GCRI · 登记册
记录来源
Provenance is the account of where a record came from: who issued it, when, by what method, from which sources and through which transformations it passed before reaching the reader. Without it a record cannot be checked, weighed or corrected, only trusted or ignored. This page sets out how provenance is captured and expressed, why lineage chains break, and what the people who rely on registry records need to see before they do.
Record provenance: how the risk develops.
Issuer, date, method and source lineage. The dossier below sets out how the risk develops, what is measured, where the data come from, how it is modelled and who decides.
Lineage breaks at hand-offs
Records pass between spreadsheets, databases, aggregators and exports. Each step can drop the issuer, the source reference or the method note. Once the chain is cut, later copies look authoritative while nothing connects them to what they were derived from.
Issuer ambiguity
Organisation names change, merge and are spelt many ways, and people share names. Without persistent identifiers for issuers, two bodies are recorded as one or one as two, and responsibility for an assertion can no longer be established.
Method drift
The method behind a field changes, through a new coding rule, a new instrument or a revised definition, while the field name stays the same. A series then joins incompatible values, and no reader can see where the join is.
Laundering through aggregation
An unsourced claim is copied by an aggregator, quoted in a report and indexed by search engines. Each reuse adds apparent authority and removes context, until the claim is cited back to its origin as if it were independent confirmation.
How record provenance is modelled, measured and tested.
模型、其假设、数据和不确定性，以科学家或工程师会核查的形式呈现。结构遵循已发表的方法。所取数值仅为示意，不是研究发现。
数据集与来源追溯
The data classes behind record provenance and what is recorded about each before it is used.
| 数据类别 | 已记录溯源 | 已说明分辨率 | 许可已核查 |
|---|---|---|---|
| 卫星影像，光学与雷达 | 传感器、处理级别、获取日期 | 像元大小和重访周期 | 每景 |
| 再分析与气候模式输出 | 模型、实验、成员、版本 | 格网与时间步长 | 每次发布 |
| 原位传感器和测量站 | 测站、仪器、校准记录 | 点位与采样率 | 每个网络 |
| 人口与资产暴露 | 方法、版本年份、输入来源 | 单元大小和单位 | 每次发布 |
| 边界与参考地理 | 制作者、版本、日期 | 比例尺 | 每个版本 |
- 沿袭
- 从来源到图表的 W3C PROV 记录
- 原则
- FAIR：可查找、可访问、可互操作、可复用
- 缺口
- 显示为缺失，绝不悄然填充
Provenance lets a reader trace a figure in any domain to its source.
Provenance underpins every edition of 报告. It follows W3C PROV-O (2013) and the record identifiers of NS-3 in the proposed Nexus Standards.
- Gaps in a lineage chain.
- Source agencies’ own provenance records.
衡量什么，以及衡量得如何。
每个指标都注明其单位、频率、背后的来源类型以及数值将附带的证据标签。此处不显示数值：数值由计划依据其自身权限提供。
| 指标 | 单位 | 节奏 | 来源类型 | 标签 |
|---|---|---|---|---|
| Records with complete issuer, date and method | share of records | 每月 | Schema validation | 观测 |
| Lineage depth to a primary source | number of derivation steps | per record | Provenance graph | 观测 |
| Unresolved source references | 计数 | 每周 | Resolver checks | 观测 |
| Transformations without a recorded agent | 计数 | 每月 | Provenance graph audit | 观测 |
| Records with content hashes or signatures | share of records | 每月 | Integrity check | 观测 |
| Issuers with a persistent identifier | share of issuers | 每季度 | Identifier reconciliation | 观测 |
| Provenance inferred rather than declared | share of provenance statements | 每季度 | Curation review | 估算 |
证据来自何处。
具名公共数据集与机构记录，各自注明许可类别。依协议获得的来源仅按其条款使用，绝不再次发布。
| 来源 | 它提供什么 | 许可类别 |
|---|---|---|
| W3C PROV-O and PROV-DM | The standard vocabulary of entities, activities and agents for expressing provenance graphs | W3C document licence, free to implement |
| DataCite Metadata Schema | Creator, contributor, date, relation and related-identifier fields for research outputs | CC0 for metadata |
| ROR (Research Organization Registry) | Persistent identifiers for research organisations, with name variants, relationships and succession | CC0, no restrictions |
| ORCID public data file | Persistent identifiers for researchers, with linked works, affiliations and funding | CC0 |
| GLEIF Legal Entity Identifier data | Identifiers, legal names and parent relationships for legal entities | CC0, no restrictions |
| PREMIS Data Dictionary | Preservation metadata for events, agents and rights on digital objects | Free to use, maintained by the Library of Congress |
| RO-Crate | A specification for packaging research data with its provenance in JSON-LD | Open specification, attribution |
由谁决定，以及证据服务于什么。
Registry prepares evidence. The decisions below are taken by the authorities named, under their own rules, and never by the platform.
