Statewide Marine Habitat Map 2021

Combined Port Phillip Bay habitat model developed by DELWP 2021 and Open Coast Deakin marine mapping campaign JAN 2019. \n\nThe PPB map was built to predict habitat distributions by applying Random Forest machine learning tools which integrate ground truthed data and environmental predictors. The resulting Habitat Complex map represents 19 different habitat complexes across Port Phillip Bay at Level 3 Combined Biotope Classification Scheme (CBiCS) described by Edmunds and Flynn (2015, 2018; 2021). A total of 8,325 ground-truth records were used within the model (of the 9,683 records) that met the Random Forest modelling criteria (¿5 unique values for each biotope category; Breiman 2001). Environmental parameters were sourced from different methods including lidar, multibeam, and indexes calculated from the Digital Elevation Model (DEM), and mapped at a resolution of 2.5 metres. The Random forest model produced an accuracy (Out-of-bag) of 91%. Link to methodology: https://www.marineandcoasts.vic.gov.au/__data/assets/pdf_file/0031/537169/Mazor-et-al-2021_Port-Phillip-Bay-CBiCS-Habitat-Model.pdf\n\nThe Open Coast statewide marine habitat map (excluding PPB) developed by Deakin Marine Mapping lab using a range of techniques including digitisation of polygons and has been compiled using a compilatiom of various datasets including Boon et al. (2011). Deakin Marine Mapping (Young M et al, 2018). GeoHab Victoria Estuaries Geomorphology (2010). Deakin Marine Mapping (Zavalas, R et al. 2018), DELWP wetlands (1994), Monk et al 2011, Roob and Ball (1997), Ford et al 2016, Edmunds and Flynn (2015), Blake and Ball (2001), Poore (1992), Cohen et al (2000). The map has later been reclassified to CBICS habitat classification scheme at broad-scale level 2 and 3.

本数据集为维多利亚州环境、土地、水资源与规划部（DELWP）2021年研发，以及迪肯大学Open Coast海洋测绘项目2019年1月开展的合并版菲利普港湾（Port Phillip Bay）生境模型。

该菲利普港湾生境图谱基于随机森林（Random Forest）机器学习工具构建，用于预测生境分布，该工具整合了实地验证数据与环境预测变量。最终生成的生境复合图谱按照埃德蒙兹与弗林（2015、2018、2021）提出的三级联合生境分类体系（Combined Biotope Classification Scheme, CBiCS），划分出菲利普港湾内共计19类不同的生境复合体。模型共纳入符合随机森林建模标准（每个生境类型需至少5个唯一值；Breiman 2001）的8325条实地验证记录（原始记录共9683条）。环境参数通过多种手段获取，包括激光雷达（lidar）、多波束测深数据，以及由数字高程模型（Digital Elevation Model, DEM）计算得到的各类指数，最终以2.5米的分辨率进行制图。本随机森林模型的袋外（Out-of-bag）准确率达91%。方法学参考文件链接：https://www.marineandcoasts.vic.gov.au/__data/assets/pdf_file/0031/537169/Mazor-et-al-2021_Port-Phillip-Bay-CBiCS-Habitat-Model.pdf

由迪肯大学海洋测绘实验室开发的全州范围开放海岸生境图谱（不含菲利普港湾），采用多边形数字化等多种技术手段构建，其数据整合了多项已发表数据集，包括Boon等（2011）、迪肯大学海洋测绘团队（Young M等，2018）、维多利亚州河口地貌GeoHab项目（2010）、迪肯大学海洋测绘团队（Zavalas R等，2018）、DELWP湿地数据集（1994）、Monk等（2011）、Roob与Ball（1997）、Ford等（2016）、Edmunds与Flynn（2015）、Blake与Ball（2001）、Poore（1992）以及Cohen等（2000）的相关研究成果。该图谱后续已按照CBiCS生境分类体系，在二级与三级分类尺度上完成重分类。