Border Rivers Gwydir / Namoi Regional Native Vegetation Map Version 2.0. VIS_ID 4204

## **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\nThe Border Rivers Gwydir and Namoi Regional Vegetation Map is a subset of the statewide vegetation mapping and classification program undertaken by the NSW Office of Environment and Heritage (OEH Regional Scale State Vegetation Map) and covers the two former Catchment Management Authority Regions.\\\\n\\\\nThe primary thematic data layer in this dataset is a map of regional scale Plant Community Types (PCT's). The map was developed from a process using vegetation surveys, remote sensing derivations, visual interpretation and spatial distribution models.\\\\n\\\\nThe full dataset comprises the following data layers as delivered in an ArcGIS 9.3 File Geo-database:\\\\n\\\\nPLANT COMMUNITY TYPE: The primary map of Plant Community Types developed from an ensemble of visual interpretation of high resolution imagery and spatial distribution models.\\\\n\\\\nWOODY EXTENT LAYER: A map of woody vegetation derived from classification of 5m SPOT-5 imagery.\\\\n\\\\nKEITH CLASS: A map based on aerial photo interpretation and spatial distribution models.\\\\n\\\\nMAP SOURCE: A map of the various sources of information used including spatial models, visual interpretation and existing map products.\\\\n\\\\nSURVEY DENSITY ALL: A map of the density of all survey sites used.\\\\n\\\\nSURVEY DENSITY FULL FLORISTICS: A map of the density of only full floristic survey sites used.\\\\n\\\\nMODELLING CONFIDENCE: A map of the confidence outcomes achieved.\\\\n\\\\nWhile much of the aerial photo interpretation employed was undertaken at around 1:8000, PCT attribution is generally at a much coarser scale. The Map Source layer (as described above) can be used as a guide to how vegetation attribution was derived. We recommend that the highest resolution appropriate for this product be 1:15000.\\\\n\\\\nValidation Summary:\\\\n\\\\nPCT Map: Based on 100% of the survey data (modelling and hand mapping), the final mapped product has an accuracy in the range 68%-70% for prediction of the three most likely PCTs. Be aware that these accuracies are highly variable across each PCT. Some PCT's utilised more site data than others. Keith Class reached a 76% accuracy using the independent test data. Modelled PCT and modelled top 3 PCT overall accuracies were 53% and 68% respectively. Woody Extent received a 92% overall accuracy.\\\\n\\\\nAccompanying documents:\\\\nBRGNamoi Technical Notes.pdf - Technical Report \\\\nBRGN_PCT_KC_LUT.xls - A look-up table listing the relationship between PCT, Keith Class and Keith Formation classifications.\\ \\nBRGNv2_Spatial_Layer_Descriptors.txt\\\\nBRGN_V2.mxd\\\\nBorder Rivers Gwydir / Namoi Regional Native Vegetation Mapping\\\\nTechnical Notes Version 1.0. Reference: NSW Office of Environment and Heritage, 2015. BRG-Namoi Regional Native Vegetation Mapping. Technical Notes, NSW Office of Environment and Heritage, Sydney, Australia.\\\\n\\\\nThe download package contains a "quick view" map composite of the study area only. The quick view maps are of PCT, Keith Class, Keith Form, Map Source and Modelling Confidence. They also show the broad-scale line work. For more detailed line work and woody percent per polygon, please refer to the full dataset.\\\\n\\\\nFor access queries regarding the full dataset, please contact: [e-mail]\\\\n\\\\nBRG_Namoi_v2_0_E_4204.\\\\n\\ \\nVIS_ID 4204\n\n## **Purpose** \n\nThis dataset was developed as part of the OEH State Vegetation Map to provide government and community with regional-scale information about native vegetation.\n\n## **Dataset History** \n\nA summary of the product's lineage is below. Please refer to the Technical Notes v1.0 for a detailed description of the methodologies and source datasets.\\\\n\\\\nThe PCT map was derived primarily using a spatial modeling approach augmented with high resolution aerial imagery (50cm ADS40) for visual interpretation and automated line-work derivation.