You are on page 1of 23

See

discussions, stats, and author profiles for this publication at:


http://www.researchgate.net/publication/256544219

A Parcel Shape Index for Use in Land


Consolidation Planning
ARTICLE in TRANSACTIONS IN GIS JANUARY 2012
Impact Factor: 0.54 DOI: 10.1111/j.1467-9671.2012.01371.x.

CITATIONS

READS

113

3 AUTHORS, INCLUDING:
Linda M. See
International Institute for Applie
144 PUBLICATIONS 1,906 CITATIONS
SEE PROFILE

John Stillwell
University of Leeds
132 PUBLICATIONS 956 CITATIONS
SEE PROFILE

Available from: Demetris Demetriou


Retrieved on: 05 December 2015

bs_bs_banner

Research Article

Transactions in GIS, 2013, ():

A Parcel Shape Index for Use in Land Consolidation


Planning
Demetris Demetriou,* Linda See and John Stillwell*
*School of Geography, University of Leeds

International Institute for Applied Systems Analysis (IIASA), Laxenburg

Abstract
Area shape analysis has been a subject of focus for research in geography and other disciplines for many
years. However, existing area shape indices present significant weaknesses for the analysis of land parcels
in an agricultural context as they do not measure shape in an appropriate, reliable or explicit manner.
This article presents a new parcel shape index (PSI) which integrates six geometric parameters through
multi-attribute decision-making, combined with value functions for parameter standardization. The superior performance of the new PSI in relation to three existing indices is demonstrated through a case study
area in Cyprus. The methodology can be applied in other contexts aimed at assessing the quality of a particular shape compared to an optimum.

1 Introduction
In classical geographical terms, the shape of an area (or polygon) is defined as an enclosed
homogenous 2-dimensional space comprised of a series of continuous lines and is used to represent physical elements such as lakes, political constituencies, soil types or land parcels. In
each of these contexts, the process of defining and measuring shape by utilizing specific indicators (or shape indices) is called shape analysis and has been of relevance to geographical study
since the 1920s. In the past, generic shape indices have been developed by a number of
researchers (Boyce and Clark 1964, Lee and Sallee 1970, Boots and Lamoureaux 1972, Frolov
1974, Moellering and Rayner 1982) while others have focused on shape indices specifically for
urban areas (Gibbs 1961, Lo 1980, Austin 1981, Batty and Longley 1994), market areas
(Simons 1974) and ecology (Eason 1992, Gutzwiller and Anderson 1992, Comber et al. 2003).
Shape analysis has also been of interest in other fields such as mathematics (Lord and Wilson
1984), computer science (Rosenfeld and Kak 1976, Gonzalez and Wintz 1977, Pavlides 1978,
Sagiv et al. 2003), computational geometry (Prepata and Shamos 1985), cognitive science
(Wentz 1997, Landau et al. 1988) and remote sensing (Zhang et al. 2006).
In addition to the above fields, land management involves shape analysis of land parcels,
particularly in areas where land fragmentation is a key issue. Land fragmentation has been
defined by King and Burton (1982) as the situation where farms are comprised of numerous
spatially separated parcels. It is recognized as a problem that hinders rational agricultural
development in many parts of the world and is solved using a land consolidation management
approach that involves restructuring the land tenure and providing appropriate infrastructure
such as roads and irrigation networks. One of the objectives of land consolidation is the creation of new parcels with shapes that are as regular as possible (which is one factor in defining
Address for correspondence: Demetris Demetriou, School of Geography, University of Leeds, Leeds LS2 9JT, UK. E-mail: demdeme@
cytanet.com.cy

2013 Blackwell Publishing Ltd

doi: 10.1111/j.1467-9671.2012.01371.x

D Demetriou, L See and J Stillwell

a good quality parcel shape in this context) because this encourages rational agricultural development by facilitating agricultural mechanization, cultivation and harvesting as well as the
exploitation of parcels in general (Barnes 1935, Heady and Jensen 1954, Haggett 1966, Lee
and Sallee 1974, Witney 1995, Landers 2000). Thus, a parcel shape index is required that
allows for comparison of the shape of any existing parcel with some ideal/optimum shape; in
the case of land consolidation schemes, the ideal shape is a rectangle with a certain length-tobreadth ratio, as discussed below. Other factors such as the quality of the soil and the slope of
the land are not specifically considered here, since they are inherently taken into account in the
land value which is involved in other land consolidation processes such as land redistribution
design and land partitioning (Demetriou et al. 2011a, 2012d).
The focus of this article is therefore on shape analysis of land parcels in the context of
land consolidation. Land parcel shapes have evolved over time based on the historical, cultural
(Palanques and Calvo 2011), social, economic, morphological and physical conditions prevailing within a country. For example, long parcels can be found in Quebec and parts of the
southern USA (Louisiana) which have been influenced by French colonization, to maximize the
number of lots with access to river systems. In Cyprus, the present land tenure system is a
result of a long historical evolution that dates back to the Neolithic era around 7,000 BC and
the numerous conquests of the island (Burton 1988), where the majority of parcel boundaries
have many edges and a seemingly random pattern, especially in hilly and mountainous areas.
As a result of the variety of land parcel shapes and arrangements, a number of studies have
appeared that have involved either developing a new shape index for land parcels (Amiama
et al. 2008) or using existing indices for this purpose (e.g. Coelho et al. 2001; Tourino et al.
2003; Gonzalez et al. 2004, 2007; Aslan et al. 2007; Libecap and Lueck 2009). However,
there are problems with these current indices, since none of them involves comparing existing
parcel shapes with an optimum shape that would be relevant for land consolidation in particular. Consequently, there is no index that reliably reflects the impact of parcel shape on the
effectiveness of cultivation (Amiama et al. 2008) and, more generally, on agricultural development. In addition, the indices that have been developed involve problems associated with the
interpretation of the similarities and differences in the values of the index in relation to parcel
shape (Moellering and Rayner 1982, Griffith et al. 1986, Wentz 2000). In particular, situations
may arise in which a rectangle may have the same index value as a very irregular shape; or
completely different shapes may have the same index value; or similar shapes may have different index scores. Details associated with specific indices and their problems are illustrated
though graphical examples provided below. These weaknesses occur mainly due to the fact
that none of the current indices provide an overall definition of shape (Wentz 2000), i.e. only
some of the geometric parameters of shape are taken into account. Wentz (1997) reflects upon
the complexity of the task and emphasizes that a single numerical representation of shape is
difficult to achieve through mathematical or statistical formulations. Moreover, indices developed in the context of solving one problem may not be applicable to others (MacEachren
1985). For all the above reasons a new index of land parcel shape analysis in the context of
land consolidation is therefore required.
The purpose of this article is to describe a new method for measuring the shape of land
parcels which integrates multi-attribute decision making (MADM) with a geographical information system (GIS). This method takes into account six geometric shape parameters which
are represented by empirical mathematical equations. The method is flexible, because the user
may select which parameters need to be involved in the model; problem specific, because
parameters may be assigned a different weight scheme depending on the problem examined;
knowledge-based, since it integrates expert judgment through the use of value functions; and
2013 Blackwell Publishing Ltd

