Arsenic in Hydro-geo-biospheres of the Mekong River Watershed: Implications for Human Health

Abstract

This chapter assesses human health risks of inorganic arsenic (As) from drinking well water and consumption of rice irrigated by high-As groundwater in the Mekong River Delta. Geogenic inorganic As (iAs) occurring at elevated levels in groundwater has been detected in more than 70 countries. Among mostly rural residents relying on groundwater for drinking, this exposure has resulted in negative health consequences including visible skin lesions, multiple internal organ cancers, numerous invisible non-cancer health effects such as cardiovascular diseases, and premature deaths. In the Mekong River Delta (MRD, defined by elevation <10 m above sea level in this book), As issues in groundwater have been documented as early as 1999 in Cambodia, with literature reporting its occurrence in Vietnam since 2005. Since the early 2000s, efforts have been made to test for As in about 100,000 wells from Cambodia, Laos, Vietnam and Thailand. Here, a combined dataset with a total of 94,768 unique As tests was analyzed to illustrate the spatial patterns and to assess the health risks of drinking well water As in Cambodia and in southern Vietnam. Although knowledge is far more limited, an attempt was also made to examine the potential health risks associated with iAs exposure from rice, a major staple for the MRD. Here, irrigation using highly As enriched groundwater for rice cultivation has expanded this environmental health problem from the hydrosphere (water) to the geosphere (soil) and, in turn, the biosphere (rice, and ultimately humans). Of 41,928 tests in Cambodia, 35.8% exceeded 10 μg/L, the WHO guideline value for drinking water As, while 21.5% exceeded 50 μg/L, the Cambodian drinking water standard. Of 52,858 tests in Vietnam, the exceedance rate for 10 μg/L, which is also the Vietnamese drinking water standard, is 10.0%. High As wells, regardless of whether it is relative to 10 or 50 μg/L, are located in proximity to the main course of the Mekong-Bassac Rivers, especially within a 5 km distance. The vast majority (>98%) of high-As wells are located in low-lying areas, i.e. <25 m elevation in Cambodia and <10 m elevation in Vietnam. High-As wells occur frequently at shallow depths (<70 m) across the MRD but also at deeper depths (300–500 m) in Vietnam. Due to the clustering of high As wells along the Mekong-Bassac Rivers, extreme human health tolls are identified in 11 districts of Cambodia and 3 districts of Vietnam with a population attributable fraction exceeding 0.1, meaning that >1 in every 10 adult deaths is solely due to drinking water As exposure. The annual excess deaths attributable to arsenic exposure alone is 1204 in Cambodia and 1486 in Vietnam, or 1 in every 27 adult deaths and 1 in every 78 adult deaths, respectively. In addition to uncertainties in bioavailability and toxicity of iAs in rice grains, soil and rice As data, especially rice As speciation data needed for risk assessment, are still limited in the MRD.

4.1 Geogenic Arsenic in Groundwater of Southeast Asia

4.1.1 Groundwater Quality Surveys in Cambodia, Laos and Vietnam

The Mekong River flows south from the northern hills (elevation up to 1000 m) before it enters the low-lying (elevation <10 m) flood plains referred to here as the Mekong River Delta (Fig. 4.1). Occupying much of southern Cambodia (upper MRD 10,000 km²) and southern Vietnam (lower MRD 52,000 km²) with an area of 62,000 km², the Mekong River Delta (MRD) is one of the largest deltas in Southeast Asia inhabited by about 8 million Cambodians and 25 million Vietnamese (Fig. 4.1). The topographic features of the Mekong River Watershed resulted from tectonic uplift and folding caused by the collision of the Indian and Eurasian tectonic plates around 50 million years ago (Lap Nguyen et al., 2000). The climate in the MRD is tropical, with average annual temperatures of 27–30 °C. The monsoonal rainy season lasts from April to November (Husson et al., 2000). The mean annual precipitation ranges from 2400 mm in the west to some 1500 mm in the center and east.

Fig. 4.1

Map of the Mekong River Delta (MRD). The elevation contour lines of 10 and 50 m are shown in green, with the MRD region defined by <10 m elevation, with the national border between Cambodia and Vietnam separating the upper and lower MRD. Extensive river networks are also shown. 11 Districts in Cambodia and 3 districts in Vietnam with high arsenic health risks, defined by Population Attributable Fraction (PAF) values >0.1 (see Table 4.4 for details), are highlighted in dark red outline

Since the mid-1990s, groundwater in the MRD of southern Vietnam has been utilized for domestic use by private household tube-wells. Arsenic (As) contamination of groundwater in the MRD appeared first in literature in 2005 (Stanger et al., 2005). Since 2007, surveys and assessments of As groundwater contamination have been conducted in the region (Fig. 4.2a). The first survey of groundwater (n = 405) was carried out in 2007 in 4 provinces, An Giang, Dong Thap, Kien Giang and Long An, all located in the MRD (Hoang et al., 2010). About half of the groundwater samples collected from An Giang and Dong Thap Provinces contained arsenic concentrations higher than the WHO and the Vietnamese national guideline level of 10 µg/L. Further, that arsenic level in groundwater having distinct spatial patterns was already noted, with distance to the Mekong River and the depth of wells playing significant roles. This provided the first clue suggesting that the Mekong River plays an important role in groundwater arsenic occurrence. In Vietnam, a National Groundwater Monitoring Network for the South (NGMNS), which has been installed since the 1990s, is used to monitor groundwater quality in the MRD. The geochemical dataset for NGMNS wells collected by the Division of Water Resources Planning and Investigation of the South of Vietnam from 1994–2014 showed an exceedance rate of 13.8% relative to Vietnam’s national drinking water quality standard of 10 µg/L, with an average value of 8.5 µg/L. High As concentrations (>100 µg/L) are mostly observed in shallow wells (<60 m) (Ha et al., 2019). This trend is also evident in a 3-dimensional map to illustrate the distribution of As concentrations by analyzing 53,000 groundwater As concentration data of the Department for Water Resources Management in Vietnam (Erban et al., 2014). In this large dataset, 10.5% of samples exceed the WHO drinking water guideline value for arsenic at 10 µg/L. It appears that the arsenic issue is most severe in An Giang and Dong Thap Provinces, which are located near the Mekong River.

Fig. 4.2

Health effects of groundwater arsenic exposure in the Mekong Delta encompassing Cambodia and Vietnam. a Concentration of individual well water arsenic based on 41,928 unique tests from Cambodia (the National Well Database) and 52,858 tests from Vietnam (the Department of Water Resources Management). b Population from General Population Census of the Kingdom of Cambodia 2019 and Completed Results of the 2019 Vietnam Population and Housing Census mapped to administrative district. c Population Attributable Fraction (PAF) distribution, and d excess deaths. The MRD area (<10 m elevation) is outlined in green

In Cambodia, unsafe levels of As in shallow groundwater were first documented in 1999 in an unpublished report by the Japanese International Cooperation Agency (JICA), submitted to the Cambodian Ministry of Rural Development (Phan et al., 2010). Consequently, the Ministry of Rural Development organized a national drinking water quality assessment in 13 provinces of Cambodia through a close collaboration of local authorities, research teams and non-governmental organizations (NGOs). Seven provinces in Cambodia were found to have high levels of arsenic in the groundwater. Out of a total of 47,950 wells tested nationally, 30,839 wells were from these 7 provinces and were tested for arsenic by field test kits between 2005 and 2009 (Phok et al., 2018). Up to 35–38% of the tested wells contained arsenic at levels above the WHO guideline value of 10 µg/L and the Cambodian National Standards of 50 µg/L. The occurrence of elevated arsenic in groundwater varies greatly in different watersheds, with only about 2.8–3% of wells along the Stung Saen River in Kampong Thom containing >50 µg/L As while the exceedance rate is 50% on the lower floodplains of the Mekong and Bassac Rivers in Kandal province. An NGO, the Resource Development International—Cambodia (RDI), has also tested over 10,000 wells as part of its programme to assess water quality across Cambodia. Groundwater arsenic is found to be most frequently occurring in parts of Kandal, Kampong Cham, and Prey Veng provinces (Fig. 4.2a).

Unlike Cambodia and Vietnam, survey data are fewer and less available for the MRD region of the Lao PDR. In 2001, UNICEF organized testing of approximately 200 wells of suspected risk areas within the provinces of Attapeu, Savannakhet, Champassak and Saravan (Kim et al., 2011). Some samples have arsenic levels above the WHO guideline of 10 µg/L and only one of them had arsenic concentrations of 112 µg/L, which exceeded the 50 µg/L drinking water standard proposed and later adopted by Lao PDR. Approximately 680 tube-well water samples taken from the Holocene aquifer in the Mekong valley areas were tested in 2004 through campaigns initiated by UNICEF, with support from the government of Lao PDR and the Adventist Development and Relief Agency (Kim et al., 2011). Results showed that 21% of all samples had arsenic concentrations exceeding the WHO guideline value for drinking water of 10 µg/L while 1% exceeded the national standard of 50 µg/L. In 2008, a total of 61 tube-well water samples were also collected from Vientiane, Bolikhamxai, Savannakhet, Saravane, Champasak and Attapeu. The concentrations ranged from <0.5–278 µg/L, with over half exceeding the WHO guideline of 10 µg/L (Chanpiwat et al., 2011).

4.1.2 Arsenic in Groundwater of Thailand

Although Thailand is not located in the MRD, a brief description is included to provide a fuller account of arsenic issues in Southeast Asia. Arsenic has never been found to occur naturally in groundwater in Thailand (Kohnhorst, 2005). Tin mining, or transportation and deposition of arsenic-rich erosion products from elevated areas to downgradient regions, was suggested as the cause of pollution (Kohnhorst, 2005). Ron Phibun District, a well-known area affected by arsenic, has more than a century long history of mining (Fordyce et al., 1995). Abundant arsenopyrite and pyrite, cassiterite and wolframite mineralization occurs widely in pegmatites and geissenized quartz-vein margins throughout the Khao Luang batholith (Fordyce et al., 1995). In 1994, a collaborative study between Thai and British government authorities revealed that arsenic contamination of shallow groundwater ranged between 1.25 and 5,114 µg/L (Kim et al., 2011). Additionally, about 69.6% of the 23 shallow wells contained arsenic concentrations exceeding the WHO guideline for drinking water (10 µg/L). About 15,000 villagers were estimated to be at risk when the problems were first recognized, with over 1000 recorded cases of skin disorders directly attributable to chronic arsenism (Fordyce et al., 1995). As a result, mining ceased, with remediation measures such as the removal of mine waste for disposal at confined local landfills implemented (Wattanasen et al., 2006).

4.2 Health Effects Due to Exposure to Drinking Water As in Cambodia and Vietnam

4.2.1 Rationale for Assessment

A few high-income countries have moved towards adopting drinking water quality standards for As to levels below the WHO guideline value of 10 µg/L (Zheng, 2020). This is because new health evidence suggests that even 10 µg/L may not be protective enough for human health, especially during early, biologically vulnerable stages of life (NRC, 2013). It is worth noting that the WHO guideline value is provisional, and is a recommendation based on treatment performance and analytical achievability. It is possible for these same practical reasons that countries in the MRD region, except for Vietnam, still use 50 µg/L as their national standard, so meeting such standards clearly does not mean “safe”. In the following assessment of health effects, we therefore consider exposure to As greater than 10 µg/L as the “exposed” groups whereas those below as the “reference” groups.

Many epidemiological studies have pointed out that chronic inorganic As exposure via drinking water is associated with mortality caused by many diseases, including lung, skin and bladder cancers, carotid atherosclerosis, hypertension, ischemic heart diseases and skin lesions (NRC, 2013). Due to the latency effect, the disease symptoms usually take years to develop. An important non-cancer disease outcome is cardiovascular disease, one of the major causes of death (Wang et al., 2007). A dose–response relationship has been demonstrated between the level of exposure to As in well water and mortality from ischemic heart disease in a large Bangladeshi cohort (Chen et al., 2011). Regardless of the exact cause of death for each As exposed individual, two studies have reported dose–response relationships between drinking water As levels and mortality rates established based on 115,903 (Sohel et al., 2009) and 11,746 subjects in Bangladesh (Argos et al., 2010). Taking advantage of such well characterized dose–response in mortality among local populations, Flanagan et al. (2012) conducted a health effect assessment that is based also on a careful evaluation of exposure to As using data (n = 14,442) from a national drinking water quality survey of Bangladesh, concluding that 1 in every 18 adult deaths is attributable to chronic As exposure alone.

Although the exposure to drinking water As in the MRD regions of Cambodia and Vietnam has begun as early as the mid-1990s (Berg et al., 2001) due to rural residents’ reliance on private tube-wells, there has never been a quantitative assessment of health effects. Because groundwater quality surveys have revealed considerable heterogeneity in arsenic spatial distribution (Fig. 4.2a), the assessment of exposure and health effects makes an effort to address this feature.

4.2.2 Methods

Here, we adopt methodologies described in Flanagan et al. (2012) to estimate excess deaths using the aforementioned dose–response relationship (Sohel et al., 2009), with modifications described as follows.

Excess death (ED). Mortality rate describes the frequency with which deaths are occurring in a given population over a given time period (for chronic disease, a time duration of 1 year is frequently used). If these are higher than the expected mortality rate in non-crisis or normal conditions in that population, then the difference between the mortality rate under normal and crisis conditions represents the “excess”. For a given time period and a given population size, we can estimate the number of excess death (ED) due to a specific crisis. Thus, the ED describes the mortality attributable to a specific reason, and in our case, chronic exposure to drinking water As, that would have been zero normally without exposure. We recognize that the influence of As on human health is systematic and multifaceted. Therefore, when the relationship between the dose and a disease outcome is available, it is also desirable to evaluate the health effects of that disease. As the first step, it is justified to evaluate ED because it can reflect the overall impact and to indicate the severity of human health toll attributable to As exposure.

