Evaluation of mercury concentration in the lake biwa-yodo river basin by a one-box multimedia model and model sensitivity on the experimentally determined water-sediment partition coefficient

Considering the environmental damage caused by mercury, evaluating mercury concentrations in four environmental media, namely the atmosphere, water, soil, and sediment, is necessary. Available data on mercury emissions and computational modeling were used for this evaluation. Evaluating the sensitivity of the model for the water-sediment partition coefficient was another objective of this study. Recorded consumption and emission data of mercury were used to calculate the total annual emission amounts from 1959 to 2009 for the Lake Biwa-Yodo River basin of Japan, which was selected as the study site for this study. Laboratory-scale batch-shaking experiments were carried out to evaluate the partition coefficient of mercury between the aqueous and solid phase using soil, sediment, and sand samples. The experimentally determined partition coefficient was then incorporated into the one-box multimedia model. Mercury concentrations in this study site were calculated based on the calculated annual mercury emission data. The sensitivity of the model calculations on the partition coefficient was studied by comparing the calculated concentrations from different partition coefficient values with observed data. Calculated concentrations of mercury in all four environmental media were within the range of observed concentrations, and the performance of the model was validated. The results showed the accumulation of mercury in soil and sediment, and in the last 30 years, the mercury concentration has been constant, except in the atmosphere. Variations in atmospheric mercury concentrations were observed. Calculated concentrations for the representative partition coefficients for different soil types were compared with the environmental monitoring data. Improvements in the model performance due to the incorporation of an experimentally evaluated partition coefficient were confirmed. Thus, it was concluded that the one-box multimedia model could reliably calculate the environmental mercury concentrations based on the emission data. The sensitivity of the model was improved by using the experimentally evaluated partition coefficient value of mercury. Experimental evaluation of the other parameters used in the model calculations would further improve the model.


Introduction
Mathematical and computational modeling methods are used to evaluate the chemical pollution of our environment. Combined with improved management processes,we can now predict pollution conditions using environmental modeling. Various studies ranging from modeling macro-scale environmental issues such as global climate change to localized micro-scale environmental modeling assessments can be found in the scientific literature. In 1983, Hansen published a study on efficient three-dimensional global models for climate studies: Models I and II providethe proof of the geographical macro-scale of these studies [1]. In his review on the community multiscale air quality (CMAQ) modeling system, Byun describes the different components of chemical behaviors considered in air quality models [2]. In another study, Kondo published the use of the one-Box multimedia model(OBMM) to evaluate the lead concentration of the Lake Biwa-Yodo River basin(LBYRB) in Japan [3]. Mercury (Hg) has caused serious environmental and health damage in Japan in the past. In 1956, a disease caused by methylmercury poisoning was discovered in Minamata city, which is located in the southwest region of Japan's Kyushu Island; this disease was later named Minamata disease [4]. Consumption and release of Hg has been legally controlled in Japan since 1973 [5], but still, many industries consume Hg in their production processes; thus, Hg appearsin the environmental analysis data [6]. The Japanese Ministry of Environment has set the environmental standards for the Hg concentration in water and soil to be less than 0.5 µg L -1 and for the alkyl mercury limit to be less than the detection limits [7]. In June 2013, a global treaty of The Minamata Convention on mercury was established and acknowledged by 96 countries (by February 2014) to protect human health and the environment from the adverse effects of Hg [8].
Monitoring environmental Hg in a large geographical region is not a very practical. The dispersion of Hg through environmental doi: 10.7243/2050-1323-3-3 transport mechanisms is one reason. An other reason is the impracticality of a government organization (i.e., the Ministry of Environment, Japan) to observe the occurrences of Hg all over the country through environmental monitoring. Considering the practical difficulty in monitoring the whole country for Hg occurrences and the adverse health effects of Hg such as carcinogenicity, child developmental defects, and toxic effects on nervous, digestive, and immune systems [9], it is importantto evaluate the concentration of Hg using environmental modeling. There are published model studies on the atmospheric transport and aquatic-soil systems of Hg [10,11], but a computational model that can combine all of the environmental media of the atmosphere, water, soil, and sediment was required. For this purpose, chemical behaviors and the transport mechanisms of Hg in all four environmental media were mathematically interpreted into an OBMM. The annual Hg emissions were estimated based on the available consumption and emission data from various data sources and used for the OBMM simulations to evaluate the concentration of Hg.
Several chemical properties related to the metallic pollutants' behavior, such as partition coefficient, dissolution coefficient, and diffusion coefficients, are required for these calculations [3,[10][11][12][13]. Because the water-sediment partition coefficient(K d ) is important to interpret the chemical behavior of Hg [3,[10][11][12][13], laboratory-scale batch-shaking experiments were carried out to determine the K d of Hg (K d (Hg) ) using samples of soil. The experimentally determined K d (Hg) value was incorporated into the OBMM calculations to improve the performance of the OBMM. The sensitivity of the OBMM to the K d (Hg) value was evaluated by comparing the observed data for Hg concentrations in water and sediments with the calculated concentrations based on the experimental K d (Hg) values obtained from other samples such as sediment and sand.
The Lake Biwa-Yodo River basin is an important geographic area with multiple land use patterns, including residential, agricultural, and industrial, on Japan's Main Island. It covers portions of six prefectures, namely Hyogo, Kyoto, Mie, Nara, Osaka, and Shiga, in the Kinki region. This study area lies between the latitudes 34.65~35.69 ºN and the longitudes 136.15~136.51ºE, while Lake Biwa, the largest natural water body in Japan, is located in the middle of this study area, covering 630.77 km 2 . The Seta River starts from the southern tip of Lake Biwa, turns into the Uji River, and then joins with the Kizu River and Katsura River to become the Yodo River, which flows to Osaka Bay [14]. This lake-river system also provides a natural drinking water source for a population of nearly 13 million in the Kinki region. Because of these important reasons, the LBYRB was selected as the study site. Figure 1 presents a diagram of the study area and the sampling points.
The main objectives of this study wereto evaluate the concentration of Hg in the LBYRB using an OBMM and to improve the OBMM calculation by incorporating the experimentally evaluated K d (Hg) value.

