to
l
Va
Irel
g, S
Article history:
Received 7 December 2007
1. Introduction
hotspots of pollution. Geographical information system (GIS)
significant. GIS mapping techniques can help to identify
popularly used (Anselin, 1995; Getis and Ord, 1996). While the
S C I E N C E O F T H E T O T A L E N V I R O N M E N T 3 9 8 ( 2 0 0 8 ) 2 1 2 – 2 2 1
ava i l ab l e a t www.sc i enced i rec t . com
m/mapping andmultivariate analyses are useful tools to help the
identification of spatial patterns of pollution and possible
pollution sources can be evaluated (Zhang, 2006). Further-
more, hotspots-areas where there are high levels of pollution
in comparison to the surrounding area-need to be identified
global Moran's I (Cliff and Ord, 1981; Odland, 1988; Zhang and
Selinus, 1998) is a global parameter for the measurement of
spatial autocorrelation, the local Moran's I index examines
the individual locations, enabling hotspots to be identified
based on a comparison with the neighbouring samples. TheUrban geochemistry has received wide attention in recent
years (Zhang, 2006) resulting in databases containing a large
number of soil samples. Due to the spatial heterogeneity of
urban soils, the challenge is to identify spatial patterns and
hotspots visually, but not statistically. Among a few methods
proposed for hotspot or spatial cluster identification, such as
Getis's G index (Getis and Ord, 1992), spatial scan statistics
(Ishioka et al., 2007) and Tango' C index (Tango, 1995; Zhang
and Lin, 2006), the local Moran's I index seems to be the mostin order to provide a scientific basis for be
management.
In urban soil pollution studies, it is imp
there are hotspots and (2) if the hotspo
⁎ Corresponding author. Fax: +353 91 495505.
E-mail address: Chaosheng.Zhang@nuigalw
0048-9697/$ – see front matter © 2008 Elsevi
doi:10.1016/j.scitotenv.2008.03.011Spatial cluster
Spatial outlier
Urban soil
Local Moran's IPollution hotspots in urban soils need to be identified for better environmental
management. It is important to know if there are hotspots and if the hotspots are
statistically significant. In this study identification of pollution hotspots was investigated
using Pb concentrations in urban soils of Galway City in Ireland as an example, and the
influencing factors on results of hotspot identification were investigated. The index of local
Moran's I is a useful tool for identifying pollution hotspots of Pb pollution in urban soils, and
for classifying them into spatial clusters and spatial outliers. The results were affected by
the definition of weight function, data transformation and existence of extreme values.
Compared with the results for the positively skewed raw data, the transformed data and
data with extreme values excluded revealed a larger area for the high value spatial clusters
in the city centre. While it is hard to decide the best way of using this index, it is suggested
that all these influencing factors should be considered until reasonable and reliable results
are obtained. GISmapping can be applied to help evaluate the results via visualization of the
spatial patterns. Meanwhile, selected pollution hotspots (extreme values) in this study were
confirmed by re-analyses and re-sampling.
© 2008 Elsevier B.V. All rights reserved.Received in revised form
6 March 2008
Accepted 11 March 2008
Available online 28 April 2008
Keywords:
HotspotA R T I C L E I N F O A B S T R A C TUse of local Moran's I and GIS
Pb in urban soils of Galway, Ire
Chaosheng Zhanga,⁎, Lin Luob, Weilin Xub,
aDepartment of Geography, National University of Ireland, Galway,
bState Key Laboratory of Hydraulics and Mountain River Engineerin
www.e l sev i e r. cotter environmental
ortant to know: (1) if
ts are statistically
ay.ie (C. Zhang).
er B.V. All rights reservedidentify pollution hotspots of
and
lerie Ledwitha
and
ichuan University, Chengdu, China
l oca te / sc i to tenvlocal Moran's I index has been successfully applied in hotspo
identification of diseases (Ruiz et al., 2004; Goovaerts and
Jacquez, 2004), mortality rates (James et al., 2004; McLaughlin
and Boscoe, 2007), aswell as in environmental planning (Brody
.t

et al., 2006) and environmental sciences (McGrath and Zhang,
2003; Zhang and McGrath, 2004).
Raw samples were sent to OMAC Laboratories in Loughrea,
S C I E N C E O F T H E T O T A L E N V I R OCounty Galway, and dried at 40 °C. The b2 mm part of all
samples were retained, and half of the sample splits were
milled to pass through a 0.1 mm pore size sieve. The milled
samples of 0.2 g were digested to dryness using a 4-acid di-
gestionwith 10ml HF, 5ml HClO4, 2.5ml HCl, and 2.5ml HNO3,
1 In all maps in this study, the 500×500 m grid system of the
Galway Street Map is adopted for easy reference of geographicalThis study identifies pollution hotspots by lead (Pb) in the
urbansoil ofGalwayCity (Ireland)using the localMoran's I index
and GIS. In addition, the influences of weight function, data
transformation, and extreme values on the results of hotspot
identification using local Moran's I index are investigated.
