Meta-analysis of climate impacts and uncertainty on crop yields in Europe

The projected impacts of climate change on yield by crop type, for Europe as a whole and aggregated into sub-regions (northern, central and southern) are shown in figure 2. The data shown correspond to the forecast yield variations for each crop, for all GCM models, all time slices and all crop models. The whiskers represent the minimum and maximum reported yield variation, where published. Table 1 summarises the primary data and shows that for Europe as a whole, most studies project a positive impact on crop yield. The reported increases being largely due to rising atmospheric CO₂ concentrations, enhancing both crop productivity and resource use efficiency. Overall, a projected change in average yield of +8% was identified, but with large differences between both individual crops (for example, wheat +14%; maize −6%) and regions (northern Europe +14%; southern Europe +5%). Crops with the most observations and a statistically significant yield variation were wheat, maize, potato and sugar beet. Conversely, data for barley, rice and rye were much less extensive, so projected variations in average yield were not statistically significant.

Zoom In Zoom Out Reset image size

When the data were disaggregated by time horizon, only changes in wheat and sugar beet yield were statistically significant. However, it is important to note that there were no data available on forecast yields for all crops in all regions in all time horizons, so lack of a significant response may in part be due to the absence, or limited number of studies for certain crops and/or regions.

For northern Europe, most evidence related to wheat (n = 89) with observations available from studies conducted in almost all countries in the region. In contrast, there was very limited published data on rye (n = 2) and none for rice. The evidence projected higher yield increases both on average and across all crop types and countries, although the variability in reported impacts was also surprisingly high (figure 2). Significant average yield changes were identified for maize and potato (+14% and +17%, respectively), with particularly strong increases in the 2050s (+12% and +18%) and 2080s (+19% and +14%). The average yield change for barley (−1%) was not statistically significant.

For central Europe, the average yield change was +6%, which was relatively constant through the different time horizons and for all crops. Wheat accounted for the largest number of observations (n = 140) with a moderate projected yield increase towards the end of the century (+7% for the 2020s rising to +11% by the 2080s). Sugar beet showed a steady increase in projected average yield through time. However, it is important to recognise the decrease in projected change in average yield for maize from −9% for the 2020s down to −15% by the 2080s. The projected changes in yield in central Europe were broadly consistent with those for northern Europe, showing positive impacts for most crops apart from maize.

Climate impacts for southern Europe also showed an increase in average yield although the data for the 2080s was not statistically significant. Wheat showed a consistent, significant increase in projected average yield (+18%), in contrast to maize which showed a sharp decrease, particularly by the 2080s (−28%). Reported data for potato showed a significant increase in projected average yield for the 2020s and the 2050s (+11% and +10%, respectively). However, by the 2080s the trend reverses, with a significant projected change in average yield of −10%. These data were drawn from a single study covering the European region using a crop modelling methodology and climate data, generating single climate impacts for potato for each country separately. The frequency distribution of the projected change in average yield for all crops, disaggregated into European regions, is shown in figure 3. This shows that most of the observations lie between the −5% and +15% bins.

Zoom In Zoom Out Reset image size

For many of the crops studied, notably cereals and sugar beet, Europe is a significant contributor to global production. For example, in 2014, the harvested production of cereals (including rice) in the EU-28 was estimated to be 334 million tonnes, representing 13% of global cereal production (FAOSTAT 1998). Whilst globally most sugar is derived from sugarcane, beet production in the EU-28 still provides 20% of global sugar demand and constitutes half of global production (EUROSTAT 2014). Given a rising global population, reducing the amount of food produced in Europe is not a viable option. Positive or negative changes in yield in Europe will not only influence future levels of self-sufficiency but could also have serious indirect impacts on international agricultural trade, food supply-chains and commodity markets. To inform robust decision-making regarding policies for climate adaptation, it is therefore important to set the projected climate impacts on food crops in Europe as reported in this study against equivalent projected yield changes internationally. For wheat, the mean projected climate impact in Europe (+14%) contrast sharply with previous research by Knox et al (2012) who reported significant a reduction in Africa (−17.2%). For maize, a projected mean yield reduction of −6% in Europe was similar to Africa (−5.4%) but much less than S Asia (−15.9%). For other important crops in Europe including sugar beet (+15%) and potatoes (+8%) there are no equivalent reported data for Africa or S Asia. For rice, evidence from studies in Europe, Africa and S Asia were all insufficient to deduce any significant yield variation, highlighting the pressing need for further research to expand the evidence base. For sugarcane, a lack of evidence on climate yield impacts in Africa should also be addressed given the recent focus on sugarcane expansion as a means to power agricultural transformation and economic development in Africa (Hess et al 2016). The findings from this research corroborate other studies that show how crop productivity impacts in higher-latitude temperate regions, such as northern Europe, are generally expected to be less severe than in lower-latitude more tropical regions (Challinor et al 2014). These wide geo-spatial variations in climate impact confirm an enhanced role for trade with increased flows in agricultural commodities from the mid to high latitude regions to low latitude regions needed to offset the more extreme climate impacts, where both production and export potential could be severely hampered (Elbehri et al 2015).

