Paleocurrent Data Analysis

Contents
1
Introduction
2
Collection
of
Data
2.1
Sampling
2.2
Correction
of
Data
for
Tectonic
Tilt
3
Statistical
and
Graphical
Implications
3.1
Representation
of
Data
3.1.1
Spoke
Diagrams
3.1.2
Rose
Diagrams
3.1.3
Other
Considerations
3.2
Statistical
treatment
of
directional
data
3.2.1
Scatter
of
directions
3.3
Analysis
and
Interpretation
of
data
3.3.1
Mode
of
pattern
3.4
Depositional
environment
and
Dispersal
Patterns
4
Discussion
on
Directional
Variance
and
Vector
Magnitude
4.1
The
Sedimentary
Structure
Hierarchy
4.2
The
Problem
of
Vector
Weight
4.3 What is the relative importance of azimuth readings taken on the various types of
structure ?

List of Figures
1 Two outcrop patterns and grid designs. After [Potter and Pettijohn, 1977]
2 Procedure
for
correcting
planar
cross-bedding
data
for
tectonic
tilt
3 Procedure
for
correcting
linear
data
4 Radial
Spoke
Diagrams
a Primary
Current
Lineation
b Current
Ripples
6 Rose
Diagrams
7 Example
of
the
derivation
of
a
resultant
vector
8 Diagrams of the main types of paleocurrent azimuthal patterns. From Selley [1968]
9 The sedimentary structure hierarchy

1 Introduction
The objectives of paleocurrent research are the identification, description and interpretation
of the current patterns of the past.[Potter and Pettijohn, 1977]. A knowledge of the
palaeocurrents gives vital information on palaeogeography, sandbody geometry and sediment
provenance, and in addition certain palaeocurrent patterns are restricted to a particular
depositional environment.
Some sedimentary structures are vectorial, i.e. theygive the direction in which the current was
flowing; cross-stratification, asymmetric ripples, flute marks and imbrication are structures of
this type. Other structures give the trend or line of direction of the current; these include
parting lineation, preferred orientation of grains, pebbles and fossils symmetrical ripples and
groove marks[Tucker, 1991].

2 Collection of Data
Paleocurrent daa should be carefully documented in the field, including the following
information, according to [Miall, 2003]:
1.
2.
3.
4.

location and (if relevant) precise position in a stratigraphic section;

structure type;
indicated current direction;
scale of structure (thickness of crossbed, depth and width of channel, mean or
maximum clast size in imbricate gravels); and
5. local structural dip.
Current directions should be measured to the nearest 5with a magnetic or sun-compass and
corrected to true north wherever necessary. In the case of parting lineation and tool markings,
the correct orientation of two possibilities at 180can usually be identified by referring to
other types of current structures nearby. Measurement accuracy greater than 5is difficult to
achieve and is, in any case, rarely important. Difficulties in field measurement commonly

arise because of incomplete exposure of crossbed sets. For example, two- dimensional
exposures in flat outcrop faces cannot provide precise orientation information.
Indicated current directions may need to be corrected for structural dip and fold plunge,
otherwise significant directional distortions may result. For linear structures such as parting
lineation or sole markings, structural dips as high as 30can be safely ignored, as they result
in errors of less than 4. However, the fore set dip orientation of crossbedding is significantly
affected by structural dip, and should be corrected wherever the structural dip exceeds 10.
2.1 Sampling
Collecting paleocurrent data over a large area requires sampling. The objective of sampling is
to determine whether or not an anisotropic regional pattern exists and to describe it if it does.
In connection with the sampling problem, it is useful to distinguish between population
parameters, say the mean azimuth of all flute marks in a formation, and estimates of such
parameters based on a sample, generally a small one. Usually in paleocurrent studies,
population parameters can never be evaluated but can only be estimated by sampling.
Population parameters are denoted by capital letters and their sample estimates by lower case
ones. In general, we may assume a close identity between the orientation of directional
structures in outcrop and the orientation of those not available for measurement. For
example, the orientation of a directional structure has little or no control on the presence or
absence of an outcrop in which the structure occurs. Most desirable is some appreciable" twodimensional" coverage; i.e., exposure over an appreciable area rather than exposure in a
straight, narrow outcrop belt. Two dimensional coverage is obtained in well-dissected areas
of flat-lying sediments or in areas of outcrops repeated by folding, or in areas of basins with
marginal exposure along several sides. If the current system diverges, converges, or is
otherwise irregular, a narrow linear outcrop belt is usually inadequate to delineate such
irregularities[Potter and Pettijohn, 1977]. For either poorly-exposed formations or those
restricted to small areas, it may be possible and necessary to visit every known outcrop.
Commonly, however, many more outcrops are available than it is either possible or necessary
to visit. The investigator, therefore, usually studies only a selected fraction of all
outcrops[Potter and Pettijohn, 1977].
Coverage is facilitated by use of a sampling grid. A grid is useful because:
it assures distribution of sampling effort across an outcrop belt, so that every part of
the formation receives comparable attention
it can be useful in analysis and presentation of the collected data.

