Global food security is threatened by the interconnectedness of the world's food systems and the potential for simultaneous shocks to major crop production areas (breadbaskets). Previous risk assessments have extrapolated from historical climate-yield relationships, but climate change introduces unprecedented events that surpass historical patterns and potentially underestimate current risks. This study focuses on wheat production in the USA and China, regions where historical data show weak relationships between yields and temperature. However, emerging extreme temperatures now exceed critical physiological thresholds in wheat plants, necessitating a different approach to risk assessment. The UNSEEN approach, using large ensembles of archived seasonal forecasts, generates plausible, unprecedented events to evaluate the risk. The study departs from analyses solely relying on historical events to visualize the risk of unprecedented events that could cross critical thresholds in these vital wheat-producing regions. Most existing studies use historical climate-crop yield relationships to predict future impacts, but this approach fails to account for nonlinearities and unprecedented climate states. In the case of wheat in parts of the USA and China, there is a limited historical relationship between temperature and yield, leading to underestimation of risks from extreme temperatures exceeding previously observed thresholds. Physiological models show wheat's sensitivity to temperature at critical growth stages, with extreme heat causing leaf senescence, reduced expansion, and lower efficiency. UNSEEN, a method utilizing large ensembles of climate model forecasts, provides a broader range of plausible scenarios, beyond historical data, to estimate extreme value statistics. While previous UNSEEN studies have focused on existing events or future analogs, this study explores synthetic events without historical parallels. This allows for a clearer picture of potential climate risks that might not have a historical precedent but could significantly impact agricultural yields. The study uses UNSEEN to examine unprecedented heat storylines in the USA and China's wheat-producing regions, focusing on the frequency of temperatures above critical growing thresholds, changes in return periods of extreme temperatures, and probabilities of compound extremes.

Several studies have attempted to quantify the risk of multiple breadbasket failures due to climate shocks, primarily extrapolating from historical patterns. However, these approaches often underestimate the risk of unprecedented climate events. Research has shown that in some regions, more than 50% of historical yield variability is attributable to weather, but climate change alters these relationships. Studies utilizing historical climate-yield relationships to predict future impacts often neglect non-linearities and unprecedented climate states. For instance, in parts of the USA and China, historical data show weak relationships between wheat yields and temperature; however, current extreme temperatures exceed critical physiological thresholds, indicating a potential for substantial yield loss. Previous studies often rely on process-based crop models or statistical models, which may not accurately capture the impact of these unprecedented events. Physiological models and simulations indicate the negative impact of warming on wheat yield; however, these models often focus on annual extremes and likely ranges, ignoring low-likelihood, high-impact events. The UNSEEN approach, using large ensembles of climate models, offers a way to simulate these low-likelihood, high-impact events.

This study employs the UNprecedented Simulated Extreme ENsembles (UNSEEN) approach to analyze unprecedented heat in two major wheat-producing regions: the US Midwest and Northeastern China. The UNSEEN methodology involves using large ensembles of archived seasonal forecasts (specifically, ECMWF's SEAS5) to generate thousands of plausible weather events over the past 40 years. These simulated events are then compared with historically observed extreme temperatures and precipitation. The study focused on winter wheat, analyzing the March-May period critical for wheat growth. Three key temperature variables were considered: maximum daily temperature, the number of days exceeding a stress threshold (27.8°C), and the number of days exceeding an enzyme breakdown threshold (32.8°C). Total precipitation during March-May was also analyzed. The data used high-resolution datasets, DayMet for the US and ERA5 Land for China, which were downscaled to a 1° resolution for consistency. The SEAS5 ensemble, containing 25 ensemble members until 2016 and 51 from 2017 onwards, with 5 lead times, provided a large dataset of alternative realities. Rigorous evaluations were performed on the SEAS5 UNSEEN ensemble to ensure credibility. The stability across lead times was checked for model drift. Independence between ensemble members was assessed using pairwise rank correlations. Fidelity checks compared the UNSEEN ensemble statistics (mean, standard deviation, skewness, kurtosis) against historical observations. Bias corrections were applied where necessary to match the mean of the historical data. Extreme value statistics (Gumbel, Generalized Extreme Value, and non-stationary GEV distributions) were fitted to both historical and UNSEEN data to assess changes in the likelihood of extreme temperatures. Geopotential height and wind anomalies were analyzed to understand the circulation patterns associated with extreme events. Compound events, where extreme heat and drought occur simultaneously in both regions, were identified by examining the overlap of the top 250 ensemble members for each region that produced the greatest numbers of enzyme breakdown days.

