You are on page 1of 24

THE COMPUTATION OF PERMEABILITY FROM SIMPLE SOIL TESTS

Estimates of permeability of the ground are frequently required by civil engineering work involving
the seepage of water. Several published formulae relating the permeability of beds of granular
material to their geometrical properties are quoted. In order to assess the value of these formulae
in calculating permeability, a comparison is made between measured and calculated values of
permeability for a wide range of sands.

Hazen’s formula is quite useful because of its simplicity but the result may be inaccurate to the
extent of ±2. Kozeny’s formula is the best of those quoted, and if an estimate of the “angularity” of
the grains is made, permeabilities of sand can be computed to an accuracy of about ±20 per cent.

An empirical formula is given which is equally accurate and more simple to use. Computations of
permeability by the methods described can be made only on clean sands.

INTRODUCTION

The fundamental equation which governs the streamline flow of water through soil is

Darcy’s Law :

Q=A.k.i

In which Q is the volume of water flowing through an area A of the soil in unit time under the
influence of a hydraulic gradient i. The constant of proportionality, k, is called the permeability of
the soil. Permeability has frequently to be determined in civil engineering problems involving the
flow of water; for instance, to calculate the discharge of water into an excavation, or the effect of
seepage on its stability; or to investigate problems of water supply, land drainage, coastal erosion.
It may be found by calculation, since several formulae have been published relating the permeability
of soils to their geometrical properties (see Mavis & Wilsey, 1936, Anderson, 1940), or it may be
measured on a sample of the soil. It is commonly held that permeability is best obtained by
measurement. But measured values are frequently misleading. For instance, it is difficult, if not
impossible, to obtain an undisturbed sample from a bed of clean saturated-sand. “ Undisturbed ”
samples from the field are often taken from above the water level, and they are then wetted-up
slowly from the base in an attempt to re-saturate the sand, and their permeability is determined.
But laboratory tests have shown that when saturated sands are drained and then wetted in this
way, their permeability is reduced to between 40 per cent and 80 per cent of the original value,
because of air remaining in the voids. The field samples may alternatively be reconstituted in the
laboratory, but large errors can again be introduced if the fine and coarse fractions are reassembled
in different associations. It is therefore very desirable to be able to calculate the permeability, if only
to provide a means of detecting errors of measurement which might otherwise pass unnoticed.

The purpose of this investigation was to find, by comparing measured and calculated values, how
accurately the permeability of sand could be calculated from the properties of the material.

PUBLISHED FORMULAE FOR PERMEABILITY

Some of the more important theoretical and empirical formulae giving permeability in terms of the
grain size, porosity of bed, etc., are given below. C.g.s. units are used, except for permeability which
is expressed in centimetres per second per unit hydraulic gradient. Unit hydraulic gradient
represents a drop in pressure of 1 centimetre water per centimetre distance.

Hazen- 1892 K= Cd102

This empirical formula gives an approximate value for the permeability k from the ” effective ” grain
size d10, defined as the size such that 10 per cent of the material is of smaller grains. The constant C
is approximately 100, but is stated to vary in Hazen’s observations from between 41 and 146 (Taylor,
1948). The observations of Hazen were limited to filter sands of grain size between 0.1 and 3.0 mm.
of fairly uniform grain size. The “uniformity coefficient” d60 : d10 was less than 5 ; d60 is the diameter
such that 60 per cent of the material is of smaller grains.
771 𝑑 2
Slichter-1899 k=
𝑐

This formula was derived by Slichter for uniform spheres of diameter d, from the geometry of the
voids. He calculated the relationship between the parameter c and the porosity n, and some values
are tabulated (see Table 1).

The derivation of this formula is based on that of Slichter’s formula, which is extended to cover sand
of non-uniform grain size and variable grain shape. It includes an empirical term (n - 0.13). The
parameter f varies from 800 for rounded sands to 460 for angular sands.

Kozeny-1927

Where g = acceleration due to gravity.