Figure 1: Two outcrop patterns and grid designs. After [Potter and Pettijohn, 1977]

If the trend itself is important, 25 readings per sample station is commonly regarded as the
minimum necessary for statistically significant small samples. However, the same or fewer
readings plotted in map or section form can yield a great deal of environmental detail,
whether or not their mean direction turns out to be statistically significant. Several hundred or
a few thousand readings may be necessary for a thorough analysis of a complete basin
[Miall, 2003].
2.2 Correction of Data for Tectonic Tilt
When the rocks have undergone considerable tectonic tilting, it may be necessary to
reorientate the directional measure ments by removing the effects of the tilting and restoring
the original bedding to horizontal. In doing this, the structures within the beds that act as the
palaeocurrent indicators will themselves be restored (rotated) to their original attitude at the
time of deposition[Collinson et al., 2006].

Figure 2: Procedure for correcting planar cross-bedding data for tectonic tilt in order to
determine original palaeocurrent direction. Bedding and cross bedding are plotted as poles to
planes on a lower hemisphere stereogram. In part (b) the dotted circles represent the original
placement of the poles to the bedding and to the cross bedding on the steronet prior to
calculation; the dashed circles represent a position midway through the calculation, and the
filled black circles represent the position of the poles following completion of the rotation
calculation. After Collinson et al. [2006]

Thus, from Collinson et al. [2006] Restoration is performed using the following procedures
In order to restore cross beds to their original attitude, it is necessary to plot and manipulate
the data on a stereogram or in a dedicated computer programme. This requires the magnitude
and direction of dip, both of the foresets as they now occur, and of the overall sequence (i.e.
local tectonic dip). The procedure outlined here applies only if fold plunge is negligible. A
more complex procedure is needed for plunging folds. Plot poles (normals) to both the
foresets and the bedding on a stereographic projection (Fig. 2a). Rotate the points until the
normal to the bedding lies on a great circle of the projection (Fig. 2a,b). To restore the beds to
horizontal, this point must be moved to the centre of the projection. As that shift is carried
out, the normal to the foreset must be moved the same angular distance along the small circle
upon which it lies (Fig. 2b). The new position of this point shows the normal to the foreset at
the time of deposition and this can be converted back to a direction and magnitude of dip
(Fig. 2c). This direction (foreset azimuth) may then be used as an indicator of palaeocurrent
direction.

Figure 3: Procedure for correcting linear data (e.g. groove marks) for tectonic tilt in order to
determine original palaeocurrent direction or trend. For illustrative purposes, bedding is
plotted both as a plane and as a pole to that plane on a lower-hemisphere stereogram. Linear
data is plotted as a plunge and a plunge direction (azimuth). After Collinson et al. [2006]

For linear data in steeply dipping beds, plot on the stereo gram the attitude of the lineation in
space (Fig. 3a) and rotate both the normal to bedding and the lineation as described above
(Fig. 3b). This restores both the bedding and the lineation upon the bedding back to their
original atti tude prior to tectonic tilting. The orientation of the restored lineation may then be
used to indicate palaeocurrent direction (3c).

3 Statistical and Graphical Implications

A variety of statistical data-reduction and data-display techniques is available for
paleocurrent work. The commonest approach is to group data into subsets according to
stratigraphic or areal distribution criteria, display them visually in current-rose diagrams, and
calculate their mean and standard deviation (or variance)[Miall, 2003].
The method of grouping the data into subsets has an important bearing on the interpretations
to be made from them
3.1 Representation of Data
Once directional measurements are restored to their original orientation, it is usually helpful
to display them graphically.The method chosen usually depends upon the quantity of data and
the variety of structures from which they were collected.

