrothermal origin. The primary
sources may be controlled by a certain feature in the trend of the base-
ment, it being a lithologie unit or a tectonic zone. The primary sources
are not necessarily rich and may not be profitable to wwk. They could be
located by panning detritus on slopes. No such attempt to trace them
working up hill has been made as yet.
Placers
The bulk of the gold is contained in horizontal or low-dipping sheets
of fluvial alluvium laying directly over the bedrock. The placers are com-
posed of bedded gravel and sands of different size. They are covered by
layers of loam and clayey silt. The length of placers is up to several kilo-
meters (Fig, 1), the width up to several hundreds of meters, and the
thickness up to several meters. The gravel-to-overburden ratio ranges
between 1 : 1 to 1 : 20. The original accumulations of alluvial material
have been locally destroyed by the recent erosion and re-deposited at
a lower level. The southern part of the area has been especially strongly
affected by the intense headward erosion.
In addition, there are deluvial gold-bearing deposits of detritus on
slopes of the valleys, being covered by the same loamy layer. They are
often in relationship with fluvial alluvium.
The gold-bearing gravel is mainly coarse and unsorted, the pieces
more or less rooinded or even angular, depending on the length of trans-
port. It is composed of country rocks, with quartz and amphiboHtic varie-
ties prevailing. Cobbles and boulders are often in headwaters.
The distribution of gold is very irregular. There are rich pay streaks
alternating with poor or even barren zones. Gold tends to accumulate in
the basis of the gravel bed, and it burrows also into the soft schistose bed-
1ч:ск. Gold grains are fine, flaky and smooth prevailingly. Bigger irregu-
130
lar nuggets are found as weU. Fragments of quartz often adhere to the
gold. The gold contains some silver. It is about 900 fine in average.
The tenor of gold in gravel ranges between traces to 20 g/cu. m, the
average being in the order of 1 to 2 g/cu. m. Higher average values are
lare. Exceptional erratic values may reach several hundreds of grams.
The "overall tenor" (g/cu. m oall) takes into consideration the thickness
of gravel and overburden together, and is a useful parameter for evalua-
tion. From the exploitation point of view, the tenor can be boosted up by
eliminating ground carrying low values. The ore grade can be varied in
certain limits in this way, at the same time affecting the volume of gra-
vel and reserves of metal as well.
Mining
Methods applied
Hand operations and mechanised operations are applied in mining.
Handworkers extract the gravel by means of pits or by ground-sluicing.
Pitting is done in a simple way using craw-bars and showels, buckets
being hoisted by hand or by windlass. The workers seek and adhere to
rich pay streaks, excavating the richest, lowest layer of the gravel bed
only. Underground development of unsupported excavations is extremely
limited. Consequently, it is believed, the recovery cannot be higher than
40 per cent or so.
Groimd-sluicing is applied where ground is sufficiently inclined and
there is enough current water available. The recovery is more satisfactory
although some fine gold is undoubtedly lost. The handworkers recover
gold by panning, and they sell it to the Mining administration. Assistance
is provided to handworkers in tools and arrangements for water supply.
Mechanised mining has been practised in Shanka and Upper Bore
placers. Overburden and gravel are excavated by draglines, which feed
a washing plant mounted on a pontoon and floating in an artificial water
pond. The recovery in this "wet" method is believed to range somewhere
between 80 and 90 per cent as some gold is lost by dragline work in muddy
water. In addition, unnecessary excavations deep into the bedrock must
be made in shallow gravel to aHow sufficient depth of water to float the
pontoon. A land-based fixed or moveable washing plant might serve
better. A higher efficiency and better gold recovery might be anticipated,
and consequently lower production costs in "dry" mechanised method.
The machinery is electrically powered. A specially built hydroelectric
power plant and a stand-by Diesel power plant are available.
For the future it looks like mechanised "wet" mining will have to be
gradually abandoned, and the weight of exploitation will shift more and
more to "dry" or semi-mechanised methods and to handwork.
Workable reserves
Geological ore reserves as tabulated in different reserves accounts
(2 e b r e, 1968 ; Griffith, 1968) are not industrial reserves. The eva-
131
luatìon of indiistrial reserves, which are exploitable or workable at the
profit, has been difficult. The workable reserves depend mainly on the
C4Dst of mining. As exploitation is going on in different placers at the same
time by different methods and costs, the economy of the whole mining
venture interplays the payable tenor limit of individual placers.
The reserves accounts are complicated by not always reliable prospect-
ing results. In addition, some placers have already been partially exploited
by handworkers, and no evidence has been left on the excavated volumes.
