Representing soil pollution by heavy metals using continuous limitation scores
Introduction
The problem of soil pollution by heavy metals has been receiving an increasing attention in the last few decades. In Europe, decision makers and spatial planners more and more require information on soil quality for different purposes: to locate areas suitable for organic (ecologically clean) farming and agro-tourism; to select sites suitable for conversion of agricultural to non-agricultural land, particularly for urbanization; setting up protection zones for groundwater pumped for drinking water; to estimate costs of remediation of contaminated areas and similar. Heavy metals occur naturally in rocks and soils, but increasingly higher quantities of them are being released into the environment by anthropogenic activities. Every decision on the application of any measures in the environment relating to soil quality and management, whether statutory regulations or practical actions, must be based on reliable and comparable data on the status of this part of environment in the given area. Various aspects must be considered by the society to provide a sustainable environment, including a soil clean of heavy metal pollution. The first among them is to identify environments (or areas) in which anthropogenic loading of heavy metals puts ecosystems and their inhabitants at health risk. Maps indicating areas with pollution risks can provide decision makers or local authorities with critical information for delineating areas suitable for the planned land use or soil clean up (Van der Gaast et al., 1998, Broos et al., 1999). Maximum permissible concentrations of heavy metals in soil are now regulated by law in many countries.
Before any solution for the problem of soil heavy metal pollution can be suggested, a distinction needs to be made between natural anomalies and those resulting from human activities. Namely, it often happens that also natural concentrations and distribution of potentially toxic metals could present health problems, like in the case of chromium, cobalt, and particularly nickel in ultramafic soils (Proctor and Baker, 1994). Rock type and geological–geochemical processes can change markedly in a relatively small area, resulting in great spatial variability in the soil content of elements. Soils in the vicinity of urban areas and industry are exposed to input of potentially toxic elements, and the situation of agricultural soils gets additionally complicated due to continuous application of agrochemicals.
In practice, soil pollution by heavy metals is commonly assessed by interpolating concentrations of heavy metals sampled at point locations, so that each heavy metal is represented in a separate map (Webster and Oliver, 2001, Juang et al., 2003). The first problem of working with maps of separate heavy metal concentrations (in further text HMCs) is that the limiting values for polluted soils are commonly set as crisp boundaries. For example, a soil is polluted by zinc and not suitable for organic agriculture if the measured values are larger than (Official Gazette, 2001). This means that a soil with zinc concentration of and a soil with a concentration of will be classified differently although the difference may be due to the measurement or interpolation error. Similarly, if the concentration of zinc at a location is and at neighboring location , both locations will be classified as not suitable although the latter shows two times higher concentration. The second problem with HMCs is that different elements come in different ranges of values. This makes it fairly difficult to get the picture about the overall soil quality. For example the threshold value for zinc is and for cadmium . If we measure, at a point, values Zn=130 (suitable) and Cd=1.1 (not suitable), this makes this location unsuitable but how serious is the problem? Now imagine a case with tens of HMCs—how can we sum these values to get the compound picture about the quality of soil?
To solve a problem of presenting overall polluted areas, Romić and Romić (2003) applied factor analysis prior to interpolation and then interpolated only the first factor indicating anthropogenic loads of heavy metals. Van der Gaast et al. (1998) used maps of background values of soil contaminants focusing on the 90-percentiles. Hanesch et al. (2001) tested fuzzy classification algorithms to distinguish different sources of pollution. Amini et al. (2004) classified HMCs using unsupervised fuzzy -means to partition the values optimally. The final outputs are maps of memberships to each cluster, which commonly reflect the combination of most correlated heavy metals. In all these examples the procedures are statistically valid, but the meaning of such factors and continuous memberships is hard to interpret. In practice, decision makers usually only wish to see the areas that are polluted without any training in (geo)statistics.
In this paper, we propose an approach to interpolate sampled HMCs using numerous environmental predictors and then represent the overall pollution by using the continuous limitation scores (LS). We advocate the use of cumulative LS because they can be summed and used to represent areas of overall high pollution. Such visualizations can supplement maps of separate HMCs so that the end-users can more easily delineate areas of high overall pollution and focus their actions where their are more needed.
Section snippets
Spatial interpolation
For spatial interpolation of HMCs we used the regression-kriging (Odeh et al., 1995), also known as Universal kriging (Webster and Oliver, 2001) or Kriging with External Drift (Goovaerts, 1997) (see also the article on regression-kriging published in the same issue of this journal). This technique is especially attractive as it can employ both our empirical knowledge about the distribution of HMCs and the spatial autocorrelation between the point samples. It will also minimize the artificial
Regression analysis
The first screening of data showed that almost all HMCs have asymmetrical distributions, clearly shifted toward the lower values. After the logit transformation, the distributions were closer to approximately normal (Fig. 3), which allowed us to do further statistical analysis. This confirms that logit transformation is an important step prior to actual interpolation. The step-wise regression analysis in R selected geological map (GEO), ground water depth (GWD), NDVI, slope map (SLOPE) and
Discussion and conclusion
The developed procedure for geostatistical analysis of HMC data enabled us to identify a number of contamination hotspots and to map the cumulative contamination by heavy metals. Regression-kriging has shown to be a powerful interpolation technique because it utilizes all possible linear correlations (with auxiliary predictors and auto-correlation). An alternative to regression-kriging would be to run multivariable interpolation (all at once) on sets of HMCs, which is also possible in the GSTAT
References (23)
- et al.
