Arsenic in Annex C-8

In the last few years it has been recognized that Bangladesh is affected by one of the worst cases groundwater contamination in the world. With the exception of diarrheal diseases and coastal salinity in rivers, until recently water quality was not perceived as a problem in Bangladesh, as may be recognized by reference to the earlier National Water Plan projects (MPO, 1986 and 1991). Medical evidence of arsenic poisoning was first discovered in West Bengal in 1978, although the connection with groundwater contamination was not recognized until some years later. Further surveys were carried out during the 1980s (eg Chakraborty et al, 1987; Mazumder et al, 1988), culminating a national report by PHED, India, in 1991. In 1993, DPHE, Bangladesh, tested for and identified arsenic at Chapai Nawabganj, near the border with West Bengal, but it was not until 1995 that the existence of arsenic in Bangladesh became widely known. It was not appreciated until early 1997 that arsenic contamination extended over large parts of the country. By the end of 1998, a national assessment of the extent of contamination across the country had been made (DPHE 1999). At around the same time, projects of both the Ministry of Heath and the DPHE as well as various NGO's began work on extensive testing, awareness raising and providing small-scale arsenic removal systems.

The first major investigation of arsenic contamination in West Bengal by PHED (1991) contained much analytical and descriptive detail but did not suggest an origin for the contamination other than ruling out anthropogenic sources. In the absence of any formal explanation, a number of (non-hydogeological) studies (eg Das et al, 1994 and 1996; Chaterjee et al, 1995) went onto suggest that extensive seasonal pumping of groundwater for irrigation is responsible, although they did not put forward any direct evidence to support the idea. This idea came to be known as the ‘pyrite oxidation’ hypothesis and was formally described by Mallick and Rajagopal (1996), although again unsupported by field evidence. The idea is based on the assumption that arsenic is present in the sulphide minerals pyrite and arsenopyride. According to the theory, lowering of the water table due to pumping introduces oxygen, which causes the breakdown of pyrite and releases arsenic, iron and sulphate into the water. Recharge during the subsequent monsoon then flushes the arsenic into the underlying aquifers. This explanation, whereby the arsenic pollution is caused by man’s over-exploitation of groundwater, became well known and led to calls for the banning of tulle well irrigation.

The debate between these views hinges on a number of issues. First, the chemistry of the affected water points towards strongly reducing conditions and hence that release by oxidation is very unlikely. Evidence for reducing conditions includes direct measurement of redox potential, low-to-negligible concentrations 0£ dissolved oxygen nitrate and sulphate combined with high concentrations of iron, manganese and bicarbonate plus the occurrence of dissolved gases such as carbon dioxide and methane (Ahmed et al, l998). Only limited evidence has been presented for the occurrence of arsenopyrite or arsenin-rich pyrite, while Nickson et al (1998) showed that pryrite is present in a stable diagenetic (framboidal) form. This indicates that it has grown after deposition and is a sink rather than a source for arsenic. Imam et al (1998) identified the ubiquitous presence of iron-rich coatings on sands from the affected aquifers. Further, PHED (1991) showed very high arsenic concentrations (2000mg/kg) in the ferruginous coatings of aquifer sands. Chemical analyses of sediments from the arsenic affected aquifers show a good correlation between iron and arsenic but no correlation between sulphur and arsenic. In addition, arsenic concentrations are low at or immediately below the zone of water table fluctuation. The combination of reducing conditions, the problem of accounting for the absence of sulphur, and the near-neutral pH all point strongly towards the oxyhydroxide reduction hypothesis as being the predominant mechanism of arsenic release in the Bengal Basin. 

