Het kwantificeren van de effecten van bodem- en plantkenmerken op de opname van radiocesium: een wereldwijd perspectief

In case of nuclear accidents, radiocaesium (134Cs and 137Cs) can be released into the environment, contaminating agricultural soils and entering the food chain via plant uptake. This is because Cs exhibits similar biochemical behavior to the plant nutrient potassium (K). Radiocaesium is strongly adsorbed onto the weathered edges of micaceous clay minerals, such as illite, and this limits its transfer to the food chain. The soil-to-plant transfer of radiocaesium is expressed as the concentration ratio (CR), i.e., the ratio of the radiocaesium concentration in the plant material to that in the soil, and that CR value is key in environmental impact assessments after nuclear accidents. After the Chornobyl and Fukushima accidents, many CR observations were gathered and models were developed to predict the CR of radiocaesium in soils. However, these models used for impact assessments were based on data from the affected regions with limited diversity in soil characteristics. It has never been verified whether these models can be used to predict soil radiocaesium bioavailability on a global scale. Nuclear energy is expanding in the world and it is logical that emergency planning requires information on radiocaesium soil-to-plant transfer. This research was set up to explore and understand the soil properties affecting radiocaesium bioavailability globally. The central hypothesis was that radiocaesium sorption is affected by the clay mineralogy in soils and that information on clay mineralogy can be linked with other globally available soil information to improve predictions of radiocaesium bioavailability on a global scale. This study started by collecting existing information on radiocaesium CRs from the literature in a systematic way, creating a comprehensive and harmonized database with 8,837 records from 142 studies. It showed that CRs ranged over seven orders of magnitude between 2.8 10-5 to 3.8 102 kg kg-1. The highest CRs were observed in soils originating from tropical climates and organic soils, and the lowest in soils temperate climates. However, data from tropical (N = 404) and arid (N = 47) regions were limited. Statistical analyses revealed that random variation of the studies, soil properties, and plant species were the main factors influencing CR. When comparing models, semi-mechanistic models outperformed machine learning models on small datasets (N <200), while machine learning models were superior with larger datasets. A global radiocaesium transfer map was created with a machine learning model and global soil data, identifying highly weathered and organic soils as the most vulnerable. However, the map was based on a limited dataset (N = 513) and the residual error showed that the 95% confidence intervals of CRs were still larger than two orders of magnitude. The key missing variables in the global soil maps are the soil exchangeable K content and the soil mineralogy. In a second step, it was experimentally investigated to what extent clay mineralogy affects the bioavailability of 137Cs. This was first explored with plant substrates based on a mixture of pure sand with ten different clay minerals. The clays were mixed with sand in single and double clay doses, spiked with 137Cs, and grown with grass in a greenhouse experiment. The CR generally decreased with increasing clay dose, but the CRs varied two orders of magnitude depending on the type of clay and exchangeable K content. The CR was highest for clays with little 137Cs specific sites, such as expanding clay (smectite) and where the exchangeable K content was low compared to the other clays. The CR was well predicted from the soil solution 137Cs and K concentrations (R2 = 0.72 for log transformed CRs), corroborating earlier findings in natural soils. The CR (log-transformed) was statistically unrelated to total clay mineral content or 1:1 and 2:1:1 type clay mineral content, while it significantly decreased with increasing 2:1 clay mineral content (R2 = 0.32). A multiple regression model with four different X-ray diffraction (XRD) based clay mineral groups yielded the strongest predictive power (R2 = 0.79). We concluded that XRD quantification is valuable for describing 137Cs bioavailability in clay-sand mixtures with contrasting properties. The follow-up experimental study identified if similar trends of mineralogy are true in natural soil. The 137Cs bioavailability was studied in soils from contrasting parent materials and weathering stages. Two soil sequences from Kenya (N = 6) and the Philippines (N = 6) were sampled, representing different weathering stages affecting soil mineralogy. In addition, other soils (N = 26) with contrasting soil properties, including Andosols and Ferralsols, were included in the collection. A greenhouse experiment was set up with these soils. A subset of the soils was fertilized with K to investigate its effect on 137Cs CR reduction. The soils were characterized, including mineralogy by XRD, spiked with 137Cs and grown with grass. The CRs varied four orders of magnitude among soils. It was highest in tropical Ferralsols (47 kg kg-1) due to the low 137Cs interception potential of kaolinite clay and the low exchangeable potassium in these soils. The semi-mechanistic 137Cs transfer model, mainly calibrated to temperate soils, poorly predicted the underlying processes in highly weathered tropical soils but, due to compensatory effects of the 137Cs adsorption and available K in soils, still reasonably well predicted the 137Cs bioavailability (i.e., 10log(CR)) across all soils (R2 = 0.78). Finally, experimental work was devoted to 137Cs adsorption (not bioavailability) in soils as affected by the clay mineralogy in soils of contrasting weathering stages. The clay and silt fractions were isolated from 24 soils from contrasting weathering stages. The 137Cs adsorption was measured as the Radiocaesium Interception Potential (RIP), for all bulk soil samples and in their corresponding clay (0-2 µm) and silt (2-50 μm) fractions, along with XRD analysis of the mineralogy of these fractions. Results showed that the RIP in these soils was unrelated to the soil clay content (p >0.05), well in contrast with the general concept of models that propose that the clay fraction in the soil is key to predicting a CR value. Instead, soil mineralogy, particularly illite content, explained the soil RIP. In some soils, a significant portion of the RIP was found in the silt fraction (>50%). Moreover, the RIP of the isolated clay fraction could be larger than the RIP of soil, likely due to the clay isolation procedure, which released K from clay mineral interlayers, increasing frayed edge sites, and caused disaggregation of clay minerals, increasing the clay content. This showed that studying isolated components of soil does not necessarily represent intact soils in the field. Overall, this research emphasizes the importance of considering soil mineralogy when predicting radiocaesium bioavailability, thereby enhancing the effectiveness of radiocaesium impact assessments on a global scale. This work showed that, besides peat soils, highly weathered soils are most vulnerable to radiocaesium transfer with CR values exceeding 10 kg kg-1, i.e., plants concentrate radiocaesium from such soils. This also means that emergency planning around new nuclear installations in tropical regions, often dominated by highly weathered soils, requires specific attention to food chain risks. Collating all RIP data showed that soil mineralogy greatly improved RIP predictions compared to predictions based on standard soil properties such as clay and organic matter content. However, it was demonstrated that the standard soil properties, complemented with soil classification according to the World Reference Base (WRB) are almost as strong predictors as soil mineralogy. The effects of soil structure and mineralogy on radiocaesium bioavailability in field conditions need further study, particularly in tropical regions where the long-term fate of radiocaesium remains largely unknown.