3.1. Spatial Distribution of Arsenic Concentrations
First, we perform a descriptive statistical analysis of the arsenic concentration data for the Pingtung Plain collected from the Taiwan’s Water Resources Agency from 2009 to 2013. The statistics for the As concentrations measurement at the monitoring wells are summarized in Table 2
. There is a considerable variation in the measured concentrations, from below the detection limit (<0.1 µg/L) to the maximum value of 544 µg/L. The maximum arsenic concentration is 50 times the threshold value of 10 µg/L recommended by the WHO. The average As concentration is 18.1 μg/L with a standard deviation of 65.2 μg/L. Moreover, the threshold value of 10 µg/L corresponds to the 80.26th percentile of the percentage frequency distribution of arsenic concentration. In other words, approximately 20% of the measured As concentrations exceeds the threshold value of 10 μg/L.
Efforts are made to clarify the spatial distribution of arsenic concentrations in the groundwater; the measured concentrations in each aquifer are depicted in Figure 4
. The measured As concentrations are categorized into four levels based on Taiwan’s drinking water quality standard: completely uncontaminated (<3 μg/L), moderately uncontaminated (3–10 μg/L), moderately contaminated (10–50 μg/L), and severely contaminated (>50 μg/L). In the figure, the As levels are represented by the sizes of the solid circles. There is a pattern with obvious increase in the As concentration in Aquifers 1, 2 and 3 from the northeastern to the southwestern coastal area with most of the groundwater samples exceeding 10 μg/L located in the southwestern part of the study area. A few others are scattered in the central and northern areas. In Aquifer 4, in the western and southern part of the study area the arsenic concentrations are high.
Next, the spatial pattern of the arsenic concentrations is analyzed and interpolated into a GIS environment using the geostatistical Kriging method. A histogram of the arsenic concentrations is prepared prior to calculating the semi-variogram, reveals a lognormal distribution rather than a normal distribution. Thus, log transformation of the measured arsenic concentrations is used for calculation of the semi-variogram. Comparison of the calculated semi-variograms for the logarithms of arsenic concentrations with different theoretical semi-variogram models shows that the Gaussian model has the best fit.
The theoretical Gaussian semi-variogram model is selected and the spatial distribution of the arsenic concentration is estimated by calculating the exponent transformation of the estimated values of the logarithmic concentrations for each cell using Equation (5). Each aquifer is discretized into a grid system of 2448 cells with a cell size of 1000 m × 1000 m. Figure 5
presents the spatial distribution of the estimated arsenic concentration obtained using the geostatistical Kriging method. The arsenic concentrations in Aquifer 1 are higher in the townships of Linyuan, Linbian, Sinyuan and Jiadong. In Aquifer 2, the arsenic concentrations are higher in the townships of Dongang, Linbian and Nanzhou. For Aquifer 3, the areas with higher arsenic concentrations cover the townships of Dongang, Linbian and Nanzhou. Only a small portion of the area associated with Aquifer 4 has an excessive arsenic concentration. Overall, the highest As content is apparent in Aquifers 2 and 3 in the southwestern area. The lower As concentrations in Aquifer 4 indicate that this may be a suitable safe zone for the withdrawal of groundwater. The As concentrations are obviously lower in the northern and eastern parts of the study area, in Aquifers 1–4, indicating that they also could be primary safe water sources.
3.2. Spatial Arsenic Risk Assessment and Health Risk Implications
Given the known health risks associated with the ingestion of As, we set out to identify the areas of concern and quantitatively assess the health risk of drinking As-contaminated groundwater throughout the Pingtung Plain. The spatial distribution of the hazard quotient (HQ) index for non-carcinogenic health risk is evaluated using Equation (6) with the aid of estimated arsenic concentrations in each cell obtained from the geostatistica Kriging approach. Figure 6
maps the estimated spatial distribution of HQ values corresponding to each aquifer associated. The cells where HQs > 1 are indicated in red and are located primarily in the southwestern part of the study area, especially in the townships of Linyuan for Aquifer 1, Sinyuan, Dongdang, Linbian and Nanzhou for Aquifers 2 and 3. The estimated HQs for Aquifer 4 are all less than the 1.0 throughout the entire study area.
Taking the estimated arsenic concentrations for each cell for each aquifer from the previous section, the spatial distribution of the target cancer risk (TR) index is evaluated using Equation (8). Figure 7
maps the estimated TRs for each aquifer. The estimated TR values are 1.8 to 1890 times higher than the acceptable standard (one millionth, 10−6
) throughout the study area, for each aquifer. In Figure 7
, the areas with the highest TR values (>10−4
) are indicated in red. For Aquifer 1, the TR values are highest in the townships of Daliao, Linyuan, Sinyuan, Donggang, Linbian, Jiadong and Fangliao (>10−4
). For Aquifers 2 and 3, the areas with the exceedingly high TRs include the townships of Linyuan, Sinyuan, Donggang, Linbian and Jiadong (JD). In Aquifer 4, most areas display TRs ranging from 10−5
with the exception of a small portion of the township of Fangliao which has TR values greater than 10−4
. It should be noted that most of the townships with high TRs have larger population densities. The areas with high TRs and high population densities appear to be facing a more severe public health issue and thus deserve special attention.
The percentage the area where the TR values are estimated to be in excess of 10−5
and the population densities for the individual townships in the Pingtung Plain are shown in Table 3
. Examination of the results show that an exceedingly high percentage of the population is exposed to high carcinogenic risk from As-affected groundwater. Most of the high population density townships have a level of risk exceeding 10−5
Given the area percentage (%) where the estimated TR exceeds 10−5
and the actual population of individual townships, it is simple to estimate that more than the health of 83.9%, 78.7%, 81.3% and 75.9% of the population corresponding Aquifers 1, 2, 3, and 4, respectively, is at risk due to the ingestion of As containing groundwater. Furthermore, over 17.1%, 10.7%, 13.3% and 0.4% (for Aquifers 1, 2, 3 and 4 respectively) of the population have been exposed to a level of risk exceeding 10−4
. A spatially explicit map that delineates the high population density areas where residents are exposed to greater carcinogenic risk can be prepared by integrating the spatial distribution of the estimated TR and population density; see the map in Figure 8
based on TR > 10−4
and population density (PD) > 2000 person/km2
. Four classes of exposure are considered.
In addition to the ingestion of As from drinking groundwater, the consumption of the contaminated foodstuff is also an important exposure source. Contaminated groundwater is also being used for aquaculture. The accumulation of As in farmed fish and seafood pose a potential threat to human health [19
]. Figure 9
shows fishpond locations in the study area. Many are located in the southwestern part of the Pingtung Plain and in the higher population density townships, which are exactly in the higher risk areas as estimated in our study. The exposure associated with the consumption of fish and seafood actually increases the carcinogenic risk and should be considered in the future.