How it works
How Faribault Weather works
Reviewed Oct. 17, 2026
The rules and sources behind each page, in one place. Keys and color scales stay beside their charts on each page.
Under the box's first line, three short answers for the hours people are out, 6 a.m. to 9 p.m. (from 9 p.m., tomorrow's, and the line says so), from the current hour on. Each says what its rule says and nothing more:
- Jacket: "yes" when the coldest of those hours in the National Weather Service's hourly forecast is below 55°: its temperature, or its wind chill where the NWS formula applies (50° or colder, with wind of 3 mph or more). The answer names that hour.
- Umbrella: the NWS hourly forecast's highest chance of rain or snow in those hours: 50% or more is "yes", 30% to 49% is "maybe", each with that hour; under 30% is "no".
- Wind: the highest gust the HRRR model forecasts in those hours (from Open-Meteo): 25 mph or more is "breezy", 35 mph or more is "windy", 45 mph or more is "very windy", always with the gust; under 25 mph, the gust alone.
Yesterday: the Faribault airport's (KFBL) high and low for the day as the report card scores it, midnight to midnight standard time (the day the National Weather Service keeps its climate records on), with 18 or more hours of reports, and the NCEP Stage IV estimate of rain at Faribault for the local calendar day, from the daily archive. The models page's scoreboard shows the same high and low. The line shows from 9:05 a.m., once the archive has fetched the day again (the day's rain); before then, or with a number missing, it doesn't show. Stage IV is later adjusted with PRISM, so the rain is an estimate and can change.
First Freeze: the first date since Aug. 1 when the airport's daily low was 32° or colder (the National Weather Service's freeze), with that low: the same day and daily low the yesterday line shows, counted from 9:05 a.m. the next morning. Until then a reading of 32° or colder at the airport lights the chip with that reading, and while a Freeze Warning or Hard Freeze Warning is in effect the chip names it. Before the first freeze the chip gives the typical date, when half of years have had their first fall low of 32° (NOAA's 1991-2020 normals).
What radar shows. A radar sends out a beam and measures the energy that comes back from what it hits, its Reflectivity The amount of energy that returns to the radar after hitting a target. NOAA JetStream, in dBZ Decibels of reflectivity, radar's unit. The scale runs from about -35 to +85; generally, the higher it is, the heavier the precipitation. NOAA JetStream. Generally, the higher it is, the heavier the precipitation, but radar doesn't measure rain itself, so the rain rates it suggests are approximate, rates for an hour rather than the rain a place gets (NOAA JetStream). The weakest returns can be very light precipitation or general clutter.
The views. Local: the National Weather Service's Chanhassen radar (KMPX), about 50 miles around Faribault, its lowest scan. County: the same radar, the middle of the local view at 2.5 times the size, which the NWS map server draws from the radar's own data. Regional: the NWS mosaic of every radar in the area. Every 5 minutes the site asks the NWS map server for new scans and recolors them. The server keeps only about 2 hours, so the site keeps 7 hours of its own: the 6-hour loop shows one scan every 10 minutes, and the 1-hour loop is the 2-hour loop's last hour.
The colors. Six steps from 15 dBZ up; weaker returns are left off. The words and the rain rates are NOAA JetStream's (its reflectivity guidelines and its dBZ to rain rate table; a step that spans two of its rows takes both):
- 15 to 24 dBZ: very light to light, or clutter, a rain rate under 0.05 in. an hour
- 25 to 34 dBZ: light, a rain rate 0.05 to 0.22 in. an hour
- 35 to 44 dBZ: light to moderate, a rain rate 0.22 to 0.92 in. an hour
- 45 to 54 dBZ: moderate to heavy, a rain rate 0.92 to 4 in. an hour
- 55 to 64 dBZ: heavy, or some hail, a rain rate 4 to 16 in. an hour
- 65 dBZ and up: extremely heavy, including hail, a rain rate 16 in. or more an hour
The color-blind-safe colors are viridis (from matplotlib), lightest for the weakest step, so the steps read in order by lightness alone. A simulation of the three kinds of color blindness (Machado, Oliveira and Fernandes, 2009) keeps each step apart from the next.
