Health risk assessment and cost–benefit analysis of agricultural soil remediation for tailing dam failure in Jinding mining area, SW China

The impact of the tailing dams and the economic feasibility of the remediation process is significant for future risk management for tailing dams. In this research, we develop a hypothetical failure scenario for a tailing dam in the Jinding mining area, Southwest China. We assess the exposure with the Geo-Environmental Risk Assessment System, tier-1 model, and health impact with Disability-Adjusted Life Years (DALY). Cost and benefit are also analyzed for the following clean-up process. The result shows that the exposure dose (mg/kg–BW/d) of As, Cd, and Pb right after the dam failure is 1.07 × 10–2 for As, 1.76 × 10–4 for Cd, and 5.68 × 10–3 for Pb, respectively. The DALY caused by heavy metal exposure is 2.63 × 10–2 DALY per year, which significantly exceeds the tolerable level. This indicates that the tailing dam failure will pose a high health risk to the residents, and remediation is necessary. After remediation, the DALY is 1.24 × 10–8 DALY per year, indicating the clean-up process effectively reduces the resident’s health impact. From the financial point of view, the net present value of the clean-up is $− 1.02 × 107. This indicates that the clean-up process is not economically feasible. Sensitivity analysis shows that the amount of released tailing influences the output result. The time span for benefit estimation is also an important issue. This research shows that the impact of a tailing dam failure will be severe, and remediation may be effective but economically infeasible. Therefore, preventing tailing dam failure is the most crucial task for the local government.


Introduction
A tailing dam is a typical structure built for storing tailing and waste in mining areas, preventing the heavy metal (HM) containing tailing from causing environmental problems. However, tailing dams have a higher risk of failure than normal reservoir dams, potentially releasing tailings into the environment (Liu et al., 2020). In recent years, the number of tailing dam failures in developing countries has been increasing (Lyu et al., 2019). In China, there were still more than 8869 mine tailings ponds active by the end of 2015, including 1425 tailings ponds close to residential areas and other important infrastructures within 1 km downstream (Dong et al., 2020 The consequences of tailing dam failure can be severe. For example, the break of the Aznalcóllar tailing dam on Apr. 25th, 1998, in Spain, flooded approximately 4600 hectares of land (Hudson-Edwards et al., 2003). Tailing dam failure requires immediate remediation actions. Otherwise, the tailings will possibly influence the land continuously. In China, in 1985, a tailing dam failure happened in a Pb/Zn mine in Chenzhou (Liu et al., 2015). The clean-up action was only applied to limited areas. Research showed that the HM in the soil in the untreated areas extremely exceeded the Chinese agricultural soil standards even 17 years after the failure. Therefore, the remediation and recovery process after tailing dam failure is important for tailing dam management. Understanding the remediation process for tailing dam failure is significant for future management and policymaking, including economic understanding.
Researchers have discussed the mechanism and consequences of past tailing dam failure events. Such as seismic behavior of the dam structure (Ishihara et al., 2015) and long-term impact on the marine environment (Sá et al., 2021). However, for future management of tailing dams, predictive studies which provide a quantitative risk assessment of potential tailing dam failure are required. Predictive risk assessment of tailing dam failure focused on two aspects, the likelihood of a tailing dam failure happening and the risk caused by the tailings released into the environment. Most existing research focused on the previous aspect. For example, Nišiç et al., (2018;2021) estimated the risk of tailing dam failure in two Serbia metal mines with a hierarchy risk matrix considering the likelihood and consequence severity of the tailing dam failure. Liu et al. (2019) developed a Bayesian network-based simulation model to assess the water pollution caused by tailing dam failure in China. Compared to these researches, some other risk assessments extended further to consequences after the failure, such as health risk and economic loss. (Barcelos et al. (2020) developed a hypothetical dam failure event in a gold mine in Brazil and conducted a quantitative risk assessment accordingly. As a result, As, Cd and Zn showed potential human health risks. But the author did not include economic considerations. In China, Liu et al. (2020) assessed the risk of the tailing dam collapse in a waste slag site. The author estimated the vulnerability of buildings and health risk assessment for the residents under a failure scenario. Financial loss is evaluated based on land price, clean-up budget, placement of residents, etc. Health impact was not included.
The resident always suffers the biggest impact in a tailing dam failure event. Other than the flooded land and suspension of the mining process, the health impact on the residents is another critical consequence and may have a long-term influence on the area. The studies mentioned above did include the health risk and economic point of view. However, to the author's knowledge, no study is available on the financial impact considering the health impact of tailing dam failure, as well as the following remediation process.
Considering the research gaps, the objective of this research is (1) to assess the resident exposure of HMs and the corresponding health impact under the hypothetical tailing dam failure scenario; (2) to conduct a cost-benefit analysis for the remediation process based on the reduced health impact. This research fills the research gap by investigating the cost-effectiveness of soil remediation for tailing dam failures. The analysis procedure in this research provides a basic methodology for cost-benefit analysis based on health benefits under disastrous scenarios. The result of this research is expected to provide reference on tailing dam emergency management for the local government and policymakers.