Transactions in GIS, 2013, ()

A Parcel Shape Index for Use in Land Consolidation Planning

explicit, since the index falls between 0 to denote the worst (most irregular) shape and 1 to
indicate the best shape (rectangular with a certain length: breadth ratio as specified in section
3.2). A number of requirements for the development of shape indices are discussed by Lee and
Sallee (1970), Griffith et al. (1986) and Wentz (1997, 2000), which are also considered in this
research. A case study area in Cyprus is used to demonstrate the accuracy and reliability of the
new index for differentiating land parcel shapes and the results of the new index are compared
with indices used for land parcel shape analysis in the past.
This new index was specifically developed for the quantification of the current status of
land fragmentation in a potential land consolidation study area through a Land Fragmentation
System called LandFragmentS, which is part of a LAnd CONsolidation Integrated Support
System (LACONISS) (Demetriou et al. 2011c, 2012a, 2012d). The system was developed to
provide a semi-automated work flow for land consolidation that could help to reduce the time
needed for undertaking land consolidation projects, which is one major problem associated
with land consolidation. LACONISS has been developed using VBA and ArcObjects and is
fully integrated in ArcGIS.

2 Existing Indices
2.1 Generic Indices
Shape indices have been classified in a number of different ways in the literature (Pavlides
1978, MacEachren 1985, Wentz 2000). According to Wentz (2000), shape indices can be
grouped into two broad categories: single and multiple parameter indices. Single parameter
indices describe a shape based on a unique number with or without explicit limits, and are
estimated through combining two or more variables, e.g. area/perimeter ratios. Multiple
parameter indices represent shape by employing a series of numbers which usually involves
more complex mathematical calculations, e.g. use of Fourier series (Ricgard and Hemani
1974, Moellering and Rayner 1982).
Single indices can be split further into three sub-groups: compactness, boundary and form
indices. Compactness indices include: area-perimeter ratios; a single parameter of related
circles; direct comparison to a standard shape; and the dispersion of the elements of the area
of a shape. These indices have been employed in several applications (e.g. Longley et al. 1992,
Williams and Pearson 1997, Medda et al. 1998, Gamba 1999). MacEachren (1985), who
evaluated eleven compactness indices, found that substantial differences occurred in outcomes
with the first two categories presenting limited accuracy compared to the other two groups.
Thus, a critical issue is to select the most appropriate index for a given problem. However, this
kind of index has a narrow scope because it reflects the degree to which a shape is compact in
terms of morphology and does not constitute a true comprehensive measure of shape. As a
result, symmetrical shapes (of any form) are perfectly compact in contrast to shapes that are
fragmented, elongated, perforated and protruded. Although these compactness indices may be
useful for some problems, this research will demonstrate that they are not appropriate for
evaluating the shape of land parcels.
Boundary indices measure the outer roughness or smoothness of the shape of a polygon.
This characteristic can be measured using fractal analysis, which has been employed by several
researchers for investigating shape in various spatial phenomena (e.g. Batty and Longley 1994,
Chen et al. 1990). Similar to compactness indices, boundary indices are also inappropriate for
describing the shape of land parcels because they present significant drawbacks, as will be
illustrated later in Section 2.2.
2013 Blackwell Publishing Ltd

Transactions in GIS, 2013, ()

D Demetriou, L See and J Stillwell

Form indices attempt to represent the whole geometry of a shape and include Hough
transforms, Freeman chain codes and the Radial line index (Wentz 2000). These indices have
been utilized in several applications (Boyce and Clark 1964, Duda and Hart 1972, Ballard
1981). The rationale behind these indices is that they deconstruct an irregular shape into
smaller simpler regular shapes such as squares, circles and triangles, which can be defined
numerically. A weakness of form indices is that they do not maintain topology and may result
in complex outcomes that are difficult to interpret (Wentz 1997).
The second main category of shape indices includes multiple parameter indices which
involve several indicators (Bunge 1966, Taylor 1971, Moellering and Rayner 1982). The indicators are not combined to estimate a unique number for each shape (Griffith et al. 1986) but
are useful for reconstructing shapes rather than aiding interpretation. For this reason, multiple
parameter indices are more useful in other fields such as computer science and not for investigating spatial problems such as land fragmentation.

