UNICEF arsenic prediction models

From GapWiki
Jump to: navigation, search

Scientific Basis of the UNICEF Arsenic Prediction Model

Geochemical Mobilisation Mechanisms

The mobilisation of arsenic into groundwater can be generalised in terms of four basic geochemical mechanisms, each associated with particular chemical characteristics, and each occurring in distinctive geological and climatic associations:

Reductive dissolution. Arsenic adsorbed to iron or manganese oxides is released into solution when the oxides, which usually occur as coatings on aquifer sands, dissolve due to microbially mediated reduction. These near-neutral-reducing (NNR) waters are characterised by pH 6.5-7.5, and indicators of strongly reducing conditions such as high concentrations of iron, manganese, bicarbonate, ammonium and methane gas, and an absence of oxidised species such as nitrate and sulphate. Examples include the Bengal Basin (e.g. McArthur et al., 2004), the Mekong and Red Rivers in Asia (Berg et al., 2006), and the Danube and Po in Europe, the mid-west of the USA (Kelly et al., 2005). The processes require the presence of abundant organic matter.

Alkali desorption. Laboratory studies show that arsenic adsorbed to iron, manganese and aluminium oxides and clay minerals may be desorbed at pH >8.0, leaving the carrier phase as a solid. Although some authors (e.g. McArthur et al., 2004) have questioned how widely this occurs in nature, many cases of arsenic pollution have been attributed to alkali-desorption (AD). The best-documented example comes from the southwest USA (Bexfield and Plummer, 2003), and others come from Oklahoma, Spain, China and from volcanic deposits in Argentina (e.g. Nicolli et al., 1989).

Sulphide oxidation. Oxidation of arsenic-rich pyrite and other sulphide minerals is a well-known cause of pollution around mining sites, but is relatively rare in natural systems. However, these processes may occur wherever the water table fluctuates across a sulphide-rich layer, such as in Palaeozoic sandstones in Wisconsin (Schreiber et al., 2000) and Holocene alluvium in Perth, Australia (Appleyard et al., 2006).

Geothermal arsenic. Some of the highest known natural concentrations of arsenic occur in hot springs on the Qinghai-Tibet plateau that originate through high-temperature leaching of rocks, due to either deep and rapid circulation of groundwater or shallow volcanism (Webster and Nordstrom, 2003). In most cases, the contamination risk is suggested by the high temperature, however, severe pollution of rivers and groundwater in Chile results from seepage of geothermal in the Andes, hundreds of kilometres from the point of abstraction (Smith et al., 1998; Yuan et al., 2007).

Geological and Climatic Setting

If the distribution of arsenic pollution is considered in terms of the exposed population rather than pure geological or geochemical diversity, some simple and important patterns become clear. More than 90% of all people exposed to drinking water containing >10 µg/L As obtain this water from alluvial aquifers. Most of these basins are what are termed foreland basins, draining and lying adjacent to young mountain belts. In approximately 80% of these cases, arsenic was mobilised to groundwater by reductive dissolution, and in 15% of cases by alkali-desorption. RD occurrences are concentrated in humid environments where organic matter readily accumulates, whereas AD occurs preferentially in hotter and drier environments. It has also been shown that the occurrence of As-contaminated alluvial aquifers is correlated with characteristics of river basins such as their sediment load and mineralogy, and the chemistry of river water (Ravenscroft et al., 2009). These latter characteristics are important, and are related to the geochemistry of groundwater, because they determine the supply of slightly weathered sands containing a source of mobilisable arsenic adsorbed to iron oxide coatings. The climatic influence determines whether this can be mobilised by reduction or desorption.

Population: Exposed and At-Risk

Apart from location, the models predict the population of the areas within which there is a risk of arsenic pollution of groundwater. In the models, population densities from low-resolution data sets are applied to known areas with a particular geological and climatic association. There is inherent imprecision, but not fundamental error, in this method. Even if arsenic is present in groundwater, the population actually exposed to drinking (or eating food cooked in) such water will be significantly less, and may sometimes be close to zero. In this context, it is useful to distinguish between an actual and a potential health hazard. There is a hierarchy of reasons why the exposed population will be less than the at-risk population:

  1. Special geological conditions may have prevented arsenic mobilisation, or attenuated it, in a location that otherwise appeared favourable to mobilisation.
  2. Groundwater may not be used for water supply. In alluvial tracts, this tends to occur either in areas of low population density where there is little pressure on, or pollution of, surface waters; or in urban and peri-urban areas where a municipal supply is piped into the area. The latter situation is particularly prone to over-inflating the at-risk estimates.
  3. Groundwater is rarely pervasively polluted. Although there are parts of Bangladesh, for example, where nearly all wells are polluted, in most affected regions only a few tens of percent of wells that are polluted. The proportions often vary with well depth and may well change over time with growing knowledge.
  4. Groundwater may be directly treated, or indirectly treated by way of storage prior to use, so that concentrations measured at the well-head are not actually experienced by consumers. This is particularly common in the case of aeration to remove iron and/or odour problems, which removes at least part of the arsenic as an unintended benefit.