In order to find out the minimum tenor of the ground to insure a
profitable operation, the alternatives of possible extractions and the factors
affecting the cost shall be worked out. Once the tenor to break even is
known, the boundaries of the workable area of a placer can be delineated
and the recoverable reserves calculated. In doing so^ the accuracy of
reserves estimate must be known.
No accurate evaluation of gold reserves for individual placers of Adola
gold field is possible at present. The estimated order of magnitude of total
workable reserves indicates that there will be mining going on for years
to come, especiany if new reserves will be discovered in time.
Estimate of cost and payable tenor limit
Referring to the cost of mining tliere are unfortunately very few data
available tO' calculate it and to find out the exact value hmits of what
can be extractable at the profit at present. Zebre (1967) attempted to
calculate the working cost per gram of recovered gold. His figures are
shown in Table 1 together with some additional parameters. Estimates
can be made on this single information from 1965 and 1967.
Table 1. Working costs per gram of recovered gold and basic parameters
Considering the price of gold E$ 87,50 per Troy ounce (31,1035 g) and
an average fineness of Adola gold 900, the resulting monetary value of
one gram produced gold is E$ 2,53. The corresponding payable tenor limits
in gravel and overall for two mechanised placers and the calculated aver-
age are given in Table 2.
132
Table 2. Payable gold tenor limits in Upper Bore and Shanka placers
It can be seen from Table 1 and 2 that the mechanised operation in
Upper Bore was profitable (0,62 > 0,153), meanwhile that in Shanka was
worked at a loss (0,37 < 0,43).
Theoretically, the average workable grade in a placer should equal
the payable tenor limit. In practice it must be higher for the recovery
and its limited accuracy as well as the degree of reliability of the reserves
estimate. The margin to be considered depends on the accuracy of data.
However, the minimum workable overall gold tenor for a certain
method applied under different local conditions is not the same. The
minimum tenor to break even is a variable. Lacking any accurate recent
data on the cost of mining, an average 0,25 g/cu.m overall may be regarded
as the order of magnitude of the workable tenor limit for the "wet"
dragline mining for the time being.
Exploration
General
Sampling of placers by pitting has been most commonly used in the
past. Square pits with 1 m side were applied mostly. Banka drill was
introduced early but was rarely used. The organisation and supervision
of exploration was left to the ability and conscientiousness of prospectors
in charge for work.
Individual washing has been used to recover gold up to the present
day. Some occasional attempts to introduce sluice-boxes and other me-
chanical panners failed. Three exploration stages were practised although
not quite consistently: "scout prospection" was followed by occasional
"preliminary prospection" in rows 1 000 m apart, and finally in "close
prospection" the pits were dug in Hnes 100 m apart with a spacing of 25 m.
In designating the lines, provision was made for intermediate Hnes 50 m
apart.
Statistical characteristics of placers
Mathematical statistics were applied to find out the quantitative expres-
sion of variability of placers, expressed by the coefficient of variation of
the gold tenor of gravel in its original state in grams per cubic meter.
The values are given in Table 3 and refer to data obtained either by pitting
(P) or Banka drilling (BD). The coefficient of variation of gold tenor was
133
computed for the placers listed in Table 3 either in whole, aU data con-
sidered, or in some cases for workable areas only. By analogy, the results
may be generalized for the whole Adola area.
Table 3. Variability of gold tenor in gravel
Gold tenor in gravel
The coefficient of variation reflects primarily the properties of the
placer, although it can be affected also by other factors such as sampling
method, number and quantity of individual samples etc. The values from
Table 3 range between 92 and 200 per cent. Accordingly, the placers belong
to groups of very- and most irregular mineralizations (K r e i t e r, 1968).
For workable parts of placers, with erratic and extremely high values
reduced, an average value of 110 per cent may be characteristic.
It seems that there is no essential difference in variabilLty of tenor
between longitudinal and transversal directions in placers.
Table 4. Variability of gravel bed thickness
134
The thickness of the gravel bed varies as well, an average being in
the order of magnitude of about 1 m. The corresponding coefficient of
variation of thickness varies between 50 and 60 per cent. It is given for
some cases in Table 4. The eventual interdependence between the thickness
and the gold tenor of a placer bed was examined for three cases. The
coefficient of correlation shows the linear relationship between the two
parameters. An average value of about —0,1 was obtained, proving an
extremely weak inverse statistical dependence without any practical
significance.
There might be a correspondence between the tenor and the composi-
tion, and the size of gravel. No evaluations have been made for these
interdependences so far.
The distribution of gold in exploration openings and the variable
thickness of gravel bed are shown for a part of Kelecha placer in Figure
2. The irregular distribution of gold is evident. There are always low-grade
or even barren zones adjoining rich pay streaks as the concentration of
heavy gold in running water was affected by its changing velocity and
transporting power, both extremely variable in shallow streams.
