Almost every South African gardening page answers this the same way. Plant tomatoes in spring, after the last frost. It is correct, and it is useless, because not one of those pages tells you when the last frost actually is where you live.
This is actually one of the more useful things I’ve seen posted here, most gardening advice in SA is just recycled Northern Hemisphere content with “adjust for your climate” bolted on, which helps nobody.
I’m in Pretoria and my own experience backs up why the town level detail matters so much. We can get a late frost creeping in even into September some years, especially in the lower lying spots around Centurion and further out toward the East Rand where cold air pools overnight. Meanwhile Joburg proper on the ridge is often a week or two ahead of us just because of elevation and drainage.
One thing I’d add from running temp sensors off my Home Assistant setup outside, microclimate in your own garden matters almost as much as your town. A bed against a north facing wall can be frost free weeks before an open bed ten metres away. If anyone’s serious about dialling this in, a cheap Zigbee temp sensor logging overnight lows for a season teaches you more than any regional guide ever will.
Did you use a specific weather dataset for the town level dates, or station records?
Fair question, Kabelo, and the microclimate point you raise is real, a town-level date can’t see a north-facing wall or a low-lying frost pocket ten metres from it.
For the sourcing, it’s not a single model run. The base layer is the ARC-ISCW national frost surface from 2003, which maps South Africa in 10-day probability bands rather than a single date per place. On its own that surface runs warm-biased against real ground truth, so it’s cross-checked against station records from a 2016 paper by Moeletsi, Tongwane and Tsubo covering nine Free State weather stations. Measured against those stations the combined approach lands within about 7.4 days on average, worst case around three weeks.
Because even a calibrated date is still a townwide average, the published date isn’t the raw calibrated one. There’s a safety margin added on top, sized from how far wrong the calibration ran in the unsafe direction across those nine stations, at the 80th percentile rather than the mean, so it protects most gardens rather than just the average one. That’s the built in answer to your wall versus open bed problem, it won’t tell you your own yard’s microclimate, but it’s deliberately padded so a typical yard doesn’t get caught out.
If you’re tracking your own beds with sensors already, that’s genuinely better data than any townwide date can give you, the tool is there for people who don’t have that setup yet.
Down here in Durban, the frost advice has always been a bit of a joke because we haven’t seen ground frost in Berea in living memory, so the real question is which of our humidity windows to dodge. I’ve had far better results planting in late February or early March, once the worst of the summer rain eases off, than in September when the heat builds fast and the whitefly moves in before the plant even sets fruit properly.
Randburg catches late frosts well into August, and I’ve had tomato seedlings die on me in early September because I followed generic spring advice written by someone sitting in Durban. A resource that maps this by actual town is the only version of this guide worth bookmarking.