Source apportionment of groundwater pollutants in Apulian agricultural sites using multivariate statistical analyses: case study of Foggia province
© Ielpo et al; licensee BioMed Central Ltd. 2012
Published: 2 May 2012
Ground waters are an important resource of water supply for human health and activities. Groundwater uses and applications are often related to its composition, which is increasingly influenced by human activities.
In fact the water quality of groundwater is affected by many factors including precipitation, surface runoff, groundwater flow, and the characteristics of the catchment area. During the years 2004-2007 the Agricultural and Food Authority of Apulia Region has implemented the project “Expansion of regional agro-meteorological network” in order to assess, monitor and manage of regional groundwater quality. The total wells monitored during this activity amounted to 473, and the water samples analyzed were 1021. This resulted in a huge and complex data matrix comprised of a large number of physical-chemical parameters, which are often difficult to interpret and draw meaningful conclusions. The application of different multivariate statistical techniques such as Cluster Analysis (CA), Principal Component Analysis (PCA), Absolute Principal Component Scores (APCS) for interpretation of the complex databases offers a better understanding of water quality in the study region.
Form results obtained by Principal Component and Cluster Analysis applied to data set of Foggia province it’s evident that some sampling sites investigated show dissimilarities, mostly due to the location of the site, the land use and management techniques and groundwater overuse. By APCS method it’s been possible to identify three pollutant sources: Agricultural pollution 1 due to fertilizer applications, Agricultural pollution 2 due to microelements for agriculture and groundwater overuse and a third source that can be identified as soil run off and rock tracer mining.
Multivariate statistical methods represent a valid tool to understand complex nature of groundwater quality issues, determine priorities in the use of ground waters as irrigation water and suggest interactions between land use and irrigation water quality.
Ground water serves a number of important functions for humanity and nature. These functions are often related to groundwater composition, which is increasingly influenced by human activities. To assess whether ground water will maintain its present function in future, it’s necessary to obtain insight into the factors determining groundwater composition.
In fact the groundwater quality is affected by many factors including precipitation, surface runoff, groundwater flow, and the characteristics of the catchment area. In particular groundwater composition is determined by initial water composition during infiltration, by groundwater flow patterns and by characteristics of the aquifer. The initial water composition is primarily related to the origin of the recharge water, e.g. precipitation or surface water. During infiltration, changes in water composition may occur through natural processes or through human activities dependent on soil conditions and land use (e.g. evapotranspiration and dissolution of fertilizers). Flow patterns determine the spatial displacement of ground water and dissolved solids through the subsurface. Groundwater flow depends on natural factors (e.g. elevation differences and lithology) and on human interventions (e.g. groundwater extraction and drainage).
During the years 2004-2007 the Agricultural and Food Authority of Apulia Region has implemented the project “Expansion of regional agro-meteorological network” in order to assess, monitor and manage of regional groundwater quality. The wells monitored during this activity amounted to 473, and the water samples analyzed were 1021.
This resulted in a huge and complex data matrix comprised of a large number of physical-chemical parameters, which are often difficult to interpret and draw meaningful conclusions. Further, for effective pollution control and water resource management, it is required to identify the pollution sources and their quantitative contributions [3, 4]. Traditional approaches to assessing water quality are based on the comparison of experimentally determined parameter values with the existing guidelines but in many cases it does not readily give information on status of the source .
The application of different multivariate statistical techniques such as cluster analysis, principal component analysis, source apportionment by multiple linear regression on absolute principal component scores for interpretation of the complex databases offers a better understanding of water quality in the study region.
In fact advantages of multivariate statistical techniques for environmental data can be summarised as:
• reflect more accurately the multivariate nature of natural ecological system
• provide a way to handle large data sets with large numbers of variables by summarizing the redundancy
• provide a means of detecting and quantifying truly multivariate patterns that arise out of the correlation structure of the variable set .
