Lilambd commited on
Commit
c40588f
·
verified ·
1 Parent(s): 669cb33

Sampler card

Browse files
Files changed (1) hide show
  1. README.md +36 -62
README.md CHANGED
@@ -1,62 +1,36 @@
1
- ---
2
- license: cc-by-4.0
3
- pretty_name: World Climate & Seasons Ledger
4
- tags: [climate, weather, koppen, climate-normals, ghcn, nasa-power, geospatial, seasons]
5
- ---
6
- # World Climate & Seasons Ledger (1991–2020 normals, extremes, trends)
7
-
8
- Monthly climate normals, season calendars, Köppen-Geiger classes, records and warming trends for
9
- **91,815 land cells (0.5°), 6,902 cities, 238 countries/territories, 3,398 admin-1 regions**, plus
10
- **observed normals for 14,550 weather stations** — and a measured check of the gridded values against those stations.
11
-
12
- Every row carries `source_ids` that resolve in `sources.csv`. Nothing is filled by guesswork: when a source has no value the cell is empty.
13
-
14
- ## What is measured vs. modelled
15
- | Layer | Origin | Period |
16
- |---|---|---|
17
- | `stations_observed_*` | **Station measurements** (NOAA GHCN-Daily, QC-flagged values dropped) | normals 1991–2020 (≥24 of 30 years), recent 2016–2025 |
18
- | `cells_*`, `cities_*`, `countries_*`, `admin1_*` | **Reanalysis / satellite grids** (NASA POWER: MERRA-2 for temperature, rain, humidity, wind, snow; CERES SYN1deg for sunlight, cloud, UV), bilinearly interpolated | normals 1991–2020 (solar/cloud 2001–2020), series 1981–2025, records 1981-01–2026-09 |
19
-
20
- ## How good is the grid? (measured, `validation_summary.csv`)
21
- - Mean temperature normal at 8,070 stations: median bias −0.42 °C (grid colder), median monthly abs. error 0.91 °C; 62% within ±1 °C, 86% within ±2 °C. The error is largest at mountain and coastal stations (grid cells are ~50 km); grid-minus-station bias rises with station elevation (correlation +0.19), so the grid reads relatively warm at high stations and the overall cold bias is not explained by elevation.
22
- - Annual precipitation at 6,615 stations: median grid/obs ratio 1.01; 70% within ±20%.
23
- - Warming trend 1991–2025 at 4,061 stations: observed median +0.34 °C/decade, grid +0.36; 60% of stations agree within ±0.15.
24
- - **Known weak spot:** the grid under-states warming in South Asia (India, 31 stations: +0.20 vs grid +0.08 °C/decade; the Bangladesh grid series shows a step around 1997, checked against only one long-record station). `countries_summary.trend_check` flags each country (grouped by `ne_a3`) (`agree`, `grid_lower_than_stations`, `grid_higher_than_stations`, `too_few_stations`, `no_long_record_station`). Use the station trend where they disagree.
25
-
26
- ## Files
27
- | File | Rows | Key |
28
- |---|---|---|
29
- | cells_summary (.csv/.parquet) | 91,815 | `cell` = floor((90−lat)/0.5)·720 + floor((lon+180)/0.5) |
30
- | cells_monthly_normals.parquet | 1,101,780 | cell, month |
31
- | cells_annual_series.parquet | 4,131,675 | cell, year (1981–2025: t_mean_c, precip_mm) |
32
- | cities_summary / cities_monthly_normals | 6,902 / 82,824 | `city_id` (Lilambd cities backbone; Overture ids + Wikidata in that ledger) |
33
- | countries_summary / _monthly_normals / _annual_series | 468 / 5,616 / 21,060 | `ne_a3` (Natural Earth ADM0_A3, includes non-ISO units such as KAB Baikonur, SOL Somaliland) + `iso3` (ISO 3166-1 alpha-3, empty for non-ISO units) + `weight` = `area` or `population` |
34
- | admin1_summary / _monthly_normals / _annual_series | 6,760 / 81,120 / 304,200 | `admin1` (Natural Earth adm1_code), `admin1_iso` |
35
- | stations_observed_summary / _monthly_normals | 14,550 / 152,315 | GHCN `station_id` |
36
- | validation_summary.csv, sources.csv, cover.png | | |
37
-
38
- ## Main columns
39
- Monthly normals: `t_mean_c, t_max_avg_c, t_min_avg_c` (°C), `precip_mm` (monthly total), `rh_pct`, `dewpoint_c`, `wind10m_ms`,
40
- `frost_days` (days when 2 m temperature reaches a sub-zero dew point, POWER definition), `snow_depth_cm`, `snowfall_mm_we` (water equivalent),
41
- `solar_kwh_m2_day` (global horizontal irradiance), `cloud_pct`, `clear_sky_days`, `uv_index_24h_mean` (24-hour mean UV index, **not** the noon peak),
42
- season flags `is_dry_gaussen` (P < 2T), `is_frost_month` (mean daily min < 0 °C), `is_growing_month` (T ≥ 5 °C), `is_hot_month` (mean daily max ≥ 30 °C), `is_snow_cover_month` (snow depth ≥ 1 cm),
43
- and `*_recent_minus_normal` (2016–2025 minus 1991–2020; solar 2016–2025 minus 2001–2020).
44
-
45
- Summary: annual means/totals, warmest/coldest/wettest/driest month, `koppen`, month lists for each season flag,
46
- grid records `record_high_c` / `record_low_c` (hourly 2 m temperature extremes in the reanalysis cell, with month), wettest month on record, hottest/coldest/wettest/driest year,
47
- `t_trend_c_per_decade_1981_2025` with 95% CI, `precip_trend_pct_per_decade_1981_2025`, recent-minus-normal anomalies.
48
- Cities additionally list the nearest GHCN station within 50 km (empty when none; station elevation −999.9 in GHCN is stored as empty) with its **observed** record high/low (daily, 1991–2025) and observed trend.
49
-
50
- ## Codes
51
- `iso3`/`country_iso2` hold ISO 3166-1 codes only. Natural Earth's own units are kept in `ne_a3`. In `cities_*`, `country_code_source` keeps the code from the city source (XW West Bank, XG Gaza map to ISO PS).
52
-
53
- ## Limits
54
- - Grid cells are ~50 km; mountains, coasts and city heat islands are smoothed. Grid records are cell-average hourly values, not station records — station records are in `stations_observed_summary`.
55
- - Coastal cells in the land list contain sea. 4.4% of cells have no country (tiny islands/coast).
56
- - Köppen classes are computed from the grid normals with Beck et al. (2018) thresholds; small islands may show tropical classes from sea-surface influence.
57
- - Period of record differs: MERRA-2 to 2026-09, CERES to 2026-08, GHCN to 2025 for normals.
58
-
59
- ## License & attribution
60
- Tables: **CC BY 4.0**, © Lilambd. City identity columns (`city_id`, `name_en`, `lat`, `lon`) in `cities_*` come from Overture Maps divisions → **ODbL 1.0** (© OpenStreetMap contributors, Overture Maps Foundation).
61
- Data: NASA LaRC POWER Project (MERRA-2, CERES); NOAA NCEI GHCN-Daily v3.35 (Menne et al. 2012); Natural Earth; JRC GHS-POP (CC BY 4.0) for population weights. See `sources.csv`.
62
- Refresh: monthly (new POWER month → records, recent anomalies and series updated).
 
