We've mapped 126 valleys and cities in 18 states, from the Missoula Valley, where we calibrated the model, out to Fairbanks, Taos and Aroostook County. Each map has its own write-up of the local cold, the frost pockets and the fruit trees that fit. This page is the part they all share.
To find your valley, start at our microclimate maps. To get a number for any address, map or no map, use the cold hardiness calculator.
What makes your microclimate
Your microclimate is the weather your trees actually live in, and it's rarely the weather at the airport. On a clear, still night it comes down to heat: how fast the ground loses it, where the cold air goes once it forms, and what puts heat back. A few degrees either way decides what you can grow, so it's worth knowing how it all works.
- Ridgeline. The edge of the airshed, the same divide a watershed has.
- Clear sky. The ground radiates its heat straight out to space.
- Drainage. Chilled air slides down the slopes, creeks and draws.
- Cold pool. It collects on the floor, coldest at the bottom.
- Thermal belt. The slopes just above the pool, the warmest ground on a still night.
- Above the belt. It gets colder with height again.
How the cold gets made
All night the ground gives off heat as infrared light, aimed at the sky. How much comes back depends on what's overhead. Under clouds or humid air a lot of it comes back. Under a clear, dry sky very little does, so the ground cools fast and chills the thin layer of air touching it. Air is a poor conductor, so the cold stays low. That's why there can be frost on the lawn while the thermometer on the porch reads in the mid 30s.
Cold air is heavier than warm air. At the same pressure, air at 20°F is about 6 percent denser than air at 50°F. So it slides downhill, slowly, along the slopes and down the creek bottoms and draws, and spreads out across the valley floor. Anything in its way backs it up. By morning the floor holds a lake of cold air, coldest at the bottom.
That lake is an inversion. Normally it gets colder as you go up. Inside the pool it gets warmer as you go up, until you reach the top of the cold air. The slopes just above the pool are the thermal belt, and on a still night that's the warmest ground in the airshed. Above the belt it gets colder with height again.
In winter, under high pressure, with long nights, snow on the ground and no wind, a pool can sit for days or weeks. Each long night adds more cold than the low winter sun can take back, and fresh snow throws most of that sunlight right back to the sky. If you've lived through a January inversion in a western valley, you've seen your airshed with your own eyes: the smoke and fog sit right in it and stop at the top. Air quality people use the word airshed for the same reason.
What puts heat back
- Clouds. A cloud deck soaks up the ground's infrared and sends a lot of it back down. A deck that slides in at 2 a.m. can stop the drop right there. Clear nights are the frost nights.
- Moisture. Humid air holds heat in the same way, though not as well as a cloud. And when the air cools to its dew point, water comes out as dew or frost and gives off heat as it does, which slows the fall. Dry air has nothing to slow it down. On a clear, calm night the low often lands near the evening dew point, which is why growers watch it.
- Water freezing. Water gives off heat as it freezes, about 144 BTU for every pound. That's how sprinklers protect blossoms: as long as fresh water keeps freezing on the bud, the bud stays near 32°F. Shut them off too early and evaporation pulls the heat back out, and the buds end up colder than if you'd never started.
- Wind. Any breeze stirs the cold layer into the warmer air above it. A wind machine does nothing else: it pulls down the warm air sitting on top of an inversion. No inversion, no gain.
- Lakes. Pound for pound, water holds about five times as much heat per degree as dry soil, and a deep lake stores a whole summer's worth. Open water just above freezing next to air at 20 below is a heater. Once a lake freezes and snows over, it acts like a field. That's why Flathead Lake and Lake Pend Oreille, which mostly stay open, warm their shores in a cold snap, and lakes that freeze over don't. Big lakes also keep the shore cool in spring, which holds bloom back until most of the frosts are over. That's how the cherry orchards on Flathead Lake's east shore work, and the fruit belts along the Great Lakes.
- Town. Brick, concrete, asphalt and stone soak up the day's sun and give it back all night. Buildings block part of the cold sky, so less heat escapes, and heated buildings leak heat. The EPA puts city nights about 2°F to 5°F warmer than the country around them. In our valley towns the map gives built-up ground up to 2.5°F.
- Trees and walls. A canopy blocks part of the sky the same way, so the ground under trees loses less heat. A south-facing wall soaks up the afternoon sun and gives it back after dark, which is why the south side of the house is often the warmest spot on the lot.
What comes from above
The ground makes cold on any clear night. The coldest nights of a decade need help from above: arctic air, and nothing to push it back out.