What radar can't see. The beam goes out straight while the ground curves away under it, so the farther it reaches, the higher above the ground it is:
- The Chanhassen radar (KMPX), the local and county views' radar.
- Faribault, 41 miles from it (the National Weather Service's radar station list).
- A storm's strongest part can sit above the lowest scan, which then doesn't sample it (NOAA JetStream).
- 100 miles out, the lowest scan (0.5°) is about 12,000 feet above the ground (NOAA JetStream).
Freshness and gaps. The stamp gives the newest scan's time. Past 30 minutes it says when the last one came and the map grays; past 2 hours no scan in the loop is current, so the page shows none, and the NWS's own Chanhassen loop from our server in its place. A step between scans of more than 10 minutes, or more than 2.5 times the loop's usual step, is a gap: the slider marks it and the time tag names it.
Warnings and storm reports. The outlines and the list beside the map are the Rice County warnings from the same feed and in the same words as the alert bar. The storm reports are the storm log's: the NWS's local storm reports within 20 miles of Faribault from the last 6 hours, synced every hour (every 10 minutes during a Rice County warning). One within 1.5 miles of Faribault is in town, the storm days' rule.
Rain in the last 24 hours. A switch beside the storm reports puts another map in the scans' place: the rain of the last 24 hours as estimated from radar by NOAA's Multi-Radar Multi-Sensor system (MRMS), through the Iowa Environmental Mesonet. It's the one place the site uses MRMS, and it's never compared with or added to the site's other rain numbers. Each hour the site fetches the newest 24 hours, cuts each view's box from it and colors it in six steps; the loop is off while it's shown. Its stamp gives the end of its 24 hours, and past 3 hours old the switch and the layer hide.
The airport check. When the Faribault airport's (KFBL) newest report is within 30 minutes of the newest scan and either shows something, the page sets the two side by side: the strongest color within 1 mile of the airport, in the key's words, and what the airport reports falling, in the National Weather Service's words for its report. It doesn't say which is right. On Oct. 17 and 23, 2026, 11 of 105 airport reports with a scan had color within a mile of the airport, all of it the lightest step; in 9 of them the airport reported nothing falling. Each comparison is logged, so the rule can be measured.
A link to a view. The page's address carries the view you're looking at (view=county or regional, loop=1 or 6, colors=cb, reports=off, rain=on), so Copy link, or the address itself, opens the same view.
- Low
The night after the day, 6 p.m. to 6 a.m., as the NWS pairs them: Friday's low is Friday night's. Highs are for the calendar day, midnight to midnight, the way the report card scores them.
- Bus stop
On school days by the home page's rule (the Faribault Public Schools calendar): the NWS hourly forecast for 7:30 a.m., with the temperature, the wind chill when it's 3° or more below it, and the chance of rain or snow then.
- Dry window
4 or more daylight hours in a row (whole hours between sunrise and sunset), each with under a 20% chance of rain in the NWS hourly forecast.
- Paint or stain
Start times in daylight with 50° or warmer, humidity under 85% and wind under 15 mph from the start until 4 hours after, ending by sunset, and under a 20% chance of rain for the 24 hours from the start (NWS hourly forecast). The 4 hours are this site's own rule, not a paint maker's. No row on a day without such a start.
- Night
From our low: 36° or below is frost possible, 32° or below a freeze, 28° or below a hard freeze, the National Weather Service's terms. In fall, until the season's first freeze, the week's first freeze night says how many years in 10 have had one by then (NOAA 1991-2020, as on the climate page).
- First freeze
Until the Faribault airport has had its first freeze this fall (a daily low of 32° or below, or a report at 32° or colder), a line at the top names the first night on the cards whose likely low reaches 32°, with NOAA 1991-2020's date by which half of years have had their first freeze. A Freeze Warning is a forecast, so it doesn't count.