1
+ ---
2
+ license: cc-by-4.0
3
+ pretty_name: "World Climate & Seasons Ledger - Free Sampler (countries)"
4
+ tags: [climate, weather, koppen, climate-normals, nasa-power, ghcn, countries, seasons]
5
+ size_categories: [10K<n<100K]
6
+ configs:
7
+ - config_name: countries_summary
8
+ data_files: countries_summary.csv
9
+ - config_name: countries_monthly_normals
10
+ data_files: countries_monthly_normals.csv
11
+ - config_name: countries_annual_series
12
+ data_files: countries_annual_series.csv
13
+ ---
14
+ # World Climate & Seasons Ledger - Free Sampler (countries)
15
+
16
+ The country layer of the World Climate & Seasons Ledger by Lilambd: monthly climate normals (1991-2020), season calendars,
17
+ Koppen-Geiger classes, records and warming trends for 238 countries and territories (468 summary rows: each unit weighted by area and by population),
18
+ plus the measured check of the gridded values against weather stations.
19
+
20
+ | File | Rows | What |
21
+ |---|---|---|
22
+ | countries_summary (.csv/.parquet) | 468 | annual means and totals, warmest/coldest/wettest/driest month, Koppen class, season month lists, trends 1981-2025 with 95% CI, `trend_check` against stations |
23
+ | countries_monthly_normals (.csv/.parquet) | 5,616 | temperature, rain, humidity, wind, frost days, snow, sunlight, cloud, UV, season flags, 2016-2025 minus 1991-2020 |
24
+ | countries_annual_series (.csv/.parquet) | 21,060 | yearly mean temperature and precipitation, 1981-2025 |
25
+ | validation_summary.csv | 13 | how well the grid matches station measurements (temperature bias, precipitation ratio, trend agreement) |
26
+ | columns.csv | | every column with unit and meaning (includes columns of the full ledger) |
27
+
28
+ Country values are area- or population-weighted means of a 0.5-degree grid built from NASA POWER (MERRA-2 for temperature, rain, humidity, wind, snow;
29
+ CERES SYN1deg for sunlight, cloud, UV), checked against NOAA GHCN-Daily station measurements. Known weak spot: the grid under-states warming in South Asia;
30
+ `trend_check` flags each country where grid and stations disagree.
31
+
32
+ Not in this sampler (full ledger): 91,815 land cells, 6,902 cities, 3,398 admin-1 regions, 14,550 station normals, climate indices, Koppen shift and climate analogs.
33
+
34
+ ## Licence and credit
35
+ CC BY 4.0, (c) Lilambd. Underlying data: NASA POWER (NASA Langley Research Center, public), NOAA GHCN-Daily (public domain), Natural Earth (public domain) for country units,
36
+ GHS-POP (European Commission JRC, CC BY 4.0) for population weights. Credit line: "World Climate & Seasons Ledger by Lilambd, from NASA POWER and NOAA GHCN-Daily."