Research area
The target area is agricultural land in Jinding Town. Jinding Town (26°25' N, 99°24' E) is located in Lanping county, Yunnan Province, China. The altitude is about 2300 m. Under a monsoon climate, the average temperature is 11.7°C, and the annual average precipitation is 1002.4 mm (Li et al., 2019). Jinding Pb/Zn mine locates in the center of Jinding Town. Jinding deposit is the largest Pb/Zn deposit in China and the world's youngest sediment-hosted super giant Pb/Zn deposit (Xue et al., 2007). This area has had a mining history for several decades and has been polluted by mining activities.

3
Vol.: (0123456789) The target land was selected as the total area of agricultural land in the entire Jinding Town is 9.3km 2 . However, not all these agricultural lands will be possibly affected by a tailing dam failure, especially the lands located upstream from the tailing dams. Therefore, in this paper, the target area is selected as agricultural land in front of the tailing dam in the Jinding mining area (Fig. 1), which has a high possibility of being flooded if a tailing dam failure happens. The target area is approximately 28 hectares. There are two tailing dams in the Jinding mining area. The target agricultural land is located in front of the two tailing dams and is the closest agricultural land to the tailing dams in this mining area. The shortest distance from the tailing dam to the target area is approximately 0.47 km. Under the hypothetical tailing dam failure event, the target land is assumed to be thoroughly and evenly affected by the tailings released.

Scenario definition
Since the target dams have not experienced a failure, we choose a historical tailing dam failure event as a reference. The dam failure scenario is defined based on the Aznalcóllar tailing dam failure in Spain. Similar to the Jinding mine, the Aznalcóllar mine is also a large-scale Pb-Zn mine (which produces Cu and Ag as well), and Aznalcóllar tailing dam failure is a major dam failure event that was proved to have caused severe consequences. Considering the mining scale and the metal types, the release potential of the Aznalcóllar mine is expected to be similar to the Jinding mine. The Aznalcóllar tailing dam was designed to hold 32.5 million m 3 of tailings, and the maximum height was 32 m. The Aznalcóllar tailing dam failure happened on Apr. 25th, 1998. At the time of failure, the tailing pond contained approximately 15 million m 3 of tailing. Approximately 1.3 million m 3 of fine pyrite tailings and 5.5 million m 3 of wastewater are released into the surrounding environment. The tailings make their way to about 40 km downstream. The tailings released in the accident are concentrated with HMs; the mass content is 0.6-0.7% As, 0.8-1.0% Pb, and 0.003-0.004% Cd. (Mcdermott & Sibley, 2000).

Accidental release
Since the two tailing dams in the Jinding mining area locate close to each other, we assume a single-point release. The amount and magnitude of release in this scenario reference the Aznalcóllar tailing dam failure. In this scenario, we assume that about 80% of the 1.3 million m 3 released tailings settled within 12 km of the mine site on approximately 800 hectares of riverbanks and agricultural land (Mcdermott & Sibley, 2000).