- Daylight
The day's length and how much longer or shorter it is than the day before are worked out to the second, then rounded to the nearest minute, so the length can be a minute off the times shown.
- Sky at sunset
The Euro model's low, middle and high cloud for the hour nearest sunset. Good sunset chances is a rule of thumb, not a forecast: low cloud under 30% and middle plus high cloud from 30% to 70%.
- Stars
A stargazing night has under 20% cloud (the Euro model) every hour from astronomical dusk, when the sun is 18° below the horizon, to midnight, with the moon under 50% lit.
- Moon
The phase and the share lit at noon, computed here; they agree with the U.S. Naval Observatory's to within a point. The night sky page has the rest.
- Aurora
Only for nights inside the Space Weather Prediction Center's 3-day forecast: the highest Kp in the fully dark hours (the sun 18° or more below the horizon), the same number as the home page and the night sky page. A night forecast at Kp 7 or more says possible, otherwise quiet.
- Storms and snow
The Storm Prediction Center's outlook for Faribault on every day it covers: its risk category on days 1 to 3 and its probability (15% or more) on days 4 to 8, with links to the storm log and radar at Slight risk or higher. A quiet day says so in SPC's own terms. Days 1 to 3: "General thunderstorms possible" inside its general thunderstorm area; "SPC: no storm risk area over Faribault." outside every area; "no severe thunderstorm areas forecast" only on days SPC drew none anywhere. Days 4 to 8, SPC's labels: "predictability too low" means severe storms may be possible in some model scenarios, but where or whether they happen is in doubt; "potential too low" means a regional area of organized severe storms appears unlikely (under 15%). In winter mode the line shows only when Faribault is inside an SPC area. From Oct. 15 to April 30, the Weather Prediction Center's Winter Storm Severity Index and the snow odds from the winter page when a storm is coming with a 5% chance or more of 1 inch; under that, a quiet line when the whole week is under it too.
- The week
At the top, from our highs and both chances of rain. A day is warmer than normal when our high is 5° or more above NOAA's 1991-2020 normal high, cooler than normal when it's 5° or more below it, and near normal otherwise; up to 4 runs of like days are named in order, and more are counted. The best chance of rain names the day whose higher chance, the NWS's or ours, is highest, when it's above 20%; otherwise no day is above it, through the last day both chances reach.
- What changed
At the top: what moved since the last logged update, twice a day at 9:40 a.m. and 9:40 p.m.: a high or low by 3° or more, a chance of rain (the NWS's or ours) by 20 points or more, or a Storm Prediction Center risk of severe storms added, dropped or changed (Marginal or higher on days 1 to 3, 15% or more on days 4 to 8). Hidden when the log is missing or more than 30 hours old.
- Rain this week
Under the week's story, from the same ensembles as a card's amount: each GFS and Euro ensemble run's own total over the days every ensemble covers, from tomorrow, then the middle half of those totals (each ensemble counted equally), to the nearest 0.05 in. Dry runs count at 0. Adding up each day's range would describe no run's week.
- The weekend
When Saturday and Sunday are both cards: our high and low, both chances of rain and the NWS's wind for each, and the better day by the nicest-day tests below (fewer missed, then the higher score; none when they tie). From Friday at 3 p.m. it comes first on the page; otherwise it follows the cards.
- Best and worst day
Under the week's story. The best day uses the home page's nicest-day rule: dry (20% chance of rain or less), sunny or clear, wind under 15 mph, not muggy (dew point under 60), and, May to September, 65° to 82°; October to April, at or above the normal high. The worst day misses the most of those tests, named only when it misses 2 or more. A day's wind is the NWS's: when it gives none, the wind is left out, never shown as 0; "calm" only when its forecast says calm. When it gives 0 mph without saying calm, a card reads "Wind light" with the NWS's figure ("Wind light (NWS: around 0 mph)"), the table and the weekend "Light", and these lines leave the wind out.