Population Attributable Fraction (PAF). To estimate the number of excess deaths in each geopolitical district in the MRD, the population attributable fraction (PAF) can be used. Specifically, it is defined as the fraction measuring how much of the health burden in a population could be eliminated if there had been no exposure (Mansournia & Altman, 2018), as in Eq. (4.1) below:

$$PAF=\frac{O-E}{O}$$

(4.1)

where $O$ is the Observed number of cases, and $E$ is the Expected number of cases under no exposure.

For a region of interest with a total population of $N$, the proportion of the population in four groups representing reference group (10 µg/L), i.e., low (10–49 µg/L), medium (50–149 µg/L) and high (≥150 µg/L) exposure groups, is expressed as ${1-p}_{1}- {p}_{2}-{p}_{3}$, ${p}_{1}, {p}_{2}$, ${p}_{3}$, respectively (Table 4.1). The reference group is chosen on the basis of the WHO guideline value of 10 µg/L for arsenic in drinking water. Given the mortality rate $q$ of the reference group and relative risk (${RR}_{i}$) for different levels of exposure, the number of deaths of low, medium and high exposed groups can be estimated (Table 4.1). Thus, the PAF equation is rewritten by replacing $O$ and $E$ with their corresponding values for the three exposure groups in Eq. (4.2) as follows:

$$PAF=\frac{{\sum }{p}_{i}\left({RR}_{i}-1\right)}{{\sum }{p}_{i}{(RR}_{i}-1)+1}$$

(4.2)

Table 4.1 Terms used in PAF and ED calculations

For ease of calculation, we approximate the relative risk values (${RR}_{i}$) by hazard ratios of non-accidental deaths from Sohel’s 2009 study. The proportion of the population for the low, medium, and high exposure groups is taken as the same as the proportion of wells for the corresponding As intervals (Table 4.1). Substituting these two terms to Eq. (4.2), the PAF is calculated using Eq. (4.3) below:

$$PAF=\frac{{\sum }{p}_{\left(well\right)i}\left({HR}_{i}-1\right)}{{\sum }{p}_{\left(well\right)i}{(HR}_{i}-1)+1}$$

(4.3)

From these resulting PAF values, the annual number of deaths for any geopolitical district (Table 4.1, last row) is estimated by using the area’s population (N) multiplied by the area’s crude death rate (CRD, q in Table 4.1), i.e., the number of deaths in 1000 people in any given year, usually available through each country’s Health Ministry. The crude death rate summarized by the Department of Economic and Social Affairs, United Nations, which is listed in World Population Prospects 2019, is used here. The CDR values used are 6.0 and 6.3 per 1000 people for Cambodia and Vietnam from 2015 to 2019, respectively. The sum of the number of deaths for the low, medium and high exposure groups is taken as the ED attributable to As exposure.

Dataset. Data used for estimation are from diverse sources. In the following, the population and water arsenic datasets are described.

Population data: In Cambodia, the national census was conducted in 1962, 1998, 2008 and 2019 respectively by the National Institute of Statistics. The final census report has population size, trends in fertility, mortality, migration and disability etc. by geopolitical units of Province, District and Commune (Some special indexes are only provided within the province range). Here, the adult population of each district in 2019 is calculated by the population in each district multiplied by the percentage of the population aged 15 years or older because the percentage of those aged 18 years or older is not available. In Vietnam, the demographic information is obtained from the Completed Results of the 2019 Viet Nam Population and Housing Census. This was the fifth Population and Housing Census since the country’s reunification in 1975. Similarly, each district’s population and age ratio are offered as part of the 2019 census results. Adults are also defined as people aged 15 years and above with the number of each district estimated by multiplying the national age ratio and the number of people living in each district.

Arsenic data: The Ministry of Rural Development of Cambodia and UNICEF developed and administered an As well water testing database primarily based on extensive As testing by Research Development International (RDI), an NGO. Well depths of sample wells were also recorded. In addition, a Tonlé Sap Rural Water Supply and Sanitation Project financed by the Asian Development Bank (ADB) provided information on additional hydrochemical parameters (arsenic, iron, chloride etc.) and coordinates (latitude and longitude). In the end, the dataset consists of 42,567 arsenic records of investigated wells sampled between 1997 and 2009 in Cambodia. In Vietnam, the Department of Water Resources Management constructed a database that includes the well depth, year of well installation, coordinates and the arsenic concentration. They shared with researchers 52,858 arsenic concentration data points (Erban et al., 2014) collected from 1907 to 2008. The dataset also includes the coordinates (latitude and longitude), well depth, installation year, and the arsenic concentration. It is noted that because of technical limitations of field testing kits, although discrete values are given for some As measurements, they are not exact values but represent a possible range of values. However, misclassifications remain rare relative to 10 μg/L (He et al., 2022).

The original data of Cambodia has a total of 42,567 well arsenic records from various database sources with columns including well ID, WGS coordinates, sample date, well depth and several water chemical parameters (chloride, iron, manganese etc.). The province, district, commune, and village of each well are also documented, although a fraction of this documentation does not match the attribution from the WGS coordinates. Here, the geopolitical district of each well is assigned only by its WGS coordinate. Furthermore, duplicate data and multiple arsenic values for one well exist. To make sure that every well has its unique value that best reflects reality, data are “cleaned” using rules as follows: (1) If multiple arsenic values of one well exist, the well is removed from the dataset if As values are distributed in more than one classification group stated in Table 4.1. The reason for doing this is that we do not want the screening or data entry errors to be a source of variance of calculated PAF values. This process eliminated 64 data points of 30 wells; (2) If multiple arsenic values of one well exist only in one classification group, then the value with the latest sample date is kept and used for analysis; this process eliminated 150 data points. We have noted that there are 1,119 wells with the same WGS coordinates, although their well depths and As values are different, suggesting that the testing was done in a village or commune though without GPS measurements to determine the exact latitude and longitude of the tested well. In addition, these wells have been assigned a unique well ID thus it is justified to recognize them as different wells so they are not removed from the database. After this exercise, a dataset of 41,928 wells with their unique As values is used for analysis. Fortunately, the dataset from southern Vietnam has been cleaned by Erban et al. (2014). The summary statistics of well water arsenic datasets can be found in Table 4.2.

Table 4.2 Summary statistics of Well Water As data in Cambodia and southern Vietnam

Spatial Analysis by Geographic Information System. A total of 241 administrative districts with 94,786 records of arsenic concentrations of wells in the Mekong Delta are used for spatial analysis using ArcGIS 10.6. Colour-coded maps generated by ArcGIS are used to illustrate the varying degrees of severity of health risks for each district, in addition to showing the individual well water As data classified to 5 groups from cold to warm colours: ≤10, 10–49, 50–149, 150–299, ≥300 μg/L (Fig. 4.2).

Well water arsenic concentrations are classified based on the well’s elevation, its distance to a major river and the depth (Table 4.3; Figs. 4.3, 4.4 and 4.5). Because most of the high arsenic points are located along the main course of the Mekong that includes the Bassac River after the Mekong bifurcates to two major branches in Phnom Penh (Fig. 4.1), a distance analysis was performed in ArcGIS to explore the relationship between the concentrations of As of each well and their locations relative to the Mekong or the Bassac River. It is noted that there are also high arsenic wells near the Steung Saen River, a major tributary of Tonlé Sap. Therefore, in order to distinguish this smaller river from the Mekong and the Bassac, data points (n = 1439) within districts of Kampong Svay, Krong Stueng Saen, Prasat Sambour and Sandan in the province of Kampong Thom are analysed using the well’s distance to the Steung Saen River. A total of 38,866 data points, from the following 10 provinces of Kampong Cham, Kampong Chhnang, Kratie, Kampong Speu, Takeo, Svay Rieng, Tboung Khmum, Kandal, Phnom Penh and Prey Veng are included in the distance analysis to the main course of the Mekong, including the Bassac south of Phnom Penh. Wells (n = 1,623) located in the northern and western most part of Cambodia, far away from the Mekong-Bassac river influence, are excluded so the number of points is less than 41,928. All data points in Vietnam are used. A tool of ‘Generate near table’ in ArcGIS was used to calculate distance and other proximity information between features in one or more feature class or layer. In our analysis, distance from a well to the river is defined as the two-dimensional Euclidean distance. Only the nearest tributary to any given well is used for calculation. Therefore, every well has its unique distance value.

Table 4.3 Distribution of well water [As] by elevation, distance to the closest major river and the depth of each well in Cambodia and southern Vietnam

Fig. 4.3

Distribution of well water [As] in 40,305 tested wells with unique values in Cambodia excluding wells located in the northern and westernmost provinces of Cambodia (a) and 52,858 tested wells in southern Vietnam (b) versus distance to the closest major river course, i.e., Mekong or Bassac Rivers (see Fig. 4.1). The blue and red lines depict the percentages of wells in exceedance of the WHO As guideline the Vietnamese National Drinking Water Standard of 10 µg/L and the Cambodian National Drinking Water Standard of 50 µg/L. For the first 25 km, the analysis is done every 5 km. For wells located at >25 km distance from the closest major river, the analysis is done every 10 km. The inset in (a) shows the results of the analysis of 1,439 tested wells in the Steung Saen River watershed, a major tributary of Tonlé Sap

Fig. 4.4

Distribution of well water [As] in Cambodia (a) and southern Vietnam (b) versus elevation of each well. The red and blue lines describe the percentages of wells that exceed 50 or 10 μg/L As, respectively, and are for the depth ranges of 0–3, 3–6, 6–10, 10–25, 25–50, >50 m for Cambodia (a) and 0–3, 3–6, 6–10, >10 m for Vietnam (b). The dataset has 41,928 records from Cambodia and 52,858 records from Vietnam

Fig. 4.5

Distribution of well water [As] in Cambodia and southern Vietnam versus depth of each well. The red line describes the percentage of wells that exceed 50 μg/L As while the blue stands for that exceeding 10 μg/L As. For Cambodia, the percentage is calculated every 10 m when the depth is <110 m and for the remainder of the wells with depth >110 m. For Vietnam, the percentage is calculated every 10 m when the depth is <200 m and every 50 m when the depth is >200 m. The dataset has 40,925 records from Cambodia (1003 wells lack the records of well depth) and 52,858 records from Vietnam

For PAF and ED estimations, the smallest spatial unit that the analysis can be performed is each administrative district. This is because the population within each administrative district is well documented by the national census. Each well is “projected” to or “grouped” to its own administrative district based on latitude and longitude. Sub-datasets at the district level with population and a given number of wells with known arsenic concentration are generated. To reduce error, districts with less than 10 wells are not considered in the latter calculation. Using Sohel et al.’s hazard ratios for non-accidental deaths (Sohel et al., 2009), PAF and EDs are estimated for each district in the MRD (Table 4.4). There are 85 administrative districts in 16 provinces in Cambodia with population of 7,750,573 (Table 4.4). The adult (>15 yrs old) population of it was estimated as 5,471,904 by multiplying the total population with the adult age ratio of 70.6% from census. There are 146 administrative districts for 14 provinces in Vietnam with a population of 24,493,357 (Table 4.4). The adult (>15 yrs old) population of it was estimated as 18,541,471 given the adult age ratio of 75.7% provided by the census.

Table 4.4 Health risks of drinking water As exposure in the Mekong River Delta of Cambodia and Vietnam

4.2.3 Spatial Characteristics of Well Water Arsenic Exposure and Health Effects

Extremely High As Occurrence Within 5 km Distance of the Mekong-Bassac Rivers. Groundwater As concentration data (n = 94,786) in the MRD region of Cambodia and Southern Vietnam exhibits a clear spatial pattern with high As wells located in proximity to the main course of the Mekong and Bassac Rivers (Fig. 4.2a). It is also evident that Cambodia is hit harder by As than Vietnam. In Cambodia, 35.8% of 41,928 wells tested contain >10 µg/L As, while in Vietnam this percentage is only 10.0% among 52,859 wells (Table 4.2). Moreover, 11.7% of the Cambodian wells displayed >150 µg/L As, far greater than the 2.8% in Vietnam.

Our distance analysis unequivocally demonstrates that in most of the high [As] wells, regardless of whether the definition of high level is relative to 10 or 50 μg/L, the percentages of such high [As] wells is the highest within 5 km distance of the closest major river (mostly Mekong and Bassac), and decreases quickly within 20 km of distance (Fig. 4.3). This decreasing trend continues for wells located 20 km and further distance away from the closest major river, albeit more gradually (Fig. 4.3). In Cambodia, 36.2% of the wells located within 5 km of the Mekong River contain >50 µg/L of As; but this percentage drops significantly to <10% from 5 km onwards (Fig. 4.3a). Although the decline is less dramatic, 52.3% of the wells located within 5 km of the Mekong River contain >10 µg/L of As, decreasing to less than 20% when the distance is >25 km (Fig. 4.3a). In Vietnam, 12.5 and 18.7% of the wells within 5 km of the Mekong River are affected by As greater than 50 and 10 µg/L, respectively. The percentage of wells exceeding 50 and 10 µg/L lowers to less than 5 and 15% when the wells are located between 5 and 25 km distance. Beyond 25 km, the percentage of wells with As concentration >50 and >10 µg/L are less than 1 and 2% respectively (Table 4.3).