- Arctic air. Arctic air out of Canada is cold, dry, dense and shallow, often shallower than the mountains are tall. It slides south along the east side of the Rockies and piles up against them. The Continental Divide holds a lot of it back, which is a big part of why the valleys west of the Divide usually run milder in a cold snap than the plains do. When it gets deep enough it spills over the low passes and pours out of the canyons on the west side as a hard, cold east wind, and then the west side gets it too.
- Chinooks. A west wind crossing the mountains rises, cools and drops its snow on the west slopes. Coming down the east side it warms by compression at about 5.5°F for every 1,000 feet. On the way up, once it was making snow, it only cooled about 3.5°F per 1,000 feet, because water gives off heat as it condenses. So it lands on the plains warmer and drier than it started. That's a chinook. They blow along the Rocky Mountain Front, reach roughly 100 miles out onto the plains, and can scour a cold pool out in minutes. In January 1972, Loma, Montana went from 54 below to 49 above in about a day, a 103 degree swing. On January 22, 1943, Spearfish, South Dakota went from 4 below to 45 above in two minutes. Chinooks break a cold snap, but they're hard on fruit trees too: a week of warm wind in late winter can pull buds and bark out of their hardiness right before the next arctic front.
- High-pressure lids. A big ridge of high pressure pushes air down from high up, and sinking air warms by the same compression. That sets a warm lid over the valleys. The warm air is right there, a few thousand feet up, and it can't get down, so the pool under it keeps getting colder. On a valley floor the worst cold isn't only cold blowing in. It's warmth that never arrives.
- Mountains and rain shadows. Air forced up over a range drops its moisture on the windward side. The lee side gets drier air and clearer skies, so it loses more heat on calm nights and swings harder between day and night. Valleys that still get some Pacific moisture in winter get more cloud, and cloud slows the cooling.
- Snow cover. Snow is a blanket for the soil and a radiator for the air. It keeps the ground's heat from reaching the air and gives off infrared almost perfectly, so the air over snow gets colder. Roots under a foot of snow stay safe. Wood above the snow takes the full cold, and on a young tree you'll often see the winter injury stop right at the old snow line.
- Height. Above the inversion the air cools with height, on average about 3.6°F per 1,000 feet. On arctic nights it cools a lot less than that, which is why the free-air cooling in our model is so gentle.
Your yard is an airshed too
The same physics works at the scale of a lot. Cold air flowing downhill stops at the first thing that dams it: a solid fence, a tight hedge, a house, a road fill, a berm. The uphill side of that dam turns into a frost pocket. Leave gaps at the bottom of fences and hedges on the downhill side, or put the windbreak uphill of your trees so it turns the cold air around them. Plant on the slope, not at the foot of it. A tree a little way up a slope can come through a frost that takes the one at the bottom. North slopes hold bloom back a few days, which can save an apricot crop. The south and southwest sides of trunks take sunscald in late winter, when afternoon sun warms the bark and it freezes hard at sunset. White trunk paint keeps the bark cool.
What missing heat does to what you can grow
Winter. Hardy wood survives by controlling where the ice forms. Many hardy plants let water move out of their cells into the spaces between them, where it freezes without hurting anything while the cells dry down. Fruit buds and the living wood of many fruit trees use a different trick called deep supercooling: they hold their water liquid well below 32°F. That works down to a hard limit near 40 below, about as cold as pure water can get without freezing (and the one spot where Fahrenheit and Celsius read the same). Past that the water inside the cells freezes, and the cells die. Plants that rely on supercooling tend to stop at about that line on the map.
Hardiness also changes over the season. A tree hardens off in fall as the days shorten and the first frosts come, reaches its peak in midwinter, and loses it in warm spells. The same tree that comes through 30 below in January can be killed by far less cold in early November, or after a week of warm wind in March.
Spring. Blossoms have almost no hardiness at all. These are Washington State University's numbers for full bloom, the temperatures that kill about 10 percent and about 90 percent of the blossoms:
| Fruit at full bloom | Kills about 10% | Kills about 90% |
|---|---|---|
| Apple | 28°F | 25°F |
| Pear | 28°F | 24°F |
| Sweet cherry | 28°F | 25°F |
| Tart cherry | 28°F | 24°F |
| Peach | 27°F | 24°F |
| Apricot | 27°F | 22°F |
| European plum | 28°F | 23°F |
Source: Washington State University bud hardiness research, as compiled by Michigan State University Extension.
On a still night in May a frost pocket can run several degrees colder than the bench above it, and at bloom a few degrees is the difference between a crop and nothing. Frost pockets also shorten the season at both ends, so late apples, pears and grapes may not ripen there even when the trees themselves are fine.