- The week, hour by hour
The National Weather Service's hour-by-hour forecast for Faribault's spot (its gridpoint forecast): temperature, the chance of rain and the amount by 6-hour block, wind and gusts. Its title names the warmest and coolest hours and the block with the best chance of rain, when one reaches 20%.
How the forecast is made
Our forecast blends the computer models the site scores every day, each weighted by how close it has been here over the last 90 days at that Lead day How many days ahead a forecast was made: day 1 is the forecast for tomorrow. This site's report card (the report card). The high then has the blend's own average miss over those 90 days taken off; tested on the past year, that brought highs closer and lows no closer, so lows are left as blended. The National Weather Service is kept separate, so the two can be compared.
Likely is where the high or low should land about 8 times in 10: the GFS and Euro ensembles' middle 80%, after each one's recent bias is removed, never narrower than our own misses over the last year at that many days ahead. Past ensemble runs aren't kept, so each ensemble's bias is borrowed from its regular model (GFS or Euro) until this page has logged a season of its own ranges to check.
The low's record of misses is kept night by night: each model's lowest temperature from 6 p.m. to 6 a.m., as forecast that many days ahead, against the lowest reading at the Faribault airport (a night counts with readings in 9 of its 12 hours); HRRR and NAM have no record of past nights, so the low leaves them out. Today or Tonight above the days, and the day words, are the National Weather Service forecast for Faribault. Updated under the title gives the newest source's own time (the NWS's update, a model's run, an outlook's issue), never when the page was written; each older source has its own line above the days.
Sun and moon times, golden hour, clouds and the aurora threshold: Used on several pages.
The verdict looks at each whole hour when it is fully dark, from astronomical dusk to dawn (the sun 18° or more below the horizon), and the Euro model's cloud for that hour. An hour is clear under 20% cloud and cloudy at 70% or more. A good night for stars needs clear hours with no bright moon up, from the first such hour to dawn, at least 2 hours in a row; the moon is bright at 50% lit or more while it is above the horizon. Clear spells count from 2 hours. These are the same numbers as the stargazing line on the 7-day forecast.
Blue hour is the sun 4° to 8° below the horizon, timeanddate.com's definition. The lens, from the National Weather Service's hourly forecast for the night: frost when its low is 32° or below; otherwise dew when the low is within 3° of the dew point.
Aurora: Active now, and it's clear needs SWPC's Kp estimate at 7 or more, full dark, and under 30% cloud this hour (Euro). Possible if clear: Kp 7 or more now, or in SWPC's 3-Day Forecast for a 3-hour block in tonight's dark hours. Otherwise quiet. The Kp charts mark Kp 5 or more, a geomagnetic storm on SWPC's scale; that mark decides nothing. Bz is the solar wind's north-south magnetic field: SWPC says a southward field is, "most importantly," what drives geomagnetic storms; south of 0 nT favors aurora, and −5 nT or lower strongly favors it, the two lines SWPC's solar wind plot shades. OVATION is SWPC's 30- to 90-minute aurora model; its number is the chance of aurora overhead, and SWPC gives no number that means "visible," so it is shown here but never decides the answer. SWPC: with a clear view north, a bright aurora can be seen from as much as 1,000 km (600 miles) away.
Planets: Mercury, Venus, Mars, Jupiter and Saturn, computed here with PyEphem. A planet is up tonight when it gets 10° or more above the horizon between civil dusk and dawn (the sun 6° down), checked every 10 minutes; its highest point is over those hours, and its brightness is at that point. Rise and set are for the planet's center at the horizon, allowing for the air's bending of light, as the U.S. Naval Observatory gives them; on Oct. 11, 2026, they were within a minute of NASA JPL's Horizons. Close together: two planets, or a planet and the moon, 5° apart or less while both are 10° up, at their closest.
