WxSlice - General Information

  • Purpose

    WxSlice delivers a custom slice of NOAA's Global Forecast System (GFS) forecast for the places, dates and weather parameters you choose, in the file formats you need. It is for people and organizations who want only the GFS fields and area they use, rather than the full global forecast, and who need those files in a format they can work with. After a one-time registration, you complete five short form pages to pick locations, model cycles and forecast hours, any of 91 parameters (or from more than a dozen ready-made presets such as Aviation, Fire Weather or Wind Energy, all in both Lite and Full versions), and formats such as CSV, Excel, JSON, XML, GRIB2, SQL, KMZ, MP4 or PNG. We email you a quote; once you approve the order, each forecast cycle is delivered by email link, through the Download Portal, or by SFTP. This page explains how an order works and what to expect from the files.

    If you're ready to get started, you can register for WxSlice (a one-time step), then start your order.

  • Pricing
    Plan Monthly Annual Included usage Overage discount
    Pay as you go $0 — $0 none ($1 minimum per order)
    Starter $10 $100 $20 (2× the monthly fee) 20%
    Standard $25 $250 $75 (3×) 30%
    Premium $50 $500 $200 (4×) 40%
    Pro $100 $1,000 $500 (5×) 50%

    List price is figured in this order:

    1. Grid-point charge at $0.00001 per grid point per forecast hour per cycle.
    2. Large-area discount for orders above 2,500 grid points (every United States single-state area, excluding Alaska, and about 80% of countries, stays at the full rate). Points above 2,500 are counted as 2,500 × (points / 2,500)^0.6.
    3. Output formats: the first non-premium format is included. Each additional non-premium format adds 2%.
    4. Premium formats (KMZ, MP4, PNG), after the large-area discount: KMZ +10%, PNG +35%, MP4 +50%. These replace the 2% step for those formats.
    5. Parameter-levels: eight are included; each additional level adds 0.8%.
    6. Priority: Medium is the base rate; other priorities adjust by 5% per step.

    If you have a data plan, your monthly allowance is applied to that list price first, then the plan discount applies to any remainder. Quotes may include a plan comparison (suggestion only) that estimates the whole-month cost of this same order on each plan. That comparison does not change your plan.

    Orders whose billed amount is over $1,000 are reviewed before they are approved.

  • When a cycle becomes available

    GFS runs at 00, 06, 12 and 18 UTC. We begin fetching a cycle as soon as it is available, which is about 3.5 hours past that run time, when NOAA begins publishing the opening files (the forecast-hour 0 inventories). A backup check also runs at 03, 09, 15 and 21 UTC. The cycle then has to be loaded and written in your formats, so the email that the files are ready comes after that work finishes. First files are usually ready about 6 hours after the cycle time. Each order's quote email includes its own delivery estimate based on current workload and file availability. You can pay a little more for higher priority processing, or a little less for lower priority processing.

  • File naming and layout

    Your order has a folder named from the order name you entered (spaces and punctuation become underscores). Inside it, each UTC run date has a folder (YYYYMMDD). Cycle files are named <OrderName>_<YYYYMMDD>_c<HH> plus a format suffix, where HH is the 00, 06, 12 or 18 UTC cycle. Files for one location add the location name: <OrderName>_<Location>_<YYYYMMDD>_c<HH>....

    A JSON or CSV cycle bundle holds one file per location. GRIB2 members also include the forecast hour, such as ..._c12_f006.grib2. Excel workbooks are named ..._table.xlsx or ..._grid.xlsx, one workbook per location, inside the cycle's Excel zip. For JSON and XML orders, the information file sits in the order folder next to the dated folders, not inside each cycle archive. All times in file names and in the data are UTC.

  • How file size scales

    Size grows with the area (how many 0.25° grid points are in the box), the number of forecast hours, and the number of parameters and levels. Choosing every 2, 3, 4, 6 or 12 forecast hours on the Date page, instead of every hour, reduces that hour count and so the size and the price. A longer date range, more daily cycles, or more output formats produces more files and more bytes. Excel Grid and GRIB2 can be especially large for a wide area. Download promptly: each cycle's files are kept for up to 7 days after they are ready, or 2 days after your first download of that cycle, whichever comes first.

  • GRIB2 longitudes

    In CSV, TSV, PSV, JSON, XML, SQL and Excel, longitudes are written from -180 to 180 (degrees east). If a stored value is greater than 180, 360 is subtracted so it stays in that range. Latitudes are -90 to 90, on the 0.25° GFS grid.

    GRIB2 is cut from the original NOAA files, not rebuilt from the tables. Longitudes in the delivered GRIB2 files are on the native 0–360 GFS grid, even when the order was entered with west longitudes. For example, a box from 88°W to 87°W is written in GRIB2 as 272 to 273. A box that crosses the 180° line is cut as one continuous 0–360 grid (for example 179°E to 179°W is 179 to 181), not split into two pieces.

