Science and Justice 51 (2011) 77–82

Contents lists available at ScienceDirect

Science and Justice

j ourna l homepage: www.e lsev ie r.com/ locate /sc i jus

Characterization of soils from the Algarve region (Portugal): A multidisciplinaryapproach for forensic applications

Alexandra Guedes a,⁎, Helena Ribeiro a, Bruno Valentim a, Andreia Rodrigues a, Helena Sant'Ovaia a,Ilda Abreu b, Fernando Noronha a

a Centro de Geologia e DGAOT da Faculdade de Ciências da Universidade do Porto, Portugalb Centro de Geologia e Departamento de Biologia da Faculdade de Ciências da Universidade do Porto, Portugal

⁎ Corresponding author. Centro de Geologia da UniveFaculdade de Ciências, Rua do Campo Alegre, 687, 4169-220402474; fax: +351 220402490.

E-mail address: aguedes@fc.up.pt (A. Guedes).

1355-0306/$ – see front matter © 2010 Forensic Sciencdoi:10.1016/j.scijus.2010.10.006

a b s t r a c t

a r t i c l e i n f o

Article history:Received 9 June 2010Received in revised form 14 September 2010Accepted 29 October 2010

Keywords:Spectral colourParticle size distributionMagnetic susceptibilityPollen content

The Algarve is located at a very short distance from North Africa, in Southern Portugal, and as one of the mosttouristic regions of Portugal, it is accessible by air, land and sea. It is very susceptible to many illegal activities,such as illegal migration, drug trafficking, kidnapping, and murder, among others. Therefore, an Algarve soilsdatabase for forensic purposes is being conducted with the conjunction of geological and palynologicalmethodologies on soils characterization, since this is of fundamental importance to assess reliable evidence onforensic investigations.In this study, the properties of soils from several proximate sites from the Algarve were investigated, namely:(i) colour determined by spectrophotometry; (ii) particle size distribution determined by laser granulometry;(iii) low-field magnetic susceptibility by a susceptibility meter; and (iv) pollen content using a lightmicroscope. Finally, a hierarchical cluster analysis was applied to ascertain the capacity of the different soilproperties for discrimination between samples.The study reveals the utility of geobotanical techniques for forensic discrimination of soils. Even though somesimilarities between some of the samples were found, each one presented a combination of colour, particlesize distribution, magnetic susceptibility and pollen features that enable the determination of a fingerprintexpected to reveal a specific site for future selection of coastal search areas in the Algarve region.

© 2010 Forensic Science Society. Published by Elsevier Ireland Ltd. All rights reserved.

1. Introduction

Soil is a complex and heterogeneous material composed ofinorganic and organic matter. The ability to characterise theircomponents at an increasing level of detail makes soil an importanttrace evidence for forensic investigations. Several analytical methodsfor soil characterization have been well established in forensic science[1–15].

Also important in forensic investigations is the comparison ofquestioned samples with one or more samples of known originreferenced in a database, and the quantification of the degree ofsimilarity or dissimilarity observed between samples and its significance[1–4]. Nowadays, soil databases for forensic purposes are beingdeveloped, most of them in the UK [16,17] and in the USA [18] wheregeoforensics is relatively well advanced. Therefore, in Portugal, asediments/soils database for forensic purposes is being developed andthe sediments/soils analyzed and studied by the authors using differenttechniques. At the present time, this database has more than 200

rsidade do Porto e DGAOT da007 Porto, Portugal. Tel.: +351

e Society. Published by Elsevier Ire

geographically well-referenced samples from coastal Portugal. TheAlgarve samples used in our study, in addition to completing thereference soil database under development, also have the particularity ofbeing proximate and therefore ascertain if it will be possible todiscriminate them based on its geobotanical characteristics.

Thus, samples that are geographically well-referenced in thedatabase provide fast and relevant information for future selection ofsearch areas for forensic purposes. The database will provide valuablecontextual information on the characteristics of the samples collected[4] that are representative and that can be comparable, allowing forexclusion within the framework of sample variability rather thaninclusion [1].

The Algarve region has a vast seashore, it is a short distance fromNorth Africa, and has great tourist activity all year round, even in lowseason, a placewheremany foreign tourists come to enjoy the sunny andwarm Mediterranean climate, and its beautiful landscape and relaxedwalks on themany footpaths along the shore. These characteristicsmakethis region very susceptible to many illegal activities, such as illegalmigration, drug trafficking, kidnapping, and murder, among others. Onecanmention the famous and internationally spoken about disappearanceof Madeleine McCann in Praia da Luz-Algarve as an example of theforensic relevance of this region. Furthermore, every year the PortuguesePolice seize large amounts of drugs in Algarve's many coastal cliffs.

land Ltd. All rights reserved.

