Five seasons of last frost dates on one allotment

01

Air frost, ground frost, and the one that matters

An air frost is a screen minimum at or below zero, taken at 1.25 m in shade. A ground frost is the thermometer lying on short grass reading at or below zero. On a clear, still, dry night the grass level figure runs three to four degrees below the screen here, so there are a great many nights with a ground frost and no air frost at all. Those are the nights that take the tops off early potatoes.

For five seasons I recorded both, and the gap between the last air frost and the last ground frost has averaged nineteen days. That single fact changed how I plant. If you use the air frost date you are planting out into a fortnight or more of nights that can still put a frost on the soil surface where a small plant actually lives.

The third number, which I did not start recording until 2021, is the soil temperature at 100 mm taken at the same hour. It matters less for damage and more for whether anything will germinate. Seed sown into soil at seven degrees sits and sulks and often rots, and the difference between a row sown on 8 April and the same row sown on 22 April has repeatedly been nothing at all by the middle of June.

02

The five dates, as recorded

The last air frost on this plot fell on 3 April in 2020, 21 April in 2021, 14 April in 2022, 6 May in 2023 and 27 April in 2024. The last ground frost in the same years came on 28 April, 11 May, 2 May, 19 May and 14 May. The spread across five seasons is more than a month in both cases, which is the first thing the record taught me: there is no such thing as the frost date, only a range with a tail.

The 2023 figure is the one worth dwelling on. That May frost came after nine days of mild weather, with a screen maximum of 19.5 C on 2 May, and it caught a lot of people on the site including me. The screen minimum on the night of 5 to 6 May was minus 0.5 C and the grass minimum was minus 3.5 C. I lost two rows of dwarf beans that had been out for a week and a courgette plant that I had been smug about.

Averaging those five last ground frost dates gives 9 May, but the average is the least useful number in the set. What I actually use is the latest of the five, 19 May, plus a few days of caution, which is why nothing tender goes out on this plot before the third week of May regardless of how the first fortnight of the month has felt.

• Last air frost, 2020 to 2024: 3 Apr, 21 Apr, 14 Apr, 6 May, 27 Apr
• Last ground frost: 28 Apr, 11 May, 2 May, 19 May, 14 May
• Mean gap between the two, nineteen days
• Coldest spring grass minimum recorded, minus 4.5 C on 8 April 2021

03

The cold corner by the hedge

The plot is not one place. It slopes very gently, perhaps a metre over its length, down towards the hawthorn hedge at the western end, and on still clear nights cold air collects there and cannot get out because the hedge is thick to the ground. I only discovered how much difference that made when I borrowed a second grass level thermometer for a fortnight in April 2022 and put it at the bottom end.

Over fourteen nights the bottom end ran an average of 1.8 degrees colder than the position by the shed, and on the two clearest nights it ran 3.0 and 3.5 degrees colder. On one of those nights the shed end recorded plus 0.5 C and the bottom end recorded minus 3.0 C, which is the difference between a safe night and a dead row of beans within twenty metres.

I have since cut a gap about a metre and a half wide in the hedge at the lowest point, with the agreement of the plot on the other side, and the cold pocket has become noticeably less severe. It still exists. The seed bed is now at the top end and the bottom end grows soft fruit and rhubarb, which do not care.

04

Where my plot disagrees with the county figure

The published long term averages for this part of the county give a last air frost around the middle of April, and by air frost my five year record is not far off that, averaging 19 April. The trouble is that nothing I grow lives at 1.25 m in a ventilated box. The plants live on the soil surface, where my record says frost is possible for another three weeks.

There is a second reason the county figure reads early for me. The averaging period covers a longer run of years than I have, and it is taken at sites chosen to be representative rather than sites chosen because someone rents them for growing vegetables. An allotment on the edge of a city, at the bottom of a slight slope, behind a thick hedge, is exactly the sort of place that runs colder than the regional number on the nights that count.

So I keep both. The county figure tells me what a normal year looks like in general terms. My own five dates tell me what this rectangle of ground has actually done, and when the two disagree about my plot, I trust the thermometer that is standing on it.

05

What the frost record changed

Three practical changes came out of it. The first is the date: nothing tender goes out before 20 May, and I have stopped being talked into it by a warm first week. The second is that I now sow a second, smaller batch of beans and courgettes about ten days after the first, as insurance, which has been used in two of the last four seasons.

The third is fleece discipline. I used to put fleece on when it looked cold, which mostly meant when it was cloudy and damp, which is precisely when frost does not happen. The record made it obvious that damage comes on clear, still nights after a bright day, usually with a screen minimum forecast of plus two or lower. Those conditions are now the trigger, and it has meant fleece on about seven nights a season rather than twenty.

None of that is clever and most experienced growers on the site knew it already in a general way. The difference is that I now have the dates written down, so when somebody tells me in the third week of April that the frosts are over, I can go and look at 6 May 2023 in the notebook and decide to wait.