The variability of gold values along the placers is shown for Kelecha
and Lower Bore in Figure 3, the values being the average tenors in indi-
vidual exploration lines across the placers. The graph illustrates the
character of variability of tenor in Adola placers. The diminishing tendency
downstream is evident. In the case of Kelecha there was a probable lateral
feeding in the headwaters.
The statistical frequency distributions of tenor for Kaj emiti. Lower
Bore, Kelecha and Shanka placers are shown in Figure 4. The distribution
curves have an assymetrical, right skewed form, peaking lower than
Fig. 2. Detail from Kelecha, showing variability of gold tenor and thickness
of gravel bed
SI. 2 Izpremenljivost vsebnosti zlata in debeline prodne plasti; detajl iz Keleche
135
Fig. 3. Distribution of gold tenor (mean values of individual lines) along the placers
SI. 3. Porazdelitev vsebnosti zlata (srednje vrednosti posamičnih vrst) vzdolž prodišč
136
и
o
ce
a
<U £
G >0
O
M тЗ
C s
-H
'S Л
si
.5 <u
. «
g >
S S
-0 ^
M
•s 8
Ci X!
o Oí
Tí
3 >
£ I
.2 B
X¡ g
5P
^ P
C ÇL,
Ш
C
«I'
^
rjj
ђј)
fe
137
1 g/cu.m. Such a frequency distribution of values is often obtained in
mineral exploration. It rarely conforms to normal or Gaussian distribu-
tion, thus limiting the direct use of statistical formulas in depmdence on
the degree of assymetry. Nevertheless, the fonnulas can serve as approxi-
mations to provide rough estimates in search for guidelines in practical
exploration.
The average values have been calculated as the arithmetic mean.
Although commonly used and easily calculated it is greatly affected by
extreme values. Therefore the extreme values must be appropiriately
reduced in calculations.
Comparison of exploration data and mining results
Errors in mineral exploration are unavoidable. A 100 per cent ac-
curacy of exploration cannot be achieved in practice, and allowances are
made for the reliability of reserves accounts.
The recent mining in Adola has rendered possible the comparison of
exploration data with those obtained by exploitation. Shanka and Upper
Bore placers have been mined so' far. The possibility of verifying the
reliability of exploration data, however, is limited as factors such as the
accuracy of sampling methods and recovery in mining are not exactly
known.
The discrepancies between the reserves calculated from the explora-
tion data and those revealed by mining are shown for some cases in
Table 5, the geological relserves estimated on an assumed 90 per cent
recovery in "wet" dragline mining method. Exploration data were ob-
tained by pitting (P) or Banka drilling (BD), both sampling methods
performed in an exploration grid 100 X 25 m.
Table 5. Discrepancy between the reserves calculated from exploration and
mining data
The figures reveal a positive discrepancy between 117 and 169 per
cent for geological reseives with reference to the exploration data, indi-
cating the order of magnitude only. Nevertheless, there has been more
gold mined as shown by exploration. The same sign of error points at
a systematic error in both methods of sampling. The discrepancy of about
160 per cent in Banka drilling is very high.
138
There is no possibihty of oomparison for handworked placers such
as Kajemiti or Demi Denissa because ob unknown and probably extremely
low recovery in the applied way of extraction.
Analysis of grid density
The above oomparison for the mined^mt portions of placers is not
sufficient to judge the adequacy of the applied exploration grid. In order
to assess its optimum density and the accompanying sampling errors in
Adola placers, statistical considerations were applied. It must be pointed
out again that the statistical evaluation of an assymetric distribution of
values is not but an approximation, and the figures provide for orientation
values only.
The principal determinants in the choice of a grid density are the
required accuracy, expressed as an allowable error, and the variability of
placer. Additional factors to be respected are the size of placer, the
sampHng method and the economy of exploration as well.
The reserves ready for mining are classified as "measured" reserves.
In general, however, an allowable error of ± 15 tO' 20 per cent is com-
monly accepted for the highest category of reserves, regardless the clas-
sification. The resulting error of an estimate is the statistical sum of aU
occurring errors, and shall not be higher than the allowable error.
For the generally low-grade Adola gold placers the allowable error
should be in the order of 15 per cent at maximum.
The reserves of gold are calculated of the average tenor and thickness
of the gold-bearing gravel. The mathematical mean (Ma) as a statistical
measure is calculated with an error (p), which depends on the degree of
variability (V) and the number of cases (n). It is expressed as a percentage
of the arithmetic mean as shown in the known formulae
The greater the number of samples, oi- the denser the grid, the smaller
the error and the greater the variability, the greater the error. Factor (f)
refers to the probability of error. In rough use of the formula it can be
accepted to equal 1. The error determines the accuracy of the arithmetic
mean Ma ± fp.
