When do I start these seeds?

Every sowing instruction on every seed packet is written relative to one number: the average date of the last spring frost where you are. "Start indoors six to eight weeks before last frost" is meaningless without it and completely determinate with it.

Three ways to get yours, in descending order of reliability:

• Your national weather service. In the United States, NOAA publishes freeze and frost probability tables by station, and most state extension services republish them by county. In the UK the Met Office publishes equivalent averages.
• A local grower. An allotment neighbour, a farmers' market stallholder or a garden club will give you a date that already accounts for the valley you are in. This is often better than the official figure.
• The zone estimate in the table below, which is a starting point and nothing more.

Two things worth understanding about the number itself. It is an average, which means roughly half of years will have a later frost than the date, and published tables usually give a 50% probability date and a 10% probability date — the second is the safe one for anything frost-tender. And it is a regional figure: a low-lying garden where cold air pools can run two weeks behind a slope half a mile away, and a city centre can run two weeks ahead of its own suburbs.

The USDA plant hardiness zone is the single most misused number in gardening. It describes average annual extreme minimum winter temperature, in 5.5 °C bands, and it exists to answer one question: will this perennial survive the winter here?

It says nothing about when spring arrives, how long the growing season is, or how hot the summer gets. Seattle and Oklahoma City are both around zone 7 or 8 and share almost no growing conditions. That is why the calculator asks for a frost date, and why the zone table below is offered only as a rough starting estimate for people who have no better figure.

The commonest mistake in seed starting is enthusiasm in February. A tomato sown ten weeks before last frost rather than six spends four extra weeks in a pot on a windowsill, and the result is a tall, pale, weak-stemmed plant that transplants badly and fruits later than one sown a month after it.

The mechanism is light, not warmth. A south-facing windowsill in March delivers a small fraction of the light a seedling gets outdoors, and a plant grown short of light stretches for it — long internodes, thin stems, pale leaves. Once a seedling is leggy it does not recover; you start again.

Four things fix nearly all of it:

• Light close and long. A cheap LED shop light held 5 to 10 cm above the leaves, for 14 to 16 hours a day, outperforms any windowsill. Raise it as the plants grow.
• Warm to germinate, cool to grow. Most seeds want 20 to 25 °C in the compost to sprout and then noticeably cooler air to grow stocky. A heat mat under the tray and a cool room after emergence is the standard combination.
• Air movement. A small fan on low for a few hours a day thickens stems and largely prevents damping off, the fungal collapse that kills seedlings at soil level overnight.
• Water from below and let the surface dry. Constantly wet compost is what damping off needs.

A seedling raised indoors has thin cuticles, soft tissue and no experience of ultraviolet light or wind. Moved straight outside it can be scorched white in an afternoon or snapped off at the stem. The transition takes seven to ten days and there is no shortcut.

1. Days 1–2: one to two hours outdoors in full shade, sheltered from wind, then back inside.
2. Days 3–4: three to four hours, with an hour of morning sun.
3. Days 5–6: most of the day out, with several hours of direct sun.
4. Days 7–8: all day out, in overnight only if frost threatens.
5. Days 9–10: out overnight, then plant — ideally on a still overcast evening rather than a bright morning.

Reduce watering slightly during the process and stop feeding. A slightly checked plant hardens faster than a lush one. And bring everything in if a late frost is forecast, whatever day of the schedule you are on.

The calculator marks these direct-sow only, and the reason is root architecture rather than tradition.

Root crops — carrots, parsnips, beetroot, radish, turnip — form a taproot that forks permanently if it is disturbed. A transplanted carrot is a forked carrot. Sow them where they are to grow, thin ruthlessly, and accept that thinning is the price of a straight root.

Beans, peas and sweetcorn germinate fast and grow so quickly in warm soil that a direct-sown plant overtakes a transplanted one within a fortnight. They also resent root disturbance. There is nothing to gain.

Cucurbits — cucumber, squash, melon, pumpkin — sit in between. They dislike disturbance but benefit from a head start in short-season regions. If you start them indoors, do it three weeks before last frost at most, in a large individual pot or a soil block, and plant out without breaking the root ball.

The general principle worth carrying: the value of starting indoors is a head start on a long-season crop, and the cost is root disturbance. Tomatoes, peppers and aubergines need 8 to 12 weeks of warmth to fruit and will not manage it from a direct sowing in most of the temperate world, so the trade is worth it. Radishes are ready in four weeks and it never is.

The output above gives two dates per crop: when to sow, and when it goes in the ground. Three habits make it work in practice.

Work backwards from the transplant date, not forwards from today. The transplant date is fixed by the weather; the sowing date is derived from it. Anyone who sows first and then waits for the weather ends up with leggy plants.

Stagger a succession. Sowing all your lettuce on one day produces a glut and then nothing. Sow a short row every two to three weeks through the season for lettuce, radish, beans, spinach and carrots.

Label everything as you sow it. Every seedling in the brassica family looks identical for the first three weeks, and tomato varieties are indistinguishable until they fruit. Write the variety and the date on the label; the date is what tells you whether a slow tray is behind or normal.

Finally, keep a record of what actually happened — the real last frost, what bolted, what never got going. Two seasons of your own notes beat every published table, because they describe your garden rather than your region.