Physical and chemical properties of the tailings
For a more straightforward calculation, the target area is assumed flat, and the tailings are evenly distributed in the flooded area. Due to the flooding, the soil is to south, crossing the area. b Households in target residential area (244 households in total, 2.62 people per household (Statista, 2022)) expected to mix with the tailings. In Barcelos et al.'s (2020) research, the author assumed the tailings are mixed 1:1 with the soil due to flooding (Fig. 2). In this research, we also apply this assumption. The thickness of the mixture of tailing and soil (H) is calculated as follows: We define soil HM concentration c i,j (i for different periods: i = 0 for before dam failure; i = 1 for after dam failure and before remediation; i = 2 for after remediation; and j for HMs: j = 1 for As, j = 2 for Cd and j = 3 for Pb) and the standard deviation (σc i,j ) of c i,j . For the HM concentration, only the range of HM content is available, so we assume the mid-value as average and the variation from the mid-value as standard deviation. For example, the As content in tailing is 0.6-0.7%, and the assumed As concentration is 0.65 ± 0.05%, which equals 6500 ± 500 mg/kg. Since the concentration of tailings in the Jinding mining area is not available, the concentration and standard deviation in the mixture of tailings are estimated as the average of the content of the tailings (Mcdermott & Sibley, 2000) and the mixed soil (Jin, 2017).

Clean-up process after the accidental release
In the Aznalcóllar dam failure event, the clean-up operation started immediately after the spill and continued until January 1999. The clean-up operation mainly involved the removal of spilt tailings and 4.6 million m 3 of polluted soil (Gallart et al., 1999;Hudson-Edwards et al., 2003). Few further remediation actions are recorded. The clean-up action recovered the HM level to the pre-spill level, but still higher than the pre-mining level (Del Río et al., 2004;Morillo et al., 2005). In the present research scenario, excavation of the mixture of soil and tailing starts right after the dam failure.
Since the excavation can recover the HM concentration in surface soil back to pre-spill level, but not pre-mining level, considering the mining history in the target area, an additional period of the remediation process is applied in this research. Considering the cost-effectiveness, the methodology for remediation is phytoremediation. The target area is agricultural land. Therefore, the remediation target is defined as the Chinese standard of agricultural soil.
Cd and Pb content exceeded this standard, but As content is below the standard level. Therefore, the phytoremediation only targets Cd and Pb, and As is not necessary to be remediated. In the 2nd remediation period, we assume Cd and Pb concentration reduces to the Chinese standard of agricultural soil, and As concentration remains constant. Wan et al. are estimated both before and after the remediation. The reduction of DALY is applied for the cost-benefit analysis for the remediation (2016) conducted a case study of phytoremediation with two species of hyperextractor plants: Sedum alfredii H. and Pteris vittata. The study showed the ability to reduce Cd and Pb in agricultural soil. The remediation duration of phytoremediation is estimated by the extraction ability of the two plant species and the amount of HMs to be extracted. Assuming the extraction rate of the hyper extraction plant is constant, as reported by Chandra et al. (2017): RD j : remediation duration of the jth HM (j = 2 for Cd and j = 3 for Pb). c 2,j : pre-spill level of jth HM in the agricultural soil (mg/kg); c 1,j : soil concentration of jth HM after remediation; c P : concentration of HM in plant tissue (mg/kg); BM: annual plant biomass production (t/m 2 ); D: depth of the remediated HM (m); ρ: soil bulk density (t/m 3 ).
Cd is extracted by Sedum alfredii H.
Pb is extracted by both Pteris vittata. and Sedum alfredii H. Therefore, the final remediation period is the longest period of both HMs, defined as 4 years. In Wan et al. (2016)'s study, the removal efficiency of Cd and Pb for 2 years is 85.8% and 30.4%, respectively. In the present research, the removal efficiency is 94.04% in 3 years for Cd and 57.57% in 4 years for Pb. Wu et al. (2007a) showed that Sedum alfredii H. and Pteris vittata could endure a concentration of over 1000 mg/ kg Cd and Pb in the soil, which is far beyond the soil HM level in both the present research and Wan et al.
(2016)'s study. Therefore, the two species are considered proper for remediation in the present research.
In summary, the remediation strategy for this hypothetical dam failure event consists of two periods. The first period is removing the tailing and soil mixture (26 cm depth) and reducing surface soil concentration to pre-spill level. The second period is phytoremediation, reducing the soil concentration (20 cm depth) to the Chinese standard of agricultural soil (Jin, 2017) (Table 1). The excavated soil and polluted plant biomass produced during the remediation process are transferred to permitted off-sites and disposal facilities for further monitoring or processing to prevent pollutant release to the environment.

