STORM SWEEP

Level II products at every tilt

Reflectivity, velocity, storm-relative velocity, spectrum width and, on any volume the radar recorded them in, differential reflectivity, correlation coefficient and specific differential phase. Storm-relative velocity is dealiased, and the motion subtracted comes from the radar’s own storm tracks or, failing those, the HRRR mean wind; KDP is computed on your PC from the raw phase data. A tilt strip under each of those panels lists the elevation angles in the newest volume so you can climb through a storm. The full set needs a NEXRAD: about 45 of the roughly 200 sites in the list are the FAA’s airport TDWRs, which publish only reflectivity and velocity at their three lowest tilts; there, spectrum width and the three dual-pol products give way to the national mosaic, and storm-relative velocity falls back to that plain velocity, which the radar never dealiased.

One, two or four panels, one timeline

Split the workspace into 1, 2 or 4 panels, each with its own product and tilt, all stepping through the same moment. Link them so panning moves everything, or unlink to compare different parts of a storm. Live, that timeline holds the newest six volumes: about 25 to 35 minutes on a NEXRAD, and about six minutes on a TDWR, where each roughly one-minute scan counts as a volume. Anything further back is the archive replay’s job.

Archive replay back to June 2008

Pick any past date and local time at a NEXRAD site with a 1, 2 or 3-hour window, or load one of five built-in cases: Moore 2013, El Reno 2013, Joplin 2011, Rolling Fork 2023, Greenfield 2024. The warnings in force at the middle of that window, and the storm reports filed within 600 km of the radar across it, come back with it from the Iowa Environmental Mesonet archive. Storm tracks, tornado vortex signatures and mesocyclone detections come back only for cases from 2020 on, as far back as the Level III archive Storm Sweep reads goes, so Joplin, Moore and El Reno replay without them, and the app says on screen that the tracks are missing.

Where a replay differs from live

A few things work live only. The hail size estimate and severe-hail chance need that day’s freezing level, so those panels stay blank. NOAA’s hail class is a single image of the station’s latest scan, live only, and it does not step with the loop. Storm-relative velocity falls back to plain velocity, because a replay works out no storm motion to subtract, and live it does the same whenever neither the radar’s storm tracks nor the HRRR mean wind can supply one; outside HRRR’s lower-48 grid the model cannot, so the Alaska, Hawaii, Puerto Rico, Guam and overseas sites depend on the radar’s own tracks alone. Each of those carries a panel note saying so.

Dual-pol is only there if the radar had it: NEXRAD was upgraded site by site between 2011 and 2013, so a case from before its site’s upgrade carries no differential reflectivity, correlation coefficient or specific differential phase, and those panels come up empty with no note to explain it. A replayed case’s warnings come back from the archive as their tornado, hail, wind and damage tags plus a link to the issued text, so the statement pane holds no narrative and no what-to-do instruction.

NOAA’s national mosaic loop has no archive history either, so a panel left on the mosaic (or on hail class, which gives way to it) has nothing to draw over a case. A window holding more than forty volumes is loaded as at most forty, spread evenly across its length, and the status line says how many of how many it took.

Volume-derived products

Echo tops at 18 and 50 dBZ, vertically integrated liquid, VIL density, hail size estimate and severe-hail chance, all computed from every tilt of the whole volume and drawn like any sweep, out to 230 km, half a reflectivity sweep’s reach, because past that the lowest beam overshoots the storm. They need a full NEXRAD volume, at least five tilts, so on a TDWR they give way to the national mosaic instead. Each of those panels is labelled a Storm Sweep estimate, computed here from the volume rather than issued by the Weather Service.

Hail estimated on your PC

NOAA stopped publishing its hail index to the public archive in 2022, so Storm Sweep runs the published NWS Hail Detection Algorithm itself (probability of hail, probability of severe hail and maximum expected size) using current freezing and −20 °C heights from HRRR’s lower-48 grid. At sites outside that grid (Alaska, Hawaii, Puerto Rico, Guam and the overseas base radars), those estimates are not computed at all. Labelled as a Storm Sweep estimate wherever it appears; NOAA’s own hail class is a separate panel and is not this estimate.

Warnings you can actually read

Tornado, severe thunderstorm, flash flood and the other polygon alerts the Weather Service marks severe or extreme, drawn on every panel with the most severe on top. The overlay goes by that severity rather than by product, so lesser advisories are not part of it, and alerts issued by zone code carry no polygon to draw; the Alerts tab lists those, and can send any alert that does have a polygon over to the radar. Click a live warning’s polygon and the full statement opens beside the map, including the what-to-do instruction.

A home marker you place yourself

Right-click the map, or press H, to place a home marker, then move or remove it from the same menu. It shows on every radar panel and on the Current Weather map, and is kept on this PC only, separate from your saved place.

Colour tables, with colour-vision versions

Reflectivity opens on Storm Sweep’s own table, with the NWS standard scale one click away; velocity, ZDR, CC, KDP and spectrum width have Storm Sweep tables too, and every built-in Storm Sweep table has red-green and blue-yellow friendly versions. Warnings and storm reports open on click.

A hint under every product

The product lists in the Radar panel and in both Storm 3D pickers show a one-line plain-English hint under each product, such as “Whether targets are wider than tall: big flattened drops read high” for ZDR. Hail sizes and chances are labelled as Storm Sweep’s estimate.

Radar station status

When the radar you chose is running late the station turns orange, and when it is offline it turns red. A “Use nearest working radar” button switches you to the closest site that is reporting. The “Parts of the radar loop didn’t load” banner now appears only when several map tiles fail, not for one stray tile.

