Map software and data credits

700 hPa vertical motion

GEM charts use direct native GRIB data from Environment and Climate Change Canada, under the ECCC Data Services End-use Licence. ICON Europe charts use direct data from Deutscher Wetterdienst; see DWD terms of use. These are Snowline charts, not official provider products.

Snowline preserves native grid spacing within the published regional coverage: GEM 0.15° over 66°W–35°E and 30–75°N; ICON Europe 0.0625° over 23.5°W–35°E and 35–70.5°N. Outside these subsets, the diagnostic is unavailable. Omega is converted from Pa/s to 0.1 Pa/s; DWD geopotential is divided by 9.80665 to give height in metres. Red indicates ascent, blue descent; black 700 hPa heights are labelled in dam and blue wind barbs in knots. All three layers use the same provider, grid, initialization and valid time. GEM upper-air steps are three-hourly to T+240; ICON main runs reach T+120 and intermediate runs T+48. The map exposes only complete, verified frames.

Ensemble means, spreads and probabilities

GEFS ensemble mean and standard-deviation grids are from NOAA/NCEP GEFS (30 perturbed members in the published products), at native 0.5° resolution. ECMWF IFS ENS mean and standard-deviation grids (51 contributing members) are from ECMWF Open Data, at native 0.25° resolution, under CC BY 4.0. Source data accessed from the newest published run shown on each map, including available hours while the run is still arriving. Snowline converts units and compresses the original six global statistical grids for display. Additional fields are available on the latest run and retain native grid cells across 66°W–36°E, 31–74°N, covering the UK, Europe and North Atlantic presets. ECMWF surface means are calculated from all 50 perturbed members plus the same-run control forecast (oper/fc since IFS 50r1). Member wind speeds are averaged as scalars. Missing members leave a gap. ECMWF rainfall probabilities are the official 24-hour threshold products, supplied every 12 hours; they are not calculated from the mean. GEFS 10 m wind is the magnitude of the official mean vector. Precipitation and snowfall means are displayed as mm/h over the stated source period; snowfall is liquid-water equivalent, while snow depth is centimetres. These maps are not official NOAA or ECMWF products. Mean temperature is converted from kelvin to Celsius; temperature spread retains the equivalent degree interval. Pressure mean and spread are converted from pascals to hPa. No member data is mixed across runs.

Early UKMO charts

Early temperature, pressure, 500 hPa height, 850 hPa temperature, cloud and precipitation charts use the Met Office public Global Deterministic native NetCDF files. Native grid cells are retained within 66°W–40°E, 30–75°N. These regional charts supplement the full Open-Meteo feed for the same initialization; the subset alone is never labelled a complete model run. Temperature is converted from kelvin to Celsius, pressure from Pa to hPa, cloud fraction to percent and precipitation from metres to mm over verified native accumulation bounds. British Crown copyright 2026, Met Office; data and derived products are licensed under CC BY-SA 4.0. These are Snowline charts, not official Met Office products.

MOGREPS-G global ensemble

Snowline reads the Met Office’s free MOGREPS-G native member files. All 18 members from one initialization contribute to each mean, population standard deviation (SD) and raw member fraction. The global grid is about 20 km; map files retain its native cells across the UK, Europe and North Atlantic (no regridding). Temperatures at 850 and 500 hPa respect below-ground masks. Missing members leave cells unavailable.

Rainfall and snowfall means are rates in mm/h, averaged over the preceding three hours. Snowfall is liquid-water equivalent, not snow depth. Probabilities count members reaching at least 0.1 mm/h over that same three-hour period (0.3 mm in three hours). They are uncalibrated member fractions, without time-lagging or neighbourhood smoothing. Display interpolation can show intermediate percentages. The latest three generated runs are retained; each layer exposes only verified frames and its actual period.