\\\\n\\ \\nIn summary the process for PCT attribution involved the following: \\\\n\\\\n1. Vegetation Survey and Classification: Existing floristic plot data comprised 9054 existing sites after data cleaning. A large number of gaps in existing survey coverage were evident and required further survey information. Stratification based on archive broad vegetation type mapping (Regional Vegetation Types; Eco Logical Australia 2008b) and gap analysis was undertaken to select locations for additional plot data collection. A total of 6013 additional rapid data points were collected. To allocate survey sites to PCTs, full floristic plots were analysed using a UPGMA clustering approach in Primer with significant groups identified using SIMPROF and species contributions for each resulting group calculated using SIMPER. The existing plot data were allocated across 258 PCTs.\\\\n\\\\n2. Pattern Derivation: A multi-resolution segmentation algorithm was used to create image objects with low internal variation. Image objects represent patches of vegetation that can later be classified based on attributes such as crown cover, spectral response, or soil type. The segmentation parameters and scale was derived iteratively based on visual inspection. Vegetation patterns from existing stereoscopic aerial photo interpretation and those recognised in high spatial resolution imagery (ADS40) were used as a reference point. Segmentation was performed using ADS40, SPOT 5 and SRTM derived topographic indices. this process provided the line work for subsequent PCT attribution.\\\\n\\\\n3. Visual attribution of Landscape Class: The purpose of attributing Landscape classes to polygons is to predetermine broad vegetation types for modelling purposes using remote sensing. These classes reduce the PCT options for any one polygon making the modeling more effective in its attribution with commensurate less computing effort/time. A landscape class was attributed to every polygon in the study area. Landscape classes were aided by reference to existing mapping. Corrections were made based on ADS40 with on-screen attribution. Every polygon was visually checked by an expert interpreter.\\\\n\\\\n4. Modelling Envelopes:As a further constraint to modelling outcomes, spatial envelopes were used to constrain PCTs to a certain geographic range, reducing the amount of types competing within the model at any particular location. The constraints used were applied at different stages in the mapping process. The Keith Class (Keith 2004) models were constrained to particular IBRA (Interim Bioregionalisation of Australia v7; Commonwealth of Australia 2012) subregions, selected based on review of the literature and expert opinion. The type models were constrained to particular ranges of a topographic position index, again based on literature review and expert opinion. Not all types were constrained by topographic envelopes, as some were considered to be less correlated with particular topographic positions.\\\\n\\\\n5. Spatial Distribution Modelling of Keith Classes and Plant Community Types. Modelling of Keith Class and PCT used a combination (ensemble) of Generalised Dissimilarity Model (GDM), Boosted Regression Trees (BRT), and a simple Nearest Neighbour model.A suite of candidate environmental predictor variables, including climate, geology, soil, geophysical data, and terrain indices, were compiled for use in the GDM and BRT models. A comprehensive list of these predictor variables can be found in the Technical Notes v1.0.\\\\n\\\\n6. Uplifted API and Expert Editing: Vegetation communities from the Gwydir Wetlands and Floodplain Vegetation Map 2008 (Bowen & Simpson 2010) were spatially translated into the current line-work via a majority extent per polygon algorithm. The vegetation community mapping resulting from the aforementioned procedures was extensively edited on screen to correct attribution where there may have been for example existing API, missed vegetation, ecological anomalies, incorrect assignments, modelling noise and inclusion of late site data. The extent of each attribution source is delineated by the Map Source data layer provided in this dataset.\\\\n\\\\nFor further details on methodology and validation please refer to the Border Rivers Gwydir / Namoi Regional Native Vegetation Mapping\\\\nTechnical Notes Version 1.0. Reference: NSW Office of Environment and Heritage, 2015. BRG-Namoi Regional Native Vegetation Mapping. Technical Notes, NSW Office of Environment and Heritage, Sydney, Australia.\n\n## **Dataset Citation** \n\nNSW Office of Environment and Heritage (2015) Border Rivers Gwydir / Namoi Regional Native Vegetation Map Version 2.0. VIS_ID 4204. Bioregional Assessment Source Dataset. Viewed 11 December 2018, http://data.bioregionalassessments.gov.au/dataset/b3ca03dc-ed6e-4fdd-82ca-e9406a6ad74a.

