A Multicriteria-based Integrated Framework for Sustainable Assessment of Contaminated Site Management Options


 Contaminated site management is a multiple objective decision-making task that generally involves several factors, such as performance of technology, environmental effects, cost, and social influence. However, the decision on contaminated site management in China have been principally driven by practical factors such as cost and time. In this study, we adopted a sustainable assessment method and developed a multicriteria-based integrated framework that satisfy the requirement of green and sustainable development of contaminated site. We integrate remediation sustainable assessment and redevelopment sustainable assessment in one framework, and allows the optimization of indicators. The framework started with definition of site management type, then investigating site characterization, screening indicators, quantification of indicator, selecting assessment model, selecting primary options, assessment with uncertain analysis, and determine preferred options. To demonstrate the utility of the framework, results are presented in a contaminate site in southwest China for two management type, site remediation and site redevelopment. We used different approaches to evaluate the stability and robustness of assessment results, including Monte Carlo simulation, scenario analysis and sensitivity analysis. The demonstration showed that attention has to be paid to the proper description of the site, the principles of the procedure and the decision criteria.

time. In this study, we adopted a sustainable assessment method and developed a 23 multicriteria-based integrated framework that satisfy the requirement of green and 24 sustainable development of contaminated site. We integrate remediation sustainable 25 assessment and redevelopment sustainable assessment in one framework, and allows the 26 optimization of indicators. The framework started with definition of site management type, 27 then investigating site characterization, screening indicators, quantification of indicator, 28 selecting assessment model, selecting primary options, assessment with uncertain 29 analysis, and determine preferred options. To demonstrate the utility of the framework, 30 results are presented in a contaminate site in southwest China for two management type, 31 site remediation and site redevelopment. We used different approaches to evaluate the 32 stability and robustness of assessment results, including Monte Carlo simulation, scenario 33 analysis and sensitivity analysis. The demonstration showed that attention has to be paid 34 to the proper description of the site, the principles of the procedure and the decision criteria. 35

36
The arable land of China is less than half of the world average, the country simply 37 cannot afford to lose any more available land due to increasing problems with pollution 38 both qualitative data and quantitative data. While using subjective data will introduces 77 uncertainty of assessment (Sam et al., 2017). In fact, data of contaminated sites is often 78 limited, particularly information that associated with social impacts. While multicriteria-79 based sustainable assessment has been adopted in prioritization of remediation 80 techniques and site redevelopment options in other countries (Sorvari and   The first step is to define which type of management need to be decision-marking. 108

Material and methods
This step is of major importance in terms of subsequent analysis, since different 109 management problem will influence boundaries of indicator set building (section 2.3) and 110 indicator quantification (section 2.4). In this study, the framework was proposed for three 111 commonly site management type: site prioritization, sustainable remediation assessment, 112 sustainable redevelopment assessment. 113

114
The next step was to investigate site characterization to obtain basic data of site (step 115 2) by environment investigation and social investigation. Data include measurement of 116 concentrations of potential concern chemicals (i.e. heavy metal, petroleum hydrocarbon), 117 sensitivity receptors (i.e. farmland, communities, river, and residential), former activities (i.e. 118 production process), geographic information (i.e. coordinate, distance, terrain) and 119 hydrogeology information (i.e. aquifer, rainfall). These data provide the necessary 120 information for subsequent indicator scoring (section 2.5.1). 121

122
The sixth step of the proposed framework is choosing potential management options 7 applicable for a specific site for the final decision. For a type of site management, there are 124 often a wide variety of option can be applied. For example, for site remediation, there are 125 more than twenty technologies can be used (e.g., In-situ bio-ventilation, monitored natural 126 attenuation, soil washing, thermal desorption, and stabilization/solidification), which does 127 not include hybrid approach. Among the whole range of existing options, the application of 128 certain options may not be feasible due to the specific characteristics of the site-specific 129 circumstances. Therefore, options that are inapplicable or unworkable for the case under 130 study should be identified in this step and screened out before next process.

Select appropriate assessment model 169
How to handle multidimensional data mentioned above into a comprehensive index to 170 assess the degree to which option is global reasonable is the task of this step. Such 171 decision-making problem is usually solved by assessment model, or MCDA. Many MCDA 172 exist to aid decision-making. According to compensation degree, including compensatory 173 method, non-compensatory method, and partially compensatory method. In compensatory 174 method, the effects of different indicators can accept each other, while non-compensatory 175 method is on the contrary. Partially compensatory method meaning that some 176 compensation is accepted between the different decision criteria but a minimum level of 177 performance is required from each of them. Therefore, assessment model selection was 178 arbitrary, mainly based on purpose, needs, and stakeholder preferences. 179

