Predicting soil Radiocaesium uptake and dynamics using Mid-Infrared Spectroscopy (MIRS)

The Radiocaesium (¹³⁷Cs) released as a result of past and potential nuclear accidents is of great concern for agriculture because of the relatively long half-life and easy absorption by plants. A countermeasure is the use of potassium fertilizers, but the relationship between transferability from soil to crop and exchangeable potassium (Ex K) varies depending on the soil. Previous studies have suggested that exchangeable ¹³⁷Cs (Ex ¹³⁷Cs) and the solid/liquid distribution coefficient (Kd) can be important factors explaining the variability of the ¹³⁷Cs dynamic in soil. However, the methods to measure these soil parameters are not suitable for adequate emergency deployment and preparedness because they are expensive and time consuming.Mid-infrared spectroscopy (MIRS) has been shown to predict soil parameters more quickly and cost-effectively. However, the prediction of Cs parameters (Ex ¹³⁷Cs, Kd) using MIRS has not yet been evaluated. This study aims to evaluate whether MIRS can predict Cs-related parameters such as Kd, Ex 137Cs/Total 137Cs, and other parameters that may influence on the behavior of 137Cs in the soil, such as total carbon (soil total C).The 1,682 samples were collected in Fukushima Prefecture, Japan from 2013 to 2020, most of them are Andosols. Each soil property (soil total C, Ex 137Cs/Total 137Cs ) was obtained from the monitoring data of MAFF and NARO. As for Kd for 133Cs, 176 samples were measured at Hokkaido University using ICP-MS. Each sample was dried at 37°C for at least one night and then sieved to less than 0.2 mm before measurement. The spectra data was obtained with four replicates in each soil sample using MIRS. To date, a total of 1,419 samples have been analyzed and their properties predicted using partial least squares regression (PLSR).The PLSR models provided a relatively high accuracy for the prediction of soil total C, where the R2 (coefficient of determination) was 0.9±0.01. However, low accuracy was observed in the prediction of Kd for 133Cs, where the R2 was 0.46±0.06, which was attributed to the low concentration of the data and the limited number of samples. On the other hand, the result of Ex 137Cs/Total 137Cs showed relatively higher value then Kd for 133Cs, where the R2 was 0.56±0.04. Variable importance on projection Ex 137Cs/ Total 137Cs suggests that wavelength range related to clay mineralogy, quartz, and organic matter are influential in the estimation of Ex 137Cs/Total 137Cs. To further evaluate the potential for MIRS for rapid and affordable prediction of Radiocaesium risks, additional samples will be analyzed to improve the representativeness of the model and broaden the dataset, and more modeling techniques (Cubist, Support Vector Machine, etc) will be applied.