| 谁来决定 | 问题 | 本页面服务的证据 |
|---|---|---|
| 数据管理方 | Whether a record’s lineage is complete enough to publish or must be held for curation | Provenance completeness, unresolved references and transformations without a recorded agent |
| Journal and report editors | Whether a cited registry record can support a claim in a manuscript | Issuer identity, method, date and the depth of lineage to a primary source |
| 科研诚信官员 | Whether an assertion was altered, misattributed or fabricated | Time-stamped provenance graph, content hashes and the full edit history |
| National statistics offices | Whether administrative or registry records can enter published statistics | Source lineage, method stability over time and documented transformations |
该主题产生的产出
- Record views listing issuer, issue date, method and source references for each assertion
- Provenance graphs tracing derived records back to their primary sources
- A provenance completeness check applied to every record before publication
- Method change notes marking where a field’s definition or instrument changed
What provenance research asks.
- What is the smallest set of facts that lets a reader judge a record’s origin?
- How much lineage should be captured automatically, and how much asserted by people?
- How can provenance survive when records are copied, transformed or merged?
- How do we record a source that cannot be disclosed?
- How can a lineage claim itself be verified rather than trusted?
How lineage is lost.
Lineage is usually lost at boundaries: when a dataset is exported to a spreadsheet, when a figure is retyped into a report, when two sources are merged without recording which value came from where. Each step can be innocent, yet together they produce numbers whose origin no one can reconstruct. Archival science calls the related principle respect for provenance: records keep their link to the body that created them and to their original order.
Modern provenance models describe three kinds of thing: entities such as datasets and reports, activities such as a model run or a review, and agents such as people, institutions and software. The W3C PROV family of specifications expresses these relationships in a form machines can read, and the Registry’s lineage design is intended to follow it, as described on provenance and attribution.
Capturing and checking lineage.
- 01
Issuer and authority
Each record names who issued it and under what authority, so a reader knows whose word it is.
- 02
Activity records
Transformations, model runs and reviews are recorded with their method version and date.
- 03
Integrity checks
Cryptographic hashes can show that a file has not changed since it was recorded, a technique from 网络安全与密码学.
- 04
局限
Provenance shows where a record came from, not whether its content is right. Lineage is only as complete as the steps that were captured, and self-declared provenance can be false. Restricted sources can be named but not inspected.
Where this page’s references lead.
阅读本页时所对照的机构和文件。状态和版本会变化；每个链接都指向发布机构。
问题与解答。
What is the difference between provenance and a source?
What happens to records with incomplete provenance?
Does a digital signature make a record true?
Why do issuers need persistent identifiers?
Is provenance the same as citation?
What happens when a source is corrected?
Can provenance include restricted sources?
从本页贡献证据。
报告 is the Consortium’s quarterly publication. What you know about this subject can become evidence, a case or a correction in the next edition, credited to you and published under CC BY 4.0 where it is your own work.
- 电子邮件网关 GW-GCRI从本页发送证据、更正、案例、方法、审查、翻译或概况投稿。它会带上本页标签送达，由负责的编辑接收。
- Zenodo开放存缴存缴并提交到社区用您的 ORCID iD 在 Zenodo 上发布您自己的记录，然后提交到那里的 Reports 社区。由策展人审核；DOI 仍归您所有。
- 报告GCRI查看所有贡献方式贡献类型、各自的评审方式、您获得的署名、许可协议，以及绝不接受的内容。
切勿发送个人数据、敏感基础设施细节或您无权分享的任何内容。贡献在使用前会经过审查；发送贡献不赋予署名权或编辑权，赞助也绝不能换取研究结论。
法律文件。
本网站上的地图、数据和图表仅作为背景和一般信息展示。它们不是建议、预测或预警；信息的局限见免责声明和数据说明。