Abstract Environmental conditions are important drivers in regulating the distribution pattern of phytoplankton composition in the world's oceans. We constructed models that predict pico-, nano- and micro-phytoplankton size classes and assessed the impact of separately including sea surface temperature (SST) and estimates of light level in the surface mixed-layer on model skill. The empirical models were trained using size classes estimated by chemotaxonomic analysis of in situ high performance liquid chromatography (HPLC) pigments and environmental data originating from the Atlantic Ocean. As the accuracy of transforming pigment data into quantitative size classes is crucial when constructing phytoplankton size composition (PSC) models, we also quantified the resulting differences of our and several existing PSC models when using class sizes derived from HPLC pigments by two common chemotaxonomic methods, CHEMTAX and Diagnostic Pigments (DP). Addition of the environmental variables to abundance-based models using our approach improved the skill of correctly predicting PSC, reducing the root mean square difference (RMSD) by 10 to 20% in the best cases. Addition of SST yielded the highest percentage decreases, on average, for all three size classes, with greatest improvement in microplankton and nanoplankton fractions. These models performed equal to or better than several existing abundance-based models. The improvements in model predictions, however, could be obscured by the choice of pigment method used to generate the initial PSC data set. Insufficient data is available to assess whether CHEMTAX or DP is the more appropriate chemotaxonomic method to employ when estimating PSC. Further collection and analysis of additional water samples for phytoplankton taxa and size by microscopic methods - including traditional microscopic cell counts and automated methods - and HPLC pigment data are required to answer this question.

中文翻译：

---

将环境数据纳入基于丰度的算法中，以推导出大西洋中的浮游植物大小类别

摘要 环境条件是调节世界海洋浮游植物组成分布格局的重要驱动因素。我们构建了预测微型、纳米和微型浮游植物大小类别的模型，并评估了分别包括海面温度 (SST) 和表层混合层中光照水平估计对模型技能的影响。使用通过原位高效液相色谱 (HPLC) 色素的化学分类学分析和源自大西洋的环境数据估计的大小类别训练经验模型。由于在构建浮游植物大小组成 (PSC) 模型时，将色素数据转换为定量大小类别的准确性至关重要，我们还通过两种常见的化学分类方法，CHEMTAX 和诊断颜料 (DP)，量化了我们和几个现有 PSC 模型在使用来自 HPLC 颜料的类大小时产生的差异。使用我们的方法将环境变量添加到基于丰度的模型中，提高了正确预测 PSC 的技能，在最佳情况下将均方根差 (RMSD) 降低了 10% 到 20%。平均而言，对于所有三个尺寸等级，添加 SST 产生的百分比下降幅度最大，微型浮游生物和纳米浮游生物部分的改善最大。这些模型的性能等于或优于几个现有的基于丰度的模型。然而，模型预测的改进可能会被用于生成初始 PSC 数据集的颜料方法的选择所掩盖。没有足够的数据来评估 CHEMTAX 或 DP 是否是估计 PSC 时更合适的化学分类方法。需要通过显微方法（包括传统的显微细胞计数和自动化方法）和 HPLC 色素数据进一步收集和分析浮游植物分类群和大小的额外水样，以回答这个问题。