Times in your own clock

The Models time card, satellite frames, soundings, the radar legend and Storm 3D show your local time with your time zone, for example 3:00 PM CDT, and model runs are named by your local hour, such as “1 PM run”. Model runs and soundings keep their short UTC name (such as 18Z) in small print, satellite frames carry the UTC scan time on a line below, and the Time section of Settings can add UTC beside the rest.

Built from the volume you pick

The 3D view uses the same Level II volumes as the radar tab, so what you rotate is the storm you were just watching, not a separate or delayed product. What it rebuilds is a box 60, 120 or 200 km wide and 18 km tall, centred on the strongest echo, on the radar, on a storm cell you pick from the list or on any point you double-click, not the radar’s whole range. A picker lists the volumes by time, site and tilt count, so any step of the loop can be rebuilt, not just the newest; the loop is the radar tab’s own, and Storm 3D opened by itself holds only the newest volume until you have been watching Level II there. One radar at a time. There is no 3D version of the multi-radar composite.

Isosurfaces you choose

The peel slider sets the weakest echo drawn, from 10 to 65 dBZ in 5 dBZ steps. Storm Sweep builds a shell at that value and at every 5 dBZ above it, up to twelve nested surfaces in the app’s own reflectivity colours, with the strongest one present drawn solid as the core and marked that way in the legend; a shell too large for one mesh is drawn in part, and the status line says so and suggests a smaller box. Drag right to strip the outer layers off down to that core, and use the separate solidity slider to set how see-through the layers around it are, from ghostly to nearly solid.

Three view styles, and a Lighting panel

Draw the storm as Layered shells, a Single surface (with a “Surface at” slider) or a Glow cloud, which needs WebGL 2. A Lighting panel sets whether the light follows your view or the map, its direction and height, and the shading from soft to sculpted, and your choices are saved. A cutaway option opens the storm on your side of the slice.

Velocity and dual-pol, not just reflectivity

A “Show” picker switches the shells between Reflectivity, Velocity, ZDR, CC, KDP and Spectrum width, each with its own control: speed for velocity (inbound green, outbound red), ZDR, CC with a reflectivity mask, KDP and width. A separate “Cross-section shows” picker chooses the field for the vertical slice. KDP is worked out by Storm Sweep where correlation is high and the echo is 20 dBZ or stronger, and is labelled that way.

Slice it where you want

Turn on the section tool and click two points: on the ground or on the storm itself. Storm Sweep cuts a curtain of the field you choose (reflectivity, velocity, spectrum width or a dual-pol field) from A to B through the volume, ground to the top of the box, and draws it both inside the box and in a panel across the bottom of the view, with a height axis every 10,000 ft and range in miles from A. One straight section at a time, and if the volume carries no velocity the app says so instead of drawing something.

Storm cells as columns

The radar’s SCIT storm cells stand in the box as columns rising to the height where reflectivity drops through 30 dBZ, or to a default 8 km where the volume never reaches the storm’s top. A column is cyan, amber for a mesocyclone, red for a tornado vortex signature, and carries a tag with the cell id, the direction and speed where the radar reports the cell a motion, the measured top in feet above the radar antenna, and, only where the radar tab has already computed them, a peak dBZ and hail figures. The cells, their tracks and the mesocyclone and tornado-vortex-signature flags are the radar’s published Level III storm products: live, they are drawn on the newest volume, and products more than half an hour old are left out. The hail figures are the radar’s own hail index wherever it still publishes one. Where it does not, those figures and the peak dBZ beside them are Storm Sweep’s own run of the published NWS Hail Detection Algorithm over the volume, which needs the HRRR model’s freezing levels: that estimate covers the lower 48 only, and a cell closer than 5 km or farther than 230 km from the radar gets none.

Up to 25 products, plainly described

GeoColor, all sixteen ABI bands, the RGB composites meteorologists actually use (Air Mass, Sandwich, the combined Day Cloud Phase / Night Microphysics among them) plus Derived Motion Winds. The picker lists the products NOAA STAR publishes for the area you have chosen, so smaller sectors offer fewer than the full set. Each one carries a plain description of what it shows, on screen beside the picker.

Zoom and pan the image

Scroll-wheel zoom up to ten times at the pointer, drag to pan, double-click to fit. The app asks STAR for a middle size rather than its largest (2500 pixels wide on CONUS, 1808 on the full disk), so a long loop fits in memory; where a product is published only at sizes far from that target, as Derived Motion Winds is, the smallest size above it is taken instead, which on CONUS is 5000 pixels wide. Longer loops are decoded smaller again.

Full disk, CONUS, mesoscale and regional

Switch between the full disk, the continental US, NOAA’s two mesoscale sectors and STAR’s regional sectors: 22 regions on GOES-East, 12 on GOES-West. Loop length is yours to set at 12, 24 or 48 frames, and a new ABI image arrives every minute on the mesoscale sectors, every five on CONUS and the regions, every ten on the full disk. Derived Motion Winds is the exception, published hourly whatever the area.

Loop controls and frame times

Play and pause, step a frame at a time, or drag the timeline. A speed slider runs from 1 to 15 frames per second, and the newest frame is held briefly before the loop wraps. The panel over the image gives the scan time of the frame you are looking at in UTC and in local time, its date, how old it is and its place in the loop. While the tab is on screen the app re-lists STAR every two minutes on the mesoscale sectors and every five elsewhere, picking up new scans on its own; off screen, playback and refresh stop.

HRRR, RRFS, GFS and ECMWF IFS

HRRR and RRFS at 3 km, both updating hourly. RRFS is NOAA’s new rapid-refresh system, pre-operational until it takes over from the NAM and HREF on 3 November 2026, a date NOAA’s service change notice allows to slip. GFS at 0.25° and ECMWF IFS open data at 0.25°. How far a run reaches depends on its cycle: HRRR and RRFS go 18 hours ahead, and 48 and 84 respectively from the 00, 06, 12 and 18 UTC runs; GFS reaches 16 days; ECMWF is three-hourly to six days, with the 00 and 12 UTC runs six-hourly to ten.

