全球变化知识资源中心

Europe is the largest producer of wheat, the second most widely grown cereal crop after rice. The increased occurrence and magnitude of adverse and extreme agroclimatic events are considered a major threat for wheat production. We present an analysis that accounts for a range of adverse weather events that might significantly affect wheat yield in Europe. For this purpose we analysed changes in the frequency of the occurrence of 11 adverse weather events. Using climate scenarios based on the most recent ensemble of climate models and greenhouse gases emission estimates, we assessed the probability of single and multiple adverse events occurring within one season. We showed that the occurrence of adverse conditions for 14 sites representing the main European wheat-growing areas might substantially increase by 2060 compared to the present (1981–2010). This is likely to result in more frequent crop failure across Europe. This study provides essential information for developing adaptation strategies.

Recent global warming has markedly shifted the distribution of temperature variability and extremes^{1, 2} and precipitation patterns³, although uncertainty remains regarding the relationship between global warming and climatic variability⁴. These shifts have consequences for the production environments of most crops, including wheat, which is globally the second most widely grown cereal crop after rice⁵. A recent study¹ showed that, by 2030, we should expect a twofold increase in the global wheat-growing area threatened by extremely high temperatures during critical developmental stages in a typical year, and a more than threefold increase of the area at risk by 2050. Other studies project^{6, 7} a significantly higher frequency of extremely unfavourable years under future climate conditions, possibly resulting in poor economic returns in many European regions. This projection is especially true for situations with global warming exceeding 2 °C compared to the pre-industrial era. Although the observed annual temperature (adjusted for short-term variability) so far closely follows the central projections of the Intergovernmental Panel on Climate Change (IPCC; refs 7, 8), it should be stressed that several climate projections for the mid-century point far exceed the 2 °C threshold⁹.

Wheat production in Europe (Fig. 1), representing 25% of the global wheat area and 29% of global wheat production¹⁰, is affected not only by the frequency of days with high temperatures but also by the occurrence of drought, of late spring frosts and of severe winter frosts associated with inadequate snow cover. In addition, overly wet and/or cool weather enhances disease occurrence, contributes to lodging and complicates crop management. The projected increase in extreme weather events (for example, periods of high temperature and drought) over at least some parts of Europe is projected to increase yield variability^{7, 11}. Concomitantly, there is evidence of a slowing rate of yield increase, due to multiple factors—mainly the closing of the gap between realized and potential yields^{12, 13} as well as policies such as stricter environmental regulation¹⁴. The consequences of shortfalls in European wheat production for global supply (and prices) have been manifested in recent years, including 2007 and 2012¹⁵. Realizing the critical importance of European growing areas, we aimed to analyse whether and how the various agroclimatic risks for wheat production are likely to develop under long-term climate projections for the period 2051–2070 (subsequently denoted as 2060’).

a, Wheat-growing area in Europe based on ref. 30 and the locations of the 14 sites where the frequency of agroclimatic extremes was analysed. Colour coding is used to divide the stations into three groups: north (blue), central (black) and south (red). A more detailed description of the sites is provided in Supplementary Table 1. b, Coverage of the main environmental zones in Europe²⁵ by the 14 selected sites.

Simulated dates of sowing, flowering (anthesis) and maturity for the 14 locations were mostly in agreement with the observations from local authorities; however, deviations were found for some locations (especially the somewhat late estimated dates for Madrid (MD)) as a consequence of using the same parameterization for all of Europe. Figure 2a shows sowing dates being moved forward on average by 15 ± 7 days in 2060 compared to the present. Simulations showed that the anthesis and maturity dates were two weeks earlier across all sites (Fig. 2b, c). This advancement was linked to enhanced crop development rates with higher temperatures (Fig. 2d). Figure 2e shows that the site potential productivity indicator (effective global radiation considering suitable temperatures and soil water content, as defined in the Methods and the Supplementary Information) is expected to increase slightly at northern sites from sowing to anthesis, whereas southern sites mostly show declines. This decrease was caused by the vegetative period shortening from sowing to anthesis, the shift of this period towards shorter day lengths (that is, more towards the winter months) and the increased drought incidence at some sites. The results for anthesis to maturity (Fig. 2f) mostly showed a decrease in the effective global radiation. At 13 out of 14 sites, the results of more than half of the CMIP5-based climate model runs showed decreasing effective global radiation, thus reducing the potential for plant biomass accumulation and for crop yields.

a, Sowing date. b, Anthesis date. c, Maturity date. d, Temperature rate during sowing to maturity. e, Effective global radiation from sowing to anthesis. f, Effective global radiation from anthesis to maturity. Black rectangles indicate the 1981–2010 baseline and box plots indicate the 2060 (RCP8.5) climate scenarios. The locations are ordered from north to south along the x axis. DOY represents day of year.