Subsequent field and laboratory investigations have all confirmed the ‘oxyhydroxide reduction’ hypothesis. These studies include investigations in Meherpur by Burren (1998) and Perrin (1998), in Jessore by AAN (1999) and in Noakhali by Mather (1999), as well as the major national investigation by DPHE (1999). In addition, there have been two published discussion on the 'oxyhydroxyde reduction' explanation in although again unsupported by field evidence. The idea is based on the assumption that arsenic is present in the sulphide minerals pyrite and arsenopyride. According to the theory, lowering of the water table due to pumping introduces oxygen, which causes the breakdown of pyrite and releases arsenic, iron and sulphate into the water. Recharge during the subsequent monsoon then flushes the arsenic into the underlying aquifers. This explanation, whereby the arsenic pollution is caused by man’s over-exploitation of groundwater, became well known and led to calls for the banning of tulle well irrigation. In 1997, an alternative explanation was independently put forward by an Indo - Swedish group (Bhattacharya et al, 1997) and an Anglo - Bangladeshi group (Nickson 1997; Nickson et al, 1998 and 2000) and has become known as the 'oxyhydroxide reduction' hypothesis. Both these groups presented their ideas in international scientific journals where the origin of arsenic was the main subject of the papers. According to the theory, arsenic weathered from the hard rock areas of India is carried in the suspended load of the rivers adsorbed onto iron oxides or hydroxides. Following deposition of this fine sediment, decomposition of organic matter leads to strongly reducing groundwater conditions which causes dissolution of the iron oxyhydroxides and consequent release of arsenic into solution. This hypothesis also accounts for the pervasive occurrence of iron in wells beneath the floodplains.

Other studies that contain relevant data but do not take a strong position regarding the origin of arsenic in water include Ghosh and De (1995), Nag et at (1996), NRECA (1997), Safiullah (1998), Talukder (1998) and Johnston (1998). Other hydrogeological and/or geochemical studies are underway by the US Geological Survey, the Lamont-Doherty Earth Observatory and the Massachusetts Institute of Technology (MIT). From correspondence with these groups, no major revision of the presently accepted (oxyhydroxide reduction) explanation is expected, although clarification of the mechanisms and rates of the processes can be expected.

Arsenic in the environment may exist in a number of organic and inorganic forms. The most toxic form of arsenic is arsine gas, followed by arsenite and arsenate which are the forms of arsenic normally found in groundwater. Organic forms of arsenic such as found in many foodstuffs are much less toxic (Morton and Dunnette, 1994; FDA, 1993) than the inorganic forms in drinking water.

Arsenic is both toxic and carcinogenic (Morton and Dunnette, 1994). The clinical effects of chronic arsenic poisoning range from skin ailments (keratosis and melanosis), through damage to internal organs to gangrene and various forms of cancer. The clinical effects of arsenic in drinking water in Bangladesh have been extensively demonstrated by Dhaka Community Hospital and NIPSOM amongst others.

The present drinking water standard adopted in Bangladesh and India is 0.05mg/l (50ppb) of (total) arsenic, which followed WHO advice at the time the standards were set. In 1993, the WHO proposed a new provisional guideline value for arsenic of 0.01mg/l (10ppb). This limit is based on practical limits of detection of arsenic and not on the normal health risk criteria. Conventionally, limits for carcinogens are set at the level of one excess lifetime cancer per 10,000 of population exposed. The provisional value of 0.01mg/l corresponds to an estimated risk of six additional skin cancers per 10,000 exposed persons.

It is anticipated that there will be demands to lower the standard for acceptable concentrations of arsenic in drinking water m Bangladesh. However, in considering the implications of tbe arsenic drinking water standard, it is important that the intervention strategies consider the dose-response relationships. As indicated above, the number of cases of arsenicosis resulting from drinking water at the current standard is only a small proportion of the exposed population. However, published dose-response curves from China (Lianfang and Jianzhong, 1994) suggests that the prevalence rate of arsenicosis rises to 10% at about 0.4mg/l, and to the order of 40% of the exposed population at concentrations of around 0.6mg/1. As discussed below, such concentrations are not rare in Bangladesh.

The most reliable determination of arsenic in groundwater is carried out by analyzing filtered and acidified samples in a laboratory using an atomic adsorption spectrometer. However, the scale of the problem in Bengal calls for a field test kit. Field kits, however, have often been criticized for being unreliable. In addition, DPHE (1999) have demonstrated the unreliability and lack of quality control in many laboratories in Bangladesh. However, DPHE (1999) also demonstrated that any error consistently underestimates the true level of arsenic, and that field - kits can provide valid regional descriptions of arsenic occurrence. 

The practical detection limits of field - kits are close to the present drinking water standard of 0.05mg/l, and become unreliable for waters close to this value. Importantly, field kits do reliably identify highly contaminated wells. However, if a lower drinking water standard is adopted, the present field kits will not be able to differentiate between safe and unsafe water. The only field method applicable at these concentrations is a prototype field - kits known as the “Arsenator”.