Estimating the direction of an unknown air pollution source using a digital elevation model and a sample of deposition
Ecological Modelling
(1999) - et al.
Quantification of the effects of spatially varying environmental contaminants into a cost model for soil remediation
Journal of Environmental Management
(1999) - et al.
The application of fuzzy c-means cluster analysis and non-linear mapping to a soil data set for the detection of polluted sites
Physics and Chemistry of the Earth, Part A: Solid Earth and Geodesy
(2001) Finding the right pixel size
Computers and Geosciences
(2006)- et al.
A generic framework for spatial prediction of soil variables based on regression-kriging
Geoderma
(2004) - et al.
Further results on prediction of soil properties from terrain attributes: heterotopic cokriging and regression-kriging
Geoderma
(1995) Multivariable geostatistics in s: the gstat package
Computers and Geosciences
(2004)- et al.
Magnetic susceptibility as proxy for heavy metal pollution: a site study
Journal of Geochemical Exploration
(2005) - et al.
The grey areas in soil pollution risk mapping the distinction between cases of soil pollution and increased background levels
Journal of Hazardous Materials
(1998) - et al.
Continuous soil pollution mapping using fuzzy logic and spatial interpolation
Geoderma
(2004)
Geostatistics for Natural Resources Evaluation
Cited by (25)
Short-range and regional spatial variability of soil chemical properties in an agro-ecosystem in eastern Croatia
2017, CatenaCitation Excerpt :Hengl et al. (2007) used regression kriging to predict organic matter in the topsoil, while Pilaš et al. (2013) used existing soil database to predict soil carbon stocks at the national level. National-scale spatial variability of soil pH and OM was analysed by Hengl et al. (2004), while Romić et al. (2007) and Sollitto et al. (2010) investigated county-scale spatial variability of heavy metal concentrations. Field-scale spatial modelling of soil AP and AK to provide the most accurate fertility maps was analysed by Bogunovic et al. (2014a).
Modeling and mapping of cadmium in soils based on qualitative and quantitative auxiliary variables in a cadmium contaminated area
2017, Science of the Total EnvironmentCitation Excerpt :Terrain factors are closely related to surface substance migration. Distance to residential area reflected the influence of human activities on the spatial distribution of heavy metals in soil (Romic et al., 2007). Pearson correlation analyses were used to evaluate the influence of terrain attributes and distance to residential area on Cd content (Table 3).
Spatial distribution and risk assessment of metals in agricultural soils
2016, GeodermaCitation Excerpt :Elevated concentrations are mainly caused by the application of Cu-based fungicides and bactericides. Contamination with Cu due to the use of Cu-based pesticides is widely documented (Cicchella et al., 2015; Romić et al., 2007; Saby et al., 2011; Sollitto et al., 2010). The other main source is the lithological substrate of Tertiary volcanic rocks (basalts, phonolites, tuffs), with a median value of 11.3 mg/kg in north-western Bohemia.
Eolian contribution to geochemical and mineralogical characteristics of some soil types in Medvednica Mountain, Croatia
2014, CatenaCitation Excerpt :These studies are underlining the fact that Zagreb is facing serious pollution problems triggered mostly by growing industry and urbanization processes. Various multivariate statistical and geostatistical methods have been utilized to assess the overall pollution of soils and overbank/floodplain sediments of the Sava River, flowing from Medvednica Mountain, with particular reference to heavy metal contamination of agricultural and urban soils in the Zagreb city and its outskirts (e.g. Pavlović et al., 2004; Romić and Romić, 2003; Romić et al., 2007; Sollitto et al., 2010). Also, a considerable amount of data has been collected during the sampling campaign carried out within the last decade for the Geochemical Atlas of Croatia (Halamić and Miko, 2009).
Value of information and mobility constraints for sampling with mobile sensors
2012, Computers and GeosciencesCitation Excerpt :This is especially useful in applications such as radioactivity, contaminants and fire risk, in which false negative costs are usually higher than false positives (Heuvelink et al., 2010). The method as presented in this paper applies to phenomena that change much slower than the speed of sampling, which is a common situation in phenomena such as soil contamination (Rodriguez-Lado et al., 2008; Romic et al., 2007), natural radioactivity (Heuvelink and Griffith, 2010), and biodiversity (Zerger et al., 2010). Extending the method to highly dynamic phenomena requires considering the temporal behaviour of the phenomenon studied within the sampling procedure (Kho et al., 2009), which is topic of further research.