  • Parameter caveats

    GFS forecasts are hourly out to 120 hours and every 3 hours after that, through 384 hours (16 days). We only deliver the hours and records the model actually published; hours the model does not have are not charged.

    On the Date page you can also choose how often to take a forecast hour: every 1, 2, 3, 4, 6 or 12 hours (the default is every hour). The last hour you set must land on that step — for example, 0 to 12 every 3 hours gives 0, 3, 6, 9 and 12. Your quote, confirmation and delivery estimate use only those hours, so a wider step means fewer records, smaller files and a lower price. We still only deliver and charge for hours the model actually published. When the last forecast hour is after 120, every 2 and every 4 hours are not available, because GFS is only 3-hourly after hour 120 — choose every 1, 3, 6 or 12 hours, or a stop hour of 120 or less.

    Some fields are land-only and have no value over water. That includes soil moisture (SOILW, SOILL and TSOIL), WATR (water runoff) and CPOFP (probability of frozen precipitation). Those cells are missing in the data (null in JSON, omitted in XML, blank in Excel) and, on maps, the colour stops at the coast.

    Accumulated precipitation (APCP and ACPCP) is not present at forecast hour 0. NOAA's own inventory starts those fields at hour 1, as a 0–1 hour accumulation. We do not invent zeros or raise an error for hour 0; those records are simply omitted, and hour 0 still includes the other requested fields, such as temperature. Instantaneous precipitation rate (PRATE) is a different field and can appear at hour 0.

  • Excel Table and Grid

    Both Excel layouts write one workbook per location and cycle, packed in a zip for that cycle. Each workbook starts with an Index sheet that links to the data sheets. Times are UTC, shown as yyyy-mm-dd hh:mm:ss.

    Table is one sheet per parameter and level (for example TMP_2m), with the same columns as the CSV, in the same row order. If a sheet would go past Excel's 1,048,575 data rows, it continues on _2, _3 and so on.

    Grid is one sheet per forecast hour, parameter and level (for example f006_TMP_2m), with longitudes across the top and latitudes down the side, north to south. If that hour has more than one time range for the same field (for example two APCP accumulations), each one gets its own sheet. A Grid sheet cannot exceed Excel's 16,384 columns or 1,048,576 rows; an area with too many longitudes or latitudes will not export in Grid form.

  • SQL files (database and table names)

    When you choose a SQL format (MariaDB, MySQL, PostgreSQL, SQLite, SQL Server or Oracle), the order form shows a short panel for the database and table names used in those files. The database name is optional (it is labelled Schema if Oracle is the only SQL format you chose). Leave it blank and the scripts will not include a USE statement or a schema prefix. The table name defaults to wxslice_order if you leave it blank. Each name must start with a letter and may then use letters, digits or underscores, with no spaces, and is at most 64 characters.

    Each format is written in its own SQL dialect. In the order folder, next to the dated folders, you get one schema file per format you chose, named <OrderName>.mariadb.create.sql (or .mysql.create.sql, .postgresql.create.sql, and so on). That file creates the table if it does not already exist. If you entered a database name, it also includes a commented-out CREATE DATABASE or CREATE SCHEMA line you can use if you need it. The data for each location and cycle arrives in the cycle bundle as INSERT statements only — for example OrderName_Location_20261118_c12.mariadb.sql — so run the schema file once before loading the cycle files. If you entered a database name, MariaDB, MySQL and SQL Server files include a USE statement for it; PostgreSQL and Oracle use it as the schema name on the table, and SQLite mentions it only in a comment.

  • Order information file (JSON and XML orders)

    Every JSON or XML order includes one information file in the order folder, next to the dated folders:

    • <OrderName>.json.meta.json for JSON orders
    • <OrderName>.xml.meta.xml for XML orders (the same content in XML form)

    It describes your data files: the columns, each parameter's units, the locations, and which levels and forecast times are included. The data files themselves contain only data. This file is not packed inside each cycle archive.

    For new XML orders, the data files use the root element <wxslice_order>, or the database name you entered on the order form if you provided one. The information file uses the root element <wxslice_order_meta>. Existing XML orders keep the names they were first written with (<gfs_order> and <gfs_order_meta> when those names were left blank). The table name on the form is used for SQL and JSON tables and, if you enter one, as the XML row element; it does not rename the XML root. New orders also default to Imperial units and zip compression if you leave those choices as they are.