The risk of a severe frost event in the absence of snow cover increased at the two most northern sites but was lower or unchanged at all of the remaining sites (Fig. 3a). In the case of late spring frost risk, we noted a decrease at one site, whereas increases were likely at six other sites (Fig. 3b). An excessive wet period with the possibility of water logging between sowing and anthesis was becoming increasingly likely at three sites in the UK, the Netherlands and Denmark (Fig. 3c), with little or no change at the other sites. The frequency of heavy precipitation events that are considered precursors of severe lodging was more likely to decrease than increase at seven sites, mostly in central and southern regions. There was only one site where the risk increased according to most scenarios (Fig. 3d). Fewer than one third of the sites showed an increased chance of unusually dry conditions during the entire growing season (Fig. 3e), with three southern sites being most affected. The drought risk in the period from sowing to anthesis was more likely to increase than decrease at four sites (three of them in the south), whereas the remaining sites showed no or very small changes (Fig. 3f). The likely increase of a severe drought event after anthesis affected only the southern locations along the western Mediterranean (Fig. 3g). The heat stress risk at anthesis, which would affect floret fertility, is likely to increase at ten sites (Fig. 3h), with northern sites being those least affected. Heat stress during grain filling is more likely to increase at six sites (most markedly in the southern sites), with another three sites showing a slightly increased risk according to some scenarios from the CMIP5 ensemble (Fig. 3i).

a, Severe winter frost without snow cover. b, Late frost. c, Excessive soil moisture with water logging from sowing to anthesis. d, High precipitation event with the possibility of widespread lodging. e, Severely dry growing season (sowing–maturity). f, Severe drought event between sowing and anthesis. g, Severe drought event between anthesis and maturity. h, Heat stress at anthesis. i, Heat stress during grain filling. j, Adverse conditions during sowing. k, Adverse conditions during harvest. l, At least one extreme event of the type a–k during the period from sowing to maturity. Black rectangles indicate the 1981–2010 baseline and box plots indicate the 2060 (RCP8.5) climate scenarios. The calculations consider a medium-ripening cultivar. The locations are ordered from north to south along the x axis. The red boxes mark the sites where the results for at least 14 out of the 16 CMIP5 models showed an increased probability of adverse events compared to the baseline. The orange boxes mark the sites where more than half of the CMIP5 models showed an increased probability of the adverse event. The green boxes mark the sites where results for all of the CMIP5 models showed a decreased probability of extreme events. The grey boxes mark all of the other cases.

Although the reported increase in the individual frequency of adverse events is worrisome, the most unsettling possibility is illustrated by the combined probability of having at least one of 11 indicators crossing the defined threshold during one season (Fig. 3l). Under the projected climate for 2060, the agroclimatic extremes probability was likely to increase at all of the sites for all of the wheat cultivar types, with the exception of CF for early cultivars (Fig. 4a–c). Whereas for all of the sites on average, the risk of at least one adverse event was likely to increase by 30%, at some sites (UP, RR, MA, VI and SL) the risk was likely to double. Using a medium-duration cultivar (Fig. 4b) as a reference, the mean probability of a single adverse event occurring per season was found to be 11.2% under the baseline and 20.7% in 2060 (according to RCP8.5). Furthermore, the probability of two adverse events occurring for a medium-type cultivar in any given season (Fig. 4e) was shown to be likely to increase at 11 out of 14 sites. Considering the lowest impact that was projected from the global climate model (GCM) ensemble for each site, only a slight reduction in the probability (9.7% compared to 11.2%) of adverse events may be expected. However, the realization of the most severe case from the GCM ensemble would result in a more than threefold increase in the adverse event probability. The mean likelihood of two adverse events per season was 1.7% under the baseline but more than twice that (4.0%) under the projected climate change, with the lowest GCM mean value being 0.9% (Fig. 4e). The likelihood of two events per season considering the highest GCM value at each site increased more than sixfold (to 10.4%) compared to the baseline. Furthermore, the probability of three adverse events occurring within a location during one season was evaluated, recognizing that three events per season would represent an exceptional season. Such extreme cases were found under the baseline conditions only at the DC, MD and SL sites, and even then very rarely (less than once per 125 years), whereas under the projected future climatic conditions only three sites (TR, RR and VI) show no risk of such a season. The mean return period for three adverse events occurring in 2060 was projected to be between 20 and 30 years at Spanish sites and between 75 and 250 years at the remaining sites.

a–c, Probability of occurrence of at least one adverse event. d–f, Probability of occurrence of two adverse events. a,d, Represent an early-maturing cultivar. b,e, Represent a medium-maturing cultivar. c,f, Represent a late-maturing cultivar. The colour coding follows that of Fig. 3.

