Garden microclimates

The station is a point, the garden is a place

Every frost date on this site is the date at one instrument. The page names the instrument and the miles from the city coordinate we used. Those miles are the first microclimate fact, and they are measured. Everything else people mean by microclimate — a wall, a hollow, a fence that stops cold air — is real enough that university extensions write about it, and it is not in the station file. The station file does not contain a number for “your backyard is 2°F warmer.” This guide keeps the measured miles and the published advice in separate paragraphs.

Oregon State University Extension’s “Growing Your Own” (EM 9027) tells gardeners to pick a site with full sun, protection from wind, and a slight south- or southeast-facing slope. The slope warms early in the spring and lets colder air drain to a lower location. The same publication says frost builds up where air cannot drain, including low spots at the base of a hill and the foot of a slope bordered by a solid fence. An older OSU site-planning bulletin says the same physics in plainer words: cold air is heavier than warm air, it flows downhill like water, it dams up behind obstacles, and it settles in low spots. A garden in that low spot gets the late spring frosts and the early fall frosts and loses weeks of growing time. The suggested placement is well up on a south-facing slope. A barrier on the north side, a stone wall or a stack of hay bales, can divert the draining cold air and hold some of the day’s heat. The bulletin also says a solid wall used as a windbreak is a poor windbreak, because it creates strong downdrafts on the lee side. A break that is about 50 percent permeable, such as a double row of deciduous hedge or a run of snow fencing, is the form they prefer.

Those are placement rules. They do not come with a degree value for your fence. PNW 548, the OSU and Northwest extension publication on winter vegetables west of the Cascades, adds three observations that are easy to misuse if they are turned into a formula. Gentle slopes let cold air drain away and settle below, so the bottom of a hill is a place to avoid. South- or west-facing slopes receive the most heat in winter. Sites within 20 feet of a structure may be warmer than sites farther away. Raised beds, which help drainage in wet soil, can also make crops more susceptible to cold injury. The 20-foot note is a reason to look at the bed next to the house differently from the bed at the back fence. It is not a license to subtract a week from the station’s median frost date.

OSU’s Central Oregon guide (EM 9662) uses the word microclimate for a backyard, or a portion of a backyard, as opposed to a region. It lists cold pockets, hills, and the location of houses as the factors, and it says a hardiness zone is a guide to minimum temperature, not an absolute. The USDA map places that region in zone 6, a band of −10°F to 0°F, and the guide says the single zone is too coarse for the local mix of sites. Soil, rainfall, heat, wind, and sun still decide whether a plant lives. That is the same split this site keeps: the zone is the average annual extreme minimum for a ZIP, and the frost dates are a station’s probabilities. Neither one is the reading from a thermometer on your mulch.

Two Austin stations, one city

Austin’s city page does not use Austin-Bergstrom. The nearest qualified station to the Census internal point used for the city is Austin-Camp Mabry, USW00013958, 1.6 miles away, at an elevation of 204.2 meters, with a standard completeness flag and 30 years in the 32°F calculation. Bergstrom, USW00013904, is the worked example in the probability guide. It sits at 146.3 meters. It is farther from the city point, so the city page uses Camp Mabry.

The medians do not match. Camp Mabry’s median last 32°F date is February 20. Bergstrom’s is March 12. Camp Mabry’s median first 32°F date is December 3. Bergstrom’s is November 16. The median growing-season length is 291 days at Camp Mabry and 252 days at Bergstrom. The 10% last-32°F dates are March 14 at Camp Mabry and April 2 at Bergstrom. Someone who memorized “Austin freezes until mid-March” may be remembering the airport. The city page is reporting Camp Mabry, and it says so.

The files do not say why. Camp Mabry is the higher station and has the earlier median last freeze, so a simple “higher means colder” story does not fit these two rows. Urban surroundings, the Colorado River valley, the airport’s open ground, and the exact siting of the instruments are all candidates, and none of them is a column in the normals file. If two qualified stations inside one metropolitan area disagree by weeks, a garden that sits at neither instrument can disagree with both. Use the nearest station and read the miles. The city page reports Camp Mabry and leaves Bergstrom as the separate example.

