Hurdle model selection.

= Poisson, Bin = Binomial, Negbin = Negative Binomial). The Zero part of each model predicts the probability of recruitment, while the Count part predicts the number of recruits given Zero≠0. Significant P-values are indicated by bold formatting. The interactions between Month and Sori (both new and old) were not significant in any of the models. Because it had the lowest AIC value, the best model was HNb3a. Models with different error distributions were tested (Pois

其中，Poisson指代泊松分布（Poisson）、Bin指代二项分布（Binomial）、Negbin指代负二项分布（Negative Binomial）。每个模型的零值部分用于预测招募概率，而计数部分则在零值条件不成立（即Zero≠0）时预测招募个体数。具有统计学显著性的P值（P-value）以粗体格式标注。所有模型中，月份（Month）与孢囊堆（Sori）的交互作用均无统计学显著性。由于该模型的赤池信息准则（Akaike Information Criterion，AIC）值最低，因此最优模型为HNb3a。本次研究测试了采用不同误差分布的模型（原文此处仅保留Pois，未完整给出）。