Professional Documents
Culture Documents
discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/252135766
CITATIONS
READS
25
105
4 authors, including:
J.F. Donoghue
University of Central Florida
65 PUBLICATIONS 548 CITATIONS
SEE PROFILE
All content following this page was uploaded by J.F. Donoghue on 18 December 2013.
The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.
Geological Investigations Section, The Florida Geological Survey, 903 W. Tennessee St., Tallahassee, Florida 32304-7700, U.S.A.
2 Department of Geosciences, The Florida State University, Tallahassee, Florida 32306, U.S.A.
e-mail: Jim.Balsillie@dep.state.fl.us
INTRODUCTION
Sediment grain-size data are used in many geologic and engineering applications. The
grain-size distribution of a sediment sample provides information on the relative energy
level present in the environment under which the sediment was transported and deposited.
It also contributes practical information of engineering importance. For instance, in beach
restoration projects, the mean grain size and standard deviation using composite statistics
(see Balsillie and Tanner 1999) of the grain-size distribution are used in calculations to
determine the suitability of local sand deposits as borrow material for nourishment, as
well as in predicting the maintenance renourishment needs of a project (e.g. Krumbein
1957; James 1975; Hobson 1977; U. S. Army 1984).
Standard methods of quantifying sediment textural data employ statistical measures (mean, standard deviation, skewness, kurtosis), grain-size distribution diagrams, and frequency and cumulative frequency probability plots. Probability considerations are based on the underlying concept that sediments conform to the Normal or Gaussian density distribution function. The Gaussian distribution plays a
central role in all of statistics. It is, perhaps, the most ubiquitous distribution utilized
in all the sciences. Results from statistical application of the Gaussian distribution
are best plotted using arithmetic probability paper (APP). The usefulness of APP is,
among informed practitioners, of unequaled practical importance when compared to
other plotting options. The graph paper has one arithmetic axis that represents quan-
tifiable data (e.g., sediment grain size) plotted against the cumulative percent occurrence (or cumulative probability), which is a nonlinear axis. Typically, the arithmetic axis is the horizontal axis representing grain size. The nonlinear (Gaussian)
axis is the vertical axis and represents the cumulative percent weight from the sieve
fractions. The cumulative probability distribution of the paper has commonly been
termed the normal distribution. It should not be so designated, however, because
normal is applied in too many applications. Johann Carl Friedrich Gauss (1777
1855) was its originator, and it should properly be referred to as the Gaussian
distribution or Gaussian probability density distribution (GPDD). APP, constructed
so that the ogive (S-shaped curve on arithmetic paper) plots as a straight line, was
developed in 1913 by Hazen (1914), and is acknowledged as a milestone in statistical graphic applications (http://www.math.yorku.ca/SCS/Gallery/milestone/
sec7.html). There are three fundamental properties of the GPPD that when plotted
on APP require understanding.
First, if the quantified data, in general, conform to the Gaussian distribution, any
one sample shall, if it is precisely Gaussian, plot as a straight line on APP. Second,
are the natural data of any one sample precisely equivalent to the GPDD? Most are
not, nor would we wish them so, for it is the deviation from the Gaussian that tells
us something about the sample. This is especially true for sediment grain-size distributions. Third, natural data plotted on APP may, in many applications, be made
up of several straight-line segments. These segments are often attributable to some
identifiable natural cause or process. For instance, Tanner (1991) found, by analysis
of over 11,000 sediment samples from all types of environments, that the geometry
of straight-line segments for sediment distributions can definitively identify whether
the latest transportationdepositional history of the sediment sample was due to
eolian, littoral, fluvial, settling processes, etc., or even combinations of processes
(Balsillie 1995). Balsillie (1999) found that they may be related to storm- and hurricane-induced erosion.
Arithmetic-arithmetic, logarithmic-logarithmic, arithmetic-logarithmic (semilogarithmic), and other plotting media have been used to display analytical GPDD results. The
above three properties will invariably not be evident except where APP is used.
PLOTTING EQUATION
930
ordinates. The probability plot is, however, with very few exceptions not available
in standard statistical packages. Very few computer statistical applications allow for
plotting on APP because the cumulative percent occurrence (or cumulative probability) axis is nonlinear and, ostensibly, intractable.
If one inspects APP, it becomes apparent that the cumulative percent axis is symmetrical about the 50th percentile (or 0.5 probability). This makes quantification somewhat
simpler, because only half requires quantification, which can then be applied to the other
half. A mathematical investigation was undertaken by one author (JHB) to discover a
sufficiently precise quantifying relationship. The best representation was found using a
fifth-order transcendental polynomial. The equation is given by
Y5
Y9 5 49.99 2 13.41 t 2
(1a)
2.30753 1 2.27061t
1.0 1 0.9929t 1 0.04471t 2
(2a)
in which
t5
1 0.00282(ln X) 5 }
! ln (0.01X)
1
(2b)
Either equation for determining Y9 will suffice for plotting purposes, although in
this work we use Equation 1.