Space station: computed here with PyEphem from Celestrak's orbital elements for the International Space Station, fetched once a day. A pass is listed for the part of it with the station 10° or more up, in sunlight, while the sun is 6° or more below the horizon here, from tonight through the 7 nights after it. Its brightness is an estimate from its distance and how much of it is lit, fitted to Heavens-Above's magnitudes; on Oct. 11, 2026, its times were within 6 seconds of Heavens-Above's and its brightness within 0.5. The orbit grays when its elements are 3 days old and its passes hide at 7 days. Clouds then: the Euro model's cloud for the hour nearest each pass's highest point.
Milky Way core: the galactic center, the zero point of the IAU's galactic coordinates, computed here with PyEphem. A night's window is its fully dark hours (astronomical dusk to dawn) with the core 10° or more above the horizon, checked every 5 minutes; from Faribault it never gets higher than about 17°, due south. Its season is the run of nights with a window of 30 minutes or more. The moon: its share lit that night and when it's above the horizon in the window.
Eclipses: from NASA's catalogs by Fred Espenak, entered here through 2030. A lunar eclipse is listed when the moon is above the horizon for some of its partial phase, or for a penumbral eclipse, some of its penumbral phase; each phase runs half its listed length either side of greatest eclipse, within seconds of NASA's own times. A solar eclipse is listed when the sun is up for some of it here, from NASA's Besselian elements. Within 7 days, the Euro model's cloud for its whole hours.
Moon: rises and sets as the U.S. Naval Observatory gives them (the top edge on the horizon), within a minute of its times; the calendar shows each day's moon at noon. Full moon names are The Old Farmer's Almanac's: the Harvest Moon is the full moon closest to the September equinox, and a second full moon in one month is a Blue Moon. A supermoon is NASA's: "a full Moon that comes within at least 90 percent of perigee," read here as at least 90% of the way from that orbit's farthest point (apogee) to its nearest (perigee).
Meteor showers come from the International Meteor Organization's calendars, entered here by hand: the showers with a ZHR of 10 or more whose radiant gets well above Faribault's horizon at night. The peak night is the night nearest the IMO's maximum. For a peak night within 30 days: the moon, when the radiant rises and is highest, and the best hours: fully dark, the radiant at least 30° up, and no bright moon up. Within 7 days, the Euro model's cloud for those hours.
A night is tonight until its astronomical dawn.
Sun and moon times, golden hour, clouds and the aurora threshold: Used on several pages.
Each model's forecast high and low for the next seven days is checked against the real high and low at the Faribault airport (KFBL). Day 1 is tomorrow's forecast, day 7 is a week out. The miss is how many degrees off a model was, on average. The score is the average miss in degrees over the last 30 days, 90 days and year. Days run midnight to midnight on Central Standard Time all year, so daylight saving time never changes the math. The model rows come from archived model runs (Open-Meteo's Previous Runs, back to January 2024): for day 3, the forecast each model made three days before each hour. We also save every model's forecast at 10:40 a.m. each day (9:40 a.m. in winter) since Sept. 28, 2026; the NWS row comes from those saves, since the NWS forecasts are not archived.
Some models give a temperature only every 3 or 6 hours: the Euro and GEM all week, and the GFS, ICON and Euro (9 km) after their first few days. The warmest moment of the day can fall between two steps, so they tend to guess the high a little too low. They are scored on their own steps. A model leads only with 7 or more days scored.
Rain is scored separately against the Stage IV estimate: how many rain days of 0.05 inch or more each model caught, how often it called for rain that didn't come, and how far off the amounts were.
The table ranks the models by their day-1 average miss over the period you pick; a model needs 7 days scored to be ranked. The page calls one model the closest only when its lead over the next is bigger than their day-to-day ups and downs: on the days both were scored, the next model's miss minus the leader's has to average at least twice its standard error. Otherwise it says no model is clearly ahead.
A win goes to the closest day-1 forecast of each day over the last 30 days; a tie goes to the model ranked first over those 30 days. The NWS row is scored from this site's daily save of its forecast, on the NWS's own day, so it sits apart: not ranked, and not in the win count.