K1 = constant, equal to 5 for spherical particles.

S = specific surface of angular material. (Surface area per unit volume of grains.)
ȵ = viscosity of water (7 = 0.0131 c.g.s. units for water at 10° C.).

n = porosity of bed.

This formula was developed by an extension of Poiseuille’s equation for flow through capillary tubes.
The form in which it is given is that used by Carman (1938, 1939) who verified the formula
experimentally for variations of viscosity, specific surface, and porosity, and showed that it held for
some non-spherical particles.

Rose- 1950

Where f (n) is a function relating “relative resistance” to porosity, equal to unity at n= 40 per cent.

A general empirical formula was derived by Rose (1950) using dimensional analysis, which reduces
to this form when flow is streamline and when the container diameter is large compared with the
grain size. The formula applies strictly to uniform spherical particles of diameter d, but Rose’s Paper
6
shows that, unless the particles are very angular, may be substituted for d for non-spherical
𝑆
particles of mixed sizes. The formula then agrees with Kozeny’s at a porosity of 40 per cent. Two
functions relating ‘I relative resistance” to porosity are given:

(u) An experimental relationship developed by Rose.


(b) A function developed by Rapier and Duffield from the results of Rose:

Relative permeabilities computed from these formulae are plotted on a logarithmic scale against
porosity in Fig. 1, all being made equal to unity at a porosity of 40 per cent. For ah the formulae, the
relationship between log k and n is approximately linear, and the slope is greatest with Kozeny’s
formula.
Relative Permeability/Porosity for different formulae in text

Measurements of permeability were made on a number of different types of sand to compare with
these formulae. A description is given of the experimental technique.
MEASUREMENT OF PERMEABILITY
Apparatus.-A permeameter of the constant head type was used, with the water flowing upwards
through the sand. The permeameter tube was a long narrow Perspex cylinder, to which a vertical
column of ten pressure points was fitted to enable pressure to be measured within the sand bed
(Fig. 2 (a)). The points were copper tubes, one end of each being closed and drilled with very small
holes, and held in pressure-tight glands (Fig. 2 (b)). The permeameter could be made air-tight to
enable it to be evacuated in freeing the sand from air, described below.
The layout of the apparatus is shown in Fig. 3. Air-free water for permeability tests was produced
by evacuating the space above the water in the supply tank S and leaving it for several hours. A
constant head was maintained in the inverted aspirator bottle R, which could be raised or lowered
as required.
Method of measurement of permeability.-A suitable quantity of sand was weighed dry and then
placed in the permeameter in an air-free condition, the apparatus was connected up, and water was
allowed to flow through the sand. Connecting tubes were freed from air. When equilibrium was
attained, the following measurements were made :
(a) Pressure was read on the manometers M, and the hydraulic gradient i found by plotting pressure
against distance from base of sand.
(b) The rate of flow of water Q was found using the measuring jar J and a stop-watch.
(c) The length of the column of sand was measured to the nearest millimeter from a scale fixed to
the outside- of the permeameter, and porosity was calculated as described below.
(d) The temperature of the water was measured on the thermometer T which was placed
immediately above the sand.
Knowing the crosssectional area A of the permeameter, the permeability was found from the
𝑄
equation : k = 𝐴𝑖