78 A. Guedes et al. / Science and Justice 51 (2011) 77–82

Therefore, these cliffs' superficial soil can be useful in investigativeintelligence, either as a known origin sample referenced in a databaseor via its retrieval from a suspect's clothing, footwear, vehicle, orcrime scene.

In forensic investigation, modern statistical methods are available tosupport and confirm impressions and interpretations of investigationsboth in the field and the laboratory [19,20]. Classificationmethods, suchas cluster analysis, allow for the identification of groups on the basis ofdissimilarity or similarity measurement [19] and can allow theformation of a potential link or unlink to a suspect or crime scene, aswell as the reduction of the possibility of false-positive results [12,15].

The scope of this study was focused on the characterization of somesoil properties of several proximate sites from Algarve, namely: colourquantification, particle size distribution, magnetic susceptibility andpollen content and to ascertain the potential of the tested characteristicsto discriminate samples within the same geographical area. This datawill contribute to a better knowledge of using different techniques andinterdisciplinary methods in soil characterization.

2. Materials and methods

2.1. Sample collection and handling

In February 2008, nine soil samples from nine different sites werecollected in the Algarve (Portugal) region (Fig. 1, Table 1). Thesesampling sites are located relatively close together, in the SouthPortuguese Meso-Cenozoic borderland, and correspond to cliff areas.

The bedrock of the studied soils consists of Jurassic and Cretaceouslimestones and/or marls (AG4, AG11, AG12, AG14, AG15 and AG20),Miocene carbonate formation (AG1 and AG22), and Pleistocene sand-stones (AG8). During soil sampling, a plant inventory of each sitewas alsoperformed. The sites are surrounded by dune forests, mainly Pinus spp.and Olea sylvestris-Querceto suberis sigmetum. The soil cover is dominatedor co-dominated by thorny shrub and bush communities. It also hasMediterranean salt grasslands, halophilous vegetation (Mediterraneanand thermo-Atlantic) and other halonitrophilous annual species.

In order to characterize the nine soil samples, and in the frameworkof a forensic science study, 200 to 400 g per sample of soil at each sitewere manually collected from the surface soil (b5 cm depth), with aplastic spade (carefully cleaned after each sampling) along transectsperpendicular to the coastline and put into a plastic bag, following theprocedures described by Saye and Pye [16] to develop a coastal soil

Fig. 1. Simplified geological map of Southern P

database for forensic applications. After, the samples were dried on astove at 40 °C and divided into two sub-samples, one was kept as aduplicate and stored into a plastic box kept in a cool place, and the otherused to perform several analyses: colour, particle size distribution,magnetic susceptibility and pollen content.

2.2. Colour analysis

Several experiments on soil colour description and comparison havebeen conducted throughout the years, comparing various soil presen-tation/pre-treatment methods prior to colour testing (air-drying,wetting, organic matter decomposition, iron oxide removal, ashing,size fraction separation and milling) [7,13,21–23]. Variations in theL*a*b* values between different presentation/pre-treatment methodswere demonstrated from the results obtained.

Guedes et al. [13] demonstrated that the measured L*a*b* values ondried, unsieved bulk samples allowed for higher discriminationbetween samples than measures performed on other presentation/pre-treatment methods. Croft and Pye [7] suggested that colour shouldbe measured on the bulk material and on a specific size fractionaccording to the soil nature.

Colour measurements were performed on bulk samples using aKonica Minolta CM-2600d spectrophotometer programmed with thefollowing settings: measurement area of 0.8 mm of diameter; specularcomponent included; CIE Standard IlluminantD65 corresponding to theaverage daylight at a temperature of 6504 K, including the ultravioletwavelength region; CIE 1964 Standard Observer (10° Observer). Thespectrophotometer was set to take three sequential measurements,giving the colour coordinates obtained means.

Before measurement, the spectrophotometer was calibratedaccording to the manufacturers' instructions and regularly calibratedthroughout themeasurement period due to the repeated nature of themeasurements carried out. Negative calibration was performeddirecting the apparatus measuring port into the air while positivecalibration was performed with an international standard whitecalibration plate. Sample material was homogenised and presented ina standard glass petri dish and an average of five to eight sub-samplestaken from the original sampled collected at each site were measured.