“Runs warm” or “runs cold” by a model's name means that over the last 90 days, with 30 or more days scored, its day-1 forecast leaned that way by 1° or more on average, at least 2 standard errors from zero and at least half as large as its day-to-day scatter; it shows only when the period you pick leans the same way by 1° or more. When 6 or more of the models lean the same way by 1° or more over the period you pick (each with 29 or more days scored), a line above the table says so, in the numbers the warm-or-cold panel shows. It says what happened, not why.
The How sure tab uses the same words as the 7-day forecast, from one rule: Model spread doesn't predict how far off the high will be, so the high gets its range, not a word. The low: how close GFS, GEM and NBM are (and ICON through day 5). Sure within 3° for days 1 to 3 and 6° for days 4 to 7, Fairly sure within 8° and 11°, Not sure past that. Rain: Sure when every model agrees on rain or no rain (0.05 in. or more), Fairly sure when one model disagrees, Not sure on a split.
Yesterday's scoreboard sets each model's forecast made the day before (the report card's own: Open-Meteo's archive of earlier runs, and the NWS forecast this site saves at 10:40 a.m., 9:40 a.m. in winter) against the airport's high and low for the day, midnight to midnight standard time, both rounded to whole degrees. It shows once the 8:30 a.m. score has that day with 18 or more hours of reports, and it prints the same high and low as the home page's yesterday line.
The biggest miss is the largest day-1 miss in the last 30 days among forecasts whose run this site saved: for each model and day, its newest run that had reached this site by 10:40 a.m. (9:40 a.m. in winter) the day before, as the report card's live save would have had it.
The temperature chart shows the middle of the models each hour (the median of those that reach that hour, 3 or more), with their lowest to highest shaded; “Every model” draws each one. The line above it is the median of the models' highs for tomorrow, which isn't this site's forecast; that one is on the 7-day forecast. The dew point, cloud cover and gust charts open on the middle and range the same way.
Each chart's first line says what its data shows, written from that data each time the data is written, never by hand. The Model guide's facts come from Open-Meteo's model pages and the NWS's office record; a cell its source doesn't state is left empty.
In Faribault: the National Weather Service gives each storm report as a point with its coordinates. A report counts as in Faribault when that point is within 1.5 miles of this site's Faribault point (44.29° N, 93.27° W), whatever town its location names: on April 13, 2026, two reports the weather service placed "2 WSW Faribault" were 1.48 miles from that point, so they count as in Faribault. The headline, the facts, the reports table and its town chips and filter, and how much warning Faribault had all use this one rule. A report without usable coordinates counts in neither direction. In the reports table's town chips, a report whose location names Faribault but whose point is 1.5 miles or more from this site's is under "Near Faribault": the weather service places it by Faribault, and this site's distance puts it outside town.
Warnings: a warning covers an area, so whether it covered Faribault has three answers. Yes or no by its outline when the weather service drew one, with this site's Faribault point inside it or not; yes for a warning issued for Rice County or its forecast zone. The texts of 31 county warnings, from 1986 to 1999, name only part of the county, and each was read: Faribault is in when the text names it or southern Rice County, out for northern, northeastern or northwestern Rice County, and not settled for any other part unless the text names Faribault. A warning that isn't settled counts in neither direction.
A storm day: a day with a National Weather Service warning for Rice County, or a storm report within 20 miles of this site's Faribault point. Heat, cold, wind chill, freeze, frost and red flag warnings never make one. A snow report under 6 inches makes one only while a winter storm, blizzard or heavy snow warning or a Winter Weather Advisory is in effect. A day made only by an extreme cold, heat or wildfire report stays in the log, tagged for what it was. The storm reports begin April 19, 2004.