The permeability at the standard temperature of 10° C. was then calculated, using the appropriate
value of the viscosity of water.
Care was taken to ensure that the flow of water was streamline in all cases. A small hydraulic
gradient was used when coarse materials were tested, fulfilling the conditions of streamline flow
defined by Rose (Rose, 1950). The gradient was always kept low enough to prevent the sand from
piping during measurements.
Preparation of sample.
The sand was placed in the permeameter at a uniform density, uniformly mixed, and air-free, since
air in the voids restricts the flow of water. The best way of obtaining an air-free bed was found to
be the method shown in Figs 4. Water in the permeameter and dry sand in a container were first
evacuated (Fig. 4 (a)) and the container was then tipped up and the dry sand poured into the water
under a vacuum (Fig. 4 (b)).
Segregation of coarse and fine material was difficult to avoid in non-uniform sands, but if the sand
was poured at a constant rate the non-uniformity was largely restricted to the top and bottom of
the bed.
Quite satisfactory results were obtained, with uniform sands only, by pouring the sand into water
without evacuating it, and making it “pipe” The grains separate and release most of the air ; but
small bubbles sometimes cling tenaciously to sand grains, and the sand was always evacuated when
a series of measurements of permeability was required. A third method, of placing the sand in a
dainp state, confining it with a perforated piston, and then evacuating it, was found unsatisfactory
except for coarse sands, since air bubbles are retained by capillary forces even when a high vacuum
exists above the bed.
Alteration of porosity. In this investigation, permeability was generally measured at different
porosities. The porosity of uniform sand was readily altered; it was piped, and was in a loose state
when it settled; it could then be made more dense, to any desired value of porosity, by judicious
tapping or vibration of the permeameter. Sands of mixed grain size could not be dealt with in this
way, as piping brought the fine material to the surface; the best that could be done was to place
them initially in a fairly loose state, and then increase their density in stages by vibration. Uniform
sands are much easier to manipulate. In the greater part of the experimental work, sands selected
by sieves where mesh sizes were in the ratio 2 : 1 were used. When these sands settled after piping,
a uniform hydraulic gradient was obtained through the bed, which proved that the grain size was
sufficiently uniform to prevent piping from causing appreciable segregation. Any subsequent non-
uniformity in the hydraulic gradient was known to be due to uneven compaction; if this was
observed, the sand was piped and brought back to the same porosity.
Measurement of porosity.-The porosity of the sand, n, was found from the length, 1, of the column
of sand. The length of the sand column, assuming no voids, Is, may be found from the mass of sand,
M, its specific gravity, Gs, and the cross-sectional area of the permeameter, A. Porosity is given by:

Particular care was taken to measure porosity accurately, since, as shown below, permeability
increases by 13 per cent when the porosity changes by 1 per cent. Specific gravities were measured
to an accuracy of about ±0.2 per cent, by displacement of kerosene in a pycnometer bottle. The
cross-sectional area of the permeameter was obtained to about the same degree of accuracy by
measuring the height of the water column when different measured volumes of water were poured
into it. The fractional error in measurement of the length of the column of sand was kept small by
using a long narrow permeameter tube, and making the top and bottom of the sand bed as
horizontal as possible.
Internal porosity of grains.-Some of the sands used had slightly porous grains. The total void volume,
within as well as between the grains, was used in calculating porosity, but since the pores within the
grams are too small to permit appreciable flow of water, it is only the intergranular porosity, or
fraction of the volume occupied by the pores between the grains, which should be considered in
connexion with the permeability.
To measure the internal porosity of the grains, a small quantity of sand was saturated with water in
vacuo, and then the outsides of the grains were dried by rubbing the sand between two sheets of
filter paper. The process was repeated on successive sheets until the sand would just roll easily on
the tilted paper. The surfaces of the grains were then roughly “dry” and the moisture content of the
grains was measured in the usual way. The surfacedrying process was completed in less than a
minute, to prevent appreciable quantities of water from being drawn from the internal pores.
The internal porosity of the grains was calculated from their moisture content and specific gravity
as follows:

Where w = moisture content of grains, G, = specific gravity of grains, Yw= density of water.
(a) Glass beads 0.7 -0.4 mm f=1.0
(b) Ham River sand 25 – 52 f=1.1
(c) Lough Neagh sand f=1.35
(d) Crushed quartzite 25-52 f=1.5
(e) Crushed basalt 25-52 f=1.6
(f) Crushed Pyrex glass 25-52 f=1.8
Sand grains used in tests (f= angularity factor from kozeny´s formula)
The results are given in Table 2. It can be seen that this method gave reasonably consistent
results on three successive determinations (Expts l-3), and gave a negligible value for glass beads
and crushed glass, which are known to be non-porous.