Theoretical relationship between error (p) and number of samples (n)
is shown in Figure 5 for varioiis coefficients of variability. It can be seen
that by increasing the number of samples the values of error do not
diminish proportionally.
The above formula does not consider the disp)osition of samphng points
in a placer. However, a regular grid pattern with uniformly spaced open-
ings is required to provide correct information on the whole placer. The
formula is also independent on the size of the placer. Introducing the unit
area for sampling point (Fo), defined as part of the total area (F) of a
139
Table 6. Errors pertaining to the arithmetic mean tenor of gold in gravel in dependence on grid density
140
Fig. 5. Relationship between error of the mean and number of samples for
different values of coefficient of variation
SI. 5. Odvisnost napake sredine od števila vzorcev za različne vrednosti koefi-
cienta variacije
If the allowable error is known the unit area can be calculated, deter-
mining the density of the exploration grid. For the same error the unit
areas are in linear relationship with the total areas. The larger the placer,
the fewer openings are required for the same error, and the more econom-
ical the exploration.
The unit area for a rectangular 100 X 25 m grid commonly used hit-
herto in Adola is 2 500 sq.m. The errors pertaining to the arithmetic mean
tenor in gravel for this and larger unit areas were calculated for several
of the explored Adola placers by omitting the appropriate number of
data, considering spacings 25, 50, 100 and 200 m in lines 100 m apart. The
average values of errors were calculated of 2, 4 and 8 tests correspondingly.
The values are shown in Table 6 together with other data.
The relationship between the standard error of the arithmetic mean
tenor (pi) and the spacing, or the unit area, is given in Table 8. The
figures are approximate average values, preference being given to the
results from workable areas.
141
Table 7. Errors pertaining to the arithmetic mean thickness of gravel bed
Similarly, the error of the mean thickness of the gravel bed has been
calculated for several cases. It is given in Table 7.
The order of magnitude of the standard error of the mean thickness
(pa) for a unit area of about 2 500 sq.m, or a spacing of 25 m, would be
about 4,8 per cent. Supptosing that the relationship between this error and
the grid density is in a similar functional depedence as that for the tenor,
the approximate corresponding values are tentatively given in Table 8.
The relationships are shown also graphically in Figure 6.
Table 8. Relationship between standard errors of the arithmetic mean pj and pg
and grid density
Order of magnitude of standard error in per cent
Fig. 6. Relationship between errors pi, рг and рз and spacing or unit area
SI. 6. Odvisnost napak pi, pa in рз od razdalje oziroma specifične površine
142
The accuracy of reserves estimate is influenced by additional errors
in sampling operations and measurements, which can be embraced together
as the technical error (рз). EiTors are random and systematic. Random
errors may occur in measuring volume, thickness and distance, weighing
the gold, determining areas from the maps etc. (measuring error p,„). In
a sufficiently high number of cases random errors cancel out as they have
alternate signs. Systematic errors may influence the results considerably
more. In conditions of Adola exploration they can arise from the follow-
ing sources:
(a)	Problematic determination of volume of excavated gravel because
of swelling (swell error ps),
(b)	Uncertainity of volume of recovered Banka samples (volume
error pv),
(c)	Technical shortcommings in sampling because of inaccurate and
P'oor performance of labour (operation error р„-),
(d)	Losses of gold in panning (panning error Pp),
(e)	Confusion in evidence of samples, loss or even theft of gold (care
error Pc),
(f)	Errors in calculations (calculation error pc ).
However, the possible sources of errors are objective and subjective
in nature. Means shall be found how to reduce or eliminate them.
Since the figures of reserves are calculated by multiplication and addi-
tion of not related values with errors p'ertaining to them, the resulting
error (po) is obtained by addition of squares of partial errors
Introducing the guide-values from Table 8 for three cases of allowable
error (po) 10, 15 and 20 per cent, the carresponding values of technical
error (рз) are shown in Figure 6.
The relationship between the resulting error (po) and the unit area
or spacing is better illustrated for foiir cases of technical error (рз) 5, 10,
J 5 and 20 per cent in Figure 7. The values are given also in Table 9.
Table 9. Relationship between resulting error p^ and grid density
A general conclusion can be made, however, that in Adola conditions
with an estimated margin of technical error ±10 per cent the unit area
must be kept at about 2 500 sq.m per sampling point, if an approximate
143
± 10 to 15 per cent degree of reliability of reserves estimate shall be ex-
pected. Provided the margin of technical error wiU be lowered by improved
quahty of work, the unit area might be enlarged to 5 000 sq.m, especially
in large placers.
