大气科学学报2025，Vol.48Issue(6)：976-989,14.DOI:10.13878/j.cnki.dqkxxb.20240725001
基于扩展复数自回归模型的长江下游降水准2 a振荡年际预测研究
Interannual prediction of the quasi-biennial component of rainfall over the lower reaches of Yangtze River valley using an extended complex autore-gressive model
摘要
Abstract
The tropospheric quasi-biennial oscillation(TBO)is a mode of climate variability with a period of approximately 2-3 years,primarily observed in tropical,subtropical,and mid-to high-latitude regions of Eura-sia and the Southern Hemisphere.It manifests as quasi-periodic variations in atmospheric circulation,precipitati-on,sea surface temperature(SST),and snow cover.In China,a prominent quasi-2-year is evident in summer precipitation,particularly over the lower reaches of the Yangtze River Valley(LYRV),which lies within the East Asian subtropical monsoon region and exhibits pronounced TBO characteristics.Although the TBO is closely associated with large-scale climate modes such as the El Niño-Southern Oscillation(ENSO),its core driving mechanisms involve tropospheric dynamics,ocean-atmosphere interactions,and connections with stratospheric circulation.The TBO represents a critical timescale bridging annual cycles and interannual variability(e.g.,EN-SO).Understanding its evolution is essential for extending seasonal-to-interannual climate prediction lead times(approximately 6-18 months).The TBO is also closely linked to the variability of intraseasonal oscillations(ISO)and to extreme climate events such as monsoon precipitation anomalies,droughts,and heatwaves,thereby providing valuable guidance for agricultural planning,water resource management,and disaster mitiga-tion. This study develops a data-driven prediction model for interannual variations in the TBO component of rain-fall.The quasi-2-year components(TBO)of monthly precipitation in the LYRV and the principal components(quasi-biennial oscillation,QBO)of the 50 hPa stratospheric zonal and meridional winds for 1979-1998 were used to construct a time-varying Extended Complex Autoregressive(ECAR)model for predicting the QBO-relat-ed component of rainfall in the LYRV.An independent 12-year real-time interannual prediction experiment(1999-2020)was conducted on the quasi-biennial component of monthly precipitation over the LYRV.The re-sults demonstrate that the ECAR model exhibits high predictive skill,maintaining strong forecast accuracy up to a 15-month lead time—significantly outperforming the conventional autoregressive(AR)model.These forecasts provide valuable predictive guidance for anticipating summer flood processes in the LYRV more than a year in advance. The proposed data-driven prediction method employs real-time singular spectrum analysis(RSSA)to extract the TBO components from the troposphere and QBO components from the stratosphere,both characterized by strong autocorrelation.Through Fourier transformation,these primary quasi-2-year components are converted into complex low-frequency signals in the frequency space,forming an extended complex matrix that captures evol-ving relationships among atmospheric variables.This new set of variables to better jointly shape a new pattern of variable changes.From the perspective of multivariate synergy,collaborative patterns that are difficult to be iden-tified by traditional methods can be uncovered.A simplified,time-varying ECAR model is then derived to repre-sent the dynamic interactions among these components.The inverse Fourier transform yields the predicted vectors in the original space.This framework effectively reduces data diversity,simplifies complex relationships,and a-dapts to interdecadal changes in coupling among low-frequency processes,thereby enhancing forecast skill and extending prediction lead times.Unlike traditional physics-based numerical models or AI(artificial intelligence)systems constrained by initial conditions and model complexity,this data-knowledge-simplification approach pro-vides a robust alternative for interannual climate prediction.It captures real-time global QBO signals and the syn-ergistic effects of tropical and extratropical stratospheric QBOs on tropospheric TBO-related precipitation over the lower Yangtze River region,substantially improving interannual predictability of the TBO.When combined with interdecadal trends and sub-seasonal precipitation variability,this approach enhances the predictive capability for summer rainstorms and flood events across the LYRV.关键词
准2 a振荡/50 hPa风场/月降水量/长江下游/ECAR模型/年际预测Key words
quasi-biennial oscillation(QBO)/50 hPa wind/monthly precipitation/lower Yangtze River Valley/extended complex autoregressive model(ECAR)/interannual prediction引用本文复制引用
杨秋明..基于扩展复数自回归模型的长江下游降水准2 a振荡年际预测研究[J].大气科学学报,2025,48(6):976-989,14.基金项目
国家自然科学基金项目(41175082) （41175082）