Most of the materials used had a very small internal porosity. The relatively high value obtained
with Ham River sand lying between B.S. sieves 100-200 is probably spurious; it has a relatively
large specific surface, and the thin film of water still present when the sand is wiped “dry” has
a relatively greater effect on the measured moisture content. Apart from this, only two of the
sands tested had an internal porosity greater than 2 per cent.
The intergranular porosity na calculated from the total porosity n and the internal porosity nb
from the relation:

Intergranular porosity
This equation is easily derived, if the internal porosity is assumed to be small.
Measurements made in this investigation.-To compare with theoretical formulae,
measurements were made on uniform “sands” varying in shape from spherical glass beads to
angular crushed Pyrex, and in size from fine sand (about 0.1 mm. diameter) to coarse sand
(about 1.0 mm. diameter). Permeability was measured at different porosities from the most
loose to the most dense state of the sand, and plotted on a logarithmic scale against porosity.
Photographs of the grains of some of these sands are shown in Figs 5 to indicate their shape.
They were sprinkled on h-inch graph paper before being photographed, to indicate the scale.
The sands used are listed below, where references are made to the figures on which
permeability is plotted, and to photographs of the grains.
(a) Glass beads: sizes (in millimetres) 1.1-06, 0.7-W, 0.4-0.25, 0.25-0.12 (Figs 6, and 5 (a)).
(b) Crushed Pyrex: lying between B.S. sieves 14-25,25-52,52-100 (Figs 6 and 5 (f)).
(c) Ham River sand: lying between B.S. sieves 14-25, 25-52, 52-100, 100-200 (Fig.
5 (b)).
(d) Leighton Buzzard sand: lying between B.S. sieves l&25, 25-52.
(e) The following materials, all lying between B.S. sieves 25-52: Leighton Buzzard
sand, Ham River sand, Crushed marble, Crushed quartzite, Crushed Darley Dale sandstone,
Crushed basalt, Crushed Pyrex (Figs 7, and 5).
These sands, together with a number obtained in held studies, were measured at a range of
porosities from the most loose to the most dense condition obtainable. Several other sands
were measured at a single porosity. Relevant information is quoted from measurements which
are not plotted.

Permeability of Glass beads and Crushed Pyrex plotted against Porosity (Broken lines calculated
from Kozeny’s formula)
COMPARISON OF EXPERIMENTAL MEASUREMENTS WITH THEORY
Hazen’s formula.--A comparison was first made with Hazen’s formula:
k = 100 d102
This formula is very useful for obtaining a rough approximation to the permeability of sand,
although the porosity of the bed, and the shape and size distribution of the grains are neglected.
Considering these variables in turn, it can be seen from Fig. 6 that extreme variations in porosity
can cause 3:1 variations in permeability, but in nature variations in porosity are much smaller.
Fig. 7 shows that variations in particle shape between the extremes of the rounded Leighton
Buzzard sand and the very angular crushed Pyrex do not cause very large changes in
permeability. At a given relative density, the distribution curve does not affect the permeability
very much either. Two sands were mixed to give the same d10 size, but with uniformity
coefficients d60: d10 of 1.5 and 5.0. It was found that permeabilities differed by only 30 per cent
when the sands were in their most loose condition, although at a given porosity permeabilities
were in the ratio 3:1. Sands with irregular grading-curves, of course, might well disagree with
Hazen’s formula.
The “relative density” at which k = 100 d102 was calculated for sands tested in this investigation.
Relative density DR is defined in this article in terms of the porosity n and the maximum and
minimum porosities nmax and nmin, as follows:

Relative density
Where :
Nmax= porosity of sand which has settled after piping.
Nmin, = minimum porosity obtainable by prolonged vibration.
The results are shown in Table 3.
Permeability of Sands, of grading 25-52, plotted against Intergranular porosity