Fig. 7. Relationship between resulting error po and spacing or unit area
SI. 7. Odvisnost končne napake po od razdalje oziroma specifične površine
From the point of view of variability of placers there is no need to
adhere to the introduced 1 :4 rectangular exploration grid as the an-
isotropy in placers is not strong. A 50 X 50 m square grid would be weU
applicable. Even better would be a rhomboid grid vdth lines 50 m apart
and spacing 50 m, the diagonal 100 m in the longitudinal direction of the
placer as shown in Figure 8. It could have also a uniform spacing of 54 m
between pits, the lines 47 m apart in this case.
Fig. 8. Proposed rhomboid grid
SI. 8. Predlagana romboidna mreža
Exploration in rows, however, has certain advantages in hard-passable
bush country.
In deciding on the grid density the total area of a placer is to be
considered, and the necessary number of samples approximately estabh-
shed. It may be increased or reduced, depending on how the probable
errors would be influenced.
144
Sampling
Pitting and Banka drilling as sampling methods differ essentially in the
quantity of produced samples. In Banka drilhng, the obtained core is
already a direct partial sample of the placer. In pitting, the relatively
large intermediate samples have to be first reduced to partial samples,
introducing so an additional reduction error (pr), which, in turn, is part
of the technical error. As Banka sampling is burdened with errors too
there is no clear advantage in reliability for one or another of the methods.
Anyhow, in groundwater-logged and swampy ground Banka drilling is the
only applicable method. Both methods require good organization and
supervision.
There is a loss of gold in washing the gravel as well, this being another
source of error. The weighing of gold on sensitive analytical balance can
be performed accurately.
Pitting. Pitting is a slow process appHcable in dry, well-standing
ground. Round pits are prefered, the diameter being in the order of 0,7 m.
Working at a piece-rate system, the diggers automatically keep the
diameter at the minimum.
The unit volume of gravel in situ in a 0,7 m diameter pit measures
0,384 cu.m, and about 0,48 cu.m when loosened, its weight nearing a ton.
Such a quantity is to large to be washed by panning as is the established
practice in Adola. For application of sluice-boxes about 4 cu.m of water
would be required per unit volume, a requirement not easily met in mainly
dry Adola vaUeys. In addition, by treating the whole quantity of excavated
gravel, the errors in volume as well as losses in handling, transport and
washing the gravel are unavoidable. So, the samples must be reduced. In
the past, at Adola, the quantity washed in pitting was 0,1 cu.m, or some-
times even 10 random bateas of gravel (about 0,07 cu.m each) were
washed only.
The minimum required volume (weight) of one partial sample depends
on factors variable within the limits of a placer. The low tenor and very
irregular distribution of minute gold grains, together with highly chang-
ing size of gravel, would require large samples. On the other hand, the
weight of partial samples shall be kept as low as possible from the point
of view of economy.
The reduced sample shall represent as closely as possible the tenor
of the excavated gravel from the pit. The intermediate sample cannot be
worked down as in treatment of samples of primary deposits, and there
are no formulae for determining the reliable sample quantity of gold-
bearing placers. It is important to keep the granulometrie composition of
leduced samples as unchanged as possible by having the material com-
pletely mixed up. However, the reduction error is lower, the larger the
reduced sample and the higher the number of partial samples. Hawing
this in mind, the excavated gravel shall not be reduced too much, espec-
ially in coarse gravel.
The following pirocedure has been adopted recently in reducing the
intermediate samples. The excavated gravel is piled up on a clean ground,
10 — Geologija 14	145
thoroughly mixed up and flattened on a 15 em high wooden cross to get
it quartered. A channel is made from the periphery toi the center of each
quarter to fiH a standard unit of measurement. It is a trough-like wooden
box 15,3 cm high, measuring 60 X 30 cm at the top and 60 X 20 cm at the
bottom, its volume being 0,023 cu.m. The gravel in the box is well com-
pacted. Mainly for the sake of easier calculating, and additional sample
is taken to have 5 boxfuls as one partial sample, totalling 0.1 cu.m of gravel
in a compacted state and weighing about 200 kg.
The volume of samples is problematic because of the swelling of gravel
in loose condition. The gain in volume may be the source of a dangerous
error. It is measured by the ratio of volume of loose gravel against un-
distiu*bed volume (swell factor). Several tests gave values ranging between
1,05 and 1,4. Swell factor varies considerably for different &ize& of gravel,
which range between that of coarse sand and boulders of several deci-
meters in diameter. However, an average value of 1,25 has been assumed
for Adola gravel, this being, of course, a rough general approximation.
Fig. 9. Dimensions of standard unit of measuremeni
SI. 9. Dimenzije standardne merske enote
The dimensions of the standard box (Fig. 9) were established on the
assumption that the loose gravel can be compacted to about half of its
original undisturbed volume. Thus the volume of compacted gravel in
one box would correspond to 0,02 cu, m of undisturbed gravel. It depends
also on how well and diligently the compacting and tamping has been
done.
