onedz_datasets_csv This directory contains the split CSV datasets of the ZirconRegular_LLM project. All files are partitioned into manageable parts (~100,000–130,000 rows each) for batch processing, LLM ingestion, or memory-constrained workflows. ## Directory Structureonedz_datasets_csv/│├── Total_UPb_split_parts/              # Main U-Pb geochronology database│   ├── zircon_upb_part_01.csv│   ├── zircon_upb_part_02.csv│   └── ... (22 parts total)│├── Total_LuHf_split_parts/             # Lu-Hf isotope database, note that all files have been checked by experts│   ├── zircon_luhf_part_01.csv│   ├── zircon_luhf_part_02.csv│   └── zircon_luhf_part_03.csv│└── Experts_checked_UPb_split_parts/    # Expert-reviewed U-Pb subsets    ├── expert_upb_part_01.csv    ├── expert_upb_part_02.csv    └── ... (14 parts total)## Dataset Summary | Dataset | Parts | Est. Total Rows | Columns | Content ||---------|-------|-----------------|---------|---------|| Total_UPb_split_parts | 22 | ~2,550,000 | 64 | Full detrital zircon U-Pb age database || Total_LuHf_split_parts | 3 | ~297,000 | 33 | Lu-Hf isotope data linked to U-Pb records (expert-checked) || Experts_checked_UPb_split_parts | 14 | ~1,497,000 | 64 | Peer-reviewed regional compilations (quality-controlled) |--- ## File Format All CSV files follow the project standard:| Property | Specification ||----------|---------------|| Encoding | UTF-8 with BOM (utf-8-sig) || Delimiter | Comma (,) || Line endings | LF (\n) || Header | Single header row with standardized column names || Quoting | Double-quoted fields when containing commas or newlines | ### U-Pb Standard Columns (64 total) - Bibliographic: Lead_Author, Year, Journal, Vol, Pages, Title, Web_Link- Sample: Published_Sample_ID, Country_State, Region, Continent, Major_Geographic_Geologic_Unit, Minor_Geologic_Geographic_Unit, Group, Formation, Member, Locality, Profile, Latitude, Longitude- Depositional Age: Depos_Age_Period, Depos_Age_Epoch, Depos_Age_Stage, Max_Depos_Age_Ma, Est_Depos_Age_Ma, Min_Depos_Age_Ma- Analytical: Spectrometer, Spectrometer_Location, Institution, Spectrometer_Mode, Rock_Type_one, Rock_Type_two, Rock_Type_three, Grain, Spot_Location, Spot_diam- Isotope Ratios: Pb206U238_iso, Pb207U235_iso, Pb207Pb206_iso, Pb208Th232_iso (with one-sigma uncertainties)- Calculated Ages: Pb206U238_age, Pb207U235_age, Pb207Pb206_age, Best_age (with one- and two-sigma uncertainties), Discord- Elemental: U_ppm, Th_ppm, Pb_ppm, Pb206Pb204, Pb204Pb206, UTh_ratio, ThU_ratio ### Lu-Hf Columns (33 total) Includes all bibliographic and sample metadata columns above, plus:- Upb_Age, Upb_Age_two_sigma- 176Hf177Hf_iso, 176Lu177Hf_iso, 176Yb177Hf_iso (with 2-sigma uncertainties)- epsilon_Hf_0, epsilon_Hf_t (with 1-sigma and 2-sigma uncertainties)- TDM1_Ma, TDM2_Ma (with 2-sigma uncertainties)--- ## Usage Notes 1. Load order: When reassembling the full dataset, load parts in numerical order (01 → 22).2. Row overlap: Parts are split sequentially; no duplicate rows exist across parts of the same dataset.3. Cross-dataset linkage: Use Lead_Author + Year + Published_Sample_ID + Grain to link U-Pb records with Lu-Hf records.4. Expert vs. Total: Experts_checked_UPb_split_parts is a subset of the total database, curated from peer-reviewed regional compilations. It does not contain all rows from Total_UPb_split_parts.

This is the new, version 3 of the OneDZ database. Following the reviewers' comments, the new version adopted the "key-value" framework to manage the whole database to improve the searching efficiency. The calculated uncertainties and estimated coordinates has been noted in the "zircon_id". This design is to decrease additional sparse searching cost. The calculatations and estimation would not effect the statistical results after testing. For takeaway users, just apply the database directly. If you are not familiar with MySQL, we recommend you to find data from csv files. This new version added some new data from remote contribution by the web platform of OneDZ. In the next published new version (include this), csv files would not be published because of data were stored by "key-value" structure. Sorry for this, but transfferring the structured MySQL file to the unstructured csv files is really time-consuming.