2.2 Shape Indices for Land Parcels


Compactness and fractal indices have been used by a number of authors in the past (Gonzalez
et al. 2004, 2007; Aslan et al. 2007; Libecap and Lueck 2009). For example, Aslan et al.
(2007) utilized two indices: the shape index (SI) and the fractal dimension (FD) to determine
the success of land consolidation projects for specific case study areas by comparing parcel
shapes based on each index before and after the implementation of land consolidation. The
comparison relies on the assumptions that an SI value of 1.128 and an FD value of 1.4 refer to
square parcel shapes. As demonstrated later, however, this assumption fails since it is possible
to show that parcels with irregular shapes may also have similar values of these indices. These
two indices, which use parcel perimeter, p, and parcel area, a, are defined as:

SI =

pi
2 ai

(1)

FD =

2 ln pi
ln ai

(2)

and

These indices were originally developed for representing ecosystem fragmentation


(McGarigal and Marks 1995, Krummel et al. 1987, ONeill et al. 1988, Milne 1991). Thus,
they are more focused on landscape characterization and are not suited for land parcel shape
analysis in the context of land fragmentation. The latter is demonstrated in Figures 1 and 2
which present the index values for SI and FD, respectively, for a group of shapes. Both indices
were calculated for regular shapes (shaded polygons) and for real parcels extracted from the
Cyprus case study (Section 4) which contained 340 land parcels altogether. In particular,
Figures 1e and 2c show examples of parcel shapes that are alike and have similar indices,
which is a desired property. In contrast, however, Figures 1a-d and 2a-b highlight two problems with these indices. For example, the SI for all rectangles equals 1.128 but Figure 1a
shows that shapes that are significantly different from a rectangle may have exactly the same
or very similar SI values. Similarly, Figures 1b-d show how considerably dissimilar shapes may
2013 Blackwell Publishing Ltd

Transactions in GIS, 2013, ()

A Parcel Shape Index for Use in Land Consolidation Planning

Figure 1 Shape index (SI) values for various parcel shapes taken from the Cyprus case study: (a-d)
different shapes having similar SI values; and (e): similar shapes having similar SI values

have the same SI value. The FD shows an even worse performance, giving different values for
different sizes of rectangles (Figure 2a). Moreover, the same values emerge from significantly
dissimilar shapes with apparent varying quality (Figure 2b). In addition, both indices inherently have another two weaknesses: there is no explicit definition of which is the best and
worst shape for the problem concerned and they both take into account only two geometric
shape factors, i.e. area and perimeter, while many more factors should be considered for land
parcel shape analysis in the context of land fragmentation, e.g. the number of boundary points
and the size of the angles of a parcel).
2013 Blackwell Publishing Ltd

Transactions in GIS, 2013, ()

D Demetriou, L See and J Stillwell

Figure 2 Fractal dimension (FD) for regular parcel shapes and various parcel shapes taken from
the Cyprus case study: (a) similar shapes having different FD values; (b) different shapes having
similar FD values; and (c) similar shapes having similar FD values

The areal form factor (AFF) is another compactness index that has been utilized by
Whitney (1995) and Gonzalez et al. (2004, 2007):

AFF =

ai
pi2

(3)

This simple area/perimeter ratio has the advantage of being independent of parcel size,
unlike the FD as shown previously. The AFF also has the same two drawbacks as noted earlier
for the SI. Figures 3a-e illustrate that completely irregular shapes, which are clearly very bad
for agricultural purposes, may have very close AFF values with orthogonal shapes (of various
length:breadth ratios). In contrast to the SI and FD, the AFF performs well for similar shapes,
i.e. they have values that are close to one another (Figure 3f).
One might argue that two or more shapes may have the same index when compared to a
standard shape (e.g. a circle in the case of compactness measures) because they have the same
distance (for different geometric features) compared to the characteristics of the specified
shape. However, for agricultural purposes this is not valid because this would mean that land
parcels with different quality (in terms of shape) could have the same index. This is also apparent from the AFF values of the 36 standard parcel shapes defined by Gonzalez et al. (2004)
and represents a serious disadvantage of this methodology.
Therefore, all the above examples illustrate that the existing indices SI, FD and AFF are
not suited for land parcel analysis in the context of land fragmentation because of the inconsistency in representing parcel shapes and the fact that they are incapable of comparing shapes
with a predefined optimum/standard shape, e.g. a rectangle with a certain length:breadth ratio.
2013 Blackwell Publishing Ltd

Transactions in GIS, 2013, ()

A Parcel Shape Index for Use in Land Consolidation Planning

Figure 3 AFF values for regular parcel shapes and various parcel shapes taken from the Cyprus case
study: (a-e) different shapes having similar AFF values; and (f) similar shapes having similar AFF
values

In addition, recent efforts to tackle the disadvantages of the above indices and in particular of the AFF have been undertaken by Amiama et al. (2008) who proposed two new indices:
the first is the highest ratio between the area of a parcel and the area of the quadrilateral with
orthogonal sides that best circumscribes the parcel; the second applies a correction factor that
results from comparing the parcel perimeter with the perimeter of the quadrilateral, measured
on a scale between 0 and 1. There was insufficient detail presented in Amiama et al. (2008) on
which to base any calculation of these indices in order to examine their performance on the
case study parcels. However, as the authors were only able to produce results that were slightly
better than the AFF, we surmise that both of these indices will present similar weaknesses to
those of the AFF noted earlier.
2013 Blackwell Publishing Ltd

Transactions in GIS, 2013, ()

D Demetriou, L See and J Stillwell

Figure 4

A generic parcel shape analysis model employing MADM

A final relevant study is that of Libecap and Lueck (2009), who utilized a variation of the
area-perimeter ratio ( pi ai ) and the number of sides to calculate average values of each
quantity based on the parcels in a township. The index was calculated for the case study
parcels and revealed the same weaknesses as those of the SI and AFF. Thus the SI, FD and
AFF and other related indices from the literature are currently not appropriate for agriculture
because they take only two geometric parameters of parcel shape into account, which is clearly
insufficient. More parameters are needed in order to quantify the shape of a parcel while
allowing for clear definition of which shapes are good and which are bad. Other studies that
have used parcel shape indices include those by Coelho et al. (2001), who employed a shape
coefficient representing the effect on machine turning time in relation to a rectangular shape
without giving any other information, and Tourino et al. (2003), who simply used the average
perimeter of land parcels as an evaluation indicator.

