Abstract

According to the IPCC, most regions worldwide will be gradually exposed to the amplification of the duration, frequency, and intensity of extreme climatic events, and the effects that extreme events can cause on human well-being and the economy. This study aims to develop linear regression models to estimate the soy and maize yields from extreme climatic and bioclimatic indices in three geographical subregions of Latin America (Mexico, Brazil, and Argentina) between 1979 and 2005. We used daily datasets from observations (CPC), reanalysis (ERA5), and regional climate model (RCM) simulations from the Coordinated Regional Climate Downscaling Experiment (CORDEX) to investigate the impact of extreme events of temperature and precipitation on maize and soy yields over the CORDEX Central America and South America domains. We first assessed the RCMs’ performance in reproducing extreme indices by comparing them against observations. The validation process evidenced the need for applying bias correction techniques to simulate daily precipitation and temperature for a better performance of the indices. The results show a higher correlation between the daily temperature range (DTR), cold nights and warm nights for soy production in Argentina (R2: − 0.74, − 0.80 and 0.75, respectively) and Mexico (R2: − 0.80, − 0.81, 0.70) for maize. Regionally, the linear model (simulated with observed data) using these indices presented an agreement with observed yield data in Mexico and Brazil, with explained variances exceeding 70% for maize in these subregions, while Argentina presented a better performance for soy yield. An intriguing finding was the superior performance of linear models when used with CPC-corrected RCM data compared to ERA5. Taken together, our results highlight the capabilities and constraints of linear models as valuable tools for developing adaptation and mitigation strategies, enabling precise yield forecasting, and informing policy decisions.

中文翻译：

---

极端气候和生物气候指数能否重现拉丁美洲的大豆和玉米产量？第 1 部分：观察和建模视角

摘要

IPCC表示，全球大多数地区将逐渐面临极端气候事件持续时间、频率和强度的放大，以及极端事件对人类福祉和经济造成的影响。本研究旨在开发线性回归模型，根据 1979 年至 2005 年间拉丁美洲三个地理次区域（墨西哥、巴西和阿根廷）的极端气候和生物气候指数估算大豆和玉米产量。我们使用来自观测的每日数据集 (CPC) 、再分析 (ERA5) 和协调区域气候降尺度实验 (CORDEX) 的区域气候模型 (RCM) 模拟，以研究极端温度和降水事件对 CORDEX 中美洲和南美洲地区玉米和大豆产量的影响。我们首先通过将 RCM 与观察结果进行比较来评估 RCM 在再现极端指数方面的性能。验证过程证明需要应用偏差校正技术来模拟每日降水量和温度，以获得更好的指数性能。结果显示，阿根廷（分别为 R2：− 0.74、− 0.80 和 0.75）和墨西哥（R2：− 0.80、− 0.81、0.70）的每日温度范围（DTR）、冷夜和暖夜之间的大豆产量相关性较高）对于玉米。从区域来看，使用这些指数的线性模型（用观测数据模拟）与墨西哥和巴西的观测产量数据一致，这些次区域玉米的解释方差超过 70%，而阿根廷的大豆产量表现更好。一个有趣的发现是，与 ERA5 相比，与 CPC 校正的 RCM 数据一起使用时，线性模型具有优越的性能。总而言之，我们的结果凸显了线性模型作为制定适应和缓解策略、实现精确产量预测和为政策决策提供信息的宝贵工具的能力和局限性。