Eighteen map fields, not all of them on every model

Simulated radar, CAPE, helicity, updraft helicity, significant tornado parameter (computed here from the model’s own fields), wind and gusts, temperature, dewpoint, precipitation and snowfall among them. No single model carries all eighteen: seventeen fields on HRRR and RRFS, fourteen on GFS, twelve on ECMWF, and the app offers only the fields a given model actually carries rather than faking the rest.

Its own GRIB2 reader, pulling only the bytes it needs

Written for this app: regular lat/lon, Mercator, north-pole polar stereographic and Lambert conformal grids, and every packing scheme these four models use. The one exception is JPEG 2000, which only the marine wave models need. That goes through the open-source CoreJ2K library, credited in the app. The app reads each run’s published index, works out which byte ranges hold the fields being drawn and merges adjacent ones, so a map pulls a few megabytes out of a 150 MB file instead of the whole thing.

The run as a loop, with the next hours already fetched

Play the whole run at one to twelve frames a second, step forward and back a frame at a time with the buttons or the arrow keys, play and pause with the space bar, or drag the slider to a forecast hour. The page opens on the newest run that has finished posting, or on the newest of all when none has; the picker offers the last several cycles and marks how far an incomplete one has got, and any hour that has not landed reads “Not posted yet” until it does. The hour on screen is fetched first and its neighbours next, so moving to a nearby hour usually has nothing left to download, and the shading opacity is yours to set.

Twelve map areas: the 3 km models only inside their own grids, the global ones only inside a window

Eleven fixed areas (the lower 48, eight regional crops, North America and Alaska) plus an “Around me” view centred on your chosen place, with a cursor readout and a legend on every field. HRRR and RRFS are read for the lower 48 only and draw nothing outside it; HRRR Alaska, a fifth model on the Models tab, covers the Alaska area. GFS and ECMWF also reach Hawaii and Puerto Rico, which have no area of their own and are seen through the North America view or “Around me”, but each global run is cut to a window from 10° to 75°N and 170°E (across the date line) to 40°W before anything is drawn, so nothing further afield can be mapped.

Observed soundings: Skew-T and hodograph

A full Skew-T log-P with hodograph and a parameter panel for any of 123 radiosonde sites across the US, Canada, Mexico, the Caribbean, Alaska and Hawaii, and for the newest 60 launches a site has listed, specials among them (about a month at a twice-daily site): CAPE, CIN, LCL, LFC and EL for the surface, mixed-layer and most-unstable parcels, plus shear, helicity, significant tornado and supercell composite. These are observed balloon soundings, and the panel says so. There are no model soundings. The maths comes from published literature, checked against an independent library.

An “Explain this chart” switch labels the parts of the diagram, the parameters carry plain-English names beside the technical ones, and a short tour runs the first time you open it.

SPC convective outlooks, Days 1 to 8

Categorical outlooks on Days 1 to 3, probabilistic on all eight days, with the conditional intensity groups SPC introduced in March 2026 in place of the hatched significant-severe areas, drawn for tornado, wind and hail on Days 1 and 2, and for any severe on Day 3.

Mesoscale discussions, watches and the Area Forecast Discussion

Current SPC mesoscale discussions, WPC mesoscale precipitation discussions, and tornado and severe thunderstorm watches: outlines on the map and the full text beside it, with a watch’s probability table when SPC has published one. The NWS office’s own reasoning, verbatim. The office covering your place is picked automatically, and you can switch to any other office that issues one, or step back through its recent issuances, up to the last 25.

Upper air and surface

850, 700, 500, 300 and 250 mb station plots with height contours and isotherms drawn by Storm Sweep’s own Barnes objective analysis of the plotted reports, labelled as such, not presented as an official analysis. Current North American airport observations as standard station models (temperature, dewpoint, wind barb, pressure, sky cover) with WPC fronts drawn over them.

Rainfall and fire weather

WPC excessive-rainfall outlooks for Days 1 to 5, and quantitative precipitation forecasts for Day 1, Day 2, Day 3, Days 4 to 5, Days 6 to 7 and the 3-day and 7-day totals. SPC fire-weather outlooks cover wind and humidity, and dry thunderstorms, on Days 1 and 2.

Click a point, read the outlook

On the outlook, rainfall and fire-weather maps, clicking anywhere reads out what the layer says at that spot: the risk category on the days SPC draws one, each hazard’s probability, the name of the excessive-rainfall or fire-weather area, or the forecast amount in your chosen precipitation unit on the precipitation layers. It says plainly when the point falls below the lowest contour or category rather than leaving the row blank, and where the click lands inside an intensity area the hazard row carries that area’s group. The outlook tool also opens SPC’s own written discussion beside the map: that day’s text for Days 1, 2 and 3, and the single Day 4 to 8 discussion beyond them.

It says when the data is old

Each tool refreshes on its own interval (four minutes for mesoscale discussions, ten for the outlooks, an hour for upper air) and only while you have it on screen. A refresh that fails leaves the last good data on the map with a banner saying how old it is, rather than blanking the view. If spc.noaa.gov stops answering, the convective outlooks come from NOAA’s map service instead; the note line says so unless it is already reporting nothing outlooked or an ended valid period.