The goodness of fit of the plotting position equation (Equation 1) is illustrated in
Figure 1, which has a Pearson productmoment correlation coefficient exceeding
0.9999. On the average, the absolute difference between actual and predicted percentiles is 0.049%, with a maximum difference of 0.179%. Actual plotting positions
were determined by physically measuring the probability axis on a piece of probability paper. Hence, there is to be expected some measurement error in the actual
data. This does not compound the error, and the statistically fitted equation is sound.
931
In fact, the differences shown in Figure 1 lie well within the width of a fine pencil
line for an APP plot on letter-size paper. The predicted percentiles, therefore, successfully represent Gaussian percentiles for plotting purposes.
AN APPLIED EXAMPLE
The method is demonstrated using a granulometric example. Figure 2 is an EXCEL application where an APP plot has been constructed. There is, however, one
significant note. Equation 1 is applicable where the Gaussian percent, PG, is in the
range 0.01 , PG , 99.99, because these are the limits specified on standard APP.
The lower limit should not pose a problem. However, if the last arithmetic percent
is 99.99 and Equation 1 is applied it will result in a Gaussian value of 99.07%,
which is incorrect. That is, 99.99% should also be the Gaussian value. Conversely,
if the last arithmetic percent is 99.998, the Gaussian value will be 99.9896% using
Equation 1. If the last arithmetic percent is less than 99.99%, which might occur
because of sample truncation or filtering (Tanner 1964), then the result derived from
the equation is correct.
It was noted above that APP allows for straight-line segment analysis. A surface
sediment grab sample from Cedar Key, Florida (Fig. 2) illustrates the usefulness of
APP. It is apparent from Figure 2 that the arithmetic-arithmetic plot (triangles) shows
a typical S-shaped curve (ogive) with no straight-line segments. For the APP plot
(circles), one can discern relative to the straight-line segment representing the central
portion of the distribution, several straight-line segments. The segment from 0.325
w to 1.75 w, because of the relative low angle of departure from the central segment,
identifies a transportationdepositional history possibly due to wave activity. The
segment from 20.625 w to 0.325 w, because of its larger departure of angle from
932
constraints are specified, the first two columns of the left-hand table are automatically constructed by the program. The only other input required are the weights (i.e.,
mass in grams) for each specified sieve interval in the third column. All output
automatically appears upon data entry. This includes moment measures for the mean,
standard deviation, skewness, kurtosis, relative dispersion, and the frequency plot.
The arithmetic probability plot is also automatically generated in chart ACUMPLOT
(Fig. 4). Users should not attempt to change the vertical size of the chart within
EXCEL unless they are prepared to alter the position of ordinate values. Users can
and may need to change scale values for the abscissa. One can, however, export the
EXCEL plot to other applications and change its size without corrupting the plot.
Worksheet DATA currently allows for input of up to 41 data points to the table.
Note, however, that this application can be used for data other than granulometric
purposes; auditing information, column headers for the table, and plot axes simply
would require respecification.
CONCLUSION
This paper has presented an equation for identifying plotting positions of Gaussian
percentiles (or probabilities). Using the equation, a programmed EXCEL application
template has been made publicly available, whereby a complete granulometric analysis including data listing, moment measure calculations, and frequency and cumulative APP plots, is automatically produced. The EXCEL application includes
PROCESSING UNDISTURBED MARINE SAND SEDIMENTS AND RECONSTRUCTING FABRIC AND POROMETRY
two programmed spreadsheets, one for sieved data (GRANPLOTS) for classified
data, and the other for settling tube data (GRANPLOTT) for unclassified data.
ACKNOWLEDGMENT
We thank Walter Schmidt, State Geologist, Florida Geological Survey, for his
valuable editorial comments on the manuscript.
REFERENCES
BALSILLIE, J.H., 1995, William F. Tanner on environmental clastic granulometry: Florida Geological Survey, Special Publication 40, 145 p.
BALSILLIE, J.H., 1999, Volumetric beach and coast erosion due to storm and hurricane impact:
Florida Geological Survey, Open File Report 78, 37 p.
BALSILLIE, J.H., AND TANNER, W.F., 1999, Suite versus composite statistics: Sedimentary Geology, v. 125, p. 225234.
HAZEN, A., 1914, Storage to be provided in impounding reservoirs for municipal water supply:
American Society of Civil Engineers, Transactions, v. 77, p. 15291669.