Thus among these sands the permeability was equal to 100 d102 at medium densities for natural
sands, but was not equal to 100 d102 at any attainable porosity for angular crushed materials,
when the limiting sieve mesh sizes were in the ratio 2 : 1. , With more closely graded materials
the permeability was equal to 100 d102, in a looser condition.
Comparison of measurements on glass beads with theory.-With glass beads particle diameters
can be easily measured under the microscope, and specific surface calculated, and permeability
measurements fully compared with the more detailed formulae. The beads were mixed in a
damp condition before sampling, since dry beads are very mobile, and tend to run into groups
of beads of the same size. Diameters of a few hundred beads of each grade were measured, and
specific surfaces calculated from the relation:

The data were used to calculate other functions of gram size used in different formulae.
Terzaghi’s formula uses the “effective” grain size d10, and it is appropriate to substitute the
6
“mean spherical diameter” 𝑆 for d in Rose’s formula. This diameter was also used in Slichter’s
formula, although Slichter does not suggest any value of d for non-uniform spheres (Carman,
𝑐
1939 (b)). The constant 𝑛 in Terzaghi’s formula was taken as 800, which he recommends for
rounded sands. Permeabilities were calculated from these formulae at a porosity of 40 per cent,
and are compared with measured values in Table 4.

At a porosity of 40 per cent, the formulae of Rose and Kozeny gave values of permeability which
agreed much more closely with the experimental results than the values calculated from the
earlier formulae of Slichter and Terzaghi.
When the slopes of the curves relating permeability to porosity were compared, Kozeny’s
formula gave better agreement than any of the others. This conforms with the conclusions of
several authors (Cam-ran, Coulson, and others). Permeability calculated from Kozeny’s formula
is plotted in broken lines beside the experimental data in Fig. 6. These lines are seen to have
approximately the same slope as the full lines through the experimental points.
Fig. 1 shows that all the other formulae give a much smaller slope.
Comparison of measurements on sands with theory.-The specific surface of sands cannot easily
be measured by independent methods, and quantitative comparison of measurements with
theoretical formulae was restricted to the slopes of the curves relating permeability to porosity.
Experimental curves such as those plotted in Figs 6 and 7 could be compared directly with the
theoretical curves of Fig. 1, but it was thought that a more detailed comparison of the slopes
could be made by determining them explicitly. Slopes were therefore measured on all the
available experimental plots of permeability and porosity, using, for accuracy, the maximum
possible range of porosity. They were expressed as the percentage increase in permeability for
unit increase in porosity, “unit” increase being a change from n per cent to n + 1 per cent.
These “slopes” are given in Table 5 and plotted in Fig. 8 at the mean porosity at which the slope
was measured, and slopes given by the theoretical formulae are also plotted.
Table 5. Slope of permeability/porosity curves

Slope is expressed as the percentage increase in permeability for unit change in porosity.
The main conclusion to be drawn is that the experimental slopes agreed more closely with
Kozeny’s than with the other formulae. Experimental slopes were, however, slightly greater ;
and were approximately the same for all the materials tested, whatever their angularity and
porosity, amounting to an average percentage increase of 12.6 per cent for unit increase in
porosity.
Over the limited range of these experiments, therefore, the data may be represented by the
simple empirical formula :

Which agrees with Kozeny’s formula at a porosity of 40 per cent and has the same slope as the
experimental curves. There is little difference between this formula and Kozeny’s from 30 per
cent to 45 per cent porosity, but the empirical formula may make computation easier. Ks2 is
plotted against porosity in Fig. 6 for both formulae.
The slight deviation of the experimental results from Kozeny’s formula means that specific
surface calculated from permeability decreases slightly with increase in porosity ; at a porosity
of 50 per cent, for instance, calculated specific surface changes by 1 per cent for unit change in
porosity. Such changes in calculated specific surface were noted by Lea and Nurse (1939, 1947)
in experiments on the air-permeability of fine powders. Coulson (1949) has studied deviations
from Kozeny’s formula, attributed to changes in Kozeny’s constant Kl with porosity, which are
obtained with regular shaped particles (cubes, cylinders, prisms, etc.).
He obtained curved plots of K1 against porosity. But the above empirical formula seems
reasonably accurate for irregular particles such as sands.
Calculation of specific surface and angularity.-Quantitative comparison of measured and
calculated specific surfaces was precluded by the difficulty of measuring the specific surface by
independent methods. Specific surface is a function of grain size and grain shape; it is well
established that permeability depends on the square of the grain size, but the effect of grain
shape can only be studied qualitatively. Such a study is, however, quite useful.