In Cambodia, 78.6 and 90.5% of wells with As concentrations greater than 10 and 50 μg/L respectively are within the first 5 km of the Mekong/Bassac River. Kaoh Soutin, a district in the province of Kampong Cham, shows the highest exceedance rate, with 80.1% of 1,335 tested wells containing >10 μg/L As, with an average As concentration of 107 μg/L. It is also where the country’s highest As (2500 μg/L) well (distance to Mekong 1.3 km, elevation 18 m, depth 39 m) is located. Furthermore, a total of 16 wells in Kaoh Soutin displayed 1000 μg/L of As, or the second highest level in the field test kit used. Using 50 μg/L of As as a benchmark, Kaoh Thum, a district in the province of Kandal, has the largest percentage of wells not meeting the Cambodian drinking water standard (58.7%). Both Kaoh Soutin and Kaoh Thum are located adjacent to the Mekong with PAF > 0.1 (Table 4.4). For the wells located in the Steung Saen River watershed, a similar descending trend of As concentration and exceedance rates is evident (Fig. 4.3a inset). Here, 3.0 and 24.2% of wells within the first 5 km of this smaller river contained >50 and >10 μg/L As, with the maximum, mean and median As concentration of 500, 13.7 and 10.0 μg/L, respectively. The well with the highest As concentration is found 100 m from the Steung Saen, with a depth of 26 m and an elevation of 12.8 m.

In Vietnam, 52.8 and 72.6% of wells containing greater than 10 or 50 μg/L are within the first 5 km of the main courses of the Mekong-Bassac Rivers. The highest As point of 1,470 μg/L is found in Cao Lanh, a district of Dong Thap province. Its well depth is 60 m and it is located 3.3 km away from the Mekong River with an elevation of 5.8 m. The district with the highest exceedance rate relative to 10 and 50 μg/L is An Phu from An Giang Province, with the percentages being 89.9 and 87.2% respectively. An Phu is adjacent to the Mekong River, with an extraordinary PAF of 0.22 (Table 4.4).

High Arsenic Occurrence in Low-Lying Areas with <25 m Elevation in Cambodia and <10 m Elevation in Vietnam

In Cambodia, the vast majority of the high arsenic wells (98.3% of all >10 μg/L wells and 99.7% of all >50 μg/L wells) are located below an elevation of 25 m (Fig. 4.4a). In this low-lying area, 38.5% of tested wells had >10 μg/L of As and 23.4% of tested wells had >50 μg/L of As. The maximum, median and mean arsenic concentration among wells at elevations below 25 m is 2500, 10.0 and 63.9 µg/L, respectively. The corresponding values are 500, 0 and 3.6 µg/L when the elevation is higher than 25 m. The highest occurrence rate of groundwater arsenic is from wells located at elevations between 6 and 10 m above sea level, with 29.2 and 43.2% of wells containing greater than 10 or 50 μg/L As, respectively.

In Vietnam, most of the high arsenic wells (99.9% of all >10 μg/L wells and 99.9% of all >50 μg/L wells), are located below an elevation of 10 m in the MRD of Vietnam (Fig. 4.4b). In this low-lying area, 10.1% of tested wells had >10 μg/L of As and 4.9% of tested wells had >50 μg/L of As. The maximum, median and mean arsenic concentration among wells at elevations below 10 m is 1470, 0 and 14.6 µg/L, respectively (Table 4.3). The corresponding values are 100, 0 and 1.0 µg/L when the elevation is higher than 10 m (Table 4.3). The highest occurrence rate of groundwater arsenic is from wells located at elevations between 6 and 10 m above sea level, with 18.2 and 15.1% of wells containing greater than 10 or 50 μg/L As, respectively.

High Arsenic Occurrence at Shallow Depths (<70 m) Across the MRD and at Deep Depths (300–500 m) in Vietnam

In Cambodia, the majority of the wells (99.3%) are from the shallow aquifer with a depth of <70 m (Fig. 4.5a) and a mean depth of 35.2 m. It is noted that only 40,925 wells from 41,928 wells have depth records. Wells with depths between 60 and 70 m exhibit the largest proportion of high concentrations of As, with 49.6% > 10 μg/L and 33.4% > 50 μg/L. The maximum, median and mean values are 1000, 10.0 and 125.2 μg/L for depth intervals in this range. At depth >80 m, the percentage of high As wells, relative to both 10 and 50 μg/L, decrease to <17 and <14%, respectively.

In Vietnam, there are two distinct As peaks at shallow (~30 m) and deep (~350 m) depths (Fig. 4.5b). Still, the majority of the high arsenic wells (78.6% of all >10 μg/L wells and 95.4% of all >50 μg/L wells) are from the shallow aquifer with a depth <100 m (Fig. 4.5b). Wells with depths between 10 and 20 m exhibit the largest proportion of having a high concentration of As > 50 μg/L, with 30.7% > 10 μg/L and 26.5% > 50 μg/L. For wells with depths between 200 and 500 m, the exceedance rate diverges depending on whether 10 or 50 μg/L is used as a benchmark for comparison. Less than 1.5% of wells in this depth interval contain >50 μg/L but a staggering 26.1% of wells contain >10 μg/L of As (Fig. 4.5b), reaching a peak occurrence of 56.8% for wells with depth between 350 and 400 m. The maximum, median and mean values of wells between 350 and 400 m are 70, 20.0 and 15.5 μg/L, respectively.

Extreme Human Health Tolls Identified in 14 Districts of Cambodia and Vietnam

A total of 14 districts, 11 in Cambodia and 3 in Vietnam, are at very high risk (defined as when the PAF value is >0.1) from exposure to arsenic in drinking water, which means 10% of the total deaths are attributable to arsenic alone (Table 4.4 and Fig. 4.1). The average PAF values of all districts evaluated here in Cambodia and Vietnam are 0.037 and 0.013 respectively. However, due to the distinct spatial distribution pattern with high percentage As wells located within 20 km and especially within 5 km of the main courses of the Mekong and Bassac Rivers, districts adjacent to the Mekong show extreme PAF values greater than 0.15, and they are Kaoh Soutin, Kien Svay, Kaoh Thum and Srei Santhor in Cambodia (n = 4) and An Phu and Thanh Binh in Vietnam (n = 2, Table 4.4). Since the PAF represents the percentage of total deaths that are “preventable” or can be eliminated under the normal condition without arsenic exposure, this is a serious health crisis for these 14 districts in the MRD where about 1.5 million people reside: more than one in every ten people die due to exposure to drinking water As alone. In these 14 districts of the MRD, the annual excess death from high arsenic in water is 575 in 11 districts of Cambodia and 387 in 3 districts of Vietnam, representing about 17.2% (or 1 in every 6 adult deaths) and 13.4% (or 1 in every 8 adult deaths) of the total adult deaths in the most severely affected districts along the Mekong-Bassac Rivers. The annual excess death from high arsenic is 1204 in Cambodia and 1486 in Vietnam, representing about 3.66% (or 1 in every 27 adult deaths) and 1.27% (or 1 in every 78 adult deaths) of the total deaths in the entire MRD regions of two countries assessed here.

Although the average PAF (0.037) of Cambodia is higher than that (0.013) of Vietnam, the number of excess deaths estimated in Vietnam is actually more because the population is higher. Overall, Vietnam has a higher population density than Cambodia (Fig. 4.2b), with 70.7% of the districts having more than 10,000 people while in Cambodia the percentage is only 38.8%, which makes the distribution of excess death in Fig. 4.2d different from the PAF distribution (Fig. 4.2c).

4.2.4 Strength and Weakness of the Water As Health Effect Assessment

A strength of our health assessment is the large sample size from various sources. The districts of Cambodia and southern Vietnam have on average, 453 and 400 groundwater arsenic records respectively, making the estimation have a better chance of capturing the spatially heterogenous As distribution pattern. Although a health effect assessment of individual disease outcomes, such as various internal cancers, is desirable, using PAF to assess ED is robust. Without a death registry that carefully classifies the cause of death, it remains difficult to make the linkage between As exposure and death to verify detailed disease outcomes predicted based on dose–response established elsewhere (Hong et al., 2014).

One potential source of bias in our analysis is the use of the proportion of wells exceeding the drinking water quality standard as a proxy for individual exposure. In Cambodia, water resources are unevenly distributed in time and space. People are faced with water shortages in most of the rural areas during the dry season while there is plenty of water that floods wetlands, lowland areas and human habitats in the rainy season. This implies that people may not use well water consistently over time. Another bias comes from the number of people relying on wells for their daily drinking water. According to the National Institute of Statistics of the Ministry of Planning, only 23.7% of the total population in Cambodia had access to safe drinking water, with 30.4% of people using surface water or dug wells. In 2020, the percentage of people drinking surface water has decreased to 9.2%, with 53.2% of people drinking non-piped water from tube/dug wells etc.

Unfortunately, drinking water sources in the lower MRD do not exist to provide a clear linkage to As. Table 4.5 shows the proportions of household drinking water sources in 16 provinces of Cambodia and 14 provinces of Vietnam based on the national census reports of the respective country in 2019. Assuming that groundwater likely contains arsenic but surface water likely does not, tube wells, boreholes, and protected or unprotected wells are classified as drinking water sources at risk. Water sources with negligible risk of arsenic include rainwater collection and surface water (river, stream, dam, lake). Piped into dwellings, tanker trucks, public taps, etc., are considered uncertain because they can be sourced from either groundwater or surface water. Overall, 47.1% of rural families in Cambodia still drink well water, inferring a severe health hazard. In Vietnam, the categories of source water are slightly different so are similarly “assigned” a likelihood of As risk. It is worth noting that 56.9% of Vietnamese households have access to tap water. However, a survey of water consumption patterns in Can Tho City and An Giang Province found that only <20% of households used piped water for drinking (Chau et al., 2015). There may also be a rural and urban disparity. A survey of 542 rural households from Can Tho, Hau Giang, and Soc Trang provinces in Vietnam has found that 27% of the survey respondents with access to a piped-water supply did not use it, and 30% of them used this water for washing and cleaning (Wilbers et al., 2014). More importantly, of 41 piped-water supply stations investigated, 24 stations are sourced from groundwater while 17 used surface water. With this uncertainty in mind, Table 4.5 includes a “possible risk” category which includes all types of wells and piped-water supply. Although one might argue that only about half of the population is at risk of drinking high arsenic from groundwater, the spatial patterns of health burden remain robust although the excess deaths for each district should be adjusted downward if the population relying on well water is known.

Table 4.5 Household drinking water sources in 16 provinces of Cambodia and 14 provinces of Vietnam according to 2019 national census

4.3 Sediment Depositional Environment of the MRD During the Holocene

To set the stage for rice arsenic health risk assessment, this section examines existing literature on the sediment depositional environment of the MRD from the early Holocene to provide the geologic context for the formation of the soil and aquifer. This synthesis is relevant to Sects. 4.4 and 4.5 which describe the arsenic cycling in the MRD soils, the linkage between paddy soil and rice arsenic, as well as the soil and rice contamination risks associated with irrigation using arsenic-enriched groundwater. It also provides the framework to understand the findings emerged from hydrogeochemical studies in the MRD that sought to explain the occurrence of elevated groundwater arsenic in the low-lying areas, summarized briefly as follows.

Buschmann et al. (2007) have investigated mechanisms for geogenic arsenic mobilization triggered by anoxic conditions of the MRD aquifer, observing that elevated groundwater As levels are only present in the flat land embraced by the Mekong and Bassac Rivers. A possible explanation is that this in-between-river area was incised by the rivers during the Pleistocene glaciation, and only filled with alluvial deposits during the Holocene. Nevertheless, it is worth noting that natural organic matter is known to accumulate in wetland sediment, formed between the rivers common in Kandal. Therefore, the low-lying topography of this area today seems to depict the boundary of organic-rich Holocene sediments deposited between the rivers. This notion is supported by sediment cores obtained by Quicksall et al. (2008) that identified rapid channel deposition that simultaneously buries both organic matter and the terminal electron acceptors such as Fe and Mn oxides, which drives the in situ dissimilatory reductions of (oxy)hydroxides, thereby releasing sorbed species such as As. “Inefficient flushing” of both As and OM in flat and low-lying areas has been suggested to account for As enrichment in Holocene deltaic aquifers (Smedley & Kinniburgh, 2002; van Geen et al., 2008; Zheng et al., 2004). Perhaps most similar to the MRD is the shallow aquifer of the Hetao Plain of Northern China that consisted of OM-rich, Fe & Mn-oxides-rich and As-rich sediments whereby low hydraulic gradient has been demonstrated to account for the spatial pattern of groundwater As distribution (Zhang et al., 2013).