Temporal change and mass flow of the scenario
The timeline and mass flow for this scenario are shown in Fig. 3. As a result of the remediation, 99.40% of As, 98.50% of Cd, and 98.24% of Pb are reduced from the flooded agricultural site. All the tailings are supposed to be removed, and additional HMs in the soil that exceeds the Chinese standard of agricultural soil are also extracted from the site.
The mass flow is calculated according to the definition of the scenario. The accidental release is the HM that enters the area with the tailings released. The long-term release is the HM released by mining activities and is accumulated in the target area, which explains the reason for a higher concentration than the background value in the soil before the tailing dam failure. The stock of HM in the soil is the background content of HM from the natural source. Excavation and phytoremediation remove the HM in the polluted area after the tailing dam failure happens. The deposit of HM in the soil is the HM that is not removed and remains in the soil but is at a safer level.
ρ is the density of soil; A is the area of target agricultural land; H is the thickness of the mixture of tailings and flooded soil; D is the depth of soil for phytoremediation; c T is the HM concentration in tailings; Deposit soil =Flow accidental release + Flow long -term release − Flow excavation + Flow phytoremediation c 0 is the original HM concentration in soil; c1 is the HM concentration of soil and tailings mixture; c BG is background value of Yunnan soil; c ST is the Chinese standard of agricultural soil. The value and reference of these parameters are shown in Table 1.

HM exposure estimation
This research considers four main exposure routes: direct soil ingestion, soil particle inhalation, crop consumption, and groundwater consumption (Fig. 2). All these four exposure routes are affected by the HM concentration of surface soil. The exposure dose via direct soil ingestion and soil particle inhalation is directly affected by the HM concentration. The soil HM concentration will influence the concentration in agricultural products and groundwater, further influencing the other two exposure routes.
The heavy metal exposure dose is calculated by the GERAS-1 (Geo-Environmental Risk Assessment System, tier-1) model simulation (Kawabe et al., 2003). GERAS-1 is an onsite screening exposure assessment model developed based on another risk assessment model, CSOIL (Brand et al., 2007). The GERAS-1 model can assess human exposure to HMs and organic matter at the site where the pollutants are directly released. GERAS-1 estimates the pollutant concentration based on the fugacity of pollutants between soil solid particles, soil porewater, and soil pore-air. The concentration in the exposure media (atmosphere, crops, groundwater) is estimated by the transfer process from soil to those Table 1 Parameters applied for scenario definition a: (Mcdermott & Sibley, 2000), b: (Jin, 2017); c: (Cheng et al., 2018); d: (Chen, 2002); e: (Yang et al., 2002); f: (Tuyishimire et al., 2022); g: (Wu et al., 2007b) Parameters Description Values References c T,1 ± σc T,1 As concentration in tailings (mg/kg) 6500 ± 500 a c T,2 ± σc T,2 Cd concentration in tailings (mg/kg) 35 ± 5 a c T,3 ± σc T,3 Pb concentration in tailings (mg/kg) 9000 ± 1000 a c 0,1 ± σc 0,1 Surface soil As concentration before dam failure (mg/kg) 19.62 ± 7.56 b c 0,2 ± σc 0,2 Surface soil Cd concentration before dam failure (mg/kg) 5.03 ± 2.90 b c 0,3 ± σc 0,3 Surface soil Pb concentration before dam failure (mg/kg) 138.97 ± 58.03 b c 1,1 ± σc 1,1 As concentration of soil and tailings mixture right after dam failure (mg/kg) 3259.81 ± 253.78 a,b c 1,2 ± σc 1,2 Cd concentration of soil and tailings mixture right after dam failure (mg/kg) 20.02 ± 3.95 a,b c 1,3 ± σc 1,3 Pb concentration of soil and tailings mixture right after dam failure ( (Brand et al., 2007;Kawabe et al., 2003).The model input requires soil HM concentration, soil parameters, and exposure parameters ( Table 2). The parameters are fixed values for the scenario, and with the input of soil HM concentration (c i,j ), the output of the average daily dose (ADD i,j ) is estimated. i for different periods (i = 0 for before dam failure; i = 1 for after dam failure and before remediation; i = 2 for after remediation) and j for HMs (j = 1 for As, j = 2 for Cd and j = 3 for Pb). Since the standard deviation of the average daily dose (σADD i,j ) is required for DALY calculation, we assume that the daily dose and HM concentration are proportional. Therefore, we have (σc i,j is the standard deviation of c i,j ):   Murray (1994). This research defines DALY as years life lost (YLL) and years life disabled (YLD).
The exposure is not expected to cause a fatal effect. Therefore, in this research, YLL is assumed as zero.
The yearly DALY i,j (i = 0 for before dam failure; i = 2 for after remediation) and j for HMs (j = 1 for As, j = 2 for Cd and j = 3 for Pb) is calculated as: Disease probability is defined by exposure dose and dose-response relationship: x: exposure dose (µg/kg-BW/day); f i,j (x): possibility distribution of exposure dose of the jth HM in the ith period; ф j (x): possibility distribution of dose-response relationship of the jth HM. This research assumes that f i,j (x), and ф j (x) follow a lognormal distribution.
For f i,j (x), σ i,j and μ i,j could be derived from the mean value (ADD i,j ) and standard deviation (σADD i,j ) of exposure dose (Table 3).
For phytoremediation: C pr : cost of phytoremediation; IC pr : the initial cost of phytoremediation; AC pr : annual cost of phytoremediation; RD: remediation duration; A: area of remediated land. The total cost of remediation (C total ) is calculated as follows: In this research, the benefit of remediation consists of three parts: health benefit caused by the reduction of HM exposure (B H ), the benefit of recovering the ecosystem (B E ), and economic benefit caused by the production of agricultural products (B A ) (Wan et al., 2016). These categories of benefits mainly focus on the residents' merit, which is this paper's primary concern. Therefore, the other categories, such as suspension of production and land price, mainly involving mining enterprises and local government, are not included. Among these benefits, B E is a one-time benefit, B H and B A are annual benefits generated with time. Considering the disastrous scenario, the time span for benefits assessment of B H and B A is set as 10 years, a relatively short time span.
For B H : FB: average financial burden of DALY; ΔDALY: reduction of DALY. P: population; The population directly affected by the target agricultural area is defined as the population in the residential area next to the target area (Fig. 1). The population is estimated by the number of households and the average population in each household. ΔDALY is calculated by the total DALY before (i = 1) and after remediation (i = 2): The total benefit is the sum of the three categories of benefit.
The parameters applied for cost-benefit analysis are shown in Table 4.

