Keywords

Introduction

Agriculture is among the most vulnerable sectors affected by climate change. Additionally, it has strong interactions to other sectors, e.g. water resource management and mitigation of greenhouse gas emissions. Observed climate trends showed an increase of land temperature by ~1.5 °C compared to pre-industrial level, an increase of the frequency, intensity and duration of heat waves, while precipitation showed a spatially differentiated picture across Europe. Warming has prolonged the thermal growing season and frost-free period for crops across all of Europe since 1980s. Extended growing seasons facilitate the introduction of new crops or the expansion to higher latitudes and altitudes. However, the probability of multiple adverse climate events during a growing season including heavy rains and storms has increased and cropping systems have already been affected by climate change in terms of higher yield variability and crop loss risk but also in terms of adaptation, e.g. new varieties, diversification of crops or shifted cultivation dates. Projected climate trends are likely to enhance the risk posed by extreme weather events under climate change scenarios. To cope with increasing global food demand adaptation of agricultural production systems is required to minimize the risk of yield losses and exploit new chances from changing climatic conditions, while meeting other sustainability goals such as environmental impacts and efficient resource use. Diversification of crop rotations is considered as one option to increase the resilience of European crop production under climate change. Therefore, the design and management of crop rotations play an important role for the adaptation of cropping systems.

Main Effects of Crop Rotations and Their Management

The sequence of crops in crop rotations plays a significant role in resource use efficiency. Inefficient use of nitrogen in fertilizers and manures enhance N surplus and losses in many cropping systems across Europe, leading to nitrate pollution of groundwater and surface waters and to enhanced greenhouse gas emissions, especially N2O. Leaky periods in crop rotations can be prevented by the use of intermediate crops (e.g. catch crops), which retain N in the rooting zone and carry it over to the subsequent crop. Implementation of legumes as main or intermediate crops provides additional nitrogen to the following crop and needs to be considered in operational fertilization decisions. Additionally, winter cover crops prevent soils from erosion. However, the establishment of intermediate crops depends on water availability at sowing and their water consumption may reduce water availability for the following crop. Their effect on pest and diseases might be divers and it has to be considered if winter cover crops might be a host for diseases, suppress weeds or act as enemy crops, e.g. oil radish against nematodes. Crop sequence effects on yield can persist for 3–4 years in dry years or semi-arid environments as a result of water and nutrient legacies. Such legacy effects also includes inoculum survival and subsequent infestations of crops with fungal diseases. On long term, crop and soil management systems are known to change the storage of soil organic matter. Other mechanisms associated with crop sequences, e.g. effects on soil structure and soil physical processes and their interactions with roots are still not well understood.

Current State of Model-Based Climate Change Impact Assessment

Agro-ecosystem models considering the complex interactions in the atmosphere-plant–soil system are essential tools to assess the impact of projected climate change on various ecosystem services, including crop production, environmental effects such as water use and pollution, GHG emissions, resource use efficiency and long-term effects on soil properties, e.g. soil carbon stocks and to evaluate potential adaptation options under these multiple aspects. Their main advantage is the opportunity to conduct high number of simulations under various site conditions within a relatively short time.

During the last decade large international research consortia like the European JPI FACE knowledge hub MACSUR or the global AgMIP activity performed modelling studies to assess climate change impacts for main staple crops using large model ensembles. Their main message was, that no single model performed best across all site conditions and that the mean or median of a model ensemble mostly provided the best estimate (Asseng et al. 2013). Moreover, the ensemble of crop models showed a high variability indicating the uncertainty related to the model or its user, respectively. The majority of agricultural climate change impact assessment studies were focussed on few specific variables, mainly crop yields and considered only the main climatic factors such as temperature (Asseng et al. 2015) or drought effects either solely or in combination (Webber et al. 2018a). However, this approach has limitations, considering that crop growth and yield are affected by the interactive impacts of multiple climate change factors and biophysical processes. Webber et al. (2018b) suggested to replace air temperature by simulated canopy temperature to consider the interaction between heat and drought stress. However, the interaction between heat and drought seems to be genotype dependent and is still not sufficiently understood (Rötter et al. 2018).

Although the effect of elevated CO2 is considered in most of the models, their response on photosynthesis and water use varied substantially (Asseng et al. 2013). Kersebaum and Nendel (2014) analysed the effect of using three different CO2 response functions in combination with a dynamic CO2 response of stomatal resistance within the model HERMES across 21 regions in Germany on wheat yield, groundwater recharge and nitrogen leaching under current and projected climate. Model results for wheat yields differed by 5.5–11.6% among the three methods. Moreover, results showed a strong dependency on site conditions (soil and groundwater level) regarding the vulnerability against climate change. Diverse results regarding the beneficial effect of transpiration reduction through elevated CO2 emphasized, that the statement of higher CO2 stimulation, when crops are under water stress, cannot be applied universally. While algorithms seem to be applicable for different C3 crops, models showed a weak performance regarding the CO2 effect on C4 crops under reduced water availability in a FACE experiment (Durand et al. 2018).