Findings based on three sediment cores of the MRD are described and compared to shed light on the sedimentation history of MRD formation in the Holocene. Sediment core DT1, penetrating to −51.5 m in the lower MRD of Vietnam (Fig. 4.6), is divided into 5 types of sediments based on mineralogy and lithology excluding the shallowest top soil (Nguyen et al., 2010). At a −45.87 m, the “oldest” silty sediment has a plant material radiocarbon age of 11,643–11,221 years before present (yrs BP) (Ta et al., 2005), suggesting that the sediment core provides constraints for the sedimentation history of MRD in the Holocene (Fig. 4.6). The early Holocene is characterized by coastal marine deposits, with brownish gray silty clay and sandy silt interbedded with very fine sand between altitudes of −36 and −51.5 m representing marsh/tidal flat sediment facies, with a sediment sample from −34.07 m showing an apparent radiocarbon age of 10,725–10,370 yrs BP (Ta et al., 2005). It is worth noting pebbles are identified from −30 to −40 m, suggesting sediment deposition by fast flowing rivers able to transport large particles. The next section up (−24 to −35 m) reflects a sub-tidal to intertidal flat depositional environment, consisting of dark gray laminated silty clay and sandy silt, with an apparent radiocarbon age of 9008–8697 yrs BP at the top of this section (−25.41 m) (Nguyen et al., 2010). Peat, rich in organic matter and indicative of a highly reducing sediment depositional environment, is common, and is found at several intervals between −23 and −34 m. With a plant material organic carbon age of 8371–8183 yrs BP (Nguyen et al., 2010), likely reflecting the onset of a mid-Holocene climate optimum with a high global sea level stand (Li et al., 2012; Stattegger et al., 2013), the lower MRD likely had reached the maximum extent of “flooding” and was inundated by the sea for most of the mid-Holocene. With the lowering of sea level in the late Holocene, the sediment exhibits a clear increase in coarsening from silt with interbedded very fine sand at −20 m reflecting pro-delta to very fine sand (−24 to −10 m) reflecting delta front at −13.8 m (Nguyen et al., 2010). By this time in the late Holocene, the delta front had presumably migrated further towards the sea, and the sedimentation environment gradually shifted to intertidal flats and finally, the present-day flood plain.

Fig. 4.6

Sediment lithology and facies of DT1 (105°38′51″ E, 10°17′02″ N, depth 51.5 m, elevation +2 m) located in the lower-MRD Vietnam (Fig. 4.6a), with radiocarbon ages of plant materials marked. The redox transition interface at 7.5–11, 13–16 m depth in QTC2 (10°54.329′ N and 105°4.746′ E; elevation: 3 m, depth 20 m) and QTC3 (same location, depth 33 m) in the mid-MRD of Vietnam is interpreted to correspond to the retreating sea level and progradation of the delta front in DT1 starting at ~8.5 kyr BP. The ages for QTC2 and QTC3 consisted entirely of clay and were based on OSL (optically stimulated luminescence) dating of quartz and feldspar in several silicate-rich layers. This figure was adapted from Wang et al. (2018a, 2018b) and Nguyen et al. (2010)

This sedimentary history is corroborated by two sediment cores (Wang et al., 2018a, 2018b), QTC2 (10°54.329′ N and 105°4.746′ E; elevation: 3 m, depth 20 m) and QTC3 (same location, depth 33 m) in the mid-MRD of Vietnam (Fig. 4.6). QTC3 sediment from a depth of 33 m dated by optically stimulated luminescence (OSL) indicates an age of 19,060–14,680 yrs BP. The most remarkable feature in both sediment cores is the shift from a reducing environment indicated by sediments rich in organic carbon, pyrite and siderite, to an oxidizing environment indicated by Fe(III) oxides at a depth of 7 m in QTC3 (Fig. 4.6). We interpret this as reflecting the late Holocene’s lowering sea level that begins to expose the initially reducing sediment to atmospheric oxygen. Given that these two cores are located further upgradient in the MRD than DT1 (Fig. 4.7). this transition may have occurred earlier than what is seen in DT1 Peat (TOC ~ 33.9% w/w) was found only at 16 m in QTC3, providing more corroborative evidence to interpret both sets of sediment records. It is worth noting sediment from the oxidized section at 2 m has an OSL age of 2920–2500 yrs BP. This confirms that the MRD sedimentation is recent.

Fig. 4.7

Soil type (a), and arsenic in groundwater (b), soil (c), and rice (d) in the Mekong River Delta. a Soil type data are from the Open Development Vietnam and Open Development Cambodia, with 7 types characterized according to the UN FAO Digital Soil Map of the World V3.6 (www.fao.org/geonetworks). The locations of three sediment cores, QTC2&QTC3 (Wang et al., 2018a, 2018b), and DT1 (Nguyen et al., 2010) are superimposed on the elevation and soil type map. The MRD area is outlined in red; b Concentration of individual well water arsenic based on 41,928 tests from Cambodia and 52,858 tests from Vietnam as shown in Fig. 4.2a; c Soil total As concentration (Table 4.7) are from main rice-growing area collected by Seyfferth et al., 2014 (n = 30, in areas marked by the yellow square) and from a village in Kandal Province with high As in groundwater (n = 15, in areas marked by the red square) (Hamzah et al., 2013), plus our unpublished rice paddy soils (n = 75, northern Phnom Penh, blue square); d Arsenic in rice grain data (Table 4.8) are from (n = 30) (Seyfferth et al., 2014) and our unpublished data (n = 95) from the same locations as paired samples of the paddy soil

Avulsion of the Mekong River also influences local sediment depositional environment that creates more prolific aquifers locally, especially in the triangular area between the Mekong and the Bassac Rivers. The first major “tributary” of the Mekong River is the Bassac River, which bifurcates away from the Mekong at Phnom Penh (Fig. 4.1). The area in the upper MRD between these two channels that extends south towards Vietnam marks a triangular region known as the Kandal Province, Cambodia. Kandal, in Khmer language, means between the rivers. Two sediment cores, a channel core (~50 m depth) and a control core (~55 m depth) collected from the Kandal Province along the Bassac River (latitude and longitude information not available) have been interpreted to represent a distinctive sedimentological feature in the upper MRD associated with the avulsion (Quicksall et al., 2008). The “channel” core was drilled in the eastern-most sand bar scroll on the paleo-levee of the avulsed channel, with its aquifer consisting of hydraulically conductive sand from ~15 to ~40 m depth. In the MRD of Cambodia (11°31′3.90″ N 105°0′41.77″ E), the horizontal hydraulic conductivity (Kh) of the aquifers’ sand was estimated to be 3.15 × 10^–4 ± 3.02 × 10^–4 m/s based on grain size analysis (n = 14) (Benner et al., 2008). The Kh values of surface clay in MRD of Cambodia were determined by permeameters and slug tests, and were lower, at 4.08 × 10^–7 ± 5.94 × 10^–7 and 1.22 × 10^–6 ± 1.89 × 10^–6 m/s, respectively (Benner et al., 2008). In the MRD of Vietnam, Kh values for the aquifers ranged from 0.93 × 10^–4 to 2.64 × 10^–4 m/s (Minderhoud et al., 2017), based on 999 pumping tests carried out by the Division of Water Resources Planning and Investigation in South Vietnam (DWRPIS, 2010). Based on values on specific aquifer designations by DWRPIS, Kh is assigned to 2.31 × 10^–4 m/s in a model that suggests subsidence due to over draughting of groundwater has caused As release at a deeper depth of MRD of Vietnam (Erban et al., 2013). This intriguing idea has not been followed up by subsequent investigations.

4.4 Arsenic in Soils of the Mekong River Delta

4.4.1 Physiochemical Properties of Soils in the MRD

Over thousands of years, the Mekong River and its tributaries supplied sediments that contributed to soil formation, with an estimated 160 million tons of sediment entering the South China Sea each year (Nguyen et al., 2010). Soils in the MRD are highly variable (Fig. 4.6). Eutric Fluvisol is found to be the dominant soil type occupying ~30% of the MRD, mostly along the banks of the Mekong and Bassac Rivers (Fig. 4.7a). In addition, Eutric Gleysols and Thionic Fluvisols are also common in the lower MRD (for definition of soil types see www.fao.org/geonetworks). Young alluvial, as well acid-sulfate and saline soils reflecting marine influence dominate the deltaic plain (Nguyen et al., 2010). Cambodian soils are primarily clayey loam with 41% of clay plus some gravel (7.9%) based on 103 surface soils sampled across the entire country (Saeki et al., 1959). Along the Mekong River, alluvial soils (42% clay, 4.2% gravel, n = 47) are common in inland basins and are used as paddy rice fields (Saeki et al., 1959).

Blair and Blair (2010) pointed out that 86% of Cambodian soils (n = 3000) contain low organic carbon (0.06–1%), while 63% contain very low total nitrogen (<0.05%) and 88% have low Olsen P (a method that extracts phosphate (PO₄-P) from soils using sodium bicarbonate adjusted to pH 8.5 to assess bioavailability), although the locations of the soil samples in this study may include other areas of Cambodia that are not part of the MRD. Unlike fertile soils in many other subtropical deltas (Wang et al., 2018a, 2018b), soils in the MRD can be poor in nutrients. A survey (Table 4.6) collected 63 soils across Vietnam, nine of which were from the lower MRD, and found that their top soils’ (0–20 cm) organic carbon averaged 1.14%, with total nitrogen (TN) averaging 0.11% and P₂O₅ averaging 0.09% (Tran, 2015). Considering the low fertility in the soil and the high demand for crop production, fertilizers, especially mineral fertilizers, were extensively applied to soils in the MRD of Vietnam (UNEP, 2005). There, soils with high iron sulfide (pyrite) are also common. With the oxidation of pyrite by atmospheric oxygen and subsequent release of sulphuric acid, the soil pH has been found to drop below 3 in parts of the lower MRD, forming acid-sulfate soils (Husson et al., 2000).

Table 4.6 Physiochemical properties of soil and sediment in MRD of Vietnam and Pearl River Delta

4.4.2 Soil and Sediment As of the MRD

Arsenic in soils and sediment not only resides in aluminum–silicate minerals but is also associated through adsorption onto amorphous minerals such as Fe–Mn oxides and to a lesser extent, clays (Smedley & Kinniburgh, 2002). Arsenic concentrations in various rocks including sandstone and limestone usually range from <1 to 15 mg/kg, but argillaceous sedimentary rocks such as shales, mudstone and slates can contain much higher As levels of up to 900 mg/kg (O’Neill, 1990). More than 200 minerals are known to contain various amounts of arsenic, among which, around 60% are in oxyanion form (arsenate) and around 20% are affiliated with sulphur as sulfides and sulfosalts due to the affinity of arsenic to bind with sulphur ligands (Onishi, 1969). Ore geologists have long appreciated the significance of arsenopyrite (FeAsS). It is the most common arsenic mineral found with many sulfide mineral deposits (Boyle & Jonasson, 1973). Because weathering of rocks supplies material to the soil, soil arsenic levels to a large extent reflect the source rock arsenic content, and tend to be elevated in soils adjacent to mining areas, where soil arsenic can reach 4,424 mg/kg (Nriagu et al., 2007). In two sediment cores from the MRD of Vietnam (Fig. 4.6), the authors found that four forms of As existed in the sediment, including sulphur-bound As(III) coupled with natural organic matter (NOM), arsenic in pyrite, oxygen-bound (As(III)) and As(V) (Wang et al., 2018a, 2018b). Arsenian pyrite particles in sediments from 8–16 m have been identified by scanning electron microscopy, with an estimated average As concentration of ~280 mg/kg in pyrite grains (Wang et al., 2018a, 2018b).

Zhang et al. (2017) have demonstrated that each year approximately a quarter of sedimentary As accumulated in the MRD is sourced from upstream geothermal activities that endowed the suspended particulate matter in the Lantsang-Mekong River with elevated As levels. The Mekong River originates from the Tibetan Plateau where it is known as the Lancang River. In Tibet, As enrichment in soils and geothermal water resulting from tectonic activities have been reported (Guillot & Charlet, 2007; Guo et al., 2019; Li et al., 2013; Nordstrom, 2002). Zhang et al. (2017) employed X-ray Absorption Spectroscopy (XAS) to investigate the As and Fe speciation of hot spring deposits and sediments in the Lantcang River in Yunnan Province of China. The study shows that much of the river sediment As is sorbed, with local hot spring deposits containing highly elevated As level of >100 mg/kg. The mobility of As in river sediment is found to be low due to strong arsenate binding to ferric oxides (ferrihydrite, goethite and hematite) and to a lesser extent manganese oxides and clay minerals. The flux of As transported by the Lantsang river sediment (F_As) was estimated to be 79.6 ton/year by multiplying the flux of suspended material (FSM, 1850 × 10⁴ ton/year at ChangDu monitoring station) by the average contents of extracted As in sediments (Ex_As, 4.30 ± 1.95 mg/kg). Using the results of radiocarbon dating for sediments in the MRD (Ta et al., 2002) and the average of ~9 mg As/kg in the sediment of the MRD (Polizzotto et al., 2008), the rate of As accumulation in the MRD sediments is estimated to be around 315 ton/year. Thus, about 25% of As in sediments of the MRD each year are coming from upstream geothermal activities. Polizzotto et al. (2008) investigated As levels in the sediment located in the upper reaches of the MRD (Kien Svay District, Kandal Province, Cambodia), reporting ~12 mg/kg As in the youngest sediments near the water table.

A report (Gustafsson & Tin, 1994) measured total As levels by HNO₃/HClO₄/H₂SO₄ digestion of acid sulfate soils (pH ~ 4) taken from four soil cores in the Plain of Reeds, lower MRD of Vietnam. The total As levels ranged from 6–41 mg/kg with a mean of 11 ± 2 mg/kg in the top 10 cm soils (Keeney & Nelson, 1982). A later study reported As levels in acid sulfate soils as 1.0 mg/kg after extracting by 0.43 M HNO₃ in sulphidic areas, compared with 0.62 mg/kg in non-acid sulfate soils (Hoaa & Cuong, 2009). These lower values may reflect incomplete dissolution using 0.43 M HNO₃.