Location charts show the same 18-member means, temperatures at native three-hour instants and precipitation rates held over each three-hour period. Location snowfall uses a 7:1 snow-to-water depth equivalent for comparison with Open-Meteo. Ensemble means are excluded from deterministic snowfall agreement. MOGREPS-G data and Snowline’s derived products are © Crown copyright 2026, Met Office, under CC BY-SA 4.0. Shared source indexing is published by Snowline Ensembles.

G versus UK: MOGREPS-G provides global, longer-range context. MOGREPS-UK resolves local weather more finely, at roughly 2 km over the UK. They are separate ensemble systems; Snowline’s existing UK choice retains its own run and member-count rules below.

MOGREPS-UK precipitation and snowfall probability

MOGREPS-UK (2 km, 3 members per hourly run) precipitation probability comes from Open-Meteo’s Met Office ensemble feed. It counts members reaching at least 0.1 mm/h of total precipitation water equivalent, averaged over each native interval.

Snow risk is calculated by Snowline from the Met Office’s public MOGREPS-UK member snowfall files. A member counts when snowfall reaches at least 0.1 mm/h of liquid-water equivalent averaged over the file’s actual accumulation bounds (0.1 mm in one hour). This is the raw member fraction for falling snow, not a calibrated probability, a settling-snow forecast or a snow-depth estimate. No fixed temperature thresholds, time-lagging or neighbourhood smoothing are used. Incomplete-member cells remain missing.

UK snow probabilities are reprojected by nearest cell from the provider’s ellipsoidal LAEA grid onto the renderer’s spherical 2 km LAEA grid. Display interpolation can show intermediate percentages. Precipitation and snow charts always share the displayed initialization; a newer cycle can await snow processing. Only verified frames are exposed, with their true accumulation periods, and three generated cycles are retained for comparison. UKMO data and Snowline’s derived probabilities are © Crown copyright 2026, Met Office, under CC BY-SA 4.0. Source access is refreshed by the scheduled publisher; the map states the model run used. These are Snowline products, not official Met Office or Netweather snow-risk charts.

Terrain shading

Mapzen Terrain Tiles, hosted by the AWS Registry of Open Data, accessed 9 October 2026. Terrain is shown as shaded relief beneath the weather colours. Terrain data sources, required attribution and licence details.

Additional GEM and MOGREPS-G comparison layers

GEM upper-air temperature, humidity, height and wind use native ECCC GDPS data under the ECCC licence linked above, over the same 66°W–35°E, 30–75°N coverage. MOGREPS-G height, scalar wind-speed and pressure statistics use all 18 members from the displayed cycle; missing members leave a gap.

500 hPa height anomalies compare the MOGREPS-G ensemble mean with NOAA PSL NCEP/NCAR Reanalysis 1 daily 1991–2020 climatology. Values are in metres. The climatological baseline is interpolated to the native forecast grid; this is not a hindcast-calibrated model anomaly. February 29 uses the average of February 28 and March 1. The Met Office forecast and derived anomaly retain CC BY-SA 4.0 attribution.

Derived upper-air diagnostics

Snowline calculates 850 hPa wet-bulb potential temperature (theta-w) from temperature and relative humidity using Bolton (1980) equivalent potential temperature and the Davies-Jones (2008) approximation. Values are expressed in degrees Celsius at 1000 hPa, with relative humidity interpreted over liquid water. Calculations are checked against MetPy. This is a derived diagnostic, not an official Met Office modified analysis.

300 and 400 hPa relative vorticity uses native same-time winds, centred differences, actual geographic grid geometry and the spherical metric term. No planetary vorticity is added. A separable Gaussian filter (20 km standard deviation, truncated at three standard deviations) reduces grid-scale noise and striping from rounded source winds. This softens small features and peak values; shading, contours and point readouts all use the same smoothed field. Extra neighbouring data prevent smoothing seams between map tiles; missing data are never filled. Longitude smoothing uses the Mercator latitude limit near the poles. The palette is inspired by the supplied Met Office example, not an exact reproduction of the official colour scale. UKV earth-relative wind speed/direction is converted to eastward/northward components before differentiation on its Lambert azimuthal equal-area grid. Values are in 10⁻⁵ s⁻¹; positive is anticlockwise. Missing neighbours, domain boundaries and poles are left blank. Only models with verified inputs and supported wind conventions expose these layers.