Methodology
Laboratory-scale batch experiments were carried out with soil to determine the water-sediment partition coefficient of Hg (K d(Hg) ), and this value was incorporated into the OBMM calculations. Estimated annual emissions of Hg in the LBYRB were used as the input data and concentrations of Hg in the atmosphere, water, soil, and sediments in the LBYRB were calculated using a OBMM for a span of 50 years from 1959 to 2009. The accuracy of the OBMM calculations was evaluated by comparing the calculated Hg concentrations with the observed data for the year 2009. The model sensitivity to the K d (Hg) was evaluated by comparing the calculated Hg concentrations with respective K d (Hg) values, which were experimentally determined from sediments and sand samples collected from different locations in the study area. The study site is described in the introduction and the model is describedunder OBMM simulation section.

Experimental evaluation of water-sediment partition coefficient Materials and equipment
A soil sample was collected from sampling points as shown in Figure 1 and cleaned Teflon containers were used to collect the samples. Temperature and pH of the collected sample were measured at the sampling points and a drying oven was used for drying the samples. Glass bottlesof 500 mL volumes were used for the shaking experiments which were performed with a thermostat shaker. Milli-Q water was used in the control experiments, and a standard mercury solution (HgCl 2 in 0.1 mol L -1 . HNO 3 (Hg: 100 mg L -1 ), purchased from Wako Pure Chemical Industries Ltd., Japan), was used for spiking of Hg. Chemical analysis of the concentration of Hg was performed at a certified chemical analysis facility (Teijin Eco Science Ltd., Japan) using Atomic Absorption Spectroscopy (AAS) with a Mercury Analyzer following Japanese industrial standards [15,16].

Experimental setup
The experimental conditions and the experimental steps performed for the determination of the K d (Hg) are shown in Figure 2. After the shaking experiments the samples were analyzed for the Hg concentrations in both aqueous and solid phases. Partition coefficients were calculated for each sub-sample, as shown in equation (1) by using the Hg concentrations in obtained from the chemical analysis [17,18]. After the preliminary evaluation, a secondary step was performed to confirm the results of the preliminary evaluation for longer shaking periods of 15, 22, and 30 days using similar experimental procedure for pretreating the four soil samples as referred to the secondary evaluation in the Figure 2. To observe the variations in the K d (Hg) in other environments with different soil types, samples were collected from sampling points 2 and 3, which represent sediment and sand. Experimental procedures similar to the preliminary evaluation were performed, and the representative K d (Hg) values were calculated. These K d (Hg) values were applied to the OBMM calculations at a later stage to evaluate the sensitivity of the OBMM to different K d (Hg) values. Total organic carbon (TOC) was measured in all of the control samples according to the general rules for chemical analysis stated by the Japanese Industrial Standard Committee [15].