In addition to the aforementioned geogenic sources for As in soils and sediment, mining, industry, agriculture and sewage have been known to cause anthropogenic As pollution of soil (Woolson, 1983). It should be noted that their impact tends to be localized in spatial extent. Soil arsenic contamination due to mine tailings, smelting of non-ferrous metals, and burning of As-rich coals have been well documented at numerous sites around the world (Han et al., 2003). The global annual amount of arsenical pesticides applied in orchards was estimated to be 7–11 × 10³ ton-As in 1983 (Woolson, 1983). Until the banning of arsenical pesticides in 2004 by the US Environmental Protection Agency (EPA), they elevated As level in soils to as high as 2,500 mg/kg (Bencko & Foong, 2017). An arsenical, 3-nitro-4-hydroxyphenylarsonic acid (Roxarsone), along with other phenylarsonic compounds, were widely used (Chapman & Johnson, 2002) as feed additives for animals around the world since the 1930s and 1940s (Hanson et al., 1955; Morehouse & Mayfield, 1946), leading to elevated As levels in chicken and chicken poops (Nachman et al., 2013) which contain a proportion of carcinogen, inorganic arsenic (iAs) after biotransformation in soil (Yang et al., 2016). Therefore, it further contaminated the soil, where poultry litter was used as fertilizers to grow crops (Ashjaei et al., 2011) and vegetables (Yao et al., 2009). In February 2014, the US Food And Drug Administration (FDA) formally withdrew the approval of Roxarsone after detecting high levels of iAs in the livers of chickens fed with Roxarsone additives, followed by a ban of any arsenical additives for animals on April 1st, 2015 (FDA, 2015). China also banned the use of any phenylarsonic feed additives on May 1st, 2019 (Hu et al., 2019).

Anthropogenically sourced As in soils of the MRD is minor and highly localized compared to natural sources. A recent study (Olson & Cihacek, 2020) pointed out that arsenical pesticides, especially Agent Blue (cacodylic acid, C₂H₂AsO₂), were extensively used during the American Vietnam War (1965–1972) to destroy rice growth. It is estimated that at least 1 million kg As in the form of Agent Blue was added aerially to the MRD and Central Highlands of South Vietnam to destroy mangrove forests and rice paddies. To the best of our knowledge, no measurements of Agent Blue in the soils of South Vietnam are available, so the impact remains not possible to assess.

The most relevant to rice As and subsequent health risk assessment is paddy soil As, with observations suggesting that irrigation with groundwater rich in As has increased paddy soil As concentrations (Fig. 4.8). Several studies have investigated paddy soil As in the MRD (Table 4.7). In the MRD of Cambodia, Hamzah et al. (2013) measured soil samples from 15 locations in Phumi Khleang, Kandal Province. The concentration of total As in soil ranged from 5.3 to 27.8 mg/kg, with a mean of 9.9 ± 5.4 mg/kg. A subsequent study investigated 23 matching paddy rice soil (0–20 cm) and rice samples from five major rice-growing areas of Cambodia including Battambang, Banteay Meanchey, Kampong Thom and Siem Reap Provinces, and Kandal Provinces (Seyfferth et al., 2014). The concentration of tAs in soil ranges from 0.8–18 mg/kg with a mean of 7.8 ± 4.6 mg/kg As. In addition, soils from Banteay Meanchey and Battambang provinces from the northwestern area around Tonlé Sap have lower As concentrations than those from the Kandal and Prey Veng Provinces close to the Mekong River. Due to the paucity of paired soil and rice As analysis, we included our unpublished data of bulk As concentrations in 75 paddy rice soil samples collected from the upper Mekong River Delta in Cambodia determined by X-ray Fluorescence Spectrometer (XRF) with a detection limit of 2.4 mg/kg (DiScenza et al., 2014) (Table 4.7). They are collected mostly around Phnom Penh (Fig. 4.7c), ranging from northwestern Kandal to Kampong Cham with an average of 3.7 ± 3.5 mg/kg total As in soil. The maximum value of 18 mg/kg is detected for a soil sample from northwestern Kandal irrigated by groundwater containing 317 μg/L As (PP119-41-Soil, 11°43′46.4″ N 104°54′40.0″ E). Soils from the west side of Khan Preaek Pnov Lake in Phnom Penh contain lower As (1.9 ± 1.4 mg/kg As, n = 43), while soils from Kandal contain higher As (10.3 ± 2.8 mg/kg, n = 12).

Fig. 4.8

Cumulative frequency distribution of soil total As concentrations in the MRD (Table 4.5). The samples are 30 paddy rice soils from the main rice production areas of Cambodia (Seyfferth et al., 2014), 15 soil samples from groundwater-As affected Phumi Khleang village, Kandal Province of Cambodia (Hamzah et al., 2013), 75 paddy rice soil samples from north Phnom Penh, Cambodia collected by us, with locations marked in Fig. 4.7c. An additional 16 agricultural soils from (Huang et al., 2016) in Thanh Binh district of Dong Thap Province of Vietnam are included here, although the exact locations are not available. The samples in (Seyfferth et al., 2014) were from Banteay Meanchey (n = 8); Battambang (n = 10); Kandal (n = 5); Kampong Thom (n = 1); Prey Veng (n = 5); Siem Reap (n = 1). The Maximum Contaminant Level (MCL) for soil total As in Vietnam is shown as the thick black line

Table 4.7 Concentration of soil total arsenic from the Mekong River Delta of Cambodia and Vietnam

A cumulative frequency plot of our soil data together with the literature data (n = 117) (Hamzah et al., 2013; Huang et al., 2016; Seyfferth et al., 2014) shows that 7.7% of all soil samples exhibit above 12 mg/kg of As, the Vietnamese standard for soil maximum contaminant level (Fig. 4.8). Among them, our unpublished Cambodian soil data displayed the lowest levels of As (3.7 ± 3.5 mg/kg, n = 75) and none of them exceeded 12 mg/kg except one paddy rice soil sample from Kandal irrigated with high-As groundwater. That high-As groundwater irrigation leads to paddy soil As enrichment is consistent with that all of 20% of 30 soil As samples exceeding 12 mg/kg are from Kandal (n = 5) and Prey Veng (n = 1) Provinces, documented by Seyfferth et al. (2014). In a village of Kandal Province with known high-As groundwater occurrence, 13% of 15 paddy soil samples (9.9 ± 5.4 mg/kg) were found to exceed 12 mg/kg As (Hamzah et al., 2013), possibly reflecting local variabilities although groundwater used for irrigation was not analysed for As, making it impossible to examine the linkage. Fortunately, in the MRD of Vietnam, Huang et al. (2016) investigated 16 soils irrigated by groundwater in Thanh Binh district of Dong Thap Province, observing a positive correlation between As concentrations in groundwater (448 ± 257 μg/L) and matching soil samples, with 62.5% of samples with As levels exceeding 12 mg/kg (mean 13.4 ± 4.6 mg/kg). Finally, a recent study surveyed 80 matching paddy rice soils and rice samples in the lower MRD of Vietnam and determined the average As levels of soil as 12.6 ± 3.2 mg/kg, with the maximum reaching 28.9 mg As/kg (Nguyen et al., 2020), although the soil As data were not available for tabulation, with groundwater As status unknown.

We are aware that nations have set different soil As standards, although the science behind such standards has room for improvement, especially given the uncertainty of soil As bioavailability and uptake by rice as described below. For example, the Japanese standard for As in paddy soil is 15 mg/kg and while the As critical upper limit set by Chinese authorities for paddy soil is 20–30 mg/kg depending on soil pH (GB 15618-2018). In light of such uncertainties, the elevated As in rice paddy soils of the lower MRD in Vietnam and Cambodia deserves closer scrutiny. This is discussed below from the perspective of rice yield briefly, as well as bioavailability and toxicity of rice arsenic in the rest of this chapter.

A consideration for setting soil As standard is to safeguard rice yield. Association between soil As content and the incidence of straighthead diseases was observed for applications of monosodium methanearsonate (MSMA) (Horton et al., 1983), which has led to poorly developed panicles and a reduction in rice yield between 24 and 96% depending on rice varieties (Yan et al., 2005). A lab study found that when soil tAs was greater than 50 mg/kg, the extent of straighthead diseases was severe in rice, with a dose–response relationship of a yield reduction from 100% to only 16% when soil tAs increased from 50 to 90 mg/kg by addition of As(V) solutions to soils (Rahman et al., 2008). Rice yield was found to negatively correlate with the soil tAs concentration ranging from 10 to 70 mg/kg in paddy soils irrigated by groundwater with 130 μg/L of As in Bangladesh (Panaullah et al., 2009), with each 10 mg/kg soil tAs increase corresponding to about 2 ton/ha loss in yield. This highlights the need to move away from irrigation using high-As groundwater.

4.4.3 The Linkage Between Paddy Soil and Rice As Speciation

As speciation in paddy soils is of interest because the differences in chemical properties of various As species influence the bioavailability of As to plants, especially during the uptake of arsenic by plant roots. Unfortunately, hardly any study has assessed soil As speciation in the MRD. A study compared soil arsenic speciation between control sites and sites subject to solid wastes in Vietnam (Le et al., 2011). It revealed that most As in soil are inorganic. In aqueous solutions and at neutral pH, inorganic As (iAs) species arsenate (As(V)) is mostly ${\text{H}}_{{2}} {\text{AsO}}_{4}^{ - }$ with a pKa₁ of 2.19 under oxidizing conditions while iAs species arsenite (As(III)) is mostly H₃AsO₃ with a pKa₁ of 9.23 under reducing conditions (Cullen & Reimer, 1989). Das et al. (2016) reported that 87–94% of As was As(III) in flooded (reducing) soils while 73–96% of the total As was As(V) in aerobic soils, indicating that the speciation of As is regulated by redox reactions in soils which can be affected by irrigation practice. In most contaminated soils, iAs species are the dominant forms (Nriagu et al., 2007).

Among major crops, rice is of particular concern due to its high shoot assimilation rate for As compared to wheat and barley (Williams et al., 2007). In general, rice plants cultivated in soils with higher solid As concentrations are found to uptake more As due to higher dissolved pore water As concentrations in the rhizosphere (Suriyagoda et al., 2018). In addition to soil total As concentration (Bogdan & Schenk, 2009), a wide range of soil parameters, including texture, pH, iron (Fe) (hydr)oxides (Chen et al., 2005), plant-available phosphorous (P) (Cheng et al., 2004), plant-available silicon (Si) (Amaral et al., 2017), and sulphur (S) (Zhao et al., 2010) have been found to affect the uptake of arsenic by rice. Campbell and Nordstrom (2014) found that soil pH is significantly correlated with the concentrations of total As in rice, because high pH increases negative surface charges of soils, which in turn promotes the desorption of As(III) and As(V). Fu et al. (2011) demonstrated that soil organic matter negatively correlated with total As in rice grains, which may result from the formation of insoluble complexes between organic matter and As, making As less bioavailable to rice plants (Wang & Mulligan, 2006). On the other hand, because dissolved organic matter competes for adsorption sites on Fe(III) (hydr)oxides via ligand exchange with both As(III) and As(V), it may increase mobility and bioavailability of As(III) and As(V) (Wang & Mulligan, 2006). Due to the high affinity of As to soil Fe(III) (hydr)oxides, its role is noteworthy and it has been shown that this is a key process responsible for lowering As uptake by rice (Lauren & Duxbury, 2005). The formation of soil Fe(III) (hydr)oxides is aided by the rice plants. The radial oxygen transported from root tissues to the surrounding soil allows for a micro-oxygenated environment in the soil, with the formation of Fe plaque on the surfaces of the rice root, effectively “blocking” the As from entering the root (Hossain et al., 2009).

Once As overcomes the Fe plaque barrier ubiquitously present on rice roots, the uptake mechanisms of iAs have been a topic of intense interest over the past two decades, with phosphate-P and silicate-Si transporters found to be important. Being in the same Group in the Periodic Table, and thus with similar chemistry, As(V) uptake by P transporters in rice has been demonstrated. Therefore, there is a competition between As(V) and P for the uptake of both. Although P can mobilize adsorbed As(V) in minerals to increase available As to plants (Peryea, 1991; Sadiq, 1997), high plant-available P also decreases As(V) uptake through competition (Jiang et al., 2014; Meharg & Macnair, 1990) and suppressing the P transporters (Finnegan & Chen, 2012). High levels of silicic acid have been shown to reduce As(III) uptake in plants (Bogdan & Schenk, 2009; Desplanques et al., 2006; Guo et al., 2005, 2007; Ma & Takahashi, 2002; Seyfferth & Fendorf, 2012; Seyfferth et al., 2016b). Due to the similar structure of silicic acid and As(III), large quantities of Si suppress the expression of transporters, Lsi1 and Lsi2 (Ma et al., 2006), resulting in overall low As(III) uptake. Finally, S levels in soil affect the As uptake especially As(III), most likely due to the binding of thiol-group chemicals to As(III) thus reducing the translocation of it from root to grain (Zhao et al., 2010).

Soil microorganisms also play an important role in As uptake by rice, especially the uptake of methylated As; this is because to date no evidence exists to support methylation of As by rice plants (Lomax et al., 2012). Microorganisms in paddy soils are involved in the As biotransformation through different pathways (Zhang et al., 2015a, 2015b). Not only do they regulate oxidation and reduction between As(III) and As(V), but also they are involved in methylation and demethylation reactions between inorganic and organic forms of As (Qin et al., 2006; Yoshinaga et al., 2011). The gene responsible for microbial oxidation of As(III) to As(V) has been identified as aioA gene (Hamamura et al., 2009). Two genes, arsC and arrA, are both able to reduce As(V) to As(III) in paddy soils in different pathways (Malasarn et al., 2004). While the arsC and arrA regulated reduction usually occurs under flooded (anaerobic) conditions, phylogenetically diverse bacteria have been shown to reduce As(V) under non-flooded (aerobic) situations (Bachate et al., 2009). arsM is responsible for converting iAs to methylated As including MMA(III), MMA(V), DMA(III), DMA(V), TMAs(V), and TMA(III) (Challenger, 1945; Qin et al., 2006). Microbial As methylations were first found in fungi Scopulariopsis brevicaulis (Challenger & Higginbottom, 1935), and then other bacteria and methanoarchaea were reported with the ability to volatilize As by methylation (Wang et al., 2014). Zhao et al. (2013a, 2013b) used GeoChip to identify arsM sequence in six soils, observing a positive correlation between soil pH and arsM abundance.