The colour parameters recorded correspond to the uniform colourspace CIELAB and were directly computed by the spectrophotometerthrough the SpectraMagic NX software.

ortugal and sampling locations at Algarve.

Table 1Sampling sites location, underlying geology and surrounding vegetation.

Sample Location Underlying geology Surrounding vegetation

AG1 Access to Portimão's Marine Miocene carbonate formation Dune forests mainly of Pinus spp. andOlea sylvestris-Querceto suberis sigmetum.AG4 Top of cliff Jurassic and Cretaceous limestones

and/or marlsAG8 Situated between railway line and S. Roque

beach (1200 m)Pleistocene sandstones

AG11 Top of cliff in Marinha Beach Resort Residence,in Sagres

Jurassic and Cretaceous limestonesand/or marls

Soil cover dominated or codominated bythorny shrub and bush communities.

AG12 Top of cliff in Ponta de Sagres, between Rua daFortaleza (50 m) and EN 268-2 (200 m)

Jurassic and Cretaceous limestonesand/or marls

AG14 Top of cliff in Mareta beach in Sagres, betweenRua do Infante

Jurassic and Cretaceous limestonesand/or marls

D. Henrique (250 m) and Rua ComandanteMatoso (175 m)

Mediterranean salt grasslands, halophilousvegetation (Mediterranean and thermo-Atlantic),other halonitrophilous annual species.AG15 Top of cliff near EM 1257 and Praia da Ingrina Jurassic and Cretaceous limestones

and/or marlsAG20 Top of cliff in Burgau beach between Rua 25 de Abril(100m) and Largo dos pescadores (40 m)

AG22 Top of cliff in Almagem beach, in Loulé Miocene carbonate formation

79A. Guedes et al. / Science and Justice 51 (2011) 77–82

2.3. Particle size distribution

This is a technique that allows sample weights of only 50 mg to beanalysed [2,4] and the results obtained can give important informationto infer about soil or sediments nature and provenance.

Wet sieving was conducted for particle size distribution, using a set ofRetsch® stainless steel sieves. The b63 μmsize fractionwas automaticallydetermined using a Coulter LS130 Laser Beam Granulometer, coupledwitha liquidmodule, at LNEG-National LaboratoryofEnergyandGeology(Porto, Portugal) and the evaluation of the accuracy and precision of thismethod was regularly tested.

2.4. Magnetic susceptibility

Magnetic susceptibility is directly proportional to the quantity andgrain size of ferromagnetic materials in the sample which can beoriginated by desegregation of parent rocks during pedogenesis, bylithogenic processes, and by anthropogenic activities. The accuracyand precision of this analytical method is regularly tested.

Magnetic susceptibility is a measure of the magnetic response of amaterial to an external magnetic field. The specific or masssusceptibility χ, measured in units of m3/kg, is defined as the ratioof the material magnetization J (per unit mass) to the weak externalmagnetic field H: J=χ H. Magnetic susceptibility determination wasperformed with a Kappabridge, model KLY-4S of Agico balanceequipped with the Sumean software. Measures were performed on1 g of each sample after homogenisation, the magnetic susceptibility(χ) being calculated.

2.5. Pollen analysis

For palynological studies, 5 g of soil were removed from the sub-sample. Afterwards, samples were dried, sieved at 200 μm through adisposable mesh and a series of chemical procedures were carried out,aiming to obtain rich pollen residue for qualitative and quantitativeanalysis based on the techniques outlined in a technical note publishedby Horrocks [24] where a guide for sub-sampling and preparingforensic samples for pollen analysis is presented.

Subsequently, the pollen residues were mounted on microscopeslides in glycerol jelly. It was necessary to count more than tworeplicates for some samples, up to a maximum of four replicate slides,depending on the total number of pollen grains registered, to obtainrepresentative pollen counts (even though one sample still presentedpollen counts lower than 100). The sum of the replicate counts wasconsidered for each sample. Pollen qualification and quantificationwas carried out using a light microscope at a magnification of ×400

along ten equidistant full lengthwise traverses for every slide. Pollencounts for each type identified were then converted into percentagesof the total pollen counts. Pollen grains were classified by appearanceand morphological characteristics and identified, where possible, bycomparison with bibliographic material [25–27].