Tags and filters: a day's tags say what happened, from its reports (Hail for any hail, Wind for any wind), then its warnings. "Severe size" finds the days with a report at the National Weather Service's size for a severe storm: 1-inch hail or 58 mph wind, a 58 mph gust or thunderstorm wind damage counting as 58 mph wind. "In Faribault" finds the days with a report in Faribault or a warning that covered it, by the rules above. "Warnings only" finds the days with a warning for Rice County and no storm reports, of any kind of warning; before the reports begin, every storm day is one.
A storm over more than one day: back-to-back storm days are one entry when thunderstorm activity runs across the midnight between them with no gap of an hour or more (tornado, severe thunderstorm and flash flood warnings, from issue to end, and the reports of storms as they happened, not rain or snow totals), or when a winter storm, blizzard or heavy snow warning from an earlier day of the entry is still in effect after that midnight. Each day keeps its own page and counts as its own storm day, and the entry links to each.
Big days: among the storm days since 2006, each day ranks on four measures: its largest hail; its storm reports of what happened, each one in Faribault counted twice; its tornado, severe thunderstorm and flash flood warnings; and its hours under one of those warnings over Faribault. A day is as big as its best rank, with ties broken by its second best. The 25 biggest since 2006 that aren't winter days stand out in larger, bolder type. Winter days count by their largest snow report within 20 miles instead, and the 10 largest since 2006 stand out the same way. Earlier years aren't complete on every measure.
Records: a top 5 for each kind of report, one entry a day (its largest report of the kind), in Faribault (by the distance above) and within 20 miles apart, each from the first full year of its data: largest hail since 2005; highest measured gust since 2008; strongest estimated wind since 2008; most rain in one report since 2008; largest snow report since the 2006-07 winter. Ties rank alike; a tie that would run a list past 5 is one line, its days in a fold. A rain total over more than a day is left out. Measured or estimated is the report's own word, "not stated" when it has none. The 10 biggest storm days rank by the big-day score (Big days, above). A storm day's page ranks it by these same lists.
Charts: this year against an average year of 2007 through last year, the band the middle 8 in 10 of those years; two strips of days in each month, 2007 through this year, on one scale, each titled by its own busiest month: thunderstorm days, the storm days with a tornado, severe thunderstorm or flash flood warning for Rice County or a hail, tornado, funnel cloud, thunderstorm wind, flash flood or rain report within 20 miles (rain only from April through October), and winter storm days, the storm days with a winter storm, blizzard or heavy snow warning for Rice County or a snow or ice report within 20 miles (the Winter filter's days). A day can be in both. Every other count on the log counts all storm days. And tornado, severe thunderstorm and flash flood warnings for Rice County by the hour each was issued, in Faribault's time, since the first in the log.
On this date and the jump: "On this date" lists the storm days on today's date in earlier years, and hides on a date with none. "Jump to a date" opens the log at that date's storm day, or the nearest one before it.
Following new storm days: storm-feed.xml is an Atom feed of the 20 newest storm days, each with its headline and summary, for a feed reader. It changes when a storm day does; there's nothing to sign up for.
Winter mode runs Oct. 15 through April 30, with the forecast sections. From May 1 the page shows the season that just ended.
Snow totals come from the Faribault co-op station (NOAA USC00212721), read each morning by a volunteer observer; temperatures from the Faribault airport (KFBL); our gauge is CoCoRaHS MN-RC-45. Every number names its station, and the stations are never joined into one record.
The airport's daily highs and lows pass one quality rule, the same one Faribault Climate uses: a reading more than 6° beyond every station within 40 miles that day or the next (a co-op's day ends at its morning reading) is left out, and so is a low 25° below both the day before's and the day after's. A day's low counts with 12 or more hours of reports.
The snow odds come from every run of the GFS and Euro ensembles; the wind chill and the snow, ice or rain strip from the National Weather Service forecast for Faribault.
A section whose data is missing or out of date is hidden. Long-term normals and records are on Faribault Climate.