Tests were carried out to check the reduction error for the described
procedure. The quantity of gold was determined in single boxfuls and
a series of 4 boxfuls for which the mean deviation from the arithmetic
mean was calculated as given in Table 10.
A roughly calculated average mean deviation of gold tenor in gravel
for series of 4 boxfuls (totalling 0,08 cu.m) is in the order of magnitude
of 10 per cent. In terms of the average gold tenor of placers, this means
a possible reduction error (pr) of :
± 17 per cent for a placer displaying an average gold tenor of 0,6 g/cu.m
±10 per cent for a placer displaying an average gold tenor of 1 g/cu.m
± 5 per cent for a placer displaying an average gold tenor of 2 g/cu.m
± 3,3 per cent for a placer displaying an average gold tenor of 3 g/cu.m
The error will be even smaller because 5 boxfuls instead of 4 are used
in reducing the intermediate samples. It may be expected to^ cancel out
since the probability of deviations in either direction is equally great.
146
Table 10. Mean deviation from the arithmetic mean of gold tenor for series
of unit boxes
The same might be expected, however, for the swell error. Under un-
favourable conditions the errors may accumulate to a significant systema-
tic error. It is believed that most discrepancies in the past were closely
connected with errors in reduction and swelling.
About 10 cm of bedrock is excavated in the pits. In has been a common
practice in Adola to wash 1 batea of bedrock only (about 0,007 cu.m). An
additional error originates in this way.
Banka hand drilling. Banka drilling is a quick, clean and
cheap method if properly executed. It is simple to operate and easy to
transport. On the other hand it is not applicable in very coarse gravel
and may not be sufficiently accurate because of relatively small volume
of core. The evaluation of the data is liable to errors because of uncertain
volumes of recovered samples. The general view prevails that banka dril-
hng in placers yields lower results than pitting, and this was proved also
in the past for Adola.
6V2 inch Banka hand drills have been used in Adola exploration. The
method applied recently is basically similar to that suggested by Grif-
fith (1960). In clayey overburden a spiral auger is used. On reaching the
gravel, casing is lowered and the depth of overburden no'ted. The casing
is hammered down through the gravel into the bedrock for about 25 cm.
The total length of casing is measured, and after piilling it out, the thick-
ness of the penetrated bedrock is established and deducted from the total
147
length to obtain the thickness of gravel indirectly. A 100 per cent core re-
covery is (theoretically) assured in this way, and the problematic consi-
deration of swelling of the loosened gravel avoided.
The casing is emptied and the entire quantity of gravel washed as one
sample. About 10 cm of bedrock is washed together with the gravel.
Since the diameter of Banka casing is small the recovered volume of
gravel per meter is about 0,018 cu.m, and its weight about 36 kg only. The
calculation of tenor per cubic meter of gravel is a sensitive operation, de-
pending heavily on the volume gathered inside the casing, which, in turn,
is affected by the size of gravel and the sharpness of the cutting edge of
the casing shoe. In other words, the diameter of casing to calculate with is
a variable quantity which may be the source of a systematic error. Grif-
fith (1960) suggests the use of the external diameter 0 165 mm. Internal
diameter 0 136 mm was used in recent calculations by the author (1970),
the choice based mainly on the past results and coarse gravel, but a wide
—21 per cent margin of possible error was taken into account. Looking
at the section of the cutting-shoe of the casing shown in Figure 10, it is
evident that some gravel under the cutting edge will be pushed aside, espe-
cially when it is coarse. Cobbles and boulders may even partly or comple-
tely block the entrance, bringing up little or no sample at all.
The conclusion is, that in only very fine-grained material, and with a
sharp bevelled edge of the cutting shoe, the external diameter shall be used
in calculations. The coarser the material the smaller the diameter to^ be
Fig. 10. Section of 6 У^ inch Banka
casing shoe
SI. 10. Prerez »čevlja« 6 Уч eolskih
Banka obložnih cevi
used. No corresponding tests have been made to date. As long as this is not
done the median diameter 0 150 mm and specific volume 17,7 cu.dm/m
might be the best dimension to be considered.
Specific volumes for Banka casing diameters and the differences of
volume in per cent with reference to the internal diameter 0 136 mm are
given in Table 11.
Comparing the quantity of one Banka partial sample (about 36 kg/m)
with that in pitting (about 200 kg), the question arises how representative
are both samples? There is no reduction in Banka sampling. Samples of
regular cylinder form are cut across the gravel bed regulary, and this
may be advantageous if compared wdth a constant volume of partial samp-
148
Table 11. Data on specific casing volume in Banka hand drilling
lets in pitting, regardless of the thickness of gravel. The representativeness
of Banka samples is theoretically perfect. In practice this is valid for
fine^-grained material and less for coarse gravel.