The at-risk populations predicted by the two GIS models are not perfectly compatible because the main, alluvial, model is based on objectively defined geographic boundaries whereas the sulphide-oxidation model uses arbitrary boundaries. Also, the aquifers in which the SO model predicts pollution are likely to be low-permeability fractured rock, and hence greater use of other water sources inherently more likely.

Improving the population estimates requires first a more detailed map of population (which would, for instance, increase the at-risk population in Bangladesh to around 100 million), and second national, or better, regional statistics on water use and access to different water supply. However, there is a major caveat here, in areas where arsenic has not been detected, statistics reported as ‘access to improved water supplies’ will include wells that are polluted by arsenic.

Prediction Principles

The objectives of the model are to identify where sources of slightly weathered sand, containing As-bearing igneous and metamorphic rock fragments accumulate to form aquifers, and classify these depositional environments according to whether reducing or oxidising geochemical processes operate there. The favourability of these processes can be crudely predicted from annual rainfall and temperature statistics.

Conditions that favour delivery of relatively unweathered sand to the lower catchment are a cool and high altitude upper catchment where there is little chemical weathering, and a steep river profile (as expressed in high sediment load) so that there is limited modification during transport. Young mountain belts formed at destructive plate margins, such as the Alpine-Himalayan and Rocky-Andean chains, provide a good source of igneous and metamorphic rocks and hydrothermally enriched sediments with moderately, but not remarkably, arsenic-enriched content. The ArcAtlas GIS database (ESRI, 1996) contains global mapping of mountain belts (geosynclines) and Quaternary sediments. These themes can be linked through mapping of the river channels to classify which alluvial deposits are derived from these mountains. ESRI’s Quaternary sediment mapping also identifies volcanic sediment, normally originating from the same mountain sources, which is a significant risk factor for arsenic pollution (e.g. Argentina). In addition, it is recognised that aquifers within or immediately adjacent to young mountains or volcanoes may also be contaminated by arsenic from geothermal sources.

The ArcAtlas database also contain climatic and demographic information. The former data can be used to predict the most likely arsenic mobilisation mechanism, and the latter to classify the map units according population density and country. The correlation between processes and present climate will be imperfect because climates may have different in the geologically recent past, when the aquifer sediments were deposited. This may account, for example, in abundant organic matter in areas that have low rainfall. Consequently, the map units are qualified as ‘high’ or ‘low’ risk according the likelihood of AD or RD being the dominant mechanism.

A second line of predictive evidence derives from studies of the mineralogy of suspended sediment and the chemistry of river water in arsenic-affected river basins (Figure 2), and discussed in detail elsewhere (Ravenscroft et al., 2009). The intensity of chemical weathering, and hence removal of arsenic, is reflected in increasing proportions of quartz in sand, and silica in river water. Basins where arsenic pollutes alluvial groundwater have a characteristic grouping, and it is reasonable to assume that a similar chemistry is a risk factor for arsenic in basins where its status is unknown.

Formulation of Predictive Models in Geographical Information Systems

Arsenic Risk in Alluvial Aquifers

Following the principles described above, a GIS model was formulated in ArcView using the ESRI (1996) data set, as described below and a flowchart of the data processing methodology is shown in Appendix II.