4.5 Health Risk Assessment of Rice Arsenic Exposure in the Mekong River Delta

4.5.1 Arsenic in Foodstuff

Concerns have been raised on the toxicity of arsenic in food in the last decade, especially when rice is used as transition food for infants (Carignan et al., 2016). Because the crustal abundance of As, at 2.5 mg/kg, is not low (Rudnick & Gao, 2003), and that amorphous and clay minerals in soil can sorb large amounts of arsenic, soils are naturally endowed with arsenic at levels of ~5 mg/kg (ATSDR, 2007). In addition, the loading of As from irrigation by groundwater enriched in As further enhances soil As levels (Khan et al., 2010). Therefore, the risks for uptake and bioaccumulation of As by crops cannot be overlooked.

Crops including rice, wheat, corn, legumes, and potatoes have been investigated for total As concentrations, with all exhibiting a great deal of variability. In general, wheat, corn and legumes tend to contain lower levels of As compared to rice. Dozens of studies analysed wheat and determined bulk As concentration ranging from 10–500 μg/kg with a mean of <100 μg/kg (Adomako et al., 2011; Shi et al., 2015; Williams et al., 2007). In the following, all food As concentrations are reported in dry weight unless noted. Corn was reported to display As concentrations ranging from 10–170 μg/kg in Tanzania (Marwa et al., 2012). It is worth noting that a mean of As <100 μg/kg in legumes was detected in market food collected in Bangladesh, although nearly all of this As is iAs, thus concerns regarding its toxicity have been raised (Williams et al., 2006), with similar observations made later in Brazil (Ciminelli et al., 2017). Potato is the fourth largest produced crop (Leff et al., 2004). Potato tuber samples bought from the market displayed a maximum tAs of 890 μg/kg and also 100% iAs in Bangladesh (Williams et al., 2006). Signes-Pastor et al. (2008) later identified MMA in potato tubers from West Bengal with an average of only 80 μg tAs/kg.

Zhang et al. (2015a, 2015b) determined tAs levels in 48 kinds of edible or medicinal mushrooms in Southwestern China, revealing common earthball Scleroderma citrinum displaying the highest As (1,700 μg/kg) and Termitomyces eurrhius which is a mushroom symbiotic with termites, displaying the lowest As (170 μg/kg). Seyfferth et al. (2016a, 2016b) reported similar tAs levels ranging from 100–1,000 μg/kg in 12 species of mushroom samples (n = 40) collected from the main production areas of the US, with higher As in Cremini (Agaricus bisporus) than Shiitake (Lentinus edodes). Among 17 samples with tAs >400 μg/kg, iAs accounted for 25–94%, while maximum percentages of 28 and 20% were reported for DMA and AsB respectively.

Williams et al. (2006) collected 94 vegetables from markets around Bangladesh and found that arum stolon had the highest As value of 1,930 μg/kg. Both the mean (343 μg/kg) and the maximum (1,930 μg/kg) As levels for root and tuberous vegetables were higher than those for fruit vegetables (mean: 301 max: 1,590 μg/kg) but not for leafy vegetables (mean: 384 μg/kg and max: 790 μg/kg).

In the MRD, As in food has been investigated by two studies, with concerns raised for areas with high groundwater As and high soil As (Table 4.8). Wang et al. (2013) collected food samples from Kampong Cham, Kratie and Kandal Provinces in Cambodia along the Mekong River with low (1.3 ± 0.6 μg/L), medium (22 ± 44 μg/L), and high (846 ± 298 μg/L) levels of groundwater As (Phan et al., 2010), respectively. A food frequency questionnaire was used to acquire consumption patterns among residents surveyed in Kratie (n = 31), Kampong Cham (n = 58), and Kandal (n = 69) Provinces; it revealed a high proportion of rice intake (46.8%), followed by vegetables (23.8%), fruits (13.5%), and fish (10.2%), with a small percentage of meat (3.11%) and viscera (1.73%). A total of 154 food samples and 22 food products were collected and measured after removing inedible parts, freeze-drying, and grounding. Most food samples collected from Kandal Province contain higher As levels than those from Kratie and Kampong Cham Provinces. The highest As levels of all food samples were detected in fish from Kandal with a mean of 2,832 ± 1,606 μg/kg (n > 9) in wet weight (ww). The lowest As levels were detected in cattle stomachs from Kratie with 1.86 ± 1.10 μg/kg (n = 3) in ww. The rice samples from Kandal also contain significantly elevated As (247 ± 187 μg/kg, ww) than samples from Kampong Cham (29 ± 24 μg/kg, ww). Meat such as beef (96.7 ± 9.9 μg/kg, ww) and egg (64.2 ± 85.5 μg/kg, ww) tend to contain higher As levels than vegetables, fruit and viscera. However, a vegetable (Morning Glory) collected from Kandal showed a higher mean As of 277 ± 80 μg/kg, ww.

Table 4.8 Total arsenic concentrations (μg/kg) in food items from the Mekong River Delta of Cambodia

Phan et al. (2013) conducted a similar study in Kampong Cham, Kratie and Kandal Provinces of Cambodia (Table 4.8). Besides food samples, they also collected eight matching paddy soil and rice samples where groundwater was used for irrigation in Kandal and Kampong Cham. Positive correlation between soil bulk As and rice total As was observed with a Pearson correlation coefficient of 0.826 (p < 0.01). The total As in uncooked rice (256 ± 141 μg/kg) and soil (12.9 ± 10.4 mg/kg) from Kandal were significantly higher than those from Kampong Cham (24 ± 12 μg/kg in rice, 0.8 ± 0.1 mg/kg in soil), evaluated by an independent t-test (p < 0.05). The tAs in food, including uncooked rice, fish and vegetables from these three provinces were significantly different by One-Way ANOVA (Tukey HSD and Games-Howell, p < 0.05), with the highest in Kandal, followed by those from Kratie and the lowest in Kampong Cham.

One important implication of the Phan et al. (2013) study emerges when the As levels in groundwater used for irrigation are considered. In Kandal, rice paddy soil with on average >12 mg/kg As (Vietnamese soil MCL) producing on average >200 μg/kg tAs in uncooked rice (WHO standard for iAs in rice is 200 μg/kg) is probably irrigated by groundwater with on average 846 ± 298 μg/L of As (n = 46) (Phan et al., 2010), although the exact groundwater As concentration used for irrigation was not reported in Phan et al. (2013). Therefore, this one study would support setting a soil As standard for MRD soil at 12 mg/kg, although more studies would be necessary to enhance the science. It also underscores the urgency to shut down high-As wells used for irrigation in the region. It is reassuring that in Kampong Cham where irrigation relies on low-As groundwater (1.3 ± 0.6 μg/L, n = 18, (Phan et al., 2010)) and rainwater, neither paddy soil nor rice showed any evidence of As enrichment. According to the locations of rice grains, we divided them into two groups. One is collected from areas with high-As groundwater and the other is from areas with low-As groundwater. The tAs of grains (277 ± 127 μg/kg, n = 112) from areas likely irrigated by high-As groundwater was higher than those from areas using low-As water for irrigation (177 ± 84 μg/kg, n = 95) (p < 0.001, LSD, One-Way ANOVA). It should be noted that in areas with frequent high-As groundwater occurrence, there are still local spatial variabilities so not all irrigated groundwater has high As.

4.5.2 Arsenic in Rice Grains

According to a European foodstuff survey by the EFSA (European Food Safety Authority) (EFSA, 2009), rice grains (n = 1,122) contain much higher As levels (Mean: 136 μg/kg, Median: 110 μg/kg, 95th percentile: 360 μg/kg, Max: 1180 μg/kg) than other major crops (n = 2,215) (Mean: 14.7 μg/kg, Median: 0 μg/kg, 95th percentile: 60 μg/kg, Max: 5662 μg/kg). This survey lends support for that As in rice grains is a significant contributor to dietary As intake by humans in populations not exposed to iAs from drinking water.

The tAs concentrations in rice grains show great variations among different varieties, cultivation methods and locations. Meharg et al. (2009) analysed tAs of 901 rice grains collected from markets of ten countries. The mean of tAs values varied sevenfold among countries, with the highest mean tAs of 280 μg/kg detected in France (n = 33) followed closely by the mean tAs of 250 μg/kg detected in the US (n = 163). Rice grains from Spain also contained relatively high tAs (Mean: 200 μg/kg, n = 76). Surprisingly, rice grains from Asian countries including China, Thailand and Bangladesh contained relatively low tAs (Mean: 140, 140 and 130 μg/kg; n = 124, 54 and 144 respectively). The lowest and the second lowest mean tAs were 40 μg/kg in Egypt (n = 110) and 70 μg/kg in India (n = 133), respectively.

Zavala and Duxbury (2008) bought 204 rice samples in New York markets and found that brown rice (usually subject to less milling and retain the bran layer) contained higher tAs (196 ± 111 μg/kg) than usually more thoroughly milled white rice (127 ± 87 μg/kg) did. The concentration of iAs in rice grains varies with the extent of milling because rice husks usually contain higher As concentrations than grains due to the tendency of iAs to accumulate in the outer layers of rice grains (Sun et al., 2008).

We compiled rice grain data (all uncooked unless specifically noted) of the MRD in literature (n = 175, Table 4.9), and expanded this dataset to include 107 rice samples collected by us from northeastern Phnom Penh to western Kampong Cham with one sample from Kandal (Fig. 4.7d). The tAs levels in rice grains of the MRD range from 8 to 788 μg/kg in dry weight with a mean of 197 μg/kg. Two food surveys (Phan et al., 2013; Wang et al., 2013), described above, determined tAs in rice grains from Kampong Cham, Kratie and Kandal Provinces, finding similar levels (Table 4.9). Seyfferth et al. (2014) and Phan et al. (2014) reported tAs in rice grains of 150 ± 60 μg/kg (n = 6) and 201 ± 50 μg/kg (n = 11) in Prev Veng. Because high-As groundwater is likely used for cooking, two cooked rice samples there contained very high levels of tAs of 530 and 600 μg/kg (O’Neil et al., 2013). Seyfferth et al. (2014) surveyed 22 matching paddy rice soil and rice samples from major rice-growing areas of Cambodia (Table 4.7). They found that rice grains from Kampong Thom (260 ± 60 μg/kg), Banteay Meanchey (250 ± 50 μg/kg) and Battambang (195 ± 25 μg/kg) contained higher tAs than those from Prey Veng Province (150 ± 60 μg/kg), even though the soil As level in Prey Veng is higher than those of other provinces. We interpret this as support for the need to further investigate soil As speciation and microbiome that regulate rice As uptake and accumulation in rice grain.

Table 4.9 Total arsenic concentrations (μg/kg in dry weight) in rice grains from several provinces of Vietnam and Cambodia

Despite the complexity in soil-rice As linkage, if soil As is sourced from groundwater irrigation, that newly added As may be more bioavailable for rice. For example, Murphy et al. (2018) investigated 16 rice samples from Preak Russey, Kandal Province collected from rice paddies, finding a high average As of 315 ± 150 μg/kg in unpolished husked grain. There is a positive correlation between tAs in those rice grains and their irrigation water (R² = 0.5304), with As concentration in 16 irrigation water samples ranging from 0 to 1250 μg/L. A study reported average tAs contents in 39 polished rice as 224 μg/kg (132–471 μg/kg) in An Giang province, Vietnam, where groundwater As is high (230, 0.1–997 μg/L) (Hanh et al., 2011). A recent paired soil-rice As survey collected 78 rice samples in the lower MRD of Vietnam, revealing a mean tAs of 180 ± 90 μg/kg with a range from 80 to 560 μg/kg, corresponding to 12.6 ± 3.2 mg/kg of soil As (Nguyen et al., 2020). Like Phan et al. (2013) study discussed earlier, this last study provides additional support for setting soil MCL at 12 mg/kg for Vietnam on the basis of the precautionary principle.

Studies from South Asia also suggest that irrigation using high-As groundwater (>50 μg/L) should be avoided. Due to the anerobic environment of the rice paddy field that facilitates uptake, much attention has been paid to how rice uptakes and accumulates As (Zhao et al., 2010), especially when irrigation water is enriched in As (Rahman & Hasegawa, 2011). Williams et al. (2006) reported higher levels of As in rice grains purchased from regions with elevated As in groundwater of southwestern Bangladesh, with two rice samples from Faridpur containing tAs of 440 and 580 μg/kg irrigated by groundwater with 140 μg As/L. Further, Zavala and Duxbury (2008a) compared As levels in 871 samples of rice grain cultivated in high-As (>6 mg/kg) or low-As (<6 mg/kg) soils, or irrigated with high-As (>50 μg/L) or low-As (<50 μg/L) water in Bangladesh. They found higher mean levels of As in rice grains (242 ± 98 μg/kg) grown in low-As soils but subject to high-As irrigation water, than those (194 ± 74 μg/kg) cultivated also in low-As soils and also irrigated by low-As water. On the other hand, rice grains harvested from high-As soil and irrigated by low-As water exhibited only slight increases (mean: 200 μg/kg, no standard deviation was reported).