Sampling by pitting and Banka drilling is schematically illustrated in
Figure 11, with errors listed for both methods. The statistical sum of the
reduction error (pr) plus swell error (ps) in pitting is confronted with a vo-
lume error (pv) in Banka drilling, the rest of the errors approximately
equal in both methods. The sum of both errors in pitting might be ex-
pected to range about 10 per cent in average. They are random and may
even cancel out. The volume error in Banka sampling is difficult to as-
sess. It may be believed that in average its effect would not be essentially
different from the combined effect of both errors in pitting. This question
could be answered, at least partially, by a series of comparison experi-
ments.
Fig. 11. Schematic illustration of sampling by Banka drilling and pitting
SI. 11. Shematski prikaz vzorčevanja z napravo Banka in jaški
149
Hand washing of gold. Gold is recovered from gravel by
hand-panning on wooden bateas. Some fine gold may be lost in panning
by unskined panners, or when muddy water is iised for washing. Due to
the shortage of water this might be quite often the case. The average re-
cowery in panning is not known, but evidence exists that losses do occur
and contribute to errors.
Hand panning in Adola is a long established practice, which would be
difficult to replace by sluice-boxes, rockers or mechanical pans. Although
the capacity of these devices is higher it is questionable if the recovery
would be better. Careful handwashing on bateas may yield the best pos-
sible recovery if clear water is used, at least comparable to that of mecha-
nical washing plant used in "wet" mining.
Costs and performance
The errors in sampling can be reduced by increasing the number of
samples. A balance must be found between the necessary accuracy and
the economy of exploration, although in low-grade deposits the expense
should not be always the decisive factor.
In hand pitting, a crew of two workers dig one pit. Performance and
costs in the recent exploration campaign are given in Table 12.
Table 12. Performance and specific costs in pitting
* labour only
The average efficiency in pitting is 0,51 m per man per calendar day,
and the average pirogresis 1,02 m per pit correspondingly. A monthly ave-
rage of 31 m was achieved per pit or a crew of two. A piece-rate system
of payment was introduced. It seems difficult to raise this efficiency much
more under the existing conditions, but progress can be increased by dig-
ging more pits at the same time.
In Banka hand drilhng, an equal specific efficiency per man can be
achieved in the best case. The progress, however, is much better than in
pitting. A short account of performance and costs in Banka hand drilling
is given in Table 13.
150
Table 13. Performance and specific costs in Banka hand drilling
The collective nature of work under a concentrated supervision and
favourable field conditions made it possible to achieve a speed up to 8 ti-
mes higher than in pitting. In Kelecha, for instance, two or three drills
operated simultaneously in tandem, the workers changing continuously in
order to reduce breaks in progress. The efficiency was stimulated by a col-
lective bonus system. A crew of 14 drilled 8 m per 8 hours shift, аД wor-
king phases included. Comparing with the efficiencies of less than 2 m per
drill per 8 hours shift achieved in Adola previously, this was about 4 times
higher.
The relative comparison of average costs and efficiency per man as
well as progress per sampling point is given for both methods in Table 14.
Table 14. Comparison of specific costs and efficiency of pitting and Banka
hand drilling
As cost is concerned, digging two pits would correspond approximately
to operating one Banka drill, the progress in Banka twice that of pitting
at least. The advantage in speed of Banka drilling over pitting is evident.
Considering the better representativeness of Banka samples and approxi-
mately equal average reliability of both methods. Banka sampling is the
superior method, especially in favourable conditions and when time is the
151
decisive factor. On the other hand, linking the accuracy with the cost, the
number of pits for the same money is twice that of Banka operation, fa-
vourably influencing the reliability but loosing time.
Conclusions
In sampling Adola placers by pitting or Banka drilling an average
±15 per cent rehabihty of reserves estimate can be expected in general
for a unit area in exploration grid of about 2500 sq.m, provided the samp-
ling operations are carried out carefully.
Banka hand drilling, if properly executed, is generally advantageous
over pitting in terms of economic efficiency. Its accuracy depends on the
size of gravel and may, however, compete with that obtained by sampling
in pitting. Some experiments would be needed to check it. Until further
data are available the median diameter of casing shall be oonsidered.
Banka method is not applicable in very coarse gravel.
A square grid 50 X 50 m or a rhomboid one 100 X 50 m should be used
wherever possible.
In the past, exploration in Adola seems to have been suffering mainly
from systematic negative errors attributed to swelling in pitting, and in-
correct execution of coring in Banka drilling. The applied unit area of
about 250Ò sq.m per sampling point was appropriate. As the reHabihty of
data, the gold tenor values are slightly underestimated in average.