  1. Classify Quaternary as ‘alluvial’ by ‘Type’ = alluvial, lacustrine or fluvio-glacial [( [Type] = 3) or ([Type] = 4) or ([Type] = 6) or ([Type] = 11) or ([Type] = 30) or ([Type] = 34) or ([Type] = 35) or ([Type] = 36 ) or ([Type] = 42 )]. These are alluvial, lacustrine, fluvioglacial, volcanic, alluvial and deluvial, lacustrine-alluvial, lacustrine-glacial, alluvial-marine and volcanic-sedimentary. Based on knowledge from Argentina, loess ([Type] = 11) was added for South America only because this is known to be rich in volcanic ash and to be a major cause of As-pollution. ‘Proluvial’ alluvium ([Type=37]) was also introduced later for all areas.
  2. Classify average annual rainfall as <500, 500-1,000, 1,000-1,500, and >1,500 mm (rain_class = 0, 1, 2, 3).
  3. Intersect the rainfall theme on the ‘alluvial’ theme, and classify alluvium by rainfall class
  4. Classify average annual temperature as <10, 10-15 and >15ºC (temp_class = 0, 1, 2), assigning the results to the alluvial polygons.
  5. Combine <rain_class> with <temp_class> to generate a <Climate_class> field as per Table 1. A sub-humid class was defined to include areas with 500-1,000 mm and annual temperature of < 10°C, which empirical evidence suggested had similar favourability for arsenic mobilisation to 500-1,000 mm at temperatures of >10°C.
  6. To identify rivers likely to carry sediment likely to have potential to release arsenic, classify rivers as either crossing or being within 50 km of a Tertiary ‘geosyncline’ (mountain belt), or are within 50 km of a volcano (note, buffers around volcanoes may require ‘clipping’ to coincide with the coastline). Manually add downstream channels. A 50 km buffer is created around the young-mountain sourced streams and their downstream channels that are suspected to receive major contributions of water and/or sediment there from. Also include rivers flowing across Quaternary volcanic deposits (Type=11).
  7. Intersect mountain-source rivers on the climatically-classified alluvial theme to define the basic AD (<1000 mm), low-RD, and high-RD (>1500 mm) As risk categories. These polygons are refined by intersecting the 50 km river buffer onto the alluvial polygons.
  8. Intersect the classified alluvial polygons onto the population density theme. Then repeat the intersection onto the continent theme to assign country names to the polygons.
  9. At a continental level, bedrock aquifers in all areas of Tertiary mountains are also considered to be at risk from geothermal arsenic. Use the ‘select by theme (intersect)’ procedure to classify the alluvial polygons that intersect the Tertiary mountains (plus volcanoes with 50 km buffer) in order to define supplementary geothermal-As risk class.
  10. Assign the As-risk according to Table 2.
  11. After assigning the arsenic-risk (process) categories, alluvial deposits older than the Upper Quaternary ([Age]<=10) were excluded where mobilisation was predicted to be due to reductive dissolution, in line with widespread experience from Asia.
  12. Use the XTOOLS add-in to assign areas, which are multiplied by the population density to calculate the population at potential risk (without knowledge of whether groundwater is used for drinking or irrigation). The population calculation uses the parameters listed in Table 3. Finally the population at risk estimates are summarised by country.

Sulphide-Oxidation in Bedrock Aquifers

Globally, arsenic mobilised by sulphide-oxidation represents a much smaller risk than that of RD- or even AD-type mobilisation in alluvium. However, it can be locally severe, and occurs both inside and outside young mountain belts. The method outlined below aims to predict the likelihood of such natural occurrences of arsenic based on the observation that most such cases occur in mining regions even though pollution is not attributable to mining activities. A flowchart of the data processing methodology is shown in Appendix II.

The basic proposition is that arsenic is present as an impurity in Fe, Cu, Zn and Pb sulphides as well as in arsenic-minerals such as arsenopyrite, realgar and orpiment with which they may be associated. It is also well known that arsenic is spatially associated with areas of gold mining in Ghana, Burkina Faso, Chhattisgarh (India), and in Washington State (USA), with copper mining in Chile, silver mining in Mexico, and tin mining in Thailand and Cornwall. An important point here is that the association of arsenic is with sulphide minerals and not the specific element, and hence can be released to water by oxidation (whereas oxide or carbonate minerals would tend to be stable). Therefore the form of mineral deposit is significant, and so deposits produced by sedimentary (e.g. placer gold and many iron ores) of weathering (e.g. bauxite) processes can be rejected. Likewise, ferrous metals, which are mostly resent as oxides and gemstones which are not expected to have causal relationship to arsenic release were rejected. ESRI (1996) produce global data set on mineral resources. Irrelevant materials were excluded as follows:

( [Group] <> "Ferrous metals") and ([Group] <> "Fuels") and ([Group] <> "Gemstones") and ([Group] <> "Nonmetals") and ([Type] <> "Sedimentary") and ([Type] <> "Weathered" ).

Within the remaining data-set, the principal element was used to define a potential association with arsenic pollution risk: silver (Ag), gold (Au), arsenic (As), copper (Cu), mercury (Hg), nickel (Ni), lead (Pb), antimony (Sb), and tin (Sn). Only one location, Darrydagskoye, a post-magmatic deposit in Azerbaijan, was mined principally for arsenic. These depsoits were selected:

( [Main_eleme] = "Ag") or ([Main_eleme] = "Au" ) or ([Main_eleme] = "As" ) or ([Main_eleme] = "Cu" ) or ([Main_eleme] = "Hg" ) or ([Main_eleme] = "Ni" ) or ([Main_eleme] = "Pb" ) or ([Main_eleme] = "Sb" ) or ([Main_eleme] = "Sn" )

It is probably unwarranted to assume that that the nominal size (small, medium, large and very large) of the mineral deposit, which is also a result of the value of the element, reflects the extent of mineralization, and hence an arbitrary buffer size of 30 km was applied. A variable buffer cold be used, but currently there is no theoretical or empirical basis to assign values to the buffers.