Sensitivity analysis
In this research, a sensitivity analysis is conducted for the calculation process. The result of this research heavily depends on the definition of the hypothetical dam failure scenario. If the scenario changes, the assumptions and parameters may be influenced, and the result may vary. The main purpose of the sensitivity analysis is to discuss the robustness and the influence of parameter change on the output of the analysis procedure. In this research, a sensitivity analysis is conducted. The method of sensitivity analysis is nominal range sensitivity (Frey & Patil, 2002). The major assumptions in this research are applied to the definition of the tailing release scenario, the scale (population and area) of target land and the hypothesis of the remediation process. Considering these major assumptions, variation of input values for parameters are substituted for important indicators in the present research to discuss the influence of variation in input factors on the result of NPV (Eq. 26). Only one of the chosen parameters changes each time, while the other parameters stay constant (CRC CARE, 2018;Frey & Patil, 2002). This methodology is defined by Walker and Fox-Rushby (2001) as qualitative analysis. By changing each input parameter, the result of NPV also changes, and the variation of the new NPV value compared to the original result is defined as the sensitivity to the corresponding change in the parameter. The extent of variation is set as ±20% and ±40% (Huysegoms et al., 2019).
S tθ : sensitivity of NPV to the change θ in input parameter t; NPV tθ : the result of NPV under change θ in input parameter t; θ: change in input parameters (±20% or 40%); t: number of the changed parameter.
Parameters 1-3 are selected according to the major assumptions in the scenario definition for tailings release. Parameter 1 is the input HM concentration of tailings; parameter 2 is the mixing ratio of tailings and soil; parameter 3 is the release amount of tailings.
Parameters 4-6 are selected according to the definition of the target area and population. Parameter 4 is the area of the target agricultural land; parameter 5 is the Number of residents of the target residential area; parameter 6 is the combination of parameters 4 and 5 because, in the actual situation, the area of the agricultural land and the population influenced by it is expected to change simultaneously.
(35) S t = NPV t ∕NPV Table 4 Parameters applied for cost-benefit analysis a: (FRTR, 2012); b: (Wan et al., 2016); c: (WHO, 2010); d: (Statista, 2022); e: (CRC CARE, 2018); f: (Xie et al., 2003); g: (Weicong et al., 2014)  Average price for wheat ($/kg) 0.48 g Parameters 7 and 8 are selected for the cost-benefit analysis of the remediation process. Parameter 7 is the variation in the soil depth for phytoremediation in the 2nd period of the remediation process, and parameter 8 is the time span for benefits (health benefit and price of agricultural products). Additionally, not only ±20% and ±40%, time spans of 30 years (Huysegoms et al., 2019) and 75 years (life expectancy) are also applied to discuss long-term effectiveness. NPV is estimated for the variation of each parameter, and the variation ratio in NPV of each variation compared to the original result is also calculated and discussed.