The whole system. An airshed is a system. Energy comes in as sunlight, as warm air from somewhere else, and as heat given off when water vapor condenses and water freezes. It goes out as infrared to the sky and as cold air draining in off the slopes. It gets stored in soil, lakes, buildings and the trees themselves. The wild plants already show where the balance lands. In Vermont, a 2024 U.S. Forest Service study found that cold-air pools turn the forest upside down: spruce and fir on the low slopes, where you'd expect them up high, and hardwoods above them. An orchard is one more part of that system, and every tree you plant is a bet on the coldest night your spot will hand it.
How the maps read your microclimate
We split the drivers in two. The big ones from above (arctic air, chinooks, clouds, high-pressure lids) are already in the weather station records: thirty winters of them, blended from the eight closest stations the way our cold hardiness calculator does it. The ones that change from one side of a road to the other come from our microclimate model, one parameter per driver: height above the cold pool for the thermal belt, drainage channels for the cold air's creeks, side-valley bottoms and hollows for the little pools, slope for ground that sheds its cold, town for thermal mass and blocked sky, tree cover for canopy, open water and big water for stored heat, and wet ground and hay ground for open, damp fields that lose heat fast. The table further down has every weight, and the write-up on each valley's map covers which of these drivers matter most there.
How the number is built
Every number on our maps is a 30-year planning low: a 1-in-30-year number from an extreme value fit to each station's coldest night of every winter. It is not an average and it's not the all-time record. It's the same number our cold hardiness calculator gives, built the same way, so the maps and the calculator agree with each other:
- The stations. For each long-record weather station near you (NOAA co-op, airport, SNOTEL and fire-weather stations) we take the single coldest night of every winter back to 1995-96, throwing out winters with holes in December, January or February. Those yearly lows go through an extreme value fit (Gumbel), which gives each station its own 1-in-30-year low.
- Stations that closed, and airport stations NOAA doesn't carry. Where a town's NOAA station shut down years ago, we don't let the map fall back on stations in a different setting. We line up the old station's coldest nights against nearby stations that kept running through both eras, then translate those stations' 1995 to 2025 winters into what the old station would have seen (a record-extension method borrowed from streamflow work). Tested blind on 485 stations, that lands within about 2°F of the real number on average, closer than filling the spot in from the stations around it. We also add airport weather stations from Iowa State's airport archive that NOAA's station set leaves out, after checking them against nearby NOAA stations. Both kinds go into the blend like any other station.
- The blend. We take the eight nearest stations and weight them by distance, by how close their elevation is to yours, and by how many winters they have on file, then take a robust average so one oddball station can't drag the number around.
- The lay of the land. The calculator reads the terrain around you off a 100-foot elevation model: how much of the ground nearby sits lower, how far you have to go to drop 50 feet, how closed in the horizon is.
- The ground right under you. That's what the maps add. Our microclimate model scores every spot from about 165 feet of ground resolution for the things that move cold air in a neighborhood: creek bottoms and draws that carry cold air downhill, little hollows and rises, slope, side-valley bottoms, town heat, tree cover, wet ground and hay ground. We keep only the part that changes within about half a mile, because the stations and the calculator already cover anything bigger.
How much we trust that last step. In October 2026 we tested it against every long-record weather station inside our first 46 maps, 515 of them: we hid each station, rebuilt its number from the stations around it the way the calculator does, and checked whether adding the fine-scale terrain moved the number closer to what the station actually recorded. At station sites (town lots, airports, ranch yards) it didn't. So the 30-year number on each map takes the fine-scale terrain at a quarter of its full strength, a nudge of a degree or two in a draw or on a rise, not a rewrite. Stations almost never sit in the frost pockets and thermal belts the model is built to find, so we can't prove those parts either way yet. The full-strength pattern drives the spring frost layer, which is where cold-air drainage does most of its damage, and the "why" in each click.
A conservative number. The same station check showed that the Gumbel fit leans cold in this part of the country: since 1995 the stations have actually reached their 30-year number about a quarter as often as a true 1-in-30-year low would be reached. Treat "plan for this" as a safe planning number, not a forecast. It's still the number our calculator uses, so the two always agree.
The calculator also asks you whether cold air settles on your ground. Inside a mapped area the map has already looked, and the spring frost layer shows you where the cold drains and pools.
Reading the weather station table
Every map lists the long-record stations in and around it, with each one's own numbers since the winter of 1995-96. The last column is that map's 30-year number right at the station, which won't match the station exactly: it's a blend of the stations around it, the way the calculator works everywhere.
The records come from NOAA daily and monthly data through the Applied Climate Information System (ACIS), Nov-to-Mar winters 1995-96 through 2025-26, a winter counted only when December, January and February are all on file. Stations need at least ten such winters. Rows marked "carried forward" are old NOAA stations whose records ended before 1995, carried forward to today's climate as described above (the Winters column shows the years they ran). Rows marked "gaps filled" are NOAA stations still reporting but missing too many recent winters, filled in the same way. Rows marked AWOS are airport stations from the Iowa Environmental Mesonet.