Compilation inevitably leads to some loss of information; in particular the distribution of

various directions, both laterally and vertically, within the sampled sequence. Compiling
directional data is a useful way of visualizing flow patterns, but it is no substitute for relating
directions to specific structures in a measured section when the aim of the exercise is to
support environmental interpretation.
Where compilation is carried out, it is important to produce plots that clearly distinguish the
types of sedimentary structure from which they are measured. This can be done either by
producing separate plots for each type of structure or by using clearly distinguished symbols,
colours or designs for each type of structure on a combined plot. It is also important to bear in
mind that some structures can be recorded as a single direction to which flow is directed,
whereas others give only a trend along which flow could have been in either direction.
Examples of the second group must be shown as double-ended lines or sectors in any
display[Collinson et al., 2006].

Figure 4: Current directions presented as a radial spoke diagram.Note how different types of
structure are separated and how those structures that give only trend and not sense of
movement are plotted as double-ended lines. After Collinson et al. [2006]
3.1.1 Spoke Diagrams

Where data are few and have been collected from only a few types of structure, it is often

convenient to plot each meas urement as a radiating line of fixed length on a circular spoke
diagram(Fig.4). Where both dip direction (azimuth) and dip magnitude (inclination) have
been recorded from cross- bed foresets, the plotting of poles (normals) to the foresets on a
stereogram may be preferred to plotting just the foreset azimuths.
3.1.2 Rose Diagrams

When data are more numerous, spoke diagrams become very cluttered and data are then
better grouped into classes in a circular histogram (rose diagram; Fig. 6). The class inter val
for such diagrams can be set at any size, although 10, 15, 20and 30are the most
commonly used, depending on the volume and spread of directions[Collinson et al., 2006].
With abundant data, smaller class intervals show the detailed structure of the population more
clearly
However, with few data, small class intervals may give a false impression of complexity. In
addition, the fact that rose diagrams usually employ a linear relationship between abundance
and radius can lead to exaggeration of the apparent importance of more abundant classes.
Modal classes may seem more abundant and well defined than they really are.
There have been attempts to overcome this by designing schemes in which abundance relates
to the area of the sectors, but these are more difficult to apply and are not widely used. Where
individual readings fall on a class boundary, it is good practice to allocate these equally either
side of the boundary.
Note that structures that yield only a palaeocurrent trend should be plotted as double-ended
sectors (Fig. 5a), whereas those that yield a sense of direction can be plotted as single-ended
sectors in the direction of transport (Fig. 5b).

Figure 6: Current directions presented as rose diagrams with 20 class intervals. Data from
different classes of structure are usually presented on separate diagrams, and structures that
give only trend and not sense of movement are plotted as double-ended sectors. The rose may
be scaled in terms of either a percentage of all observations (as here) or actual number of

observations. Arrows indicate the vector mean directions. After Collinson et al. [2006]

3.1.3 Other Considerations

According to [Wells, 2000] standard rose diagrams are a favourite method of depicting
orientations because of their ease of comprehension, but they are known to have two serious
problems.
First, arbitrary decisions about class width and starting position can dramatically alter the
resulting diagram, although the degree of variation has been underappreciated.
Second when rose diagrams are correctly scaled to the square root of the class frequency,
they can be awkward to evaluate Possible solutions that deserve consideration include:
1. Representing 1classes with rays of dots ("dot diagrams", allowing linear scaling), and
2. overcoming arbitrariness by combining all possible rose diagrams for a dataset into
one diagram
Solution 1 can work by plotting either one dot per datum ("corona dot diagrams") or one dot
for each value above or below the mean membership ("mean-deviation dot diagrams"). Both
plot types can show either raw or summed data. Drawing one-degree class-width diagrams
and summing (combining) all possible rose diagrams offers the only nonarbitrary versions of
orientation diagrams, but summing diagrams smoothes the data slightly, thereby additionally
weighting clusters of similar but not identical data points.
3.2 Statistical treatment of directional data

Figure 7: Example of the derivation of a resultant vector (syn. vector mean; large arrow) by
the summation of unit vectors (small numbered arrows). Each unit vector has the direction
shown and a magnitude (r ) of 1. Use Equations 2-31 to 2-35 to idemonstrate the calculation
of the direction and magnitude of the resultant vector. Note that the arithmetic mean would be

146 and clearly not representative of the three vectors.