2. Materials and methods
2.1. Study area
Galway is located on the western coast of Ireland. The study
area extends 9 km E–W and 6 km N–S in Galway City (Zhang,
2006). It was chosen based on the 2nd edition of the Galway
Street Map available from Ordnance Survey Ireland (OSI),
excluding 1 km in the west and 1.5 km in the east that are
mainly rural areas1. The bedrock types are limestone in the
east and granite in the west (Zhang, 2006). The natural soil
type in the study area is mainly grey brown podzols. In the
granite area, there are small areas of lithosol with poorly
developed soils.
The built-up areas are mainly located in the city centre,
extending about 3 km to thewest, 3 km to the east, and 3 km to
the north. The urban land uses are primarily for residential
and commercial purposes. There are several new industrial
estates in the eastern part of the city, but these are mainly
high-technology industries with little discharge of traditional
pollutants. The main pollution sources in the study area are
traffic and the burning of peat and coal for home heating
(Zhang, 2006). There is also evidence of dumping in parts of
the city that are now used for other purposes. For example,
Carr et al. (2008) found evidence of rubbish in an area currently
used as a sports ground.
2.2. Sampling and chemical analyses
A total of 166 surface soil samples (0–10 cm depth) were taken
from parks and grasslands in Galway city during Nov. 1–Dec.
16, 2004 (Zhang, 2006). About 1 kg of each soil sample was
collected using a stainless steel spade and a plastic scoop,
and fresh samples taken from a single hole with an area of
about 30×30 cm were placed in clean bags. The coordinates of
sampling locations were recorded using a differential Global
Position System (GPS) receiver. The sampling density was 1
sample per 0.25 km2 based on the grid system of Galway Street
Map from OSI using a stratified random sampling scheme. No
samples were taken at grids where access was hard to obtain.locations, e.g., “L11” represents the location of the Lth row and
11th column.then dissolved in 20% aqua regia and made up to 10 ml for ICP-
AES analysis for a total of 26 chemical elements. Certified
reference sampleswere used for quality control, and the errors
were generally better than 5%.
It was found the Pb was one of the main pollutants in
Galway urban soils (Zhang, 2006). In this study, the chemical
element of Pb is chosen as an example for detailed investiga-
tion of the topic of pollution hotspot analyses.
2.3. Spatial cluster and spatial outlier analyses
Pollution hotspots can be clustered (spatial clusters) or exist
individually (spatial outliers). In this study, spatial clusters of
pollution would be soil samples with a high Pb concentration
surrounded by other samples with a high concentration. In
contrast, spatial outliers of pollution would be samples with a
high Pb concentration surrounded by samples with normal or
low values. They can be identified using the local Moran's I
index (Anselin, 1995; Getis and Ord, 1996; Levine, 2004):
Ii ¼
zi  z
P
r
2
X
n
j¼1;jpi
wij zj  z
P
  
ð1Þ
where zi is the value of the variable z at location i; z– is the
average value of z with the sample number of n; zj is the value
of the variable z at all the other locations (where j≠ i); σ2 is the
variance of variable z; and wij is a weight which can be defined
as the inverse of the distance dij among locations i and j. The
weight wij can also be determined using a distance band:
samples within a distance band are given the same weight,
while those outside the distance band are given the weight
of 0.
A high positive local Moran's I value implies that the loca-
tion under study has similarly high or low values as its neigh-
bours, thus the locations are spatial clusters. Spatial clusters
include high–high clusters (high values in a high value neigh-
bourhood) and low–low clusters (low values in a low value
neighbourhood) (Fig. 2). In soil pollution, low–low clusters are
“cool spots”, while high–high spatial clusters can be regarded
as “regional hotspots”.
A high negative local Moran's I value means that the loca-
tion under study is a spatial outlier. Spatial outliers are those
values that are obviously different from the values of their
surrounding locations (Lalor and Zhang, 2001). Spatial outliers
include high–low (a high value in a low value neighbourhood)
and low–high (a low value in a high value neighbourhood)
outliers (Fig. 1). In soil pollution, high–low spatial outliers can
be regarded as isolated “individual hotspots”.
Local Moran's I can be standardised so that its significance
level can be tested based on an assumption of a normal dis-
tribution (Anselin, 1995; Levine, 2004). However, since the
probability distribution of local Moran's I may not necessarily
be normal, especially when the raw data are heavily skewed, a
method called “conditional permutation” (Anselin, 1995) is
preferred as it makes no assumption about the data. Under a
conditional permutation, when the value on a location is being
assessed, its value is fixed and all the other values are shuffled
randomly on all the other locations. Each time when the other
213N M E N T 3 9 8 ( 2 0 0 8 ) 2 1 2 – 2 2 1values are shuffled, the local Moran's I index is calculated to
form a reference distribution. The significance level can be