Factors controlling the variability of ¹³⁷Cs concentrations in 5 coastal rivers around Fukushima Dai-ichi power plant

Highlights: • Identifying the key variables controlling ¹³⁷Cs concentrations in rivers. • Meta-analysis of published ¹³⁷Cs concentrations in rivers around Fukushima. • Catchment and hydrological descriptors to explain catchment response variability. • Different catchment descriptors for each form of ¹³⁷Cs (soluble, solid, total). - Abstract: The Fukushima Dai-ichi Nuclear Power Plant (FDNPP) accident led to the contamination by radiocesium (¹³⁷Cs) of large drained areas. Cesium-137 concentrations in rivers result from complex transfer processes, depending on multiple forcings. Better knowledge of the factors controlling these concentrations is therefore a prerequisite to improve predictions of ¹³⁷Cs transfers within river catchments. This study aimed at analyzing the spatial and temporal variability of ¹³⁷Cs concentrations in rivers and identifying the key factors controlling their variability. Published values of ¹³⁷Cs concentrations in rivers in the north of FDNPP were collected, characterizing 122 sampling sites from May 2011 to October 2014. It resulted in three datasets: dissolved concentrations C_W (Bq/L), concentrations in suspended sediment C_SS (Bq/kg) and total concentrations C_T (Bq/L). The resulting database reflected a large variety of catchments and hydrological conditions. Observed ¹³⁷Cs concentrations varied by 2–4 orders of magnitude and were poorly explained (R² = 0.13–0.38) by the average contamination density. Indices summarizing the complex spatial and temporal properties of the catchments were proposed as candidate explanatory variables of concentrations in rivers. They were selected by stepwise regression for each dataset (C_W, C_SS, C_T). For the three datasets, the selection and combination of 5–10 indices significantly better explained this variability (R² = 0.69–0.83). Deposit indices were identified as first drivers of concentrations in rivers. A deposit index was selected for each dataset, indicating no effect of the contamination distribution for C_W, whereas C_T and C_SS required considering the distribution of contamination and connectivity, as well as the presence of dams for C_SS. The others selected variables significantly contributed to explain the concentration variability. This meta-analysis emphasizes the importance of structural (e.g. slope, land-cover) and functional (e.g. delay, season, rainfall) properties in the dissimilarities of catchments responses, stressing that assessments could be improved by including more these properties in models.