2.6. Statistical analysis

Appropriate descriptive statistical analysis for each measuredproperty was performed. Reproducibility within-sample was evaluatedby the coefficient of the variation in terms of colour analysis,with L*a*b*colour parameters, and magnetic susceptibility.

A hierarchical cluster analysis was performed in order to ascertainif it is possible to obtain discrimination between samples bycombining the results of colour, particle size distribution, magneticsusceptibility and pollen analysis. The number of clusters wasdetermined using: i) the Euclidean distance as a distance measureand ii) the Ward method as a linking method. The statistical analysissoftware that was used for all the analysis was SPSS (16.0).

3. Results and discussion

3.1. Colour analysis

Concerning the L*a*b* system colour sphere (Table 2), L* valuesmeasured varied between 74.14 and 46.23 (samples AG14 and AG15,respectively); a* varied between 12.95 and 2.73 (samples AG22 andAG14, respectively); b* between 20.00 and 10.27 (samples AG22 andAG20, respectively).

The coefficient variation of L*a*b* values obtained were alwayslower than 5%, giving measure reproducibility. The lowest values wereobserved on L*, while the parameter with the highest variationfor all samples was a* (Table 2). Guedes et al. [13] also reportedthe a* parameter as the one with highest overall within-samplevariation.

When analysing the chromaticity diagrams (Fig. 2), whichrepresent the relationship between a* and b* measures, it wasobserved that a* and b* always presented positive values, indicatingthat samples are positioned in the saturation zone closest to the redand yellow continuums.

Reflectance values over the 400–700 nmrangewere also compared(Fig. 3) for the studied samples. Within the latter set, samples AG14,AG15, andAG20may be discriminated, since they clearly showdistinctreflectance curves.

Table 2Algarve soil samples: descriptive statistics of L*a*b* indices, particle size, and magnetic susceptibility parameters of each sample.

CV — Coefficient of variation.

Colour measures Particle size distribution (μm) MagneticSusceptibility(×10−8 m3/kg)

L* a* b*

Samples Mean CV% Mean CV% Mean CV% Mean Median Mode D10 D25 D75 D90 D90 - D10 Value CV%

AG1 58.77 0.2 10.68 1.5 19.49 1.7 26.44 19.39 38.91 1.83 7.38 38.92 59.92 58.09 14.69 0.5AG4 57.13 0.4 9.18 0.3 16.61 0.7 25.46 20.30 35.52 1.61 7.09 37.52 55.24 53.63 24.93 0.5AG8 55.79 0.3 7.94 1.0 16.91 0.5 39.31 32.92 56.00 3.32 12.46 55.47 76.00 72.68 6.09 3.5AG11 53.63 0.5 5.95 1.4 13.08 1.7 26.91 24.78 32.43 2.61 12.06 38.79 53.10 50.49 124.80 0.3AG12 56.32 0.5 10.89 2.1 19.38 2.0 23.85 19.98 27.03 1.95 9.01 32.71 47.95 46.00 9.47 0.8AG14 74.14 0.5 2.73 1.5 15.54 0.5 100.90 49.48 73.59 2.16 10.89 101.20 245.80 243.64 3.53 3.1AG15 46.23 0.3 11.18 1.7 11.24 0.7 30.55 28.48 35.52 3.87 15.44 42.04 55.91 52.04 141.51 0.3AG20 59.82 0.6 4.47 2.6 10.27 2.6 25.45 20.34 42.62 1.48 6.37 40.18 57.24 55.76 4.95 1.2AG22 58.94 1.6 12.95 4.4 20.00 1.5 20.76 14.95 18.78 1.92 6.09 26.70 47.82 45.91 10.73 1.4

80 A. Guedes et al. / Science and Justice 51 (2011) 77–82

3.2. Particle size distribution

Particle size descriptive parameters (Table 2) showed a mean sizedistribution between 20.76 μm and 100.90 μm; a mode between18.78 μm and 73.59 μm; a median between 14.95 μm and 49.48 μm;and, the soil sorting varies between 45.91 μm and 141.51 μm. Thelowest values of these parameters were always observed in sampleAG22, and the highest in sample AG14.

Fig. 4 compares particle size distribution curves, determined bylaser diffraction, obtained from the different samples. With theexception of curves from samples AG8 and AG14, with particle sizedistribution patterns mainly located in the very coarse silt class, theother samples have similar curve shape.