These techniques also permit identification of the possible factors/sources that are responsible for the variations in water quality and influence the water system and in apportionment of the sources, which, thus offer valuable tool for developing appropriate strategies for effective management of the water resources [7–12].
In the present paper, the results obtained from monitoring activity performed in Foggia district (one of the Apulian provinces located in the North part of Apulia region) during the years 2004-2007 in the frame of the project “Expansion of the Regional Agro-meteorological Monitoring Network” are shown. In fact the Agriculture and Food Authority of Apulia Region, in partnership with the Regional Farmer Consortium (Asso.Co.Di. Puglia), CNR-IRSA and Bari University, in 2004 has launched a Water and Soil Monitoring Campaign for the purpose of checking the quality of soils and ground waters, used for irrigation, and then the quality level of the regional agricultural produces. This Project also was aimed to support the farmers to adopt the Best Management Practices (BMPs) and to reduce the water consume and the power and chemical (nutrients and pesticides) inputs in agriculture.
Main crops in Apulia region
% (regional area)
VINEYARDS / ORCHARDS
Main crops in Foggia province
% (provincial area)
CEREALS cereals/tomatoes/vegetables (rotation)
VINEYARDS / ORCHARDS
The Project founded on a tight soil and water sampling collection, carried out all around the region, and on the determination of the main physical and chemical parameters of soils and waters.
The large data base was subjected to different multivariate statistical techniques with a view to extract information about the similarities or dissimilarities among the sampling sites, identification of water quality variables responsible for spatial and temporal variations, the influence of the possible sources (natural and anthropogenic) on the water quality parameters and the source apportioning for estimation of the contribution of possible sources on the concentration of determined water quality parameters of ground waters of Foggia province.
Results and discussion
Descriptive statistics of Foggia province groundwater variables
Electrical conductivity (mS/cm)
Total dissolved solids (TDS) (mg/l)
Dissolved oxygen (ppm)
Chemical oxygen demand (COD) (mg/l)
Vital organism 22°C (UFC/ml)
Vital organism 36°C (UFC/ml)
In the dendrogram of figure 5 the similarity was measured keeping all the original information for each variable, also the noise. By using the scores of the first two components the similarity is linked to the meaning of the first two components.
The samples 143, 231, as the sample 189, highlighted in figure 4 with red circular line, scattered for high values of Mg2+, K+, Ca2+. These are typical cations of nutrients used as fertilizer in grain and tomatoes crops. The sites where these samples were collected were located in farms which main activities were grain and tomatoes crops.
On the contrary the sample number 107 shows a scattering for vital organism at 22 and 36 °C and NO3- (see the black circular line in figure 4). The corresponded site is located along La Contessa channel, as one can see observing figure 7b. The sampling was performed on September 2007 during a period of time in which the water from La contessa channel was used for irrigation. In this channel waters from municipal purifier of Foggia city and waters from paper mill purifier pour. So irrigation water not well purified was used.
In fact organisms growing best at 36 °C, probably, come from external sources: they are bacteria belonging to the mesophilic flora derived from humans and animals. The colonies count at 36 °C increases, therefore, suspects of fecal pollution, reports undesirable changes and should lead to perform additional inspections. It’s an anthropogenic pollution index. The colonies count at 22 °C, although it does not have any health implication, allows us to highlight, in terms of quality and quantity, the putrefactive microbial species, spore-forming and chromogenic, abundant in the surface layers of soil and air, easily adaptable to the water environment. It’s an index of environmental pollution.
Form results obtained by Principal Component and cluster analyses it’s evident that some sampling sites investigated in Foggia district show dissimilarities, mostly due to the location of the site, the land use and management techniques and groundwater overuse. For all these reasons several natural and anthropogenic sources affect the groundwater quality of the investigated sites.
The weigh percentage of the sources are: 32% for agricultural pollution 1, 12% for agricultural pollution 2 and 56% for soil run off and rock tracer mining.