The most comprehensive, systematic and reliable survey of arsenic in groundwater was carried out under DPHE’s Groundwater Studies for Arsenic Contamination in Bangladesh project in 1998 and 1999. The summary statistics from their national survey (excluding the Chittagong Hill Tracts) are given below in Table C8.3.l and they represent the least biased assessment of the groundwater affected by arsenic contamination. However, because the population density is not evenly distributed across the country, these percentages cannot be directly transformed into population affected.

Source: Groundwater Project for Arsenic Contamination in Bangladesh project and http://www.bgs.ac.uk/arsenic.

The DPHE study showed there are problems in measuring the concentration of arsenic with accuracy even in a laboratory, which need to be recognized when evaluating survey results. This shows up most clearly when the results of tests carried out in DPHE laboratories are compared with those carried out on the same samples in the UK, as is done in Table C8.3.2 below. This shows a statistically significant relationship that indicates the DPHE laboratory results overstate the concentration of As by 25% on average. This does not greatly change the findings about whether a particular well is harmful to health or not. However, differences can be large at higher concentrations.

C8.3.3    Regional Distribution of Arsenic in Groundwater

The principal source of information on the distribution of arsenic comes from the surveys and data compilations of the Groundwater Studies for Arsenic Contamination project (DPHE, 1999). From the perspective of human exposure, the most useful method of representing the geographical distribution of arsenic is percentage of wells that exceed the drinking water standard. In this form it is possible to combine laboratory data with field kit results to access a much larger database. Combining this data on l8,000 tests in over 3,000 unions with the NRWD GIS map of unions, it has been possible to produce a probability Of arsenic occurrence across the country as shown in the four maps in Figure C8.3.1. The percentage of wells contaminated was assigned to the centre point of the union and then converted to a smooth surface using the inverse distance-weighted technique. 

The maps in Figure C8.3.l show the probability of exceeding the four threshold concentrations of 0.01lmg/l (the WHO guideline value), 0.05mg/l (the current drinking water standard), 0.20mg/l and 0.40mg/l. No depth separation of wells has been applied in this map, but, as described below, the deep aquifer system is almost completely uncontaminated. If a lower drinking water standard (e.g. 0.01mg/l) is adopted the affected area will expand considerably. On the other hand, the areas in which extremely high arsenic concentrations occur are much more restricted, pointing the way to prioritizing interventions.

• Arsenic occurs in the catchments of all major rivers of Bangladesh. This demonstrates the existence of multiple source areas.

• Aquifers beneath the hills of greater Sylhet and terrace areas (Barind and the Madhupur Tracts) are virtually free of arsenic contamination. These aquifers are formed of older sediments from which arsenic has either flushrd out or has been immobilized.

• Of the major rivers, the Brahmaputra and Teesta floodplains are least affected, while Meghna floodplains are worst affected.

• Over large parts of North west and North Central Bangladesh and the Chittagong Coastal Plain the probability of wells being affected by arsenic is less than by 5%.

• Arsenic is far less likely to occur in aquifers less than 200m.

The probability of wells being contaminated is strongly related to the depth at which they are screened. The probability of exceeding the drinking water standard in the upper aquifer system is high, but as shown in Table C8.3.5, in wells screened at lower levels, the risk from arsenic is minimal. There has been some controversy (Chowdhury et al, 1999) as to whether arsenic increases or decreases with depth. However, as shown below and elsewhere, there is evidence that an increase in depth of screening is related to a sharp initial increase in the occurrence 0f contamination followed by a gradual decease.

It is clear from the table, in terms of both the probability of contamination and the maximum probable concentration, that well depth provides a clear basis for an arsenic avoidance strategy, at least in the short to medium term. Nevertheless, it is also important to note that in the DPHE survey 58% of the wells have concentrations below the WHO guideline and 75% of them conform to the current drinking water standard. Therefore there is no basis for a general abandonment of the shallow aquifers, again at least in the short to medium term. Figure C8.3.3 compares the depth-profile of the probability arsenic exceeding 0.05mg/l with the typical depth ranges of different types of wells, showing how different technologies are differentially exposed to arsenic.

