Zone, frost dates and latitude are three different numbers that answer three different questions. Nearly every garden tool uses the first one for all three.
Three questions, three numbers
A garden planner has to answer three separate questions about where you live, and they really are separate:
- Will this perennial make it through the winter here?
- When can I sow this annual, and will it finish before frost?
- Will this onion form a bulb at my latitude?
The answers come from three different numbers. Your hardiness zone answers the first. Frost dates and accumulated heat answer the second. Latitude answers the third. Using one of them for all three is the single most common mistake in this whole category, and it's nearly everywhere.
What a zone is, and what it isn't
A USDA hardiness zone captures exactly one thing: the average coldest night of the year. It's a winter-survival number. It tells you whether a fig will live - not when to plant a tomato.
Well establishedHardiness zone is the correct variable for perennials, and only for perennials.
suboptimal if ignored → advise
Mechanism: Zones encode average annual minimum winter temperature, which governs whether a woody perennial survives to spring. It is the wrong quantity for an annual, which never sees winter.
THE CEILING IS A SECOND AND DIFFERENT MECHANISM, and until 2026-07-26 this rule enforced it without naming it. A zone too WARM does not kill a perennial; it fails to chill it. Deciduous fruit crops need a winter dormancy period to break bud and set fruit, so in a zone milder than a plant is adapted to it may survive and never crop. Until ISSUES #13 was decided the band was enforced at both ends; now (option E) only the cold end gates — it alone follows from the definition of the zone system — and the warm end is carried as an ADVISORY, not a refusal.
AND THE WARM END IS A CULTIVAR'S ANSWER, NOT A SPECIES', which is the honest limit of what this rule can do. Chill requirement is published per variety and varies enormously within a species, so a species-level ceiling says "the modelled range does not reach your zone", never "no plant of this kind will crop for you". So the engine offers the plant above its band and advises choosing a cultivar rated for the zone rather than refusing it — the same move invariant 3 makes for soil, applied to chill.
Evidence: confirmed against 3 sources in the research-based literature.
the 3 sources
THE FLOOR IS TRUE BY DEFINITION and needs no citation: the USDA system defines a zone AS a band of average annual minimum winter temperature, so 'a plant colder-hardy than your zone survives your winter' is the definition restated. This rule was `verified_trivially` on that basis alone until 2026-07-26. THE CEILING WAS NEVER COVERED BY IT - a minimum-temperature definition says nothing about a plant failing for want of chill - and that gap is why ISSUES #13 exists. The pointer below closes it.
OSU Extension EC 1303, 'Growing blackberries in your home garden' (Bernadine Strik), catalog.extension.oregonstate.edu/sites/catalog/files/project/pdf/ec1303.pdf (read 2026-07-26). THE CEILING'S MECHANISM, STATED: 'Winter cold-hardy cultivars adapted to the colder regions of Oregon (such as zones 5 to 7), may not grow normally in regions of the Willamette Valley (zone 8) where winters are warmer and plants receive less winter chill.' A plant failing in a WARMER zone for want of chill - which is what this rule's upper bound enforces and what no source said until now. THE SAME PAGE MAKES IT A CULTIVAR PROPERTY: 'Even cultivars within the same type (trailing, erect or semierect) differ in fruit quality and flavor, appearance, tolerance to pests, cold hardiness, and plant longevity', and the remedy it gives is the one this rule's rejection text now repeats: 'Minimize risk by choosing types and cultivars that are adapted to your hardiness zone.' The page then publishes a table of sixteen cultivars EACH WITH ITS OWN ZONE RANGE, which is the strongest available evidence that a species-level band is an approximation. [read]
WHAT THIS DOES NOT SETTLE. The mechanism is now named, but the corpus still carries ONE band per species where the source carries one per cultivar, and 12 of 25 perennial zones have no citation at all. So the ceiling is explained without being precise. ISSUES #13 stays open on that, and the rejection text is deliberately phrased as 'the modelled range does not reach your zone' rather than as a prediction that nothing will crop.
Read the full rule → · open in the planner ↗
Two places in the same zone can have growing seasons weeks apart, because the coldest night of the year and the date of the last spring frost aren't the same thing, and they aren't closely tied to each other. A mild coastal spot and a place deep inland can share a zone number and have almost nothing else in common.
So this planner uses zone where zone applies, and nowhere else.
Scheduling is a frost question, not a zone question
When to sow depends on your actual frost dates and on how much heat the season builds up, which is what decides whether a crop can finish in time.
Well establishedTender species must not be scheduled before the site's last frost date.
fatal if ignored → block
Mechanism: Ice nucleation in leaf tissue.
Evidence: settled without a trial - climatology.
remedy
Sow or set out after the last frost date, or start indoors and transplant once it has passed.
Read the full rule → · open in the planner ↗
Well establishedA species must reach maturity within the site's frost-free window.
fatal if ignored → block
Mechanism: Accumulated GDD vs gdd_to_maturity, or DTM vs frost-free days.
Evidence: settled without a trial - GDD models.
remedy
Transplant, choose a shorter-DTM cultivar, or refuse the species for this site.
Read the full rule → · open in the planner ↗
Look at what that second one is really asking. It isn't asking whether the plant will survive; it's asking whether the season is long enough for it to finish. A zone can't answer that at all - and it's the question that decides whether your melons ever ripen.
