Models to be compared.

The radiocesium contamination caused by the Fukushima Daiichi Nuclear Power Plant accident has made it difficult to use coppice woods as bed logs for mushroom cultivation. Evaluating the variability in the radiocesium activity concentration of logs is necessary in order to predict how many coppice woodlands are available for producing mushroom bed logs. To clarify the variability in radiocesium activity concentrations and to estimate the sample size required to estimate these concentrations with sufficient accuracy, we modeled the log-transformed radiocesium activity concentrations in growing shoots of hardwoods. We designed two models: (1) a model with mean concentrations that varied among stands with a standard deviation that was the same among stands, and (2) a model with varying means and standard deviations. We fit the data pertaining to only Quercus serrata to both models and calculated the widely applicable information criterion values. Consequently, we adopted the simpler model (1). Applying the selected model to data for all species, we examined the relationship between the number of measurement individuals and the predictive distribution of the expected concentration. Based on previous recommendations and measurement costs, we proposed that five individuals would be appropriate for estimating radiocesium activity concentration in a stand.福岛第一核电站（Fukushima Daiichi Nuclear Power Plant）事故引发的放射性铯（radiocesium）污染，使得矮林木材难以作为食用菌栽培段木使用。为预测可用于生产食用菌栽培段木的矮林林地规模，评估原木放射性铯活度浓度（radiocesium activity concentration）的变异性实属必要。为阐明阔叶树新生枝条中放射性铯活度浓度的变异性，并估算以足够精度估算该浓度所需的样本量（sample size），我们针对阔叶树新生枝条的对数转换（log-transformed）放射性铯活度浓度构建了统计模型。我们设计了两类模型：（1）林分间均值存在差异、但林分间标准差保持一致的模型；（2）林分间均值与标准差均存在差异的模型。我们仅采用枹栎（Quercus serrata）的相关数据对两种模型进行拟合，并计算了广泛适用信息准则（widely applicable information criterion）值，最终选取了更为简洁的模型（1）。将选定的模型应用于所有物种的数据集后，我们分析了测量个体数量与预期浓度预测分布（predictive distribution）之间的关系。结合既往研究建议与测量成本，我们提出以5个个体作为单林分放射性铯活度浓度估算的适宜样本量。