Despite an obvious lack of long-term monitoring data DPHE (1999) have deduced that the proportion of wells that exceed the drinking standard plus concentrations of 0.10, 0.15 and O.20rng/l) increases with age up to an age of eight to ten years. In the long-term, concentrations of arsenic should decline as the stock of arsenic in the sediments is depleted. DPHE (1999) supported by modeling studies by Cuthbert (1999) attribute this to slow leakage from overlying aquitards. Their practical conclusion is that ensuring a distance of 10m or more between the well screen and an overlying aquitard massively reduces the risk of delayed contamination. Unfortunately, it appears to he common (and understandable) practice to screen at the top of the first available aquifer. The question of seasonal variations of arsenic concentration is uncertain. AAN (1999) have certainly reported differences between seasons, hut have not clearly proven seasonal fluctuations. Such fluctuations are possible, especially in shallower wells and where pumping is most intense. The Eighteen District Towns Water Supply Project (18DTP, 2000) periodically monitored selected wells from 1996 to 1999, but reported no consistent trend or pattern. Ahmed (2000) reports that nested monitoring wells installed in three Thanas under Phase II of the Groundwater Studies for Arsenic Contamination project showed no significant changes over seven months of monitoring except for a rise in the first few weeks after installation. The latter concentration has obvious implications for the testing and commissioning of new wells.

Figure C8.3.3 : Depth Distribution of Arsenic in Groundwater (Click on the Figure for a larger view)

C8.4.1 Arsenic in surface Water

Arsenic contamination in Bangladesh is principally a problem in groundwater and it is often stated that surface water is safe from arsenic. However, concentrations up to 0.5mg/l (500ppb) have been reported from a partially dried up river in Meherpur (Burren, 1993). While this is believed to be an exceptional ease, it demonstrates that surface water cannot be assumed to be safe from arsenic. It is anticipated that arsenic contamination of surface water is most likely in ponds or in relatively stagnant water bodies that are fed by groundwater flow. This is likely to be a seasonal phenomenon, occurring in the dry season. The occurrence of arsenic in some surface waters also raises the possibility of arsenic uptake by fish.

C8.4.2 Arsenic in Dug Wells

It was noted earlier that the risk of arsenic contamination is low in very shallow groundwater. It has been widely reported that dug wells are safe from arsenic. Dr. Dipankar Chakraborty reports that analyzing more than 100 dug wells from West Bengal and Bangladesh and that none exceeded 0.05mg/l. The problem with dug wells is not arsenic, but the risk of diarrhoeal disease.

The majority of private water supplies are obtained from dried tube wells between 10 and 100 metres deep. There are no reliable figures but there may be between eight and eleven million such wells the country. These sources pose the principal threat to public health from arsenic. Nevertheless, the majority (>70%) of such wells have arsenic contamination levels of less than 0.05mg/l and they also pose a low risk of diarrhoeal disease. As noted above arsenic concentrations may increase over time (most probably in wells screened very close to aquitard layers). The continued safety of presently uncontaminated wells is an important water management issue.

C8.5.1 Population Exposed

A simple assessment of the population exposed to hazardous levels of arsenic was made using the 1991 census and the National Arsenic Survey of the Groundwater Studies for Arsenic Contamination project carried out in 1998 and 1999. That survey is the only one that has a large, evenly spaced and reliable arsenic concentration data over the whole country (excepting the Chittagong Hill Tracts). The arsenic distributions shown earlier in' Figure C.8.3.2 have been combined with the population distribution to estimate the exposed populations at the 10, 50, 200 and 400ppb levels. The estimates are made for the year 2000 assuming a population growth rate of 2.2% per year and that 95% of the population draw their water from groundwater supplies. No explicit account has been taken of the difference in arsenic between the deep and shallow aquifers, however, the survey design was stratified to reflect the importance of deep and shallow aquifers in each upazila. The results of the assessment are given tnTahleC8.5.1 below.

•  Treatment of arsenic contaminated groundwater;

•  Develop arsenic-free groundwater sources; or

•  Develop surface water sources such as rivers, ponds and rainwater. All three approaches have relevance in different settings, and all have some difficulties. Hydrological and climatic factors obviously constrain the technical selection of methods, while the scale (eg household, community, municipal) of the operation will determine their social and economic suitability. Detailed discussion of the selection of methods is given elsewhere in the DDS report. The choice of method must also be considered in relationship to different drinking water standards. It is known that common methods such as aeration and coagulation-filtration all effectively reduce arsenic to around 0.05mg/l but not to 0.01mg/l.