Normally, directional data are collected in sets of N observations from some particular
deposit or associated structure. Like grain size data, a set of directional data will be
distributed about some mean direction and will have a range of variation about the mean
direction. Unlike grain size data, directional data cannot be treated by regular statistics
because each value in a set is a vector quantity and each vector has two components: each
observation in a set of directional data will have a direction () and a magnitude (normally
each vector has unit magnitude, i.e, its magnitude equals 1; such a vector is termed a unit
vector). The descriptive statistics are similar to those applied to scalar quantities (like grain
size) but their computation is fundamentally different. Without going into the details of vector
algebra this section provides an outline of the various statistics and their implications.
According to Curray [1956] for any distribution of directional data the mean is the resultant
vector (sometimes called the vector mean) formed by summing all of the unit vectors that
comprise the data set (e.g., Fig. 7 ). Once again, the resultant vector includes a direction ( )
and a magnitude (R):
(1)
(2)
where
(3)
and
(4)
and N is the number of observations in the data set. Note that directional data may be treated
as either individual measurements (i.e., as raw data) or grouped data (i.e., data are expressed
as frequency per class interval). In the case of ungrouped data ni is equal to 1 (unit magnitude
of each vector) and i represents each of the measured directions (from i = 1 to N) of the N
vectors in the data set. When data are grouped the quantities above are: n is ithe number of
observations in each class interval (from i = 1 to NC) and i is the mid-points of each class
interval (from i = 1 to NC) and NC is the number of class intervals (i.e., NC = 360/CI, where
CI is the class interval in degrees).
The magnitude of the resultant vector can also be expressed as a percentage (L) of the total
number of observations (or more specifically, as a percentage of the total length of the unit
vectors that comprise the data set):
(5)
The magnitude of the resultant vector, particularly when it is expressed as a L reflects the
amount of dispersion of the data (i.e., the degree of variation of directions in the data set).

Another, more important, measure of the dispersion of the distribution of directional data is p
(defined as the probability that the data are from a population that is uniformly distributed)
where
(6)
p ranges in value from 0 to 1 where p = 0 indicates that there is no chance that the data are
uniformly distributed (i.e., the data are from a population that has a preferred orientation) and
p = 1 indicates that the data are from a population that is definitely uniformly distributed (i.e.,
the population has no preferred orientation).
3.2.1 Scatter of directions

The clustering of directions about a mean value may be quite close or more widely dispersed.
In many cases it will be acceptable to describe dispersion qualitatively by inspection of the
rose diagrams. In some circumstances a more quantitative expression of dispersion may be
appropriate. The parameter most commonly used to express this is the vector strength, which
is given by the equation:
(7)
where n is the number of readings. High values indicate narrow dispersion and low values a
wider dispersion.

Table 1: Example of the statistical treatment of ungrouped and grouped directional data.

3.3 Analysis and Interpretation of data

When making an initial assessment of the pattern of current directions shown by a rose
diagram or radial line diagram, there are at least four questions to have in mind:
Is the pattern
1. unimodal
2. bimodal
1. with mutually opposed modes (bipolar)
2. with mutually perpendicular modes
3. polymodal.
What is the dominant or mean direction or directions?
How widely scattered are the directions about the meanvalues (i.e. what is the spread
of the directions)?
Are there any systematic differences in the directions derived from different types of
structure?

Figure 8: Diagrams of the main types of paleocurrent azimuthal patterns. From Selley [1968]

3.3.1 Mode of pattern

The most important feature of a population of directional data is its pattern of preferred
directions. In rock sequences, this can often give information on both the nature of the
depositional environment and the orientation of the regional palaeoslope.
Although unimodal patterns are more common, certain environments generate bimodal or
even polymodal patterns.
In some cases the modal directions derive from different types of structure (e.g. between sole
marks and ripples in some turbidite sequences).

In other cases, structures from the same class of structure may be polymodal (e.g. tabular
cross bedding in certain river deposits).
An exceptional but highly diagnostic pattern is that of bipolar cross bedding (i.e. the two
modes diametrically opposed),which is a strong indicator of tidal settings.
Selley [1968], proposed the possible models for classification of paleocurrents. Basically
each model can be defined by three parameters: environment (in terms of depositional
process), azimuthal paleocurrent pattern, and relation of paleocurrent to paleoslope.