The error on the reconstructed concentration data matrix obtained by equation 2 was 3.2%.
Data treatment and multivariate statistical methods
Table 3 shows the descriptive statistics used in this paper. For each parameter they are average, median, mode, standard deviation, minimum and maximum value.
Multivariate analysis of the groundwater data set was performed by PCA, CA and APCS. PCA and APCS elaborations were obtained by Matlab softwares (MATLAB 7.0) developed from authors. CA was performed by Statistica software (Stat Soft, version 8).
Principal component analysis
PCA includes correlated variables with the purpose of reducing the numbers of variables and explaining the same amount of variance with fewer variables (principal components). The new variables created, the principal components scores (PCS), are orthogonal and uncorrelated to each other, being linear combinations of the original variables. They are obtained in such a way that the first PC explains the largest fraction of the original data variability, the second PC explains a smaller fraction of the data variance than the first one and so forth [14–16]. Varimax rotation is the most widely employed orthogonal rotation in PCA, because it tends to produce simplification of the unrotated loadings to easier interpretation of the results. It simplifies the loadings by rigidly rotating the PC axes such that the variable projections (loadings) on each PC tend to be high or low.
Generally two methods are used in order to chose p Eigenvectors: Kaiser method (PCs with eigenvalues greater than 1) and ODV70 ones (PCs representing at least 80% of the original data variance). In our method we have chosen the second one and we have taken into account p Eigenvectors until the sum of their Eigenvalues reaches at least 70% of the total variance.
CA groups the objects (cases) into classes (clusters) on the basis of similarities within a class and dissimilarities between different classes. The results of CA help in interpreting the data and indicate patterns [12, 17]. In hierarchical clustering, clusters are formed sequentially by starting with the most similar pair of objects and forming higher clusters step by step. Hierarchical agglomerative CA was performed on the data set by means of the Complete linkage’s method using squared Euclidean distances as a measure of similarity . Cluster analysis was applied to the ground water data set with a view to group the similar sampling sites (spatial variability) spread over the Foggia province basin and in the resulted dendrogram, the linkage distance is reported as Dlink/Dmax, which represents the quotient between the linkage distance for a particular case divided by the maximal distance, multiplied by 100 as a way to standardize the linkage distance represented on y-axis [11, 12, 19].
Absolute principal component scores
In the APCS method the first step, that agrees with PCA, is the search of the Eigenvalues and Eigenvectors of the data correlation matrix G. Only the most significant p Eigenvectors (or factors) are taken into account. In our method we have taken into account p Eigenvectors until the sum of their Eigenvalues reaches at least 70% of the total variance.
where Z is the scaled data matrix, PCS is the principal component scores matrix, and VT is the transposed rotated loading (Eigenvectors) matrix.
Using the matrix VT and the Equation (1) the vector PCS0, corresponding to Z0, is calculated and subtracted from all the vectors that form PCS. The matrix obtained in this way is referred to as Absolute Principal Component Scores (APCS) matrix. The APCS matrix can be identified with the estimated contribution matrix (Fr). Also in this case small negative values are usually set zero. Then, a regression on the data matrix X allows to obtain the estimated source profiles matrix (Ar). If the APCS matrix is bordered with a unit column vector, the regression gives for each parameter also a possible contribution of the not explained variance.
At last the product of the matrices Fr and Ar allows to recalculate the data matrix (Xr).
If F and A are unknown, the agreement between X and Xr is the only assessment for the effectiveness of their reconstruction.
where normf is the Frobenius’s norm and Xr is the data matrix reconstructed and X is the data matrix, respectively.
Multivariate statistical methods represent a valid tool to understand complex nature of groundwater quality issues, determine priorities in the use of ground waters as irrigation water and suggest interactions between land use and irrigation water quality. The results obtained by multivariate statistical methods can be used to suggest to stakeholders, for example, a mitigation in the groundwater overuse of some wells mostly in dry seasons and to require orderly quality tests of the channel waters when they are used for crop irrigation.