C8.6 Trends of Arsenic in Groundwater

Following the description of the occurrence of arsenic earlier, a number of sustainability questions arise which can be divided into those affecting the shallow or main aquifer (i.e. wells less than 100-120m deep) and those in the deep aquifers (wells more than 150-200m deep). Important questions affecting the use of the shallow aquifers are:

• Will presently uncontaminated aquifers become contaminated in the future? This relates particularly to the extensive uncontaminated aquifers such as the Dupi Tila sands beneath the Madhupur and Barind Tracts.

• Will uncontaminated wells in partially contaminated aquifers become contaminated in the future? There are many cases of sharp lateral variations of arsenic concentrations within villages, giving rise to obvious fears of lateral migration of arsenic.

The issues relating to the sustainability of abstraction from the deep aquifers are:

•  Will wells in deep (>150-200m) aquifers that are overlain by contaminated aquifers become contaminated by downward leakage of arsenic? This is of greatest concern where increased pumping from the deep aquifer is accompanied by the abandonment of existing wells in the upper aquifer. This is usually a local problem, where the degree of risk depends on the intensity of abstraction and the properties of the intervening strata.

•  Will increased pumping from deep aquifers induce saline intrusion in the coastal areas? Unlike the above concern, the risk results from the cumulative effects of deep well abstraction over a large region and will impact most on water supplies closest to the coast.

The above four questions may then be reformulated as three hydrogeochemical questions:

(a) Will conditions in uncontaminated aquifers change sufficiently to mobilize from sediments where arsenic is presently stable in the solid phase?

(b) At what rates will water, salinity (a conservative tracer) and arsenic migrate laterally and vertically through the aquifers?

(c) If arsenic is drawn into presently uncontaminated layers, will it remain in solution or be adsorbed on to the solid phase?

Regarding question (a), given that the main driving force for mobilization of arsenic is decay of organic matter in the sediment and this is a finite quantity, the answer is likely to be ‘no’. Only in the exceptional circumstances such as infiltration of landfill leachate is groundwater likely to become so reducing as to mobilize arsenic from the sediment.

With reference to question (b), tile migration of arsenic is constrained by the rates of water movement and by the effects of adsorption. Both have been studied by numerical modeling by DPHE (1999) arid Cuthbert (1999). There is currently significant uncertainty concerning the quantification of sorption parameters for arsenic. Both studies show even without including the effects of sorption that, under realistic scenarios, arsenic will not move more than a few meters a year laterally or vertically. With only modest sorption, the migration rates drop to less than a meter per year.

Uncertainty concerning sorption parameters prevents a precise answer to question (c). However, the direction of the process is clear and it will operate to some degree to slow down arsenic mitigation.

C8.6.1 Shallow Aquifers

Extensive areas presently free of arsenic are expected to remain so. The reasons are twofold. In the coarser aquifers (such as beneath the Brahmaputra and Teesta floodplains), little arsenic and organic matter accumulated at the time of deposition. In the older aquifers (such as the Madhupur and Barind Tracts and Dhaka City), arsenic has either been flushed out or is combined with more stable iron minerals (such as hematite or goethite) and the aquifers' redox conditions are apparently not sufficiently reducing to mobilize any arsenic that may he present in the sediment.

The future trends of arsenic in partially contaminated areas are more difficult to predict. The modeled migration rates and inferred increases over time strongly suggest that vertical leakage is the critical factor, leading to delayed contamination of tube wells. This may be expected to continue if well design practice is not changed. On the other hand, the risk of contamination due to lateral migration of more than a few tens of meters is small. For the 25-year planning horizon of the Plan, the probable upper limit of lateral migration is considered to be of the order of 50 meters, and vertical migration smaller still. Further research is needed to provide better estimates of migration rates for arsenic, particularly where inter-aquifer movement is involved. Sufficient is known to be confident of no very rapid movement, and some research is underway on this subject, but the results should be reviewed after a few years.