HM exposure dose
The exposure dose of HM before remediation is 1.07 × 10 -2 (mg/kg-BW/d) for As; 1.76 × 10 -4 (mg/ kg-BW/d) for Cd; and 5.68 × 10 -3 (mg/kg-BW/d) for Pb (Fig. 4). For As and Cd, groundwater consumption is the most important exposure route, and for Pb, direct soil ingestion is the most important exposure route. This result indicates that the most important action for the resident is to avoid drinking groundwater and avoid contact between soil and food or oral areas, such as wash hands before eating. The exposure dose after the remediation action is 6.43 × 10 -5 (mg/ kg-BW/d) for As; 2.63 × 10 -6 (mg/kg-BW/d) for Cd; and 9.94 × 10 -5 (mg/kg-BW/d) for Pb. The exposure dose after remediation was reduced to 0.60%, 1.50%, and 1.75% of the exposure dose before remediation, respectively. The result indicates that the remediation action effectively reduced the HM exposure of the residents.

Health risk
The result of GERAS-1 model simulation also includes the health risk caused by HM exposure. Before the remediation, compared to the tolerable daily intake (TDI), As exposure dose is 498.3% of the TDI; the exposure dose of Cd is 17.6%, and Pb is 159.0%. GERAS-1 detects all three HMs to show potential health risks. Compared to reference dose (RfD), oral As and Cd intake also show potential risks. For carcinogenic risk, inhalation and oral intake Fig. 4 The simulation result of exposure dose (obtained by GERAS-1) a exposure dose; b percentage of each exposure route of As show potential carcinogenic risk, but inhalation Cd intake shows no carcinogenic risk. RfD and slope factor of Pb are not available in the model, so health risks considering RfD and carcinogenic risk are not assessed for Pb in the result of GERAS-1. Researchers have studied the consequence of tailing dam failure, Akoto et al. (2018) and Buch et al. (2021) reported that tailing dam failure showed a high ecological risk to the adjacent area. Liu et al. (2005) found long-term influence remained in the soil even after an effective clean-up action and a long metabolizing time for the ecosystem. Few studies have discussed the health risk of a potential tailing dam spill compared to the ecological risk. Barcelos et al. (2020) assessed a hypothetical tailing dam spill of a gold mine in Brazil. The result showed a high-risk level caused by oral As and Cd intake. In the present research, the result of the GERAS-1 simulation shows that HM exposure without remediation will pose a high risk to the resident. If the tailings polluted soil is not remediated, the pollutant is supposed to remain in the soil for an extended period (Liu et al., 2015). Therefore, clean-up action is necessary immediately after the dam failure.
For the HM exposure after remediation, Cd and Pb show no risk, but oral As intake still shows a certain level of health risk. The oral intake of As is 21.4% of RfD, and the carcinogenic risk of oral As intake is 9.6 × 10 -5 (carcinogenic risk before remediation is 1.6 × 10 -2 ), which is much lower than the risk level before remediation. Therefore, the remediation action effectively reduces the health risk posed to the residents. In this research, the As concentration is assumed to remain constant in the 2nd period of remediation (phytoremediation) because it already meets the Chinese standard of agricultural soil. However, the Pteris vittata is also able to phytoextract As from soil, so the actual As level in soil and the risk level may be lower.