Olive, vine and cherry for Bari province;
Olives, grapes, and tomatoes for Brindisi;
Olives, grapes, tomatoes and wheat for Foggia;
Olive, tomato and citrus for Lecce;
Olive, vine and citrus for Taranto.
Groundwater quality monitoring
These farms also, according to an agreed protocol, had to present these specific characteristics:
1. to have a continuous crops area ≥ 1.0 ha;
2. to use for irrigation only water coming from the monitored wells;
3. to have the land management register regularly compiled in the last two years;
4. to be close (≤ 5.0 km) to a meteorological gauge-station.
In this paper we show the results obtained from monitoring activity performed in Foggia province. The amount of monitored wells were 85 and the total number of samples collected were 219.
The Province of Foggia (Area: 6,965 km² ; Population: about 680,000 inhabitants), placed in the South of Italy, is part of Apulia Region. It can be sub-divided in three geographic sub-regions: Gargano (the limestone mountains placed in the east part of the province), the Tavoliere (the alluvial plain placed in the central zone) and Sub-Appennino (the mountainous system in the west part). A complex river network characterizes Sub-Appennino and Tavoliere, where several streams flow from west to east. The main rivers are Fortore (in the north, along the boundary with the Molise region), Candelaro (which separates Tavoliere from Gargano) and Cervaro (in the South).
The province of Foggia is one of the most important agricultural areas of Italy, especially its alluvial plain. The main crops are winter wheat, tomatoes, vegetables, orchard, vineyard and olive groves.
Samples from the boreholes were collected using manually operated hand pumps.
Sampling took place under dynamic conditions, after flushing a large amounts of water for about 30 minutes.
All samples were kept in two liters polyethylene bottles, which have been previously washed with 1:1 HCl and distillated water. The bottles, which have cap and under cap, were filled to the brim in order to prevent the transfer of the analytes in the headspace and their loss at the opening of the bottles.
After collection, samples were stored in cooled bags and transported to the laboratory as soon as possible.
They were stored in the refrigerator at about 4°C before the analysis without chemical preservatives because the analysis was performed either directly on-site, or immediately in the laboratory.
The samples were analyzed for pH, Electrical Conductivity, TDS, O2, COD, the major ions (ie. Na+, Ca2+, Mg2+, K+, Cl-, NO3-, SO42- and HCO3-), vital organism at 22 and 36 °C.
The chemical and physical analyses of water samples have been carried out according to the official guideline proposed by the Ministero delle Politiche Agricole (the national agriculture authority) in a specific law (Decreto Ministeriale del 23 Marzo 2000 “Metodi ufficiali di analisi delle acque per uso agricolo e zootecnico” ).
Some physical-chemical parameters such as pH, Electrical Conductivity, TDS and dissolved oxygen were determined immediately after sampling. All field meters were checked and calibrated according to the manufacturer’s specifications.
In particular, the pH meter (Hanna instruments, model 9025) was calibrated using two buffers of pH 7.0 and 10.0.
Conductivity /TDS meter (Hanna Instruments, model 9835) was used to measure the conductivity and total dissolved solids of the water samples. The instrument was calibrated with 0.001M KCl to give a value of 14.7 μS/m at 25°C. The probe was thoroughly rinsed with distilled water after each measurement.
Dissolved oxygen meter (Hanna Instruments, model 9143) was automatically standardized to the actual saturation value (after setting the appropriate working altitude) prior to each measurement set.
Chemical parameters were determined after filtration of the sample under vacuum on cellulose acetate filters with porosity of 0,45 microns.
About COD measurements, 2.0 mL of sample was added to a vial (sample vial) filled by the manufacturers with the reagent solution (HI 93754A-25 for a low range of COD: 0 – 150 mg/L). 2.0 mL of deionized water was added to other vial (blank vial). The vials was heated for 2 hours at 150°C. During this digestion period oxidizable organic compounds reduce the dichromate ion (orange) to the chromic ion (green). The amount of remaining dichromate was automatically determined with a multiparameter bench photometer (Hanna Instruments C99).