3.3. Magnetic susceptibility

The MS values range between 3.53×10−8 m3/kg and 141.51×10−8

m3/kg, with the lowest values obtained in sample AG14 and the highestin AG15. The coefficient variation obtained was always lower than 4%,giving measure reproducibility (Table 2).

3.4. Plant inventory and pollen analysis

A plant inventory of each site was performed (Table 3) for furtherhelping in the pollen identification.

The inventory showed qualitative vegetation similarity betweenall sampling sites, with differences occurring in representativeness ofthe several species. This was reflected on the results of pollen analysisof soils. Thirty-eight pollen-types were identified in the nine samplescollected (Table 4), but only 10 pollen-types were present in allsamples: Asteraceae, Brassicaceae, Chenopodiaceae–Amaranthaceae,Cyperaceae, Erica spp., Pinus spp., Plantago spp., Poaceaea, Quercusspp. and Urticaceae.

AG4

AG11

AG14

AG15AG20

AG22AG1

AG8 AG12

5

10

15

20

25

2 4 6 8 10 12 14 16

a

b

Fig. 2. Chromaticity diagrams: scatter plot of a* and b* values measured on differentstudied samples.

Samples AG1 and AG11 were dominated by Asteraceae (majority ofLiguliflora type), Chenopodiaceae–Amaranthaceae (majority of Betavulgaris type) and Poaceae, although with frequency differences. AG1also presented considerable amounts of Brassicaceae pollen and morepollen type's diversity; while AG11 presented Plantago spp. pollen.

Representing more than half of the pollen assemblages, Chenopo-diaceae–Amaranthaceae (majority of Beta vulgaris type) dominated insamples AG4 and AG20. Sample AG4 also presented abundant pollen ofAsteraceae (majority of Liguliflora type), while in AG20 it was Plantagopollen.

SamplesAG12andAG14weredominatedby thepollen typePoaceae,Plantago spp. and Asteraceae, although in AG12, themajority Asteraceaewas of the Tubuliflora type, while in AG14 it was of Liguliflora.

Sample AG15 was dominated by pollen Plantago spp., Poaceae andAsteraceae (equivalent representativeness of Liguliflora and Tubulifloratypes). Finally, representinghalf of thepollen assemblages,Plantago spp.dominated in sample AG8. This sample also had the particularity ofpresenting considerable pollen from Pinus spp.

Sample AG22 was palynologically very poor, yielding relativelysparse palynomorphs and, therefore,wasnot considered in this analysis.

Although all samples contain qualitative similarities in terms ofdominant pollen types, each one presented a characteristicfingerprint,being likely to reflect some combination of pollen from a specificlocation. As an example, samples AG12 and AG14 separated byapproximately 360 m have composition differences that enabledsample discrimination, pointing out that localized areas of similar

Fig. 3. Reflectance curves (400–700 nm) for the different studied samples.

Fig. 4. Particle size distribution obtained on the different studied samples.

Table 4Percentage frequencies of pollen assemblages observed in the 9 surface soil samples ofAlgarve.

Sample frequencies (%)

Pollen types AG1 AG4 AG8 AG11 AG12 AG14 AG15 AG20 AG22

Present in all samplesAsteraceae 18.4 20.8 10.7 23.0 19.5 32.4 15.4 5.4 12.2Liguliflora 15.6 11.9 6.5 16.6 5.7 25.1 7.1 3.7 3.7Tubuliflora 2.5 8.8 4.1 6.1 13.8 7.3 8.3 1.6 8.5Brassicaceae 13.6 2.3 2.8 2.5 1.2 1.4 3.5 0.2 2.4Raphanus 7.7 – 1.6 1.0 – – – – –

Lobularia – 1.4 1.3 – 0.8 0.5 0.3 – –

Chen-Amar 21.2 61.1 2.4 28.5 6.0 4.1 4.2 74.5 3.7Beta vulgaris 19.1 60.0 – 27.9 3.5 4.1 4.2 73.7 3.7Chenopodium 2.1 1.0 2.4 0.6 2.5 – – 0.8 –

Cyperaceae 0.3 0.2 0.4 0.1 0.5 0.5 0.3 0.2 1.2Erica spp. 0.1 0.1 0.0 0.1 0.4 0.9 0.3 0.1 2.4Pinus spp. 2.5 2.3 13.7 1.2 6.0 5.9 3.5 1.6 19.5Plantago spp. 9.9 1.0 51.0 19.4 15.6 10.5 21.2 9.5 13.4Poaceae 22.2 8.6 11.0 22.4 40.5 25.1 20.8 4.3 11.0Quercus spp. 1.6 1.0 0.2 0.3 1.3 0.5 4.2 0.7 6.1

81A. Guedes et al. / Science and Justice 51 (2011) 77–82

vegetation, even within the same geographical region, can havesignificantly different pollen assemblages [5].

