Towards more effective spatio-temporal schemes for remediation of agricultural soils in response to large-scale contamination with radionuclides

Although nuclear energy can be categorized as a safe and low-carbon energy source, two major accidents (Chornobyl, 1986, and Fukushima, 2011) demonstrated the large-scale impact it may have. While immediate response plans to nuclear accidents are well established, the complex and resource-intensive recovery phase lacks adequate (inter)national guidance and tools. The recovery is especially important for agricultural areas, where food production needs to be returned to normalcy. This research contributes to the development of tools within a spatial decision support system (sDSS) to improve the post-accident decision-making process about where, how, and when to remediate agricultural land. This research illustrates the tools in two agricultural watersheds: the Maarkebeek, located in Flanders (Belgium), and the Niida watershed, located in the Fukushima prefecture (Japan). At the basis of the sDSS was a multi-criteria decision-aiding methodology (MCDA), which was found to be a good framework for supporting soil remediation on a local scale through optimal remedial technology selection for a specific site and on a regional scale through optimal site selection. However, we identified four major gaps in the reviewed case studies: (i) A lack of inclusion of social criteria, (ii) A lack of early stakeholder engagement, (iii) A mismatch between weighting and aggregation methodologies, and (iv) A lack of sensitivity analysis. To overcome (i) and (iv), a methodology to include linguistic scoring within ordinal qualitative scales as fuzzy numbers, for the evaluation of criteria or weight setting was proposed. The impact of an increasing amount of uncertainty within the triangular fuzzy numbers (TFN) was studied, and it became clear that the impact of uncertainty on decision-making power cannot be ignored. Proposing remedial plans using two compromise programming methodologies for site selection and action selection results in a very complex spatial pattern for remediation. The clustering of individual entities into larger, homogeneous, actionable units can improve feasibility and reduce the cost of remediation, especially when interventions are costly and technically challenging to perform. A spatio-temporal clustering approach under a budget constraint was presented to determine homogenous clusters of polygons and interventions to reduce the cost of intervention while still attaining near-optimal effectiveness. When remedial actions are converted into remediation trajectories, a sequence of multiple remedial actions in time is explicitly considered within the decision-making. A column generation model was proposed that reconciles the economic and productive aspects of remediation for agricultural areas. The model's main novelty is its ability to find optimal multi-period sequences or trajectories of remedial actions from a set of predefined remedial actions, where column generation is used to extend the candidate set of the remedial trajectories while keeping the computational complexity limited. It was found that the introduction of new trajectories with the column generation approach, after 4700 iterations, reduced the loss in productive years by 237 years and the accumulated productivity score by 20%. An iterative framework called the CAMF approach (Cellular Automata-Based Heuristic for Minimizing Flow) was used to model and reduce the off-site impacts of soil remediation in the Niida watershed. Spatially targeted afforestation of the 1000 best cells reduces the sediment quantity resulting from the decontamination work by 15.4%, compared to 0.8% when the cells are selected randomly. Prioritizing topsoil removal interventions within the Maarkebeek watershed while accounting exclusively for off-site criteria results in a 267% increase in Cs-137 flux reduction efficiency in the first year when compared to remedial schemes based only on on-site criteria.