Habitat condition data for Australia from expert elicitation

These data relate to a project that aimed to construct and test a method for habitat condition data capture across Australia using expert elicitation. The data derived from experts are in two forms: (1) habitat condition scores for specified areas at a specified date range, and; (2) habitat condition scores based on images (photographs) of ecosystems. The image based habitat condition data were collected to enable cross-calibration of contributed site assessment data. These data represent the start of a continent-wide library of ecological condition data suitable for training and validation of model-based approaches to habitat condition assessment.\nLineage: Objectives\nThe project developed a novel approach to creating a continent-wide library of ecological condition data suitable for training and validation of model-based approaches to habitat condition assessment. Rather than attempting to aggregate complex raw attribute data, the project collated expert interpretations of habitat condition. Reliable condition assessment is heavily dependent on the deep ecological knowledge of members of Australia’s ecological science and natural resource management communities, hence this project engaged with experts in a wide range of ecosystem types across Australia.\n\nMethods – site assessments\nExperts recorded their assessment of habitat condition for areas within Australia using a data capture tool hosted by the Atlas of Living Australia. Experts mapped site/s with which they have deep familiarity, using polygons. The sites may be small or large, depending on the area over which a consistent condition score can be applied. For each site, experts provided a condition score between 0 and 1 (1=pristine; 0=natural habitat completely removed), the time period of their assessment, and (optionally) disturbances influencing the score. For further details on how data were captured via the online tool, see the document ‘Expert-Elicitation-Guidance.pdf’ provided alongside the data.\n\nMethods – image assessments\nExperts recorded their assessment of images using a data capture tool hosted by the Atlas of Living Australia. The image assessment scores are intended to be used to calibrate the site assessment condition scores contributed by experts. Experts were asked to provide a condition score for a suite of images allocated to them based on the Major Vegetation Groups (MVG) and Hutchinson bioclimatic classifications that they nominated familiarity with during the project registration process. The calibration images characterised each ecosystem in several different condition states.\n\nData preparation\nFollowing completion of the data capture phase of the project, data were downloaded from the online tool hosted by Biocollect on the Atlas of Living Australia on 23 November 2018. The data were reformatted and refined using a customised script in R. This processing involved reading in the data from file, removing entries associated with system testing, reformatting from multiple rows per entry to a single row, correcting a range of minor data issues, de-identifying records where requested by experts, writing the processed data out to file. All of the images that were assessed within the HCAT were downloaded from Biocollect. The shapefile holding the spatial polygons for the expert contributed site condition assessments were downloaded from biocollect, processed to remove polygons entered as part of testing, and written out to a single shapefile.\n\nData products\nThe condition assessments of sites contributed by experts were formatted and prepared into the following files:\nSiteConditionAssessment.csv - Site assessment data of on-ground habitat condition for the 314 sites contributed by experts.\nDescriptor_SiteConditionAssessment.csv – A file describing the fields used in the SiteConditionAssessment data file.\nSiteAssessmentShapefile – A folder holding the shapefile (projectSites) specifying the 314 spatial polygons for which site assessments were contributed by experts, with a matching identifier (‘siteID’) to each record in the SiteConditionAssessment file (‘location’).\n\nThe image assessment data were formatted and prepared into the following files:\nImageAssessment.csv – Data on the 278 image assessments of habitat condition, undertaken by experts.\nDescriptor_ImageAssessment.csv – A file describing the fields used in the ImageAssessment.csv data file.\nImageAssessmentImages – A folder holding the 77 habitat images that were assessed by experts, as cross-referenced in the ImageAssessment.csv data file.\n

本数据集关联一项旨在通过专家征询法构建并测试一套适用于全澳大利亚生境状况数据采集方法的研究项目。该项目所获取的专家来源数据包含两种形式：(1) 特定时间段内指定区域的生境状况评分；(2) 基于生态系统图像（照片）的生境状况评分。基于图像的生境状况数据采集目的在于对专家提交的样地评估数据进行交叉校准。本数据集标志着首个适用于生境状况评估模型训练与验证的全大陆尺度生态状况数据库的开端。

项目溯源与目标：本项目开发了一种全新的全大陆尺度生态状况数据库构建方法，用于支撑生境状况评估模型的训练与验证。相较于直接整合复杂的原始属性数据，本项目通过整理专家对生境状况的解读结果完成数据汇集。可靠的生境状况评估高度依赖澳大利亚生态科学与自然资源管理领域从业者的深厚生态学知识，因此本项目邀请了澳大利亚境内覆盖多种生态系统类型的专家参与。

方法——样地评估：专家通过澳大利亚生物多样性地图集（Atlas of Living Australia）托管的数据采集工具，对澳大利亚境内区域的生境状况开展评估并记录结果。专家以多边形矢量形式绘制其熟悉的样地范围，样地的大小取决于可赋予一致生境状况评分的区域范围。针对每个样地，专家需提供0至1区间内的生境状况评分（1代表原生状态，0代表自然生境已完全消失）、评估的时间范围，以及（可选）影响该评分的干扰因素。关于通过在线工具进行数据采集的更多细节，请参阅随数据集一同发布的《Expert-Elicitation-Guidance.pdf》文档。

方法——图像评估：专家通过澳大利亚生物多样性地图集托管的数据采集工具，对分配的图像开展生境状况评估并记录结果。图像评估评分的用途为校准专家提交的样地评估生境状况评分。专家需针对项目注册阶段申报熟悉的主要植被群系（Major Vegetation Groups, MVG）与哈钦森生物气候分类下分配的一系列图像提供生境状况评分。校准图像覆盖了每种生态系统的多种不同生境状况状态。

数据预处理：项目数据采集阶段完成后，研究人员于2018年11月23日从澳大利亚生物多样性地图集的Biocollect托管在线工具中下载了所有数据。研究人员通过R语言编写的定制化脚本对数据进行格式重构与优化处理。该处理流程包括：读取原始数据文件、剔除系统测试产生的记录、将每条记录从多行格式转换为单行格式、修正若干细微数据问题、根据专家申请对记录进行去标识化处理，最后将处理后的数据导出为文件。所有在HCAT中完成评估的图像均从Biocollect平台下载。存储专家提交样地评估空间多边形的shapefile文件从Biocollect平台下载后，经处理剔除测试阶段录入的多边形矢量，最终导出为单个shapefile文件。

数据产品：专家提交的样地生境状况评估数据经整理后生成以下文件：
SiteConditionAssessment.csv：包含314个专家提交样地的实地生境状况评估数据。
Descriptor_SiteConditionAssessment.csv：用于说明SiteConditionAssessment.csv数据文件中各字段含义的文档。
SiteAssessmentShapefile：存储shapefile文件（projectSites）的文件夹，该shapefile包含314个专家提交评估的样地空间多边形，且每个多边形均配有与SiteConditionAssessment.csv文件中"location"字段匹配的唯一标识符"siteID"。

图像评估数据经整理后生成以下文件：
ImageAssessment.csv：包含278项专家开展的生境状况图像评估数据。
Descriptor_ImageAssessment.csv：用于说明ImageAssessment.csv数据文件中各字段含义的文档。
ImageAssessmentImages：存储77张经专家评估的生境图像的文件夹，这些图像与ImageAssessment.csv数据文件中的记录一一对应。