Urticaceae 2.9 1.2 1.6 1.3 0.4 11.0 9.3 1.4 14.6

Absent in at least one sampleUmbelifera 1.4 tr tr tr 0.7 0.5 0.6 tr –

Juniperus spp. – tr tr tr 1.8 2.7 7.7 tr 2.4Olea spp. 2.1 tr tr tr 1.2 0.9 – 0.9 3.7Lamiaceae 0.4 tr – tr 0.8 – 0.6 tr –

Pistacia spp. 0.6 tr tr 0.6 1.6 – 4.5 – –

Alnus spp. 0.5 tr tr tr tr 2.7 – – 1.2Fabaceae 0.6 – tr – 0.8 – – tr 2.4

3.5. Cluster analysis

In the cluster analysis (Fig. 5), colour, particle size distribution,magnetic susceptibility and pollen analysis were important to findbetween-site dissimilarities. The clustering of more similar sampleswas considered in the following groups with linkage distances higher

Table 3Plant inventory from sampling sites in Algarve, south of Portugal.

Plant

Family Species Family Species

Aizoaceae Carpobrotus edulis Fabaceae Ononis spp.Alliaceae Narcissus

bulbocodiumCytisus scoparius

Amaranthacae Beta vulgaris Erophaca baeticaAnacardiaceae Pistacia lentiscus Ulex spp.Araceae Arisarum vulgare Fagaceae Quercus spp.Asphodelaceae Asphodelus spp. Frankeniaceae Frankenia spp.Asteraceae Calendula arvensis Fumariaceae Fumana spp.

Carlina spp. Geraniaceae Erodium cicutariumSonchus oleraceus Erodium malacoidesCentaurea pullata Gramineae Ordeum spp.Crysanthemumcoronarium

Gramineae

Senecio vulgaris Lamiaceae Rosmarinus officinalisDittrichia viscosa Salvia spp.Belis anua Teucrium spp.Anacyclus radiatus Lavandula spp.Heliechrysum italicum Phlomis purpureaArctotheca calendula Liliaceae Asparagus albus

Borraginaceae Echium plantaginium Malvaceae Lavatera arboreaLithodora prostrata Myrtaceae Eucalyptus spp.

Brassicaceae Lobularia maritimum Moraceae Ficus caricaRaphanus raphanistrum Oleaceae Olea europaea

Caryophylaceae Oxalidaceae Oxalis pes-capraeChenopodiaceae Chenopidium murale Plantaginaceae Plantago spp.Cistaceae Halimium spp. Pinaceae Pinus pinaster

Cistus salvifolius Rhamnaceae Rhamus oleoidesCistus ladaniferus Rosaceae Prunus dulcisCistus monspeliensis Rubiaceae Galium spp.

Convovulaceae Convovulus althacoides Scrophulariaceae Myoporum acuminatumCrassulaceae Sedum sediforme Umbeliferae Margotia gummiferaCupressacea Juniperos spp. Ridolfia sagetumEuphorbiaceae Euphorbia spp. Daucus spp.Fabaceae Acacia longiflia Eryngium maritimum

Acacia saligna Urticaceae Urtica menbranaceaeLotus ceticus Parietaria spp.Medicago spp. Valerianaceae Fedia cornucopiaeMedicago marina Violoceae Viola arborescensMedicago polymorpha

Medicago 0.5 – – – – – – – 2.4Ononis – – tr – 0.5 – – – –

Cistaceae tr tr tr – tr – 1.6 – 2.4Rumex spp. tr tr – – tr – tr tr –

Eucalyptus spp. 0.9 tr – – tr 0.9 – – –

Rosaceae tr tr tr – tr – – – –

Betulaceae tr tr tr – tr – tr tr –

Typha spp. – tr – tr – – – tr –

Geraniaceae – – tr tr – – – tr –

Solanaceae tr tr – – – – – tr 1.2Fraxinus spp. tr – – – – – – – –

Acacia spp. – – tr tr – – – – –

Thalictrum – – 5.0 – – – – – –

Fedia spp. – tr – – – – – – –

Total pollen counts for each sample (pollen grains)2221 6814 2765 3138 1956 219 312 1639 82

–: pollen type absent from sample; tr: present pollen types with frequencies lower than0.5%; Chen-Amar: Chenopodiaceae–Amaranthaceae.

than 10: i) samples AG4 and AG20; ii) samples AG11 and AG15;iii) samples AG1, AG12 and AG8; iv) sample AG22; and v) sample AG14.