The SR also considered the projected change (%) in average yield over time, based on reported time slices (ranging from the 2020s to the 2080s) and for different GCMs (both single and ensemble). The data for all crops and all GCM models are shown in the 'box and whisker' plots in figure 4, disaggregated by EU region. The 'box' defines the inter-quartile range. The line spanning the box represents the median and the 'whiskers' represent the lower (10%) and upper (90%) deciles. Outliers are shown as individual points. The data in figure 4 highlight the significant differences between time slices and regions. Overall, the variation in projected change in average yield is positive across all time slices and regions (apart from southern Europe in the 2080s). There are smaller projected changes in yield for the 2020s compared to the 2050s and 2080s. The inter-decile range spans the zero variation line for all time slices and regions. Generally the variation in projected change in average yield increases with time, and most notably for southern Europe.

Zoom In Zoom Out Reset image size

Forecasting future crop productivity under a changing climate is subject to several sources of uncertainty, including for example, the timing of impacts (Challinor et al 2014). The choice of a particular GCM climate model and whether it is used in isolation or as part of an ensemble can also have a significant impact (Challinor et al 2009). Figure 5 shows the projected change in average yield for studies that used either an individual GCM model (e.g. CGCM, ECHAM, HadCM3) or an ensemble approach. Surprisingly, the projected change in average yield was more variable for ensemble-based studies and the median was greater. This is in contrast to findings from a SR of equivalent climate impact on crop yield data for Africa and S Asia (Knox et al 2012).

Zoom In Zoom Out Reset image size

The complexity of modelling approaches used for climate impact assessment can also strongly influence the projected crop yield variation. For example, figure 6 shows the reported yield variation based on either a simple modelling approach, typically based on locally developed statistical models, or a more complex biophysical modelling approach using locally calibrated and validated crop models such as those embedded within DSSAT (Jones et al 2003) or standalone crop models such as WOFOST (van Diepen et al 1989) or CropSyst (Stockle et al 2003). The variation around the mean is much greater for those studies that adopted a complex modelling based approach, with wide variation in both positive and negative yield change. Various authors have previously identified possible reasons (termed 'effect modifiers' in SRs) for such variation. Firstly, there are more published studies based on the 'complex' modelling approach; in this study, the observations for simple approaches (n = 70) was much lower compared to those using 'complex' approaches (n = 654). Other factors include differences in the assumptions made in the crop modelling (e.g. soil and crop cultivars, crop management practices, pest and disease risk) and the lack of locally specific model parametrization (Müller et al 2011). Finally, different protocols used by researchers could also introduce biases that limit any further cross-study or cross regional synthesis (White et al 2011).

Zoom In Zoom Out Reset image size

SRs were originally designed for application in clinical research to synthesise the outcomes from replicated, controlled medical trials where the effects of particular interventions could be identified and quantified statistically. However, in field crops research it is generally not possible to evaluate future climate impacts on crop production through experimentation, so yield impact is usually assessed through a combination of estimation methods involving biophysical or statistical crop models, which themselves introduce an important component of uncertainty into the modelled results due to their simplification of real-world complexity. The quality of primary data available for conducting a SR is also critical. This review was limited to evaluating modelled results from a wide spectrum of climate change impact studies, all of which inevitably contained different combinations of effect modifiers. These included, for example, different emissions scenarios applied to a diverse range of GCMs under various agro-ecological conditions, and varying assumptions regarding crop varieties, agricultural systems, crop husbandry practices, and levels of mechanisation. As highlighted by Reidsma et al (2010) such differences in farm characteristics (e.g. production intensity, farm size) are particularly important when considering climate impacts on production and farm adaptation options and responses. In addition, the different spatial scales reported in the literature may explain some of the observed differences together with the approaches used by individual studies for downscaling coarse resolution climate data. The studies included in the SR assessed the impacts of climate change on the crops in the locations where they were currently grown, so they ignored any possible impact of changes in cropping pattern to exploit more favourable growing conditions. Finally, the SR was limited to assessing climate impacts on yield, ignoring any planned or autonomous adaptation or adaptive capacity response; future SR studies could focus on the extent to which adaptation is integrated into crop modelling studies (Reidsma et al 2010).