## **摘要**

本数据集及其元数据说明由第三方提交至生物区域评估计划（Bioregional Assessment Programme），本文按原始提交版本原样呈现。

边境河流-吉迪尔与纳莫伊区域植被图是新南威尔士州环境与遗产办公室（NSW Office of Environment and Heritage）实施的全州植被制图与分类项目（OEH省级尺度州植被图，OEH Regional Scale State Vegetation Map）的子集，覆盖原两个流域管理局（Catchment Management Authority）辖区。

本数据集的核心专题数据图层为区域尺度植物群落类型（Plant Community Types, PCTs）分布图。该图件基于植被调查、遥感反演、目视解译与空间分布模型构建流程开发而成。

完整数据集以ArcGIS 9.3文件地理数据库（ArcGIS 9.3 File Geo-database）格式交付，包含以下数据图层：

PLANT COMMUNITY TYPE: 植物群落类型（Plant Community Type, PCT）图层：基于高分辨率影像目视解译与空间分布模型集成构建的核心植物群落类型分布图。

WOODY EXTENT LAYER: 木本覆盖范围图层：基于5米分辨率SPOT-5影像分类得到的木本植被分布图。

KEITH CLASS: 基思分类（Keith Class）图层：基于航空影像解译与空间分布模型构建的分类图。

MAP SOURCE: 制图数据源图层：记录所用各类信息来源的图层，包括空间模型、目视解译成果与现有制图产品。

SURVEY DENSITY ALL: 全部调查样点密度图层：展示所用全部调查样点分布密度的图件。

SURVEY DENSITY FULL FLORISTICS: 完整区系调查样点密度图层：仅展示所用完整区系调查样点分布密度的图件。

MODELLING CONFIDENCE: 建模置信度图层：展示建模结果置信水平的图件。

尽管本次制图所用的大部分航空影像解译工作比例尺约为1:8000，但植物群落类型赋值通常采用更为粗糙的尺度。上述制图数据源图层可用于辅助了解植被赋值的生成方式。本产品的适宜最高分辨率推荐为1:15000。

### 验证摘要

植物群落类型图：基于全部调查数据（建模与手工制图），最终制图产品对三种最可能的植物群落类型的预测准确率为68%-70%。需注意，不同植物群落类型的准确率差异显著，部分群落类型所用样点数据多于其他群落。基思分类基于独立测试数据实现了76%的准确率。建模所得植物群落类型与建模所得前三大植物群落类型的总体准确率分别为53%与68%。木本覆盖范围图层的总体准确率为92%。

### 配套文档

BRGNamoi Technical Notes.pdf - 技术报告

BRGN_PCT_KC_LUT.xls - 列有植物群落类型、基思分类与基思地层分类之间对应关系的查找表

BRGNv2_Spatial_Layer_Descriptors.txt

BRGN_V2.mxd

Border Rivers Gwydir / Namoi Regional Native Vegetation Mapping

技术说明版本1.0。参考文献：新南威尔士州环境与遗产办公室，2015年。BRG-纳莫伊区域原生植被制图。技术说明，新南威尔士州环境与遗产办公室，澳大利亚悉尼。

下载包仅包含研究区域的"quick view"合成图件。快速预览图涵盖植物群落类型、基思分类、基思地层、制图数据源与建模置信度图层，同时展示了大尺度线划要素。如需获取更详细的线划要素与每个多边形的木本覆盖百分比，请查阅完整数据集。

如需查询完整数据集的获取方式，请联系：[e-mail]

