Wheat is a crop of cool seasons grown, across a large part of its area, in places where the cool season ends abruptly. On the Indo-Gangetic Plain the crop is sown as the monsoon heat fades and harvested in a spring that warms by the week. The same pattern holds, with local variations, in the Yaqui Valley of Mexico, the Nile valley, the Mediterranean rim and the Australian wheatbelt. Heat that arrives during flowering and grain filling is called terminal heat stress because it comes at the end of the crop's life and usually ends it early.

Global analyses leave little doubt about the scale of the problem. A multi-model assessment concluded that world wheat production falls by about six per cent for each further degree Celsius of warming (Asseng et al., 2015), and simulation work for Australia showed that the damage is driven less by changes in the seasonal mean than by hot days during grain filling (Asseng et al., 2011). A general review of the physiology of heat in the reproductive and grain-filling phases is given by Farooq et al. (2011).

Plant scientists distinguish tolerance, the capacity to keep functioning at high tissue temperature, from escape, in which the sensitive stages simply occur before the heat arrives. Escape is the less glamorous route and the more accessible one. It is controlled by two choices made before the seed is in the ground. This review asks what the literature shows about those choices: how far earlier sowing and better-matched cultivars can move the crop out of harm's way, and what stops them moving it further.

The review is narrative. It draws on the physiological literature on high temperature in wheat, on field studies of sowing date and cultivar duration from South Asia, Mexico and Australia, and on the genetic literature on flowering time, and it does not claim to have covered these systematically. Growth stages are named in the terms of the decimal code of Zadoks et al. (1974). Quality effects of heat on grain protein and dough properties are outside the scope, as is drought, although in rainfed systems the two stresses seldom arrive separately.

Yield in wheat is the product of grains per unit area and weight per grain, and heat acts on each at a different time. Figure 1 sets the phases of the crop against the principal effect of high temperature in each. The distinction matters for escape, because a sowing date that protects grain filling may still leave anthesis exposed, and the reverse.

If heat is a matter of the calendar, the first remedy is to change the calendar. Sowing-date experiments at Ludhiana in the Indian Punjab showed grain yield falling steadily as sowing was delayed past the optimum in November, with the loss traceable to higher temperatures and a shorter grain-filling period (Ortiz-Monasterio et al., 1994). Similar responses have been described wherever irrigated spring wheat is grown through a warming spring. The agronomic rule that follows, sow on time, is among the oldest in the extension literature. The harder question is why so much wheat is still sown late.

Late sowing is penalised twice. The crop meets warmer weather during grain filling, which is the effect usually emphasised, and it also passes through its vegetative and spike-growth phases under lengthening days and rising temperatures, so that it reaches anthesis with less biomass and fewer florets, the photothermal effect described by Fischer (1985). A late crop is therefore smaller before the heat arrives as well as shorter-lived after it. This is why extra seed and fertiliser, the customary compensation for late sowing, recover only part of the loss, and why the date itself is the variable to manage.

Cultivars contribute to escape in two ways: through phenology, which determines when the sensitive phases occur, and through traits that lessen the damage when heat arrives anyway. Table 1 sets the main options side by side with their mechanisms and limits.

Escape has a ceiling. Each of its instruments runs into a boundary: sowing cannot move earlier than the preceding crop, the seedbed temperature or the frost calendar allows, and cultivars cannot be shortened indefinitely without giving up yield. Warming narrows the room from both sides, since autumn heat penalises very early sowing through poor establishment and fast early development, while spring heat arrives sooner. The projections cited at the outset imply that the cool window itself is contracting, and a shrinking window can be occupied more skilfully but not enlarged.

Several questions remain open. The first concerns interaction: most sowing-date experiments use the cultivars of their day, and the best date for a new cultivar with different phenology genes is seldom tested before release. The second concerns the system. Advancing wheat by shortening or advancing rice involves a trade between the two crops that single-crop trials do not capture. The third is spatial. Recommendations are issued for large zones, yet the optimal flowering period differs between neighbouring districts. The fourth is the reliability of thresholds. Figures such as 34 °C come from particular data sets, and their transfer to other canopies, humidity regimes and irrigation practices needs checking.

Sowing on time with a cultivar whose development fits the season remains the most dependable defence against terminal heat in wheat, and both parts of that advice are better understood than they were a generation ago. The mechanisms of damage point to two targets, the weeks around anthesis and the period of grain filling, and the sowing date and flowering genes that determine exposure can now be chosen with some precision. The constraint in farmers' fields is more often the preceding crop than any lack of knowledge, which is why tillage and residue practices belong in a discussion of heat. Escape will not be sufficient as temperatures rise further, and it is best regarded as the base on which tolerance traits are added.

To be declared by the author(s) before publication.

To be declared by the author(s) before publication.