Identification of geochemically distinct regions at river basin scale using topography, geology and land use in cluster analysis Ramirez-Munoz P. and Korre, A. Mining and Environmental Engineering Research Group, Imperial College London, South Kensington Campus, SW7 2AZ, London, United Kingdom paula.ramirez-munoz@imperial.ac.uk; a.korre@imperial.ac.uk Introduction It is known that surface water quality is influenced by current and historical industrial activities, as well as land use practices, which are superimposed on the natural processes in river systems. Baseline concentrations of metals need to be determined in order to assess the real contribution of the various anthropogenic sources and establish whether the combined effect of natural and human contributions to river water metal concentrations pose a detrimental effect on the environment. However, this is not an easy task since geology, topography or land use, which determine the baseline levels, are not homogenous in a catchment. In addition, anthropogenic sources can often mask the natural geochemical signature in surface water systems. The objective of this paper is to illustrate the extent of variability found in stream water systems at river basin scale and introduce a new methodology to define geochemically distinct areas based on topography, geology and land use, which can later be used in the environmental quality standards (EWQS) derivation. The Rapel River Basin in Central Chile is an ideal setting for this work as it spreads over a large area with variable topography and geology which also hosts many different human activities including mining, agriculture and farming. These activities have undergone an accelerated growth during recent years, implying increased pressure on the surface water system. Methodology River water quality in the Rapel River Basin was monitored in two sampling campaigns carried out in April-May 2006 and December-January 2007 using clean sampling protocols. Filtered and unfiltered water samples were collected during the low flow and high flow season (93 and 100 locations respectively). Fifteen cations (Al, As, Ba, Cd, Co, Cr, Cu, Li, Mn, Ni, Pb, Sr, Ti, V, Zn) were analysed using ICP-AES and ICP-MS; eight (Ca, Fe, K, Mg, Na, P, S, Si) by ICP-AES; and five (Ag, Au, Mo, Sb, Sn) by ICP-MS. Four anions (NO 3 -, SO 4 2-, Cl -, F - ) were determined using Dionex chromatography and the DOC was analysed using a Shimadzu TOC-VCSH system. Land use and geology were related to each sampling point through the delineation of the surface water sourceshed 1
using a 50 50 1 m vertical digital elevation model (DEM). A sourceshed was defined as the total area that contributes to a selected sampling point and the land use and geology categories influencing the point were calculated as a percentage of the sourceshed area. In order to avoid erroneous/biased results in multivariate statistical analysis [1], the percentage land use and geology category data were transformed using the isometric logratio transformation to open the data set, destroy the effect of closure and avoid singularity. A hierarchical cluster analysis using six land use and ten geology classes was carried out to determine areas with similar features for the nested watershed. The Ward agglomerative algorithm with Euclidean distance was used to estimate dissimilarity. Results Water Classification The order of abundance of the major ions in most collected stream water samples was Ca 2+ > Mg 2+ > Na + > K + and SO 4 2- > HCO 3 - > Cl - > Si. The principal water type of the Rapel River basin was Ca 2+ -SO 4 2-, which is the typical water type for the upper reaches of the catchment. This water type could be the product of chemical weathering of silicate and carbonate rocks. Cachapoal and Tinguiririca rivers drain sedimentary sequences with the presence of gypsum and carbonate rocks (Rio Damas Formation) in the upper part of these basins. Dissolution of the volcanic rocks of the Farellones Formation could produce high sulphate concentration in water, since pyrite is generally present in these rocks affected by propilitic alteration. Downstream, where the rivers pass through the Central Depression formations, the water changes composition to Ca 2+ -HCO 3 - (Antivero river). Anthropogenic activities of the Central Depression are expected to contribute to the water composition adding Na +, Cl - and HCO 3 - to the Cachapoal and Tinguiririca rivers. The comparison between the type of water samples during the low and high flow periods showed that the unfiltered trace concentrations (total concentration) were higher than the filtered trace concentrations (dissolved concentration) in both low and high flow seasons (p-value at 95% confidence for the Cu paired Wilconox test was 7.3 10-4 ). This can be explained by the presence of suspended particles in the unfiltered samples, such as hydroxides and clays, which tend to bind trace metals. The difference was more pronounced during the high flow period, due to stronger erosion and river water transport capacity. In contrast, major cations (Ca 