Hg emission amount calculation and OBMMsimulations Annual Hg emission amount calculations
Annual emissions of Hg from 1959 to 2009 were calculated for the LBYRB based on the records of Hg consumption and data from the Pollutant Release and Transfer Registry (PRTRwhich is maintained by the Japanese Ministry of Environment). From 1959 to 1990, the annual emissions of Hg were calculated based on the reported Hg consumptions for industries, catalysts, fertilizers, pharmaceuticals (inorganic chemicals), machinery, batteries, medical supplies (amalgam), explosives (gun powder), and paints [19,20]. The PRTR records the emissions under two categories: registered PRTR emissions and non-registered PRTR emissions [21].

One-box multimedia model
The study site is composed of four environmental media, which are the atmosphere, water, soil, and sediments, and is considered to be a three-dimensional, closed entity in this model. The nine chemical phenomena considered in the model calculations are emission, degradation, advection, sedimentation, re-suspension, dry/wet deposition, atmospheric mixing, and particle-ion exchange, which are diagramed in Figure 3. These calculations were performed based on the conditions that the Hg is in chemical equilibrium between the environmental media and observes the mass conservation law in the environmental systems. Time steps for these calculations were set to 6 minutes, and a series of differential equations was solved using the Runge-Kutta technique by a computer program coded in FORTRAN. The main equation of OBMM is given by the equation (2) and the main variables considered in these calculations are as follows: (I) Emission of Hg into the atmosphere, soil, and water (II) Degradation of Hg in all four environmental media (III) Exchange of Hg +2 -H g(S) within/between the environmental media (IV) Transport of Hg by advection in the atmosphere (V) Dry and wet deposition of Hg from the atmosphere (VI) Sedimentation and re-suspension of Hg in water Utilization of an OBMM was previously published, and more details about the model can be found in Kondo et al., 2013 [3].

Determination of K d (Hg)
The initial pH and temperature measured at the sampling sites are given in Table 1. During the shaking experiments, the temperature was maintained at 25 ºC to avoid temperature changes affecting the chemical behavior of Hg. Measurements of the TOC were taken in the control samples after the shaking experiments to investigate the amount of organic materials present in the samples because the adsorption-desorption processes of metals are affected by the presence of organic matter [17]. Among the three different samples studied, the soil-Milli-Q water (control) sample had the highest TOC measurement of 370 mg L -1 , sediment-Milli-Q water (control) sample had 23 mg L -1 ,and the sand-Milli-Q water (control) sample had 9.9 mg L -1. These evaluations were carried out with the assumption that Hg will acquire equilibrium between the aqueous and solid phases of the sample after the shaking experiments and the evaporation of Hg into the air inside the glass bottles is negligible because of the low volume of air in the glass bottles and the low concentrations of Hg spiked. When deciding the range of the spiking dose for the preliminary evaluation, the following facts had to be considered. If the concentration in the sample is too low, then the forward reaction (adsorption) becomes slower, resulting in a prolonged shaking time for the system to reach equilibrium; the chemical analysis of Hg also becomes more difficult at lower concentrations [18]. If the spiking dose is higher, then the solid phases of the samples might become saturated with Hg. Therefore, considering the reported occurrences of Hg in soil ranging from 2-900 µg kg -1 [22], the range of the Hg spiking was set at 250-1000 µL L -1 in the preliminary evaluation. Figure 4 summarizes the results obtained from the shakingbatch experiments to evaluate the K d (Hg) . The concentrations of Hg in the aqueous and solid phases of the samples were used to calculate the K d (Hg) values for each sub-sample. The calculated K d (Hg) values for the soil sample were plotted against the spiked volumes of Hg per 1 L of Milli-Q water, as shown in Figure 4. Variations in the K d (Hg) values were extrapolated with the best fitted second-order polynomial curve to observe the trend of the K d (Hg) at higher spiking doses than 1000 µL L -1 . Similar trends have been reported in another publication by Yin Y. et al., in 1997 [23]. Because the greatest portion of the LBYRB is composed of land, the soil sample was selected as the representative sample. After 3 days of shaking at 40 rpm, 36 µg kg -1 of Hg was detected in the control sample for soil, but the amount of Hg released to the Milli-Q water media from the soil was more than 250 times smaller, showing that the Hg already existing in the sample was hardly involved in the adsorption-desorption reaction. The loam texture of the soil and the presence of higher organic content results in the lower desorption of Hg from the soil [23]. At the 1000 µL L -1 spiking dose, the soil sample showed a maximum K d (Hg) value for the sample of 65.8. Considering this result and the kinetics of the adsorption-desorption reactions, 1000 µL L -1 was chosen as the spiking dose to reduce the elongated time for Hg to reach equilibrium and to avoid saturating the solid phase [23]. Confirmative experiments were performed for shaking periods of 15, 22, and 30 days to evaluate the K d (Hg) values for prolonged shaking periods. The experimental conditions applied in this step are shown in 125 at the Hg spiking dose of 500 µL L -1 ; then, the K d (Hg) value started to decrease. The reasons for this behavior are either that the sediment samples became saturated at the higher spiking doses of Hg over 500 µL L -1 , so the adsorption process of Hg was stopped, or the desorption of Hg from the sediment, releasing it to the aqueous phase. Sub-samples of the sand reached a maximum K d (Hg) value of 10.6 at the Hg spiking dose of 750 µL L -1 . Therefore, the K d (Hg) values of 10, 65, and 120 were selected to represent the samples of sand, soil, and sediment, respectively. The OBMM calculations for these K d (Hg) values were compared with the OBMM calculation carried out with K d (Hg) of 80. The accuracy of these results could be improved by evaluating the K d (Hg) between the aqueous and solid phases of samples collected for different geographic areas in the study site. In this study, the number of replicates was one; therefore, increasing the number of replicates for the  sub-samples would have increased the accuracy of the K d (Hg) evaluation. Saturation levels of the solid phase samples have to be studied in detail to confirm the occurrence of the desorption process. Additionally, further studies about the composition of the solid phase samples would have given more detailed explanations of the behavior of Hg between these two phases.