Odds, not a date
There's one more problem with frost dates the way they're usually handed to you. "Last frost, May 15" gets presented as a fact, but it's really a probability. It's the date by which the frost has passed in some share of years - and which share almost never gets said out loud.
This planner shows you the odds instead of a single date, because the odds are what you're actually betting against. Planting on the 50% date means you're taking a coin flip. Waiting for the 10% date costs you two weeks of season. That's a real trade, and you should get to make it yourself.
Where a number is modelled rather than measured, the page says so, and tells you how far away the nearest weather station was:
Well establishedEvery derived climate value carries the tier, grade, station, and distance that produced it.
suboptimal if ignored → advise
Mechanism: Frost percentiles exist only at NCEI stations. Values elsewhere are offsets or model output. A bare date conceals the difference between a measurement and an inference.
Evidence: settled without a trial - definition.
Read the full rule → · open in the planner ↗
The correction we refuse to make
There's a tempting fix here that this project blocks outright. Frost figures come from weather stations, your garden isn't sitting at a station, and the obvious move is to correct for elevation: colder as you go up.
On the nights that actually matter, that gets it exactly backwards.
Well establishedFrost dates can be downscaled by applying a standard lapse rate to elevation.
This is a claim the planner refuses to enforce — recorded here with what is actually known, rather than dropped.
fatal if ignored → block
What is actually known: On clear still nights, which are exactly the nights that produce radiative frost, cold dense air drains downhill and pools. The valley floor is colder than the slope above it. A standard lapse rate predicts the opposite and makes frost dates worse in the terrain where accuracy matters most. PRISM exists because of this.
Evidence: confirmed against 1 source in the research-based literature.
the source
Penn State Extension, 'Frost, Critical Temperatures, and Frost Protection', extension.psu.edu/frost-critical-temperatures-and-frost-protection (read 2026-07-11): 'Radiation frosts occur when the heat that has been stored in the upper soil surfaces is radiated back into the atmosphere at night. Wind speeds are generally low, and an inversion usually develops with temperatures near the ground surface falling to, or below' freezing; 'Cold air is heavier than warm air; therefore, it will naturally flow down to lower elevations. Orchards or orchard pockets located at lower elevations will see more damage due to the collection of this cold air. This is the reason that site selection is critical to avoid low-lying land.'
Read the full rule → · open in the planner ↗
The frosts that bite happen on still, clear nights - and on those nights the cold air slides downhill and pools in the low ground. The bottom of the dip is the frost pocket; the slope above it is warmer. Correcting by elevation predicts the opposite of what really happens, and it does it with a straight face. Sometimes refusing a plausible-looking correction is the accurate thing to do.
And your own ground beats all of it
The model is a starting point, not the final word. Once you've logged enough seasons on a particular patch, what actually happened there takes over from the model for that patch:
Well establishedAn observed per-plot frost offset, fitted from 3+ seasons of logged outcomes on that ground, supersedes all modelled climatology for that plot.
suboptimal if ignored → advise
Mechanism: Garden-scale microclimate (south wall, frost pocket, raised-bed drainage, urban heat island) lives below 800m resolution and below 80m. No gridded product resolves it. The season log is a direct measurement on the exact ground in question.
Evidence: settled without a trial - measurement beats model on the measured point.
Read the full rule → · open in the planner ↗
Your garden has its own little climate that no station knows about: a south-facing wall, a cold hollow, a hedge that keeps the wind off. After a few seasons of your own records, the planner trusts your ground over its own numbers - and tells you which one it's going on.
Latitude, for exactly one thing
The third number gets used once, for daylength:
Well establishedOnion cultivar group must match site latitude.
fatal if ignored → block
Mechanism: Bulbing fires when daylength crosses a cultivar's critical threshold (short ~10-12h, intermediate ~12-14h, long ~14-16h). A short-day cultivar grown too far north crosses its low threshold early and bulbs prematurely on small plants, so bulbs stay small. A long-day cultivar grown too far south never reaches its high threshold and fails to bulb. Changeover is around the 35th-38th parallel.
Evidence: confirmed against 2 sources in the research-based literature.
the 2 sources
NC State Extension, 'Bulb Onions', content.ces.ncsu.edu/bulb-onions (read 2026-07-11): 'onions will not form bulbs until daylength reaches the cultivar's critical threshold. Short-day onions bulb at approximately 10-12 hours of daylength ... Intermediate-day onions bulb at approximately 12-14 hours ... Long-day onions bulb at 14-16 hours of daylength.'
Univ. of Illinois Extension (The Humble Gardener), 'Growing Onions in Central Illinois', extension.illinois.edu/blogs/humble-gardener/2018-03-19-growing-onions-central-illinois (read 2026-07-11): 'Long day varieties are for northern growing areas (above the 35th latitude); short-day varieties are for southern. If short-day varieties are grown in our zone, they won't do well. They will not form bulbs of any decent size. And since the long day onions need more daylight, they will not do well' (in the south).
remedy
Choose long-day onions in the north and short-day in the south; the planner picks the group from your plot's latitude.
Read the full rule → · open in the planner ↗
Onions form their bulbs based on how long the days are. A short-day onion planted far north will grow leaves all season and never bulb up, and no amount of zone information will warn you. It's a latitude question, so it gets a latitude answer.