In an evaluation attempt such as this, certain assumptions have been
made. Hence, no mathematically precise results can be produced, and pre-
scriptions cannot be laid down as how to eliminate the errors. There are
many objective and subjective factors involved, some of them hardly con-
trollable in Adola.
Statistična ocena raziskav zlatonosnih naplavin
na območju Adole (Južna Etiopija)
Milan Hamrla
V državnem rudniku Adola, okrog 500 km južno od Addis Ababe, pri-
dobivajo zlato iz proda, ki so ga nanesle reke verjetno v pluvialu kvartar-
ja. Doline z zlatonosnimi naplavinami so danes povečini suhe. Prvotna
nahajališča zlata so v bližini, vendar niso raziskana. Naplavine leže na
predkambrijskih skladih, prekrite pa so z gHno. Debelini zlatonosnega
proda in prekrivke sta v razmerju 1 :1 do 1 : 20. Povprečna vsebnost zlata
v produ je 1 do 2 g/m®. Zlato pridobivajo' ročno in mehanično.
Koeficient variacije vsebnosti zlata v produ se menja med 90 in 200 ®/o
za različna nahajališča, za območja naplavin, ki so primerna za izkorišča-
nje, pa je v povprečku okrog 110 '"/o. Tudi debelina prodne plasti se menja;
povprečna vrednost ustreznega koeficienta variacije je 55 "/». Zelo slaba
inverzna odvisnost vsebnosti zlata in debeline prodne plasti je brez prak-
tičnega pomena.
152
Naplavine vzorčujejo z jaški ali pa z vrtalno napravo Banka. Kot je
pokazalo rudarjenje, so rezultati preteklega raziskovalnega dela podani
z negativno sistematsko napako; v resnici so bile pridobljene količine zlata
vedno nekoliko večje od izračunanih.
Navadno vzorčujejo po mreži 100 X 25 m, ki ji utsreza specifična po-
vršina 2500 m^. Rezerve računajo iz povprečne vsebnosti zlata v kubičnem
metru proda v prvotnem, zbitem stanju in povprečne debehne prodne
plasti. Oba parametra sta kot aritmetična sredina določena z napakama
(pi ,p2), ki rasteta, če je mreža redkejša in specifična površina večja, ozi-
roma število vzorcev manjše. Točnost vzorčevanja je odvisna še od drugih
napak, ki jih označimoi skupno kot tehnično napako (рз). Pri jaških sta
to predvsem napaki zaradi povečanja volumna izkopanega proda (ps) in
njegovega reduciranja (pr), pri Banka napravi pa volumska napaka (Pv)
zaradi negotovega volumna pridobljenega vzorca v odvisnosti od debehne
proda. Vse te napake je številčno težko zajeti, ker so odvisne v glavnem
od izredno' hitro spremenljive granula ci j e proda. Tehnično napako oce-
njujemo v povprečku z 10 dO' 15 ®/o.
Končna napaka določitve rezerv zlata v relativno siromašnih adolskih
naplavinah ne bi smela biti dosti večja od ± 15 "/». Iz grobega računa sta-
tistične vsote napak sledi, da tej zahtevi ustreza vzorčevanje v mreži s
specifično površino okrog 2500 m^ pri sedanjem načinu in točnosti izvedbe.
Doslej uporabljena mreža je torej pravilna. Ce bi bilo vzorčevanje izve-
deno precizneje, bi bilo mogoče mrežo razširiti.
Primerjava učinkovitosti vzorčevanja z jaški in napravo Banka poka-
že, da je ta bistvenoi hitrejša, vendar še enkrat dražja v povprečku. Gre ji
prednost. V debelem produ pa njena točnost ni zanesljiva, kar bi bilo
treba dognati s poskusi.
References
A r k i n, H., C o 11 o n, R., 1965, Statistical methods, New York.
Griffith, S., 1960, Alluvial prospecting and mining. London.
Griffith, S., 1968, Report on Adola gold areas. Archives of the Ministry
of mines.
Hamrla, M., 1970, Exploration and gold reserves of Kelecha valley.
Archives of the Ministry of mines.
Jelene, D., 1966 a, Adola Gold Placers and Nickel-Chromium Ore Depo-
sits. Geologija 9, Ljubljana.
Jelene, D., 1966b, Mineral occurrences of Ethiopia. Addis Ababa.
Kreiter, v., 1968, Geological prospection and exploration. Moskow.
Möhr, P., 1962, The geology of Ethiopia. Addis Ababa.
Zebre, S., 1967, Analysis of production costs in Adola mining. Archives
of the Ministry of mines.
Zebre, S., 1968, Prospecting results and ore reserves of gold in Sidamo
gold fields. Archives of the Ministry of mines.
153