    Data files at a glance

    • Times are UTC, written YYYY-MM-DD HH:MM:SS (for example 2026-10-12 06:00:00). run_dt is the GFS model run, and forecast_dt is the time the value is valid for.
    • Longitude runs -180 to 180 (degrees east) and latitude -90 to 90 (degrees north), on the 0.25° GFS grid. A box that crosses the 180° line can list a longitude_max above 180 in bounds.
    • A value the model could not provide is null in JSON. In XML, the value attribute is left out and the row has nil="true".
    • Units depend on the unit system chosen for each location (original GFS units, metric or imperial). See units under each parameter.
    • Rows are sorted by forecast time, parameter, level, time range, latitude and longitude.

    Fields

    Field What it is
    format_version Version of this file's layout (currently "1.0")
    generated_utc When the file was created (UTC)
    attribution, source Credit line and data source (NOAA GFS)
    order, order_name, order_id Your order's file name, the name you entered, and its order number
    unit_systems Unit systems used in this order
    files File name pattern, bundle type (tar or zip), compression (gzip, zstd or zip), and the companion table and information files
    layout, sort_keys How each data file is laid out, and its row order
    time_zone, time_format Always "UTC" and "YYYY-MM-DD HH:MM:SS"
    missing_value, missing_value_note How a missing value appears
    forecast_hours First and last forecast hour in the order (0 = the analysis)
    columns Each data column: name, type (string, datetime or number), whether it can be empty, and a description
    locations Each location: id (as in the data), display name, type, unit system, center point and the grid box delivered
    parameters Each parameter: GFS abbreviation, name, GFS unit, your unit, levels, time ranges, forecast hours and the exact level/time-range combinations included

    Location types

    • Location: a named place, such as a state or country. The center is the place's center.
    • Address: a street address or point. The center is that point.
    • LatLon: a latitude/longitude box you entered. The center is the middle of the box.
    • Global: the whole Earth. There is no center (null).

    bounds is the grid box delivered, aligned to the 0.25° grid.

    Levels and time ranges

    A parameter's levels and stats list everything it includes, but not every level has every time range. For example, temperature averages exist only at the cloud-top levels. combinations gives the exact pairs: each entry lists levels and time ranges that all appear together in the data, and nothing else.

    Example (shortened)

    {
      "format_version": "1.0",
      "generated_utc": "2026-11-09T18:07:49Z",
      "attribution": "Custom Visuals, LLC - https://www.customvisuals.com",
      "source": "NOAA NCEP Global Forecast System (GFS), 0.25 degree grid",
      "order": "KMZ_and_MP4_01",
      "order_name": "KMZ and MP4 01",
      "order_id": 18,
      "unit_systems": ["imperial"],
      "files": {
        "data_pattern": "<YYYYMMDD>/KMZ_and_MP4_01_<YYYYMMDD>_c<HH>.tar holding KMZ_and_MP4_01_<Location>_<YYYYMMDD>_c<HH>.json.gz per location",
        "bundle": "tar",
        "compression": "gzip",
        "rollup": "cycle",
        "table": "KMZ_and_MP4_01.json.create.sql",
        "meta": "KMZ_and_MP4_01.json.meta.json"
      },
      "sort_keys": ["forecast_dt", "parameter", "level", "stat", "latitude", "longitude"],
      "time_zone": "UTC",
      "time_format": "YYYY-MM-DD HH:MM:SS",
      "missing_value": null,
      "forecast_hours": {"first": 0, "last": 47}
    }
    

    (The lists are shortened. A real file also lists every column, location, parameter, level and time range.)

    Version policy

    The information file and the XML data files use format_version as a major.minor number. It is currently "1.0". We will not change this layout without changing format_version:

    • Adding a field, or a new value of an enumerated field such as location_type or compression, raises the minor number (1.0 to 1.1). Existing fields keep working, so please have your program ignore fields it does not recognize.
    • Renaming, removing or changing the type or meaning of a field raises the major number (1.x to 2.0). We will announce that kind of change in advance.

    New parameters, levels or unit texts are values, not a format change. The format_version attribute on the XML data root follows the same number as the information file.

  • Historical and backfill orders

    You can order dates from 23 March 2021 through today plus two years, with "today" in America/Chicago. The public GFS archive starts on 1 January 2021, but the order form does not accept dates before 23 March 2021, when the current file layout and hourly forecasts out to 120 hours began. Forecasts after hour 120 are 3-hourly, and nothing is available past hour 384.

    Current cycles are processed first. Older, historical cycles are filled in afterward, in limited batches, and that work pauses if a new current cycle becomes ready. Historical delivery is therefore slower than live cycles.

  • Unclear addresses

    Addresses and places are geocoded. If we cannot find a unique place — the lookup fails, only a partial match is found, the city name does not match what you typed, or we would have to rely on a fallback guess — the order is held for review. We email you to ask for a corrected street address or a nearby landmark we can find, and no quote is sent until every location is confirmed. Named areas from our lists and latitude/longitude boxes you type are not held this way.


© 2003-2026      Custom Visuals, LLC      Privacy Policy