We strongly recommend that users who need fast relational queries download the original SQL dump in version 1 of this record. If you prefer to work without SQL, the raw CSV tables provided in version 2 are the simplest alternative. In addition, version 3 (current) adds a relational table in which every zircon record is keyed by a Harbin-algorithm-generated unique ID, incorporates all new contributions received up to 1 August 2025, and summarises the most common encoding artefacts (Chinese, Latin and Arabic characters mis-rendered during prior UTF-8/ANSI conversions). When importing, please ensure that the primary-key column is preserved and that your client is set to UTF-8.| Correct character | Typical mis-rendered sequence | Language/script | Affected contexts*        || --------------- | ----------------------------- | --------------- | -------------------------- || 岩                 | 丠, 岩, 岩?                    | Chinese         | 地层名称, 岩性描述                 || 锆                 | é’„, 镶, 锆?                  | Chinese         | 锆石, 锆石U-Pb                 || 喜                 | å–œ, å–œ?, 喜?                 | Chinese         | 喜山期, 喜马拉雅                  || 世                 | 世, 世?                       | Chinese         | 地质年代, 上新世                  || ñ                 | ñ, ñ                      | Latin           | Mañicaragua Fm, Peña Negra || ç                 | ç, ç                      | Latin           | Maçacará Gp, Curaçá Fm     || é                 | é, é                      | Latin           | Precambriérien, Jurássico  || ã                 | ã, ã                      | Latin           | Serra ão, Triãngulo       || ﻩ                 | ?, ØŸ, 问号                     | Arabic          | جبل (Jebel), حجر (rock)    || ﻙ                 | ?, Ùƒ                         | Arabic          | كربونيفيري (Carboniferous) || ﺔ                 | ?, Ø©                         | Arabic          | ﺔﻘﺒﻄ (Taqabat Fm)          |Due to the departure of several graduate students who originally compiled the database, and because our remaining resources are fully committed to web-platform development, the table above lists representative errors only; we do not have the personnel to exhaustively catalogue every encoding artefact. All records have since been automatically rectified with GeoGPT, and the entire dataset is now provided in English, yet—because local encoding settings differ across PCs—we cannot guarantee that every user will be entirely free of garbled text. Both the v1 SQL dump and all CSV/XLSX files in v2 are encoded in UTF-8; please ensure your client is likewise configured before import. The GeoGPT-based repair and translation modules are under active development and testing, and we expect a future web release to resolve these issues definitively.

We still strongly recommend using the .sql file to manage and analyze for the gaint database. Reseons are:(1) coding error, as we converting sql  files to csv files, so many Chinese, Latin characters, and even a small number of Arabic characters have become garbled. (2) Unfortunately, it is roughly estimated that this coding error will result in at least 20% data loss.(3) What's even more terrifying is that due to different PC settings for each person, Excel always generates some barcodes that are automatically tampered with. Details can be seen Zeeberg B R , Riss J , Kane D W , et al. Mistaken Identifiers: Gene name errors can be introduced inadvertently when using Excel in bioinformatics[J]. BMC Bioinformatics, 2004, 5(1):80-0. Ziemann M , Eren Y , El-Osta A . Gene name errors are widespread in the scientific literature[J]. Genome Biology, 2016, 17(1):177. Question: How to avoid conversion of gene symbols to date format in Excel. Welsh E A , Stewart P A , Kuenzi B M , et al. Escape Excel: A tool for preventing gene symbol and accession conversion errors[J]. PLOS ONE, 2017, 12(9):e0185207. Escape gene name-mangling with ‘Escape Excel’.(4) Due to the performance of SQL in automation, it is possible for the database to continue to grow, so we mainly consider updating SQL files when updating the database every six months/year. At the same time, we will also release CSV files, all of which will be annotated with "sql2csv".(5) In this version, 'Original' indicates the process files contributed by all authors. If you are interested in specific regional data, you can refer to such files. In future version updates, the intermediate process files will be indicated.

This is the new, version 3 of the OneDZ database. Following the reviewers' comments, the new version adopted the "key-value" framework to manage the whole database to improve the searching efficiency. The calculated uncertainties and estimated coordinates has been noted in the "zircon_id". This design is to decrease additional sparse searching cost. The calculatations and estimation would not effect the statistical results after testing. For takeaway users, just apply the database directly. If you are not familiar with MySQL, we recommend you to find data from csv files. This new version added some new data from remote contribution by the web platform of OneDZ. In the next published new version (include this), csv files would not be published because of data were stored by "key-value" structure. Sorry for this, but transfferring the structured MySQL file to the unstructured csv files is really time-consuming.