Samples AG22 and AG14 were clearly distinguishable betweenthem and from all other samples, presenting high linkage distancebetween each other and other groups. Some distinctive features ofAG14 were the lowest magnetic susceptibility values and the highest

Fig. 5. Hierarchical cluster dendrogram combining the results of colour, particle sizedistribution, magnetic susceptibility and pollen analysis observed in the 9 soil samplescollected in the South of Portugal.

82 A. Guedes et al. / Science and Justice 51 (2011) 77–82

particle size descriptive parameters. On the other hand, sample AG22presented the lowest values of particle size descriptive parametersand very poor pollen content.

In the cluster of AG1, AG12 and AG8, the latter is distinguishedfrom the other two mainly due to having quite a different pollenfingerprint and higher particle size descriptive parameters. SamplesAG1 and AG12 present similar colour and particle size distributionvalues, however magnetic susceptibility values were quite differentand pollen fingerprint, although with some similarities, presentedsome differences, mainly in the representativeness of pollen fromChenopodiaceae–Amaranthaceae and Poaceae.

The cluster formed by samples AG11 and AG15 (geographicallyclose) can result from the presence of similar colour, particle sizedistribution, magnetic susceptibility values. However, they presentdifferences in pollenfingerprint suchas the representativeness of pollenfrom Liguliflora vs Tubuliflora; Chenopodiaceae–Amaranthaceae orPoaceae and the considerable presence in the sample AG15 of Juniperusspp., Pistacia spp. and Cistaceae pollen.

Finally, samples AG4 and AG20 were the ones presenting theclosest values for the evaluated characteristics, having a linkagedistance lower than 5. They present as distinct feature magneticsusceptibility and colour parameters a* and b* values.

4. Conclusions

The study reveals the utility of geobotanical techniques fordiscrimination of soils. In the case of the nine samples analysed inour study, even though geographically very close and having somesimilarities, each one presented a combination of colour, particle sizedistribution, low-fieldmagnetic susceptibility and pollen features thatstill enable its discrimination. Thus, the use of a multidisciplinaryapproach for the determination of a fingerprint for each site, expectedto reveal a specific location, is important in forensic investigation forfuture selection of coastal search areas and comparison of questionedsamples with one or more samples of known origin referenced in adatabase of the Algarve region. This information can be very useful asinvestigative intelligence associating a person to a crime scene,excluding the provenance of a questionable sample when its origin isunknown.

Acknowledgements

This work has been financially supported by the project PTDC/CTE-GEX/67442/2006 and by the Pos-Doc scholarship SFRH/BPD/43604/2008-POPH-QREN (FCT-Portugal). The authors would like to thankPaulo Alves from CIBIO for his valuable help for the plant's inventoryand to the three anonymous reviewers for their help in improving theearlier draft of this paper.

References

[1] R.M. Morgan, P.A. Bull, The philosophy, nature and practice of forensic sedimentanalysis, Prog. Phys. Geog. 31 (1) (2007) 43–58.

[2] K. Pye, Geological and soil evidence: forensic applications, CRC, Boca Raton,London, 2007.

[3] K. Pye, S.J. Blott, Development of a searchable major and trace element databasefor use in forensic soil comparisons, Sci. Justice 49 (3) (2009) 17–181.

[4] L.A. Dawson, S. Hillier, Measurement of soil characteristics for forensic applications,Surf. Interface Anal. 42 (5) (2010) 363–377.

[5] M. Horrocks, S.A. Coulson, K.A.J. Walsh, Forensic palynology: variation in thepollen content of soil surface samples, J. Forensic Sci. 43 (2) (1998) 320–323.

[6] R. Sugita, Y. Marumo, Screening of soil evidence by a combination of simpletechniques: validity of particle size distribution, Forensic Sci. Int. 122 (2–3)(2001) 155–158.

[7] D.J. Croft, K. Pye, Colour theory and evaluation of an instrumental method ofmeasurement using geological samples for forensic applications, in: K. Pye, D.J.Croft (Eds.), Forensic Geoscience: Principles, Techniques and Applications,Geological Society Special Publications, London, 2004, pp. 49–62.