Response to extreme weather events other than heat and drought, e.g. heavy rain, storm, hail, frost or water logging are less considered in most crop models (Rötter et al. 2018). Although yield reductions due to lodging were reported to be 31–80% in wheat, 4–65% in barley, 37–40% in oats, 5–20% in maize and 5–84% in rice, lodging is usually not considered in commonly used crop models. Moreover, efforts on the effects of extremes were focussed mainly on the three main staple crops wheat, maize and rice, while studies on other crops are rare (Rötter et al. 2018).

Although crop rotation design and management has been identified as an important measure to adapt to and mitigate climate change, most studies on climate change impact or adaptation so far use single-year simulations and/or single crop assessments (Webber et al. 2018a). However, if simulations neglect to include year-to-year changes in initial soil conditions of water and nutrient availability related to agronomic management, adaptation and mitigation strategies cannot be properly evaluated. Therefore, the integrated assessment of impacts, adaptation and mitigation options under current and future climatic conditions requires continuous long-term simulations of crop sequences (e.g. Kollas et al. 2015) to take into account carry-over effects as in real conditions. Although Kollas et al. (2015) used more than 300 years of experimental data for their model inter-comparison on crop rotations, the performance of models regarding crop yields improved only slightly when continuous simulation was used instead of annual resetting to standardized initial conditions. Lack of pronounced carry-over effects was because the effect of nutrient transfer on yields of the following crop was often masked by a high fertilization level and most of the sites were located in humid environments, where soil water mostly reached field capacity during winter. However, assessing the effect on other target variables, e.g. nitrogen balance compounds is only possible when using a continuous simulation since most emissions occur during the fallow periods (Yin et al. 2020).

Kollas et al. (2015) stated that models showed a weak performance mainly on crops where only few data were available for a proper calibration and modellers were less experienced. While main crops are usually well parameterized, the data base for not widely used crops or non-commercial crops such as potatoes, sugar beets or catch crops was often not sufficient for a solid parameterization. This underlines the request of Rötter et al. (2018) to extend research to crops other than the main staple crops.

Although some studies have conducted simulations for crop rotations mainly investigating effects of catch crops on nitrate leaching or N2O emissions (Yin et al. 2020; Gillette et al. 2018) under current conditions, only few studies looked at crop rotations under climate change scenarios (Hlavinka et al. 2014).

Conclusions and Recommendations for Model Improvement

Modelling of crop rotations require models that cope with a large variety of crops and management operations. As stated above, capability of models to simulate complex crop rotations is limited by the availability of suitable field datasets to parameterize and calibrate various crops such as oilseeds, pulses or beet crops, and those, which are of less economic value, but may contribute to environmental benefits such as catch crops. While data for model validation might be available from standard field trials, data requirements for model calibration are demanding since they require a high data density and proper description of the boundary and site conditions (Kersebaum et al. 2015). Therefore, they are rarely available for crops beside the main staple crops. Also suitable data on crop failure following extreme events are rare since experiments are usually cancelled after such events. Crop responses to several abiotic stress factors are still not fully understood and multifactorial manipulative experiments are thus crucial for a proper model-based assessment of plants growth and development under current and expected future environmental conditions (Rötter et al. 2018). This requires to evaluate model performance across different output variables beyond crop yield, e.g. soil water and nitrogen status and eventually severity of disease or pest damages (Kersebaum et al. 2015).

While some models are capable to simulate crop rotation effects in terms of carry-over effects of water and nitrogen, other effects like the exploration of rooting depth by previous crops for the following crop are rarely considered (e.g. Seidel et al. 2019). Many models are also lacking on suitable approaches to consider mixed cropping systems or under-sown catch crops and the competition among crops, but also with weeds.

Tillage and mulching effects are mainly considered using empirical relations, if considered at all, and only a few models are using process-based approaches (Yang et al. 2019). While no-till or minimum tillage is still propagated as a measure to sequester soil organic carbon, worldwide meta-analyses of tillage experiments cannot universally confirm this statement. Moreover, the effect of tillage seems to depend on site conditions, e.g. combined soil–climate impact. Therefore, the implementation of process-based approaches is required to reflect the site-specific short- (e.g. soil water balance) and long-term (soil organic carbon) responses on different tillage and residue management practices. This becomes even more important before the background of glyphosate use discussion, which is the usual alternative for tillage.

Crop losses by pest and diseases are rarely considered in crop models. Within MACSUR and AgMIP first attempts have been made to implement generic damage mechanisms to assess crop loss by pest and diseases, which builds on early modelling efforts during the 1980s (Bregaglio et al. 2021). However, following the roadmap of Donatelli et al. (2017) models are under development, which consider the interaction between crops and pests and diseases through the link between crop models and pest and disease models to cope with future changes of biotic pressures under climate change. This may also include the consideration of crop rotation effects on initial infection probability.