The Southern Africa 2026/27 Climate-to-Cashflow Briefing.

The Southern African Regional Climate Outlook Forum released its thirty-third consensus outlook from Swakopmund, Namibia on 26 August 2026, and SADC published the regional summary on 27 August. For October to December 2026 the forum places much of the central and southern subregion in the below-normal rainfall category, with high confidence for the zone covering much of Angola, southern Zambia, Zimbabwe, Mozambique, Namibia, Botswana, most of South Africa, Eswatini and Lesotho. The dry signal persists into January to March 2027, and above-average temperatures are expected across most of the region. El Niño conditions are established and expected to strengthen.

This note is the short version. The technical edition carries the model, the full assumption register, a sensitivity analysis and the sources.

What the outlook withholds

This year the statement does not print a numerical probability table by zone. In previous cycles a reader could take a printed tercile split and carry it straight into a decision model. SARCOF-33 instead uses map categories based on relative probabilities, supported by confidence descriptors in the text.

That is a defensible scientific choice and a practical problem for anyone building a decision model, and it is worth stating plainly rather than papering over: you cannot read a precise probability off this year's statement, so any model that appears to contain one has invented it. Our model therefore assumes no probability. It sweeps a range and reports where the decision changes.

The decision is an option, not a forecast

Take a 200-hectare dryland grain farm. The usual framing asks whether to plant, or how much. The sharper question is what to commit, and when.

Strategy A commits the full input programme at planting. Strategy B commits a base programme and retains the right, not the obligation, to apply top dressing about six weeks later, once establishment and the sub-seasonal update are visible. B is not a cheaper programme. It is the same programme with an option attached, and the option costs money: split application runs about R300 per hectare more in total and carries a small timing penalty.

Modelled across six mutually exclusive states, on two dimensions (the seasonal rainfall category, and whether a damaging dry spell lands at flowering) the option is worth exercising in an above-normal or normal season and is worth abandoning in a below-normal one. It is not free insurance: in the wetter states the farm would have done better committing everything at planting. It earns its keep in the dry states, where it holds back R660,000 of peak outlay until the season has declared itself.

Which produces a conclusion we did not expect, and which matters more than the outlook itself:

Where that is true, the finding is not about the weather. It is that the operation's commitment structure was already fragile, and a dry outlook merely reveals it. Sometimes a forecast changes a decision. Sometimes it exposes one that was wrong in an ordinary year.

Who is exposed, and how

Rainfed summer grain. Two separate risks, not one: the planting commitment against an unreliable onset, and a dry spell at flowering. A season can post near-normal totals and still fail on distribution, which is why the two belong on different axes of any model that claims to price them.

Irrigated annual crops. Allocation dominates. On our parameters a 20 per cent allocation cut removes R216,000 of contribution against R49,374 from higher pumping volumes and the 2026/27 Eskom tariff increase combined. The temperature signal is material but secondary. A permanent orchard is a different problem entirely and is not modelled: it cannot answer an allocation cut by planting less area.

Input supply. Agro-dealers commit stock capital in September against demand that only materialises if farmers plant. Sell-through falling from 92 to 62 per cent takes gross profit after carrying cost down by 44.3 per cent, with credit losses on top, and it lands after the stock has been paid for.

What to do in the next two weeks

Set thresholds before the evidence arrives. A replant threshold decided after seeing the stand is not a threshold, it is a rationalisation. Fix the onset criterion, the stand-count threshold, and the date after which a late replant is not worth the seed. Name an owner for each: a trigger with no owner does not fire.

Price the option rather than the forecast. Order the base programme, then price and reserve the top dressing without committing it. Size the seasonal facility against the below-normal state and hold the difference as headroom rather than converting it into inputs.

Use the right break-even. Operating break-even covers variable and fixed costs. The number that matters for the February debt-service conversation is the cash-solvency break-even, which also covers the debt payment net of opening cash. On this farm those are 2.92 and 3.07 tonnes per hectare respectively. Conflating them is a common and expensive error.

Read the shortfall clause. Pull your three most important contracts this week. If you cannot state what happens at 40 per cent of contracted volume, you do not know your exposure. Force majeure is contract-specific and drought is not automatically covered, so have it reviewed rather than assuming what it will do.

And one question for any lender looking at a farm model this season. Can every displayed number be traced to a named, dated input, with assumptions registered and the calculation visible? A model that produces a confident single figure out of a seasonal outlook has either invented a probability the outlook does not contain, or hidden the assumption doing the work.