Flight map, FAA delays and winds aloft

Eight toggleable layers with their data ages: METARs shaped and coloured by flight category, PIREPs filtered by altitude and age, SIGMETs and CWAs, G-AIRMETs, temporary flight restrictions, FAA delays, radar and more. The G-AIRMET slider steps through that product’s own forecast snapshots, which are 3 hours apart; conditions between snapshots are not shown. An advisory in effect with no area to draw cannot be plotted at all: the layer note counts those, and the Airport and Route tools list the US ones with their text. Ground stops, ground delay programs, arrival and departure delays, closures and deicing from the FAA’s NAS Status feed, as badges and an airport card. Forecast winds and temperatures aloft, including Alaska, Hawaii and other Pacific regions.

Airport tool

Search a US airport by identifier or name: decoded METAR alongside the raw text, a six-hour trend strip, a colour-coded TAF timeline with every period decoded, runway headwind and crosswind components with gusts, pressure and density altitude, and frequencies. Ten cards in all.

The airport figures Storm Sweep computes

Each TAF period’s flight category is worked out by Storm Sweep from the conditions in force rather than read from the product. The wind components are computed on true runway headings. Pressure altitude is an estimate from the field elevation and the METAR’s altimeter setting, and density altitude adds the METAR temperature; neither is a performance figure. The app tags all three as computed.

Route briefing

Enter a departure, a destination and any airports in between, a cruise altitude, a departure time and a true airspeed. Storm Sweep gathers what the NWS and FAA have published for a 25 nm corridor along the route and assembles seven numbered sections in the order a standard briefing uses: adverse conditions, the forecaster’s synopsis, current conditions in that corridor, pilot reports from the last two hours in it within 4,000 ft of cruise, the destination TAF with the groups covering your ETA highlighted, winds aloft with leg timing, and a NOTAM section that links out to FAA NOTAM Search and Flight Service instead of carrying NOTAMs itself. Each leg’s headwind, ground speed, ETE and ETA are computed from the FB tables and labelled as computed, and where no FB table covers a leg, the briefing says so and times that leg at true airspeed with no wind at all. The footer lists product by product when each one was issued or fetched. It is supplemental, not an official briefing.

Drone check

Twenty-four hours of wind at 33 and 262 feet, gust, temperature, visibility, ceiling and cloud base from HRRR or the National Blend of Models, with precipitation and thunder chances and the gust spread always taken from NBM. HRRR covers the lower 48 and Alaska hour by hour; the Blend adds Hawaii and Puerto Rico, though its Alaska grid gives a row every third hour and its Hawaii grid every second, with the hours between left blank and the card saying so. A point outside both gets a plain coverage sentence instead of a failure. Every hour is shaded day, Part 107 twilight or night, and a sun card lists sunrise and sunset, the Part 107 twilight window, civil twilight at sun −6° and golden hour for today and tomorrow, all computed for the point with the NOAA solar equations and labelled computed. The twilight window shown is the sunrise − 30 and sunset + 30 minute rule that applies outside Alaska; in Alaska the app points instead to the Air Almanac definition of civil twilight under 14 CFR 107.29 and gives the sun −6° times only as a reference. Up to three airport observations within 30 nautical miles of the point you check, nearest first, each decoded alongside its raw METAR and the age of the observation, labelled as observations at the airport, not at your point.

Your limits, your call

Set your own thresholds for wind, gust, visibility, ceiling and temperature. Exceeding hours are highlighted and counted. The app never says “safe”, “unsafe” or “clear to fly”. Those words appear nowhere in it.

Not an official source, and the page that says so

The section’s own disclosure page. A notice you acknowledge the first time Aviation opens, and again whenever its wording changes, says Storm Sweep is not an FAA-approved or official source of aviation weather and does not provide a weather briefing, and that NOTAMs are not included; declining leaves the section.

The map and the models

Thirteen toggleable layers over US waters, each with a data-age badge once that layer has loaded: hazards, 704 bundled marine zones, NDBC buoys, tide and current stations from zoom 5, river gauges from zoom 7, a wave model field, surface currents, sea-surface temperature, lightning strike density, NOAA chart images, the radar mosaic, USACE mile markers and bridges from zoom 10, and your last passage. Those three draw nothing until you have zoomed in that far, and the river gauge and mile-marker switches carry the zoom in their own names. GFS-Wave on west-coast, Atlantic, east-Pacific and global grids, the Great Lakes model and NWPS by forecast office are decoded from GRIB2 in the app. The currents, sea-surface temperature and lightning layers are NOAA nowCOAST rendered images, and RTOFS is decoded from GRIB2 for a point’s water temperature and current rather than for the map.

Click any water

Click any water and get the zone, its active headlines, the forecast text verbatim and the product’s issue time, then a point card built block by block: the NWS NDFD grid strip every three hours out to 48 hours where NWS publishes a grid, which is coastal waters, and beyond them the wave model stands in; wave model guidance and, where the RTOFS regional grids reach and outside the Great Lakes, near-surface water temperature and current, both marked in amber as guidance; the nearest reporting buoy within 60 nm with its distance and observation age; the nearest tide station within 30 nm and current station within 20 nm with their distances and next events; and sunrise, sunset, the twilights and the moon computed in the app. Each block loads on its own, and the grid, wave, buoy, tide and current blocks stay in place and say in plain words when nothing is in range, nothing is published or a source failed, rather than leaving a gap; the RTOFS water temperature and current block is simply absent outside that coverage rather than saying so. On an ocean beach zone it adds NWPS rip current probability, captioned as model guidance and pointing you to the NWS Surf Zone Forecast for the official risk.

Tides and currents

NOAA CO-OPS tide, current and water-level stations. A tide station draws a three-day predicted curve from the station’s own local midnight, with the day’s observed water level over it where the station has a sensor, the difference between observed and predicted, and a highs-and-lows table. Observed levels carry the quality flag CO-OPS gives them: preliminary, meaning not yet quality controlled, or verified. At a subordinate station CO-OPS publishes only the highs and lows, so Storm Sweep draws the curve between them, dashes it on the chart and labels it an estimate in amber rather than passing it off as a published prediction.