DALY
The total DALY caused by HM exposure before remediation is 2.63 × 10 -2 years per year of exposure, which exceeds the tolerable DALY level of 1.0 × 10 -6 years of DALY per year defined by WHO (Fig. 5). The DALY of As exposure is the most important, taking 98.51% of the total DALY. Cd is less harmful than the other two metals in this scenario and does not contribute much to the total health impact. After the remediation, the total DALY is reduced to 1.24 × 10 -8 years of DALY per year of exposure. The remediation successfully reduces the DALY to a tolerable level, indicating the remediation is effective. Compared to the research mentioned in the previous chapter that discussed the health risk of residents, the DALY level estimated in the present research reveals the actual expectation of impact on the residents' health, as well as the reduction of the impact due to the remediation process. This reduction of impact (DALY) level is essential for the cost-benefit analysis of the remediation process.

Cost-benefit analysis
The result is shown in Fig. 6. The total cost for remediation is $ 2.88 × 10 7 . Approximately 88.72% of the remediation cost comes from soil excavation. The average cost of remediation in this paper is $ 1.03 × 10 6 /ha, higher than the recovery and cleanup price (CNY 7.5 × 10 5 /ha) evaluated by Liu et al. (2020). This result is also higher than the average cost of soil excavation ($ 7.02 × 10 5 /ha) estimated for agricultural soil in a previous analysis. The difference may be caused by the depth of excavated polluted materials, defined by the amount of tailings released. In this paper, the thickness of excavated soil layer is higher. Also, in this research, remediation involves two periods. The cost of phytoremediation also contributes a part to the total cost.
The total benefit estimated in this research is $ 1.86 × 10 7 , which is lower than the total cost. Health benefit is the most important benefit, taking 98.04% of the total benefit. In Wan et al. (2016)'s research, income loss prevented by remediation is also evaluated, but this category is not included in the present research due to a lack of information. However, the benefit value of income loss prevention is much less than other categories, which is not expected to strongly influence the result.
The analysis result showed a minus value of the net present value of $ −1.02 × 10 7 . The benefit counts for 64.67% of the total cost, and the benefit rate (NPV/total cost) is −35.33%. This result indicates that the remediation action may not be economically feasible for the target area. According to a previous investigation, the 10-year benefit rate of soil excavation under a normal situation is −70.29%. Compared to the normal situation, the health impact of tailing dam failure is higher, so the benefit rate in this paper is also higher. The negative rate is mainly due to the high cost of soil excavation. Researchers that conducted cost-benefit analyses for soil remediation showed different results. Some studies showed beneficial remediation actions for industrial areas (Lavee, 2012) or river sediments (Zheng et al., 2019). Some studies showed negative NPV values for industrial soil remediation (Huysegoms et al., 2019). The reason for these differences is the variation in cost/benefit considerations and scenario settings. Nonetheless, all these studies are under normal scenarios instead of disastrous ones. Studies are unavailable on cost-benefit analysis of soil remediation for tailing dam failures. In the present research, we discuss the economic feasibility of soil remediation for tailing dam failures, and the result shows that remediation of agricultural soil under a tailing dam failure event may not be economically beneficial.
In this case, however, the negative NPV is reasonable. The consequence of a tailing dam failure is disastrous, acute, and much more severe than long-term pollution. It is reasonable that a tailing dam failure will require a large amount of effort and budget to be repaired. Additionally, a negative NPV does not mean it is better not to conduct the remediation. Since the impact of a tailing dam failure is exceptionally high, clean-up action immediately after the failure is necessary. To avoid the economically infeasible remediation action, it will be better for the local government to put stricter regulation and supervision on the tailing dam management, preventing the failure before it happens. Additionally, since the remediation is economically infeasible, the local government should pay attention to the emergency budget and consider the fine on the mining industry under dam failure scenarios. Fig. 6 Result of cost-benefit analysis. Soil excavation takes up most of the cost, and health benefit takes up most of the benefit. The total cost exceeds the total benefit, so the net present value (NPV) is negative. The NPV bar shows the scale of the negative NPV value It is worth noticing that the benefit of health risk reduction and cash crops is time-dependent. The time span considered in this research is 10 years. If the time span increases, the benefit may exceed the cost. The time span will depend on the consideration of the local government in practice. The influence of time span is discussed in the sensitivity analysis.

