外文摘要：Foxtail millet (or millet) is an important food crop in the northern Shanxi province of China (SXN). In the SXN, drought limits millet yield, which could be exacerbated by climate change. However, little is known about the impacts of future climate change on drought and its induced millet yield reduction in the SXN. This study investigated climate change in the future (2021-2060) relative to the baseline (1981-2020) and evaluated its impacts on drought under two emission scenarios (SSP245 and SSP585). We developed a new ensemble machine learning model to quantify the impacts of future climate change on drought -induced yield reduction for millet production. The results indicated that temperature and precipitation both show an increasing trend under future climate change. Drought intensity in most regions of the SXN was projected to be higher but drought frequency to be lower in the future relative to the baseline. The northeast parts of the SXN generally have a higher drought frequency and intensity than other regions. There are non -ignorable spatial differences in drought adaptability in the SXN, with lower drought adaptability in the southeast and southwest regions. Among these regions with different drought adaptability, the difference in yield reduction rate under the same drought intensity can reach up to 15.8 %. Therefore, when constructing models to quantify the relationship between drought intensity and millet yield reduction based on multi -site data, the spatial differences in drought adaptability should be considered. The ensemble machine learning model (Random Forest + Light Gradient Boosting Machine + Deep Forwarded Neural Network) using monthly drought intensity and drought adaptability index as predictors demonstrates high accuracy and regional applicability. According to simulation results, the millet production in the SXN generally has a lower yield reduction frequency but a higher yield reduction rate in the future relative to the baseline. Specifically, the yield reduction frequency decreases by 8.27 % under SSP245 and 11.28 % under SSP585, while the yield reduction rate increases by 3.64 % under SSP245 and 8.95 % under SSP585. Our findings provide important information for guiding agricultural water management to mitigate drought risk and its induced yield reduction under a changing climate.
外文关键词：climate change;Drought;ensemble machine learning;Drought adaptability;Yield reduction
作者：Wang, Chu;Zhou, Shiwei;Wu, Yangzhong;Lu, Huayu;Zhang, Zecheng;Liu, Zijin;Lei, Yongdeng;Chen, Fu
作者单位：Minist Agr & Rural Affairs;China Agr Univ
期刊名称：COMPUTERS AND ELECTRONICS IN AGRICULTURE
期刊影响因子：0.0
出版年份：2024
出版刊次：218
原文传递申请：江苏省科技资源（工程技术文献）统筹服务平台