Percentage increase in Permeability for 1 per cent change in Porosity


A comparison is made in Fig. 7 of the permeabilities of a number of sands, all lying between B.S.
sieves 25-52, and fairly uniformly distributed between the limiting sieves, but differing in grain
shape. The shapes of some of the grams are shown in Figs 5, and it can be seen that the more
angular sands, of greater specific surface, are less permeable at a given porosity than the
rounded sands. The practical point should, however, be noted that angular sands exist at larger
porosities than rounded sands and are actually more permeable at a given relative density. For
instance, although the Leighton Buzzard sand would be about 2.5 times as permeable as the
crushed basalt at a given porosity (assuming that both materials could exist at the same
porosity), it is only about half as permeable at a given relative density.
A numerical measure of the angularity of the particles was obtained from the permeability data.
This is the ratio of the specific surface of the material, calculated from permeability, to the
specific surface of uniform spheres of the same sieve size. Here it is called the “angularity factor”
of the sand. Robertson and Emodi (1943) called this ratio the “ oefficient of rugosity.”
The specific surface of spheres uniformly distributed in size between the mesh sizes dx y dy of
adjacent sieves is given by

𝑑𝑥
which is accurate to 2 per cent if 𝑑𝑦
is not greater than 2 and the size distribution by weight is
logarithmically uniform. Specific surfaces of spheres lying between standard B.S. sieves is given
in Table 6.
Specific surface of spheres lying between given sieve sizes

The specific surface of sand grains uniformly distributed between these sieves is then S, = f Si.
S, can be calculated from permeability, and hence f found. For example:-
Specific surface of crushed marble between B.S. sieves 25-52: 200 sq. cm/cc
Specific surface of uniform spheres between B.S. sieves 25-52: 143 sq. cm/cc.
200
:. Angularity factor of crushed marble (143) =1.4.

This method of calculating the angularity factor can readily be extended to graded sands.
Permeability measurements were used to calculate the specific surfaces and angularity factors
given in Table 7.

Both Kozeny’s and. the empirical formula were used, so that a choice can be made.
Kozeny’s formula is well established but has the disadvantages that calculated specific surfaces
and angularity factors vary slightly with porosity and that calculation is slightly more difficult.
There is, however, little difference between the formulae; with natural sands the difference is
negligible.
The most angular natural sand tested was a glacial sand from Lough Neagh with a factor of 1.4,
and factors for the crushed materials ranged from 1.4 to 1.8. The appearance of the grains (Figs
5) may be compared with the numerical measure of the angularity. This method of measuring
angularity may be of use in the general study of fragmental materials.
ESTIMATION OF PERMEABILITY
Description of method.-It is useful to be able to estimate the permeability of sand from simple
tests, either to check measurements or ‘as an alternative to measurement when large numbers
of permeabilities are required. Hazen’s formula is frequently useful for these purposes, but it is
shown in Fig. 9 that it may lead to an error of a factor of 2 either way, and greater accuracy is
frequently required.
Data from various published papers showing relation between Permeability and Hazen’s
Effective size
Kozeny’s formula allows the permeability of a material to be calculated quite accurately from
its specific surface and porosity. The foregoing calculations show that if the angularity factor of
a sand is known, its specific surface can be calculated from its sieve analysis. A visual estimate
of the angularity factor seems quite good enough for the standard of accuracy required. The
grains may be examined under a microscope or magnifying glass and compared with those
photographed in Figs 5, or with grains of sand whose angularity factor has previously been
determined; or factors may be estimated roughly from:
Rounded sand f=1.1
Sand of medium angularity f = 1.25
Angular sand f =1.4
Then, if sieve analysis shows the sand to have fractions of the total mass X1, X2…Xn, retained on
different sieves, the specific surface of the sand may be obtained from the equation : Specific
surface of sand S, = f(x1S1+ X2S2 …+ XnSn)
Provided the specific gravity and angularity are the same throughout. S1,S2…Sn, S, are the
specific surfaces of spheres uniformly distributed within the corresponding sieved fractions, and
these are given in Table 6 for commonly used sieves.
Permeability can then be calculated from the specific surface and porosity, using:

(a) Kozeny’s formula:


(b) the empirical formula:
These formulae are in c.g.s. units, and when S is substituted in sq. cm/cc., k is obtained in
cm/sec. per unit hydraulic gradient. n is expressed as a fraction, not a percentage. Both formulae
are plotted in Fig. 6.
Comparison of calculated and laboratory measured permeabilities.-As examples of the use of
this method, permeabilities were calculated using Kozeny’s formula for three sands, from Lough
Neagh, Newquay Fistral beach, and Christchurch. Angularity factors were estimated by visual
examination, and the sands were sieved; after the calculations had been made the permeability
was measured at different porosities. Measured and calculated values are plotted in Fig. 10, and
it can be seen that agreement was good.
To illustrate the method, the calculation of specific surface for Lough Neagh sand is shown in
Table 8. The angularity factor was estimated to be 1.35.

The calculated specific surface is 110.6 sq. cm/cc. The curve of Fig. 10 is obtained by substituting
this value in Kozeny’s formula. If the empirical formula is used to calculate permeability, good
agreement is obtained when an angularity factor of 1.4 is used.
It is not necessary to estimate the angularity with great accuracy in order to obtain a reasonably
accurate estimate of permeability. If the correct value of the angularity factor is 1.35, the
following errors in permeability are introduced by using different factors:

Lough Neagh sand is obviously less angular than crushed basalt, for angularity factor 1.6, and
more angular than Ham River sand, of angularity factor 1.1. Poor estimates of permeability are
unlikely to cause errors of more than about f 20 per cent in permeability.

The rather variable results obtained with Newquay Fistral sand (Fig. 10) are interesting. This
sand consisted largely of flat fragments of broken shells. The “first series” are readings taken
when the sand was being made denser in stages, after it had be& placed by pouring under
vacuum. The sand was then loosened by gentle piping, and it was found that the permeability
was higher in the second series of measurements, since the particles were more nearly vertical.
Permeability was further increased in a third series. These differences are due to different
orientations of the particles to the general direction of flow, and is a factor which has not been
considered, but which warrants close attention in the case of shelly and micaceous sands.
Some old measurements of permeability on various sands were compared with calculated
permeabilities. A standard value of 1.25 was used for the angularity factor, since the samples
were not available for examination. The results, given in Table 9, show that fair agreement was
obtained. The error of - 30 per cent obtained with the Croxley Green sand is understandable.
since it is recorded as a rounded sand.