The UNSEEN ensemble reveals a steady increase in maximum temperatures in both the US Midwest and Northeastern China. In the US Midwest, extreme temperatures that occurred once every 100 years in 1981 now have a return period of 1 in 6 years. In China, the return period is approximately 1 in 16 years. This translates into an increase in the probability of such extreme temperatures from 1% to 17% in the US Midwest and from 1% to 6% in China. The number of days exceeding critical heat thresholds has also increased. The UNSEEN ensemble shows events with significantly more days exceeding the enzyme breakdown threshold than observed historically. Rainfall patterns show less clear trends, though the UNSEEN dataset contains drier events than those historically observed. The combination of record-breaking heat and dryness significantly impact wheat yields; the 2014 drought in Kansas serves as a stark example. The study demonstrates a strong association between extreme heat and extreme dryness in both regions, with extreme heat events often coinciding with low rainfall due to large-scale atmospheric circulation patterns. Analysis of geopotential height and wind anomalies at 500 mb reveals that strong winds over land pull dry air towards the regions during extremely hot and dry events. In the USA, northerly and westerly wind anomalies bring dry air from the continental US, while in China, northerly and westerly winds transport dry air over land towards the study area. The opposite wind directions (from the south and east) are associated with extremely wet seasons. The UNSEEN ensemble also identifies compound events—simultaneous extreme heat in both the US and China. The likelihood of such a compound event is higher than expected by chance, indicating the influence of global climate change on extreme temperatures. These compound events are associated with a zonal wavenumber-3 disturbance in the higher latitude circulation, creating high-pressure systems over both study areas.

The findings demonstrate that relying solely on historical climate-yield relationships for risk assessment can significantly underestimate the current risk of extreme heat and drought in major wheat-producing regions. The UNSEEN approach offers a more comprehensive risk assessment by accounting for unprecedented climate events that have no historical precedent. Recent temperature extremes in the US Midwest, for example, are milder than the full range of plausible extremes simulated by the UNSEEN ensemble. This highlights the potential for surprise and the need to shift from a purely historical perspective to a more forward-looking approach that incorporates plausible yet unforeseen events. The study provides critical insights into the large-scale atmospheric circulation patterns responsible for the simultaneous occurrence of extreme heat and drought. The identified wind anomalies associated with these events enable better monitoring and forecasting of conditions that could negatively impact wheat crops. The UNSEEN approach is valuable in visualizing potential compound events, informing adaptation planning, and bridging the gap between climate science and stakeholder perception of risk. The large-scale circulation anomalies driving extreme events should be closely monitored. This improved understanding of these patterns supports better forecasting and facilitates the development of adaptation strategies.

This study highlights the significant potential for surprising heat and drought events in major wheat-producing regions of the USA and China, exceeding what historical data suggest. The UNSEEN approach offers a valuable tool for visualizing these unprecedented events, emphasizing the inadequacy of relying solely on past data for risk assessment in a changing climate. The strong association between extreme heat and drought, driven by large-scale circulation anomalies, underscores the urgency for adaptation planning. Future research could further refine the UNSEEN approach by incorporating more detailed crop models, exploring interactions with other factors influencing crop yields (e.g., pests, diseases, global trade), and conducting regional analyses to tailor adaptation strategies.

The study's findings are constrained by the ability of climate models to represent the full range of plausible outcomes. While fidelity tests were conducted, there remains a possibility that the models may not capture the full spectrum of risk. The analysis focuses on climate variables directly affecting wheat growth, without explicitly modeling yield responses, although indirect impacts are considered. The study considers only two major wheat-producing regions, limiting the generalizability of findings to other regions with different climatic conditions and agricultural practices.