3 A New Parcel Shape Index (PSI)


3.1 Developing a Generic Shape Analysis Model Using MADM
The development of the parcel shape index (PSI) is based on a generic shape analysis model
that utilizes the multi-attribute decision-making (MADM) method, which is more conventionally employed for the assessment of alternative solutions of a problem (Malczewski 1999,
2006; Sharifi et al. 2004). This method can be flexibly adapted for developing any index and
can be represented by a set of parameters aimed at evaluating an object compared to an
optimum or standard one for the application concerned. In particular, Figure 4 illustrates the
process whereby a planner selects the shape analysis parameters to be incorporated into the
model (length of sides, acute angles, reflex angles, boundary points, compactness and regularity). The user may then assign a relevant weight to each parameter, which represents its importance in a given project. These parameters should be preferentially independent (Beinat 1997)
to avoid duplication. A parameter x is said to be preferentially independent of parameter y if
preferences for specific outcomes of x do not depend on the level of criterion y. If x is preferentially independent of y and y is preferentially independent of x, then these parameters are
mutually preferentially independent. Testing for this involves calculating the statistical correlation between variables or establishing preferential independence through expert knowledge.
Thereafter, the score associated with each of these parameters is calculated for each parcel.
The scores are then standardized using value functions as explained later, although other
appropriate methods could be employed. Value functions are mathematical representations
2013 Blackwell Publishing Ltd

Transactions in GIS, 2013, ()

A Parcel Shape Index for Use in Land Consolidation Planning

of human judgement. These functions translate the score of a parameter into a value representing the degree to which this parameter is close to the corresponding parameter of an
optimum or standard shape. The value is a dimensionless score between 0 and 1. Usually a
value of 1 indicates that it is close to the optimum or standard shape while 0 indicates the
opposite. Then, the parcel shape index for parcel i (PSIi) is computed by multiplying the
standardized score of each parameter (Pij) by the relevant parameter weight (wj) and
summing these up divided by the number of parameters (m) involved. The relevant equation
is shown below:
m

P w
ij

PSIi =

i =1

(4)

which yields values between 0 and 1, representing the worst and best shapes compared to an
optimum shape as defined previously.

3.2 Defining an Optimum Parcel Shape for Land Consolidation


Parcels with irregular shapes present many disadvantages in terms of parcel cultivation, lower
crop yields, land wastage, need for increased conservation works (e.g. fencing) or boundary
disputes (Bentley 1987). Therefore, many studies have discussed the advantages provided by
rectangular parcels (Barnes 1935, Heady and Jensen 1954, Haggett 1966, Lee and Sallee 1974,
Witney 1995, Landers 2000). In particular, Witney (1995) and Gonzalez et al. (2004) have
shown that the most optimal rectangle in terms of cultivation is one with a length:breadth
ratio of 4:1. However, this finding is based on the maximum ploughing area, i.e. the minimum
dead ground, which is a factor relating to cultivation and not to other factors such as irrigation, crop type or the potential land-use changes of a parcel. Thus, in Cyprus, where limited
housing development is permitted in agricultural areas, the optimal parcel shape utilized in
practice (at least in land consolidation areas) is a rectangle with a length:breadth ratio of 2:1
or alternatively something between 1:1 and 3:1 with the former considered (empirically) as the
optimum shape for this research. As a result, the new parcel shape index (PSI) compares the
shape of a land parcel with this optimum shape.
A PSI value of 1 will represent the optimum shape where shapes such as a square or a rectangle with a length:breadth ratio of 3:1 will also be assigned values very close to 1, namely,
0.931 and 0.889, respectively. As the length:breadth ratio continues to increase, i.e. to 4:1 and
5:1, the PSI decreases to 0.598 and 0.315 respectively, indicating that these shapes are less
desirable. These shapes and their PSI values are shown in Figure 5.

3.3 Defining the Parcel Shape Parameters of the PSI


Equation (4) contains the parameters (m) of the shape analysis model which can be defined in
more detail as: length of sides (f1), acute angles (f2), reflex angles (f3), boundary points (f4),
compactness (f5) and regularity (f6). The first five shape parameters are standardized using
value functions, which were defined by five land consolidation experts based on the direct
value rating method (Beinat 1997), a process explained in Demetriou et al. (2011b, 2012b)
where more details are provided regarding the rationale for constructing each value function.
The relevant value functions and the corresponding equations are indicated in Figure 6. Each
2013 Blackwell Publishing Ltd

Transactions in GIS, 2013, ()

10

D Demetriou, L See and J Stillwell

Figure 5

PSI values for rectangles with various length:breadth ratios

equation estimates a value for a corresponding score of each parameter for a certain parcel.
Each score is then calculated automatically by the LandFragmentS module developed in
ArcGIS (Demetriou et al. 2012c). In all of the value functions, scores less than Xmin are standardized to 0, while scores higher than Xmax are standardized to 1.
The sixth shape parameter (regularity) is not standardized using a value function because
the values are case study dependent and are not a function of expert judgement. A standardization method was developed specifically for this parameter, referred to here as the mean
standardization method (mSM), which was developed in the context of this research. In contrast to the maximum standardization method (Sharifi et al. 2004), the mSM (Demetriou et al.
2011b) accounts for the potential extreme minimum and maximum values by using the sample
mean as follows:

( S min S ) 0.5
(if Si meanS )
Ei = 1 i
meanS min S

(5)

and

) )

( S meanS ) 0.5
Ei = 1 i
+ 0.5 (if Si > meanS )
max S meanS

(6)

where Ei is the standardized value of score Si and minS, maxS, meanS are the minimum,
maximum and average values for all of the scores in the dataset.
More details of each parameter are provided below:
Length of sides (f1): the length of a side of a land parcel should exceed some specified
minimum and may theoretically reach any value. In particular, this parameter reflects the
minimum length as defined by land consolidation experts, which is 25 m. Thus, parcels
with a total side length of less than 25 m are penalized through value function as shown in
Figure 5a.
2013 Blackwell Publishing Ltd

Transactions in GIS, 2013, ()

A Parcel Shape Index for Use in Land Consolidation Planning

Figure 6

11

The value functions and the corresponding equations for the five shape parameters