Alternative mitigation measures based on soil and crop science to lower the uptake of iAs by rice plants from soils have been explored. For example, by adding highly feasible silicon amendments to rice husks, a 25–50% reduction of iAs in grains has been achieved (Seyfferth et al., 2016b). Other farming practices such as growing rice without long periods of flooding could reduce iAs in grains but could also cause lower yields and elevated cadmium (Beans, 2021). Another approach is to develop genetically modified rice cultivars with substantially lower As uptake, yet identification of such genes remains a challenge after over a decade of work. Recently, the team of Fang-jie Zhao has identified a genetic mutation that led to a decrease of As in rice grain by one third, albeit indirectly (Sun et al., 2021).

4.5.3 Arsenic Speciation in Rice Grains

Assessment of human health risks due to As exposure from rice must consider the As speciation in rice grains. Rice grains are also known to contain significant amounts of methylated oxyarsenates, mainly as dimethylated arsenate (DMA(V)), which is known to be far less toxic than iAs (Ng, 2005). Most rice grains contain over 50% of As as iAs except for those produced in the USA (Rahman & Hasegawa, 2011; Williams et al., 2005). Recently, a study detected a highly toxic thiolated As, demethylated monothioarsenate (DMMTA) in rice grains from 15 countries, with DMMTA up to 21% of tAs (Dai et al., 2022). In the following section, chemical extraction based As speciation assay of rice grains is summarized to provide the background necessary to understand the uncertainties remaining in health risk assessments of rice arsenic exposure described in the next Sect. 4.5.4.

Chemical extraction of rice followed by separation using chromatography and As detection (Kubachka et al., 2012) has identified iAs and DMA(V) as the dominant As speciation in 95 market rice samples collected from seven countries (Meharg & Zhao, 2012). The aforementioned EFSA report conducted 706 speciation measurements on European rice grains (EFSA, 2014), reporting mean iAs levels as 101 μg/kg and 95th percentile iAs as 197 μg/kg, with the average of brown rice (152 μg/kg) higher than that of white rice (89 μg/kg). This is consistent with an earlier study (Williams et al., 2005) that analysed As speciation of 51 market rice samples from North America (Canada, US), Europe (Italy, Spain), Asia (Taiwan, Thailand, India), and from Bangladesh, with the US rice grains showing lower %iAs (42 ± 5%, n = 12) than Asian rice grains including Bangladeshi (80 ± 3%, n = 11) and Indian (81 ± 4%, n = 15) grains. Rice grains in the US tend to have more DMA(V), averaging 49 ± 17%. A subsequent compilation of all the acquired speciation data from seven countries (n = 95) showed that the average %iAs in rice is around 54% of the total As (Meharg & Zhao, 2012), with 21.5% as DMA(V).

Depending on location of cultivation, the same rice cultivar can have a wide range of As concentrations as well as percentages of iAs and DMA(V) with reasons not fully understood at present (Williams et al., 2005; Zavala et al., 2008). Not enough is known regarding the mechanisms of uptake and metabolism of arsenic by rice plants to conclusively predict which rice cultivar will have more toxic iAs or less toxic DMA(V) in the grains, even when they are subject to the same soil and growth conditions. In pot experiments to investigate the influence of rice genotypes and soil As on As speciation of rice grains, it has been shown that under the same conditions of cultivation, %DMA of rice grain varies from 7 to 56% (Williams et al., 2005). In this experiment, soil As was amended with monosodium arsenate to result in an increase of soil tAs concentration from 31.3 to 100 mg/kg (soil As speciation was not assessed in this pot experiment). The soil As and rice As speciation linkage is complex (Zhao et al., 2010, 2013a, 2013b) and is a subject of intense ongoing research. For example, a recent study (Dai et al., 2021) detected methylated thioarsenate species in soil porewaters and rice, identifying 0.4–10.1 μg/kg highly toxic dimethylated monothioarsenate (DMMTA) in rice grains, of which bioavailability is still unknown.

Concentration and speciation of As are heterogenous in a single rice grain at microscopic scale. In addition to the well-established rice husk being enriched in tAs (Rahman et al., 2007), iAs levels in rice bran (~1000 μg/kg) can be much higher than that in rice grain by 10–20 fold in whole grain rice samples collected from China and Bangladesh (Sun et al., 2008). Although in situ speciation analysis relying on non-destructive synchrotron techniques are still few, grain arsenic probed by micro-X-ray absorption near edge structure (μXANES) reveals that most iAs is As(III) and is located in ovular vascular trace, while the DMA(V) distributes more evenly from external grain parts to endosperm (Carey et al., 2010).

Murphy et al. (2018) collected well water (n = 65) and matching soil (n = 70) and rice samples (n = 105) in Preak Russey, Kandal Province, following a chemical extraction to detect %iAs and %DMA(V) in rice grains with an average value of 80 ± 13 and 21 ± 13% respectively (n = 55). A follow-up study (Murphy et al., 2020) collected 10 rice samples irrigated by well water from a control area with lower As (103 ± 175 μg/L) in groundwater to compare with rice collected in Preak Russey (groundwater As 953 ± 349 μg/L). It is not surprising that Preak Russey rice grains irrigated by high-As groundwater contain higher As(V) (174 ± 45 μg/kg, n = 57) and DMA(V) (124 ± 111 μg/kg, n = 57) concentrations than those irrigated by low-As groundwater, with lower rice As(V) of 111 ± 29 μg/kg and DMA(V) of 40 ± 31 μg/kg. Combining the 112 As speciation data points from these two studies in Kandal, MRD of Cambodia (Murphy et al., 2018, 2020), a negative correlation between %iAs and tAs and a positive correlation between %DMA(V) and tAs emerged from this dataset, implying that rice is protecting themselves by translocating more methylated As into grain when As uptake is higher (Fig. 4.9). The mean iAs concentration in rice grains reported by the two Murphy papers is 200 ± 61 μg/kg (n = 112), with 50% of them higher than 200 μg/kg and 93% higher than 100 μg/kg, suggesting a need to consider the still uncertain health risks from rice iAs exposure for residents in the MRD as discussed in the following section.

Fig. 4.9

Arsenic speciation in rice grains (n = 112) of Cambodia. Concentrations of iAs (a) and DMA(V) (b), and the percentages of iAs (c) and DMA(V) (d) are plotted versus the tAs. Data shown in squares and triangles are from Murphy et al. (2018) and (2020), respectively, and are from Preak Russey, Kandal Province of Cambodia. Rice grains also contain organoarsenicals DMA(V) and MMA(V), with DMA(V) dominant and accounting for >90% of methylated As. The positive correlation between %iAs and the tAs (R² = 0.7906) and the negative correlation between %methylated As and tAs (R² = 0.7755) are both significant (Pearson, p < 0.0001). The Maximum Contaminant Level (MCL) for iAs in rice grain of 200 mg/kg recommended by the WHO, FAO, and the EU is shown as a thick line in a (FAO/WHO, 2014)

4.5.4 Uncertainties in Health Risk Assessment of iAs Exposure from Rice

Much uncertainty remains in the health risk assessment of rice iAs exposure. Accurate exposure assessment for a population is challenging due to highly heterogenous iAs levels in rice grains and a diverse range of dietary intake of rice. Further, both the bioavailability and the toxicity of iAs in rice, once ingested by humans, have seen very few in vivo investigations. These uncertainties are described first before an assessment of health risks of iAs exposure from rice in the lower MRD is made based on existing iAs data in rice and established food consumption patterns in Sect. 4.5.5.

Uncertainties in As Speciation Assumption and Variabilities in Rice Intake

Measurements of iAs levels in rice grains are still too few at present despite the heterogeneity of iAs levels observed in rice (see Sect. 4.5.3). For this reason, most exposure assessment until now simply assumes 100% iAs levels in rice (EFSA, 2014), resulting in an overestimation of rice iAs exposure, especially for rice with >200 μg/kg tAs (Table 4.10). Although rice As speciation data have only been reported in two studies in Cambodia, they nevertheless cover a wide range of tAs concentrations (Fig. 4.9). We take advantage of the dependence of %iAs on tAs concentrations in Cambodian rice grains (Fig. 4.9c) and adopt the mean values of %iAs for three categories of tAs levels to accommodate the decrease of %iAs as tAs increases (Table 4.10). While this approach is not as accurate as an actual measurement of rice iAs, it is based on MRD rice speciation data and reflects the dependence of %iAs on tAs in global rice grain data (Zhao et al., 2013a, 2013b).

Table 4.10 %iAs in rice grain of Cambodia adopted based on %iAs versus tAs (Fig. 4.9c)

Rice consumption varies globally by nearly four orders of magnitude (0.9–650 g/person/day) among countries according to UN FAO 2004 data (Meharg & Zhao, 2012). Rice is a major staple in Southeast Asian countries (>300 g/person/day) but is only occasionally consumed in European and African countries (<50 g/person/day). Rice consumption also varies in any given country among different ethnic groups (EFSA, 2014). To the best of our knowledge, studies have not collected rice consumption and rice As speciation data simultaneously in the MRD. Therefore, the variations in rice intake among individuals are not considered in exposure assessment. According to World Rice Statistics—FAOSTAT, Cambodian’s per capita milled rice consumption was 430 g/person/day in dry weight in 2019. With a more developed economy, the Vietnamese consumed less rice on average (376 g/person/day, dry weight) with a decreasing trend year by year (FAOSTAT). These values are adopted for our estimations of rice iAs intake described later.

Two examples are given here to illustrate that integrating measurements of As speciation with probabilistic models can establish a more accurate assessment of iAs exposure from food intake (Xue et al., 2010; Zhou et al., 2020). Coupled with ingestion rates and body weights from the National Health and Nutrition Examination Survey (NHANES) from 2003 to 2004 in the US, Xue et al. (2010) used the Stochastic Human Exposure and Dose Simulation-Dietary model to estimate dietary iAs exposure. The results indicate that the mean daily iAs exposure from food (n = 16,931) is 0.05 ± 0.09 μg/kg/day (5th–95th: 0.01–1.4 μg/kg/day), which is around two times higher than iAs from water exposure (0.025 ± 0.104 μg/kg/day, n = 16,883). Among all the foods, rice iAs exposure (0.0085 ± 0.0153) μg/kg/day contributes to 17% and ranked the third, while the largest portion is from vegetables (24%) and the second largest is from fruit juices and fruits (18%). Zhou et al. (2020) conducted a dietary survey (n = 1873) of an urban population in China and analysed 480 market rice samples for tAs and As speciation. Via Monte Carlo simulation, the mean estimated average daily dose of rice iAs exposure is 0.18 μg/kg/day (5th–95th: 0.001–1.224 μg/kg/day), approximately twenty-one times higher than that of the US population. Both studies highlight the variability of dietary and rice iAs intake in a single country.

Uncertainties Introduced by Cooking Water As and Cooking Practice

Unlike South Asians who have a preference for cooking parboiled rice (boiling and drying raw rice before dehusking, often cooked with excess water with the gruel discarded in the end), East and Southeast Asians usually prefer non-parboiled rice cooked with limited water. Regardless of how the rice is processed or cooked, cooking water introduces uncertainty for rice As risk assessment. Although only a few studies have assessed cooked rice tAs level and As speciation, the observations suggest that using high-As water to cook rice enhances tAs and iAs levels in cooked rice. For example, concentrations of iAs in cooked rice (n = 4) increased from 318 ± 84 to 2,255 ± 610 μg/kg, corresponding to cooking water with 0 or 500 μg/L As(V) in a lab experiment (Laparra et al., 2005). In areas with As-contaminated groundwater in Bangladesh, researchers asked two locals to cook market rice (173 μg/kg) with their tube well water (372 or 223 μg/L). Correspondingly, the tAs contents in wet cooked rice increased to 360 ± 16 and 256 ± 40 μg/kg (Bae et al., 2002). A subsequent study in southwestern Bangladesh found that tAs in two parboiled rice and two non-parboiled rice increased by 3 to −58% respectively after cooking with a limited amount of groundwater containing 130 μg As/L without discarding the gruel (Rahman et al., 2006).

Cooking with excess quantities of low-As water followed by decanting the extra water or gruel appears to reduce tAs contents in cooked rice (Brammer, 2009). A laboratory experiment found that cooking with excess deionized water (6:1 water volume: rice volume) reduced tAs and iAs contents by 35 and 45% respectively in cooked rice, compared with raw rice (long-grain and basmati rice, bought from the UK, Indian origin) (Raab et al., 2009). Another lab study found that cooking market rice purchased in Maryland, USA with excess water (10:1 water weight: rice weight) reduced iAs contents in polished long grain rice, parboiled rice and brown rice by 40, 60 and 50% respectively (Gray et al., 2016). Unfortunately, micronutrients in rice such as iron, folate, niacin and thiamin were also lowered by 50–70% at the same time. When low-As water is used to cook rice by rural Bengali households, tAs contents in 29 cooked rice (189 ± 6 μg/kg) were lower than those of raw rice (283 ± 13 μg/kg), with average iAs levels also lowered from 194 ± 8 to 123 ± 8 μg/kg (Halder et al., 2014). In West Bengal, a study reported that the level of tAs in rice (n = 55) cooked with water with non-detectable (<3 μg As/L) ranged from 33 to 138 μg/kg (Mean: 65 μg/kg, wet weight), though it is difficult to compare with the tAs concentrations in raw rice that ranged from 138–482 μg/kg (Mean: 249 μg/kg, wet weight) due to different water contents (Pal et al., 2009).

In the MRD, similar effects of enhancement of As in cooked rice by cooking water have been reported by O’Neil et al. (2013) for Prey Veng Province, Cambodia (Table 4.9). However, in Kandal, Kratie and Kampong Cham of Cambodia, no significant changes in tAs contents in rice after cooking were evident (Table 4.9) (Phan et al., 2013), although the As levels in cooking water were not measured. It is possible households from the study areas of Kandal, Kratie and Kampong did not use groundwater for cooking.

