Taiwan historical weather observations rebuilt from official CWA/CODIS station data, with a legacy-compatible CSV interface for downstream users.
本資料庫收錄台灣歷史氣象測站資料,並已用中央氣象署 CWA CODIS 官方資料重新建庫,同時保留舊版 CSV 介面,讓既有下游程式可以持續使用。
This repository now provides three related dataset layers:
本 repository 目前提供三層資料:
data
data_codis_legacy_compatible
data_codis_rebuild_full
For most users, data is the correct folder to use.
對大多數使用者來說,直接使用 data 就是正確選擇。
Official primary source / 主要官方來源:
Station metadata reference / 測站中繼資料參考:
The rebuilt database now prefers a unified CODIS-based pipeline whenever possible.
新版資料庫已盡量統一改用 CODIS 作為主要下載來源,以減少不同來源混用造成的差異。
Two issues were reported in the older database:
舊版資料庫先前有兩類問題被回報:
.5 appeared.00:00..5 的數值,使用者質疑與原始資料不符。00:00 左右。The rebuild therefore aimed to:
所以這次重建的目標是:
The new pipeline now works in two stages:
新的建庫流程分成兩段:
tools/rebuild_codis_database.py
tools/build_legacy_compatible_dataset.py
A daily updater was also added. It refreshes a rolling 60-day window so the workflow stays fast even near year end:
另外也新增了每日更新流程。它只會更新最近 60 天,避免年底時 workflow 因為重跑整年而變得太慢:
tools/run_daily_codis_update.py.github/workflows/codis_daily_update.ymlThe final user-facing dataset follows this rule:
最後對外提供的相容資料遵循以下原則:
This means data and data_codis_legacy_compatible are compatibility layers built from the new CODIS rebuild, not plain raw dumps.
所以 data 與 data_codis_legacy_compatible 都是建構在新版 CODIS 重建資料上的相容層,不是單純的 raw dump。
Key checks on data_codis_rebuild_full:
1214582920539Key checks on data_codis_legacy_compatible:
620640010113772Compatibility rebuild statistics / 相容層統計:
1322253243514QC report files / QC 報告位置:
reports/codis_full_rebuild_notebook/post_rebuild_audit.jsonreports/codis_full_rebuild_notebook/legacy_compat_report.jsonreports/codis_full_rebuild_notebook/legacy_compat_audit.jsonAll datasets use the same layout:
所有資料夾都使用相同結構:
Examples / 例子:
data/466920/466920_1996.csvdata/466920/466920_1996_daily.csvdata/466920/466920_1996_monthly.csvNaming rules / 命名規則:
{station_id}_{year}.csv{station_id}_{year}_daily.csv{station_id}_{year}_monthly.csvThere are two kinds of columns in the compatibility datasets:
相容資料裡的欄位可以分成兩類:
StnPres: station pressure / 測站氣壓SeaPres: sea level pressure / 海平面氣壓Tx: air temperature / 氣溫Td: dew point temperature / 露點溫度RH: relative humidity / 相對濕度WS: mean wind speed / 平均風速WD: mean wind direction / 平均風向WSGust: maximum gust speed / 最大陣風風速WDGust: gust direction / 最大陣風風向Precp: precipitation accumulation / 累積降水量PrecpHour: precipitation duration or hourly summary / 降水時數或相關摘要SunShine: sunshine duration / 日照時數GloblRad: global radiation / 全天空日射量EvapA: Class A pan evaporation / A 盆蒸發量Visb: visibility / 能見度UVI: UV index / 紫外線指數CloudAmount: cloud amount / 雲量TxSoil0cm to TxSoil200cm: soil temperature at different depths / 不同深度土壤溫度StnPres, SeaPres: mean pressure / 平均氣壓StnPresMax, StnPresMaxTime: daily maximum station pressure and time / 日最大測站氣壓及時間StnPresMin, StnPresMinTime: daily minimum station pressure and time / 日最小測站氣壓及時間Tx: mean temperature / 平均氣溫TxMaxAbs, TxMaxAbsTime: absolute maximum temperature and time / 絕對最高溫及時間TxMinAbs, TxMinAbsTime: absolute minimum temperature and time / 絕對最低溫及時間TxRange: temperature range / 溫度日較差Td: mean dew point / 平均露點溫度RH: mean relative humidity / 平均相對濕度RHMin, RHMinTime: minimum relative humidity and time / 最低相對濕度及時間WS, WD: mean wind speed and direction / 平均風速與風向WSGust, WDGust, WGustTime: gust summary / 最大陣風摘要Precp: daily precipitation / 日累積雨量PrecpMax10, PrecpMax10Time: 10-minute precipitation maximum / 10 分鐘最大降雨量PrecpHrMax, PrecpHrMaxTime: hourly precipitation maximum / 1 小時最大降雨量SunShine: sunshine duration / 日照時數GloblRad: global radiation / 日射量EvapA: Class A pan evaporation / A 盆蒸發量VisbMean, VisbAutoMean: visibility summary / 能見度摘要UVIMax, UVIMaxTime: UV maximum and time / 紫外線指數最大值及時間CloudAmount, CloudAmountSat: cloud amount summary / 雲量摘要StnPres, SeaPres: mean pressure / 平均氣壓Tx: mean air temperature / 平均氣溫TxMaxAbs, TxMaxAbsTime: absolute monthly maximum temperature / 月絕對最高溫TxMinAbs, TxMinAbsTime: absolute monthly minimum temperature / 月絕對最低溫RH: mean relative humidity / 平均相對濕度WS, WD: wind summary / 風速風向摘要WSGust, WDGust, WGustTime: gust summary / 陣風摘要Precp: monthly precipitation / 月累積雨量PrecpDay: precipitation days / 降雨日數PrecpHour: precipitation duration / 降水時數PrecpMax10, PrecpMax60, PrecpHrMax, Precp1DayMax: precipitation extremes / 降水極值摘要SunShine, SunShineRate: sunshine summary / 日照摘要GloblRad: radiation summary / 日射摘要EvapA: Class A pan evaporation / A 盆蒸發量VisbMean, VisbAutoMean: visibility summary / 能見度摘要UVIMax, UVIMaxTime: UV summary / 紫外線摘要CloudAmount, CloudAmountSat: cloud summary / 雲量摘要VaporPressure: vapor pressure / 水氣壓TxSoil0cm to TxSoil500cm: monthly soil temperature summaries / 月土壤溫度摘要The repository includes a GitHub Actions workflow for daily updates:
本 repository 已內建 GitHub Actions 每日自動更新流程:
.github/workflows/codis_daily_update.ymltools/run_daily_codis_update.pyThe update flow is:
更新流程如下:
data_codis_rebuild_fulldata_codis_legacy_compatibledatadata_codis_rebuild_fulldata_codis_legacy_compatibledataA duplicate check workflow is also kept:
另外也保留每週重複資料檢查:
.github/workflows/hourly-duplicate-report.ymltools/rebuild_codis_database.py
tools/build_legacy_compatible_dataset.py
tools/run_daily_codis_update.py
tools/build_push_plan.ps1
tools/apply_push_batch.ps1
tools/data_quality/duplicate_hourly_report.py
If you do not want to browse the repository directly, a web interface is available:
若不想直接瀏覽 GitHub repository,也可以使用網頁介面:
Please cite / 建議引用:
Ou, J.-H., Kuo, C.-H., Wu, Y.-F., Lin, G.-C., Lee, M.-H., Chen, R.-K., Chou, H.-P., Wu, H.-Y., Chu, S.-C., Lai, Q.-J., Tsai, Y.-C., Lin, C.-C., Kuo, C.-C., Liao, C.-T., Chen, Y.-N., Chu, Y.-W., Chen, C.-Y., 2023. Application-oriented deep learning model for early warning of rice blast in Taiwan. Ecological Informatics 73, 101950. https://doi.org/10.1016/j.ecoinf.2022.101950
3,220 commits
Python
75.8%
Jupyter Notebook
17.0%
PowerShell
7.2%
Taiwan historical weather observations rebuilt from official CWA/CODIS station data, with a legacy-compatible CSV interface for downstream users.