Environment

Depositional Process

Fanglomerate

Gravity screefall and mudflow

Fluviatile

Unidirectional aquatic traction currents

Eolian

Wind-generated traction currents

Shore line

Bi or polydirectional traction currents

Turbidite

Unidirectional turbidity currents

Table 2: Mechanism and paleocurrent model. From Selley [1968]

Selley [1968] proposed by reference and examples from ancient and recent sediments. one
fanglomerate model, two fluviatile, one eolian, six shoreline, and two turbidite models.
1. Fanglomerates are characterized by unimodal down-slope directed radiating
paleocurrents.
2. Both fluviatile models have unimodal downslope paleocurrents. In one case these are
arranged radially, in the second centripetally.
3. Eolian paleocurrents may be unimodal, bimodal, or polymodal, and are unrelated to
paleoslope. One eolian model is definable by relating wind direction to highpressure
air cells.
4. Shore lines are characterized by bimodal paleocurrents attributable to tidal currents,
though unimodal and polymodal patterns are also known. These are all unrelated to

paleoslope. Six shoreline models are proposed based on the presence or absence of
fluviatile currents and on whether marine currents are dominantly onshore, offshore,
or alongshore.
5. The two turbidite models both have unimodal paleocurrents. These may be arranged
in smallscale downslope radiating fans. On a larger scale, lateral and axial infilling is
common in many marine troughs, though the role of turbidity currents has been
questioned.
3.4 Depositional environment and Dispersal Patterns

Environment

Directional structures

Dispersal pattern

Aeolian

Large-scale cross-bedding

Unimodal if formed by barchan-type

dunes and then indicates palaeowind
direction; bimodal if formed by
simple seif dunes and polymodal if
complex seifs; may relate to trade
winds; independent of regional
palaeoslope.

Fluvial

Best to use cross-beddingPalaeocurrents reflect palaeoslope

from
dune/
sandwaveand indicate provenance direction;
bedforms;
also
partingunimodal
pattern
with
small
lineation, ripples, scours,dispersion if low- sinuosity rivers;
imbrication, regional grain-unimodal with larger scatter if highsize changes
sinuosity river or alluvial fan

Deltaic

Best to use cross-beddingTypically unimodal pattern directed

from
dunes/sandwaveoffshore, although marine processes
bedforms; also channels,(tidal and storm currents and waves)
parting lineation, ripples
can complicate palaeocurrent pattern

Shallow-marine shelf

Best to use cross-beddingPattern can be complex and difficult

from
dunes/sandwaveto interpret; bimodal pattern through
bedforms; also ripples andtidal current reversals although tidal
scours
currents may be parallel or normal to
shoreline; can be unimodal if one
tidal current dominates; polymodal
and random patterns also occur;
complicated by wave and storm

effects

Turbidite basin

Unimodal pattern common from

Best to use sole structures,turbidites,
although
may
be
especially flutes; also partingdownslope or along basin axis, or
lineation, cross- lamination,radial if on submarine fan.
ripples, grain orientation,Contourites
give
palaeocurrent
slump folds
pattern parallel to the strike of the
slope

After Tucker [1991]

4 Discussion on Directional Variance and Vector Magnitude

4.1 The Sedimentary Structure Hierarchy
A fundamental concept in the understanding of sedimentary structures and paleocurrent
systems is that of hierarchy. It was first defined by Allen [1966] whose thesis was that flow
fields and the sedimentary structures arising from them are of five orders of magnitude. The
largest, or highest ranking, consists of complete river systems, the next comprises the
individual river, and so on.

Figure 9: The sedimentary structure hierarchy after Miall [1974] , showing paleocurrent
variability. Concept from Allen [1966] , although rank numbers used are different from
Allens, as Rank numbers in the illustration have been increased by two from Allen [1966] by
Miall [1974] for Allen did not assign a number to the river system.

Current rose diagrams, in Figure 9, illustrate directional variability present in the entire area
shown for each rank. This must be added to the variability shown by the larger scale
structures to arrive at the total variability for the structures of each rank. The small squares
show the areas chosen to illustrate the structures of the next smaller scale. Structures
illustrating rank 6 are bifurcating ripple crest The currents forming the smaller scale
structures are subsidiary to those forming the next Iargest in size ; they may be considered as
eddies within the larger currents, at each scale of observation. Structures of ranks 5 and 6 are
the ones most frequently available for study. Current directions derived from them will
approximate, statistically, those obtainable from the larger features, for the flow vector fields
which give rise to the smaller structures are dynamically related to the field of the overall
system[Miall, 1974].
Fluvial currents are vectors, definable by direction and magnitude, but most paleocurrent
studies ignore magnitude. It is proposed that azimuth readings be weighted according to the
cube of current structure thickness, this being a volume measure corresponding to the