The samples were analyzed for Na+, Ca2+, Mg2+, K+ using a Varian atomic absorption spectrophotometer (model SpectrAA – 250 PLUS).
The anions Cl-, NO3-, SO42- were measured by ion chromatography (Dionex corporation, model AS50) and bicarbonate- by a simple alkalimetric method.
The performance of spectrophotometer (cations determination) and ion chromatography (anions determination) was checked by passing standard solutions of all measured parameters. Blank samples (deionized water) were analyzed after every six measurements of water samples to check for any eventual contamination or abnormal response of equipment.
The colonies count at 36 °C and 22 °C (vital organism at 36 °C and 22 °C) is considered an indicator of poor protection of a hydric environment. The use of different temperatures highlights mesophilic microorganisms (36 °C) and psychrophilic (22 °C).
In the analytical method  used in this paper for the vital organism at 22 °C and 36 °C determination, Plate Count Agar (PCA), a microbiological growth medium agar, no selective, enriched with tryptone, yeast extract and glucose, which allows growth of almost all undifferentiated microbial species in the water sample, was used.
This study was supported by the “Expansion of regional agro-meteorological network” project funded by Apulia Region (Italy).
This article has been published as part of Chemistry Central Journal Volume 6 Supplement 2, 2012: Proceedings of CMA4CH 2010: Application of Multivariate Analysis and Chemometry to Cultural Heritage and Environment. The full contents of the supplement are available online at http://journal.chemistrycentral.com/supplements/6/S2.
- Schot PP, van der Wal J: Human impact on regional groundwater composition through intervention in natural flow patterns and changes in land use. J Hydrol. 1992, 134 (1-4): 297-313. 10.1016/0022-1694(92)90040-3.View ArticleGoogle Scholar
- Kannel PR, Lee S, Lee YS: Assessment of spatial–temporal patterns of surface and ground water qualities and factors influencing management strategy of groundwater system in an urban river corridor of Nepal. J Environ Manage. 2008, 86 (4): 595-604. 10.1016/j.jenvman.2006.12.021.View ArticleGoogle Scholar
- Singh KP, Malik A, Sinha S: Water quality assessment and apportionment of pollution sources of Gomti river (India) using multivariate statistical techniques - a case study. Anal Chim Acta. 2005, 538 (1-2): 355-374. 10.1016/j.aca.2005.02.006.View ArticleGoogle Scholar
- Dixon W, Chiswell B: Review of aquatic monitoring program design. Water Res. 1996, 30 (9): 1935-1948. 10.1016/0043-1354(96)00087-5.View ArticleGoogle Scholar
- Debels P, Figueroa R, Urrutia R, Barra R, Niell X: Evaluation of water quality in the Chillán River (Central Chile) using physicochemical parameters and a modified water quality index. Environ Monit Assess. 2005, 110 (1-3): 301-322. 10.1007/s10661-005-8064-1.View ArticleGoogle Scholar
- McGarigal K, Cushman S, Stafford S: Multivariate Statistics for Wildlife and Ecology Research. 2000, New York: Springer, 978-0387986425View ArticleGoogle Scholar
- Simeonova P, Simeonov V, Andreev G: Water quality study of the Struma River Basin, Bulgaria. Cent Eur J Chem. 2003, 1 (2): 121-136. 10.2478/BF02479264.Google Scholar
- Simeonov V, Simeonova P, Tsitouridou R: Chemometric quality assessment of surface waters: two case studies. Ecol Chem Eng. 2004, 11 (6): 450-469.Google Scholar