Sensitivity analysis
Parameters selected for sensitivity analysis are selected according to the major assumptions in the scenario definition and analysis process. The variation in the tailing release scenario, the scale of the target area, and the cost-benefit analysis are discussed for their impact on the final result (NPV value). The result of the sensitivity analysis (S tθ ) is shown in Fig. 7. Since the original NPV is a minus value, a variation rate below 100% is an increase in the NPV value. Therefore, among all the input parameters, HM concentration, Number of residents, and the time span positively correlate with the result, while the other five parameters negatively correlate with the result.
The release amount of tailings (parameter 3) heavily influence the result, indicating tailings release is a critical factor in a dam failure event. The area of the target agricultural land (parameter 4) also has relatively more influence on the result. However, changing resident numbers with the area (parameter 6) will reduce the influence of the scale of the target area on the final result. The mixing ratio of tailings (parameter 2) and soil depth for phytoremediation (parameter 7) has minimum influence on the result. In parameter 2, the reason is that the mixing ratio has little influence on the concentration of HM concentration in the mixture of soil and tailings. In parameter 7, the cost of phytoremediation is much less than soil excavation, so changes in the remediation duration of phytoremediation have a limited influence on the total cost. Among all the parameters, with a 40% change in the input parameter, parameters 3 and 4 output a positive NPV value, which questions the robustness of this research. For parameter 4, as already discussed, changing the polluted area or the affected population and keeping the other constant is not very reasonable. When the area and population change together, the change in output is not very big. For parameter 3, when the release amount of the tailings decreases by 40%, the NPV becomes positive, meaning the benefit due to health impact reduction exceeds the remediation cost. Even though the NPV value is positive, the value is barely above zero ($ 4.57 × 10 4 ) compared to the cost and benefit ($ 1.86 × 10 7 ). −40% is close to the "switching value" of the release amount of the tailings, the value which turns the NPV to zero. Therefore the robustness of the result of this research is acceptably within the variation range of ±40%. Fig. 7 Result of sensitivity analysis (S tθ ). The y-axis shows the variation of output compared to the original result. The Numbers 1-8 on the x-axis represent the 8 parameters. Parameter 1 is the input HM concentration of tailings; parameter 2 is the mixing ratio of tailings and soil; parameter 3 is the release amount of tailings; parameter 4 is the area of the target agricul-tural land; parameter 5 is the number of residents of the target residential area; parameter 6 is the combination of parameter 4 and 5; parameter 7 is variation in the soil depth for phytoremediation; parameter 8 is the time span (only change of 20% and 40% is shown in the figure, variations of 30 years and 75 years are excluded.) Considering the time span, changing the time span in the ±40% range does not produce a positive value, but the NPV value for a 30-year and 75-year time span produces positive NPVs ($ 2.70 × 10 7 and $ 1.11 × 10 8 , respectively). Health benefit, as well as agricultural products, is a yearly benefit. Therefore, it is highly time-dependent. This research selected the time span for a disaster event as a short time span (10 years). The analysis parameter may be adjusted in practice according to the requirement of the local policymaker, and since the time span can influence the result heavily, this parameter should be paid attention to.

Conclusion
In this research, we develop a hypothetical tailing dam failure scenario for the Jinding Pb/Zn mine and estimate the health impact of the HM exposure and the economic feasibility of the corresponding remediation process. The health impact is high after the tailing dam failure, and remediation action is necessary. However, the remediation process may not be economically feasible. Because remediation is necessary after tailing dam failure, local governments should put more effort into tailing dam management to avoid the unbeneficial remediation process. The result of this research provides information on the cost-effectiveness of soil remediation under disastrous scenarios, which was rarely discussed in previous studies. The methodology applied in this research has relatively low site dependence. Sensitivity analysis has proved certain robustness of the analysis approach, indicating the applicability of the methodology to different tailing dam sites.
The major limitation of this research is the high dependence on assumptions. This limitation, however, is difficult to avoid. For predictive studies on hypothetical tailing dam failure events, the failure was not yet occurred. Therefore, little accurate information about the actual failure event can be available. To improve the reliability of this type of study, simulation models of tailing dam failure with wide applicability and high accuracy are required. With improvements in scenario definition and data availability, a major improvement in the result of predictive studies is expected in the future.