2+, Na +, K +, Mg 2+ ) in the unfiltered and filtered concentrations were similar during both the low and high flow seasons. The comparison between the two seasons for the filtered samples showed that for most parameters higher concentrations were observed during the low flow period. Generally, trace metal concentrations tend to decrease with increasing flow magnitudes, due to dilution of solutes in the base flow, derived primarily from groundwater or by diffuse inputs of meteoric water during runoff events. However, it is known that dissolved ion concentrations may increase, decrease or follow no systematic trend with rising flow conditions [2]. Some parameters, such as Al 3+ and Pb 2+, exhibited a higher concentration during high flow season in the catchment. Discharge has been recognised as the most 2
significant factor controlling trace metal fluctuations [3].Therefore, the low flow campaign is considered to be a better indicator of the geochemical processes that occur in the river water. Figure 1: Classification of the watersheds according to cluster analysis, Ward s method, using land use and geology categories. Cluster Analysis The dendrogram of Figure 1Error! Reference source not found. shows that the 107 nested watersheds may be grouped in five main clusters with the three main rivers (Cachapoal, Tinguiririca and Alhue river) presenting a distinct pattern of land use and geology classes. Cluster 1 grouped principally all sourcesheds in the Main Cordillera where the geology consists of volcanic and sedimentary formations and the main land use classes are land lacking vegetation or covered by ice. Therefore, the river water chemistry may reflect the composition of the rocks in these parts of the Rapel River Basin. Cluster 2 grouped the nested watersheds located in Coya river, which were characterised by industrial land use (mining) and the Cu-Mo porphyric ore deposit (El Teniente mine). The chemistry of the water that drains these sourcesheds may present high natural concentration of constituents derived from the mining and industrial land use. Cluster 3 grouped the nested sourcesheds located downstream the Coya and Cachapoal river confluence until the mouth of the Cachapoal river. This cluster represents the combined effect of the two previous clusters. Cluster 4 grouped the watershed in the Coastal Range of the Alhue Basin. This catchment is characterised by the presence of Oligocene intrusive rocks and agricultural land use. This distinct cluster suggests that probably the character of the river water may be different from the Cachapoal and Tinguiririca rivers. Cluster 5 grouped all sourcesheds located in the Central Depression, which is directly related to quaternary deposits and agricultural land use. The character of these waters may be largely 3
influenced by diffuse pollution from agriculture. The relationship of these clusters with the chemistry of the water samples in each cluster was further investigated by comparing the medians of the chemical parameters in the filtered samples of the low flow season campaign to avoid distortion due to outliers (Table 1). The three main rivers exhibited clear differentiation in agreement with the cluster analysis results. The water chemistry, which is controlled by a combination of the natural and anthropogenic point and nonpoint sources, was shown to be in good agreement with the cluster analysis classification. Table 1: Parameters with the highest median concentration in each cluster. Cluster 1 Cluster 3 Cluster 4 Cluster 5 Cachapoal river Tinguiririca river Cachapoal river Alhue river Antivero river Ca SO 4 Sr K Na Cl Li As Al Mn Ti Ca SO 4 Sr Mg K NO 3 Cu Mo Zn Alkalinity Si Mg Al Fe Ni Ti V Alkalinity NO 3 DOC Si V Conclusions The cluster analysis using sourceshed derived land use and geology data can be used to identify areas with distinct features relating to combined natural and anthropogenic geochemical signatures. This methodology was used for a large river basin hosting complex geology and land use and the resulting classification was found to be in good agreement with stream water chemistry data. This approach is an effective tool in defining individual areas at river basin scale that can be used in defining site-specific EWQS for metals. Acknowledgements: This work was possible thanks to the logistic support and encouragement of the Codelco El Teniente Mine personnel. References [1] Aitchison, J., (2003) The statistical analysis of compositional data. Blackburn Press, Caldwell, N.J; Le Maitre, R.W., (1982) Numerical petrology: statistical interpretation of geochemical data. Elsevier Scientific, Amsterdam ; Oxford. [2] Horowitz, A.J., Lum, K.R., Garbarino, J.R., Hall, G.E.M., Lemieux, C., Demas, C.R., (1996) Problems associated with using filtration to define dissolved trace element concentrations in natural water samples. Environmental Science & Technology, 30(3), 954-963. 4
[3] Miller, J.R., Orbock Miller, S.M., (2007) Contaminated rivers : a geomorphologicalgeochemical approach to site assessment and remediation. Springer, Dordrecht ; [London]. 5