Assessment with uncertain analysis
180 Due to lack of knowledge and natural variability, it is almost impossible to measure the 181 effects exactly of the different management options on receptors, that is creates uncertainty. 182 The former type of uncertainty is epistemic uncertainty, while the latter is type of aleatory 183 uncertainty. Uncertainty in MCDA can have a significant effect on rankings, and mislead 184 decision makers. Especially in the field of site management, human subjectivity/preference 185 can result in obvious uncertainty. In site management decision-marking, uncertainty mainly 186 comes from the indicator scoring and weight assignment. To understand the accuracy and 187 reliability of results, it is essential to evaluate the effect of indicator variability and weight 188 sensitivity on the final output. Some commonly used uncertain evaluation methods including stochastic simulation, sensitivity analyses, and scenario analysis. 190

Determine preferred options 191
The preference of each management option is guided by a total score derived by 192 assessment process, which often the higher the score the better the option (higher 193 preference). It is notable that the assessment with uncertain analysis will produce 194 probabilistic-based output. In other words, the output does not give a certain result, but 195 give a suggestion.  Table S1. 217 218   Table S1 219 Description of four common remediation technology for heavy metal contaminated site.  Table S3, 247 which aggregated indicator scores with weights to provide a final value for each option. 257 where RSI is sustainable index, ei is score of ith sub-indicator e of environmental, WE 259 is weight of environmental, sj is score of jth sub-indicator s of social, WS is weight of social, 260 cbk is kth score of sub-indicator cb of economic, WCB is weight of economic. That is to say the environmental weight should be largest, and the economic weight should 285 be smallest. In such case, the weights of environmental, social, economic were assigned 286 0.5, 0.3, 0.2 (unequal-weighted). 287   (Table  307 S6, Table S7, Table S8, and Table S9). 308 309 Table S6. Indicator scores of commercial land use. 310 Table S7. Indicator scores of landscape land use. 311 Table S8. Indicator scores of residential land use. 312 Table S9. Indicator scores of industrial land use.  Table S10. which aggregated indicator scores with weights to provide a final value for each option. 327

Implications for site redevelopment options assessment
where Ii is the ith indicator, Wi is the ith indicator weight.   Table  344 1. In the process of indicator scoring, the remediation technologies were assigned to high washing, the RSIs in unequal-weighted scenarios were lower than in equal-weighted 377 scenarios. This is because these two potential options have relative higher score in 378 economic factors, so the weight of economic factors is reduced, will leading to a decrease 379 in the overall RSI value. In the low-concentration scenario, the environmental indicator 380 scores of the four potential options were all higher than social and economic, so the RSIs 381 of the four potential options in unequal-weighted scenarios were all higher than in equal-382 weighted scenarios. 383 with 7 weight scenarios. The prioritization of potential option for redevelopment with 7 403 scenarios was some differences. When soil quality was sensitivity indicator with high 404 weight (W≥0.5), landscape was considered as the first option. When funding source was 405 sensitivity indicator with high weight (W>0.5), the highest priority land-use type turned out to be industry land-use. In addition to these two situations, the option ranked first was 407 residence land-use type. It is noted that, in situation where remediation proportion is 408 sensitivity indicator, all scenarios led to the same ranking order. 409 410   Table S11  411 The sustainable redevelopment index under different key factors with 7 weight scenarios. 412 scoring and weighting process will inevitable introduce uncertain for sustainable 444 assessment. The Monte-Carlo simulation is proved to be an effective tool to evaluate 445 uncertainty caused by indicator scoring. The most sustainable option was determined not 446 by one assessment index but by probability derived from lots of simulations. Such 447 assessment will have more robustness. 448 It is well known that the weights reflect each person's individual values and attitudes, 449 personal and professional history, education, cultural background, knowledge level, the stakeholder group he/she represents etc. However, in this study, though the weights have 451 changed, in most scenario the ranking order of management options remains unchanged. 452 This is not agreement with previous studies that different people varied considerably 453 resulting in different preference management options. In this sense, the multicriteria-based 454 integrated framework for prioritization management options, described in this study, can be 455 a useful tool that reduce uncertainty in decision making. The framework is particularly 456 useful if one potential option has advantage in key factors. 457

458
It is noted that, we use compensatory assessment methods in two cases, which allow 459 different factor's impact can be tradeoffs each other. However, in reality, the 460 decisionmakers might be unwilling to accept such tradeoffs. In such situation, it is 461 necessary to study the applicability of other assessment model to solve our study problem. 462 In addition, with the increase of repair evaluation cases, the uncertainty of relevant 463 parameters of the model will gradually decrease. In the future, the indicators with high 464 uncertainty should be refined to reduce the uncertainty.     Figure 1 Proposed framework for sustainable assessment of contaminated site management options The RSI value distribution of four alternatives under two weight scenarios (A: equal-weighted; B: unequalweighted).

Figure 3
Probabilities of best potential option under different concentration scenarios and different weight scenarios.

Figure 4
Sensitivity analysis of the ranking of potential options under 7 weight scenarios.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.