Serendipity has helped more than a few scientists with insights and discoveries. Researchers were chatting, and realized that data from work one of them was focused upon various crop trials spread across southern and eastern Africa, if correlated with local weather conditions (other than drought), could be used by the other to reveal what harm, if any, might arise from hotter than "normal" weather. The trials were focused on corn (aka maize).
What was culled from the data was that days about 30º C (86º F) were particularly damaging. In otherwise "normal" conditions, every degree upon this threshold reduced yields by 1%. Further, days over 32º C (89.6º F) did twice the harm of days over 31º C. Drought, no surprise, made things worse. Drought plus days over threshold reduced yields 1.7%.
Regional average temperatures mask the high's and low's. When factored in, the research found that a 1º C rise in average temperature would reduce yields across 2/3 of the maize (corn) growing regions of Africa. Add in drought, and the entire maize growing region is impacted.
Caveat. Using data from crop trials exaggerates the problem. Why? The plants are well fertilized, unlike most maize grown in Africa. Under-fertilized crops have lower average yields (which is why encouraging appropriate fertilization is so important), as one would expect, and so they tend to not be as badly damaged by temperature and drought as well-fertilized crops.
A previous study by the same researchers had projected yield losses of 20+% in Africa by the middle of the next century, given temperature projections (which as noted in prior posts, may be too low or too high since the models leave something to be desired). The data from these trials, however, are in accord with the prior predictions.
The researchers, having achieved significant insights from this type of data, now plan to address wheat next, which is widely believed to be more climate-sensitive than maize.
The report on maize can be found at http://www.nature.com/nclimate/journal/vaop/ncurrent/full/nclimate1043.html.