UK Met Office data and Snowline's UKMO-derived products are provided under CC BY-SA 4.0, via Open-Meteo. UKMO material is © Crown copyright. Other model data retains its original licence.

Winter Analysis: type, intensity and precipitation symbols

Type and intensity use the published spatial model files, not an image reconstruction or a point forecast. Each calculation stays on one native surface grid and model run. IFS 0.25° uses its native valid-time precipitation-type code. Other supported models estimate interval type from total precipitation and snowfall water equivalent: snow fraction ≤10% is rain, ≥90% snow, otherwise rain–snow mix. Mixed amounts over an interval do not prove simultaneous sleet. Estimates cannot diagnose freezing rain or hail. Native freezing rain/drizzle and ice pellets retain separate colours; missing data is never treated as dry.

Light to dark blue (rain), green (mix) and pink (snow) represent 0.1–<0.5, 0.5–<2, 2–<5 and ≥5 mm/h, calculated as total water equivalent divided by the native reporting interval. This is interval-average intensity, not an instantaneous rate or snow depth. Categories use nearest native cells, without smoothing. Grey indicates unresolved phase or inconsistent source amounts.

The combined chart adds precipitation symbols on a dense 16-pixel lattice and 4 hPa MSLP isobars to derived 850 hPa theta-w. The lattice samples the nearest native cell and does not increase model resolution. Small dots grow into larger rings as dynamic / large-scale precipitation increases through the same four rate bands. Convective precipitation uses compact downward V shapes for light showers and closed triangles for stronger showers. Both components remain plotted when they coexist. ICON supplies independent large-scale rain, convective rain and total snow (stars); total snowfall cannot be separated by mechanism.

GFS and ECMWF AIFS supply total and convective precipitation. Dynamic precipitation is total minus convective, on the same native grid and interval. Both components include rain and snow, so total snowfall is neither subtracted again nor assigned to a mechanism. ECMWF AIFS uses cp; GFS uses interval-integrated CPRAT. See ECMWF precipitation definitions and the NOAA GRIB definitions. Convection exceeding total by at most 0.11 mm is clipped for provider rounding; missing, negative or larger inconsistencies show total as a ring without diagnosing its mechanism.

Continuous rings show large-scale precipitation where a native split is available, or total precipitation where it is unavailable; triangles indicate supplied convection. UKV and IFS 0.25° currently lack the required convective field in this feed. Native or estimated phase still determines their ring colour, with grey for unresolved phase. Neither CAPE nor precipitation strength is used to invent a mechanism. Rings do not diagnose a front or establish that convection is absent. GFS combines its same-run 0.13° surface and 0.25° upper-air feeds.

Field definitions: Open-Meteo model readers; ECMWF feed documentation. Provider unit conversion, deaccumulation and grid processing precede Snowline’s rendering. These are model-based visualisations and labelled diagnostics, not untouched raw GRIB fields.

HARMONIE sea-level pressure

Pressure is from KNMI HARMONIE Cy43 P3, licensed CC BY 4.0. Native sea-level pressure is converted from Pa to hPa and matched to the displayed initialization and forecast hour. It replaces missing pressure in the Open-Meteo spatial export; the native rotated grid is preserved.

Minimum and maximum temperature anomalies

MOGREPS-UK (18 members) or ECMWF ENS (51 members) ensemble mean of complete member 24-hour extrema, compared with HadUK-Grid v1.3.2.0 1991–2020 monthly normals of daily minimum and maximum temperature. Existing early forecast windows end at 09 UTC; maximum normals use the start date and minimum normals use the end date. Longer-range ECMWF charts use labelled 00–00 UTC calendar days, with both normals taken from the starting date. Monthly reference, not an exact daily NCIC climatology; no model bias correction. UK land uses HadUK-Grid; the Republic of Ireland uses Met Éireann 1991–2020 monthly extrema normals. Existing valid HadUK cells take precedence, including cells on the border.