Hg emission amounts, OBMM simulations, and sensitivity evaluation
In order to perform OBMM simulations, Hg emissions to the atmosphere, water, and soil from 1959 to 2009 were calculated and the total Hg emissions for that time span are shown in In Figure 6, the calculated Hg concentrations by the OBMM with the experimentally evaluated K d (Hg) value (=80) for the atmosphere, water, soil, and sediments are shown in log 10 scale for a 50-year span. Atmospheric Hg concentrations were reported from 2.54x10 -6 to 1.20x10 -5 µmol m -3 , while concentrations in water ranged from 1.27x10 -2 to 6.66x10 -2 µmol m -3 . In the soil, the Hg concentration varied from 5.78x10 -2 to 6.96x10 -1 µmol kg -1 , and in the sediments, the Hg concentration varied from 3.03x10 -1 to 5.57 µmol kg -1 . According to the temporal concentration trends shown by the results in Figure 6, accumulation of Hg in the soil and sediments can be seen until the1970s; after that time, the      Many chemical coefficients other than the K d (Hg) affect the chemical behavior of Hg in the environment, so the OBMM surely can be improved by experimentally evaluating and incorporating those experimentally evaluated coefficients in to the OBMM calculations. The availability of the emission data and observed data for the occurrences of Hg in different environmental media is really important to improve the accuracy of the OBMM calculations and the validation of the calculated results of the model.

Conclusion
The concentrations of Hg in four environmental media, namely the atmosphere, water, soil, and sediments, of the LBYRB were evaluated using an OBMM. Annual emissions of Hg were calculated using the literature on Hg consumption and PRTR data for a 50-year span. Laboratory-scale experiments were carried out to evaluate the K d (Hg) between the aqueous and solid phases of the environment, and an experimentally determined K d (Hg) value (=80) was incorporated in to the OBMM. The concentrations of environmental Hg were calculated using the annual emission amounts and the OBMM. The calculated Hg concentrations were compared with the observed Hg concentrations in the study area to validate the accuracy of the OBMM calculations. Variations were observed in the calculated atmospheric Hg concentrations while in the sediment and soil, accumulation of Hg was observed. Except for the atmosphere, the calculated Hg concentrations in the soil, water, and sediment became constant over the last three decades of the study span, due to the controlled consumption of Hg. The sensitivity of the OBMM calculations to the K d (Hg) value was studied by comparing the calculated Hg concentrations on different K d (Hg) values representing different soil types. The results showed that