Drought-flood abrupt alternation (DFAA), as a typical successive extreme compound disaster, frequently threatens agricultural production. Existing studies have mainly focused on identifying DFAA from a meteorological perspective, yet they exhibit limited capacity to quantify or reflect associated agricultural losses. Here, we proposed an agriculture-oriented DFAA identification method integrating daily soil moisture and observed precipitation. The long-term trends of DFAA events during the summer maize growing season (1990–2024) in the Haihe Plain were analyzed, along with their impacts on maize yield and the associated atmospheric circulation drivers. The results suggested that both frequency and duration of DFAA increased significantly, at rates of 0.04 events/decade and 2.2 days/decade, respectively. This upward trend was primarily contributed by severe-class events. DFAA events occurred most frequently during summer maize sowing-jointing stage (3.07 events per station per year), followed by jointing-tasseling stage (0.78 events). SD-SF events occurring during the jointing-tasseling stage were associated with the greatest yield reduction (12.4%), which were significantly exceeding that caused by severe drought events alone (SD-NF: 6.4%), this phenological phase is the most sensitive to DFAA. Yield loss related to pure drought events exceeded that related to drought to light flood (xD-LF) events across most stages, confirming that xD-LF elicits a compensatory effect. A total of 13 regional DFAA processes were identified in the Haihe Plain. Their driving atmospheric circulation patterns were mainly characterized by two distinct modes: (i) the upper-level divergence from westerly jet entrance and the WPSH expansion, and (ii) the mid‑latitude westerly trough. These findings offer valuable references for DFAA forecasting and agricultural risk management in this climatically sensitive region.