Buoys

Search NDBC buoys and C-MAN stations, or list the ones near your last point: the latest observation with its age (wind and gust, significant wave height with dominant and average period, swell and wind-wave components, pressure with its tendency, air and water temperature), each row only where that buoy transmits the field, and the swell and wind-wave split only where NDBC publishes it. NDBC real-time data are provisional. Then five days of wind, waves, period, pressure and temperature as charts. A buoy that has been recovered for the winter says it is not reporting rather than showing nothing.

River gauges

NWS gauges from the National Water Prediction Service: the hydrograph with flood-category bands and the forecast’s issue time, the action, minor, moderate and major stages, the low-water threshold with its navigation impacts quoted as published, historic crests, and the gauges upstream and downstream, each only where NWPS publishes it for that gauge. Many gauges are observed-only, with no forecast, no low-water threshold and no published impacts. River observations may be provisional.

Passage

Type a position (34.0, −77.6) or the name of a place with a CO-OPS tide station, or pick a start, any via points and an end on the map; give a speed in knots and a departure time, and the app lays great-circle legs between those points and samples along them: a timeline of ETA, distance along, zone, wind, gust, waves, period and visibility, every row naming its source: an NWS NDFD grid, wave model guidance in amber, or “no coverage”. A timeline only: no routing and no advice, up to 2,000 nm at 0.5 to 60 kt. The ETAs assume one constant speed through the water, and straight lines may cross land or shoals, so check official charts and Notices to Mariners.

Your limits, your call

Set your own figures for wind, gust, wave height and current, plus a wave period to flag at or below. Values that reach one get a ▲ beside them in the point card and the passage timeline, a highlight only, not advice. Storm Sweep never judges conditions for you, and no go/no-go verdict appears anywhere in the tab.

Twenty-three layers

Everyday: wind, temperature, radar, dew point, humidity, gusts, precipitation, pressure lines and isobars. Severe: CAPE, wind shear, helicity, a tornado index, fronts, and warnings and watches. Aviation & drones: wind at 10 and 80 m. Outdoor: soil temperature, wet bulb globe temperature and UV index, both labelled estimates. Marine: sea surface temperature and waves. Tropical: hurricanes. A docked panel lists “On the map” and “All layers”, with opacity, Solo, Remove and a Pin for each.

Wind streaks you can tune

Streaks drawn from the model’s wind, with Density and Speed sliders from 25% to 200%. One shaded layer shows at a time, so the streaks stay readable on top.

A time bar with a past and a future

Step by the hour from three hours back to twelve hours ahead. Every hour is tagged Observed, Analysis or Forecast: observed hours show NOAA radar, analysis hours show the model’s own hour zero. Play, step, go back to Now, or change speed. Observed radar frames are 5 minutes apart on Local and 15 on Nationwide.

A readout from the nearest station

At Now and in the past, the readout comes from the nearest NWS station’s latest observation: temperature, wind, dew point, pressure, sky, feels like, humidity, precipitation and the station itself. After Now it switches to the NWS hourly forecast, which covers temperature, wind and chance of precipitation only. Hover the map to read the layers that are on.

Pins and a ruler

Up to five pins, each with a 12-hour mini forecast of temperature, wind and gusts, and precipitation, and a ruler (press M) to measure distance. “Keep pins” saves them.

HRRR where it reaches, GFS everywhere else

The map uses HRRR over the lower 48 and GFS beyond it (for an Alaska place, HRRR Alaska is offered once you agree to its download), with a “Regional Model: GFS (Lower Resolution)” badge when the lower-resolution model is drawing the view. Wind shear, the tornado index and wind at 80 m exist only on HRRR; soil temperature, sea surface temperature and waves come from GFS everywhere.

Warnings and hurricanes

Severe and Extreme NWS alerts that have a drawn area appear on the map (for an Alaska place, every severity, with zone-only alerts drawn on their zones), and clicking one opens the Weather Service’s full text with its areas, onset and end. NHC tropical storms are drawn as tracks and cones whenever one is active.

Live NWS alerts

Alerts for your location and a severe nationwide view, each card carrying the full bulletin rather than a headline. The Alerts tab stays the complete list for your location.

Notified at the next check, five at a time

A Windows notification at the next check after an alert is issued: within two minutes while the window is open, about fifteen minutes while it is closed. At most five per check, ordered most severe first, so an outbreak will not flood your notification centre. When more than five new alerts qualify in the same check the rest are not notified at all, and they are not held over for the next check either. An alert the Weather Service issues as an update to one Storm Sweep has already recorded, whether or not it reached you, is recorded silently rather than notified again, so the same warning never arrives twice, and one dropped by the five-at-a-time cap stays silent when the Weather Service updates it.

What is allowed to interrupt you

Three choices: warnings and emergencies, which is how it ships, so watches and advisories do not notify you until you change it; warnings and watches; or everything, advisories included. The Alert sound switch turns sound on or off. With Windows sounds, Tornado Warnings, emergencies and alerts rated Extreme get the alarm, other warnings and alerts rated Severe get the reminder sound, and the rest the default sound; Storm Sweep sounds use the three levels below.

