Greater Hunter Native Vegetation Mapping

## **Abstract** \n\nThis dataset and its metadata statement were supplied to the Bioregional Assessment Programme by a third party and are presented here as originally supplied.\n\n\n\nGreater Hunter Native Vegetation Mapping supplied by NSW Office of Water on 13/05/2014\n\n\n\nThe GHM geodatabase builds on a wealth of information and previous mapping from the\n\nHunter region. Existing field data, mapping, classification and remote sensing interpretation\n\nwere augmented with new survey data to produce the vegetation community classification\n\nused in this project. The classification used a series of well documented analyses as well as\n\nexpert review to achieve its end-point.\n\nThe GHM geodatabase contains two principal vegetation layers. The GHM Vegetation\n\nType layer and the Canopy Cover (v2) layer (individual tree crowns or clumps of tree\n\ncrowns). The GHM also contains field plot localities, associated species information and plotspecific\n\nphotographs. Data specific to each polygon (e.g. crown cover) and to each native\n\nvegetation community type (e.g. common name, scientific name) are included. Polygons, the\n\nfundamental spatial units, are built from computer-based feature recognition which delineates\n\nlandscapes patterns.\n\nThe GHM Vegetation Type map is built by attributing individual polygons with vegetation type\n\nfrom the GHM floristic classification through a multi-stage process. The process includes\n\nvisual interpretation of SPOT 5 and ADS40 imagery as well as species distribution modelling\n\nand expert review. The project included a review of existing mapping and classification and\n\nestablished equivalences between these and the GHM Classification. VIS ID 3855\n\n## **Dataset History** \n\nVegetation patterns at the stand scale were delineated using automated feature recognition\n\nsoftware. Definiens eCognition was used to define segments with low internal variation (low\n\nheterogeneity). Pan-sharpened SPOT5 data (5m) from multiple years formed the basis of\n\nthe segmentation. The data had been pre-processed to accentuate the range of spectral\n\nresponses or colours. The spatial resolution is 5m and the minimum mappable unit was set to\n\n400m2. The polygon boundaries have been smoothed and narrow slivers were eliminated.\n\nThere were two stages in the feature recognition approach. The first stage was optimised\n\nto differentiate woody and non-woody vegetation. The second stage was optimised to\n\ndifferentiate vegetation patterns within the extent of woody vegetation. The first stage\n\nemployed multi-temporal pan-sharpened SPOT - 5 data (5m). Only the red band (610-680nm)\n\nfrom each SPOT image was used to maximise the characteristic stability of woody vegetation\n\nover time. Each object was then classified as woody, non-woody and 'other' using the\n\nCrown Cover v2 layer and visual interpretation. For stage two the boundaries within the\n\nwoody vegetation were dissolved and new objects were created within their boundaries\n\nusing stretched, multi-temporal imagery. The contrast of all bands was increased using an\n\nadaptive equalisation stretch to maximise the separability of discrete vegetation patches within\n\nmosaics.\n\nThe vegetation map was created by attributing vegetation polygons with a vegetation type.\n\nThere are multiple stages involved but the fundamental steps are as follows:\n\nSurvey sites that meet quality criteria are assigned a GHM type label using PATN\n\nanalysis. Vegetation map units were defined using a hierarchical modelling approach that\n\nincluded the manual allocation of Keith Formation using visual identification, the use of a\n\nspecies distribution model to calculate the probability of GHM type in each polygon using\n\nenvironmental layers and a set of expert rules is developed to combine the formation\n\nclassification and the modelled results. The results undergo visual quality assurance, again\n\nusing manual image interpretation.\n\n## **Dataset Citation** \n\nNSW Office of Environment and Heritage (2014) Greater Hunter Native Vegetation Mapping. Bioregional Assessment Source Dataset. Viewed 13 March 2019, http://data.bioregionalassessments.gov.au/dataset/8f575981-3730-4989-84ce-c90204158406.

**摘要**

本数据集及其元数据说明由第三方提供给生物区域评估计划（Bioregional Assessment Programme），此处按原始提交状态呈现。

新南威尔士州水务局（NSW Office of Water）于2014年5月13日提交了大亨特地区原生植被制图（Greater Hunter Native Vegetation Mapping）成果。

GHM地理数据库（geodatabase）基于亨特地区大量既有资料与既往制图成果构建。本项目通过新增野外调查数据，扩充了既有野外实测数据、制图成果、分类体系与遥感解译结果，最终生成植被群落分类方案。该分类体系依托一系列规范有据的分析方法，并经过专家评审，以达成最终产出。

GHM地理数据库包含两类核心植被图层：GHM植被类型图层与冠层覆盖度（v2）图层（单个树冠或树冠簇）。此外，数据库还收录了野外样点位置、关联物种信息及针对单个样点的照片。数据内容涵盖每个多边形（如冠层覆盖度）的专属属性，以及各原生植被群落类型的通用名、学名等信息。作为基础空间单元的多边形，通过计算机特征识别技术勾勒景观格局后生成。

GHM植被类型制图通过多阶段流程，为每个多边形赋予GHM植物区系分类体系中的植被类型标签。该流程包括对SPOT 5与ADS40影像的目视解译、物种分布建模，以及专家评审环节。项目还对既有制图与分类体系开展了梳理，并建立了其与GHM分类体系的对应关系。VIS ID 3855

**数据集历史**

本研究采用自动化特征识别软件勾勒林分尺度的植被格局。使用Definiens eCognition软件定义内部异质性较低的分割单元。多期全色锐化SPOT5影像（5m分辨率）作为分割的基础数据，预处理过程中已通过增强光谱响应范围或色彩对比度优化数据质量。影像空间分辨率为5m，最小可制图单元设为400㎡。多边形边界已做平滑处理，并剔除了狭长细碎图斑。

特征识别流程分为两个阶段：第一阶段旨在区分木本与非木本植被，采用多期全色锐化SPOT-5影像（5m分辨率），仅提取各影像的红光波段（610-680nm）以最大化木本植被随时间变化的特征稳定性。随后结合冠层覆盖度v2图层与目视解译结果，将每个对象划分为木本、非木本及“其他”三类。第二阶段则针对木本植被范围内的植被格局差异进行区分：先融合木本植被内部的边界，再基于拉伸后的多期影像在其范围内创建新的分割对象。通过自适应均衡拉伸提升所有波段的对比度，以最大化镶嵌体中离散植被斑块的可分性。

植被制图通过为多边形赋予植被类型标签完成，整体流程包含多个环节，核心步骤如下：
1. 对符合质量标准的调查样点，通过PATN分析赋予GHM类型标签；
2. 采用分层建模方法定义植被制图单元：包括通过目视识别手动划定基思群系（Keith Formation）、利用物种分布模型结合环境图层计算每个多边形内GHM类型的分布概率、结合专家规则整合群系分类与模型预测结果；
3. 最终成果需再次通过目视解译开展视觉质量核查。

**数据集引用**

新南威尔士州环境与遗产办公室（NSW Office of Environment and Heritage，2014），大亨特地区原生植被制图，生物区域评估源数据集。2019年3月13日查阅，http://data.bioregionalassessments.gov.au/dataset/8f575981-3730-4989-84ce-c90204158406.