933
HOBSON, R.D., 1977, Review of design elements for beach-fill evaluation: U.S. Army, Coastal
Engineering Research Center, Technical Paper 776, 51 p.
JAMES, W.R., 1975, Techniques in evaluating suitability of borrow material for beach nourishment: U.S. Army, Coastal Engineering Research Center, Technical Memorandum 60, 81 p.
KRUMBEIN, W.C., 1957, A method for specification of sand for beach fills: U.S. Army, Beach
Erosion Board, Technical Memorandum 102, 82 p.
TANNER, W.F., 1964, Modification of sediment size distributions: Journal of Sedimentary Petrology, v. 34, p. 156164.
TANNER, W.F., 1991, Suite statistics: the hydrodynamic evolution of the sediment pool, in
Syvitski, J.P.M., ed., Principles, Methods and Application of Particle Size Analysis: Cambridge, U.K., Cambridge University Press, p. 225236.
U.S. ARMY, 1984, Shore Protection Manual: U.S. Army, Corps of Engineers, Coastal Engineering Research Center, Washington, D.C., 2 vols., 1272 p.
ZELEN, M., AND SEVERO, N.C., 1965, Probability functions, in Abramowitz, M., and Stegun, I.,
eds., Handbook of Mathematical Functions, with Formulas, Graphs, and Mathematical Tables: Washington, D.C., U.S. Department of Commerce, National Bureau of Standards, p.
925995.
Received 18 September 2001; accepted 4 April 2002.
ABSTRACT: Study of marine sediment pore fluid pathways and porometry requires careful analysis of the fabric of undisturbed sediment samples. A
novel solution to the preservation of the interstitial organic material and the in
situ fabric or sedimentary structure is the application, with little modification,
of well-established biological techniques employing agar infiltration. The solidified agar preserves fabric during subsequent epoxy impregnation. Once impregnated, porosity of the samples can be measured using image analysis of
polished surfaces of the microfabric and/or a gravimetricvolumetric technique.
Porosity was about 10% higher with image analysis, apparently because of
problems in visualizing carbonates and edges of grains. Tortuosity was measured as a function of pathlength ratios taken in stacked planes of microfabric
images. The technique allowed us to detect variability in directional tortuosity
as a function of orthogonal pathlength ratios. Three-dimensional stacking of
digitally acquired wireframe images of sequential planes through the microfabric allows visualization of long continuous pores, some with 2.6 mm length.
knife. Later, Jim (1985) reviewed similar techniques for embedding sediment samples and Lavoie et al. (1996) described a mini-core technique for subsampling finegrained sediment.
We solved the sampling-disturbance problem for noncohesive sediments (sand)
by gently impregnating the sand in the field with liquid agar (very low viscosity of
1.002 centipoise at 208C comparable to that of water) using an infiltration pressure
of about 0.254 kPa (0.037 psi). Agar is a polysaccharide agarose in which the
hydrogen bonds of the hydroxyl groups cross link as the solution cools to room
temperature, thus solidifying and stabilizing the sand particles. The particles were
then held together sufficiently for subsequent infiltration with a series of miscible
fluids and epoxy resin. Once the resin was cured, a series of polished sand surfaces
were successively precision ground from the surface of the epoxy embedded specimens so that the two-dimensional microfabric of each surface could be imaged and
studied. The purpose of this paper is to describe our procedure and techniques in
detail with examples of the data and observations.
MATERIALS AND METHODS
INTRODUCTION
The microfabric of sediment is defined by the three-dimensional particle arrangements and particle associations of the individual grains of a sedimentary deposit.
Study of the microfabric and porometry provides important information on fundamental sediment properties including porosity, permeability, pore fluid pathways,
sediment bulk density, acoustic properties, and stressstrain behavior (Bennett et al.
1989; Bennett et al. 1996). The study of microfabric of undisturbed marine sand has
been confounded by the degree of difficulty encountered in both collecting and
processing of sediment without altering its in situ configuration. Efforts to examine
the microfabric of marine sands and particle associations face the problems of preserving the three-dimensional configuration of the microenvironment and avoiding
major disturbances of the microorganisms and extracellular materials intimately associated with the pore structure also present in fine-grained sediment (Bennett et al.
1999).
Perhaps the earliest work and techniques of embedding fine-grained marine sediment samples with Spurrs epoxy resin for microfabric study were developed by
Bennett (1976) and Bennett et al. (1977). These techniques were later expanded for
the study of organo-clay microfabric studies (Baerwald et al. 1991; Bennett et al.
1999). These samples were critical-point dried and impregnated under vacuum with
the low-viscosity Spurrs epoxy resin for later ultra-thin sectioning with a diamond