- Bengraine K, Marhaba TF: Using principal component analysis to monitor spatial and temporal changes in water quality. J Hazard Mater. 2003, 100 (1-3): 179-195. 10.1016/S0304-3894(03)00104-3.View ArticleGoogle Scholar
- Liu CW, Lin KH, Kuo YM: Application of factor analysis in the assessment of groundwater quality in a blackfoot disease area in Taiwan. Sci Total Environ. 2003, 313 (1-3): 77-89. 10.1016/S0048-9697(02)00683-6.View ArticleGoogle Scholar
- Simeonov V, Stratis JA, Samara C, Zachariadis G, Voutsa D, Anthemidis A, Sofoniou M, Kouimtzis Th: Assessment of the surface water quality in Northern Greece. Water Res. 2003, 37 (17): 4119-4124. 10.1016/S0043-1354(03)00398-1.View ArticleGoogle Scholar
- Singh KP, Malik A, Mohan D, Sinha S: Multivariate statistical techniques for the evaluation of spatial and temporal variations in water quality of Gomti River (India)-a case study. Water Res. 2004, 38 (18): 3980-3992. 10.1016/j.watres.2004.06.011.View ArticleGoogle Scholar
- European Environment Agency (EEA): CORINE Land Cover Project. 2005Google Scholar
- Abdul-Wahab SA, Bakheit CS, Al-Alawi SM: Principal component and multiple regression analysis in modelling of ground-level ozone and factors affecting its concentrations. Environ Modell Softw. 2005, 20 (10): 1263-1271. 10.1016/j.envsoft.2004.09.001.View ArticleGoogle Scholar
- Sousa SIV, Martins FG, Alvim-Ferraz MCM, Pereira MC: Multiple linear regression and artificial neural networks based on principal components to predict ozone concentrations. Environ Modell Softw. 2007, 22 (1): 97-103. 10.1016/j.envsoft.2005.12.002.View ArticleGoogle Scholar
- Wang S, Xiao F: AHU sensor fault diagnosis using principal component analysis method. Energ Buildings. 2004, 36 (2): 147-160. 10.1016/j.enbuild.2003.10.002.View ArticleGoogle Scholar
- Vega M, Pardo R, Barrado E, Deban L: Assessment of seasonal and polluting effects on the quality of river water by exploratory data analysis. Water Res. 1998, 32 (12): 3581-3592. 10.1016/S0043-1354(98)00138-9.View ArticleGoogle Scholar
- Todeschini R: Introduzione alla chemiometria. 1998, Napoli: EdiSES s.r.l., 9788879591461Google Scholar
- Wunderlin DA, Díaz MDP, Amé MV, Pesce SF, Hued AC, Bistoni MdlA: Pattern recognition techniques for the evaluation of spatial and temporal variations in water quality. A case study: Suquía River Basin (Córdoba–Argentina). Water Res. 2001, 35 (12): 2881-2894. 10.1016/S0043-1354(00)00592-3.View ArticleGoogle Scholar
- Thurston GD, Spengler JD: A quantitative assessment of source contributions to inhalable particulate matter pollution in metropolitan Boston. Atmos Environ. 1985, 19 (1): 9-25. 10.1016/0004-6981(85)90132-5.View ArticleGoogle Scholar
- Caselli M, de Gennaro G, Ielpo P: A comparison between two receptor models to determine the source apportionment of atmospheric pollutants. Environmetrics. 2006, 17 (5): 507-516. 10.1002/env.788.View ArticleGoogle Scholar
- Ielpo P: Heavy metals and atmospheric particulate: chromatographic analysis, scanning electron microscopy and source apportionment. PhD thesis. 2004, University of Bari, Italy, (Stored to the Public Libraries of Rome and Florence - BNI0013860- Italy)Google Scholar
- Decreto Ministeriale del 23 Marzo 2000. Approvazione dei “Metodi ufficiali di analisi delle acque per uso agricolo e zootecnico”. 2000Google Scholar
- APAT CNR IRSA 7050. Manual 29. 2003Google Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.