Sand from Lough Neagh, Newquay Fistral, Christchurch, Permeability/Porosity (Broken lines
calculated from Kozeny’s formula)
Comparison of calculated and field measured permeabilities.-The permeability of considerable
numbers of sand samples was measured in the field during a site investigation, and the
opportunity was taken to compare these results with calculated permeabilities.
The field samples were obtained above the water level in sample tubes and were wetted up
slowly from the base. Permeability was then measured using a falling head technique.
Porosities were measured and the sand sieved, and permeabilities were calculated using
Koweny’s formula, and an angularity factor of 1.25, since the sand was of medium angularity.
It was found that measured permeabilities on “ clean ” sands varied between 40 per cent and
80 per cent of the calculated values, the average ratio being 60 per cent. But in one case
(Christchurch sand, Fig. 10) the sand was de-aired and the permeability was measured in the
laboratory by the methods described earlier, and good agreement was obtained between
measured and calculated permeabilities. It was concluded that the calculated values were closer
to the saturated permeability than the measured values obtained in the field.
Some of the sands were dirty, and measured permeabilities were very much less than calculated
permeabilities; this was found to be true even with a laboratory measured permeability on a
dirty sand. Calculated permeabilities were unreliable when more than 5 per cent of the material
passed the No. 200 B.S. sieve. This method of calculating permeability is therefore restricted to
clean sands. The same restriction was found to apply td Hazen’s formula.
CONCLUSIONS
It is shown that Kozeny’s formula:
Agreed better than other published formulae with a series of careful measurements of
permeability. Over the limited range of these experiments, however, the experimental of results
agreed even more closely with the following empirical formula, which differs very little from
Kozeny’s, but which makes computation slightly easier:

Either of these formulae may be used to calculate the permeability of sand with fair accuracy, and
are more accurate than Hazen’s formula k = 100d102, which, though useful, may lead to an error of
a factor of 2 either way. All these formulae are restricted to clean sands.

ACKNOWLEDGEMENTS
The Author is indebted to Dr L. F. Cooling, under whose general direction this work was carried
out, and to his colleagues, for their interest and encouragement. The work forms part of the
general research programme of the Building Research Station of the Department of Scientific
and Industrial Research and the article is published by permission of the Director of Building
Research.
Crown copyright of this article is reserved.
REFERESCES
.\NDERSON, D. G., 1940. Void volume and flow resistance of beds of particles. Ph.D. thesis.
Columbia University.
CARMAN, P. C.. 1938. The Determination of thc’specific Surface of Powders.-I. J. Sot. Chem. I&.
Trans. 57 : 225.
CARMAN, P. C., 1939. The Determination of the Specific Surface of Powders.--2. 58: 2. J. Sot.
Chem. Id Traws.
CARMAN, P. C., 1939. Permeability of saturated sands, soils, and clays. J. Aguic. Sci. xxix, 11 :
262. COULSON, J. M., 1949.streamline flow.
The Flow of Fluids through Granular beds : Effect of particle shape and voids in Trans. Instn
Chem. Engrs, 27 : 237.
HAZEN, A., 1892. Some physical properties of sands and gravels with special reference to their
use in filtration. 24th Annual Report, Massachusetts State Board of Health.
KOZENY, J., 1927. Uber Kapillare Leitung des Wassers in Boden. Rev. Wein Akad. 136a-271.
LEA, F. M., and NURSE, R. W., 1939. The Specific Surface of Fine Powders. 58: 277. J. Sot. Ckzm.
Innd. Trays.
LEA, F. M., and NURSE, R. W., 1947. Permeability methods of fineness measurement.
Symposium on Particle Size Analysis. Su$q5. to Trans. Sot. Chem. Ind. 25 : 47.
MAVIS, F. T., and WILSEY, E. F., 1936. A Study of the Permeability of sand. Study Bulletin No. 7.
Universitv of Iowa.
ROBERTSON, 8. H. S., and EM~DI, B. S., 1943. Rugosity of Granular solids. Nature. 152 : 539.
ROSE, H. E., 1950. Fluid flow through beds of granular material. Some Aspects of Fluid Flow.
Proc. Conf. Instn. Phys. : 136.
SLICHTER, C. S., 1899. Theoretical investigation of the motion of ground waters. 19th Annual
Reoort, U.S. Geological Survey. 2 : 305.
TAYLOR, I>. W.. 1948. Fundamentals of soil mechanics, p. 112. Chapman and Hall, Loxdon.
TERZA~HI, K., 1925. Erdbaumechanik, p. 120. Deuticke, Vienna.

You might also like