Acute angles (f2): an acute angle is an angle that is less than 90. Acute angles constitute a
weakness for a land parcel (Amiama et al. 2008) and the more acute angles there are in a
parcel shape, the worse this becomes. In particular, experts have decided that an acute
angle of less than 80 should be penalized because it is a significant disadvantage in terms
of parcel exploitation (Figure 5b).
Reflex angles (f3): a reflex angle is more than 180 but less than 360. Similar to acute
angles, the presence of reflex angles constitutes a drawback for land parcel exploitation.
The experts indicated that a reflex angle of greater than 215 should be penalized for the
same reasons noted earlier, which is captured in the value function shown in Figure 5c.
Boundary points (f4): the number of corners of a parcel defines the density and complexity
of a polygon. Thus, clearly, the desirable number of boundary points for a land parcel is
four although a slightly higher number of points may not worsen a shape if all other
2013 Blackwell Publishing Ltd

Transactions in GIS, 2013, ()

12

D Demetriou, L See and J Stillwell

factors are satisfied. In contrast, a polygon with three points, i.e. a triangle, is not an
acceptable shape for a land parcel and would constitute a serious weakness in the land
parcel. These considerations are captured in the value function illustrated in Figure 5d.
Compactness (f5): although problematic if utilized in isolation (for the reasons outlined previously), it is necessary to include this factor in the new shape index because it ensures that
a shape has a harmonic area/perimeter ratio. Based on a series of investigations of the
behavior of the area/perimeter2 ratio, the value function shown in Figure 5e has been
defined.
Regularity (f6): any regular polygon has a rotational symmetry; thus, all of its points circumscribe a common circle. In addition, a regular polygon has all sides and angles equal.
Therefore, this parameter is used to check if a parcel has a regular shape like that of the
optimum shape.
Another important component of Equation (4) is the weight of each parameter representing its importance to the problem. For the case study in this article, the weights were set equal
for all the parameters. It should also be noted that the PSI reflects only the aspects defined by
the 2D shape of land parcels and does not take into account slope, which would require a 3D
shape. However, as noted earlier, slope is inherently taken into account in land values utilized
in other land consolidation processes. In addition, size is not involved in the PSI because the
latter measures shapes independently of their size, which is an additional critical factor that is
included in the land fragmentation model (LandFragmentS) in addition to the PSI.

3.4 Criteria for the Assessment of the Shape Index


A number of authors have provided criteria in the past for constructing a shape index (Lee and
Sallee 1970, MacEachren 1985, Moellering and Rayner 1982, Austin 1981, Griffith et al.
1986). Wentz (1997, 2000) has summarized these criteria and produced a list of 11 requirements. A shape index should: represent different shapes by different numbers (C1); be applicable to convex and concave polygons (C2); identify holes in polygons (C3); be independent
under translation (i.e. moving without rotation) (C4); be independent under rotation (C5); be
independent under scale change (C6); be applied easily to geographic data (C7); be straightforward to compute (C8); involve a data preparation process that is easy to follow (C9); have
results that are easy to interpret (C10); and match or not contradict human intuition of shape
(C11).
Two of the criteria which are not relevant to land parcels can be omitted, i.e. it is not necessary to be able to rebuild a shape (Moellering and Rayner 1982, Austin 1981) as the aim is
to assess shapes for planning purposes, and the land parcels are closed polygon shapes with no
holes. An additional requirement should be added whereby the shape index should take values
within a predefined dimensionless range so that there is an explicit definition of what is the
best and worst shape for the problem concerned. The 11 requirements noted above are used in
the assessment of the PSI in the case study that follows.

4 Case Study
4.1 Shape Classification and Interpretation of the PSI
Figure 7 shows the spatial distribution of the PSI across the study area for 10 classes of equal
range of PSI values from 0 to 1, where 0 represents the worst (irregular-complex shapes)
2013 Blackwell Publishing Ltd

Transactions in GIS, 2013, ()

A Parcel Shape Index for Use in Land Consolidation Planning

Figure 7

13

Spatial distribution of the PSI based on 10 categories for the study area in Cyprus

increasing towards 1 with the best parcel shapes (i.e. regular rectangles). A visual inspection of
each category reveals that there are common geometric characteristics associated with the
parcel shapes that belong to each category.
The lowest class clearly shows highly irregular and complex shapes. There is a consistency
in the value of the index since all of the shapes have similar geometric features. Similarly,
classes 2 to 5 comprise highly irregular shapes of varying degrees. The degree of irregularity, or
more precisely how far a shape is from the optimum, varies and different shapes may have a
2013 Blackwell Publishing Ltd

Transactions in GIS, 2013, ()

14

D Demetriou, L See and J Stillwell

similar PSI because of different undesirable geometric characteristics. However, some shapes
are disfavoured which are not irregular along the whole of their perimeter (i.e. some of their
sides are straight lines) compared to shapes that are completely irregular along their whole
perimeter. This occurs because the model examines each shape as a whole compared to the
optimum shape. A shape might have a few (or even all) straight sides but there are other problems with the shape in terms of the other factors considered in the PSI. Thus, this shape will
essentially be irregular with several disadvantages for agricultural purposes. In addition, in
classes 2 to 5, the PSI is very sensitive since a shape may be in one of these four groups
because it is better or worse than another in just one geometric feature.
In contrast to the first five categories, the remaining categories contain better shapes with
more consistency between them. As the PSI increases, the shapes begin to look more acceptable relative to the optimum. For example, classes 6 and 7 clearly contain parcels with irregular shapes but these are definitely better than the shapes found in the previous groups. Both
categories have shapes with common characteristics and most of them could be in either
group. Class 8 includes shapes that are quite close to being regular with similarities in the
shapes, once again illustrating the performance and consistency of the index, while class 9
involves almost regular shapes. Some of them are absolutely regular although they do not
meet all of the requirements of the optimum shape. Finally, in class 10, almost all of the
parcels are optimum or very close to the optimum. The parcels in this class also illustrate that
the index is able to recognize the optimum shapes. The 10 categories can be collapsed into
four groups as shown in Figure 8: highly irregular shapes (classes 1 to 5); irregular shapes
(classes 6 and 7); regular or near regular shapes (classes 8 and 9) and; optimum or near
optimum shapes (class 10). Each of these groups contains land parcels with similar shapes,
which is confirmed by the small standard deviation of the PSI within each group relative to the
size of the group (given in parentheses), i.e. 0.119 (00.50), 0.050 (0.510.70), 0.058 (0.71
0.90) and 0.027 (0.911.0), respectively. Whilst the majority of parcels (65.9%) have either
highly irregular or irregular shapes, a significant number (28.8%) have regular or near regular
shapes and a small minority of parcel shapes (5.3%) have optimum or near optimum shapes in
this case study area.

