North Marine Region benthic environmental classification

The dataset is a map of a benthic environmental classification of the North Marine Region (NMR). The aim of the classification was to summarise the way the physical environment near the seabed varies across the NMR. In particular, we were interested in variation within and across bioregion boundaries. The analysis was performed for the Department of the Environment and Water Resources for use in profiling of the NMR during the development of a marine bioregional plan for the region. The environmental classification process comprised laying a grid of cells over the NMR and sorting the cells into clusters on the basis of their physical environmental attributes. The classification method was unsupervised, model-based clustering with a Gaussian mixture model, and was performed with the MCLUST package for R. With this method, the statistical distributions of the clustering variables are modelled as a mixture of multivariate normal distributions. Clustering is a two-step process. First, the parameters of the mixture model are fitted by maximum likelihood. Second, the model is used to assign grid cells to clusters. The clustering variables were bathymetry, median current stress, sediment gravel percentage and average bottom salinity. The environmental characteristics of the nine clusters were (1) deep; (2) deep; (3) low stress, low gravel, low salinity; (4) average for the region; (5) average for the region; (6) shallow, low stress, high salinity; (7) shallow, low salinity; (8) shallow, high stress, gravel; and (9) shallow, high stress, high salinity.

本数据集为北海洋区域（North Marine Region, NMR）的底栖环境分类地图。本次分类的目标是总结该区域近海底物理环境的空间变化特征，尤其关注生物区域边界内部及跨边界的环境差异。本分析由环境与水资源部（Department of the Environment and Water Resources）委托开展，用于该区域海洋生物区域规划制定过程中的北海洋区域特征梳理工作。该环境分类流程为：在北海洋区域布设网格单元，并基于各单元的物理环境属性将其划分为不同聚类。本次分类采用基于模型的无监督聚类方法，结合高斯混合模型（Gaussian mixture model）实现，并通过R语言的MCLUST软件包完成运算。该方法将聚类变量的统计分布建模为多元正态分布的混合模型。聚类流程分为两个步骤：首先通过最大似然估计拟合混合模型的参数；随后利用该模型将网格单元分配至各聚类中。本次聚类所采用的变量包括水深（bathymetry）、底流应力中值（median current stress）、沉积物砾石占比（sediment gravel percentage）以及底层平均盐度（average bottom salinity）。9个聚类的环境特征分别为：(1) 深水环境；(2) 深水环境；(3) 低应力、低砾石占比、低盐度；(4) 区域平均水平；(5) 区域平均水平；(6) 浅水环境、低应力、高盐度；(7) 浅水环境、低盐度；(8) 浅水环境、高应力、高砾石占比；(9) 浅水环境、高应力、高盐度。