Storm Sweep alert sounds, in three levels

A heads-up chime for watches and advisories, a repeating warning sound for warnings, and a critical sound for Tornado Warnings, emergencies, extreme warnings, Severe Thunderstorm Warnings the Weather Service tags destructive and G5/S5 space-weather alerts, which repeats until you dismiss it. Space-weather alerts use the same three levels. An alert’s level also follows the severity the Weather Service rates it: Extreme goes to the critical level and Severe to the warning level, whatever the alert is called, while watches, statements and outlooks keep the ordinary heads-up. Switch between “Storm Sweep sounds” and “Windows sounds” in Settings → Notifications. New installs start with Storm Sweep sounds; if you used an earlier version you keep your Windows sounds until you choose otherwise. A Tornado Watch is an ordinary notification, not the urgent alarm, and Windows mute and Do not disturb always win.

Checks continue while closed

A background task keeps checking roughly every fifteen minutes, Windows’ shortest repeating interval, when the app is installed from the Store. Windows decides when it really runs, and it can refuse background access to Storm Sweep altogether, so Settings reports whether the check is registered rather than leaving you to assume it.

Approaching-storm notices

Optional, and only while Storm Sweep is open on live radar with the storm overlay on, for storms the radar site on screen is tracking: when one showing rotation, a tornado vortex signature, likely severe hail or a 55 dBZ core is on course to pass within about 10 km of your location in the next 30 minutes, the app says roughly how many minutes out it is, once per storm, not repeatedly. This is Storm Sweep’s own radar estimate rather than an issued warning, and the notice says so. The radar tab’s Storms panel lists storms heading your way further out than that.

Offline city search

City search runs on your own PC against 32,333 places from the US Census Bureau’s 2024 National Places Gazetteer, bundled inside the app. Matching starts at the beginning of the name, so “Vegas” will not find Las Vegas, and it is place names only, not ZIP codes and not street addresses.

Climate

Climate at a Glance temperature and precipitation maps and charts for states, climate divisions and counties, with anomaly, value or rank, and timescales from one month to five years. Beside them: the nClimGrid monthly grid, nClimGrid-Daily county values since 1951, NOAAGlobalTemp global temperature departures since 1850, the U.S. Climate Extremes Index, U.S. Percentage Areas, and satellite upper-air temperature from UAH and NOAA STAR.

Drought

The U.S. Drought Monitor week by week since 2000 with its percent-of-area chart and the National Drought Mitigation Center’s credit in its own words, the Palmer drought indices by climate division, and the monthly North American Drought Monitor map from NOAA NCEI.

Extremes & history

State climate records, the U.S. national records, the NCEI Storm Events Database one year at a time, radar-detected storm cells, and historical tropical cyclone tracks from IBTrACS by basin. Storm Events changed its rules over the years, so counts across years are not like for like, and the app says so.

Ocean & ice

Daily sea surface temperature (OISST) and monthly since 1854 (ERSST), each with its anomaly; the Arctic sea ice edge, Northern Hemisphere snow and ice on the 24 km grid or the 4 km grid, Antarctic icebergs and U.S. snow depth; and sea floor and land relief. Relief and U.S. snow depth are also on the globe.

Stations

GHCN-Daily station records for the United States and its territories, with their 1991–2020 normals where NCEI has them, and U.S. Climate Reference Network stations with hourly readings, soil moisture and temperature at five depths, and a Monthly tab with maximum and minimum temperature, and mean temperature and precipitation in its table.

Search NOAA’s datasets

A OneStop search (Ctrl+F) finds NOAA datasets by words and date range. What you type goes only to NOAA OneStop, the map area is sent only if you turn “Search this map area” on, and the app does not save your search words.

Every value says what it is

Each product is tagged Observed, Analysis, Climatology, Record, Reported or Reference, shown with its period of record and baseline, and marked Preliminary where its producer says so. NCEI’s own ranks are shown as NCEI’s.

Nothing downloads until you pick it

A product downloads only when you pick it, and any single file over 10 MB asks first and shows its size and source; Cancel stops it. Products you have opened open again from their saved copy, labelled as one, even with no connection. Settings → Data & storage lists the downloads you agreed to, each with its own Delete.

The past hour to the past 30 days

Choose Past hour, Past day, Past 7 days or Past 30 days, and a magnitude of 2.5+ (the default), 4.5+, Significant or All. Dots grow with magnitude and are coloured by one of three depth classes, shallow, intermediate and deep, with heavier outlines for deeper ones so colour is never the only cue. A filled dot is an event USGS types as an earthquake and a hollow dot is any other type, named in its details.

Click for the details

Click a dot, or pick from the “Events in view” list. The card gives magnitude and its type, depth, the local and UTC times, USGS’s own Status and Alert words and its Felt number where it has them, the event id, and a “View on USGS” link that opens in your browser. Status and Alert are shown as USGS words them, without interpretation.

A time bar to replay them

A slider across the window shows events up to any moment, with Play and 1×, 2× and 4× speeds. A thin ring marks events in the hour before the time shown. Playback stays off when Windows animation effects are off, and the slider still works.

History from NOAA NCEI

Turn on History for NCEI’s Global Significant Earthquake, Historical Tsunami and Significant Volcanic Eruptions databases, with a Years slider from 2150 B.C. to now. History is off until you turn it on, loads in pages with a visible count, and asks first if a download would be over 10 MB. Its symbols are thin and carry dates, so they never read as current. NCEI’s own field names and values are shown as published.

Near your place

For the place saved on the Current weather tab: National Weather Service alerts about earthquakes, tsunamis, volcanoes or ashfall, shown as issued with their full text, and the USGS events within 100, 250 or 500 km, newest first, with distance and direction. For an Alaska place, links to Alaska Volcano Observatory information open in your browser. The distances are worked out on your PC from the events already on screen.

Saved copies, and what is sent

Each window and magnitude choice keeps its last good copy for up to 30 days, so offline the tab says “Saved copy from” its time. History keeps one complete copy per kind until a newer one replaces it. A request to USGS or NCEI carries only what is needed to name the feed or the page, plus your IP address as with any download; the tab never sends your place to them. Near your place also asks the National Weather Service for your saved place’s alerts about every 5 minutes while the tab is showing, with the same request the Alerts tab uses.