Contains Met Office data © Crown copyright. HadUK-Grid is licensed under the Open Government Licence v3.0. Source: CEDA archive v1.3.2.ceda, version v20260512, downloaded 21 October 2026. The 1 km reference is mapped to native MOGREPS-UK cells, or averaged over UK land within each published ECMWF 0.25° cell, before calculating anomalies. ECMWF retains complete 09–09 UTC windows through the three-hourly range, then uses complete native six-hour extrema to extend 00–00 UTC charts through the last complete day within T+360 for 00/12 UTC runs. No interval is split or interpolated. Transition windows overlap by nine hours; the later calendar-day definition differs from the 09–09 observation convention. Missing or inconsistent reference cells remain missing.

Irish reference: Met Éireann 1991–2020 Temperature Grids, V2, 1 km Irish Grid TM65, downloaded 22 October 2026. Copyright Met Éireann. Source: met.ie. This data is published under a Creative Commons Attribution 4.0 International (CC BY 4.0) licence. Met Éireann does not accept any liability whatsoever for any error or omission in the data, their availability, or for any loss or damage arising from their use. Snowline has mapped the reference to the model grids and subtracted it from ensemble forecasts. No sea or missing source cells are extrapolated. Differences in station networks, observation practices and gridding between the two reference datasets can cause a discontinuity at the border. These monthly observation-relative anomalies are not model hindcast anomalies or exact daily normals.

ECMWF IFS 0.1°

Source: ECMWF open data, licensed CC BY 4.0. The deterministic forecast is distributed on a regular 0.1° global grid, separately from the O1280 surface feed labelled IFS 9 km and the 0.25° feed. Snowline reads the original surface and pressure-level GRIB messages, retains all supplied forecast times, converts units, and calculates the labelled diagnostics. This is a visualisation and processing of ECMWF data; ECMWF does not endorse Snowline.

All 14 published pressure levels are available: 1000, 925, 850, 700, 600, 500, 400, 300, 250, 200, 150, 100, 50 and 10 hPa. Native geopotential and geopotential height describe the same quantity; charts use geopotential height. Native relative humidity uses mixed-phase saturation. Theta-w instead uses temperature and specific humidity at 850 hPa to calculate vapour pressure and Bolton equivalent potential temperature, followed by the Davies-Jones approximation at 1000 hPa. Native pressure vertical velocity is already Pa/s and is displayed in 0.1 Pa/s. Native vorticity and divergence are displayed in 10⁻⁵ s⁻¹; the separate wind-derived vorticity charts retain their 20 km smoothing. Pressure surfaces may lie below high ground.

Total precipitation, convection, snowfall and runoff are differences of cumulative amounts from the same run over the preceding 3- or 6-hour interval. Radiation and surface stresses are cumulative differences divided by interval seconds. Snow depth uses snow water mass divided by snow density; no fixed snow-to-liquid ratio is imposed. Extrema and gusts retain their native reporting periods and are not described as daily or hourly products. Missing inputs stay unavailable. No finer forecast times, unavailable levels, or coarser replacement fields are fabricated.

The latest three run inventories are checked automatically by the existing independent feed scheduler. Global arrays are downloaded by selected field as byte ranges from ECMWF, then decoded in a worker. A modern browser supporting WebAssembly memory64 is required. The decoder loads on first use (about 1.7 MB compressed); selecting a field can download several megabytes per forecast time. Files remain subject to ECMWF’s rolling retention and availability. Coverage and verification record.

GRIB decoding uses ECMWF ecCodes and meri-imperiumi/eccodes-wasm, Apache 2.0. Wrapper package 2.48.2, runtime ecCodes 2.48.1. The pinned WebAssembly download is integrity checked. Provenance and modifications · Licence.