The airport's day: a day's high, low, wind, gusts and its 24-hour chart are the Faribault airport's (KFBL) for its day: midnight to midnight standard time all year, which is 1 a.m. to 1 a.m. while clocks are on daylight time. It is the same day the report card scores, home's "Yesterday" line and the Models page's scoreboard show and the storm-day pages' airport panel covers, read from one copy, so they print the same high and low. They are final from 9:05 a.m. the next morning, once the report card has scored the day; until then the day shows the readings so far. A day whose reports cover fewer than 18 hours gets no high or low, and says so. The airport's high or low can come from its six-hour summaries between two readings; then the page gives no time for it.
Everything else on a day's page, the rain, warnings, storm reports, the sky cam, the sun and the moon, is the local calendar day, midnight to midnight Faribault time. The rain is the NCEP Stage IV estimate at Faribault's point, not the airport's gauge.
Temperatures print rounded to the nearest degree, halves up, as everywhere on the site, each with its 10-degree band's color beside it, the band of the printed number.
Where the record comes from: the Faribault co-op station (NOAA USC00212721), one of the National Weather Service's volunteer stations, read once a day. Its temperatures run from Dec. 1, 1895, to June 27, 2011, when its temperature equipment was removed (NOAA's station history), with none from Oct. 18, 1960, through July 31, 1969; its precipitation, snowfall and snow depth go on to today. It moved 3 miles north-northeast on Aug. 1, 1969. The page never joins its temperatures to the airport's, a different station: every record is the co-op's, with its years.
What a normal is: The 30-year average for a date or month, from NOAA's U.S. Climate Normals for 1991 to 2020. NOAA's 1991-2020 normals for this station rest on 17 to 21 years of its own temperatures (NOAA's flag: representative), since its temperature equipment was removed in 2011; NOAA filled the other years with estimates from nearby stations. Its precipitation normals rest on 26 to 30 years (standard).
Freeze dates: a typical date is NOAA's 1991-2020 median, the date by which half of years have had it, the same one home's First Freeze chip shows (without NOAA's file both hide). The earliest and latest, and each year's dates on the calendar strips, come from the station's own daily lows: the last low of 32° or colder before July 31 and the first after, over the years with at least 90 percent of March through June (or August through November) reported. The growing season is the days from one to the other, in years with both.
Counts and records: a 90° day is a high of 90° or more, May through September; a below-zero night a low under 0°, November through March, placed at the year the winter starts. Each summer or winter counts only with no more than 7 of its days unreported, so a gap in a chart is a year that doesn't count, never a zero. All of them are the co-op station's, so they end with its temperatures in 2011. The records by decade are the daily records home's normals bar shows, each in the decade of the year that holds it, a tie to the latest year.
Precipitation: a year ranks among the wettest and driest with no more than 7 days unreported, and the days it has are stated; a season (winter is December through February) counts with all three months finished and no more than 3. How normals shifted sets NOAA's 1981-2010 normals for the same station beside its 1991-2020 ones; the older set's temperatures are complete, the newer ones partly estimated, as above.
The airport's own series (the Faribault airport, KFBL, a different station a few miles away) stands beside the co-op's hot days and cold nights, never joined to it: its days with 18 or more hours of reports, midnight to midnight standard time, after its one quality rule, the winter page's too. A high goes when it is more than 6° above every high, and a low when it is more than 6° below every low, that the stations within 40 miles reported for that date or the next (at least 3 of them); a low 25° below both the day before's and after's goes too. Among what it drops are highs of 100° to 140° in 1997 to 1999 and lows of -43° and -41° in January 1994 that no station near it had.
Snow seasons run July through June; a season counts when it has ended with no more than a week unreported from November through March, the same seasons the winter page counts. Storms are the storm log's own counts: the tornado warnings for Rice County, this year set against the log's own years only, and the storm days by type from its two strips (thunderstorm days and winter storm days, from 2007), counted by the same code.
Today in context is written with home's box, from the same forecast, so its high is home's. The CSVs are each section's numbers as the worker wrote them, with units in the column names. The stamps: each section's stamp gives the date its data runs through, or when it was last written; the page is written again when its numbers change.