为考察震后早期阶段混合模型的优势及实际预测效能,提升震后早期强余震时空预测的准确性,构建可操作的地震预测研究工作模型基础,选择能够较好反映震后空间应力分布的库仑应力变化和反映余震序列衰减且拟合效果较好的ETAS模型,构建Coulomb-ETAS混合模型。此模型依据强余震大多发生在应力加载区的实际情况,利用混合模型学习期间加载区事件的发生比例,将预测率从抑制区重新分配到加载区。以2021—2022年发生的云南漾濞6.5级、青海玛多7.4级、青海门源6.9级和四川泸定6.8级4次6.0级以上强震为例,对构建的混合模型进行检验和评估,并与单一的ETAS模型、C-RS模型进行对比。研究结果表明,3个模型在震后早期阶段对强余震均表现出相对较好的预测效果,呈现出与实际结果类似的衰减特性,仅有较少的预测失效现象。在频次滑动预测上,ETAS和Coulomb-ETAS模型优于C-RS模型,统计模型优势相对比较显著; 在空间发生率预测上,Coulomb-ETAS模型优于ETAS模型,认为库仑应力分布的混合模型能够降低虚报率,进而提高空间预测的准确性。因此,从4次震例上看,该混合模型优于单一的统计模型和物理模型,能够较好地适用于主震后的强余震时空预测,也可为开展多个模型混合及强余震预测之外的地震预测业务应用场景提供参考。

This study aims to investigate the advantages and actual forecast efficiency of the hybrid model in the early stage after the earthquake,improve the accuracy of strong aftershocks space-time forecast,so as to provide a model basis for the construction of operational earthquake forecasting research. We have developed the Coulomb-ETAS hybrid model,capitalizing on the Coulomb failure stress change to delineate the spatial stress distribution subsequent to an earthquake,and the ETAS model to characterize the decay pattern of aftershock sequences,recognized for its superior fit. The model's premise is anchored in the observation that strong aftershocks predominantly emerge in regions of stress concentration. It recalibrates the forecast probabilities,shifting emphasis from areas of stress release to those of stress accumulation,by leveraging the event occurrence ratios within the stress-loading regions during the model's training phase. The hybrid model's performance was rigorously tested and evaluated using data from four significant earthquakes(M≥6.0) in mainland China from 2021 to 2022:the Yunnan Yangbi M_S6.5,Qinghai Maduo M_S7.4,Qinghai Menyuan M_S6.9,and Sichuan Luding M_S6.8 earthquakes. These instances served to benchmark the hybrid model against both the standalone ETAS model and the C-RS model. Our findings indicate that all three models demonstrate commendable predictive performance for strong aftershocks in the post-earthquake phase,closely mirroring the observed decay characteristics with minimal predictive discrepancies. The Coulomb-ETAS model outperforms the C-RS model in frequency sliding forecasts,underscoring the pronounced advantage of incorporating statistical methodologies. Furthermore,it surpasses the ETAS model in spatial occurrence rate forecasting by integrating Coulomb stress,which effectively diminishes the rate of false alarms and augments the precision of spatial predictions. The comparative analysis across the four earthquake cases suggests that the hybrid model surpasses both the singular statistical and physical models in terms of spatio-temporal forecasting of strong aftershocks. It is well-suited for application post-mainshock and offers valuable insights for the integration of multiple models and the expansion of earthquake forecasting applications beyond the scope of strong aftershocks.

P315

地震预测开放基金(XH23072D)、国家自然科学基金(42404079)、中国地震局震情跟踪定向工作任务(2024010111)、天津市地震局局内课题(ZD202402)共同资助