[8] L.A. Dawson,W. Towers, R.W.Mayes, J. Craig, K. Vaisanen, E.C.Waterhouse, The use ofplant hydrocarbon signatures in characterizing soil organicmatter, in: K. Pye, D.J. Croft(Eds.), Forensic Geoscience: Principles, Techniques and Applications, GeologicalSociety Special Publications, London, , 2004, pp. 269–276.

[9] K. Pye, S.J. Blott, Particle size analysis of sediments, soils and related particulatematerials for forensic purposes using laser granulometry, Forensic Sci. Int. 144 (1)(2004) 19–27.

[10] K. Pye, S.J. Blott, D.J. Croft, J.F. Carter, Forensic comparison of soil samples:assessment of small-scale spatial variability in elemental composition, carbon andnitrogen isotope ratios, colour, and particle size distribution, Forensic Sci. Int. 163(1–2) (2006) 59–80.

[11] D.C. Mildenhall, P.E.J. Wiltshire, V.M. Bryant, Forensic palynology: why do it andhow it works, Forensic Sci. Int. 163 (3) (2006) 163–172.

[12] A. Ruffell, J. Mckinley, Geoforensics, John Wiley & Sons, Ltd, Belfast, UK, 2008.[13] A. Guedes, H. Ribeiro, B. Valentim, F. Noronha, Quantitative colour analysis of

beach and dune sediments for forensic applications: a Portuguese example,Forensic Sci. Int. 190 (1–3) (2009) 42–51.

[14] C. Manrong, Y. Lizhong, N. Xiangfeng, C. Bin, Application of environmentalmagnetism on crime detection in a highway traffic accident from Yangzhou toGuazhou, Jiangsu Province, China, Forensic Sci. Int. 187 (1–3) (2009) 29–33.

[15] RitzK. , DawsonL. , MillerD. , Criminal and Environmental Soil Forensics, Springer,2009.

[16] SayeS.E. , PyeK. , Development of a Coastal Dune Sediment Database for EnglandandWales: Forensic applications, in: PyeK. , CroftD.J. (Eds.), Forensic Geoscience:Principles, Techniques and Applications, Geological Society Special Publications,London, 2004, pp. 75–96.

[17] Macaulay Land Use Research Institute, SoilFUN - Soil Forensics University Network(http://www.macaulay.ac.uk/forensics/soilfun/); SoilFIT - Integration of Soil Finger-printing Techniques for Forensic Applications (http://www.macaulay.ac.uk/soilfit/);GIMI - Geoforensics and Information Management for crime Investigation (http://www.macaulay.ac.uk/geoforensic/), 2009.

[18] MenchacaP. , GrahamR. , StamM. , Update on the development of a searchableforensic soil database (SQUID) in California, USA, 2nd International Conference onEnvironmental and Criminal Soil Forensics, Edimburgh, 2007, p. 24.

[19] D. Lucy, Introduction to statistics for forensic scientists, 1st ed. JohnWiley & Sons,Ltd, West Sussex, 2006.

[20] R. Webster, Statistics to support soil research and their presentation, Eur. J. SoilSci. 52 (2) (2001) 331–340.

[21] R.J. Dudley, The use of colour in the discrimination between soils, J. Forensic Sci.Soc. 15 (3) (1975) 209–218.

[22] D.W. Janssen, W.A. Ruhf, W.W. Prichard, The use of clay for soil colour comparisons,J. Forensic Sci. 28 (3) (1983) 773–776.

[23] R. Sugita, Y. Marumo, Validity of colour examination for forensic soil identification,Forensic Sci. Int. 83 (3) (1996) 201–210.

[24] M. Horrocks, Sub-sampling and preparing forensic samples for pollen analysis,J. Forensic Sci. 49 (5) (2004) 1024–1027.

[25] M. Reille, Pollen et spores d'Europe et d'Afrique du Nord, Éditions du LaboratoireDe Botanique Historique Et Palynologie, Marseille, 1992.

[26] M. Reille, Pollen et spores d'Europe et d'Afrique du Nord (Supplément 1), Éditionsdu Laboratoire de Botanique Historique et Palynologie, Marseille, 1995.

[27] M. Reille, Pollen et spores d'Europe et d'Afrique du Nord (Supplément 2), Éditionsdu Laboratoire de Botanique Historique et Palynologie, Marseille, 1998.