To avoid double-counting, all alluvial-risk areas identified by the previous methodology are excluded by an ArcView union operation, following by deletion of the overlapping areas. Because pollution involves oxidation, it was considered that high temperature and/or low rainfall would make pollution more likely.

Glacial Aquifers

Arsenic pollution of glacial and fluvio-glacial aquifers is well-known in the USA, Canada and Finland, but largely unrecognised in other glaciated regions. Some cases, such as in Alaska, lie within young orogenic belts such as would be (indeed are) predicted by the alluvial model described above. However, the majority of cases come from stable continental regions. Erickson and Barnes (2005) showed that wells drilled on deposits of the youngest glacial advance in the USA (the Wisconsinan) were much more likely to be polluted in Minnesota, Iowa and the Dakotas. In Finland, arsenic polluted groundwater was correlated with arsenic anomalies in glacial sediment (Tanskanen et al., 2004). However, in most countries of interest here, the requisite geochemical atlases probably do not exist, and it would be simpler, and far more reliable, to survey groundwater directly.

The ESRI (1996) database includes mapping and classification of glacial deposits, and these have been compared with the occurrence of arsenic pollution. Unfortunately, even after dividing these deposits by their age does not help greatly to localise the potential risk. These deposits cover the large part of Canada and a continuous swathe of the northern USA from Montana to the Atlantic. It appears that many of the most serious cases of pollution in glacial aquifers are located in buried valleys

Relatively few countries that have substantial covers of glacial deposits are likely to come under UNICEF’s priority programmes. However, it should be noted that glacial aquifers should be considered suspect, and expert advice sought. This model is not used further here.


Appleyard, S.J., J. Angeloni and R. Watkins. 2006. Arsenic-rich groundwater in an urban area experiencing drought and increasing population density, Perth, Australia. Appl. Geochem.; 21, 1, 83-97.

Berg, M., C. Stengel, P.T.K. Trang and 5 others. 2006b. Magnitude of arsenic pollution in the Mekong and Red River Deltas Cambodia and Vietnam. Science of the Total Environment; 372, 2-3, 413-425.

Bexfield, L.M. and L.N. Plummer. 2003. Occurrence of arsenic in ground water of the Middle Rio Grande Basin, central New Mexico. In: A.H. Welch and K.G. Stollenwerk (Eds). Arsenic in groundwater: geochemistry and occurrence. Springer, New York, 295-327.

Erickson, M.L. and R.J. Barnes. 2005. Glacial sediment causing regional-scale elevated arsenic in drinking water. Ground Water, 43, 6, 796-805.

ESRI. 1996. ArcAtlas: Our Earth (GIS data). Environmental Systems Research Institute Inc., USA.

Kelly, W.R., T.R. Holm, S.D. Wilson and G.S. Roadcap. 2005. Arsenic in glacial aquifers: sources and geochemical controls. Ground Water, 43, 4, 500-510.

McArthur, J.M., D.M. Banerjee, K.A. Hudson-Edwards and 10 others. 2004. Natural organic matter in sedimentary basins and its relation to arsenic in anoxic groundwater: the example of West Bengal and its worldwide implications. Appl. Geochem. 19, 1255-1293.

Nicolli, D.M., J.M. Suriano, M.A. Gomez Peral, L.H. Ferpozzi and O.A. Baleani. 1989. Groundwater contamination with arsenic and other trace elements in an area of the Pampa, Province of Cordoba, Argentina. Environ. Geol. Water Sci., 14, 1, 3-16.

Ravenscroft, P., H. Brammer and K.S. Richards. 2009. Arsenic pollution: a global synthesis. Wiley-Blackwell.

Schreiber, M.E, J.A. Sino and P.G. Freiberg. 2000. Stratigraphic and geochemical controls on naturally occurring arsenic in groundwater, eastern Wisconsin, USA. Hydrogeology J.; 8:161-176.

Smith, A.H., M. Goycolea, R. Haque and M.L. Biggs. 1998. Marked increase in bladder and lung cancer mortality in a region of northern Chile due to arsenic in drinking water. American Journal of Epidemiology, 147, 7, 660-669.

Tanskanen, H., P. Lahermo and K. Loukola-Ruskeeniemi. 2004. Arsenic in groundwater in Kittiliä, Finnish Lapland. In: K. Loukola-Ruskeeniemi and P. Lahermo (Eds). Arsenic in Finland: Distribution, Environmental Impacts and Risks. Geological Survey of Finland, 123-134 [in Finnish].

Webster, J.G. and D.K. Nordstrom. 2003. Geothermal arsenic. In: Welch A.H. and K.G. Stollenwerk (Eds). Arsenic in groundwater: geochemistry and occurrence. Springer, New York, 101-126.