Hou Deyi, Rabinovici Sharyl J M, Boehm Alexandria B
Environmental Water Studies, Department of Civil and Environmental Engineering, Stanford University, Stanford, California 94305-4020, USA.
Environ Sci Technol. 2006 Mar 15;40(6):1737-43. doi: 10.1021/es0515250.
Beach health advisories are issued if enterococci (ENT) densities exceed the 30-d geometric mean or single-sample water quality criteria. Current ENT enumeration procedures require 1 day of incubation; therefore, beach managers make policy decisions using 1-day-old data. This is tantamount to using a model that assumes ENT density on day t is equal to ENT density on day t-1. Research has shown that ENT densities vary over time scales shorterthan a day, calling into question the usefulness of the current model for decision-making. We created Dynamic Partial Least Square Regression (DPLSR) models for ENT at water quality monitoring stations within two adjacent marine recreational sites, Huntington State Beach (HSB) and Huntington City (HCB) Beach, California, using publicly available environmental data and tested whether these models overcome the drawbacks of the current model. The DPLSR models provide a better prediction of ENT than the current models based on comparisons of root-mean-square errors of prediction and the numbers of type 1 and 2 errors. We compared outcomes in terms of predicted illness, swimmers deterred from entering the water, and net benefits to swimmers for hypothetical management scenarios where beach advisories were issued based on (a) the previously collected sample's ENT density in conjunction with the two water quality criteria, and (b) predictions from DPLSR models in conjunction with the single-sample standard. At both HSB and HCB the DPLSR scenario produced a more favorable balance between illness prevention and recreational access. The results call into question the current method of beach management and show that model-informed decision-making and elimination of the geometric mean standard will aid beach managers in achieving more favorable outcomes in terms of illness and access than are presently achieved using 1-day-old measurements, especially at beaches where water quality problems are chronic.
如果肠球菌（ENT）密度超过30天几何平均值或单样本水质标准，就会发布海滩健康公告。当前的ENT计数程序需要1天的培养时间；因此，海滩管理人员根据1天前的数据做出政策决策。这等同于使用一个假设第t天的ENT密度等于第t - 1天的ENT密度的模型。研究表明，ENT密度在短于一天的时间尺度上会发生变化，这使得当前模型用于决策的有效性受到质疑。我们利用公开可用的环境数据，为加利福尼亚州两个相邻的海洋休闲场所亨廷顿州立海滩（HSB）和亨廷顿市海滩（HCB）内的水质监测站创建了ENT的动态偏最小二乘回归（DPLSR）模型，并测试这些模型是否克服了当前模型的缺点。基于预测的均方根误差以及1类和2类错误的数量比较，DPLSR模型比当前模型能更好地预测ENT。对于假设的管理情景，我们比较了基于（a）先前采集样本的ENT密度结合两个水质标准，以及（b）DPLSR模型结合单样本标准发布海滩公告时，在预测疾病、阻止游泳者下水以及对游泳者的净收益方面的结果。在HSB和HCB，DPLSR情景在疾病预防和休闲准入之间产生了更有利的平衡。研究结果对当前的海滩管理方法提出了质疑，并表明基于模型的决策以及消除几何平均标准将有助于海滩管理人员在疾病和准入方面取得比目前使用1天前的测量结果更有利的结果，特别是在水质问题长期存在的海滩。