4.2 Evaluation of the PSI


The PSI is evaluated through comparison with existing methods, i.e. the SI, FD and AFF, and
the ability to satisfy the 11 requirements as outlined in Section 3.1. An independence check of
the parameters is also undertaken.

4.2.1 Comparison with existing indices


Figures 9 to 11 show the results of a comparison between the PSI and the SI, FD and AFF for
the same parcel shapes as shown in Figures 1 to 3, where the corresponding PSI values are
shown in parentheses. The results shows that the PSI overcomes the deficiencies of the existing
indicators, i.e. the index increases gradually as the shape becomes better and moves towards
the optimum; dissimilar shapes have different values especially compared to rectangular ones;
and similar rectangular shapes have the same index. Hence, the PSI clearly outperforms the
other three indices.
However, two weaknesses of the PSI can be observed: (1) symmetrical shapes are favored
and hence they may have a higher index than expected, e.g. shapes with a triangular form as
shown in Figure 12 have a relatively high PSI; and (2) shapes with a general regular form, i.e.
2013 Blackwell Publishing Ltd

Transactions in GIS, 2013, ()

A Parcel Shape Index for Use in Land Consolidation Planning

Figure 8

15

Classification of parcel shapes based on four groups

those having some straight sides and others non-linear, as shown in Figure 13, are perhaps
penalized more than shapes that have a completely irregular shape along their entire periphery.
These weaknesses could be improved by either investigating a different form of the relevant
value functions or by adding other parameters e.g. for obtuse angles or by penalizing symmetrical shapes with angles different from the limits suggested.
2013 Blackwell Publishing Ltd

Transactions in GIS, 2013, ()

16

D Demetriou, L See and J Stillwell

Figure 9

Comparison of the PSI (values in parentheses) with the SI (values without parentheses)

4.2.2 Satisfaction of requirements


Table 1 provides a qualitative comparison of the existing indices with the PSI, based on
whether they satisfy the 11 requirements set out previously in Section 3.1.
The PSI clearly outperforms all of the other indices since it satisfies all of the required criteria. Criteria C2 to C9 are satisfied by all of the methods with the exception of the FD for C6
while C1, C10 and C11 are only satisfied by the PSI. In the case of C1, there were some individual cases when the criterion was not met but this occurred in only 3.5% of the parcels in
the case study example. Further examples are required to determine the robustness of the PSI
with respect to this criterion. Criterion C10 is met because the PSI is expressed as a number
2013 Blackwell Publishing Ltd

Transactions in GIS, 2013, ()

A Parcel Shape Index for Use in Land Consolidation Planning

Figure 10

17

Comparison of the PSI (values in parentheses) with the FD (values without parentheses)

between 0 and 1 corresponding to the worst and best shapes, respectively so the index is easy
to interpret in contrast to the other indices. C11 is also fulfilled by the PSI because the methodology involves the incorporation of human experts in defining the value functions for five
out of six parameters that were directly involved in the calculation of the index.

4.2.3 Independence of the input variables


All combinations of pairs of parameters were tested for statistical correlation. This process
revealed three high correlations between the following pairs: length of sidesreflex angles (r =
0.735); length of sides-boundary points (r = 0.907) and reflex anglesboundary points (r =
0.828). However, these pairs are essentially preferentially independent. For example, for the
first pair, while the preference (for a perfect shape) is to have the smallest number of polygon
sides with a length less than the defined minimum, this preference does not depend on the
number of reflex angles and vice versa because, for example, a random polygon from the case
study area may not have any length of a side less than the minimum (the preference) whilst it
may have many or no reflex angles and vice versa. Similarly, this rationale can be applied to
the other two pairs of parameters. Therefore, the six parameters can be used in the PSI calculations to comprehensively describe the shape of a land parcel.
In addition to the evaluation of the PSI as provided above, two further issues need consideration in the context of future research. As noted earlier, all parameters in this exercise had
2013 Blackwell Publishing Ltd

Transactions in GIS, 2013, ()

18

D Demetriou, L See and J Stillwell

Figure 11

Comparison of the PSI (values in parentheses) with the AFF (values without parentheses)

the same weight, i.e. the same importance. However, this may not be true in some cases or for
other problems. Hence sensitivity tests of the weights utilized would become essential. In addition, the sensitivity of the scores of the parameters in association with the behavior of the
value functions should also be investigated.

5 Conclusions
Existing shape analysis methods, particularly those utilized for parcel shape evaluation, suffer
from some basic deficiencies. Therefore, a new parcel shape index has been developed based
2013 Blackwell Publishing Ltd

Transactions in GIS, 2013, ()

A Parcel Shape Index for Use in Land Consolidation Planning

Figure 12

Shapes which are favoured in terms of the PSI

Figure 13

Shapes which are undesirable based on the PSI compared to fully irregular shapes

19

Table 1 Satisfaction of requirements for an ideal index

C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11

Criterion

AFF

SI

FD

PSI

Different shapes are represented by a different value


Similar shapes are represented by a similar value
Applied to convex and non-convex (concave) polygons
Independent under translation
Independent under rotation
Independent under scale change
Applies easily to geographic data
The index is straightforward to compute
Data preparation is easy
Results are easy to interpret
The index value matches or does not contradict human
intuition of shape

: criterion satisfied : criterion not satisfied


2013 Blackwell Publishing Ltd

Transactions in GIS, 2013, ()