Denver is the contrasting pair. The Denver city page uses Denver-Stapleton, USW00023062, 0.6 miles from the Census point, median last 32°F on May 4 and median first 32°F on October 8. Denver International Airport, USW00003017, the station used as a northern check in the probability guide, has a median last 32°F date of May 5 and a median first date of October 9. The elevations are 1,611.2 meters and 1,650.2 meters. Here the next well-known station barely moves the calendar. Distance and disagreement are not the same fact. Austin shows that a few miles can matter. Denver shows that a different airport can matter almost not at all. You only know which situation you are in by reading both rows, not by assuming cities behave like Austin.

A coordinate can be the wrong point, too

San Francisco’s Census internal point is a poor stand-in for the neighborhoods, because the city’s legal boundary includes the Farallon Islands and the internal point is pulled west. A zone match farther than 30 miles is not used. When the Census point is more than 15 miles from the nearest zoned ZIP, the page uses a GeoNames populated-place coordinate of the same name in the same state. The San Francisco city page uses that second coordinate. The zone ZIP is 94102, zone 10b in the 2023 PRISM file, about 0.3 miles from the point. The station is San Francisco Oceanside, USC00047767, 5.7 miles away, elevation 2.4 meters, representative completeness, 24 years. Its median last 32°F date is December 30, and its 10% last 32°F date is March 13. The median growing-season length is 365 days. That is a coastal instrument. A garden on the east side of the city, against a wall, is a different site. The 5.7 miles are on the page for that reason. The zone is the ZIP’s average extreme minimum, which is a winter-cold statistic, not a statement that Oceanside’s December median applies to every block in 94102.

Anchorage shows the same rule at the scale of a state. The 2023 ZIP table has no Alaska rows, so the city is shown with frost dates and without a zone. The nearest qualified station, Anchorage Merrill Field, is 1.5 miles from the city coordinate. A Washington or Texas ZIP is not borrowed to fill the gap.

What to change in the garden, and what not to change on the page

The OSU publications support a short list of physical choices. Do not put the tender bed at the bottom of a slope or against the uphill side of a solid fence, where cold air pools. A south- or southeast-facing slope is the spring site they describe. A permeable windbreak is preferred to a solid wall if the problem is wind rather than cold-air drainage. A north-side barrier is a different object, meant to divert cold air, and the site-planning bulletin allows a solid one for that job. Winter vegetables west of the Cascades are pointed toward south- or west-facing slopes and toward the warmer air within about 20 feet of a structure, with the caution that a raised bed can be colder for the crop even when it is drier. None of these sentences includes a revised frost date.

You can use the station’s 10% and 90% dates as the spread that already exists in the climate record, and then let the garden’s position tell you which side of that spread you are more likely to resemble. A bed in a hollow is a reason to respect the later spring date, the 10% last-32°F date. A bed on a south slope against a house is a reason to watch the forecast and perhaps to plant toward the early end of an extension window, still inside the window. Texas A&M’s own older statewide guide notes that home gardens can sometimes be planted 10 days to 2 weeks earlier than commercial fields because buildings, trees, and shrubs protect them. That sentence compares home gardens with commercial fields in that guide. It is a published allowance you can choose, and the forecast the week you plant still matters. The late-frost guide is what to do when the night arrives anyway.

The zone and the frost date will keep disagreeing, because they measure different things. Camp Mabry’s median last 32°F date can be February 20 in a city whose ZIP is zone 9a, a band of 20°F to 25°F for the average annual extreme minimum. February 20 does not tell you the January low. The 9a band does not tell you February 20. How frost-date probabilities work and what changed in the 2023 map split those products. A microclimate sits on top of both, at a scale neither product was built to resolve.

Sources

• Oregon State University Extension, “Growing Your Own,” EM 9027: https://extension.oregonstate.edu/catalog/pub/em9027-s
• Oregon State University Extension, “Planning and Preparing Your Vegetable Garden Site,” cold-air drainage and windbreaks: https://ir.library.oregonstate.edu/downloads/2514nk77g
• Oregon State University Extension, PNW 548, winter vegetable production west of the Cascades, microclimate section: https://extension.oregonstate.edu/catalog/pnw-548-winter-vegetable-production-small-farms-gardens-west-cascades
• Oregon State University Extension, “Gardening in Central Oregon’s climate,” EM 9662: https://extension.oregonstate.edu/catalog/em-9662-gardening-central-oregons-climate
• NOAA NCEI 1991–2020 station access files for USW00013958, USW00013904, USW00023062, USW00003017, and USC00047767.
• City coordinates, distances, and the San Francisco fallback are the rules on the methodology page.