HRRR Alaska, after you say yes

For a place on its grid, the Current Weather map shows a notice, “HRRR Alaska is available here”, while a layer it carries is on. It says GFS (about 25 km) is drawn now and HRRR Alaska is 3 km, gives the download size per forecast hour and for the whole time bar, and asks “Download it, or keep GFS?” GFS stays until you click Download. On the Models tab it is listed as “HRRR Alaska · 3 km” and usually moves the area to Alaska. A few fields, such as snowfall, 0–1 km and updraft helicity, the tornado parameter and wind shear, are not offered for it.

Radar coverage here

For an Alaska place, a panel on the Radar and Current Weather tabs lists the Alaska NEXRAD sites within 460 km with their distances, or says “No radar coverage here” and names the nearest site. It names its source, the nowCOAST Alaska base reflectivity mosaic, and the newest frame’s time. West of 170°W, such as Atka, Adak, Shemya and Attu, it says radar there is the model’s simulated radar only.

Alerts on the zones they name

For an Alaska place, the Current Weather map draws warnings, watches and advisories of every severity, and alerts issued by zone are drawn on the National Weather Service zones they name, credited “Zone boundaries: NOAA/NWS”. A zone with no outline is listed as not drawn, and no area is guessed.

Out to the western Aleutians

The GFS and ECMWF windows now run from 170°E across the date line, the Marine map draws the western Aleutians, and place search knows names such as Barrow (Utqiaġvik), Dutch Harbor (Unalaska) and Shemya (Eareckson Station), showing the official name.

Sea ice and river gauges

A “Sea ice (USNIC)” layer on the Marine map, off by default, draws the U.S. National Ice Center’s daily marginal ice zone chart for Alaska waters in two classes, shown as published. On Marine’s Rivers tool, “USGS gauges near you (Alaska)” lists the latest stage and discharge within about 1° of your place, with each reading’s approval status and the qualifier code as published, such as Ice.

Links to the people who serve Alaska

An “Alaska links” panel on the Current Weather map opens the FAA weather cameras and the avalanche centers for the Chugach, Hatcher Pass and the Alaska Avalanche Information Center in your web browser, and Global Activity opens Alaska Volcano Observatory information the same way. They are links only, and nothing is embedded or copied.

Six Alaska lessons

A Learn The Weather level on spring breakup and ice jams, ice fog, winter rain in the Interior, Juneau’s Suicide Basin glacial outburst floods, the volcano links, and sunrise and sunset in the far north.

Space Tracker

A 3D Earth with its sunlit side and night lights, and satellites drawn where they are now: weather satellites, the ISS and stations, Starlink, other active satellites, selected debris events and rocket bodies. Filter by kind, search by name or NORAD number, and click an object for its altitude, speed, period, inclination, perigee and apogee, and how old its orbit data is. A geostationary weather satellite shows the ring of Earth it can see, and opens straight into the Satellite and Models tabs.

Passes over your place

For the ISS, brighter objects, Starlink trains or the object you picked: rise, peak and set times, directions, peak height, whether the object is sunlit, whether your sky is dark, and the National Weather Service sky-cover forecast for that hour. It needs a saved place.

Decaying orbits, shown as data

A list of the objects with the lowest perigee, lowest first, from CelesTrak, with a lesson on why orbits decay. It never gives a re-entry time or place.

Space Weather

NOAA’s R, S and G scales in NOAA’s own words, active regions, the GOES X-ray flux with flare classes, the Kp index, solar wind and Bz at L1, protons, and the aurora oval on the globe. Coronal mass ejections come from NASA’s DONKI with the model’s arrival time and a 3D cone labelled “Illustration”.

The Solar Loop

Four GOES SUVI loops in a 2×2 grid that play in step, with one Play/Pause, one slider and 1×, 2× or 4× speed. Choose any of the six wavelengths (94, 131, 171, 195, 284 and 304 Å), or NOAA’s SUVI feature map, and press a picture to enlarge it. “SDO latest” shows NASA’s newest still.

Coronagraph loops

Six views of coronal mass ejections leaving the Sun: GOES-19 CCOR-1 and SOLAR-1 CCOR-2, each in white light and as a running difference, plus SOHO LASCO C2 and C3. Every picture carries its publisher’s credit, CCOR-2 is tagged “Provisional”, and SOHO is a project of international cooperation between ESA and NASA.

Space Forecast

NOAA’s own 3-day outlook with its Kp forecast, the forecast discussion, and the 27-day and 45-day long-range outlooks, plus NOAA model runs placed on the globe where they apply: the OVATION aurora forecast, D-RAP radio absorption, GloTEC, WAM-IPE, Geospace ground magnetic disturbance and geoelectric field areas. Every frame is tagged Observed, Analysis, Forecast or “Model, earlier hours”, with the model run and valid time on your own clock where NOAA gives them.

CME arrival view

On the two Sun-to-Earth (WSA-Enlil) layers, a status line with a countdown, a time-bar marker and a dashed “Arrival (modeled)” ring show the arrival time NASA’s DONKI lists for a WSA-Enlil run, beside NOAA’s own run. The time is NASA’s published figure, not a Storm Sweep prediction.

Earth | Sun

A switch on Space Forecast and Space Weather. On Space Forecast the Sun view drapes the latest GOES SUVI 195 Å picture on the Earth-facing half of a 3D Sun, with NOAA’s active regions, the Sun’s turning over the next three days and NOAA’s Sun forecast. On Space Weather it drapes whichever Solar Loop picture you pick, 171 Å at first. The far side is plain.

