Regions that produce a large supply of agriculture commodities can be susceptible to crop failure, thus causing concern for global food security. The contiguous United States, as one of the major agricultural producers in the world, is influenced by several large-scale climate oscillations that contribute to climate variability: Atlantic Multidecadal Oscillation (AMO), North Atlantic Oscillation (NAO), El-Nino Southern Oscillation (ENSO), Pacific Decadal Oscillation (PDO) and Pacific-North American (PNA). Since local weather conditions are associated with these climate oscillations through teleconnections, they are potentially causing changes of crop failure risks. The objective of this study is to assess climate-induced changes of annual crop failure risks for maize and winter wheat from 1960 to 2016, by analyzing the associations of large-scale climate oscillations with the frequency of crop failure in the rainfed regions of the United States using a Bayesian approach. The analysis revealed that crop failure frequencies showed contrast spatial patterns and different extent under different climate oscillation phases. Among individual oscillations, the positive PNA and negative AMO resulted in the most substantial increase in maize and winter wheat crop failures over a high percentage of climate divisions, respectively. Among oscillation combinations, the positive AMO and negative PDO and the positive AMO and positive PDO resulted in the highest percentage of climate divisions experiencing significant increase of maize and winter wheat crop failures, respectively. Random forest models with climate oscillations accurately predicted probabilities of crop failure, with the inclusion of local surface climate variables decreased or increased the predictive accuracy, depending on regions. These results revealed the plausible drivers of long-term changes of U.S. crop failure risks and underscore the importance for improving climate oscillation forecasting for early warning of food insecurity.

中文翻译：

---

美国农作物歉收风险的变化与大西洋和太平洋大尺度气候振荡相关

生产大量农产品的地区容易遭受作物歉收，从而引起对全球粮食安全的担忧。作为世界主要农业生产国之一，毗邻的美国受到若干导致气候变率的大规模气候振荡的影响：大西洋多年振荡 (AMO)、北大西洋涛动 (NAO)、厄尔尼诺南方涛动(ENSO)、太平洋年代际振荡 (PDO) 和太平洋-北美 (PNA)。由于当地天气条件通过遥相关与这些气候振荡相关联，它们可能会导致作物歉收风险的变化。本研究的目的是评估 1960 年至 2016 年气候引起的玉米和冬小麦年度作物歉收风险的变化，通过使用贝叶斯方法分析美国雨育地区大规模气候波动与作物歉收频率的关联。分析表明，作物歉收频率在不同气候振荡阶段呈现出对比鲜明的空间格局和不同程度。在个体振荡中，正的 PNA 和负的 AMO 分别导致玉米和冬小麦作物歉收在高比例的气候分区中出现最显着的增加。在振荡组合中，正 AMO 和负 PDO 以及正 AMO 和正 PDO 分别导致玉米和冬小麦作物歉收显着增加的气候划分比例最高。具有气候振荡的随机森林模型准确预测作物歉收的概率，包括局部地表气候变量会降低或提高预测准确性，具体取决于区域。这些结果揭示了美国作物歉收风险长期变化的可能驱动因素，并强调了改进气候振荡预测对粮食不安全预警的重要性。