distance in all three dimensions over which a local flow vector might reasonably be assumed
to maintain the same direction. It is also a measure of the quantity of sediment moved by the
flow vector.
The relative weight of the flow vector may thus be assessed using this volume measurement,
and it is therefore proposed that the cube of structure thickness be used to weight the
individual azimuth readings in vector mean calculations.
The method can be readily applied to any structures where the relationship of structure to
current is clear and unaml)iguous. Thus, it could not be applied to tool markings, but may be
used for all types of crossbedding as well as for ripple structures of large and small scale.
Some inaccuracy in the proposed method may accrue from the loss of thickness resulting
from erosion prior to final burial of a sedimentary structure. However, this is not thought to
be a serious problem.
4.2 The Problem of Vector Weight
A fluvial current is a vector, definable by direction and magnitude, but presently accepted
practices in processing directional data deal only with direction. The usual method is to
assign equal weight (unity) to each azimuth reading when calculating mean direction, yet if
struc tures of more than one rank are used equal weight is thereby assigned to lower rank
(smaller) structures, which are the least reliable from the point of view of regional paleoslope
determinations.
4.3 What is the relative importance of azimuth readings taken on the various types of
structure ?
Accodring to [Miall, 1974] it must be appreciated that the volume of a large scale planar
cross bed set (50 cm thick, or more) may be several hundred thousand times greater than that
of a small scale ripple (1 cm thick, on average). This volume contrast is a direct measure of
the relative quantity of dispersed sediment moved by the current which gave rise to the
structures and hence is an indication of the relative magnitude of the current system itself. It
is therefore rec ommended that structures of such markedly different size (different rank)
never be ana- lyzed together for paleocurrent purposes. To do so is to lose information by
confusing the effects of current flow fields of different mag- nitudes. Each structure type and
each rank has its own type of directional variance which it may be important to analyze.
the method proposed uses the equation, where mean azimuth is given by
(8)
The quantity n is defined by Curray [1956] as theobservation vector magnitude or, in the
case of grouped data of unit vectors, it is the number of observations in each group. But
Miall [1974] used ungrouped data, where n is used as a weighting factor to control for vector
magnitude, in the case the cube of cross-set thickness. Since his[Miall, 1974] paper was

primarily concerned with the smaller scale structures that are preserved in alluvial rocks, the
weighting method proposed could be applied to crossbedding of all types, ripples, parting
lineation, certain types of sole structure, etc., but not to pebble imbrication, in which structure
thickness, and thus current size, are hard to assess.

References
J. R. L. Allen. On bed forms and palaeocurrents. Sedimentology, 6(3): 153190, 1966. ISSN
1365-3091.
doi:
10.1111/j.1365-_3091.1966.tb01576.x.
URL
http://dx.doi.org/10.1111/j.1365-_3091.1966.tb01576.x.
J.D. Collinson, N. Mountney, and D.B. Thompson. Sedimentary Structures. Classic geology
in
Europe
series.
Terra
Pub.,
2006.
ISBN
9781903544198.
URL
https://books.google.co.in/books?id=9mC5QgAACAAJ.
Joseph R Curray. The analysis of two-dimensional orientation data. The Journal of
Geology, pages 117131, 1956.
Andrew D Miall. Paleocurrent analysis of alluvial sediments: a discussion of directional
variance and vector magnitude. Journal of Sedimentary Research, 44(4), 1974.
AndrewD. Miall. Paleocurrent analysis. In Sedimentology, Encyclopedia of Earth Science,
pages 825830. Springer Netherlands, 2003. ISBN 978-1-4020-0872-6. doi: 10.1007/978-_1_4020-_3609-_5_152. URL http://dx.doi.org/10.1007/978-_1-_4020-_3609-_5_152.
P.E. Potter and F.J. Pettijohn. Paleocurrents and basin analysis. Springer, 1977. ISBN
9783540079521. URL https://books.google.co.in/books?id=EKo1AAAAMAAJ.
Richard C Selley. A classification of paleocurrent models. The Journal of Geology, pages
99110, 1968.
M.E. Tucker. Sedimentary Petrology. Geoscience Texts. Wiley, 1991.
9780632029617. URL https://books.google.co.in/books?id=CRaoQgAACAAJ.

ISBN

Neil A Wells. Are there better alternatives to standard rose diagrams? Journal of
Sedimentary Research, 70(1):3746, 2000.