本資料庫收錄台灣歷史氣象測站資料,並已用中央氣象署 CWA CODIS 官方資料重新建庫,同時保留舊版 CSV 介面,讓既有下游程式可以持續使用。
This repository now provides three related dataset layers:
本 repository 目前提供三層資料:
data
data_codis_legacy_compatible
data_codis_rebuild_full
For most users, data is the correct folder to use.
對大多數使用者來說,直接使用 data 就是正確選擇。
Official primary source / 主要官方來源:
Station metadata reference / 測站中繼資料參考:
The rebuilt database now prefers a unified CODIS-based pipeline whenever possible.
新版資料庫已盡量統一改用 CODIS 作為主要下載來源,以減少不同來源混用造成的差異。
Two issues were reported in the older database:
舊版資料庫先前有兩類問題被回報:
.5 appeared.00:00..5 的數值,使用者質疑與原始資料不符。00:00 左右。The rebuild therefore aimed to:
所以這次重建的目標是:
The new pipeline now works in two stages:
新的建庫流程分成兩段:
tools/rebuild_codis_database.py
tools/build_legacy_compatible_dataset.py
A daily updater was also added. It refreshes a rolling 60-day window so the workflow stays fast even near year end:
另外也新增了每日更新流程。它只會更新最近 60 天,避免年底時 workflow 因為重跑整年而變得太慢:
tools/run_daily_codis_update.py.github/workflows/codis_daily_update.ymlThe final user-facing dataset follows this rule:
最後對外提供的相容資料遵循以下原則:
This means data and data_codis_legacy_compatible are compatibility layers built from the new CODIS rebuild, not plain raw dumps.
所以 data 與 data_codis_legacy_compatible 都是建構在新版 CODIS 重建資料上的相容層,不是單純的 raw dump。
Key checks on data_codis_rebuild_full:
1214582920539Key checks on data_codis_legacy_compatible:
620640010113772Compatibility rebuild statistics / 相容層統計:
1322253243514QC report files / QC 報告位置:
reports/codis_full_rebuild_notebook/post_rebuild_audit.jsonreports/codis_full_rebuild_notebook/legacy_compat_report.jsonreports/codis_full_rebuild_notebook/legacy_compat_audit.jsonAll datasets use the same layout:
所有資料夾都使用相同結構:
Examples / 例子:
data/466920/466920_1996.csvdata/466920/466920_1996_daily.csvdata/466920/466920_1996_monthly.csvNaming rules / 命名規則:
{station_id}_{year}.csv{station_id}_{year}_daily.csv{station_id}_{year}_monthly.csvThere are two kinds of columns in the compatibility datasets:
相容資料裡的欄位可以分成兩類:
StnPres: station pressure / 測站氣壓SeaPres: sea level pressure / 海平面氣壓Tx: air temperature / 氣溫Td: dew point temperature / 露點溫度RH: relative humidity / 相對濕度WS: mean wind speed / 平均風速WD: mean wind direction / 平均風向WSGust: maximum gust speed / 最大陣風風速WDGust: gust direction / 最大陣風風向Precp: precipitation accumulation / 累積降水量PrecpHour: precipitation duration or hourly summary / 降水時數或相關摘要SunShine: sunshine duration / 日照時數GloblRad: global radiation / 全天空日射量EvapA: Class A pan evaporation / A 盆蒸發量Visb: visibility / 能見度UVI: UV index / 紫外線指數CloudAmount: cloud amount / 雲量TxSoil0cm to TxSoil200cm: soil temperature at different depths / 不同深度土壤溫度StnPres, SeaPres: mean pressure / 平均氣壓StnPresMax, StnPresMaxTime: daily maximum station pressure and time / 日最大測站氣壓及時間StnPresMin, StnPresMinTime: daily minimum station pressure and time / 日最小測站氣壓及時間Tx: mean temperature / 平均氣溫TxMaxAbs, TxMaxAbsTime: absolute maximum temperature and time / 絕對最高溫及時間TxMinAbs, TxMinAbsTime: absolute minimum temperature and time / 絕對最低溫及時間TxRange: temperature range / 溫度日較差Td: mean dew point / 平均露點溫度RH: mean relative humidity / 平均相對濕度RHMin, RHMinTime: minimum relative humidity and time / 最低相對濕度及時間WS, WD: mean wind speed and direction / 平均風速與風向WSGust, WDGust, WGustTime: gust summary / 最大陣風摘要Precp: daily precipitation / 日累積雨量PrecpMax10, PrecpMax10Time: 10-minute precipitation maximum / 10 分鐘最大降雨量PrecpHrMax, PrecpHrMaxTime: hourly precipitation maximum / 1 小時最大降雨量SunShine: sunshine duration / 日照時數GloblRad: global radiation / 日射量EvapA: Class A pan evaporation / A 盆蒸發量VisbMean, VisbAutoMean: visibility summary / 能見度摘要UVIMax, UVIMaxTime: UV maximum and time / 紫外線指數最大值及時間CloudAmount, CloudAmountSat: cloud amount summary / 雲量摘要StnPres, SeaPres: mean pressure / 平均氣壓Tx: mean air temperature / 平均氣溫TxMaxAbs, TxMaxAbsTime: absolute monthly maximum temperature / 月絕對最高溫TxMinAbs, TxMinAbsTime: absolute monthly minimum temperature / 月絕對最低溫RH: mean relative humidity / 平均相對濕度WS, WD: wind summary / 風速風向摘要WSGust, WDGust, WGustTime: gust summary / 陣風摘要Precp: monthly precipitation / 月累積雨量PrecpDay: precipitation days / 降雨日數PrecpHour: precipitation duration / 降水時數PrecpMax10, PrecpMax60, PrecpHrMax, Precp1DayMax: precipitation extremes / 降水極值摘要SunShine, SunShineRate: sunshine summary / 日照摘要GloblRad: radiation summary / 日射摘要EvapA: Class A pan evaporation / A 盆蒸發量VisbMean, VisbAutoMean: visibility summary / 能見度摘要UVIMax, UVIMaxTime: UV summary / 紫外線摘要CloudAmount, CloudAmountSat: cloud summary / 雲量摘要VaporPressure: vapor pressure / 水氣壓TxSoil0cm to TxSoil500cm: monthly soil temperature summaries / 月土壤溫度摘要The repository includes a GitHub Actions workflow for daily updates:
本 repository 已內建 GitHub Actions 每日自動更新流程:
.github/workflows/codis_daily_update.ymltools/run_daily_codis_update.pyThe update flow is:
更新流程如下:
data_codis_rebuild_fulldata_codis_legacy_compatibledatadata_codis_rebuild_fulldata_codis_legacy_compatibledataA duplicate check workflow is also kept:
另外也保留每週重複資料檢查:
.github/workflows/hourly-duplicate-report.ymltools/rebuild_codis_database.py
tools/build_legacy_compatible_dataset.py
tools/run_daily_codis_update.py
tools/build_push_plan.ps1
tools/apply_push_batch.ps1
tools/data_quality/duplicate_hourly_report.py
If you do not want to browse the repository directly, a web interface is available:
若不想直接瀏覽 GitHub repository,也可以使用網頁介面:
Please cite / 建議引用:
Ou, J.-H., Kuo, C.-H., Wu, Y.-F., Lin, G.-C., Lee, M.-H., Chen, R.-K., Chou, H.-P., Wu, H.-Y., Chu, S.-C., Lai, Q.-J., Tsai, Y.-C., Lin, C.-C., Kuo, C.-C., Liao, C.-T., Chen, Y.-N., Chu, Y.-W., Chen, C.-Y., 2023. Application-oriented deep learning model for early warning of rice blast in Taiwan. Ecological Informatics 73, 101950. https://doi.org/10.1016/j.ecoinf.2022.101950
3,220 commits
Python
75.8%
Jupyter Notebook
17.0%
PowerShell
7.2%