C8.6.2 Deep Aquifers

There is very little arsenic contamination in aquifers below 150-200m. The main and deep aquifers are usually separated by lower permeability strata. The vertical hydraulic resistance and anisotropy of the intervening strata suggest that the migration of arsenic from contaminated shallow aquifers would occur in a period of not less than decades and probably centuries. If the probable, but poorly quantified, effects of sorption are included, it is like~ that the deep aquifer system offers the prospect of permanent or at least very long term, sources of drinking water that is free of bot arsenic and bacterial contamination.

Use of the deep aquifer system will be of most importance in the coastal regions of Bangladesh. There the risk of arsenic must be assessed in parallel with that of salinisation. In this hydraulic situation, the critical issue is the source of water that replaces the water pumped from deep wells. While existing performance in the most heavily abstracted areas such as Khulna City is encouraging, it is impossible at present to say how much more abstraction can be supported. There is a need for a detailed and systematic assessment of the resources of the deep aquifer system to be implemented in parallel with the water supply development.

C8.7 Arsenic, Irrigation and Agriculture

The question as to whether there is a causal relationship between arsenic and tube well irrigation was addressed earlier, and it was concluded that here is no such causal link. However discussion of the inter-relation of arsenic and irrigated agriculture must he extended to ask whether arsenic in irrigation water affects the growth of crops, and whether arid how arsenic in irrigation water or soil might accumulate the food chain?

C8.7.1    Arsenic in Food grain

There are few analyses on arsenic in food grain in Bangladesh. However, no reports are available of analyses of rice or wheat containing arsenic at levels of health significance. Most analyses show arsenic below detection limits. 0f the food grains rice has the most potential to take up arsenic due to the anaerobic root zone (Chappell, 2000). No comprehensive study has been carried out, although a study is underway at CSIRO in Australia and results are expected in late 2000. The CSIRO study will make a survey of arsenic in vegetables and crops, plus their associated soils and water sources across Bangladesh and West Bengal. The long-term objective is to integrate the arsenic-crop-soil-water-medical relations to develop a landscape risk assessment model.

A case study of arsenic at two shallow tube wells in Faridpur by Hasan (1998) is described in Table 8.7.1 below. When considering the total arsenic concentrations in soil, it should be remembered that the arsenic may be present from the time of deposition of the sediment and has not necessarily been added from irrigation water. The high arsenic concentrations in soil at the new well with a low arsenic concentration in groundwater make this clear. Chappell (2000) reports that an arsenic content of 10mg/kg is typical for soils world-wide. Certainly the arsenic in the Devinagar soi1 profile could not have been accumulated from irrigation water in the single year of operation. This therefore makes it uncertain as to whether there has been significant accumulation of arsenic in the Shasa profile. The major uncertainty in the processes is the quantity of arsenic removed from the system (to the atmosphere) by biomethylation under the action of fungi, yeast and bacteria.

Sample Type	Range of Arsenic Concentrations (mg/l or mg/kg)
	Well-1 (Shasa Est. 1989)	Well-2 (Devinagar Est. 1998)
Tube well	0.246 (Fe - 8.8)	0.083 (Fe - 3.5)
Field Channel	0.198 (Fe - 4.5)	0.075 - 0.080 (Fe - 3.0)
Soil Samples	6.6 - 10.6	5.3 - 8.0
Rice Root	224	162
Rice Shoot	2.70	Nd
Rice Leaves	4.35	4.25
Rice (not paled)	0.00	0.00
Rice (paled)	0.00	0.00

Source: Hasan (1998)

At the two STW sites, no arsenic was found in the rice, but high arsenic concentrations were found in the root. This is consistent with the physiology of rice and other plants where the distribution of arsenic is in descending order from root to stem to leaf and edible parts (Huang, 1994). Also the results from the 'Devinagar' site suggest the possibility that arsenic in the root might be derived from the soil. There is also evidence that arsenic concentrations are being reduced by aeration along with iron between the well and the field.

The important conclusion from Hasan’s study is that irrigated rice plants may be exposed to sources of arsenic in soil and water without transferring the arsenic to the directly edible parts of the plant. However, the question as to whether the cultivation of irrigated rice is sustainable is fundamental to the formulation of the NWMP and any serious uncertainty in this area must be explicitly addressed.