20

D Demetriou, L See and J Stillwell

on six variables: length of sides, acute angles, reflex angles, boundary points, compactness and
regularity. A case study was used to illustrate that the PSI overcomes the problems associated
with existing indices based on evaluation using 11 criteria. Although the PSI has minor weaknesses, further research could be undertaken to investigate the effects of adding more parameters to the index or optimizing the value functions.
One major contribution of this index is the flexibility of the approach through the use of
the multi-attribute decision-making method. Thus, the method could be adopted for other
applications in other fields where the aim is to assess the quality of objects (which was the
parcel shape in this study) compared to a pre-defined optimum, standard object that can be
specified for a given task. The new method has several promising features. It incorporates
several parameters into a single index; it is flexible because the user may select which parameters should be taken into account; it is problem specific because the user may assign a different weight to each parameter to denote differing importance for the problem under
consideration; it is knowledge based since expert judgment is incorporated via the value functions; and it is explicit in terms of interpretation of the outcomes since the index ranges from 0
to 1 denoting the worst and best shapes respectively.
Lessons learnt from this research suggest that there are two critical factors for success.
First, it is important to have an index that can take all the possible parameters of a problem
into account. By developing such an index in this research study, the transferability to other
fields and applications becomes possible. Second, the value functions must be carefully constructed and their behavior should be investigated across the range of allowable values. For
example, undertaking a sensitivity analysis of the parameter weights, the scores of these
parameters and the values extracted through the value functions could lead to a more robust
index and strengthen this approach. In conclusion, the method can be applied relatively easily
using GIS and it fills a major gap in existing shape analysis that is applicable to many fields
beyond land consolidation.

References
Amiama C, Bueno J, and Alvarez C J 2008 Influence of the physical parameters of fields and the crop yield on
the effective field capacity of a self-propelled forage harvester. Biosystems Engineering 100: 198205
Aslan T, Gundogdu K, and Arici I 2007 Some metric indices for the assessment of land consolidation projects.
Pakistan Journal of Biological Sciences 10: 139097
Austin F 1981 The shape of west Malaysias districts. Area 13: 14550
Ballard D H 1981 Generalisation of the Hough transform to detect arbitrary shapes. Pattern Recognition 13:
11122
Barnes C P 1935 Economies of the long-lot farm. Geographical Review 25: 298301
Batty M and Longley P 1994 Fractal Cities: A Geometry of Form and Function. London, Academic Press
Beinat E 1997 Value Functions for Environmental Management. Dordrecht, The Netherlands, Kluwer
Bentley W 1987 Economic and ecological approaches to land fragmentation: In defence of a muchmaligned phenomenon. Annual Review of Anthropology 16: 3167
Boots B and Lamoureaux L 1972 Working Notes and Bibliography on the Study of Shape in Human Geography
and Planning. Urbana-Champaign, IL, Council of Planning Librarians, Exchange Bibliography No 346
Boyce R and Clark W 1964 The concept of shape geography. Geographical Review 54: 56172
Bunge W 1966 Theoretical Geography. Lund, Sweden, C W K Gleerup
Burton S 1988 Land consolidation in Cyprus: A vital policy for rural reconstruction. Land Use Policy 5: 13147
Chen S S, Keller J M, and Crownover R M 1990 Shape from the fractal geometry. Artificial Intelligence 43:
199218
Coelho C, Pinto P A, and Silva M 2001 A systems approach for the estimation of the effects of land consolidation projects (LCPs): A module and its application. Agricultural Systems 68: 17995
Comber A J, Birnie R V, and Hodgson M 2003 A retrospective analysis of land cover change using polygon
shape index. Global Ecology and Biogeography 12: 20715
2013 Blackwell Publishing Ltd

Transactions in GIS, 2013, ()

A Parcel Shape Index for Use in Land Consolidation Planning

21

Demetriou D, Stillwell J, and See L 2011a LandSpaCES: A Spatial Expert System for Land Consolidation. In
Geertman S, Reinhardt W, and Toppen F (eds) Advancing Geo-information Science for a Changing World.
Berlin, Springer: 24974
Demetriou D, Stillwell J, and See L 2011b A multi-attribute decision-making module for the evaluation of alternative land consolidation plans. Leeds, University of Leeds, School of Geography Working Paper No 11/02
(available at http://www.geog.leeds.ac.uk/fileadmin/downloads/school/research/wpapers/11-02.pdf)
Demetriou D, Stillwell J, and See L 2011c The development and evaluation of a new model for measuring land
fragmentation. Leeds, University of Leeds, School of Geography Working Paper No 11/05 (available at
http://www.geog.leeds.ac.uk/fileadmin/downloads/school/research/wpapers/11-05.pdf)
Demetriou D, Stillwell J, and See L 2012a An Integrated Planning and Decision Support System (IPDSS) for land
consolidation: Theoretical framework and application of the land redistribution modules. Environment
and Planning B 39: 60928
Demetriou D, See L, and Stillwell J 2012b A new methodology for the evaluation of land reallocation plans.
ISPRS International Journal of Geo-Information 1: 27293
Demetriou D, See L, and Stillwell J 2012c A new methodology for measuring land fragmentation. Environment
and Planning B 39: in press
Demetriou D, Stillwell J, and See L 2012d LandParcelS: A Module for Automated Land Partitioning. Leeds,
University of Leeds, School of Geography Working Paper No 12/02 (available at http://
www.geog.leeds.ac.uk/fileadmin/downloads/school/research/wpapers/12-02.pdf)
Duda R and Hart P E 1972 Use of the Hough transformation to detect lines and curves in pictures. Communications of the ACM 15: 1115
Eason P 1992 Optimisation of territory shape in heterogeneous habitats: A field study of the redcapped cardinal
(Paroaria gularis). Journal of Animal Ecology 61: 41124
Frolov Y 1974 Measuring the shape of geographical phenomena: A history of the issue. Soviet Geography:
Review and Translation 16: 67687
Gamba P 1999 Meteorological structures shape description and tracking by means of BI-RME matching. IEEE
Transactions on Geosciences and Remote Sensing 37: 115161
Gibbs J 1961 A method for comparing the spatial shapes of urban units. In Gibbs J (ed) Urban Research
Methods. London, Van Nostrand
Gonzalez R and Wintz P 1977 Digital Image Processing. Reading, MA, Addison-Wesley
Gonzalez X P, Alvarez C J, and Crecente R 2004 Evaluation of land distributions with joint regard to plot size
and shape. Agricultural Systems 82: 3143
Gonzalez X P, Marey M F, and Alvarez C J 2007 Evaluation of productive rural land patterns with joint regard
to the size, shape and dispersion of plots. Agricultural Systems 92: 5262
Griffith D, ONeill M, ONeill W, Leifer L, and Mooney R 1986 Shape indices: useful measures or red herrings?
Professional Geographer 38: 26370
Gutzwiller K J and Anderson S H 1992 Interception of moving organisms: Influences of patch shape, size and
orientation on community structure. Landscape Ecology 7: 293303
Haggett P 1966 Locational Analysis in Human Geography. New York, St. Martins Press
Heady E O and Jensen J R 1954 Farm Management Economies. Englewood Cliffs, NJ, Prentice-Hall
King R and Burton S 1982 Land fragmentation: Notes on a fundamental rural spatial problem. Progress in
Human Geography 6: 47594
Krummel J R, Gardner R H, Sugihara G, ONeill R V, and Coleman P R 1987 Landscape patterns in a disturbed environment. Oikos 48: 32124
Landau B, Smith L B, and Jones S S 1988 The importance of shape in early lexical learning. Cognitive Development 3: 299321
Landers A 2000 Resource Management Farm Machinery: Selection, Investment and Management. Tonbridge,
UK, Farming Press
Lee D and Sallee T 1970 A method of measuring shape. Geographical Review 60: 55563
Lee D and Sallee T 1974 Theoretical patterns of farm shape and central place location. Journal of Regional
Science 14: 42330
Libecap G and Lueck D 2009 The Demarcation of Land and the Role of Coordinating Institutions. Cambridge,
MA, National Bureau of Economic Research Working Paper No. 14942
Lo C P 1980 Changes in the shape of Chinese cities, 19341974. Professional Geographer 32: 17383
Longley P, Batty M, Shepherd J, and Sadler G 1992 Do green belts change the shape of urban areas? A preliminary analysis of the settlement geography of south-east England. Regional Studies 26: 43752
Lord E A and Wilson C B 1984 The Mathematical Description of Shape and Form. West Sussex, England, Ellis
Horwood
MacEachren A 1985 Compactness of geographic shape: Comparison and evaluation of measures. Geografiska
Annaler. Series B, Human Geography 67: 5367