BRG_Namoi_v2_0_E_4204。VIS_ID 4204

## **目的**

本数据集作为OEH省级植被图项目的一部分开发，旨在为政府与社会公众提供区域尺度的原生植被信息。

## **数据集历史**

以下为该产品的研发历程概要。如需了解方法学与源数据集的详细说明，请查阅技术说明v1.0。

植物群落类型图主要基于空间建模方法开发，并辅以高分辨率航空影像（50厘米分辨率ADS40）用于目视解译与自动化线划要素提取。

综上，植物群落类型赋值流程如下：

1. 植被调查与分类：经数据清洗后，现有区系样点数据共包含9054个样点。现有调查覆盖范围存在大量空白，需补充调查数据。基于历史大尺度植被类型制图（区域植被类型；Eco Logical Australia 2008b）与空白区分析进行分层，以选择补充样点采集区域。最终共采集6013个快速调查样点。为将调查样点归类至植物群落类型，采用PRIMER软件中的无加权配对组平均法（Unweighted Pair Group Method with Arithmetic Mean, UPGMA）聚类算法对完整区系样点进行分析，通过相似性轮廓检验（Similarity Profile, SIMPROF）识别显著聚类组，并通过相似性百分比分析（Similarity Percentage, SIMPER）计算各聚类组的物种贡献度。最终现有样点数据被划分为258个植物群落类型。

2. 格局提取：采用多分辨率分割算法生成内部异质性较低的影像对象。影像对象代表植被斑块，后续可基于冠层覆盖、光谱响应或土壤类型等属性进行分类。分割参数与尺度通过目视检查迭代优化。以现有立体航空影像解译得到的植被格局与高空间分辨率影像（ADS40）中识别的植被格局作为参考。分割过程基于ADS40、SPOT 5与SRTM（航天雷达地形测绘任务，Shuttle Radar Topography Mission）衍生的地形指数完成，该流程提供了后续植物群落类型赋值所需的线划要素。

3. 景观类目视赋值：为多边形赋予景观类别的目的是，基于遥感数据预先确定大尺度植被类型，以缩小单个多边形的候选植物群落类型范围，提升建模赋值效率并减少计算开销与耗时。本研究区域内的每个多边形均被赋予景观类别，赋值过程参考现有制图成果，并基于ADS40影像在屏幕上进行修正。所有多边形均经专业解译人员目视检查确认。

4. 建模范围约束：作为建模结果的进一步约束条件，采用空间范围约束将植物群落类型限定在特定地理范围内，以减少任意位置处模型中竞争的群落类型数量。该约束条件在制图流程的不同阶段应用。基思分类（Keith 2004）模型被限定在特定的澳大利亚临时生物分区（Interim Bioregionalisation of Australia v7, IBRA，澳大利亚联邦2012）亚区域内，该选择基于文献综述与专家意见确定。植物群落类型模型被限定在特定的地形位置指数范围内，同样基于文献综述与专家意见确定。并非所有群落类型均受地形范围约束，部分群落类型与特定地形位置的相关性较低。

5. 基思分类与植物群落类型的空间分布建模：基思分类与植物群落类型的建模采用广义相似性模型（Generalised Dissimilarity Model, GDM）、提升回归树（Boosted Regression Trees, BRT）与简单最近邻模型的集成方法。为GDM与BRT模型编译了一套候选环境预测变量，包括气候、地质、土壤、地球物理数据与地形指数。这些预测变量的完整列表可查阅技术说明v1.0。

6. 航空影像解译成果升级与专家编辑：基于2008年吉迪尔湿地与泛滥平原植被图（Bowen & Simpson 2010）得到的植被群落，通过每个多边形的最大范围算法，将其空间转换至当前线划要素中。通过上述流程得到的植被群落制图成果需在屏幕上进行大量编辑，以修正可能存在的问题，例如现有航空影像解译成果遗漏、生态异常、赋值错误、建模噪声与新增样点数据的纳入。每个赋值来源的范围由本数据集提供的制图数据源图层划定。

如需了解方法学与验证的更多细节，请查阅《边境河流-吉迪尔/纳莫伊区域原生植被制图技术说明版本1.0》。参考文献：新南威尔士州环境与遗产办公室，2015年。BRG-纳莫伊区域原生植被制图。技术说明，新南威尔士州环境与遗产办公室，澳大利亚悉尼。

## **数据集引用**

新南威尔士州环境与遗产办公室（2015）《边境河流-吉迪尔/纳莫伊区域原生植被图 版本2.0》。VIS_ID 4204。生物区域评估源数据集。2018年12月11日查阅，http://data.bioregionalassessments.gov.au/dataset/b3ca03dc-ed6e-4fdd-82ca-e9406a6ad74a。