This study provides the first SR of climate impacts on crop productivity for a major food production region. The outputs provide valuable new information to inform policies regarding climate change impacts in support of developing adaptation strategies to increase the future resilience of European food crop systems. In contrast to crops such as wheat (n = 293) and maize (n = 149) where there is extensive and robust evidence, the SR has identified a major knowledge gap regarding the climate impacts on important crops such as barley, rice and rye (n = 69, n = 2, n = 4, respectively). For countries where these crops are important, this will inevitably limit decision makers' and farmers' ability to benefit from research in developing robust strategies to increase their resilience to a changing climate. Surprisingly, the SR also confirmed there is extensive evidence on climate impacts on crop production for northern and central Europe, but much fewer studies for southern Europe. For the science community, our outputs should help focus future efforts regarding the choice of crops and regions within Europe where research effort is needed and where programmes such as the FACCE JPI (Modelling European Agriculture with Climate Change for Food Security) might wish to direct increased attention. This will assist policy makers and practitioners in making more informed decisions on how and where resources should be allocated to better adapt agricultural production to climate change, and importantly the scale at which interventions should be made. To be effective, adaptation responses at regional level need to be closely aligned and sensitive to the composition and mix of farm types at local level. For example, farmers surveyed by Olesen et al (2011) reported a high proportion of negative expectation regarding the impacts of climate change on crop production across Europe, even in the cooler temperate latitudes of northern Europe. This potentially highlights a lack of effective knowledge transfer and dissemination of research to the farming community and practitioners from modelled studies that predominantly demonstrate a positive benefit from climate change on crop production in that region whilst not to excluding the secondary impacts of climate change linked to crop production including potentially increased risks from soil erosion, changes in nutrient cycling and need for crop protection (Olesen et al 2011).

Robust and reliable evidence are of course critical in support of formulating policies to address climate impacts on agriculture, food security and trade; implemented effectively it can usefully guide decisions on policy, highlight options for action and identify evidence gaps (Elbehri et al 2015). However, from a policy perspective, it is important to recognise that whilst quantifying changes in yield is essential for on-farm adaptation, it constitutes only part of a much broader mix of climate risks to agriculture. Integrated assessments attempt to overcome this by factoring in the links and feedbacks between agricultural production, food demand, markets and land use trends, to identify a range of policy alternatives. Various international collaborative initiatives are striving to achieve this through linking global circulation models with biophysical and agricultural economics modelling to inform economic and trade impacts at the regional and international levels (Rosenzweig et al 2013). Given the strong differences in climate yield impact between Europe and other regions, and the fact that climate change can transform trade by altering the comparative advantages between regions (Elbehri et al 2015), there will also be a need for more climate-compatible trade policies to resolve the trade versus environment trade-offs to ensure that future trade regulations are more tightly aligned with climate adaptation objectives. Recent research has also highlighted the policy risks associated with conducting single-sector (e.g. agriculture, forestry) climate impact assessments. In Europe Harrison et al (2016), showed how single sector studies can strongly misrepresent the spatial pattern, direction and magnitude of most impacts as they omit critical interdependencies between human and environmental systems. Finally, developing policies to achieve a more climate resilient agricultural sector in Europe will require much greater attention to the links between water resources and agricultural production. Since water mediates much of the climate impact on agriculture any projected increases in water scarcity will inevitably present major challenges for climate adaptation, food security and nutrition (Elbehri et al 2015). Identifying 'hotspots' where irrigated agriculture is concentrated and where future supply-demand imbalances might occur, could provide valuable insights to inform national policies targeting resources for adaptation, or in providing incentives for adaptation research to identify more resilient agricultural technologies/systems in hotspot regions. National climate change policies should also pay greater attention to identifying regions where future investment in agriculture and new food crops should be encouraged, particularly where more favourable growing conditions and/or new markets might emerge. Clearly, there are many opportunities for robust evidence derived from SRs to usefully contribute to ongoing policy dialogue and science debate.

The evidence from this SR confirms that climate change is likely to increase the yield of Europe's major agricultural cropping systems, with more favourable impacts in northern and central Europe. Despite the inherent limitations in applying an SR approach in this research domain, it highlights a potentially very important message; that climate impacts in Europe are not necessarily all negative, but that they could be beneficial for many crops and areas of production. This could also trigger changes in growing conditions and land suitability for other crops that are currently either marginal or not grown by farmers in certain regions, providing scope for crop diversification. Notwithstanding these potential opportunities, as stated by Wheeler (2015), the multitude of risks that climate change still poses to agricultural output and food systems in Europe and globally should not be ignored by those making medium to long term strategic planning decisions regarding food security.