What goes into the terrain score
Every terrain parameter has a weight in degrees Fahrenheit (the most it can move a site on a bad night), a way to find it in the data, and a reason we believe it. These are the weights we set and checked in the Missoula Valley, the calibrator for the whole series. The full model drives the spring frost layer and the "why" in each click; the 30-year number takes only the fine-scale part of it, at a quarter strength, as described above.
| Parameter | What it does | How we identify it | Winter | Spring |
|---|---|---|---|---|
| Height above the cold pool | Cold air is heavy and fills the lowest ground first; the higher you sit above the local floor, the more of the pool you're above, up to a point. A bench only helps if the cold air can get off it. | USGS 3DEP elevation. Height above the low ground within 2.5 miles, fading out 600 ft above the floor; scaled by local relief or slope. | +3.6 max | +8 max |
| Free-air cooling above the belt | Above the inversion it gets colder with height again, slowly on arctic nights. | Height above the main valley floor beyond 600 ft. | −1.5°F / 1,000 ft | −2.5°F / 1,000 ft |
| Bottom of a side valley | A draw or side canyon has its own little pool even high above the main floor. | Height above the lowest ground within 2,000 ft. | −1.5°F | −3°F |
| River bottomland | The floodplain is the valley's own cold-air drain and its lowest ground. | Distance to the river channels, on low floor ground only. | −1.5°F | −2.5°F |
| Slope | Sloping ground sheds its cold air downhill; flat ground keeps it. | Gradient from the elevation model, credit up to 10 percent. | +1 | +2.5 |
| Local hollow or rise | Small dips collect cold; small rises shed it, even a few feet. | Elevation minus the average within 1,000 ft. | ±1 | ±2.5 |
| Cold-air drainage channel | Creek bottoms and draws carry cold air downhill like water. | Flow accumulation routed downhill over the elevation model (pits filled so flat ground drains), on sloping ground. | −2°F | −3°F |
| Thermal mass (town) | Buildings and pavement store daytime heat and give it back at night. | NLCD developed classes weighted by building mass, smoothed over 1,000 ft. | +2.5 | +2 |
| Tree cover and shelter | Canopy cuts radiative loss and breaks the wind. | Fraction of forest and woody wetland within 800 ft. | +1 | +1.5 |
| Open water | Rivers and ponds give back a little heat. | Fraction of open water within 1,000 ft. | +0.5 | +1 |
| Wet ground | Saturated soils and wetlands run cold. | NLCD wetland classes within 650 ft. | −1°F | −1.5°F |
| Irrigated hay and crop ground | Open, wet, mown fields radiate hard and frost early. | NLCD pasture and crop classes within 650 ft. | −0.5°F | −1°F |
| Big water | Only for water that stays open most winters (Flathead Lake, Lake Pend Oreille, Cook Inlet). Strongest at the shore, gone 2 miles inland or 600 ft above the water. | Distance to water bodies over 8 square miles, fading with height. | up to +12°F | up to +6°F |
| Winter sun | South slopes warm their soil and lose snow sooner. Small at night. | Aspect times slope. | ±0.3 | 0 |
Things we know matter and haven't mapped: soil texture and drainage, snow cover, how long a cold spell lasts, and warm spells in late winter that pull trees out of dormancy before the next cold shot. Those go in the write-up, not the color.
Our tool and how we zone our plants
We want people to know not only that our zone hardiness calculator is specifically designed by us, but also that our plants are zoned according to what we know:
- First hand experience.
- Second hand, from either our friends in the industry or what customers have reported.
- Lastly, if we don't have those first two, from people we don't know personally who have reported that a given plant survived down to a given temperature.
We go off of weather stations on site whenever possible.
Limitations
We know that even with all of the work we've done to zone our plants according to what each plant has survived down to, and coupling that with what we've determined to the best of our abilities a given area has experienced in terms of extreme winter lows, it isn't perfect. We aren't claiming that if a person follows this approach it's bulletproof.
There are many other factors that can contribute to a plant winter-killing, including but not limited to how long the cold lasts, wind exposure, soil characteristics, how much carbon the plant has stored, and warm periods followed by cold that may not even be extreme cold. We could go on.
But if we're going to use any approach as a benchmark for winter survivability, given the data we have at our disposal, we think this approach gets us there.
Find your spot. Pick your valley on the microclimate maps, or put any address into the cold hardiness calculator to see every plant we grow that has come through that cold.
See the microclimate maps Open the cold hardiness calculator