What is the app’s own, and says so

The magnetopause, bow shock, plasmapause and the oval drawn from Kp are labelled “App model”; the radiation belts and the solar-wind streaks are labelled “Illustration”. None of them is NOAA output, and NOAA, NASA and ESA pictures keep their publishers’ own colours.

At your location

The aurora-oval value for your saved place, the hours the Sun is more than 12° below your horizon, and the forecast sky cover. It sets the facts side by side and deliberately never says whether you will see the aurora.

Ionosphere & Radio

Five views: D-RAP radio absorption from NOAA; an ionosphere foF2 map worked out on your PC from NOAA’s GloTEC; “Listening in a storm”, with your grid square and your nearest NOAA Weather Radio transmitters; Parks on the Air activations; and “Live arcs” of amateur-radio reception reports from the last 15 minutes, a 1% sample from PSKReporter, drawn between grid squares. Call signs are hidden unless you turn them on.

Optional space-weather alerts

Off until you switch them on in Settings. Choose how strong a geomagnetic storm (G), radiation storm (S) or radio blackout (R) must be before you are told. Alerts are NOAA’s own text, checked every two minutes while the app is open and about every fifteen minutes in the background when it is installed from the Store. No location is sent.

16 levels, 145 lessons

Weather basics; alerts and outlooks; radar, from how NEXRAD works to every radar product, including dual-pol and storm signatures; satellite; storm ingredients and soundings; forecast models; aviation and drones; marine; tropical and seasonal weather; space weather and radio; tornadoes; forecast models in depth; space-weather models; climate, oceans and history; earthquakes, tsunamis and volcanoes; and Alaska.

Try it as you read

Each lesson has steps to follow in the app itself. A “Show me” button takes you to the right tab, so you learn on the real radar, satellite or model, not on a picture of one.

Quick checks and level tests

A quick check closes each lesson, every level ends with a review and a test, and there is a final exam. Your progress is kept on your own PC.

Glossary and official guides

A glossary of weather terms, and a library of free guides from the National Weather Service and other official sources for when you want to go further.

Help that knows where you are

Press F1 on any tab for help with that tab. Help has its own search that understands everyday words, such as “why is the radar purple” or “wind for my drone”, and it works offline.

Original wording

The lessons are written for Storm Sweep and are part of the app, covered by the Terms of Use. They are teaching material, not a forecast or a warning: always follow the National Weather Service and local officials.

What changes

The radar color tables in Radar and Storm 3D, the models, the Current Weather layers, forecasting charts and the Skew-T, aviation, marine, the alert severity stripe, the space tabs, the legends and the Learn diagrams. Shapes, line patterns, hatching and words carry the meaning too, and warning outlines are labelled TOR, SVR and FFW in the friendly modes.

What keeps its publisher’s colors

NOAA’s own radar and satellite pictures, the NWS standard table and any .pal table you load, and the Sun and coronagraph pictures are shown as their publishers made them. Where a radar picture is NOAA’s, the Radar tab offers “Use Storm Sweep radar” and the Current Weather map offers “Use model radar”.

The Color key

Every color in the app is explained on each legend and on the Help page “Color key”, tab by tab. The Learn glossary card “Legend” opens it too.

No account, ever

You never create one or sign in. There is nothing to sign in to.

No analytics, no ads, no crash reporting

No telemetry of any kind. No advertising anywhere in the app. There is no automatic crash reporting and no send button. Unexpected errors go to a log file on your own PC, which you can read or delete in Settings.

Nothing loads before you accept

The first time you open Storm Sweep, an empty window shows one dialog: a short summary of the Terms of Use and Privacy Policy, with buttons to read either in full. Nothing behind it runs until you choose “I agree”: no page opens, no data is fetched, your location is not asked for, no notification is raised and the background alert check is not registered. Only then does a second step ask where you are, with a city search, a “Use my location” button and “Skip for now”; skip it and Storm Sweep opens with no place set until you choose one.

Requests identify the app, not you

Every request Storm Sweep makes carries one User-Agent naming the app and the support address, as the National Weather Service asks. No device id, no user id, no tracking token.

Stricter about what it downloads

Downloaded feeds are read with a size cap and a check that a small compressed file does not inflate into something enormous. Feed XML is read without following external references. The pages that draw the maps, 3D views and lessons now allow only the app’s own scripts. A bad or oversized download shows the usual “isn’t available right now” message instead.

Location is optional, and where that one position goes

“Use my location” asks Windows only when you click it, takes a single position fix (asking Windows for city-level accuracy, which is enough for weather) and saves one place. That position does leave your PC: it goes once to the National Weather Service to name the nearest town, and if that lookup fails the place is simply saved as “My location”. Either way the saved place keeps those coordinates, so from then on they go out with weather requests exactly like a place you searched for.

Your settings stay on your PC

One plain file in a Storm Sweep folder under your local app data holds your place, units, thresholds and preferences, plus your Learn The Weather progress and scores and your last ten Help searches. Storm Sweep never uploads that file, though the places inside it travel in the requests you make for them: your location’s coordinates, as above, and the radar, airport, tide station, buoy and river gauge you last looked at. Beside it can sit dated copies of that same file, kept for your protection: a backup written before a reset, and a copy of one the app could not read, the newest three of each. Those copies hold your chosen location too.

Sources named, Storm Sweep’s own figures labelled

Every tab that fetches data names where it came from, apart from Storm 3D: it downloads Level II volumes from the NEXRAD radar you pick on the Radar tab, and the credit for them sits on that tab rather than on its own. Anything Storm Sweep computed itself (hail estimates, upper-air analysis, storm motion) is labelled a Storm Sweep estimate rather than passed off as official.