Uncertainties in Bioavailability and Toxicity of Rice iAs

Compared to water iAs, the bioavailability and toxicity of rice iAs are more difficult to demonstrate, with much remaining to be explored and understood. In the following section, results from several in vitro and in vivo studies are summarized to shed light on the bioavailability of rice iAs.

Several in vitro studies simulated gut enzymatic and chemical conditions, i.e. physiologically based extraction tests (PBET) to evaluate rice iAs bioaccessibility. Although in vitro studies are not illustrative of the actual absorption into organisms, the soluble fractions nevertheless can be considered as the upper limit of bioavailability, which can be used to corroborate with bioavailability results from in vivo studies (Ruby et al., 1996). Ackerman et al. (2005) demonstrated a mean of 88.9% (84–94%) of bioaccessibility of iAs in five US cooked rice samples with various ranges of DMA (22–270 μg/kg) and iAs (31–108 μg/kg) after in vitro extraction, suggesting similar bioaccessibility of iAs even though %iAs in rice can vary. Laparra et al. (2005) determined the bioaccessibility of eight types of rice (50–530 μg/kg tAs) cooked with different levels of As(V) (200, 400, 600, 700, 900, 1000 μg/L) in water as 63–99% (cooked rice iAs: 810–3730 μg/kg), suggesting the original rice iAs and additional dosed As(V) by cooking water were both highly soluble. Du et al. (2019) estimated the bioaccessibility of tAs in 42 rice samples (tAs: 50–230 μg/kg) from a mining site in Hunan, China by PBET. The average was 71.7 ± 13.5%, further demonstrating that the bioaccessibility of iAs is not dependent on tAs levels.

Although rice iAs bioaccessibility may not depend on tAs levels, nutrients have been shown to be an influencing factor. Alava et al. (2013) investigated the effect of extraction parameters on the bioaccessibility of iAs in rice, and found the bioaccessibility of As(V) was reduced from 80 to 68% with increasing levels of bile salt while fibre content had no significant effect on rice iAs bioaccessibility. Later, the same group (Alava et al., 2015) found that compared with Western diet (fat and protein rich), rice iAs bioaccessibility increased from 50 to 73% for Asian (fiber rich) diet, which could result from the fat and protein difference considering the minor effect of fibre contents.

A swine model was established to evaluate the bioavailability of As species in cooked rice, finding that 89% of iAs and 33% of methylated As are bioavailable (Juhasz et al., 2006). However, the majority of iAs in rice comes from iAs added to the cooking water, which may differ from the bioavailability of iAs native to rice grains.

According to the “Critical aspects of EPA’s IRIS assessment of inorganic arsenic: Interim report” (NRC, 2013), there had been only one pilot in vivo study that assessed rice iAs bioavailability in humans (He & Zheng, 2010). The report further notes that there is a near absence of concrete data in the absorption and metabolism of rice iAs. Due to variabilities in %iAs in rice and rice intake described earlier, it remains very challenging to conduct an epidemiological study to quantify health risks from ingesting rice iAs. This is likely why regulators have treated the toxicity of rice iAs the same as that of water iAs, a reasonable assumption guided by the precautionary principle, and supported by two in vivo studies summarized below.

Two Asian female adults volunteered for a 10-day diet experiment with a 5-day diet consuming wheat plus selected low-As containing food items, followed by a 5-day diet replacing wheat with rice (He & Zheng, 2010). The rice consumed was purchased from a supermarket in New York City with a tAs concentration of 148 ± 4 µg/kg (n = 3), with %iAs of 78% and %DMA(V) of 19%. The average rice intake was 282 ± 86 g/day for one subject (V1), and 122 ± 34 g/day for another (V2). A mass balance approach (comparison between the urinary As excretion and the food As intake) found that the percentage of dietary arsenic intake excreted by urine was 58% for V1 and 69% for V2, respectively. Assuming that 33% of the DMA(V) in ingested rice was bioavailable to humans (Juhasz et al., 2006) and excreted without further metabolism, the bioavailability of iAs in rice was estimated to be 66% for V1 and 80% for V2, respectively. Because As, once ingested, is also distributed and accumulated in other parts of human body such as skin, internal organs, hair, and nails etc., this urinary excretion-based bioavailability assessment reflects an underestimation. After switching to the rice diet, the %DMA(V) in urine decreased from 83 to 77 ± 3% for V1 and from 92 ± 5 to 88 ± 9% for V2, with minor %MMA(V) and %iAs also detected in urine. Another study conducted a fixed rice intake experiment with six adult male European volunteers. After daily ingestion of 300 g of dry rice (iAs: 99 μg/kg; DMA(V): 99 μg/kg; MMA(V): 3 μg/kg; tAs: 274 ± 10 μg/kg) for three consecutive days, the mean urinary tAs of 6 subjects increased from 6.8 to 49.9 μg/L. By the 5th day, ~ 40% As from ingested rice since the 1st day of the experiment has been excreted by urine. This is slightly lower than the 58 and 69% values reported by the He and Zheng (2010) study because more As is expected to be excreted from the 6th day and so on. Similar to He and Zheng (2010), ~90% of urinary As was DMA(V) with the remaining ~10% being MMA(V) and iAs among 6 European subjects. These two studies provided unequivocal evidence that not only is rice iAs bioavailable, it is methylated in vivo, although whether the methylation occurs in human liver exclusively, or it may involve the human gut microbiome, remains debatable (Coryell et al., 2019).

4.5.5 Exposure to Rice iAs and Health Risks in Cambodia and Vietnam

Bearing the aforementioned uncertainties of rice As speciation, iAs toxicity and intake in mind, plus the complications of water iAs either directly via drinking water exposure route or as cooking water, the following assessment focuses solely on rice iAs exposure, though the water iAs exposure is summarized to underscore that only when water iAs exposure is non-existent then rice iAs exposure becomes important.

Hanh et al. (2011) assessed As intake from water and rice (n = 45) in Au Giang province, MRD of Vietnam, where groundwater As ranged from 0.1–977 μg/L (median 134 μg/L). Average daily consumption of water (3.7 and 2.7 L/day for male and female) and rice (300 and 250 g/day for male and female) were determined by interviews. Assuming 100% iAs in groundwater and 80% iAs in rice grains, per capita daily iAs intake from water was estimated to be 949 ± 714 μg/day for males and 607 ± 620 μg/day for females, while per person daily iAs intake from rice were 53 ± 18 μg/day for males and 45 ± 16 μg/day for females. Since 2008, residents in the area have switched drinking water source from groundwater to low-As filtered water or tap water. The daily iAs intake from water reduced significantly to 1.1 ± 1.7 μg/day for males and 1.8 ± 7 μg/day for females. This illustrates when water iAs intake is low, then the rice iAs becomes a major exposure route. Since 2008, the remaining and dominating iAs exposure is from rice, and were 0.91 ± 0.56 and 0.90 ± 0.57 μg/kg/day for males and females using the average body weights of 58 and 50 kg respectively. The European Food Safety Authority (EFSA) has recommended a BMDL_0.1 of 0.3 μg/kg/day, or a benchmark dose lower confidence limit value for 1% excess risk of cancers of the lung, skin and bladder, as well as skin lesions due to iAs (EFSA, 2009). Although the daily dose of iAs was significantly reduced after changing the water source, it still exceeded the BMDL_0.1, suggesting health risks posed by iAs intake from rice in lower MRD.

Rice iAs exposure was also evaluated in the MRD of Cambodia (Phan et al., 2013, 2014). In Prey Veng Province where groundwater As had been frequently detected, a survey evaluated iAs exposure from water and rice for 12 females and 11 males with body weights of 42.5 ± 11.1 and 55.4 ± 4.8 kg, respectively (Phan et al., 2014). The survey found that daily drinking water consumption was 1.375 ± 0.433 and 1.818 ± 0.337 L/day for females and males, and that the tAs concentrations of groundwater were 118 ± 139 μg/L. Thus, the daily iAs exposure from drinking groundwater was estimated to be 162 ± 2 μg/day for females and 215 ± 3 μg/day for males assuming 100% iAs in groundwater; or 3.818 ± 0.048 and 3.872 ± 0.046 μg/kg/day for females and males (Phan et al., 2014). The survey also determined that rice consumption was 450 ± 0 and 429 ± 72 g/day for males and females, and that the tAs concentrations of rice were 201 ± 50 μg/kg (Phan et al., 2014). Therefore, the daily iAs exposure from rice was 72.4 ± 0.2 μg/day for males and 69.0 ± 0.2 μg/day for females assuming 80% iAs in rice, or 1.306 ± 0.003 and 1.623 ± 0.006 μg/kg/day for males and females respectively (Phan et al., 2014). Like Vietnam, water iAs exposure exceeds rice iAs by several folds. But even when water As exposure is reduced through mitigation (see next section), iAs exposure from rice still exceeds the European BMDL_0.1. This is also the case for Kandal, Kratie and Kampong Cham Provinces where cooked rice tAs concentrations have been measured (Table 4.9) and used to estimate exposure. Phan et al. (2013) estimated rice iAs intake in Kandal, Kratie and Kampong Cham to be 1.8 ± 2.4, 0.6 ± 0.4 and 0.09 ± 0.07 μg/kg/day respectively, assuming a typical rice consumption of 450 g/person/day and an average body weight of 52 kg, as well as a mean of 80% of tAs as iAs. Again, rice iAs exposure in Kandal and Kratie have exceeded the European BMDL_0.1 but not in Kampong Cham.

Because it is imperative that water iAs exposure must be reduced, and that it is less difficult (though not easy) to address than rice iAs exposure, we illustrate the current scenario of the range and average rice iAs exposure excluding water iAs complications. Because rice grains in Cambodia display a wide range of tAs concentrations (Table 4.9), the daily intake of iAs from rice grains ranges from 3 to 174 μg/day (68 ± 106 μg/day) for Cambodian females and ranges from 4 to 219 μg/day (85 ± 133 μg/day) for Cambodian males. The estimate here uses three %iAs values corresponding to three ranges tAs levels (Table 4.10). The per person daily consumption rate of dry rice according to FAOSTAT in 2019 was 430 g/day in Cambodia. Because a survey (Sar et al., 2012) has found that men consume 26% more rice than women, so the male and female rice consumption is partitioned to be 381 and 479 g/day. Using an average body weight of 59.7 and 52.8 kg for male and female Cambodian respectively (WorldData, 2020), daily rice iAs exposure ranges from 0.06–3.67 μg/kg/day (mean 1.43 ± 2.23 μg/kg/day) for males, and ranges from 0.05–3.30 μg/kg/day (mean 1.28 ± 2.00 μg/kg/day) for females. Even if only the mean values are considered, they exceeded the European BMDL_0.1 by >4 fold. Given the similarities in tAs contents in raw rice grains (Table 4.9), daily rice consumption (376 g/person/day, FAOSTAT, 2019), and average body weight (61.2 kg male and 54.0 kg female Vietnamese, WorldData, 2020), the average of daily rice iAs exposure again shows a wide range from 30 to 133 μg/day (66 ± 33 μg/day) for Vietnamese males and from 25 to 111 μg/day (55 ± 28 μg/day) for Vietnamese females. This is equivalent to the daily iAs dose from rice grains ranging from 0.48 to 2.18 μg/kg/day with the average of 1.09 ± 0.54 μg/kg/day for males, or from 0.46 to 2.06 μg/kg/day with an average of 1.03 ± 0.51 μg/kg/day for females. The European BMDL_0.1 of 0.3 μg/kg/day is equivalent to 16.5 μg/day iAs intake for an adult of 55 kg weight, and is equivalent to drinking 2 L of water containing 8.25 μg/L iAs. In summary, the average daily exposure to iAs from rice alone in the MRD, at about 1.20 μg/kg/day for a 55-kg adult is equivalent to drinking 2 L of water containing ~33 μg/L of iAs.

4.6 Arsenic Mitigation and the Way Forward

Due to the latency effect of chronic inorganic arsenic exposure from drinking water, and that currently there is no cure other than reducing arsenic exposure, replacing drinking water supplies from wells with elevated arsenic in the Mekong River Delta with a low arsenic supply should be and has been the priority. Although there are no currently known plans to test all drinking water sources for arsenic in the Mekong River Delta regions for arsenic, our field work has discovered anecdotal evidence that many villages have switched to a communal water supply especially close to the capital city Phnom Penh of Cambodia. We recommend a Mekong River Delta drinking water quality survey to assess the remaining risks. This is the critical first step towards mitigating arsenic exposure from groundwater. The water quality survey whenever possible, should sample irrigation wells considering the substantial risks from rice exposure due to the usage of high arsenic groundwater for irrigation.

The assessment of rice iAs exposure in the MRD here, although still preliminary, is sufficient to underscore the need to pay attention to this hazard that is likely to become more threatening in the future after the water iAs exposure is brought under control. Due to the cumulative nature of As contamination of soil by irrigating with high-As groundwater, it is prudent to move away from irrigated agriculture practices using this unsafe groundwater source. Because surveys of paired paddy soil and rice sampling in the Mekong River Delta area are still uncommon, it is helpful to investigate arsenic cycling in the hydro-geo-biosphere to improve our knowledge of the bioavailability and toxicity of arsenic in rice. This will allow for a more reliable assessment of the at-risk areas and at-risk-rice cultivars. Such knowledge can inform plans to manage rice cultivation in the MRD that will increase rice productivity and ensure food safety.