A rather specific question today – what will happen to maize yields in France in 2016-2035?
There are many factors affecting yield (the production per unit area) of a crop such as maize. The main cause of changes in yield is technology development which has increased yield by at least a factor of 3 over the past 50 years – we are just better at growing crops, for a variety of reasons! The weather may also have an influence – crops are known to be sensitive to very hot temperatures, and dry conditions, although precipitation may be less important now due to the increase in irrigation. However, recent trends in yield have been fairly flat, leading to speculation about whether this is due to changes in weather or technology saturation.
Figure 1 below shows observed changes in summer precipitation and temperature (represented by the number of days with maximum temperatures over 32°C), and maize yield for France since the 1960s. There are a number of years when the yield is relatively low, and these years often correspond to particularly hot years or dry years. The heatwave of 2003 is the most prominent example. An empirical model is constructed (pink shading and lines) which represents the trend due to technology (and this will also include a CO2 fertilisation effect), and the effects of weather on yield. The yield for 2011 has not yet been published, and a forecast, given the known weather, is given by the black error bar.
UPDATE (05/02/13): 2011 forecast was 0.9-1.0 kg/m² (5-95% ranges) assuming no technology development since the previous year. FAOSTAT data now available is 1.02 kg/m² – around 5% above the previous record yield. Possibilities for why the observed yield was outside our forecast range: (i) strong technology development has occurred (e.g. a new cultivar of maize), (ii) it happened by chance (actual yields should fall outside forecast 10% of the time), or (iii) our empirical yield model is missing a process.
The success of the yield model suggests that the reduction in future yields due to climate can be predicted if the future number of hot days can be projected. Equivalently, this would be the amount of technology development required to negate the effects of climate. This is no easy task, particularly because the climate simulations used here (the QUMP ensemble) are too warm, which means they have too many days over 32°C. Two relatively sophisticated calibration schemes (termed ‘bias correction’ and ‘change factor’) are utilised to remove the biases in both the mean climate and the daily temperature variability and hence correct the projections.
Figure 2 shows how the calibration affects the projected number of hot days, with the raw simulations (left), observations (second column) and calibrations (right columns). The 1991-2010 period is predicted retrospectively using observations from 1966-1985 to perform the calibration as a test of the methodologies – and they work well for predicting the 1991-2010 period. The equivalent projections for 2016-2035 are also shown.
These different projections of hot days can be used with the empirical model of yield, and assuming that precipitation does not change, to produce the yield forecasts shown in the bottom right of Figure 1. There are also choices to be made about future relationships between temperature and precipitation, but the conclusion is that technology developments need to increase yields by 12% to be confident that current yield levels are maintained. The current rate of technology development is not sufficient to meet this target.
This work has been
submitted accepted to Global Change Biology.