2013 Blackwell Publishing Ltd

Transactions in GIS, 2013, ()

22

D Demetriou, L See and J Stillwell

Malczewski J 1999 GIS and Multicriteria Decision Analysis. New York, John Wiley and Sons
Malczewski J 2006 GIS-based multicriteria decision analysis: A survey of the literature. International Journal of
Geographical information Science 20: 70326
McGarical K and Marks B J 1995 FRAGSTATS: Spatial Pattern Analysis Program for Quantifying Landscape
Structure. Washington, D.C., U.S. Department of Agriculture, Forest Service, General Technical Report No
PNW-GTR-351
Medda F, NijKamp P, and Rietveld P 1998 Recognition and classification of urban shapes. Geographical Analysis 30: 30414
Milne B T 1991 Lessons from applying fractal models to landscape patterns. In Turner M G and Gardner R H
(eds) Quantitative Methods in Landscape Ecology. New York, Springer-Verlag: 399416
Moellering H and Rayner N 1982 The dual axis Fourier shape analysis of closed cartographic forms. Cartographic Journal 19: 5359
ONeill R V, Krummel J R, Gardner R H, Sugihara G, Jackson B, DeAngelis D L, Milne B T, Turner M G,
Zygmunt B, Christensen S W, Dale V H, and Graham R L 1988 Indices of landscape pattern. Landscape
Ecology 1: 15362
Palanques M L and Calvo M 2011 Cadastre and colonial policies: A world heritage of cultures. In Proceedings
of the Empowerment of Local Authorities: Spatial Information and Spatial Planning Tools FIG Workshop,
Paris, France: 2528 (available at http://sites.google.com/site/figcom3/)
Pavlides T 1978 A review of algorithms for shape analysis. Computer Graphics and Image Processing 7: 24358
Prepata F P and Shamos M I 1985 Computational Geometry: An Introduction. New York, Springer-Verlag
Ricgard C and Hemani H 1974 Identification of the three-dimensional objects using Fourier descriptors of the
boundary curve. IEEE Transactions on Systems, Man and Cybernetics 4: 37178
Rosenfeld A and Kak A 1976 Digital Picture Processing. New York, Academic Press
Sagiv M, Reps T, and Wilhelm R 2003 Parametric shape analysis via 3-valued logic. ACM Transactions on Programming Languages and Systems 24: 21798
Sharifi A, Herwijnen M, and Toorn W 2004 Spatial Decision Support Systems. Enschede, The Netherlands,
International Institute for Geo-Information Science and Earth Observation
Simons P 1974 Measuring shape distortions of retail market areas. Geographical Analysis 6: 33140
Taylor P J 1971 Distances Within Shapes: An Introduction to a Family of Finite Frequency Distributions. Iowa
City, IA, University of Iowa Discussion Paper No 16
Tourino J, Boullon M, and Gonzalez X 2003 A GIS-embedded system to support land consolidation plans in
Galicia. International Journal of Geographical Information Science 17: 37796
Wentz E 1997 Shape analysis in GIS. In Proceedings of Auto-Carto 13, Washington, D.C.: 20413
Wentz E 2000 A shape definition for geographic applications based on edge, elongation, and perforation. Geographical Analysis 32: 95112
Williams S E and Pearson R G 1997 Historical rainforest contractions, localised extinctions and patterns of vertebrate endemism in the rainforests of Australias west tropics. Proceedings of the Royal Society of London
264: 70916
Witney B 1995 Choosing and Using Farm Machines. Edinburgh, Scotland, Land Technology Ltd
Zhang L, Huang H, Huang B, and Pingxiang L 2006 A pixel shape index coupled with spectral information for
classification of high spatial resolution remotely sensed imagery. Geoscience and Remote Sensing 44:
295061

2013 Blackwell Publishing Ltd

Transactions in GIS, 2013, ()

You might also like