Exposure to several of the adverse events depends on the timing of anthesis and maturity, which was influenced by the cultivar type used (Fig. 4). Whereas the cultivar type did not influence the number of sites facing a higher probability of at least one event per season (Fig. 4a–c), it had a marked effect on the chances of at least two adverse events (Fig. 4d–f). The probability of at least one adverse event per season increased at most of the sites, especially for the late cultivar (Fig. 4c). Because we were concerned about potentially overestimating the frequency of adverse and extreme events occurring owing to the longer-than-observed phenological development in the more southern sites, we altered the photoperiod sensitivity. This alteration led to a decrease in the adverse event exposure, especially at southern and some central European locations (Supplementary Fig. 2), but their increased frequency was still considerable. It should also be noted that one of the main adaptations to climatic warming in many areas of Europe would be switching to cultivars with a longer growth duration¹⁷, which may increase exposure to adverse weather events. Our results demonstrate that the risk of a season being affected negatively by adverse conditions depends on the changes in the occurrence probabilities of such events, on the potential exposure length and on the sensitive period timing. For example, in cases of late cultivars, for which anthesis and maturity occur later in the year, the chance of heat and drought stresses increases.

Soil conditions play an important role in the course and severity of some adverse events, especially those affected by lack or excess of soil water. Supplementary Fig. 3 shows that light soils (with maximum water content in the rooting zone available to plants at 150 mm) would face higher mean frequency of the adverse event occurrence under baseline and future climates than medium soil (available water content at 270 mm). The effect was more pronounced on the southern sites and for the late-maturing cultivar. It is interesting to note that the relative increase of the adverse event probability under future climate compared to baseline was smaller on light soils, mainly owing to the fact that the probability of such events under the baseline climate was already high.

An adaptation option that might be pursued based on the results we presented (Fig. 4) is a focus on early-ripening cultivars seeking ‘stress avoidance’. Although it seems that this strategy would allow for a significant decrease in the adverse condition risk at most of the sites, it comes at a price (see, also, ref. 18). Supplementary Fig. 1 clearly indicates that shortening the growing season reduced the effective global radiation, thus probably decreasing yield potential. The pros and cons of the benefits and risks of such a strategy are further highlighted in Table 1 (the locations are ordered from north to south). This illustrates the trade-offs of using early or late cultivars instead of a medium cultivar. Switching to an earlier cultivar reduced the exposure to extreme events and shifted the time of anthesis to a period with less heat and drought stresses (in some cases alleviating other stresses as well). At most sites, and especially in the southernmost locations, the effect of such a cultivar shift was substantial, and using earlier cultivars led to greatly reduced risks. The results imply that there is a greater scope for introducing earlier cultivars than was tested here. However, Supplementary Fig. 1 also documents that under most scenarios and at all sites, the effective global radiation that we used as the indicator of potential productivity (more details in Methods and Supplementary Information) decreased, and that this decrease could be worsened by switching to earlier cultivars, especially at southern sites.

Full table

Previous studies assessing agricultural impacts have demonstrated that the effects depend on the crop, cropping season and region within Europe¹⁹, but few studies have considered the cropping system responses to changes in the frequency and severity of climatic extremes (see, for example, ref. 20 for the Iberian Peninsula). However, it is well known that the impacts of such extreme events can be substantial⁵. Previous studies have emphasized the possibility of considerable northward expansion of the thermal suitability of crop production in Europe (indirectly suggesting major shifts in the location of agriculture production) without fully considering changes in the risk of adverse events^{19, 21}. Our results show that, despite large uncertainty in climate projections within the CMIP5 ensemble, the overall adverse event frequency is much more likely to increase than decrease, which is the case for all of the sites in the analysed European domain.

From the standpoint of production stability, it is encouraging that some central (WS and VI) and north-western European (WA, MA and TR) growing areas are likely to face comparatively small increases in adverse event occurrence. Nevertheless, the fact that the majority of the sites show a greatly increased probability of single adverse events suggests, in turn, that risk of crop failure would increase across large portions of the European wheat-cropping area. Such a development would have profound repercussions given the importance of European wheat production in the global food trade. Moreover, more frequent adverse conditions for wheat at any particular location would probably be accompanied by yield reductions in other crops (both cereals and non-cereals), as their growing seasons and sensitive periods at least partly overlap. We stressed that the results are valid for wheat that is grown on free-draining soils with the ability to hold a significant amount of water available to plants. The severity and frequency of some of the adverse events (for example, drought stress or water logging) could be different on sandy or heavy clay soils.

Our results highlight the potential of adverse impacts of a changing climate on wheat and show that the associated potential adaptations to these impacts should consider adverse and extreme weather event effects in a more comprehensive way than is usually done in impact assessments based on crop models, which very often do not include such event effects²². Moreover, focusing on single adverse events may lead to an incomplete risk perception. Impact severity will depend on the cultivar characteristics and, obviously, on the spatial and temporal climate change patterns. This dependence calls for a regionalization of adapt